article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
hadron spectroscopy has undergone a great renaissance in recent years @xcite .
the new findings include : low - lying excitations of d and b mesons , long - awaited missing states and new states near 4 gev / c@xmath0 in the charmonium spectrum , charmed and bottom baryons , and evidence for doubly charmed baryons .
the light hadron sector remains also restless reporting new scalar mesons or showing a deep theoretical interest in the high energy part of both the meson and baryon spectra . in this talk
we center our attention in the heavy baryon spectroscopy as well as some anomalous states present in the light baryon spectra .
heavy baryons containing a single heavy quark are particularly interesting .
the light degrees of freedom ( quarks and gluons ) circle around the nearly static heavy quark .
such a system behaves as the qcd analogue of the familiar hydrogen bounded by the electromagnetic interaction . when the heavy quark mass @xmath1 , the angular momentum of the light degrees of freedom is a good quantum number .
thus , heavy quark baryons belong to either su(3 ) antisymmetric @xmath2 or symmetric @xmath3 representations .
the spin of the light diquark is 0 for @xmath2 , while it is 1 for @xmath3 .
thus , the spin of the ground state baryons is @xmath4 for @xmath2 , representing the @xmath5 and @xmath6 baryons , while it can be both @xmath4 or @xmath7 for @xmath3 , allocating @xmath8 , @xmath9 , @xmath10 , @xmath11 , @xmath12 and @xmath13 , where the star indicates spin @xmath7 states
. therefore heavy hadrons form doublets .
for example , @xmath8 and @xmath14 will be degenerate in the heavy quark limit .
their mass splitting is caused by the chromomagnetic interaction at the order @xmath15 .
these effects can be , for example , taken into account systematically in the framework of heavy quark effective field theory . the mass difference between states belonging to the @xmath2 and @xmath3 representations
do also contain the dynamics of the light diquark subsystem , hard to accommodate in any heavy quark mass expansion .
therefore , exact solutions of the three - body problem for heavy hadrons are theoretically desirable because they will serve to test the reliability of approximate techniques : heavy quark mass expansions , variational calculations , or quark - diquark approximations .
.masses , in mev , of charmed and bottom baryons . [ cols="^,^,^,^,^,^,^,^,^,^,^,^ " , ] although the value of @xmath16 might vary depending on the configurations involved in each @xmath17 coupling we shall use for the sake of simplicity the same value in all cases .
the @xmath18 results for @xmath19 mev are numerically detailed in table [ t2 ] where the values for @xmath20 and for ( @xmath21 in the chosen @xmath22 channel as well as their probabilities to give @xmath18 are also displayed .
as can be checked the improvement of the description with respect to @xmath23 is astonishing .
all the predicted @xmath18 masses lye very close to the pdg average masses for the anomalies . in fig .
[ fig3 ] the @xmath18 values for @xmath24 mev are drawn as compared to the experimental mass intervals .
( boxes ) .
n.c . means non - cataloged resonance.,width=415 ] we interpret these results as providing strong quantitative support to our former qualitative description of the anomalies .
regarding their nature a look at the probabilities reveal they are mostly meson - baryon states .
actually a meson - baryon probability greater or equal than @xmath25 can serve as a criterion to identify an anomaly .
nonetheless the coupling to the @xmath26 component is essential to lower their masses making them more stable against decay into @xmath27 .
it should be emphasized that similar results could be obtained for any other spectroscopic @xmath28 model through a fine tuning of the value of @xmath29 ( note that the small value of @xmath30 as compared to @xmath20 and @xmath31 provides an _ a posteriori _ validation of our method)@xmath32 this comes from the expression of the eigenvalues where it is clear that even for @xmath33 one gets @xmath34 which according to our @xmath22 choice is much closer to the pdg mass of the anomaly than @xmath20 .
this means that concerning the mass of the anomalies the coupling of meson - baryon to @xmath26 components may play the role of a general _ healing mechanism _ for spectroscopic models .
we have studied the heavy baryon spectra by means of the faddeev method in momentum space .
these results should be highly valuable both from the theoretical and experimental points of view .
theoretically , it should be a powerful tool for testing different approximate methods to solve the three - body problem .
experimentally , the remarkable agreement with known experimental data make our predictions highly valuable as a guideline to experimentalists .
heavy baryons constitute an extremely interesting problem joining the dynamics of light - light and heavy - light subsystems in an amazing manner . while the mass difference between members of the same su(3 ) configuration , either @xmath2 or @xmath3 , is determined by the perturbative one - gluon exchange , the mass difference between members of different representations comes mainly determined by the dynamics of the light diquark , and should therefore be determined in consistency with the light baryon spectra .
there is therefore a remnant effect of pseudoscalar forces in the two - light quark subsystem . for light baryons
we propose that @xmath35 components , in the form of @xmath36 wave meson - baryon channels which we identify , play an essential role in the description of the anomalies , say baryon resonances very significantly overpredicted by three - quark models based on two - body interactions . as a matter of fact by considering a simplified description of the anomalies as systems composed of a free meson - baryon channel interacting with a three - quark confined component we have shown they could correspond mostly to meson - baryon states but with a non - negligible @xmath26 state probability which makes their masses to be below the meson - baryon threshold .
the remarkable agreement of our results with data in all cases suggests the implementation of meson - baryon threshold effects as an essential physical mechanism to give account of spectral states poorly described by constituent quark models . | we discuss recent results on heavy and light baryon spectroscopy . address = dpto .
de fsica fundamental , universidad de salamanca , spain address = dpto . de fsica terica , universidad de valencia ( uv ) and ific ( uv - csic ) , spain address = dpto . de fsica terica , universidad de valencia , spain address = escuela superior de fsica y matemticas , instituto politcnico nacional , mexico |
magnetically ordered spin systems with coupled electron - nuclear spin oscillations attracted general interest for a long time owing to their very peculiar nonlinear properties . in such systems
the dynamics of the paramagnetic nuclear spin system is governed by the exceptionally strong ( of the order 10@xmath3 - 10@xmath4oe ) hyperfine field from the magnetically ordered electron spin system . in turn , the motion of electron spins is affected by the hyperfine field from the nuclear spins .
this results in coupling between oscillations of electron and nuclear spin systems . from the microscopic point of view
, one can describe such systems by pointing out that the coupled oscillations of the nuclear spins with the magnetically ordered electron subsystem introduce a long distance correlation between the precession of different nuclear spins .
@xcite this correlation can be described in terms of nuclear spin waves , i.e. n - magnons . from this point of view
the hyperfine interaction leads to the hybridization of , on the one hand , usual spin waves in the magnetically ordered electron subsystem , i.e. e - magnons , and , on the other hand , the precession of a nuclear spin around the direction of the hyperfine field . fig .
[ fig:1 ] shows examples of spectra for hybridized e - magnons ( frequency @xmath5 ) and n - magnons ( frequency @xmath6 ) in a weakly anisotropic antiferromagnet mnco@xmath0 . at zero wave number , @xmath7 , the frequency of n - magnons to a good approximation
is given by @xmath8 where @xmath9 is the frequency of precession of the nuclear spin in stationary hyperfine field , that is when the oscillations of the electron spin system caused by coupling to the nuclear spins are neglected , and @xmath10 is a quantity that depends on a number of parameters including the temperature of the nuclear subsystem . at large
@xmath11 the spectrum of this branch approaches @xmath12 .
thus , the correlation between motion of different nuclear spins is significant only for small enough @xmath11 , or in other words over a distance of typically 10@xmath13 - 10@xmath14 lattice sites . for the existence of @xmath15-magnons
the frequency difference between @xmath16 and @xmath9 must satisfy the condition @xmath17 where @xmath18 is the spread in the frequency @xmath9 due to inhomogeneities of the atomic and magnetic structures of the sample ( inhomogeneous broadening ) and due to relaxation processes ( homogeneous broadening ) .
there are a few crystalline materials which satisfy this condition .
most notable examples are the weakly anisotropic antiferromagnets with @xmath19mn as the magnetic ions , such as mnco@xmath20 , csmnf@xmath20 , rbmnf@xmath20 , kmnf@xmath20 , etc
. the magnetic properties of these materials have been extensively studied by the continuous wave ( cw ) nuclear magnetic resonance ( nmr ) , @xcite double electron - nuclear magnetic resonance , @xcite as well as pulsed nmr methods .
@xcite in the static magnetic field of @xmath21koe and at two values of temperature of @xmath22k ( dashed - point line ) and @xmath23k ( solid line ) .
the dashed lines show corresponding spectra in the case when coupling between electron and nuclear oscillations is neglected.,scaledwidth=48.0% ] an equivalent macroscopic description of coupled electron - nuclear spin systems is also possible .
@xcite owing to correlated motion of a huge number ( typically @xmath24 ) of spins , both the magnetically ordered electron spin subsystem and the paramagnetic nuclear spin subsystem can be described by a corresponding macroscopic net magnetization vectors @xmath25 and @xmath26 . the hyperfine interaction energy is given by @xmath27 , where @xmath28 is the hyperfine interaction constant .
like for any two classical coupled oscillators , the spectrum of coupled precession of @xmath25 and @xmath26 vectors exhibit repulsion between the precession frequencies of two uncoupled spin subsystems .
this results in two branches of a hybridized magnetic resonance , i.e. the quasi nuclear ( q - nmr ) and quasi electron ( q - afmr ) modes of precession .
the frequency of q - nmr mode is approximately given by @xmath29 where @xmath30 ( @xmath31 is the magnitude of the hyperfine field and @xmath32 is the gyromagnetic ratio of the nuclear spin ) , @xmath33 is the magnitude of the nuclear magnetization vector , and @xmath34 is the magnitude of the projection of the vector @xmath35 on the direction of stationary hyperfine field @xmath36 .
the last term in eq .
( [ eq : freq ] ) is usually referred to as the dynamical frequency shift , or frequency pulling .
note that for @xmath37 , that is for negligible deflection of the nuclear magnetization vector from the equilibrium position , the frequency of q - nmr mode approaches the bottom of the n - magnon spectrum , see eq .
( [ eq : bot ] ) . from the equivalent microscopic point of view
, the magnetization deflection may be described by a density of non - equilibrium magnons excited by a pulsed or cw radio - frequency ( rf ) pumping .
we note that the frequency of q - afmr mode decreases with the increase of the density of non - equilibrium magnons .
this means that the e - magnons have an attractive interaction and the homogeneous precession is unstable .
in contrast , the frequency of q - nmr mode increases with deflection of nuclear magnetization from equilibrium .
this shows the repulsive interaction between n - magnons and leads to stability of the homogeneous precession .
since the n - magnons obey bose - einstein statistics , there arises an important question about whether the non - equilibrium magnons can form a bose - einstein condensate ( bec ) at rf pumping that creates a sufficiently large density of non - equilibrium magnons .
the peculiar non - linear properties described above are very similar to magnetic properties of the superfluid phase of @xmath14he .
therefore , in this section we would like to review main experimental findings in this system , in particular the formation of a bec of non - equilibrium magnons .
the first study of nonlinear magnetic resonance in superfluid @xmath14he was performed at the kapitza institute in moscow and revealed a long lived induction decay signal after a pulsed nmr pumping in @xmath14he - b.@xcite this signal was explained as a formation of a domain with a homogeneously precessing nuclear magnetization .
@xcite this state corresponds to the phase - coherent precession of magnetization even in a strongly inhomogeneous static magnetic field and corresponds to the global minimum of the ginzburg - landau potential for a non - equilibrium magnon system .
later , this state was associated with the formation of magnon bec , and it is possible to show a close analogy between the formation of atomic bec and magnon bec .
@xcite more recently , another type of magnon bec was observed in thin films of yttrium iron garnet ( y@xmath20fe@xmath38o@xmath39 ) in a static magnetic field applied along the plane of the film .
@xcite in this case the minimum of the global energy corresponds to the spin waves with the wave numbers @xmath40@xmath41 .
this magnon bec state corresponds to a high - amplitude interference pattern of stationary spin waves with wave vectors @xmath42 and @xmath43 , and has some similarities with the charge density waves in superconductors .
we will not consider this type of magnon bec here .
the condition for magnon bec formation follows directly from bose - einstein statistics according to which the occupation number for the magnon state with the wave number @xmath11 is given by @xmath44^{-1 } , \label{eq : be}\ ] ] where @xmath45 is the chemical potential and @xmath46 is the effective temperature of the nuclear system . as usual
, the condition for the formation of a bec with @xmath7 corresponds to @xmath47 .
the atomic bec forms when the temperature of an atomic system decreases below the critical temperature @xmath48 at which @xmath49 . in the case of a magnon system , for which @xmath45 is always zero in thermal equilibrium , one needs to create non - equilibrium magnons by external rf pumping to increase their chemical potential . correspondingly , at a given temperature @xmath46 the bec is formed when condition @xmath50 is reached .
the critical density @xmath51 of the magnons which satisfy the above condition can be written as @xcite @xmath52 where @xmath53 is the density of thermal magnons at a given temperature @xmath46 .
the critical density of magnons is related to the angle @xmath54 of deflection of the macroscopic nuclear magnetization vector from its equilibrium orientation as @xmath55 , where @xmath56 is roughly the number of the nuclear spins in a sample . since typically @xmath57 , the critical angle @xmath58 needed to reach bec can be as low as a few degrees .
@xcite the next question is the stability of non - equilibrium magnons .
if homogeneous precession is not stable , the thermalization of the magnon system leads to the increase of the temperature of magnetic system and the critical density of magnons given by eq .
( [ eq : bec ] ) increases .
however , if coherent precession of magnons is supported by the minimum of the magnetic energy potential , then the magnons form the bec which lives a relatively long time .
thus , the stability of homogeneous precession of nuclear magnetization is the crucial point in the formation of bec .
next , let us discuss the fundamental difference between magnon bec state and usual magnetic resonance in magnetically - ordered media . usually in the analysis of magnetic resonance in a sample
, we must take into account the inhomogeneous broadening of the nmr line .
the magnetic resonance signal is formed by a set of relatively independent magnetic oscillators which have different local frequencies . in the case of magnetically - ordered media
the correlation between these oscillators appears on the distances of the effective magnetic interactions , i. e. @xmath59-@xmath60 interatomic distances . in the magnetically - ordered media considered above , the precession frequency depends on the deflection angle of the magnetization , therefore on the density of pumped non - equilibrium magnons and interaction between them
this interaction can be attractive or repulsive . in the first case ,
the well - known effects of instability in atomic bec @xcite and magnon bec @xcite take place .
we are interested in the second case when it is energetically favorable to conserve the homogeneity of precession .
when the density of non - equilibrium magnons reaches the critical value , they form condensate described by a single macroscopic wave function . the magnons in bec precess with the same frequency and with a small spreading of the phase which compensates for the inhomogeneity of the external conditions . in the bec state ,
the transverse component of magnetization is described by a wave function @xmath61 , where @xmath62 is the magnitude of the projection of the magnetization vector on the plane perpendicular to the stationary hyperfine field and @xmath63 is the phase of the wave function at @xmath64 .
this wave function describes all properties of the spin superfluidity . the best system for investigation of magnon bec and spin superfluidity is superfluid @xmath14he - b .
in this system , there were found many phenomena which have one to one correspondence with the analogous phenomena of mass superfluidity and superconductivity .
for example , critical supercurrent and phase slippage @xcite , the josephson spin - current effect @xcite , as well as spin - current vortices @xcite and goldstone modes @xcite , were found . during the last 30 years
, five different magnon bec states were found in the @xmath14he superfluids .
comprehensive reviews can be found in the literature@xcite .
unlike @xmath14he - b , the a - phase of superfluid @xmath14he has a different magnetic properties .
the superfluid @xmath14he - a is an easy - plane two sublattice liquid antiferromagnet , which magnetic properties are described by a classical antiferromagnetic hamiltonian .
homogeneous precession of the magnetization in bulk @xmath14he - a is not stable due to the attractive interaction between the magnons .
such an instability was predicted theoretically @xcite and also observed experimentally @xcite .
later it was suggested by bunkov and volovik @xcite that a reorientation of the orbital momentum of @xmath14he - a could lead to a repulsive interaction between the magnons and to their bec formation .
this suggestion was confirmed in experiments with @xmath14he - a immersed in aerogel which had been squeezed along the magnetic field . in this case
, aerogel orients the orbital momentum of @xmath14he - a along the magnetic field .
@xcite in such a configuration the magnon bec was observed .
@xcite the observation of magnon bec in @xmath14he - a is important for the subject of this paper because the dynamic properties of nmr in antiferromagnets , as discussed here , are similar to the dynamic magnetic properties of superfluid @xmath14he - a in squeezed aerogel .
the magnon bec in the system with suhl - nakamura interaction was predicted @xcite and subsequently found in csmnf@xmath0 and mnco@xmath0 .
@xcite this observation sparked a discussion about magnon bec in solid magnets .
the main objection was related to the question of magnetic relaxation .
it is customary to describe the nuclear magnetic relaxation by the phenomenological bloch equations according to which the deflected magnetization must thermalize on a time scale @xmath65 .
however , pulsed nmr experiments showed the formation of a relatively long lived induction decay signal , similar to one observed in the superfluid @xmath14he - a .
@xcite this indicates that the nuclear magnetization is not described by the bloch equations .
in particular , the magnitude of the magnetization vector does not change under magnetization deflection and relaxation .
this article provides detailed theory and experimental confirmation of this effect . in section
ii , we provide a full macroscopic description of coupled electron - nuclear spin systems based on the landau - lifshiz - gilbert equations linearized with respect to the transverse components of the electron magnetization vector . in section iii we describe measurements of the nonlinear nmr signals in a weakly anisotropic easy - plane antiferromagnet mnco@xmath0 and provide detailed comparison with our theory . some details for the derivation of the coupled landau - lifshiz - gilbert equations
are given in the appendix .
it is convenient to start with the expression for spectrum of coupled electron - nuclear spin waves .
@xcite in the case of the weakly anisotropic antiferromagnets , the spectrum is given by ( see fig . [ fig:1 ] ) @xmath66^{1/2}. \label{eq : couplmagn}\end{aligned}\ ] ] in the above equation , the frequency @xmath67 of a low - frequency branch of electron spin waves obtained by ignoring the hyperfine interaction of electrons with nuclei is given by @xmath68 where @xmath69 is the gyromagnetic ratio of the electron spin , @xmath70 cm is the inter - atomic distance , and the exchange ( @xmath71 ) and dzyalushinskii - moriya ( @xmath72 ) effective magnetic fields will be defined in the next section .
we also use notation @xmath73 where @xmath74 and @xmath75 are the magnitudes of the electron and nuclear magnetization vectors .
it describes the gap in the spectrum of the e - magnon branch @xmath76 at @xmath77=0 ( the so called hyperfine gap ) .
this value can be considered as a measure of influence of the nuclear spins on the electron spin oscillations . in particular
, the critical value of external magnetic field @xmath78 defined by @xmath79 separates the regions of weak ( @xmath80 ) and strong ( @xmath81 ) influence of nuclear spins on the dynamics of the electron spins . for mnco@xmath20
, we have @xmath82koe at @xmath83k .
the @xmath7 mode of n - magnon branch @xmath84 corresponds to the homogeneous precession of nuclear spins and is usually excited by the homogeneous rf magnetic field at the frequency @xmath85 .
an important feature of the resonance frequency @xmath86 is its dependence on the projection of nuclear magnetization vector @xmath26 on the direction of the hyperfine field @xmath87 , see eq .
( [ eq : freq ] ) .
this prevents us from obtaining large deflection angles @xmath88 by the rf excitation at the resonance frequency @xmath89 given above .
the reason for this is the deviation from the resonance condition @xmath90 at increasing @xmath54 .
however , it is possible to overcome this restriction by using the rf pumping at frequency in the interval @xmath91 . such rf pumping does not excite the homogeneous precession of @xmath35 , that is @xmath7 mode of n - magnons , if the frequency detuning satisfies the condition @xmath92 , where @xmath93 is the amplitude of the rf magnetic field . on the other hand , in the frequency band of n - magnons
there is a wave vector @xmath94 such that @xmath95 , see fig .
[ fig:2 ] .
the momentum conservation prohibits direct resonant excitation of such magnons by the homogeneous rf pumping . however , their indirect excitation is possible through intermediaries which are able to bypass the requirements of the momentum conservation , as was suggested earlier .
@xcite the analysis showed that the appearance of such n - magnons decreases the energy of nuclear spin system , therefore further leads to their gradual accumulation . in turn , the accumulation of n - magnons will lead to the increase of the defection angle and modification of the frequency band @xmath96 of n - magnons . as a result ,
the n - magnons with frequency @xmath97 for given @xmath98 become out of resonance with the rf pumping .
the further evolution of such non - resonant n - magnons can proceed according to two possible scenarios .
the first one corresponds to their thermalization , which increases the temperature of the nuclear spin system , i.e. the heating scenario .
the heating of the nuclear subsystem decreases the magnitude of the magnetization vector @xmath35 .
the nuclear temperature increases until the bottom of the n - magnon spectrum reaches the rf pumping frequency , that is @xmath99 , where @xmath100 corresponds to the new temperature of the nuclear subsystem .
the stationary state is characterized by a small amplitude of homogeneous precession of @xmath35 having a significantly smaller magnitude . in another scenario , two @xmath15-magnons with wave vectors @xmath101
can mutually transform their @xmath102 to the value @xmath103 which corresponds to the new resonance condition @xmath104 .
thus , the rf pumping and accumulation of magnons continues , leading to further increase of deflection angle @xmath54 , and so on .
the accumulation of magnons and increasing of deflection angle stops when the bottom of the frequency band of n - magnons reaches the rf pumping frequency @xmath105 see dash - dotted line in fig .
[ fig:2 ] . in this scenario ,
all accumulated n - magnons which provide the value of the deflection angle in the above equation are transformed into n - magnons with @xmath7 , which form a bec with anomalously large magnon density . for zero deflection angle ( solid line ) and for deflection angle given by eq .
( [ eq : beta0 ] ) ( dash - dotted line ) .
the frequency of rf pumping @xmath89 is indicated by a dashed line.,scaledwidth=49.0% ] in the next section we provide the quantitative macroscopic description of the steady states for both scenarios .
the scenario with heating corresponds to the bloch relaxation model which is described by the phenomenological longitudinal and transverse relaxation times .
the scenario with deflection of magnetization vector @xmath35 which conserves the vector magnitude corresponds to the indirect relaxation of n - magnons via the electron spin systems .
this mechanism appears as a consequence of the coupled oscillations of @xmath106 and @xmath35 magnetization vectors .
the preliminary results of this theoretical description were discussed earlier .
@xcite our theoretical analysis of a coupled electron - nuclear spin system in the easy - plane antiferromagnet mnco@xmath0 is based on the classical landau - lifshitz - gilbert equations for macroscopic magnetization vectors of magnetic sub - lattices.@xcite as usual , such an analysis of quantum spin systems is justified because the quantum - mechanical equations of motion reduce to the classical equations for the precession of the macroscopic magnetization vectors after averaging over the quantum - mechanical density matrix operator . following the previous work , @xcite
we describe the dynamics of coupled electron - nuclear spin system by equations for two electron ( @xmath107 and @xmath108 ) and two nuclear ( @xmath109 and @xmath110 ) magnetization vectors of two sub - lattices @xmath111 @xmath112 where @xmath113 and @xmath114 ( @xmath115 and @xmath116 ) are the effective magnetic fields acting on each electron ( nuclear ) sub - lattice , and @xmath117 and @xmath118 ( @xmath119 and @xmath120 ) are the relaxation terms for the electron ( nuclear ) magnetization in each sub - lattice .
the effective magnetic fields are given by @xmath121 the thermodynamic potential of the whole spin system including its interaction with the external static and rf magnetic fields is given by @xmath122 the exchange interaction between electron spins of the two sub - lattices is described by @xmath123 where @xmath124 describes the exchange interaction and @xmath125 is the sample volume .
the energy of magnetic anisotropy with dedicated axis @xmath126 chosen as the positive @xmath127 direction is described by @xmath128.\ ] ] in the antiferromagnet mnco@xmath20 considered here , this anisotropy is the easy - plane which corresponds to the condition @xmath129 .
the dzyaloshinskii - moriya interaction which is responsible for non - collinearity of the @xmath107 and @xmath108 sub - lattices , see fig .
[ fig : geom ] , is described by @xmath130 where @xmath131 is a constant which characterizes the strength of dzyaloshinkii - moriya interaction .
the energy of @xmath107 and @xmath108 magnetizations in the external static magnetic field @xmath132 applied in the positive @xmath133 direction and ac magnetic field @xmath134 applied along axis @xmath135 is described by @xmath136 in what follows , we neglect direct interaction of nuclear magnetization vectors @xmath137 and @xmath138 with the static and rf magnetic fields compared to the much stronger hyperfine field .
however , note that the rf magnetic field still strongly couples to the nuclear spins via the @xmath134-induced oscillations of the electron magnetization . the hyperfine interaction between the electron and nuclear spins , which for simplicity we assume to be isotropic , is described by @xmath139 at @xmath134=0 , the condition of minimum of the functional ( [ eq : phi ] ) defines the equilibrium orientation of the vectors @xmath140 and @xmath141 ,
see fig .
[ fig : geom ] .
the small angle @xmath142 in this figure is determined from the equation @xmath143 where @xmath144 is the magnitude of the dzyaloshinskii - moria field , @xmath145 is the magnitude of the electron magnetization vector in each sub - lattice , and the exchange field is given by @xmath146 the typical angle @xmath142 is of the order @xmath147 rad . the solutions of eqs .
( [ eq : m]-[eq : phihyp ] ) for the small deviations of @xmath140 and @xmath141 from their equilibrium orientations were investigated earlier .
@xcite in this case , the eigenmodes of coupled electron - nuclear oscillations can be found from eqs .
( [ eq : m]-[eq : m ] ) linearized with respect to the transverse components of magnetization vectors @xmath140 and @xmath141 and by taking the rf field @xmath148 and relaxation terms to be zero .
as expected , the corresponding eigenfrequencies coincide with the frequencies of coupled electron - nuclear spin waves , see eq .
( [ eq : couplmagn ] ) , taken at @xmath7 .
the electron and nuclear ac susceptibilities can be obtained from the same linearized equations by including the homogeneous rf magnetic field @xmath148 .
@xcite it what follows , we are interested in finding solutions of eqs .
( [ eq : m]-[eq : phihyp ] ) for arbitrarily large deviations of vectors @xmath141 from their equilibrium orientations due to rf magnetic field at frequencies close to the q - nmr resonance . following the previous work ,
@xcite it is convenient to introduce a separate reference frame @xmath149 ( @xmath150 ) for sublattice @xmath151 ( @xmath152 ) obtained from the original frame @xmath153 by a clockwise rotation around @xmath135-axis by angle @xmath142 ( by angle @xmath154 ) , see fig .
[ fig : geom ] , and chose new variables according to @xmath155 where @xmath156 , @xmath157 , @xmath158 ( @xmath159 , @xmath160 , @xmath161 ) , @xmath162 , are components of electron ( nuclear ) magnetization vectors in the corresponding frame .
the solutions of eqs .
( [ eq : m]-[eq : m ] ) in these new variables can by found by linearizing eqs .
( [ eq : m ] ) with respect to transverse components of electron magnetization vectors , whose deviation from the equilibrium orientation is assumed to be small .
this assumption is valid for sufficiently large magnetic fields @xmath80 where the oscillation of electron spins is only weakly affected by the hyperfine interaction with nuclei , see discussion in the previous section .
the details of derivation are given in the appendix .
finally , is is convenient to write the result in the coordinate system @xmath163 where the components of the net nuclear magnetization @xmath164 , @xmath165 and @xmath166 are related to @xmath167 , @xmath168 and @xmath169 by @xmath170 which is somewhat similar to the usual rotating wave transformation used to describe traditional nmr .
@xcite by omitting the relaxation terms in eqs .
( [ eq : m]-[eq : m ] ) the equations for the net nuclear magnetization become ( see appendix for details ) @xmath171 @xmath172 @xmath173 where @xmath174 is proportional to the amplitude of the rf magnetic field @xmath175 and @xmath10 is the parameter discussed previously ( see eqs .
( a14 - 15 ) in the appendix for the precise definition of @xmath174 and @xmath10 ) .
since the relaxation terms were neglected , the above equations can be used for description of the dynamics of the magnetization vector @xmath35 only on the time intervals @xmath176 which are small compared to the nuclear magnetic relaxation time .
note that , as was discussed previously from the microscopic point of view , the presence of the frequency pooling term @xmath177 does not allow us tp obtain large deflection angles of @xmath35 from the equilibrium orientation by using a short rf pulse , as typically can be done in conventional nmr ( the rabi flopping ) .
@xcite indeed , for typical experimental conditions we have @xmath178 , therefore the orientation of the axis of rotation of vector @xmath35 in the frame @xmath163 changes quickly to be nearly parallel to @xmath179 after applying the rf pulse .
however , as we show below it is possible to obtain large angles of deflection by cw nmr .
the description of the stationary state of the spin system under cw rf pumping requires to take into account the proper form of relaxation terms in eqs .
( [ eq : m]-[eq : m ] ) .
as discussed previously , the interaction between transverse components of the electron and nuclear magnetization vectors significantly changes the frequency of the nuclear spin oscillations .
another consequence of this interaction is the appearance of the relaxation of the transverse component of vector @xmath35 .
the indirect relaxation of this type has been studied previously but for relatively small amplitudes of @xmath35-vector oscillations when the nonlinear phenomena did not play any significant role .
@xcite here , we consider the nonlinear equations of motion for the nuclear magnetization taking into account its indirect relaxation via the electron spin system .
typically , the relaxation of the transverse component of magnetization occurs due to fluctuating magnetic fields acting on different spins and inhomogeneity of the external static magnetic field acting in the electron system . in what follows ,
we use an assumption that , because the nuclear spins experience an enormously large hyperfine field @xmath180-@xmath181oe , we can neglect any other fields including fluctuating magnetic fields responsible for the relaxation of vectors @xmath182 .
we can note immediately that the approximation @xmath183=0 in eqs .
( [ eq : m ] ) conserves the magnitude of the vectors @xmath184 , and therefore the net magnetization vector @xmath35 .
thus , this approximation excludes the possibility of the heating of nuclear spin system under any processes including its excitation by rf pumping . using this approximation together with the assumption of small deviations of electron magnetization vectors from their equilibrium orientations at @xmath80 , the equations of motion for the vector @xmath35 become ( see appendix for details ) @xmath185 @xmath186 @xmath187 where the nuclear transverse relaxation time @xmath188 is related to the electron transverse relaxation time @xmath189 by eq .
( [ a13 ] ) in the appendix .
the cw nmr signals are described by the stationary state solutions of the above equations .
the latter are given by @xmath190 @xmath191 where @xmath192 , which determines the longitudinal component of @xmath35 , is the root of the fourth degree polynomial equation @xmath193 and @xmath194 is given by @xmath195 for the sake of qualitative comparison with our theory , let us also consider the traditional bloch approach in which the nuclear relaxation is described by introducing two phenomenological relaxation times , i.e. the longitudinal magnetization relaxation time @xmath196 and transverse magnetization relaxation time @xmath65
. @xcite adding the corresponding relaxation terms into eqs .
( [ eq : m ] ) , we obtain the phenomenological bloch equations @xmath197 @xmath198 @xmath199 note that the above equations are obtained by disregarding the electron spin relaxation terms @xmath200 in eqs .
( [ eq : m ] ) .
the stationary state solutions of ( [ eq : mxbloch]-[eq : mzbloch ] ) can be written as @xmath201 @xmath202 where @xmath203 , as defined earlier , is given by the roots of the third degree polynomial equation @xmath204 to distinguish between the two approximations which lead , on the one hand , to eqs .
( [ eq : mxstead]-[eq : mzstead ] ) , and , on the other hand , to eqs .
( [ eq : mxsteadbloch]-[eq : mzsteadbloch ] ) , we need to compare the continuous nmr signals predicted by these equations with the experimentally observed ones .
this is done in the next section .
in the experiment , we studied cw nmr signals in an easy - plane antiferromagnetic mnco@xmath0 sample at temperatures below 1 k. the sample was in the form of a rhombus - shaped plate about @xmath205 mm thick , with diagonals of approximately @xmath206 and @xmath207 mm .
the magnetic anisotropy axis @xmath208 was perpendicular to the plane of the plate .
the measured mass of the sample was about @xmath209 g . to excite continuous nmr signals in mnco@xmath0 by external rf pumping
, the sample was placed in a high - quality factor rf resonator of a split - ring type .
@xcite the resonator and sample were oriented such that the external static magnetic field @xmath210 and the rf magnetic field @xmath211 of the resonator were perpendicular to each other and to the anisotropy axis @xmath208 of the sample .
the resonator was made of high - conductivity oxygen - free copper and had a resonant frequency of about @xmath212mhz and a quality factor of about 600 measured at @xmath1k . the input rf signal transmitted through a cryogenic low - loss semi - rigid coaxial cable
was coupled to the resonator via a loop made at the end of the cable , and reflected signal was measured at different values of static field @xmath77 and input rf power @xmath213 using a network analyzer .
the signals proportional to the in - phase and quadrature ( with respect to the rf field in the cavity ) components of ac magnetization induced in the sample could be extracted from the reflected signal . in order to avoid complications associated with the strong coupling of the spin system to the electro - magnetic mode of the resonator , which in particular leads to the normal mode splitting in the resonator reflection spectrum , @xcite we used the frequency of rf pumping of @xmath214mhz , that is significantly detuned from the resonant frequency of the resonator . in this so called dispersive regime of coupling ,
the in phase ( @xmath215 ) and quadrature ( @xmath216 ) components of the ac magnetization are related to the complex amplitude ( phasor ) of the reflected signal according to @xmath217 where @xmath218 is the complex amplitude of the signal reflected from an empty cavity , @xmath219 is imaginary unit , and @xmath220 is a proportionality coefficient that depends on characteristics or resonator and its coupling to the rf transmission line .
figure [ fig : lowp ] shows the nmr signals proportional to @xmath216 ( absorption signal ) and @xmath215 ( dispersion signal ) measured at @xmath221mk as a function of static field @xmath77 at low input rf power of @xmath222dbm . at such a low power ,
the deviation of nuclear magnetization from its equilibrium orientation is very small .
correspondingly , the frequency pooling is very close to its value for the equilibrium spin system , therefore nonlinearity of the nmr signal is negligible .
these signals are used to find the fitting parameters which define the dependence of the frequency pulling and nuclear magnetic relaxation rate on the value of @xmath77 .
it should be noted that although these signals were measured in the linear regime of nmr they are not ideally symmetric .
this suggests that the sample used in the experiment is not a single crystal , which is also confirmed by its careful visual examination under the microscope .
it is likely that the sample consists of at least two or three mono - crystals of different size with somewhat different orientations of their lattices with respect to each other . at @xmath221mk and at low input rf power of @xmath223dbm .
dashed lines are the ac magnetization components @xmath216 and @xmath215 calculated as described in section iii(c).,scaledwidth=48.0% ] figures [ fig : abs ] and [ fig : disp ] show the absorption and dispersion signals measured at @xmath221mk and at different input rf power levels .
it is clear that with increasing power the nmr frequency shifts towards lower values of @xmath77 due to the frequency pulling .
both the absorption and dispersion signals attain very asymmetrical shape .
in addition , both signals show strong hysteresis upon reversing the direction of the field sweep . in figure [ fig : highpower ] we show the absorption and dispersion signals obtained at the highest input power of @xmath224dbm used in this experiment . in the next section
we show comparison of the measured absorption an dispersion signals with predictions of our theory , see eqs .
( [ eq : mxstead]-[eq : mzstead ] ) .
in addition , we compare the experimental results with predictions based on the phenomenological bloch equations , see eqs .
( [ eq : mxsteadbloch]-[eq : mzsteadbloch ] ) . at @xmath221mk and at several values of input rf power
( shown in the figure ) during down - field sweep .
dashed lines represent values of @xmath216 calculated as described in section iii(c).,scaledwidth=48.0% ] at @xmath221mk and at several values of input rf power ( shown in the figure ) during down - field sweep .
dashed lines represent values of @xmath215 calculated as described in section iii(c).,scaledwidth=48.0% ] mk at highest input rf power of @xmath224dbm during up - field ( dashed lines ) and down - field ( solid lines ) sweeps .
the dash - dotted line is the result of numerical calculations for @xmath203 as described in section iii(c).,scaledwidth=48.0% ] in order to compare the results presented in figs .
( [ fig : abs]-[fig : disp ] ) with predictions of our theory , first we need to find relation between the ac magnetization components @xmath216 and @xmath215 and components of nuclear magnetization @xmath164 and @xmath165 described by eqs .
( [ eq : mxstead]-[eq : mystead ] ) .
it is important to realize that due to the much larger magnitude of the electron magnetization vector @xmath225 compared to that of the nuclear magnetization vector @xmath226 , whose ratio is approximately @xmath227 , the main contribution to the observed ac magnetization comes from the oscillations of vector @xmath106 .
we can find needed relations using relation between the complex amplitudes of @xmath228 and @xmath229 , as given by eq .
( [ a7]-[a9 ] ) in the appendix . according to fig .
[ fig : geom ] and eqs .
( [ eq : sin]-[eq : magmm ] ) , the vector component of electron magnetization @xmath230 , which defines the observed nmr signals , can be expressed in terms of @xmath228 as @xmath231 assuming the time dependence of the rf magnetic field in the form @xmath232 , the absorption @xmath216 and dispersion @xmath215 signals are given by , respectively , the imaginary and real parts of the complex amplitude of @xmath233 .
the expressions for @xmath215 and @xmath216 in terms of oscillating components of the nuclear magnetization vector can be found from the above equation using eqs .
( [ a7]-[a9 ] ) in the appendix .
using the transformation ( [ eq : rwt ] ) and expressing results in terms of nuclear magnetization vector components @xmath164 and @xmath165 we obtain @xmath234 , \nonumber\\ & & m_a= \frac{h_n}{h } \left ( 1-\frac{\omega^2}{\omega_e^2 } \right)^{-1 } \left [ m_y + \frac{\omega}{\omega_n\omega_pt_{2n } } m_x \right ] .
\label{eq : mdma } \end{aligned}\ ] ] in the above equations , the non - resonant contribution @xmath235 to the dispersion signal is given by @xmath236 it arises from the direct coupling of the electron magnetization vector to the rf magnetic field having amplitude @xmath93 .
the resonant ( at the nmr frequency ) contribution comes from the terms containing @xmath164 and @xmath165 .
the main contribution to the dispersion ( absorption ) signal comes from the @xmath164 ( @xmath165 ) component of the nuclear magnetization vector .
note that it is enhanced by the amplification factor @xmath237 given by eq .
( [ a11 ] ) in the appendix .
the components @xmath164 and @xmath165 are given by equations ( [ eq : mxstead]-[eq : mystead ] ) and are expressed in terms of the dimensionless parameter @xmath203 determined from eq .
( [ eq : mzstead ] ) .
it is instructive to discuss the general form of the solutions of this equations .
( [ eq : mzstead ] ) is a fourth order polynomial equation with respect to @xmath238 .
one of the roots of eq .
( [ eq : mzstead ] ) , viz .
@xmath2392 , corresponds to unstable state with antiparallel orientation of the vectors @xmath35 and @xmath240 .
the other three roots are the solutions of a third order polynomial equation which is obtained from ( [ eq : mzstead ] ) by excluding the root @xmath2392 .
we are interested in the real - valued roots and their dependence on @xmath77 , which is defined by the corresponding dependencies of @xmath241 , @xmath242 and @xmath243 given by eqs .
( [ a10]-[a13 ] ) in the appendix . at low rf excitation powers
such that @xmath244 the equation ( [ eq : mzstead ] ) has only one real root given by @xmath245 using this expression in eqs .
( [ eq : mxstead]-[eq : mystead ] ) we obtain @xmath246 at fixed rf frequency @xmath89 the position of the resonance in static magnetic field @xmath77 is determined by the dependence of @xmath10 on @xmath77 given by eq .
( [ a12 ] ) in the appendix .
the resonance value of field @xmath247 is defined by @xmath248 the width of the resonance @xmath249 , which can be defined as the distance between the maximum and minimum of the dispersion signal , is related to @xmath188 as @xmath250 the dependence of @xmath188 on @xmath77 is given by eq .
( [ a13 ] ) in the appendix .
it is convenient to express the dependence of @xmath10 and @xmath188 on @xmath77 in terms of adjustable parameters @xmath251 and @xmath252 as @xmath253 and determine the numerical values of @xmath251 and @xmath252 at given temperature from the comparison of experimental data taken at low rf excitation power and the ac magnetization components @xmath216 and @xmath215 calculated using eqs .
( [ eq : mdma ] ) and ( [ eq : linear ] ) .
such comparison is shown in fig .
[ fig : lowp ] where the @xmath216 and @xmath215 calculated using @xmath254mhz(koe)@xmath13 and @xmath255mhz(koe)@xmath256 are plotted by the dashed lines . here , we used @xmath257koe , @xmath258mhz , and @xmath259mhz .
the value of @xmath93 was chosen to be sufficiently small to ensure a linear nmr regime .
the corresponding values of @xmath260 and @xmath249 are approximately @xmath261koe and @xmath262koe , respectively .
the width of the resonance corresponds to the nuclear relaxation @xmath188 time of approximately @xmath263s
. for arbitrarily large rf excitation powers there exists a range of @xmath77 where there are more than one real - valued roots of eq .
( [ eq : mzstead ] ) .
the numerical solution of this equation for a large value of @xmath93 is shown by the dash - dotted line in fig .
[ fig : highpower ] . in this calculation , we used @xmath264 and @xmath243 given by eqs .
( [ eq : omegap]-[eq : t2n ] ) with numerical values of @xmath251 and @xmath252 given above .
the interval @xmath265 corresponds to the region where three real roots exist .
the merging point of two real roots ( where they become complex ) corresponds to the phenomenon which in mathematics is called _
catastrophe_. we restrict ourselves to the analysis of the upper branch @xmath266 which has the critical point at @xmath267 , see fig .
[ fig : highpower ] .
the state of the lower branch @xmath268 is destroyed at the fields @xmath269 due to instabilities caused by the excitation of @xmath15-magnons which are in resonance with the rf pumping field .
@xcite in order to make comparison between the experimental data presented in figs .
[ fig : abs]-[fig : disp ] and the ac magnetization components @xmath216 and @xmath215 calculated using eqs .
( [ eq : mxstead]-[eq : mzstead],[eq : mdma ] ) we need to establish correspondence between , on the one hand , the values of rf excitation power used in the experiment and , on the other hand , the numerical values of @xmath93 used in the calculations .
we used the value of @xmath270koe in fig .
[ fig : highpower ] to define the value of @xmath271oe at the excitation power of @xmath272dbm .
the lineshapes of @xmath216 and @xmath215 calculated for the corresponding values of @xmath93 and using eqs .
( [ eq : omegap ] ) and ( [ eq : t2n ] ) are plotted in figs .
[ fig : abs ] and [ fig : disp ] , respectively , by the dashed lines .
as seen in these figures , we obtain a very good quantitative agreement between our experimental data and predictions of our theoretical model based on eqs .
( [ eq : mxllg]-[eq : mzllg ] ) .
note that the conservation of the magnitude of vector @xmath35 , which follows from our theory , allows us to express @xmath266 in terms of the deflection angle @xmath54 between the vector @xmath35 and its equilibrium orientation @xmath273 at the critical field @xmath274koe in fig .
[ fig : highpower ] corresponding to the input rf power of @xmath272dbm the deflection angel has value @xmath275degrees . for the sake of comparison with the predictions of the bloch model based on equations ( [ eq : mxbloch]-[eq :
mzbloch ] ) it is instructive to consider dependence of the absorption signal @xmath216 at @xmath267 on the input rf power @xmath213 . note that the critical field @xmath276 corresponds to the resonant condition @xmath277 from eqs .
( [ eq : mxstead]-[eq : mystead ] ) for the steady state solutions this gives @xmath278 and @xmath279 , therefore from the second line of ( [ eq : mdma ] ) and eqs .
( [ a10]-[a13 ] ) in the appendix we obtain @xmath280 in order to get rid of the field - independent factor in front of the right - hand side of the above equation , it is convenient to normalize this expression to the value of @xmath281 at one of the values of the input rf power used in the experiment .
for example , we choose the lowest power of @xmath282dbm used in fig . [
fig : abs ] .
the expression for the normalized absorption signal becomes ( for the sake of simplicity we neglect @xmath283 ) @xmath284 where @xmath285 , @xmath286 and @xmath287 are the corresponding values of @xmath216 , @xmath288 and @xmath93 at the input rf power @xmath289 .
note that the normalized absorption signal depends only on the values of the critical field @xmath288 and the ratio @xmath290 , which can be readily determined in the experiment .
figure [ fig : comp ] shows dependence of @xmath291 on @xmath292 obtained from the experimental data shown in figs . [ fig : abs ] and [ fig : highpower ] ( opened squares ) , as well as theoretically predicted dependence given by eq .
( [ eq : alpha ] ) ( dashed line ) . as a function of quantity @xmath292 defined by eq .
( [ eq : alpha ] ) obtained from the experimental data presented in figs .
[ fig : abs ] and [ fig : highpower ] ( opened squares ) . the lines are the corresponding dependences predicted by our theory ( dashed line ) and by the conventional bloch approach ( dash - dotted line).,scaledwidth=48.0% ] now we compare this result with predictions of the heating scenario based on the phenomenological bloch equations ( [ eq : mxbloch]-[eq : mzbloch ] ) .
note that similar to eq .
( [ eq : mzstead ] ) , the third order polynomial equation ( [ eq : mzsteadbloch ] ) with respect to @xmath238 can have three real - valued roots in certain ranges of the static magnetic field @xmath265 .
this may result in the nonlinear nmr signals similar to those described earlier .
@xcite using condition ( [ eq : reson ] ) satisfied at the field @xmath267 , we obtain from eqs .
( [ eq : mxsteadbloch]-[eq : mzsteadbloch ] ) @xmath293 note that , unlike in our proposed theory , the bloch model does not conserve the magnitude of magnetization vector . indeed , from eq .
( [ eq : blochhc ] ) we have @xmath294 typically , we have @xmath295 , therefore @xmath296 can be significantly less than @xmath75 already for @xmath297 . using eqs .
( [ eq : reson],[eq : blochhc ] ) it is convenient to write @xmath298 as @xmath299 where @xmath300 .
note that using the above equation , the values of @xmath301 at @xmath267 can be determined from the experimental values of @xmath302 .
using the second line of eq .
( [ eq : mdma ] ) we obtain expression for the absorption signal at the catastrophe field @xmath303 and using eqs .
( [ a10]-[a11 ] ) from the appendix and definition of @xmath304 , see eq .
( [ eq : alpha ] ) , we obtain for the normalized absorption signal @xmath305 where @xmath306 is the value of @xmath301 at the critical field for the input rf power @xmath289 . again , note that the term in front of @xmath304 on the right - hand side of the above equation depends only on the value of the critical field @xmath267 , which can be determined from the experiment , and does not depend on the choice of the values for @xmath196 and @xmath65 .
the dependence of the normalized absorption signal on @xmath304 given by eq .
( [ eq : alphabloch ] ) is plotted in fig .
[ fig : comp ] by a dash - dotted line . clearly , the conventional bloch approach fails to adequately account for the behavior observed in the experiment . using eq .
( [ eq : omega1t1t2 ] ) , it is straightforward to estimate the value of @xmath298 at the critical field @xmath267 for the maximum input rf power @xmath224dbm . using the value of @xmath274koe ,
see fig .
[ fig : highpower ] , we obtain @xmath307 . thus , the bloch theory predicts significant reduction of the magnitude of the nuclear magnetization vector due to the heating .
this does not agree with our experimental results .
we suggest the description of the coupled electron - nuclear spin systems by the coupled landau - lifshitz - gilbert equations which take into account an indirect relaxation of nuclear spins via the electron spin subsystem . in our theory
the magnitude of the nuclear magnetization is conserved for arbitrary large excitation powers , which is drastically different from the conventional heating scenario based on the phenomenological bloch approach . in particular , we consider the coupled electron - nuclear spin motion in the easy - plane antiferromagnetic crystals with a weak magnetic anisotropy in the basal plain .
the equations of motion are solved analytically assuming small deviations of the electron magnetization vectors from their equilibrium orientations , which allows us to linearize the corresponding equations of motion for the electron subsystem .
the obtained solutions are compared with the experimental nonlinear nmr signals obtained in a mnco@xmath0 sample at temperatures below 1 k and good quantitative agreement is found .
this suggests that at high excitation rf powers used in the experiment the nuclear magnetization vector can be deflected by very large angles exceeding 70 degrees .
this result is drastically different from predications of the standard heating scenario based on the phenomenological bloch approach . in particular , the latter predicts significant reduction of the magnitude of the nuclear magnetization vector under conditions of our experiment .
the proposed theory brings together the properties of nonlinear nmr in , on the one hand , the magnetic systems in solids considered here and , on the other hand , the superfluid @xmath2he where the magnitude of the nuclear magnetization vector is also conserved .
this provides the theoretical background for the mechanism of formation of a bec of non - equilibrium magnons in mnco@xmath0 .
this topic is currently under the theoretical and experimental investigation .
* acknowledgements * the work of l. v. a and d. k. was supported by an internal grant from the okinawa institute of science and technology graduate university and jsps kakenhi ( grant no .
26400340 ) . the work of yu .
m. b. and r. r. g. was supported by the russian science foundation ( grant rsf 16 - 12 - 10359 ) .
part of the work is done by the order of faso of the rf : subject `` spin '' no 01201463332 , grant no 15 - 9 - 2 - 49 ; subject " electron no 01201463326 , grant no 15 - 8 - 2 - 10 .
. m. b acknowledges support from the okinawa institute of science and technology graduate university as a visiting scientist .
we thank jason ball for proof - reading the manuscript .
the system of coupled equations ( [ eq : m]-[eq : m ] ) contains twelve dynamical variables corresponding to three components for each of four magnetization vectors @xmath140 and @xmath141 .
this system conveniently splits into two independent subsystems , each consisting of six equations , if we introduce new dynamical variables according to @xcite where the upper indexes are defined as @xmath309 $ ] , @xmath310 $ ] , @xmath311 $ ] , and @xmath156 , @xmath157 , @xmath158 ( @xmath159 , @xmath160 , @xmath161 ) , @xmath162 , are components of electron ( nuclear ) magnetization vectors in the corresponding reference frames @xmath312 and @xmath313 defined in fig . [
fig : geom ] .
the components @xmath314 correspond to the high - frequency branch of oscillations of the vectors @xmath315 , which is weakly affected by the interaction between electron and nuclear spins . for this reason ,
only the low - frequency branch of @xmath315 is of interest for nmr .
this branch is described by the components @xmath316 and @xmath317 .
it is convenient to mark the indexes @xmath318 , @xmath319 and @xmath320 as , respectively , @xmath321 , @xmath237 and @xmath322 . in what follows
, we consider the static magnetic fields @xmath80 at which the deflection angles of the vectors @xmath315 from their equilibrium orientation are small , see discussion in section ii , which corresponds to the following approximation for the components of vector @xmath25 where @xmath333 is the frequency of uniform precession of vector @xmath25 neglecting the interaction between electron and nuclear spins .
the condition ( [ a5 ] ) is well satisfied for typical rf frequencies close to the nmr frequency , and thus significantly lower than electron - spin resonance frequency .
the equations ( [ a3 ] ) can be simplified further . noting that , since @xmath334 and @xmath335 , we can neglect the last two terms in the right - hand side of the first equation in ( [ a3 ] ) . moreover , since typically @xmath336-@xmath337oe and @xmath338-@xmath339oe , we can neglect @xmath340 comparing with @xmath71 . using these approximations , we obtain from ( [ a3 ] ) as usual , it is convenient to look for the solutions of the above equation in the complex form @xmath342 and @xmath343 , where @xmath344 and @xmath345 are the complex amplitudes ( phasors ) of oscillating components @xmath346 and @xmath347 . using the rotating wave approximation ,
@xcite it is straightforward to obtain to obtain the above equations , we used expansion in small parameter @xmath355 .
the influence of the relaxation is described by the imaginary part of @xmath356 in ( [ a9 ] ) . note that the above equations satisfy the condition @xmath357 which means the absence of dissipation at @xmath358 .
it is important to fulfill this strict requirement imposed by the theory of irreversible processes . in the above equations , the harmonic dependence of the components @xmath347 , @xmath360 and @xmath361 due to the rf magnetic field ( [ a4 ] ) comes from the terms containing @xmath346 . as usual , it is convenient to get rid of the harmonic time - dependent terms by applying the transformation given by eq .
( [ eq : rwt ] ) .
note that in terms of transformed coordinates @xmath164 , @xmath165 and @xmath166 the complex amplitude of @xmath347 appearing in eq .
( [ a9 ] ) can be written as @xmath362 . using this transformation and eqs .
( [ a7]-[a13 ] ) , it is straightforward to obtain eqs .
( [ eq : mxllg]-[eq : mzllg ] ) discussed in the main text . | we present a new theoretical description of the coupled electron - nuclear spin systems which takes into account an indirect relaxation of nuclear spins via the electron subsystem . in our theory
the magnitude of the nuclear magnetization is conserved for arbitrary large excitation powers , similar to the landau - lifshitz - gilbert model of relaxation .
this is drastically different from the conventional heating scenario based on the phenomenological bloch equations .
the predictions of our theory are compared with the experimental nonlinear nmr signals obtained in a weakly anisotropic antiferromagnetic mnco@xmath0 sample at temperatures below @xmath1k and good quantitative agreement is observed .
the proposed theory brings together the properties of magnetic nuclear resonance in , on the one hand , the magnetic systems considered here and , on the other hand , the superfluid @xmath2he where the magnitude of the nuclear magnetization vector is also conserved . |
early type galaxies ( the family consisting of elliptical and lenticular galaxies , hereafter e+s0 galaxies ) represent one of the most important diagnostics within the cosmological picture of galaxy formation and evolution .
e+s0 galaxies and bulges comprise most of the baryonic ( stellar ) mass ( 50% or more of all stellar spectral types ) in the present - day universe ( e.g. , * ? ? ?
* ) and reside in the most massive dark matter haloes with halo masses up to @xmath7 at @xmath8 , hence these systems are the key ingredients for understanding the star formation and the mass assembly history of the galaxy population as a whole and for the structure formation starting from the early primordial stages of the universe .
early hierarchical structure formation models based on cold dark matter particles predict different evolutionary paths and assembly time
scales for the densest environments of clusters of galaxies and the lowest densities of isolated galaxies ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . from typical age variations of 43% in the overall stellar content between cluster and field galaxies , a late mass assembly for massive e+s0 galaxies of smaller sub units
is expected with the involved physical processes ( mainly star formation rates , gas cooling and heating ) poorly understood . as a consequence ,
the simplified baryonic recipes are interpreted and implemented differently in the simulations @xcite . in this paradigm , field
early type galaxies are assembled through merging and accretion events of smaller clumps and fragments on longer time scales up to the recent past @xcite .
the environment is therefore suggested to play an important role to constrain the formation and evolution of e+s0 galaxies and their internal properties .
these predictions have significant imprints for the abundance and mass distribution of an existing or even fully established population of old massive systems at @xmath9 @xcite and also for recent claims of the detection of their possible progenitors at high redshifts of @xmath10 @xcite . over the past few years
theoretical semi - analytic models of galaxy formation have been improved dramatically . by including a secondary star burst mode to the previous quiescent star formation history , i.e. allowing a fraction of gas to be directly transformed into stars in merging galaxies @xcite , an increasing amount of supernova feedback and/or an additional effect of active galactic nuclei ( agn ) feedback @xcite or virial shocks of infalling gas in a spherical halo @xcite to suppress the remaining star formation activity ,
a much better agreement with the observations has been achieved .
the suppression of star formation in massive halos using primarily agn feedback has a strong impact at the top of the luminosity function ( lf ) that results in a far less overproduction of massive galaxies at @xmath8 than earlier simulations , while still providing a good description of the rest of the observed lf up to @xmath11 @xcite .
however , there is still a non - negligible excess of local luminous galaxies detectable in the local simulated lf , which models adjust by including additional internal dust obscuration or a relative inefficiency of the stellar mass of merging galaxies that results in a population of free - floating intergalactic stars @xcite .
previous studies have quantified to detect the deviations from passive evolutionary tracks as predicted by semi - analytic models in the evolution of the fundamental plane ( fp ; * ? ? ?
* ; * ? ? ?
this relationship is constructed among the basic parameters of e+s0 galaxies , their size ( the half - light effective radius @xmath12 in kpc ) , stellar kinetic energy ( central velocity dispersion @xmath13 in km s@xmath14 ) and luminosity ( average surface brightness within @xmath12 , @xmath15 , in units of mag arcsec@xmath16 ) and provides a powerful tool to measure the the evolution of the mass - to - light ( @xmath5 ) ratio as a function of redshift without being subject to uncertainties in the photometric model estimates , such as initial mass function or age and/or metallicity degeneracy . using this observational technique cluster early
type galaxies formed the majority of their stars at redshifts @xmath17 and their stellar populations follow a slow passive evolution since then ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
however , the cosmic formation epoch and evolutionary history of isolated field galaxies remains unclear .
indeed , galaxy formation models based on a hierarchical structure growth envision for a given galaxy mass a dependence on environment @xcite . at a given redshift this environmental effect results in explicit age variations between e+s0 galaxies in clusters and in low density regions , where cluster ellipticals show an average luminosity weighted age of 9.6 gyr , whereas field e+s0s being by @xmath184 gyr younger and displaying a broader age spread than their counterparts in clusters @xcite .
in addition , brighter ( and hence more massive ) field galaxies ( @xmath19 ) should feature on average @xmath182 gyr younger ages than fainter ones and comprise more metal rich @xmath20}}\,\le 0.1 $ ] dex populations than their cluster representatives @xcite , but no differences in their element abundance ratios ( @xmath21/fe ) .
so far , for the field galaxy population inconsistent results have been derived .
some studies suggest an evolution similar to early type galaxies in clusters @xcite , but others prefer a faster mass assembly over a shorter time period since @xmath6 @xcite .
recently , @xcite performed a comprehensive investigation of early - type field galaxies within the goods - n field and detected evidence for a mass dependent evolution .
similar results were derived by @xcite and @xcite from an analysis of field e+s0 galaxies in the cdfs . however , although these samples were drawn from the same field they rather found incomparable galaxy sizes and velocity dispersions for individual objects ( see [ sec : litcmp ] for more details ) . moreover
, the different target selection procedures among the studies ( e.g. , magnitude , colour , morphology ) might induce some inconsistency in the final sample , for example selecting galaxies with the reddest colours would favour the more massive systems where variations are less pronounced or using low signal - to - noise ( @xmath22 ) spectra limits the detection post - starburst galaxies or the contamination by bulge - dominated spiral galaxies .
furthermore , difficulties may arise from the small number statistics of some works and the impact of cosmic variance remains yet to be unknown and hard to distinguish from the used sample selection ( e.g. , * ? ? ?
* ; * ? ? ?
* but see [ sec : litcmp ] ) .
there is clear evidence that the interpretations of measurements are highly sensitive to selection effects and high - quality spectroscopic data is stringent necessary to allow a careful analysis of the properties of galaxies at high redshift . in the present work
, we aim to analyse in detail the assembly history and the stellar populations of a carefully selected sample of 24 early - type field spheroidal galaxies down to @xmath1 in two deep sky surveys , the fors deep field and the william herschel deep field , by using a combination of high signal - to - noise vlt spectroscopy together with deep ground - based multi band imaging and high - resolution space - based photometry .
although our study is restricted to small number statistics , our data is based on high - quality spectra providing both precise kinematic and stellar population properties . in a future work
, we are aiming to disentangle both difficulties of high redshift investigations related to small number statistics and cosmic variance ( a. fritz et al .
2009a , in preparation ) .
the structure of this paper is as follows . in section 2 , the sample selection , observations and the kinematic analysis of the spectroscopic data
are described .
the photometric observations together with the derivation of structural parameters , galaxy morphologies and spectrophotometric properties are presented in section 3 . here
also a discussion on available x - ray data is given .
the scaling relations for field early - type galaxies are presented in 4 , and the implication of these results on their formation is discussed in 5 .
a comparison with previous works is given in 6 , followed by a discussion in 7 .
the main conclusions and results are summarised in 8 . throughout this article
the concordance cosmology for a flat , low - density universe is assumed with @xmath23 , @xmath24 and @xmath25kms@xmath14mpc@xmath14 . unless otherwise noted ,
all magnitudes are given in the vega - system .
@xmath26@xmath27 colour - colour diagram for the whdf field galaxies .
spiral candidates ( squares ) are displayed with the final observed field early - type galaxies ( circles ) and compared to evolutionary tracks for e / s0 with different formation redshifts of @xmath28 ( solid line ) and @xmath29 ( dotted line ) predicted by the burst bc96 models .
for the bc96 models a cosmology with @xmath30kms@xmath14mpc@xmath14 and @xmath31 was adopted . ]
target objects were selected based upon deep @xmath32 and @xmath33 multi - band imaging surveys of the fors deep field ( fdf , * ? ? ? * ) and william herschel deep field ( whdf , * ? ? ?
* ) , respectively .
the 50% completeness limits of the fdf and whdf are @xmath34 and @xmath35 , respectively .
early - type candidates were selected according to their ( i ) total apparent magnitude @xmath36 , ( ii ) spectrophotometric type , ( iii ) elongated structureless appearance and ( iv ) photometric redshift ( only available for the fdf ) .
the constraint in the apparent brightness ensured a sufficient signal - to - noise of @xmath22@xmath3715 in the absorption lines , mandatory for a robust determination of velocity dispersions and line - strengths measurements .
for this reason , faint early - type candidates with apparent magnitudes @xmath38 were included in more than one mos setup and the individual spectra combined after data reduction .
spectrophotometric types and estimated redshifts were selected from the fdf photometric redshifts catalogue of more than 3800 objects @xcite . only candidates with an early ( e / s0 ) model spectral energy distribution ( sed ) were considered and the photometric redshifts were restricted to @xmath39 . based on the photometric redshifts , the elliptical candidates were spread out on different ccd positions among the spirals , which were observed for a similar project studying the tully fisher relation of late type galaxies , for each mos setup in the observed spectral wavelength range such that either the -feature passband ( @xmath40 ) or the g - band ( @xmath41 ) was included . for a reliable star / galaxy separation , targets with insignificant photometric redshift and large uncertainty ( i.e. , @xmath42 ) were discarded .
field candidates were selected from a deep fors2 @xmath43-band reference image consisting of 10 seeing averaged @xmath43-band images ( between 0.470.50arcsec fwhm ) with a final integration time of 3000sec and a point - spread - function ( psf ) of 0.49arcsec fwhm . in case of the fors1 observations ,
the target selection was based on a fors1 @xmath43-band reference image with a psf of 0.52arcsec fwhm . in the whdf
, field objects were drawn from a deep over five individual exposures averaged @xmath43-band reference image with a total integration time of 1500sec and a psf of 0.6arcsec fwhm . to avoid misclassifications , detections of the source extractor package ( sextractor ;
* ) that exhibited a star classification parameter ` star`@xmath370.9 and @xmath44@xmath45@[email protected] were most likely bonafide stellar objects and were therefore rejected .
apart from photometric redshifts , the target selection for the whdf field early - type galaxies was performed with the same constraints as for the fdf field elliptical galaxies .
redshifts of field early - type candidates were estimated using a combination of colour - colour diagrams and apparent magnitudes .
figure [ fig : cmdbrbi ] illustrates the target selection of the whdf field elliptical candidates in the @xmath26@xmath27 colour - colour diagram .
the total whdf spectroscopic data and the actually observed whdf field ellipticals are compared to evolutionary tracks for e / s0 with different formation redshifts of @xmath28 ( solid line ) and @xmath29 ( dotted line ) as predicted by the passive evolution models for a simple ( single burst ) stellar population of bruzual & charlot .
note , that the bc96 models have been computed for a slightly different cosmology with @xmath30kms@xmath14mpc@xmath14 and @xmath31 . in the @xmath26@xmath27 colour - colour diagram
elliptical and spiral galaxies @xcite are well separated up to @xmath47 ( indicated by the arrows ) and follow the predictions of the evolutionary models quite well . to avoid a selection of high redshift elliptical galaxies with @xmath48 , the apparent @xmath49 magnitude was used as an additional constraint . although the bc96 models represent more a generalisation of the various stellar populations of early - type galaxies , together with the cut in magnitude they provide a good separation between high redshift interlopers and lower redshift galaxies which we were interested in . more complex model star formation histories ( cf .
* ) are beyond the scope of this paper . to allow a better estimate of the star formation histories of the galaxies in our sample , we will utilise absorption line indices and other spectral diagnostics in a future work ( a. fritz et al . 2009b , in preparation ) .
as shown in figure [ fig : cmdbrbi ] , galaxies populating a narrow range in the colour - colour diagram of @xmath50 and @xmath51 have been selected as bonafide field ellipticals and successfully verified .
one object ( id14 ) initially selected as an early - type candidate with @xmath52 , @xmath53 and @xmath54 ( @xmath55 ) is an intermediate - type sc spiral galaxy and was therefore discarded .
it turned out that this galaxy is a spiral galaxy with a clearly visible disc on the acs images which were not available at the time of the target selection .
spectroscopic observations of the sample of field early - type galaxies were performed simultaneously with those of the late - type galaxies which were subject to the investigation of the evolution of the tully fisher relation in the fdf @xcite and in the whdf @xcite .
as the target selections and instrument configurations were very similar the two individual data sets are discussed in combination .
multi - object spectroscopy ( mos ) was conducted at the very large telescope ( vlt ) using the focal reducer and low dispersion spectrograph 2 ( fors2 ; * ? ? ?
* ) instrument in september and october 2000 , july and october 2002 and fors1 in december 1999 and october 2001 . during these observations ,
a total of 220 galaxy spectra have been acquired under mostly excellent seeing conditions . using the grism ` 600r ` with the order separation filter ` gg435 + 81 ` and slit width of 1.0arcsec resulted in spectral resolutions of @xmath56 for the fdf observations .
for the whdf run , the volume phased holographic grism ` 600ri ` was chosen , which is even more efficient at redder wavelengths ( @xmath578000 ) than the grism ` 600r ` and has a resolution of @xmath58 .
both versions of fors instruments offer 19 slitlets in the mos configuration , 9 with slit lengths of 22arcsec , 8 with slit lengths of 20arcsec , whereas the uppermost and lowermost slits have lengths of @xmath1811arcsec .
the total integration time for each setup was 2.5hours . seeing conditions ranged between 0.43arcsec and 0.92arcsec fwhm , with a median of 0.76arcsec . in order to limit the corrections of atmospheric absorption , the airmass was constraint to @xmath59 .
fors was operated at standard resolution setup , i.e. ccd read out in one - port mode . in low gain mode
( using port a ) fors1 offers a gain of 3.51@xmath60/adu and a read - out - noise ( ron ) of ron=7.21@xmath60 , resulting in a spatial scale of 0.2arcsec / pixel for the fdf spectra . after
the fors2 detector system upgrade which was available for science use in march 2002 , the standard setup offers a gain of 0.70@xmath60/adu and ron of 2.7@xmath60 .
the spatial sampling in the final whdf spectra was 0.25arcsec / pixel .
more details on the observations can be found in @xcite and @xcite .
table [ fdfeobs ] summarises the spectroscopic observations of the field early - type candidates in the fdf .
columns give the date of the observation , the mean airmass and mean differential image motion monitor ( dimm ) seeing value during the observations , the number of e / s0 galaxies and anonymous galaxies ( additional fill up objects with no @xmath61 ) for each mos setup .
galaxies which were observed twice are counted only once in the total number of objects .
a total of nine mos setups yielded a final sample of 39 e+s0 galaxies , regardless of their environment ( see further discussion in [ sec : zdis ] ) .
table [ whdfeobs ] gives an outline of the observations for the field early - type galaxies in the whdf . in total , six different mos setups with various mos mask orientations depending on the interested position angles of the spiral galaxies were observed , in order to fulfill both requirements for spirals , disc inclination angles of @xmath62 and restricted misalignment angles of @xmath63 between apparent major axis and slit orientation ( cf .
each setup was splitted into three single exposures of 3000sec with a total integration time of 2.5 hours . and
typically comprised 23 early - type candidates .
two additional objects ( one in setup 4 and one in setup 1 ) fell by coincidence into a slit of a target galaxy .
one object ( id810b ) turned out to be a field s0 galaxy at @xmath64 .
the other object is a background spiral galaxy at @xmath65 and thus was discarded .
therefore , the total sample of whdf elliptical galaxies comprises eleven field early - type galaxies .
lcccrc tel./instrument & date & airm . &
dimm & e+s0 & anon . & ( m / d / y ) & & & & vlt / fors2 & 09/27/00 & 1.21 & 0.51 & 5 & 1 + & 09/27/00 & 1.33 & 0.43 & 5 & - + & 10/04/00 & 1.40 & 0.81 & 6 & 1 + & 10/05/00 & 1.36 & 0.80 & 5 & 2 + & 10/05/00 & 1.28 & 0.74 & 8 & - + & 10/06/00 & 1.15 & 0.66 & 7 & 2 + & @xmath66 & & & 28 & 6 + vlt / fors1 & 12/00/99 & 1.08 & 0.66 & 1 & 0 + & 10/12/01 & 1.43 & 0.76 & 6 & 1 + & 10/14/01 & 1.07 & 0.89 & 5 & 0 + & 10/12/01 & 1.38 & 0.82 & 5 & 3 + & @xmath66 & & & 11 & 4 + clccrc mos & date & airm . & dimm & e+s0 & anon . &
( m / d / y ) & & & & 1 & 07/10 , 10/03/02 & 1.18 & 0.64 & 1 & 1 + 2 & 08/07 , 09/11/02 & 1.19 & 0.73 & 2 & - + 3 & 08/04 , 08/07/02 & 1.51 & 0.92 & 3 & - + 4 & 08/04 , 10/02/02 & 1.13 & 0.89 & 3 & 1 + 5 & 09/10 , 09/12/02 & 1.18 & 0.91 & 2 & - + 6 & 09/12 , 10/4 - 5/02 & 1.14 & 0.85 &
2 & - + & @xmath66 & & & 11 & 1 + image reduction followed the standard procedure of overscan subtraction , flat fielding , sky background subtraction , and wavelength calibration as described in @xcite and @xcite . because spectra particularly near to the edges of the mos frame exhibited spatial distortions , images of each individual slitlet were extracted from the full frame after bias subtraction allowing the typical two - dimensional image reduction of long - slit spectroscopy .
the night sky background was subtracted by iteratively fitting each ccd column separately using a polynomial fit of first or zero order . during the wavelength calibration , for the dispersion relation typical rms values of 0.04 at a dispersion of @xmath181.62 / pixel
were achieved .
after the wavelength calibration , spectra of the respective exposures were summed up and the spectral rows corresponding to the full - width at half maximum ( fwhm ) of an object were averaged along the dispersion direction using a horne - based algorithm @xcite . we also compared this method to a different slit extraction technique using equal - weight extractions which yielded to the same velocity dispersion measurements within their uncertainties @xcite . in those cases where a galaxy was observed in two different mos setups , the wavelength ranges covered did not match each other exactly leading to varying final count rates at different wavelengths .
wavelength units , the inset shows the derived quantities . ]
internal velocity dispersions ( @xmath13 ) and radial velocities ( @xmath67 ) of the galaxies were computed using an updated version of the _ fourier correlation quotient _ ( fcq ) method @xcite , as described in @xcite and @xcite .
a brief summary follows .
the algorithm is based on the deconvolution of the peak of the template - galaxy correlation function with the peak of the autocorrelation function of a template star . in comparison to other procedures @xcite ,
the main advantages of the fcq method are the possibility to _ a posteriori _ chose the most appropriate function to approximate the line - of - sight - velocity - distribution ( losvd ) , a very high robustness against template mismatch and a high insensitivity due to metallicity differences in the spectra .
for sufficient high @xmath22 spectra of 10 per pixel , even velocity dispersions of @xmath68 , with @xmath69 being the instrumental resolution , can be determined with a high degree of accuracy . for each galaxy , @xmath13 and @xmath67
were measured around the spectral wavelength range of the g4300-band ( hereafter g - band ) , the and the h@xmath70 feature in fourier space by convolving a template spectrum with a gaussian to match the width of absorption lines in the observed galaxy spectrum of interest .
spectral regions that suffered from strong night sky emission lines , sky subtraction residuals defects or the a- or b - band absorption caused from the terrestrial atmosphere were carefully masked out during the fitting process . in 10 out of 24 cases , at least two different wavelength regimes could be analysed . for all early - type galaxy
candidates the signal - to - noise in the spectra was sufficient to yield very stable stellar velocity dispersion measurements ( see appendix [ sigcheck ] for a discussion ) .
deviations in @xmath13 from template to template are small and ranged from 2 to 20 km s@xmath14 .
since only one good template star ( sao162947 ) was available that was observed under the same fors instrumental configurations as the galaxies , three more k giant stars ( sao32042 , sao80333 & sao98087 ) were used which had been observed with mosca at the 3.5m - telescope at the calar alto observatory ( * ? ? ?
* a. fritz et al .
2009c , in preparation ) .
these templates have sufficiently resolved spectra with instrumental resolution of @xmath71 km s@xmath14 around and @xmath72 km s@xmath14 around the g - band . ] to determine the instrumental resolution of the galaxy spectra , the widths of 1214 unblended and not - saturated hgar / ne emission lines of the respective wavelength calibration frames were measured at wavelengths corresponding to and the g - band separately . in case of the fdf observations ,
the hgcd lamp has been used additionally to the he , ar and ne lamps to warrant better efficiency in the blue wavelength range which is particularly neccessary for the higher redshift spiral galaxies .
typical instrumental resolutions for the observed spectra fall in the range of @xmath73 km s@xmath14 . to ensure a reliable error treatment for the velocity dispersions and radial velocity measurements and to derived the signal - to - noise of each galaxy spectra monte - carlo simulations for different stellar templates over a range of input @xmath13 and @xmath22 ratios were performed @xcite .
for @xmath74 per the velocity dispersions are systematically underestimated on average by 4% , but for s / n@xmath378 per the effect becomes negligible . to check the reliablility and accuracy of our velocity dispersion measurements ,
serveral independent tests have been performed .
these verifications are presented in [ sigcheck ] of the appendix .
an example of the velocity dispersion measurement for a field elliptical galaxy in the whdf with typical s / n spectrum at @xmath75 is given in figure [ fig : fcqgs ] .
several template star spectra were smoothed to the same resolution as the galaxy spectrum to derive @xmath13 .
the velocity dispersion for this galaxy is @xmath76 km s@xmath14 derived within the extracted spectral region around the feature with a signal - to - noise level of @xmath77 @xmath14 ( see figure [ fig : fcqgs ] ) . in this case , the final @xmath78 was averaged over three individual measurements @xmath79 km s@xmath14 , @xmath80 km s@xmath14 and @xmath81 km s@xmath14 which are in excellent agreement as the scatter between the values varies only within their uncertainties .
the measured luminosity weighted average velocity dispersions @xmath82 were aperture corrected as @xmath83 ( similar to * ? ? ? * but adopting a slope of -0.06 ) to yield central velocity dispersions within a circular aperture of radius @xmath84 . for each galaxy , its aperture during the observations was individually computed using the slit width , the number of extracted rows over which the spectrum was averaged and the respective pixel scale ( see * ? ? ?
* for a description ) .
] ] figure [ fig : fwzdb ] shows the redshift distribution of all spectroscopically confirmed field early type galaxies detected within the fdf and whdf .
early type galaxies in the fdf that were spectroscopically identified to be members of a cluster at @xmath85 ( see next paragraph ) are indicated as well .
the vlt spectra of these cluster e+s0 candidates are of similar high quality to the spectra obtained for the field galaxies .
the 13 field e+s0 galaxies in the fdf encompass a range in redshifts of @xmath86 with a mean of @xmath87 ( @xmath88 ) and the 11 field galaxies in the whdf cover a redshift space of @xmath89 with a mean of @xmath90 ( @xmath91 ) .
the total field sample of 24 e+s0 galaxies exhibits a mean redshift of @xmath92 and a median of @xmath93 .
it turned out that the southwestern corner of the fdf most probably covers the outskirts of a galaxy cluster at @xmath94 @xcite .
based on the radial velocity measurements for the elliptical galaxy candidates in the fdf , the lower limit for the velocity dispersion of the cluster is @xmath95 km s@xmath14 , which is most likely only a lower limit as the cluster centre is not located on the fdf but only its outskirts . allowing for a spread in redshift space of @xmath96 , which corresponds to @xmath186 times the typical velocity dispersion of a rich galaxy cluster ( @xmath97 km s@xmath14 ) , it is found that a total of 13 spiral galaxies and 15 early - type galaxies based on their spectroscopic redshifts are likely to be members of this cluster .
this is also visible through a small break in redshift space at @xmath98 for the 341 galaxies at @xmath99 ( 144 with @xmath100 ) of the total spectroscopic catalog @xcite .
the spatial distribution of all early - type field galaxies in the fdf is shown in figure [ fig : fdfed ] .
potential cluster members are additionally denoted by squares . adding the possible spiral cluster candidates
, it is suggestive that the cluster has an elongated shape .
if the center indeed is positioned to the southwestern corner of the fdf , the distribution of the galaxies will be a striking example of the morphology density relation @xcite with the early type galaxies mainly populating the dense , inner region and the late type galaxies located at larger clustercentric radii .
unfortunately at a redshift of @xmath101 , the mg5170 absorption line is strongly affected due to the terrestrial absorption of the b band , which makes an accurate measurement of the internal galaxy velocity dispersions impossible .
for this reason , early - type galaxies which are possible members of a cluster at @xmath85 will be discarded from the further analysis .
nevertheless , these e+s0 cluster galaxies will be used as a reference for comparison reasons at a later stage , see [ stpops ] .
the coordinates , redshifts and rest - frame @xmath102 colours of the cluster galaxies are presented in table [ ccoord ] .
lccc i d & z@xmath103 & @xmath21 & @xmath104 + & & ( j2000.0 ) & ( j2000.0 ) + fdf-1161 & 0.3959 & 01 05 51.689 & -25 45 45.965 + fdf-4285 & 0.3956 & 01 06 02.961 & -25 44 55.832 + fdf-5011 & 0.6538 & 01 06 05.549 & -25 48 48.886 + fdf-5908 & 0.2237 & 01 06 08.796 & -25 48 06.816 + fdf-6307 & 0.4513 & 01 06 10.208 & -25 44 28.225 + fdf-6336 & 0.4593 & 01 06 10.334 & -25 45 16.169 + fdf-6338 & 0.4100 & 01 06 10.329 & -25 49 15.565 + fdf-6439 & 0.3955 & 01 06 10.669 & -25 42 44.437 + fdf-7116 & 0.4567 & 01 06 12.961 & -25 43 25.565 + fdf-7459 & 0.5386 & 01 06 14.172 & -25 45 04.429 + fdf-7796 & 0.4089 & 01 06 15.428 & -25 48 02.843 + fdf-8372 & 0.2290 & 01 06 17.554 & -25 44 50.593 + fdf-8626 & 0.4081 & 01 06 18.705 & -25 48 10.266 + whdf-92 & 0.3988 & 00 22 42.293 & 00 18 27.750 + whdf-111 & 0.7418 & 00 22 39.393 & 00 18 33.610 + whdf-158 & 0.4825 & 00 22 42.873 & 00 18 44.700 + whdf-173 & 0.4507 & 00 22 37.386 & 00 18 50.560 + whdf-318 & 0.3973 & 00 22 43.251 & 00 19 52.580 + whdf-437 & 0.2599 & 00 22 22.977 & 00 20 34.840 + whdf-508 & 0.3977 & 00 22 47.040 & 00 21 01.450 + whdf-749 & 0.2130 & 00 22 20.162 & 00 22 50.460 + whdf-810 & 0.2133 & 00 22 34.984 & 00 24 14.120 + whdf-946 & 0.2123 & 00 22 35.664 & 00 24 04.960 + whdf-810b & 0.2118 & 00 22 34.286 & 00 24 09.670 + lccccc i d & z@xmath103 & @xmath21 & @xmath104 & @xmath105 & @xmath106 + & & ( j2000.0 ) & ( j2000.0 ) & ( mag ) & ( rest ) + fdf-822 & 0.3357 & 01 05 50.371 & -25 48 19.922 & 23.17 & 2.11 + fdf-980 & 0.3340 & 01 05 50.986 & -25 48 07.142 & 24.01 & 2.07 + fdf-1114 & 0.3345 & 01 05 51.520 & -25 48 05.462 & 22.70 & 2.15 + fdf-1265 & 0.3340 & 01 05 52.103 & -25 47 47.219 & 23.33 & 2.12 + fdf-1371 & 0.3356 & 01 05 52.579 & -25 48 34.480 & 22.52 & 2.21 + fdf-1697 & 0.3365 & 01 05 53.736 & -25 48 13.855 & 22.63 & 2.24 + fdf-1836 & 0.3325 & 01 05 54.221 & -25 48 47.082 & 25.16 & 3.84 + fdf-2580 & 0.3390 & 01 05 57.080 & -25 48 08.880 & 22.67 & 2.18 + fdf-3051 & 0.3335 & 01 05 58.775 & -25 48 23.506 & 22.43 & 2.28 + fdf-3365 & 0.3312 & 01 05 59.842 & -25 48 13.029 & 23.92 & 2.04 + fdf-3909 & 0.3378 & 01 06 01.661 & -25 46 46.379 & 22.19 & 2.11 + fdf-4030 & 0.3360 & 01 06 02.098 & -25 47 33.302 & 23.48 & 2.18 + fdf-5362 & 0.3318 & 01 06 06.786 & -25 47 34.088 & 25.22 & 3.80 + fdf-5446 & 0.3355 & 01 06 07.131 & -25 45 39.069 & 23.67 & 1.76 + fdf-6457 & 0.3360 & 01 06 10.720 & -25 47 46.366 & 25.28 & 1.61 + lccccccccccc i d & z@xmath103 & @xmath13 & @xmath22 & @xmath105 & @xmath107 & @xmath108 & @xmath109 & @xmath110 & @xmath111 & @xmath112 & @xmath106 + & & ( km s@xmath14 ) & ( @xmath14 ) & ( mag ) & ( kpc ) & ( arcsec ) & ( mag arcsec@xmath16 ) & ( @xmath113 ) & ( @xmath113/@xmath114 ) & & ( rest ) + fdf-1161 & 0.3959 & 125@xmath11516 & 19 & 23.41 & [email protected] & 0.40 & [email protected] & [email protected] & 0.38 & @[email protected] & 1.87 + fdf-4285 & 0.3956 & 159@xmath11516 & 29 & 23.19 & [email protected] & 0.40 & [email protected] & [email protected] & 0.51 & @[email protected] & 1.87 + fdf-5011 & 0.6538 & 82@xmath11513 & 15 & 24.84 & [email protected] & 0.50 & [email protected] & [email protected] & 0.87 & [email protected] & 1.94 + fdf-5908 & 0.2237 & 286@xmath11505 & 127 & 19.71 & [email protected] & 2.50 & [email protected] & [email protected] & 1.03 & @[email protected] & 1.87 + fdf-6307 & 0.4513 & 185@xmath11514 & 29 & 23.52 & [email protected] & 0.59 & [email protected] & [email protected] & 0.74 & @[email protected] & 1.91 + fdf-6336 & 0.4593 & 140@xmath11543 & 8 & 24.69 & [email protected] & 0.50 & [email protected] & [email protected] & 0.96 & [email protected] & 1.99 + fdf-6338 & 0.4100 & 164@xmath11519 & 22 & 23.89 & & & & & & & 1.88 + fdf-6439 & 0.3955 & 200@xmath11515 & 41 & 22.44 & [email protected] & 0.35 & [email protected] & [email protected] & 0.42 & @[email protected] & 2.07 + fdf-7116 & 0.4567 & 123@xmath11521 & 28 & 24.21 & [email protected] & 0.25 & [email protected] & [email protected] & 0.28 & @[email protected] & 1.70 + fdf-7459 & 0.5386 & 172@xmath11523 & 20 & 24.28 & & & & & & & 1.85 + fdf-7796 & 0.4089 & 106@xmath11533 & 15 & 23.65 & [email protected] & 0.75 & [email protected] & [email protected] & 0.57 & @[email protected] & 2.02 + fdf-8372 & 0.2290 & 161@xmath11509 & 57 & 21.62 & [email protected] & 0.40 & [email protected] & [email protected] & 0.46 & @[email protected] & 2.01 + fdf-8626 & 0.4081 & 379@xmath11537 & 27 & 21.45 & [email protected] & 1.45 & [email protected] & [email protected] & 0.90 & @[email protected] & 1.95 + whdf-92 & 0.3988 & 183@xmath11513 & 31 & 22.53 & @[email protected] & 0.18 & [email protected] & [email protected] & 0.15 & @[email protected] & 2.06 + whdf-111 & 0.7418 & 172@xmath11512 & 23 & 22.88 & [email protected] & 0.91 & [email protected] & [email protected] & 0.16 & @[email protected] & 2.05 + whdf-158 & 0.4825 & 134@xmath11514 & 37 & 21.28 & [email protected] & 1.28 & [email protected] & [email protected] & 0.26 & @[email protected] & 1.85 + whdf-173 & 0.4507 & 184@xmath11516 & 37 & 22.51 & [email protected] & 0.88 & [email protected] & [email protected] & 0.60 & @[email protected] & 2.08 + whdf-318 & 0.3973 & 211@xmath11514 & 32 & 22.25 & [email protected] & 1.91 & [email protected] & [email protected] & 1.17 & [email protected] & 2.02 + whdf-437 & 0.2599 & 128@xmath11508 & 57 & 21.67 & [email protected] & 1.10 & [email protected] & [email protected] & 0.63 & @[email protected] & 1.89 + whdf-508 & 0.3977 & 138@xmath11539 & 9 & & [email protected] & 2.92 & [email protected] & [email protected] & 1.35 & [email protected] & 2.01 + whdf-749 & 0.2130 & 130@xmath11511 & 52 & 21.10 & & & & & & & 2.66 + whdf-810 & 0.2133 & 186@xmath11510 & 83 & 20.28 & [email protected] & 1.54 & [email protected] & [email protected] & 0.77 & @[email protected] & 2.00 + whdf-946 & 0.2123 & 107@xmath11505 & 49 & 21.63 & [email protected] & 1.47 & [email protected] & [email protected] & 0.82 & @[email protected] & 2.07 + whdf-810b & 0.2118 & 140@xmath11519 & 20 & 21.59 & [email protected] & 0.39 & [email protected] & [email protected] & 0.46 & @[email protected] & @xmath1160.47 + figure [ fig : whdfed ] illustrates the spatial distribution of the early - type field galaxies in the whdf .
apart from three galaxies which are located in the northern part of the image , all field ellipticals are homogenously distributed over the entire field . to test for possible group candidates , radial velocity measurements of
the three galaxies close in projection resulted in a dispersion in radial velocities of @xmath117 km s@xmath14and a group velocity dispersion of @xmath118 km s@xmath14at @xmath119 .
looking at the individual acs galaxy images , these objects are not visually associated to each other through tidally distorted outer structures ( e.g. tidal arms , collisional rings , interaction signatures ) or very nearby other companions .
for this reason , both independent approaches give strong evidence that the three objects are not likely members of a group but isolated field galaxies .
the coordinates and redshifts of the galaxies for which velocity dispersions were measured are given in table [ fwcoord ] .
velocity dispersions , s / n ratios , total johnson @xmath2-band magnitudes , sizes , surface brightnesses , dynamical masses , @xmath5 ratios , and rest - frame @xmath102 colours of the galaxies are presented in table [ phy ] . to estimate the amount of recent star formation present in our field galaxies we measure the spectral diagnostic features of @xmath1203727 and @xmath1204101 in the observed wavelength range of the spectra . for the indices @xmath1203727 and @xmath1204101 we adopt the passband definitions by @xcite and @xcite , respectively .
uncertainties in the line index measurements account for poisson noise as well as possible noise variations in the spectra . in the subsequent analysis
we always refer to rest - frame equivalent widths in units of .
the spectroscopic measurements of our galaxies will further be discussed in [ sec : sfd ] , where the implications of the recent activity signatures on the stellar population properties will be assessed
. llcccc field & date & area & @xmath121 & m@xmath122 & & arcmin@xmath123 & [ sec ] & @xmath124 fdf & 20/07 - 29/09/02 & @xmath125 & 2360 & 25.50 + whdf & 14/08 - 23/08/03 & @xmath125 & 2450 & 26.30 +
using the advanced camera for surveys ( acs ) wide field channel ( wfc ) , _ hubble space telescope ( hst ) _ observations of the fors and william herschel deep fields were carried out during cycles 11 and 12 . to cover the @xmath126arcmin@xmath123 sky areas of the fdf and whdf , four wfc pointings were aligned as a @xmath127 mosaic .
a single acs wfc chip offers a field of view of @xmath18@xmath128arcsec@xmath123 and a pixel scale of @xmath180.049arcsec .
each pointing was split into two exposures and has a total exposure time of 2360s ( fdf ) and 2450s ( whdf ) through the f814w ( @xmath43 ) filter .
table [ acsobs ] gives an overview of the hst / acs observations of the fdf and whdf .
the column @xmath121 represents the total exposure time for each object and m@xmath122 lists the 50% completeness limit for @xmath124 detections of extended sources within the two deep fields .
data reduction was performed using the standard calacs pipeline and the unprocessed images ( including bias and dark subtraction , flat fielding , distortion and astrometric correction ) were used for an additional cosmic ray rejection filtering by combining the two exposures of each visit to produce the final mosaic .
structural parameters of half - light ( effective @xmath129 ) radii , ( effective ) surface brightnesses ( @xmath130 within @xmath129 ) and total magnitudes and shapes of the galaxies were derived with the galfit package @xcite . for the convolution of the surface brightness model profiles ,
an average psf was constructed using @xmath1820 unsaturated stars with @xmath131 which were normalized to the same central flux and afterwards median averaged .
two - dimensional surface brightness profiles of the galaxies were analysed with four different fitting functions , by a pure classical de vaucouleurs @xmath132 , a pure @xcite with variable srsic exponent @xmath133 ( @xmath134 ) , and each of those profiles in combination with an exponential disc component .
the best fit to the observed light distribution of a galaxy was assessed by comparing the residual images and reduced @xmath135 values .
note that by definition a profile consisting of a srsic plus disc component results in lower @xmath135 numbers than using a single @xmath132 model alone .
therefore , preference was given to those model profiles with yielded to the lowest residual structures in the residual images . in 40% of the cases
this was a pure classical de vaucouleurs profile , in 60% of the cases this was a srsic plus disc component .
figure [ sbfr ] presents examples of the results of the surface brightness modelling for our field galaxies . for each galaxy , the original image , the best 2d model fit and the residuals from an @xmath132 or @xmath136 + exponential disc component are shown . for internal consistency
, the surface brightness distribution of the field galaxies were also modelled with the gim2d package by @xcite and the algorithm by @xcite .
a detailed discussion on differences between these techniques can be found in @xcite and @xcite .
total magnitudes derived with all three methods agree well within the errors as well as with those values measured on the ground - based @xmath43-band images of the fdf and whdf with the sextractor package .
observed magnitudes of the galaxies were calibrated on to the vega system using the the synthetic transformation by @xcite : @xmath137 where the first term is the observed magnitude expressed as total count rate per electrons and seconds in the source system , tmag and tcol being the observed magnitude in the target system and the @xmath138 , @xmath139 , and @xmath140 the transformation coefficients .
we adopt our @xmath27 aperture colors and for the coefficients @xmath141 , @xmath142 , and @xmath143 . for the fdf and whdf a galactic absorption of @xmath144 and @xmath145
was derived , respectively .
aperture colours were measured on the ground - based @xmath32 and @xmath33 band photometry within a 2 diameter aperture .
galaxy morphologies were classified using a combination between a visual inspection by one of us ( a.f . ) , the results of the surface brightness profile analysis ( i.e. interpreting the residuals from the @xmath132 luminosity profile fitting ) and an independent analysis using gim2d ( see * ? ? ?
* for a description ) .
the information of all three independent approaches was combined to yield the best and most reliable decision on the final morphologies of the galaxies .
figure [ fdfacsthumb ] and figure [ wdhfeacsthumb ] display hst / acs thumbnail images in the f814w filter of the fdf and whdf field early - type galaxies , respectively .
table [ fdfacsrem ] and table [ whdfacsrem ] list the morphologies and noteworthy remarks of special features for the individual galaxies .
the first column of each table shows the galaxy i d , the second indicates the galaxy morphology and the last column gives additional information about special features to the respective galaxy . in total , the sample is divided into 13 field ellipticals ( e and e / s0 ) and nine field lenticulars ( s0 to sa bulges ) .
the two classes comprise the sub - types of 6 e , 1 e / s0 , 3 s0 and 2 sa bulges in the fdf , and 4 e , 2 e / s0 , 1 s0 , 1 s0/sa and 2 sa bulges in the whdf .
no acs image is available for the galaxy id6338 in the fdf , based on its spectrum this object has a morphology of a spiral galaxy ( see also table [ fdfacsrem ] ) .
the findings from the luminosity profile fitting provided a consistency check for the visual inspection and resulted in the same classification scheme except for one field galaxy in the whdf , id810b : s0 ( visual ) , e ( model fit ) .
lcl i d & morp & note + 1161 & s0 & + 4285 & e & galaxy with dust lane , merger of s0 & s ? + 5011 & e & + 5908 & e & + 6307 & e & model too boxy , residuum at core + 6336 & sa & sb ? , peculiar structure , warps + 6338 & s & no acs image , no analysis , morph based on spec + 6439 & s0 & + 7116 & e & hot pixel + 7459 & e & no analysis + 7796 & sa & peculiar structure , sf in upper spiral arm ?
, warps + 8372 & s0 & hot pixel + 8626 & e / s0 & close to edge of chip + lcl i d & morp & note + 92 & s0 & two background ?
objects in south + 111 & e / s0 & external halo in upper north part , two objects in north - west + 158 & sa & + 173 & s0 & + 318 & e & close to star , star in center + 437 & e & face on , faint star in north - east , [ oii]3727 emission + 508 & sa & peculiar , dust pattern ? , rot . ? ,
[ oii]3727 emission ? ,
star + 749 & e / s0 & no acs image , morphology based on spectrum + 810 & e & boxy isophotes , stars in center , [ oii]3727 emission + 810b & e & discy structure , faint , [ oii]3727 emission + 946 & s0/sa & lensed object ( maybe in projection ) in north - east + @xmath146-corrections were calculated by convolving respective filter transmission curves with the redshifted spectral energy distribution ( sed ) of an early - type galaxy template from @xcite as described in @xcite and @xcite .
no separation between elliptical and s0 types was done because both exhibit similar ( optical and uv ) spectral shapes , balmer discontinuities ( 4000 ) and absorption features for @xmath147 which are indistinguishable from their broadband optical and near - infrared colours . for internal consistency ,
synthetic spectral sed templates of an 12gyr elliptical galaxy which were generated through evolutionary synthesis modelling by @xcite were also utilised .
typical deviations in the @xmath146-corrections between the seds of the two template libraries are small @xmath148 . depending on the redshift of the object
the most suitable filter was chosen to transform to rest - frame johnson @xmath2-band .
the @xmath149 filter was selected for galaxies in the fdf with @xmath150 , @xmath151 for @xmath152 and @xmath153 for @xmath154 . for field objects in the whdf with @xmath150 a transformation to @xmath2 rest - frame
was conducted based on the @xmath2 passband and for @xmath155 the @xmath156 filter was utilised .
therefore , the @xmath146-correction was much less sensitive to spectral type than using a global filter transformation that is restricted to a specific filter , e.g. of @xmath157 . for reasons of consistency ,
the rest - frame magnitudes were checked by transforming the observed f814w - magnitudes into rest - frame johnson-@xmath2 .
deviations between magnitudes were small ( differences were less than their errors ) .
total errors on the absolute @xmath2-band magnitudes for the galaxies were computed as the quadratic sum of individual errors in the respective ground - based photometry filter ( either @xmath2 , @xmath49 or @xmath43-band ) , the error in the @xmath146-correction and the uncertainty in the galactic absorption correction .
thanks to the very deep imaging of the fdf and whdf , the photometric errors in the ground - based photometry are very small for all filter passbands ( median uncertainties @xmath158 ) . for the fdf and whdf field sample
, the total absolute error falls into the range @xmath159 with a median of 0.15mag . in figure [ mrs3 ] the distribution of the early - type field galaxies in the fdf and whdf in absolute johnson @xmath2-band magnitude
are shown . only the rest - frame magnitudes derived with the fdf and whdf photometry
are shown .
the 24 e+s0 field galaxies cover a range in absolute @xmath2-band magnitude of @xmath160 ( @xmath161 ) .
the dashed arrow corresponds to the mean value ( @xmath162 ) , whereas the median is denoted by the solid arrow . ] ] to acquire independent complementary information on nuclear activity in our early - type galaxy sample we checked the archive for possible x - ray counterparts of our field galaxies . in the selection process of a suitable sky area for the fdf , one particular constraint was that the field had to be devoid of strong radio and x - ray sources to avoid potentially present galaxy clusters at medium redshifts .
chandra _ and _ xmm - newton _ observatories have not observed the sky region covered by the fdf ; therefore , we can not look for the presence of possible active galactic nuclei ( agns ) in these field galaxies .
for the whdf two _ chandra _ surveys are available ( pi : t. shanks , pid 02900499 ) . a deep , high resolution acis - i imaging survey ( 72.13 ksec ) of the central field and a shallow survey ( 10 fields with exposure times of 10 ksec each ) with the acis - s around narrow line galaxies that have previously been detected by rosat .
the acis - i covers the spectral range between 0.2 to 10 kev and has a spatial resolution of @xmath180.492 or better , with psf variations less than 0.2 across the total [email protected] field - of - view @xcite . through a cross - correlation with our optical photometry we identified 28 objects as x - ray point - sources above a @xmath164 source detection limit ( 95% confidence level ) .
the detected counts for each source were converted into fluxes , assuming a power law of the photon spectral flux distribution of @xmath165 with an intrinsic photon index of @xmath166 and column density of the galactic absorption of @xmath167 cm@xmath16 .
a photon index of @xmath166 is typical for x - ray emission that has its origin from star formation .
the on - axis sensitivity limits of the ( unabsorbed ) flux in the soft ( @xmath169 kev ) and hard ( @xmath170 kev ) bands are @xmath171 ergs cm@xmath16 s@xmath14 and @xmath172 ergs cm@xmath16 s@xmath14 , respectively . these fluxes correspond to a detection limit in x - ray luminosities emitted in the soft band of @xmath173 ergs s@xmath14 and in the hard band of @xmath174 ergs s@xmath14 .
the _ chandra _
exposures are deep enough to both fully resolve the x - ray background ( xrb ) at about @xmath175 ergs cm@xmath16 s@xmath14 in the @xmath176 kev energy band and detecting all the hard x - ray sources within the whdf . very recently a shallow _ xmm - newton _ survey ( pi : m. guainazzi , pid 0407030101 ) targeted a similar sky region ( 27.91 ksec ) .
however , these observations cover only the southern part of the whdf and are at a significantly coarser angular resolution of @xmath1815than the _ chandra _ detections . using the same photon spectral flux distribution configuration as above the sensitivity limits of the ( unabsorbed ) flux in the @xmath169 kev energy band are @xmath177 ergs cm@xmath16 s@xmath14 and in the @xmath170 kev range @xmath178 ergs cm@xmath16 s@xmath14 .
the soft and the hard fluxes correspond to detection limits of @xmath179 ergs s@xmath14 and @xmath180 ergs s@xmath14 .
figure [ fig : chandra ] shows the distribution of _ chandra _ and _ xmm - newton _ detections in the whdf .
three galaxies that were observed by _
chandra _ were additionally detected by us on the _ xmm _ images ( see figure [ fig : chandra ] ) .
apart from two clear chandra detections of two sc spiral galaxies ( id438 ) at @xmath181 and ( id635 ) at @xmath182 ( no maximum rotation velocity @xmath183 could be derived for both objects ) , four field early - type galaxies might have associated x - ray counterparts . these objects in the whdf with possible chandra and/or xmm counterparts are listed in table [ xrct ] .
there is only one rather likely candidate for a galaxy with a central agn ( id437 ) as this object shows also weak emission . in case of the other candidates either bright
( id173 ) or bright and extended ( id92 or 158 ) _ xmm _ sources close to the galaxies are found that might be associated to the optical counterparts .
non of the early - type galaxies in our sample is surrounded by a diffuse halo - gas , thus no object appears to be an extended x - ray source , hence the influence of agn activity for our galaxies is small or even negligible .
we will discuss the possible impact of x - ray detections together with our results in section [ agn ] . in appendix
[ xrayp ] , we assess in more detail contributions due to low - luminosity agn ( llagn ) and/or emission from soft , diffuse x - ray plasma or or low - mass x - ray binaries ( lmxbs ) in case of our galaxies .
rccll i d & morp & chandra / xmm & association & note 92 & s0 & yes?/no & maybe & xmm bright , extended + 158 & sa & no / yes ? & not likely & xmm bright , extended + 173 & s0 & yes?/yes ? & likely & xmm bright + 437 & e & yes?/yes ? & maybe & [ oii ] emission + 438 & sc & yes / no & yes & @xmath181 , no @xmath183 + 635 & sc & yes / no & yes & @xmath182 , no @xmath183 +
to derive the kinematic and/or the spectrophotometric evolution of early - type galaxies at intermediate redshift , it is crucial to carefully select a sample of early - type galaxies at low redshift which can be utilised as a reference for the purpose of comparison .
the coma cluster at @xmath184 is one of the best studied local rich clusters of galaxies and represents with a look back time of only @xmath180.3 gyr ( corresponding to 2% of the age of the universe ) the present - day universe .
a number of previous works in the literature on both cluster and field early - type galaxies ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) used this cluster for the purpose of comparison of the fundamental plane at @xmath185 and it therefore provides a reliable and widely accepted local reference when addressing evolutionary questions . in order to minimise systematic uncertainties arising from filter transformations , distance determinations and selection effects , we utilise the well defined coma sample in the johnson @xmath2-band by ( * ? ? ?
* hereafter sbd93 ) which comprises 39 early - type cluster galaxies ( splitted into 25 e and 14 s0s ) .
this data compilation is a sub set of the original `` 7 samurai '' sample @xcite which contains 59 e+s0 galaxies in the virgo and coma cluster ; but for the latter no morphological information is available . for deriving the evolution of the @xmath5 with redshift we use the fp coefficients as published by @xcite .
early - type galaxies follow a scaling relation between their luminosity and internal velocity dispersion , also known as the faber jackson relation @xcite .
the relationship can be used to constrain the formation and evolution of e+s0 galaxies but it requires accurate velocity dispersion measurements as small uncertainties in the kinematics transfer into large errors in the magnitudes , hence the measured luminosity evolution .
figure [ fig : fjr ] shows the faber
jackson relation ( fjr ) in the @xmath2-band for the field early - type galaxies in the fdf and whdf , compared to the local coma sample by sbd93 . for this scaling relation ,
galaxy magnitudes were measured with sextractor on the deep ground - based images with @xmath180.60arcsec fwhm in the @xmath43-band , hence the whole field sample can be considered .
as the median uncertainty in our velocity dispersion measurements is only 15 km s@xmath14and our magnitude limit reaches @xmath186 ( about one magnitude fainter than previous studies ) , we can put reliable implications on the evolution of our distant field galaxies using the fjr . a linear @xmath135-fit to the restricted coma reference sample yields : @xmath187 which is indicated as a dashed line in figure [ fig : fjr ] . assuming that the local slope holds valid for our distant galaxies ( i.e. there is not evolution in the slope of the fjr )
, we analyse the mean residuals from the local fjr , shown as the solid line in figure [ fig : fjr ] . because of the on average younger light - averaged ages of the stellar populations in the distant galaxies , for a given velocity dispersion
the field galaxies are on average brighter by @xmath188 mag ( median @xmath189 mag ) than their local counterparts .
the @xmath190 scatter of the offsets in the fjr for the distant galaxies is @xmath191 .
the two objects that are slightly offset from the rest of the distant field galaxies show both signs of star formation , one of them being also the galaxy with the highest redshift in our sample ( cf .
[ sec : sfd ] ) .
the measured evolution is in very good agreement with a slow , passive evolution of the stellar populations as suggested by single - burst simple stellar populations models with a high formation redshift of the bulk of the stars @xcite . for formation redshift of @xmath192
these evolution models predict an average increase in the johnson @xmath2-band of @xmath193 mag at @xmath194 .
similar as for local studies , our distant galaxies show weak evidence for a larger scatter at fainter luminosities ( @xmath195 ) , that might be indications for an evolution of slope in the fjr .
we interpret this effect as a stronger evolution of the lower - mass galaxies .
however , apart form the mass dependence , an additional small contribution to the offset of lower - luminosity galaxies might be the result of a weak dependence of the velocity dispersion on luminosity of the early - type galaxies that was previously found within a large sample of early - type galaxies drawn from data release 4 of the sloan digital sky survey ( sdss , * ? ? ?
this luminosity dependence becomes detectable in all fp projections and is most pronounced for faint galaxies in the sdss around @xmath196 which corresponds to about @xmath197 . at @xmath198
our field distant galaxies show a diversity , half of the sample indicates a strong luminosity evolution whereas the other half a mild evolutionary trend .
an un - constrained linear @xmath135-fit to our field galaxy sample yields : @xmath199 however , our sample is too small and the dependence on luminosity too weak to test possible luminosity effects on the slope of the fjr at @xmath194 in greater detail . to verify our results and to test the underlying stellar populations in even more detail , we converted the extinction - corrected @xmath26 rest frame colors of the field galaxies in our sample into @xmath5 ratios in the @xmath3-band following a similar approach as in @xcite .
absolute @xmath3-band magnitudes were derived by using the synthetic stellar population models by @xcite .
the fjr in the @xmath3-band for the field e+s0 galaxies in the fdf and whdf is presented in figure [ fig : fjrk ] .
the early - type galaxies are compared to the local relationship of cluster early - type galaxies as found by @xcite . for a given velocity dispersion
the luminosities of the distant field galaxies ( excluding id111 at @xmath200 ) are on average brighter by @xmath201 mag ( median @xmath202 mag ) than their local representatives .
the moderate luminosity evolution found for the field e+s0 is in excellent agreement to the findings derived using the @xmath2-band magnitudes .
less - massive galaxies display a larger offset with respect to the local relationship , confirming the faster evolution of less - massive galaxies in the @xmath2-band .
furthermore , lower - luminosity ( mass ) galaxies show a faster evolution in luminosity than higher - luminosity ( mass ) galaxies .
this point will be addressed in greater detail in [ stpops ] . to study the change of surface brightness evolution for the fp sample at a fixed size , the magnitude
size relation was constructed . the kormendy relation ( kr )
represents the projection of the fundamental plane along the velocity dispersion onto the photometric plane .
as only the structural parameters are used , samples have larger scatter in the kr than in the fp but results should complement and endorse findings obtained with the fp . the kr of the field galaxies is presented in figure [ fig : krb ] . for a given size
the surface brightnesses of the distant field galaxies ( excluding the sa id508 ) are on average brighter by @xmath203 mag arcsec@xmath16 ( median @xmath204 mag arcsec@xmath16 ) .
one object ( id508 ) that does not follow the kr is a spiral galaxy , has a high surface brightness and also shows possible signs of rotational support in its two - dimensional spectrum ( cf .
[ fp ] ) .
we confirm the power of the kr to separate passive quiescent galaxies from galaxies with blue colors as found by @xcite . in consistency with passive evolution models
, the fading of the stellar populations implies that the galaxies will evolve to brighter surface brightnesses with increasing redshift , hence moving away from the local kr of present day early - type galaxies .
there are no indications of an evolution in the sizes of our field galaxies , see figure [ fig : krb ] .
we emphasise that it is hard to separate an evolution in the size of the galaxies from an evolution in the surface brightness as the photometric parameters of half - light radius and surface brightness are correlated and enter the fp and its projections in combination as @xmath205 , with @xmath206 ( e.g. , * ? ? ?
* ; * ? ? ?
the measurements and results derived for the kr depend critically on the selection procedures applied to the sample galaxies . for example , two studies targeting the same sky region obtained inconsistent results , @xcite claims an evolution in the size of galaxies in the cdfs whereas @xcite does not support this finding .
most likely this effect is due to differences or constraints set in the selection process , where the former primarily selected fainter objects whereas the latter observed a more homogenous population of early - type galaxies . in case of the low - mass galaxies a differential evolution
can be produced by a combined effect of large spectroscopic apertures and substantial rotational support in the target objects .
in particular , the combination of the large slit - apertures with low s / n galaxy spectra by @xcite can cause miss - classifications of spirals with unresolved possible rotational support as detectable in the emission lines .
this effect points to a bias in the selection procedure .
a further discussion on possible selection effects is presented in [ sec : litcmp ] .
early - type galaxies in the nearby universe follow a tight fundamental relationship which is defined by their basic observables as @xmath207 where @xmath108 is the effective ( half - light ) radius given in kpc , @xmath13 denotes the internal velocity dispersion in km s@xmath14and @xmath208 is the surface brightness in mag arcsec@xmath16 , related to @xmath15 as @xmath209 . for the fp coefficients we performed a principal component analysis and derived for the slopes and the intercept @xmath210 , @xmath211 and @xmath212 , which are in good agreement with the fp coefficients by sbd93 ; @xcite . from the sample of field galaxies of @xcite we measured that the intercept of the fp is offset by @xmath213 from the relation as given in equation [ fprel ] .
figure [ fpfw ] shows the fp for the 21 early - type field galaxies in the fdf and whdf in the rest - frame johnson @xmath2-band , compared to the 39 early - type galaxies in the coma cluster by sbd93 .
the local reference is indicated with small squares , whereas the distant field galaxies are shown as the large circles and stars .
filled symbols denote ellipticals , open symbols s0 galaxies and sa bulges .
the 21 fdf and whdf early type field galaxies are morphologically classified into 9 e and 12 s0 .
five ellipticals and six s0s originate from the fdf and four es and six s0s are located in the whdf .
two galaxies are not visible on the acs images and for one elliptical galaxy in the fdf no structural parameters could be derived .
the distant field galaxies cover a redshift range of @xmath214 with a median of @xmath215 .
two outliers in the fp , id6336 ( @xmath216 ) and id508 ( @xmath217 ) , reveal on the acs images a significant disc component and early spiral morphology .
galaxy id6336 has an extraordinary strong h@xmath218 absorption , which points to a young luminosity - weighted average model age under assumption of a single stellar population . on the acs images
, id508 shows a peculiar structure which might indicate that this object is an ongoing merger .
both galaxies were classified as sa bulges hence these early - type spirals offer at least a low level of star formation which increases their total luminosity and displaces their position beyond the tight plane as established by early type galaxies .
rejecting the two outliers , for the total sample of 19 early - type field galaxies in the fdf and whdf an average luminosity evolution of @xmath219 with a median of @xmath220 is found .
similar results are measured for 17 cluster ellipticals at @xmath221 @xcite with similar look - back times ( @xmath185 gyr ) , which show an average brightening of their stellar populations of @xmath222 , compared to their local counterparts .
overall , the distribution along the fp and the measured luminosity evolution is similar for cluster and field galaxies at @xmath221 and within the cluster sample no differences between ellipticals and s0 galaxies were found .
however , the sample of distant cluster galaxies lacks galaxies with small sizes and has a magnitude limit of @xmath223 , whereas the field galaxy sample comprises no ellipticals with large sizes ( i.e. cd galaxies ) and is more than one magnitude fainter ( @xmath1 ) .
a comparison of the sizes and luminosities of these two data sets is therefore not adequate and would cause misleading results .
table [ fptabf ] gives a summary of the results for the fp of our early - type field galaxies .
lrcccc sample & @xmath224 & @xmath225 & @xmath226 & @xmath227 & @xmath228 fdf & 11 & 0.066 & 0.075 & @xmath1160.235 & 0.136 whdf & 10 & 0.199 & 0.214 & @xmath1160.667 & 0.289 fdf+whdf & 21 & 0.130 & 0.089 & @xmath1160.279 & 0.227 fdf+whdf@xmath229 & 19 & 0.170 & 0.126 & @xmath1160.393 & 0.197 e & 9 & 0.075 & 0.075 & @xmath1160.230 & 0.121 s0 & 12 & 0.171 & 0.211 & @xmath1160.643 & 0.281 s0@xmath229 & 10 & 0.255 & 0.241 & @xmath1160.735 & 0.219 s0@xmath230 & 9 & 0.214 & 0.211 & @xmath1160.643 & 0.185 low lum.@xmath231 & 11 & 0.056 & 0.082 & @xmath1160.255 & 0.180 high lum .
& 10 & 0.211 & 0.165 & @xmath1160.515 & 0.254 high lum.@xmath232 & 9 & 0.164 & 0.126 & @xmath1160.393 & 0.219 low mass@xmath233 & 11 & 0.099 & 0.089 & @xmath1160.279 & 0.221 high mass & 10 & 0.163 & 0.126 & @xmath1160.393 & 0.240 + + + + + the thickness of the fp appears to vary between the individual field galaxy samples . for the fdf an average rms scatter of 0.136 is detected , whereas the whdf show a larger 1@xmath13 dispersion of 0.289 .
this effect is partly caused by three high redshift objects @xmath234 in the whdf , which are all disc - dominated galaxies , without these object the rms scatter is 0.202 . dividing the field sample with respect to velocity dispersion at log@xmath235 or luminosity at @xmath236 , no strong variations in the amount of evolution between the sub
samples are detected .
however , the comparisons rely on the assumed cutoffs in @xmath13 and @xmath237 because outliers , such as early sa spiral galaxies , can strongly influence the derived luminosity evolution . in the next section
, possible differences between the stellar populations of elliptical and lenticular galaxies will be addressed .
looking at differences between the morphological types of _ field _ early - type galaxies , the s0 galaxies display a stronger evolution than the elliptical galaxies .
assuming that the slope of the local reference holds valid for the distant galaxies , the 9 field ellipticals show in edge - on projection of the fp a zero - point offset @xmath238 which corresponds to a brightening in their stellar populations of @xmath239 . by comparison ,
the nine field lenticulars exhibit an offset in the zero - point with respect to the local coma galaxies @xmath240 which corresponds to an average evolution of @xmath241 .
note , that this fp offset is very similar if we include the two sa bulges and results in the same amount of evolution if also the highest redshift lenticular galaxy is accounted for , see table [ fptabf ] for details . errors on the zero - points of elliptical and s0 galaxies were individually derived as @xmath242 where @xmath243 denotes the total error which enters the fp in the rest - frame @xmath2-band and @xmath244 is the uncertainty computed through an iterative bootstrap re sampling of the data points 100 times @xcite . limiting the s0 galaxies to redshifts @xmath245 and rejecting the two sa bulges , the same results are derived ( see table [ fptabf ] ) . in addition , the field ellipticals obey a tight fp with a small 1@xmath13 scatter of 0.121 , whereas the s0 types have a larger dispersion of 0.219 ( omitting the two sa bulges ) .
this gives further evidence that the lenticular galaxies have a larger range of different stellar populations .
a faster evolution of the @xmath5 ratios ( see next section [ stpops ] and equation [ dml ] for a description of the connection between the fp and @xmath5 ratios ) of the s0 galaxies with respect to elliptical galaxies was also found for _ cluster _ early - type galaxies at @xmath246 @xcite , which are in agreement with a low formation redshift @xmath247 . as the slope of the @xmath248 relation for these distant cluster galaxies is steeper than those for the coma cluster galaxies ,
this can be interpreted as a dependence of the evolution on galaxy mass , and hence a downsizing effect in the evolution of these galaxies .
the evolution of the fp for the early - type fdf and whdf field galaxies in the rest - frame johnson @xmath2-band is illustrated in figure [ fpfw ] .
early - type galaxies were binned in redshift space to investigate their location within and along the edge - on fp as a function of redshift and to test the effects of possible outliers .
ellipticals are represented as filled , s0s and sa bulges as open symbols . for each morphological type
the number of galaxies within a redshift bin is indicated in the brackets .
small squares denote the coma galaxies and the straight line is a principal component fit to the local sample @xcite .
figure [ fpfw ] clearly shows the evolution of the fp for the distant galaxies with respect to the local coma fp .
the offsets of the distant field galaxies increase with redshift but the scatter appears to be mainly amplified for lenticular galaxies .
the two galaxies with a `` positive '' evolution in the last two panels are the sa bulges id6336 and id508 and show a clear disc on the acs images .
these objects can mimic on average a weaker evolution in luminosity for the whole galaxy population ( cf .
table [ fptabf ] ) .
regardless of the sa bulges , the s0s exhibit a larger scatter at higher redshift @xmath249 than the ellipticals which suggests that the stellar populations of e and s0 types are different and that lenticulars are a more heterogeneous group . an explanation could be that these s0 galaxies resemble post starburst galaxies as no strong emission lines were detected in their spectra .
this statement will be further addressed in [ sec : sfd ] .
the observed evolution of the @xmath5 ratio as derived from the fp depends on the age of the stellar population of the galaxies . in general
, the luminosity of a young stellar population becomes rapidly fainter when the massive and bright stars which have a short lifetime disappear .
for an old population comprising mainly low mass stars the dimming of the luminosity proceeds on a more gradual evolutionary path . as a consequence of this
, a stellar population formed at lower redshift will evolve faster than one generated at high redshift . in the following
, the observed evolution of the @xmath5 ratio will be compared to simple stellar population models of a single burst formed at redshift @xmath250 .
the models have been generated following the description of the analytic models by @xcite .
the luminosity evolution of a single - age stellar population can be described by a power law as @xmath251 where @xmath252 is the stellar formation time which corresponds to a formation redshift of @xmath250 ( e.g. , * ? ? ?
the coefficient @xmath253 depends on the imf , the metallicity and the chosen passband in which the luminosity is measured . in the stellar population models by @xcite , a normal imf with @xcite slope , solar metallicity and @xmath254 for the rest
frame @xmath2-band was adopted .
model predictions are shown in figures [ dmmlrbz ] to [ dmlrbz ] for the field elliptical and s0 galaxies in the rest
frame @xmath2-band .
note that the predicted evolution of the models is independent of @xmath255 as the age dependence of the @xmath5 ratio is a power law .
two different model tracks for a single stellar population with formation redshift @xmath256 ( lower line ) and @xmath257 ( upper line ) are indicated as the dotted lines in figures [ dmmlrbz ] to [ dmlrbz ] . to assess the evolution of the stellar populations in the early type galaxies
, the evolution of the fp zero - point @xmath258 for the distant galaxies can be directly converted into an evolution of the average @xmath5 ratio as @xmath259[dml ] where the average evolution of @xmath260 was derived via the average zero - point offset of the distant fp @xmath261 from the local coma fp @xcite . in a recent study of 27 field early type galaxies between @xmath262 a mass dependent evolution
was found @xcite .
this sample of high redshift galaxies showed that the evolution of low mass field galaxies with a characteristic mass of @xmath263 is faster than their more massive counterparts .
similar results have been drawn from field early type galaxies at @xmath6 from the goods - n sky region @xcite and from the k20 survey @xcite .
the 21 field early type galaxies in our sample are located at a lower redshift range of @xmath264 .
but it is very interesting , if already at a look back time of @xmath185 gyr a possible evidence for a mass dependent evolution or at least a trend with galaxy mass can be detected .
for this reason , the average evolution of the @xmath5 ratio of the field elliptical and s0 galaxies in the fdf and whdf was derived .
figure [ dmmlrbz ] displays the average evolution of the @xmath5 ratios for the field galaxies in the rest
frame @xmath2-band compared to the early type cluster galaxies taken from @xcite . for a compilation of massive early type galaxies with @xmath265 in rich clusters of galaxies taken from the literature these authors deduced an average formation redshift of @xmath266 , which is indicated as the solid line .
all calculated model tracks are forced to go through the average @xmath5 ratio of the field galaxies at @xmath267 by @xcite , shown as the small dot in figures [ dmmlrbz ] and [ dmlrbz ] .
the field fdf and whdf sample was splitted according to the characteristic mass at @xmath268 , which gives 13 lower mass and 8 higher mass galaxies .
the average evolution of the higher mass field galaxies is @xmath269 at a redshift of @xmath270 .
lower mass galaxies indicate on average a stronger evolution of @xmath271 at @xmath272 .
the slight deviation in the redshift is solely due to the not uniform number of galaxies in each sub sample .
massive field galaxies are in compliance with a slow evolution as derived for the massive cluster galaxies and a high formation redshift for the bulk of the stars of @xmath266 . in contrast ,
less massive field galaxies in the fdf and whdf with @xmath263 evolve faster and follow an evolutionary track between @xmath273 which is similar to the less massive cluster galaxies from @xcite , which is indicated as the open square in figure [ dmmlrbz ] .
this result gives evidence that already at a look back time of @xmath185 gyr differences in the measured @xmath5 evolution for the fdf and whdf field galaxies can be revealed . to translate the evolution in the @xmath5 ratios into a specific formation redshift @xmath250 for the field galaxies ,
the evolution of the @xmath5 ratios for the individual field early type galaxies are investigated .
figure [ dmlrbz ] illustrates the offset from the local fp in the rest
frame @xmath2-band for elliptical and s0 galaxies in the fdf and whdf . a comparison sample of 27 field early - type galaxies , splitted into less - massive @xmath263 ( blue , open symbols ) and more - massive field galaxies @xmath265 ( red , filled symbols ) , at high redshift @xmath262 by @xcite
is also shown . again
, model tracks for a single - age stellar population with formation redshift 1 ( lower line ) and 2 ( upper line ) are displayed as dotted lines .
all model tracks were normalised to the average @xmath5 ratio of the field galaxies at @xmath267 by @xcite , indicated as the small black dot .
the evolution of cluster galaxies with masses @xmath265 can be best approximated assuming a formation redshift of @xmath266 , which is denoted as the solid line in figure [ dmlrbz ] . for our field
early - type galaxies we measure an evolution of @xmath274 .
although the scatter in the offsets from the local fp is large , a clear trend for a faster evolution of field s0 galaxies is found .
on the other hand , field elliptical galaxies appear to form two separate groups with respect to @xmath5 ratios . to assess the formation redshift @xmath250 , we split our field galaxy sample into a low- and high - mass sample at @xmath268 .
an unrestricted linear @xmath135-fit to the low - mass early - type field galaxies with @xmath263 gives @xmath275 this would correspond to a formation redshift of @xmath276 .
the error on @xmath250 denotes the observed ( 1@xmath13 ) scatter in the @xmath277 relation . for the higher - mass galaxies with @xmath265 a linear @xmath135-fit yields @xmath278 for the massive early - type field galaxies , the evolution of the @xmath5 predicts a mean star formation epoch for the bulk of their stellar populations of @xmath279 . in comparison to the lower - mass counterparts ,
a higher formation redshift is significant on the 2@xmath13 level based on the uncertainty of the high - mass galaxies .
the lack of field galaxies with masses of @xmath265 introduces a larger uncertainty on the @xmath250 value . by comparing the derived formation epoch of the high mass cluster galaxies with numbers for cluster galaxies in the literature
a good agreement is found .
based on a compilation of clusters out to redshift of @xmath280 , @xcite deduce @xmath281 .
this study gives an upper limit for the formation redshift of massive cluster galaxies of @xmath282 ( 1@xmath13 ) .
the result obtained for the massive field galaxies here is within this upper limit , which suggests that these field galaxies have old passively evolving stellar populations similar to their massive counterparts in clusters .
besides the scatter of the individual galaxies , lower - mass field galaxies have a stronger evolution in their @xmath5 ratio .
the mean stellar population ages of the less - massive field galaxies are younger , with lower formation redshift of @xmath283 .
this suggests a _ mass dependent luminosity evolution _ with a stronger evolution for lower mass galaxies with @xmath263 .
+ ] ] the rest - frame optical colours offer a powerful tool to investigate the origin of the scatter in @xmath284 .
figure [ fig : biml ] displays the @xmath5 ratios as a function of the rest - frame @xmath102 colour for our early - type field galaxies . as expected from stellar population model predictions , galaxies with lower @xmath5 ratios should have bluer colors than galaxies with higher @xmath5 .
indeed , lower - mass galaxies ( with low @xmath5 ratios ) are on average bluer than higher - mass galaxies ( with high @xmath5 ratios ) .
galaxies with low @xmath5 ratios follow the predictions of a single stellar population with solar metallicity by @xcite , indicated as the dotted line in figure [ fig : biml ] and form an extension to the higher redshift sample by @xcite .
field galaxies with higher @xmath5 ratios populate a transition zone to the color and @xmath5 ratios regime as defined by the local sdss sample . in figure
[ fig : bizw ] the rest - frame @xmath102 colour as a function of redshift is displayed for our early - type field galaxies . as a comparison
, we also indicate the median @xmath102 colour derived for early - type cluster galaxies in the fdf ( see also [ sec : zdis ] ) . if we compare the figures [ dmlrbz ] and [ fig : bizw ] , strong similarities can be found . the strong correlation between @xmath5 and galaxy colour and a similar evolution in @xmath5 and in colour indicates that the observed evolution and scatter in @xmath5 are intrinsic to the galaxies and due to differences in their stellar populations .
note , that in this figure also galaxies for which no @xmath5 ratios could be derived are shown .
if the sample is equally split according to the characteristic mass separation as introduced in figure [ dmmlrbz ] , lower - mass galaxies with @xmath285 have on average bluer colours ( median colour of @xmath286 than more - massive galaxies with @xmath287 which feature redder colours of @xmath288 . to test the detected mass dependent evolution even further , figure [ mlmfw ] displays the @xmath289 ratio of the early - type field galaxies as a function of mass @xmath290 .
the galaxies were divided according to their redshift into two different bins .
the left panel shows the @xmath291 for all field galaxies , whereas in the right panel the field objects are restricted to a sub - sample of 14 high - redshift galaxies . as a comparison , the nearby coma cluster sample ( small dots ) and the high - redshift field e+s0 galaxies at @xmath6 of @xcite ( triangles )
are indicated together with linear @xmath135-fits to both samples .
all data points have been corrected for maximum progenitor bias @xmath292 .
ellipticals , in particular below @xmath293 , follow the local relation whereas the s0 galaxies have on average lower @xmath5 ratios and match the intermediate
mass range of the @xmath6 field e+s0 galaxies .
lenticular galaxies show a larger scatter and the two outliers in the fp represent the two sa bulges ( id6336 and id508 ) that are clearly offset from the rest of the sample .
the difference between field elliptical and s0 galaxies found in the fp can be explained by a difference in their formation epochs .
ellipticals , especially massive ones , were generated at earlier times and follow rather moderate quiescent evolutionary tracks . on the contrary ,
lenticular and sa bulges show more diverse stellar populations which are younger and have been formed at recent times with formation redshifts @xmath283 . in figure
[ fig : biml ] we have shown that the mass - to - light ratio correlates well with the rest - frame @xmath106 color .
as independent diagnostic tools for star formation , we also compare the equivalent widths of and to stellar population models .
using the and star formation indicators ( see [ sec : ssf ] ) allows us to probe different time scales of star formation episodes in our sample galaxies . assuming a passive evolution , for systems with old stellar populations ( average age of @xmath571 gyr ) with no emission lines in their spectra , the equivalent widths of balmer absorption lines decrease continuously over time .
within the first @xmath294 yr of the generation of new stars in these galaxies the equivalent width of h@xmath104 increases rapidly up to its maximum value around 10 .
afterwards the balmer line decreases steadily until the stellar population reaches an age of 2 gyr .
this can be interpreted that substantial star formation events ( @xmath1830 @xmath295 yr@xmath14 ) occurred in the first @xmath294 yr of the systems with subsequent weak star formation thereafter . in case of the emission line ,
it offers an diagnostic that is sensitive to ongoing or very recent star formation episodes ( @xmath29650 myr ) .
the evolution of the absorption line strength with redshift of our field galaxies is presented in figure [ fig : hdel ] .
we compare our measurements to the determinations of field galaxies up to @xmath6 by @xcite and galaxies with strong h@xmath104 absorption in the gemini deep deep survey ( gdds ) by @xcite . for consistency with previous studies
we define systems with @xmath297 as being galaxies with strong h@xmath104 absorption .
two sets of different stellar population models by @xcite are also indicated .
compared to the h@xmath104-strong galaxies , the field galaxies in the whdf and fdf exhibit lower absorption features .
non of our systems exhibit very strong equivalent width , in particular all galaxies gave @xmath298 .
therefore , they are most likely not systems with strong h@xmath104 absorption .
they follow the evolution track of the composite stellar evolution model consisting of an initial burst of sf at @xmath299 and a secondary sf burst at @xmath300 .
a single stellar population model consisting of a main star burst forming 90% of the total stellar content at @xmath301 is consistent with three galaxies .
all these galaxies show both stronger absorption and additional 3727 emission in their spectra and have experienced a very recent episode of star formation ( see next section [ sec : rsf ] for further details ) .
six galaxies in our sample feature 3727 emission lines in their spectra .
two of these are spiral bulges ( id158 and id508 ) , three are of intermediate type s0/sa ( id946 ) and e / s0 ( id111 and id173 ) and one is an elliptical galaxy ( id437 ) . for these objects ,
the main questions of interest are the star formation rate ( sfr ) , the mass fraction involved in the star formation and the duration of the star formation episode . in oder to convert the 3727 equivalent width to star formation rates , we assume that the equivalent width measured within the aperture is a representative diagnostic for the global value of the galaxy
. the observed luminosity @xmath302()@xmath303 of the emission line can be derived as : @xmath304}})_{\rm obs}=(1.4\pm0.3)\times10^{29}\frac{l_b}{l_{b,\odot}}\ , { \rm ew}({\mbox{[o\,{\scriptsize ii } ] } } ) \ [ { \rm ergs\ s}^{-1}],\ ] ] with @xmath305 and @xmath306 @xcite .
the observed ( measured ) @xmath302()@xmath303 is related to the intrinsic @xmath302()@xmath307 luminosity as @xmath308}})_{\rm int}=3.11\times10^{-20}\,l({\mbox{[o\,{\scriptsize ii}]}})^{1.495}_{\rm obs}.\ ] ] equation [ lilo ] takes into account the intrinsic reddening . to derive the total sfr in units of solar masses per year
, we use the relation established by @xcite : @xmath309}})_{\rm int}\ \ [ m_{\odot}\,{\rm yr}^{-1}].\ ] ] since we do not know the oxygen abundances of our galaxies , we have chosen to adopt the average calibration by kewley et al . using our observed equivalent widths , we derive for the six emission - line galaxies a median sfr of 6.2 @xmath310 . for the four e+s0 galaxies we find a median sfr of 3.7 @xmath310 , whereas the sfr for two sa bulges is 15.5 @xmath310 . instead , transforming our
observed equivalent widths to rest - frame and assuming the relationship reported by @xcite with a global extinction correction of e(h@xmath21)=1 mag , we find for the whole sample a median sfr of 2.5 @xmath310 .
the sfrs for the e+s0 and sa galaxies are 2.5 and 9.1 @xmath310 , respectively .
if we correct the absolute @xmath2-band magnitude of our galaxies using the luminosity difference of the distant galaxies and the low - redshift reference sample as derived from the fp , we get a mean mass of these galaxies of @xmath311 ( median @xmath312 ) . assuming an 1 gyr burst of star formation with the mean sfr found for the line
would have involved between 3% to 10% ( median about 5% ) of the stellar mass of the field galaxies .
we would expect that the involved mass in the sf burst should appear to be strongly dependent on the galaxy mass .
however , our sample statistics of galaxies with recent sf are too small and the galaxies are spread among different redshifts to give further constraints .
to further constrain the duration and the mass fraction involved in the star formation episode we compared our measurements to sets of ` toy ' models that consist a mix of two stellar populations @xcite .
line emission are either caused by a small mass fraction ( 3% ) involved in the star formation burst suggesting @xmath313 , or if a larger mass fraction ( 10% ) is adopted the episode should have been started earlier @xmath314 .
in general , line emission is evidence for star formation activity or an agn which increases the average luminosity of the galaxy and thereby causing the system to be offset from the local fp relation .
in contrast to what is expected , the galaxies with detected ongoing sf in our sample indicate no offset along the edge - on projection of the fp but a tight relation . based on our measurements
we estimate that the stellar mass fraction involved in the sf episode must be in order of 25% to result in a visible effect within the fp .
the emission - line galaxies in our sample indicate no signs of interactions or merging events that could trigger the star formation episodes in these objects .
we therefore find no evidence for a connection between star formation activity and interaction signatures or mergers . ]
lccccccl reference & @xmath315 & @xmath315 & @xmath315 & @xmath316 & @xmath317 & @xmath224 & method ; selection + & reported & high mass & low mass & & & & + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) & ( 8) + treu et al .
( 2001 ) & @xmath318 & @xmath319 & @xmath320 & @xmath321 & 0.29 & 19 & @xmath322 ; @xmath323 , magnitude , colour + van dokkum et al .
( 2001 ) & @xmath324 & @xmath325 & @xmath326 & @xmath327 & 0.42 & 18 & @xmath322 ; @xmath323 , magnitude + treu et al . ( 2001 , 2002 ) & @xmath328 & @xmath329 @xmath329 @xmath329 & @xmath329 @xmath329 @xmath329 & @xmath330 & 0.38 & 29 & @xmath322 ; @xmath323 , magnitude , colour + van dokkum & ellis ( 2003 ) & @xmath331 & @xmath332 & @xmath333 & @xmath334 & 0.56 & 27 & @xmath322 ; @xmath323 , magnitude + gebhardt et al . ( 2003 ) & @xmath335 to @xmath336 & @xmath337 & @xmath338 & @xmath339 & 0.64 & 21 & @xmath322 ; magnitude , absorption lines + van de ven et al . ( 2003 ) & @xmath340 & @xmath341 & @xmath342 & @xmath321 & 0.54 & 21 & lensing isothermal , lenses + rusin & kochanek ( 2005 ) & @xmath343 & @xmath329 @xmath329 @xmath329 & @xmath329 @xmath329 @xmath329 & @xmath329 @xmath329 @xmath329 & 0.54 & 21 & lensing complex ; lenses + van der wel et al .
( 2005 ) & @xmath344 & @xmath345 & @xmath346 & @xmath347 & 0.90 & 27 & @xmath322 ; @xmath323 , magnitude + treu et al .
( 2005 ) & @xmath348 & @xmath329 @xmath329 @xmath329 & @xmath329 @xmath329 @xmath329 & @xmath349 & 0.56 & 141 & @xmath322 ; @xmath323 , magnitude + this work & @xmath350 & @xmath351 & @xmath352 & @xmath353 & 0.40 & 24 & @xmath322 ; @xmath323 , magnitude , colour +
over the past few years , several groups have measured the @xmath5 evolution of field early type galaxies at different redshifts , either using direct mass measurements that involve the internal stellar kinematics of the galaxies @xcite or modelling the stellar kinematic structure and mass profile from the separation of multiple gravitationally lensed images of background galaxies @xcite . in the following we will compare these previous results in the literature with our findings , and comment on possible inconsistencies that exist between the datasets .
table [ pres ] gives a compilation of the @xmath5 evolution as reported by various groups based on different datasets .
for comparison , we also list the methods and selection criteria of each of the studies , where @xmath323 corresponds to a selection based on morphologies . in the methods we present the two most popular techniques in measuring the internal kinematics of e+s0 galaxies , either using the stellar velocity dispersion @xmath322 or using separation of gravitational lensed images via an isothermal model ( lensing isothermal ) or more complex algorithms ( lensing complex ) . besides the magnitude limited samples , there are two studies that analysed the same sample of lensing galaxies @xcite . the predictions based on lensing properties of early - type galaxies agree with one another to within their quoted uncertainties .
most of these studies report that early - type lens galaxies appear to have nearly isothermal mass density profiles ( e.g. , * ? ? ?
@xcite took a different approach and described the mass distribution using statistical model assumptions where the galaxy populations are described by a range of star formation redshifts allowing for a scatter created by deviations from homology or spread in formation epochs .
both methods of lensing and direct mass measurements agree quite well within their uncertainties , although lensing appears to predict a slightly slower @xmath5 evolution of the galaxies . from comparison with other results
, we found that differences in the fitting technique can change the amount of evolution in @xmath354 for each data sample by up to 26% . using direct kinematic measurements ,
two groups have drawn samples of field galaxies from the cdfs to study their evolution at @xmath355 @xcite .
however , these different teams found rather inconsistent galaxy sizes and velocity dispersions for both individual and same objects .
discrepancies arise most likely because these studies used different fitting algorithms for both galaxy parameters ( sizes and kinematics ) .
@xcite argue that their smaller sizes are due to a selection effect of compact galaxies .
these objects are rather faint and have low velocity dispersions ( 1/3 of the galaxies have @xmath356 km s@xmath14 ) and thus could be either biased to early - type spirals or low - luminosity early - type galaxies containing faint discs that are unresolved .
this argument gets support by their small measured srsic indices .
in addition , the large spectroscopic apertures used in their work may cause to underestimate a substantial rotational support in the case of faint low - luminosity galaxies . the most likely explanation between the differences of @xcite and @xcite might be the selection of the target objects and the different analysis tools , where the former primarily selected fainter galaxies whereas the latter included also higher - luminosity ( hence more - massive ) galaxies .
additional selection effects may arise from small number statistics , noise introduced by cosmic variance or the use of shallow imaging data which results in a contamination by bulge - dominated spiral galaxies .
further , low-@xmath22 spectra can not disentangle galaxies with post - starburst spectra ( i.e. , h@xmath104-strong galaxies ) from spectra of normal passive galaxies .
unfortunately our current analysis is limited with respect to the sample size .
however , effects due to imperfections of data quality can completely be ruled out in case of our investigation , as our sample is based on deep imaging photometry and high @xmath22 galaxy spectra .
apart form the work of @xcite , our study comprises the field sample with the lowest average masses of galaxies . therefore
, the lower - mass range where effects of differences in the evolution should be more pronounced can be investigated in greater detail .
in the nearby future , we will address both important restrictions of number statistics and cosmic variance that impact high redshift studies ( a. fritz et al .
2009a , in preparation ) .
some previous studies have investigated the possible influence of the nuclear activity on the evolution of early - type galaxies .
@xcite find no differences in the evolution of the fp for bl lac objects and radio galaxies with a central agn from normal early - type galaxies up to @xmath357 .
however , their sample consists of high - luminosity galaxies , hence it is biased towards high - mass galaxies , in particular @xmath358 . @xcite detected in an fp analysis of field early - type galaxies up to @xmath6 that 4 out of 11 galaxies at @xmath359 ( 36% ) and one out of 16 for @xmath360 show agn activity .
although their galaxies with agn are twice as massive than galaxies without agn they show similar @xmath5 evolution and ( no blue ) rest - frame @xmath102 colours .
the authors suggest that the difference between agn and non - agn galaxies may become more apparent at higher redshift @xmath361 .
a similarly low number of agn galaxies was detected within the sample of @xcite , who identified out of 26 e+s0 x - ray counterparts nine agn based on x - ray luminosity ( xagn ) , 16 llagn ( see section [ xray ] and appendix [ xrayp ] for these group definitions ) , but only two out of 13 -emitting e+s0 galaxies as secure agn .
the total early - type galaxy sample of @xcite has a quite low contribution of agn activity ( 16% ) and none of the spectra shows signs of significant sfr , with an upper limit of @xmath1830 @xmath295 yr@xmath14 .
we find three candidates in our field galaxy sample ( @xmath36213% ) that show weak evidence for hosting an agn ( see section [ xray ] ) .
our observations suggest a very mild evolution in the absolute numbers of agn hosts up to @xmath363 .
this is consistent with the results of the stellar mass function of aegis field galaxies @xcite , where the abundance of x - ray selected agn host galaxies ( @xmath36410@xmath365ergss@xmath14 ) decreases roughly by a factor of two from @xmath366 to @xmath367 , but remains relatively flat across the stellar mass range of @xmath368 .
however , possible contributions due to llagn and/or emission from soft extended x - ray halos or lmxbs could have influence on the activity of our individual galaxies as well . for our field galaxies we measure an average contribution of lmxbs to the x - ray emission in the @xmath3-band of @xmath369(lmxb)@xmath36218%@xmath1156% .
the x - ray luminosities in our distant galaxies show a weak excess in the soft diffuse x - ray plasma emission , depending on the established local @xmath370 relationship , between @xmath3627%@xmath1153% to @xmath36226%@xmath11515% compared to the local reference .
the increased luminosities are most likely originating from a net effect of star formation and/or llagn .
a more detailed discussion of these findings and their implications can be found in the appendix [ xrayp ] .
blue ( central ) cores in galaxies are not a stringent indicator for a central agn power source . for example , within the field early - type galaxy sample of @xcite only 8% show signs of central blue rest - frame colours with @xmath371 , and only two of these objects ( 1% ) appear also to host an agn based on their high x - ray luminosity ( @xmath5710@xmath365ergss@xmath14 ) .
this suggests that the colour information of the galaxies is not a very powerful technique to distinguish active from non - active galaxies .
as there is no multi - band hst photometry for our sky fields available , we thus can not investigate the colour gradients in our e+s0 galaxies . to establish and explore the evolutionary link between e+s0 galaxies at high redshift and their local descendants ,
we assume that e+s0 galaxies comprising passive ( red and `` dead '' , non - starforming ) old stellar populations at high redshift will only be subject to a pure passive evolution in luminosity through @xmath8 .
our approach is justified by the colour evolution of galaxies populating the red sequence following a pure passive evolution @xcite . in the next step
, we adopt the luminosity dimming as derived from the evolution of the fp ( see [ stpops ] ) as @xmath274 and correct for progenitor bias .
this latter effect also compensates for a possible ( morphological ) transformation of galaxies which end up as being classified as passive local e+s0 galaxies .
massive field e+s0 galaxies show a smaller offset from the @xmath5-relation , whereas less massive ones evolve faster which is supported by younger stellar populations ( see section [ sec : zf ] ) .
independent of the luminosity of our galaxies , we apply this relationship to all galaxies in our sample and then investigate a possible dependence of the evolution on luminosity and the physical mechanisms that drive our measured mass dependent evolution .
the luminosity of the distant e+s0 galaxies is then decreased by a rate of @xmath372 to obtain their absolute ( rest - frame ) @xmath2-band magnitude at @xmath8 as @xmath373 this luminosity is the resulting characteristic brightness at @xmath8 to which the galaxies would fade , assuming a pure passive evolution .
the rate of luminosity evolution in equation [ levp ] is consistent with the predictions of stellar population synthesis models that formed the bulk of the stars in e+s0 galaxies at @xmath374 @xcite . in figure
[ fig : mbz ] we compare the evolution of the rest - frame luminosity @xmath375 ( corrected for expected passive evolution ) and stellar mass as a function of redshift for various field early - type galaxy samples .
several implications can be drawn from this diagram .
first , there is a clear trend of @xmath375 becoming more luminous with increasing redshift for the field e+s0 galaxies in the fdf and whdf ( large squares ) .
our results are in good agreement with the findings for field e+s0 galaxies up to @xmath6 by @xcite , shown as small triangles together with their associated errors ( @xmath376 scatter ) .
these measurements have also been corrected with equation [ levp ] , but using the evolution as derived from their fp .
moreover , the brightening detected for our galaxies indicates that the most massive e+s0 galaxies are the first to build up the red sequence whereas less massive ones join at more recent epochs ( @xmath377 ) .
the observed trend of both the evolutionary corrected luminosity and total stellar mass is consistent with that of a characteristic redshift evolution of the so - called `` quenching '' mass @xmath378 @xcite .
the quenching mass limit @xmath379 is a diagnostic that traces the total baryonic ( stellar ) mass scale and a direct by product of the mechanism that drives downsizing .
it is defined as the fraction of star forming galaxies ( usually 1/3 of the total fraction ) which contributes to the total ( star forming ) mass budget . above this mass threshold @xmath379 ,
star formation in stellar systems is suppressed and the majority of galaxies have already been transformed into passive , quiescent galaxies .
the global trend of a downward evolution of the transitional mass with increasing age of the universe , where @xmath379 decreases by a factor of @xmath185 from @xmath380 , is well reproduced by measurements based on the combo-17 and deep2 redshift surveys @xcite .
this transitional ( crossover ) mass is computed in different redshift bins for 50% late - type ( spirals and peculiars ) morphologies ( dashed line in figure [ fig : mbz ] ) , 50% galaxies with blue @xmath381 colour ( dot - dashed line ) , and 50% galaxies with -inferred sfrs above 0.2@xmath310 ( dotted line ) .
these trends can also be fairly well described by the solid line in figure [ fig : mbz ] , which indicates the average evolution of the passively evolved lf for early - type galaxies from the combo-17 , deep2 and sxds with redshift @xcite at which the bulk ( @xmath4560% ) of the e+s0 galaxies are already in place .
while our observed field galaxies are in good agreement with the model predictions using blue @xmath381 color criterion , in particular for fainter ( hence lower mass ) individual field galaxies ( @xmath382 , @xmath285 ) , models with high sfr ( as measured from the emission strength ) are not consistent with the observations .
there is a weak increase of the quenching mass @xmath379 with redshift which is already apparent at lower redshifts @xmath245 , which continuous up to higher redshifts @xmath383 .
this trend suggests that star formation is more effectively suppressed in more luminous ( massive ) systems .
strong suppression of star formation could be triggered from agn feedback of weak low luminosity ( radio - mode ) agn that have red optical colours and are situated in very massive dark matter haloes @xcite . ]
figure [ fig : mbz ] suggests further that the evolution is a strong function of luminosity and mass and the evolution depends on the ( evolution corrected ) luminosity and stellar mass range of each specific galaxy sample under consideration .
evidence for this trend is also accumulated by studies of the redshift evolution of the lf , which argue that the number density of massive e+s0 galaxies ( @xmath287 ) remains constant out to @xmath6 @xcite . for less massive galaxies ( @xmath285 ) , the lf becomes progressively underpopulated with increasing redshift , which could point to the fact that the progenitors of the massive local e+s0 galaxies are still experiencing a period of active star formation .
these galaxies would be too blue to fulfill the requirements in the lf selection at higher redshift , and could gradually build up the red sequence of massive e+s0 galaxies by a successive quenching of the star formation at the less massive end of the lf ( @xmath285 ) . in this picture , a fraction of blue ( spiral ) galaxies experience a rather slow , gradual morphological transformation into red , ` dead ' quiescent e+s0 galaxies by a progressive quenching of star formation since redshifts @xmath384 .
using a combination of high signal - to - noise intermediate - resolution vlt spectroscopy , deep @xmath32 ground based photometry , and high - resolution acs imaging , we have investigated the properties and evolution of 24 field early - type ( e+s0 ) galaxies covering redshifts @xmath0 down to a magnitude limit of @xmath1 .
the targets were selected from the southern fdf and northern whdf based on morphology , luminosity , spectrophotometric type , and photometric redshift or broad - band colours . scaling relationships of the faber - jackson relation ( fjr ) , and kormendy relation as well as the fundamental plane ( fp )
have been constructed to assess the evolution of the stellar populations for the galaxies in our sample .
the formation epoch for the bulk of the stars in these systems was constrained by comparing the evolution of the @xmath5 ratios to different stellar population models .
to reveal the origin of the star formation in some of our galaxies , we have exploited the _ chandra _ and _ xmm - newton _ data archives .
the main results of this work can be summarised as follows : * the mass - to - light ratio of the field early - type galaxies evolves on average as @xmath274 , which is faster than for cluster early - type galaxies .
this is consistent with field e+s0 galaxies comprising on average younger stellar populations than those of their cluster counterparts .
* the evolution as derived from the scaling relations of the fjr , the fp and hence the effective mass - to - light ratio shows a strong dependence on the mass .
less - massive systems with @xmath263 display a faster @xmath5 evolution of @xmath385 , whereas more - massive galaxies with @xmath386 evolve much slower as @xmath387 .
this implies that for the most massive field galaxies ( @xmath386 ) the majority of their stellar populations have been formed early at @xmath388 . for less - massive galaxies the evolution in the @xmath5 ratios
can be translated into a lower formation redshift of their stars of @xmath389 , which is at variance with the predictions of single - burst stellar population models .
a fraction of 5%10% in the total stellar mass of lower - mass galaxies ( @xmath263 ) must have been formed at more recent epochs . at @xmath194 , an accelerated evolution for less - massive galaxies ( @xmath390 ) than for more - massive ones ( @xmath269 )
is significant on the 2@xmath13 level . *
the morphological sub - classes of early - type galaxies , elliptical and lenticular ( s0 ) galaxies , appear not to represent a homogenous group but follow different evolutionary tracks .
lenticular galaxies feature on average a stronger luminosity evolution ( @xmath391 ) , bluer rest - frame colours and more diverse stellar populations and hence evolve faster than elliptical galaxies ( @xmath392 ) . however , the differences between elliptical and lenticular galaxies are only moderate significant and difficult to disentangle from their individual scatter .
note that for deriving the scatter we took into account random errors as well as systematic uncertainties ( see [ eso ] ) .
interestingly , the scatter appears to be mainly amplified for lenticular galaxies . *
evidence for secondary star formation activity in some of our galaxies is provided by the detection of emission or strong balmer absorption as well as bluer rest - frame @xmath102 colour diagnostics .
the absorption line strengths in our galaxies implies that the residual ( low level ) star formation of the galaxies accounts for 5% to 10% in the total stellar mass budget of these systems .
* we have found three possible candidates that could harbour an central agn source .
however , the trigger of the star formation in our galaxies is most likely either stellar origin or the result of agn feedback from weak low luminosity ( radio - mode ) agn with red optical colours @xcite . * over the past @xmath186 gyr , there is the trend of a slow decreasing quenching mass , which characterises a crossover in stellar mass above star formation in galaxies gets suppressed .
this suggest that our systems experience a gradual suppression in their star formation processes .
our results favour a scenario where star formation is not immediately suppressed for less - massive halos and hence works on longer time scales in lower - mass galaxies .
the evolution of the @xmath5 ratio of our field early - type galaxies favours a downsizing formation scenario , similar to the detected decrease in luminosity of rapidly star forming field galaxies since @xmath3931 @xcite . in this ` top - down ' picture , massive galaxies are dominated by quiescent ( red , old ) stellar populations and evolve passively on longer timescales .
less - massive systems , however , exhibit more extended star formation histories with mass hosting star formation processes decreasing as times evolves .
both , the mass assembly and star formation are accelerated in massive stellar systems that are preferably located in high density environments .
in contrast , in less - massive ( smaller ) systems , these physical processes work on longer timescales . moreover , our results suggest a continuous mass assembly of field early - type galaxies during the last 5 gyr .
this is supported by recent studies of field galaxies up to @xmath6 @xcite .
an accelerated evolution in the @xmath5 ratios of the less - massive galaxies with respect to more - massive ones was also found for _ cluster _ early - type galaxies @xcite .
the lack of a significant age difference between field and cluster early - type galaxies as expected from hierarchical formation models @xcite , and the dependence of the evolution on galaxy mass suggests that environmental effects play a rather minor role in the formation of early - type galaxies .
more important to the evolutionary history of a stellar system are its internal properties , such as chemical composition or element abundance ratios . in a next step
, we will combine our precise fp measurements with an analysis of the absorption line strengths in our field early - type galaxies .
this will provide strong constraints on the evolution of stellar populations and the chemical enrichment history of heavy elements in these systems .
we would like to thank the referee prof .
g. gavazzi for insightful comments and a constructive review which improved the clarity of the manuscript .
we thank the eso paranal staff for the professional and efficient assistance during the spectroscopic observations and the fdf team for the contributions to the first fdf sample analysis , in particular j. fliri and a. riffeser ( both usm mnchen ) for the cosmic ray removal on the hst / acs images of the fdf .
we are grateful to j. heidt ( lsw heidelberg ) for taking pre - images of the whdf .
we also thank prof .
k. j. fricke ( gttingen ) for encouragement and prof .
r. bender ( usm / mpe mnchen ) for kindly providing his fcq program .
would like to acknowledge support from prof .
w. w. zeilinger and the institut fr astronomie der universitt wien where he was a visiting researcher .
this work has been supported by the volkswagen foundation ( i/76520 ) .
a.f . acknowledges partial support from grant hst - go-10826.01 from stsci .
stsci is operated by aura , inc .
, under nasa contract nas 5 - 26555 .
a.b . acknowledges financial support by the bmbf / dlr ( 50or0301 and 50or0404 ) .
appenzeller , i. , fricke , k. j. , frtig , w. , et al .
1998 , the messenger , 94 , 1 balogh , m. l. , morris , s. l. , yee , h. k. c. , carlberg , r. g. , & ellingson , e. 1997 , apj , 488 , l75 baugh , c. m. , cole , s. , frenk , c. s. , & lacey , c. g. 1998 , apj , 498 , 504 bell , e. f. , & de jong , r. s. , 2001 , apj , 550 , 212 bell , e. , mcintosh , d. h. , katz , n. , & weinberg , m. d. 2003 , apjs , 149 , 289 bell , e. f. et al .
2004 , apj , 608 , 752 bender , r. , 1990 , a&a , 229 , 441 bender , r. , saglia , r. p. , ziegler , b. , belloni , p. , bruzual , g. , greggio , l. , & hopp , u. 1998 , apj , 493 , 529 bender , r. , et al .
2001 , in eso astrophysics symp . , deep fields , ed . s. cristiani , a. renzini , & r. e. williams , springer , p. 96
benson , a. j. , ellis , r. s. , & menanteau , f. 2002 , mnras , 336 , 564 bertin , e. , & arnouts , s. 1996 , a&as , 117 , 393 birnboim , y. , dekel , a. , & neistein , e. 2007 , mnras , 380 , 339 borch , a. , et al .
2006 , a&a , 453 , 869 bower , r. g. , benson , a. j. , malbon , r. , helly , j. c. , frenk , c. s. , baugh , c. m. , cole , s. , lacey , c. g. 2006 , mnras , 370 , 645 bhm , a. , ziegler , b. l. , saglia , r. p. , et al .
2004 , a&a , 420 , 97 bhm , a. , & ziegler , b. l. 2007 , apj , 668 , 846 bruzual , g. a. , charlot , s. 2003 , mnras , 344 , 1000 bundy , k. , et al . 2006 , apj , 651 , 120 bundy , k. , et al .
2008 , apj , 681 , 931 chapman , s. c. , smail , i. , blain , a. w. , & ivinson , r. j. 2004 , apj , 614 , 671 cimatti , a. , daddi , e. , & renzini , a. 2006 , a&a , 453 , l29 cole , s. , lacey , c. g. , baugh , c. m. , & frenk , c. s. 2000 , mnras , 319 , 168 conselice , c. j. , et al .
2007 , mnras , 381 , 962 cowie , l. l. , songaila , a. , hu , e. m , cohen , j. g. 1996 , aj , 112 , 839 croton , d. j. , et al . 2006 ,
mnras , 365 , 11 de lucia , g. , kauffmann , g. , & white , s. d. m. , 2004 , mnras , 349 , 1101 de lucia , g. , springel , v. , white , s. d. m. , croton , d. , & kauffmann , g. 2006 , mnras , 366 , 499 desroches , l .- b .
, quataert , e. , ma , c .- p .
, & west , a. a. 2007 , mnras , 377 , 402 diaferio , a. , kauffmann , g. , balogh , m. l. , white , s. d. m. , schade , d. , & ellingson , e. 2001 , mnras , 323 , 999 di serego alighieri , s. , et al .
2005 , a&a , 442 , 125 djorgovski , s. , & davis , m. 1987 , apj , 313 , 59 dressler , a. 1980 , apj , 236 , 351 dressler , a. , lynden - bell , d. , burstein , d. , davies , r. l. , faber , s. m. , terlevich , r. , & wegner , g. 1987 , apj , 313 , 42 faber , s. m. , & jackson , r. e. 1976 , apj , 204 , 668 faber , s. m. , wegner , g. , burstein , d. , davies , r. l. , dressler , a. , lynden - bell , d. , & terlevich , r. j. 1989 , apjs , 69 , 763 faber , s. m. , et al .
2007 , apj , 665 , 265 fabbiano , g. 1989 , ara&a , 27 , 87 ferreras , i. , et al .
2005 , apj , 635 , 243 fritz , a. , ziegler , b. l. , bower , r. g. , smail , i. , & davies , r. l. 2005 , mnras , 358 , 233 fritz , a. 2006 , phd thesis , university of gttingen fritz , a. , ziegler , b. l. , bhm , a. , balogh , m. l. , bower , r. g. , smail , i. , & davies , r. l. 2006 , in _ the fabulous destiny of galaxies : bridging past and present _ , eds . v. lebrun , a. mazure , s. arnouts & d. burgarella , paris : frontier group , p. 525
gavazzi , g. , pierini , d. , boselli , a. 1996 , a&a , 312 , 397 gavazzi , g. , bonfanti , c. , sanvito , g. , boselli , a. , scodeggio , m. 2002 , apj , 576 , 135 gebhardt , k. , faber , s. m. , koo , d. c. , et al .
2003 , apj , 597 , 239 glazebrook , k. , et al .
2004 , nature , 430 , 181 granato , g. l. , de zotti , g. , silva , l. , bressan , a. , & danese , l. 2004 , apj , 600 , 580 heidt , j. , appenzeller , i. , gabasch , a. , et al .
2003 , a&a , 398 , 49 hopkins , p. f. , bundy , k. , hernquist , l. , & ellis , r. s. 2007 , apj , 659 , 976 horne , k. 1986 , pasp , 98 , 609 jrgensen , i. , franx , m. , & kjrgaard , p. 1995 , mnras , 276 , 1341 jrgensen , i. , franx , m. , & kjrgaard , p. 1996 , mnras , 280 , 167 jrgensen , i. , chiboucas , k. , flint , k. , bergmann , m. , barr , j. , & davies , r. 2006 , apj , 639l , 9 kauffmann , g. , 1996 , mnras , 281 ,
487 kauffmann , g. , & haehnelt , m. , 2000 , mnras , 311 , 576 kennicutt , r. c. , jr .
1992 , apj , 388 , 310 kewley , l. j. , geller , m. j. , & jansen , r. a. 2004 , aj , 127 , 2002 kim , d .- w . , & fabbiano , g. 2003 , apj , 586 , 826 kim , d .- w . , & fabbiano , g. 2004 , apj , 611 , 846 kinney , a. l. , et al .
1996 , apj , 467 , 38 kriek , m. , et al .
2006 , apj , 649 , l71 le borgne , d. , et al . 2006 , apj , 642 , 48 mcgrath , e. j. , stockton , a. , canalizo , g. 2007 , apj , 669 , 241 metcalfe , n. , shanks , t. , campos , a. , mccracken , h. j. , & fong , r. 2001 , mnras , 323 , 779 monaco , p. , murante , g. , borgani , s. , & fontanot , f. 2006 , apj , 652l , 89 mller , c. s. , fritze-v.alvensleben , u. , fricke , k. j. , & calzetti , d. 2001 , ap&ss , 276 , 799 noll , s. , et al .
2004 , a&a , 418 , 885 osullivan , e. , forbes , d. a. , ponman , t. j. 2001 , mnras , 328 , 461 osullivan , e. , ponman , t. j. , & collins , r. s. 2003 , mnras , 340 , 1375 pahre , m. a. , 1999 , apjs , 124 , 127 peng , c. y. , ho , l. c. , impey , c. d. , & rix , h .- w .
2002 , aj , 124 , 266 rusin , d. & kochanek , c. s. 2005 , apj , 623 , 666 saglia , r. p. , bender , r. , & dressler , a. 1993 , a&a , 279 , 75 ( sbd93 ) saglia , r. p. , et al .
1997 , apjs , 109 , 79 salim , s. , et al . 2007 , apjs , 173 , 267 salpeter , e. e. 1955 , apj , 121 , 161 sarazin , c. l. , irwin , j. a. , & bregman j. n. 2000 , apj , 544 , l101 sargent , w. l. w. , schechter , p. l. , boksenberg , a. , & shortridge , k. 1977 , apj , 212 , 326 srsic , j. l. 1968 , atlas de galaxias australes ( crdoba , argentina : observatorio astronomico , univ .
crdoba ) simard , l. , et al .
2002 , apjs , 142 , 1 sirianni , m. , jee , m. j. , bentez , n. , et al .
2005 , pasp , 117 , 1049 somerville , r. s. , & primack , j. r. 1999 , mnras , 310 , 1087 somerville , r. s. , primack , j. r. , & faber , s. m. 2001 , mnras , 320 , 504 thomas , d. , maraston , c. , bender , r. , de oliveira , c. m. 2005 , apj , 621 , 673 tinsley , b. m. 1980 , fundamentals of cosmic physics , 5 , 287 tonry , j. , & davis , m. 1979 , aj , 84 , 1511 treu , t. , stiavelli , m. , bertin , g. , casertano , s. , & mller , p. 2001
mnras , 326 , 237 treu , t. , stiavelli , m. , casertano , s. , mller , p. , & bertin , g. 2002 , apj , 564 , l13 treu , t. , et al .
2005 , apj , 633 , 174 van de ven , g. , van dokkum , p. g. , franx , m. 2003 , mnras , 344 , 924 van der wel , a. , franx , m. , van dokkum , p. g. , rix , h .- w . , illingworth , g. d. , & rosati , p. 2005
, apj , 631 , 145 van dokkum , p. g. , & franx , m. 1996 , mnras , 281 , 985 van dokkum , p. g. , & franx , m. 2001b , apj , 553 , 90 van dokkum , p. g. , franx , m. , kelson , d. d. , & illingworth , g. d. , 2001a , apj , 553 , l39 van dokkum , p. g. & ellis , r. s. 2003 , apj , 592 , l53 van dokkum , p. g. , & stanford , s. a. 2003 , apj , 585 , 78 weisskopf , m. c. , brinkman , b. , canizares , c. , garmire , g. , murray , s. , van speybroeck , l. p. 2002 , pasp , 114 , 1 woo , j. h , urry , c. m. , lira , p. , van der marel , r. p. , & maza , j. 2004 , apj , 617 , 903 woo , j. h , urry , c. m. , van der marel , r. p. , lira , p. , & maza , j. 2005 , apj , 631 , 762 worthey , g. , & ottaviani d. l. 1997 , apjs , 111 , 377 ziegler , b. l. , bhm , a. , fricke , k. j. , et al .
2002 , apj , 564 , l69 ziegler , b. l. , thomas , d. , bhm , a. , bender , r. , fritz , a. , & maraston , c. 2005 , a&a , 433 , 519
] to test the accuracy and the possible internal systematic effects of velocity dispersion measurements using different absorption features , galaxies where multiple @xmath13 determinations could be obtained are compared . in case of five field early - type galaxies in the whdf @xmath13
was derived based on the and the g - band feature .
figure [ cmprdsmgw ] shows the formal relative uncertainty @xmath394 as a function of velocity dispersion ( upper panel ) and as a function of the @xmath22 ( lower panel ) per for the early type galaxies in the whdf where both measurements are available .
the lower panel is divided according to the @xmath22 of the feature of interest , @xmath22 values from the lines are denoted as circles , whereas the @xmath22 based on the g band is indicated as triangles .
@xmath394 was computed as the relative difference in velocity dispersion values of the with the g - band line , normalised to the reference velocity dispersions @xmath395 .
both features were free from any contamination due to sky lines and did not suffer from any residual sky background artefacts .
the relative uncertainties cover the range @xmath396 with a median of @xmath397 .
the @xmath13 determinations are in extremely good agreement and show similar absolute errors between 8 to 18 km s@xmath14 . even for the lowest @xmath22 ratios the formal uncertainty of 5% is less than the typical individual error in the @xmath13 measurement .
this limit is also the smallest formal uncertainty that can be reached with the instrumental resolution and the template stars .
no trend of increasing systematic errors on the @xmath13 values with low @xmath22 can be found .
results from the monte - carlo simulations showed that the effects of systematic errors on @xmath13 become important at @xmath74 per ( cf .
[ sec : sig ] ) . as this limit is below the average @xmath22 in the spectra of our low
luminosity galaxies , the formal uncertainty therefore becomes negligible even for the lowest derived velocity dispersions in the galaxy samples .
a comparison of the absolute velocity dispersions for the whdf ellipticals is presented in figure [ cmpsidou ] in the next section . ] for a total of five field galaxies in the whdf sample which are included in two or more different mos masks , it is possible to confirm the internal reliability of the spectral reduction and analysis .
a total of nine repeat observations was acquired for the whdf ellipticals .
two galaxies were observed thrice ( id92 and id111 ) and id173 four times .
the velocity dispersions have been measured for both individual setups and the agreement between these two determinations is very good with a median offset in @xmath13 at the 4% level .
figure [ cmpsidou ] compares the velocity dispersion measurements for the five whdf field galaxy spectra which were observed with two or up to four different mask setups .
one galaxy with four repeat observations is denoted with two additional large circles . in general ,
velocity dispersions of the secondary measurements are based on lower @xmath22 spectra and therefore show larger errors . the final measurement with the smallest error yielded @xmath398 km s@xmath14 ( @xmath399 ) , which represents also the smaller deviation with respect to the reference line .
for the whdf field ellipticals the relative difference between reference and secondary measurements is @xmath400 with @xmath401 . in order to increase the signal to noise in the final spectra ,
these repeat observations have been co added .
the repeated @xmath13 measurements in different slitlet configurations prove that our kinematic measurements are very stable and that internal effects such as varying spectral wavelength range and slitlet position on the fors spectrograph can be neglected .
early - type galaxies in the nearby universe with luminosities typically @xmath402ergss@xmath14 most likely harbour an agn source in their centre @xcite . below this flux limit at @xmath403ergss@xmath14 ,
the main contributions to the total x - ray luminosity of a galaxy could be a variety of sources , such as from low - luminosity agn ( llagn ) , high - mass and low - mass x - ray ( stellar ) binaries ( lmxbs ) and emission fluxes from the diffuse soft x - ray plasma that surrounds the galaxy ( @xmath404kev ) .
lmxbs are associated with evolved old stellar populations ( a stars ) and a common feature in local bulge systems in the field and globular clusters . as both e and s0 galaxies comprise homogenous stellar populations , their x - ray binaries ( xrb ) populations
should also be uniformly distributed among the galaxy populations . in the past
it was proposed that lmxbs might even dominate the global x - ray emission of early - type galaxies @xcite .
however , this issue is hard to quantify as lmxbs can not be identified as individual point sources . on the other hand for fainter x - ray systems observations
argue that lmxbs account for a significant high fraction ( from 20% up to 50% ) of the x - ray emission of the host galaxy @xcite , which becomes a non - negligible factor when estimating the dynamical masses of their stellar host systems . in oder to identify in our sample of galaxies with possible associated x - ray counterparts ( see [ xray ] ) those systems with nuclear activity or ongoing star formation , an effective way would be to classify our galaxies schematically based on their measured x - ray luminosity . using the x - ray flux measurements for our objects
, we can give constraints on the various sources that contribute to the x - ray emission of our galaxies .
the two main groups of interests are systems with nuclear activity based on the x - ray luminosity ( xagn ) and galaxies with low - luminosity nuclear activity ( llagn ) and/or ongoing star formation . for the old quiescent stellar populations in early - type galaxies , pointlike x - ray sources
are very likely to be entirely resembled by lmxbs .
the contribution due to lmxbs is direct proportional to the stellar mass of a galaxy . in general , these binary systems ( @xmath405 sources per galaxy ) account for a smaller but non - negligible fraction ( @xmath29625% ) of the total galaxy x - ray luminosity @xcite . using the relationship between total @xmath2-band and x - ray luminosity by @xcite as @xmath369(lmxb)=@xmath406 ( in units of ergss@xmath14 )
, we measure an average contribution of lmxbs of @xmath36217%@xmath1157% .
the evolution corrected absolute @xmath2-band luminosities of our galaxies fall in the range @xmath407 with a median of 10.47 . as a comparison , introducing our @xmath3-band measurements of @xmath408 ( average 11.22 ) and a similar relation as above but for the @xmath3-band luminosity of @xmath369(lmxb)=@xmath409 ( ergss@xmath14 ) , we detect an lmxb source contribution of @xmath36218%@xmath1156% .
local early - type galaxies show a correlation of their soft x - ray luminosity with their stellar @xmath410 ( e.g. , * ? ? ?
excluding dwarf galaxies , brightest cluster / group galaxies and galaxies with agn hosts , @xcite found for 184 nearby e+s0 galaxies log(@xmath369)=@xmath411log@xmath412 . because of the evolved stellar populations in the local e+s0 galaxies , any detected excess with respect to this empirical relation
would imply a measure of star formation processes or llagns in our distant galaxies . applying the above relationship to our sample ,
we derive an excess in diffuse plasma fraction of @xmath36226%@xmath11515% compared to the local value . using instead the relation by @xcite @xmath413 , the fraction of soft x - ray plasma decreases to @xmath3627%@xmath1153% .
both measurements give evidence for a weak excess in the x - ray luminosities in our distant galaxies ( @xmath414 ) compared to the local reference .
the increased luminosities are most likely the result of a combination of star formation and/or llagn .
however , a possible contribution from lmxbs and soft diffuse x - ray plasma emission to the x - ray luminosities can not be ruled out completely .
to clearly distinguish between the various contributions a detailed analysis of the x - ray spectra is necessary which goes beyond the scope of the present work . | we explore the properties of 24 field early type galaxies in the redshift range @xmath0 down to @xmath1 in a sample extracted from the fors deep field and the william herschel deep field .
target galaxies were selected on the basis of a combination of luminosity , spectrophotometric type , morphology and photometric redshift or broad band colours .
high signal to noise intermediate
resolution spectroscopy has been acquired at the very large telescope , complemented by deep high - resolution imaging with the advanced camera for surveys onboard the _ hst _ and additional ground based multi band photometry .
all galaxy spectra were observed under sub
arcsecond conditions and allow to derive accurate kinematics and stellar population properties of the galaxies . to clarify the low level of star formation detected in some galaxies , we identify the amount of active galactic nuclei ( agn ) activity in our sample using archive data of _ chandra _ and _ xmm - newton _ x - ray surveys .
non of the galaxies in our sample were identified as secure agn source based on their x - ray emission .
the rest - frame @xmath2 and @xmath3-band scaling relations of the faber - jackson relation and the fundamental plane display a moderate evolution for the field early type galaxies .
lenticular ( s0 ) galaxies feature on average a stronger luminosity evolution and bluer rest - frame colours which can be explained that they comprise more diverse stellar populations compared to elliptical galaxies .
the evolution of the fp can be interpreted as an average change in the dynamical ( effective ) mass - to - light ratio of our galaxies as @xmath4 .
the @xmath5 evolution of these field galaxies suggests a continuous mass assembly of field early - type galaxies during the last 5 gyr , that gets support by recent studies of field galaxies up to @xmath6 .
independent evidence for recent star formation activity is provided by spectroscopic ( emission , ) and photometric ( rest frame broad band colors ) diagnostics .
based on the absorption feature we detect a weak residual star formation for galaxies that accounts for 5%10% in the total stellar mass of these galaxies . the co - evolution in the luminosity _ and _ mass of our galaxies favours a downsizing formation process .
we find some evidence that our galaxies experienced a period of star formation quenching , possible triggered by agn activity that is in good agreement with recent results on both observational and theoretical side .
[ firstpage ] cosmology : observations galaxies : elliptical and lenticular , cd galaxies : evolution galaxies : formation galaxies : kinematics and dynamics galaxies : stellar content |
there are several numerical approximations schemes in the time direction for the numerical solution of nonlinear parabolic equations ( systems ) .
linearized ( semi)-implicit schemes are the most popular ones since , at each time step , the schemes only require the solution of a linear system .
however , time - step size restriction condition is always a key issue in analysis and computation . for many nonlinear parabolic systems , error analysis of finite element methods ( or finite difference method ) with linearized semi - implicit schemes in the time direction
often requires certain time - step conditions .
see @xcite for the navier - stokes equations , @xcite for the nonlinear joule heating problems , @xcite for flows in porous media , @xcite for viscoelastic fluid flow , @xcite for the kdv equations , @xcite for the ginzburg - landau equations , @xcite for the nonlinear schrdinger equations and @xcite for some other equations .
such time - step size restrictions may result in the use of an unnecessarily small time - step size and extremely time - consuming in practical computations . to study the error estimate of linearized ( semi)-implicit schemes , the boundedness of numerical solution ( or error function ) in the @xmath3 norm or a stronger norm is often required .
if a priori estimate for the numerical solution in such a norm can not be provided , one may employ the induction method with an inverse inequality to bound the numerical solution , such as @xmath5 where @xmath6 is the finite element solution , @xmath7 is the exact solution and @xmath8 is certain projection operator .
a time - step size restriction arises immediately from the above inequality , particularly for problems in three dimensional spaces .
most previous works follow this idea .
a new approach for unconditionally optimal error analysis of a linearized galerkin fem was pesented in our recent work @xcite , also see @xcite , where the error function is split into two parts , the spatially discrete error and the temporally discrete error , @xmath9 where @xmath10 is the solution of a corresponding time - discrete parabolic equations ( or elliptic equations ) .
optimal estimates for the second term can be obtained unconditionally in a traditional way if suitable regularity of the solution of the time - discrete system can be proved .
more recently , unconditionally optimal error estimates were established for a nonlinear equation from incompressible miscible flow in porous media . in @xcite ,
analysis waw given only for a linear fem and a low - order galerkin - mixed fem , respectively . in this paper
, we consider a general nonlinear parabolic equation ( or system ) @xmath11 in a bounded and smooth domain @xmath12 in @xmath0 ( @xmath13 or @xmath14 ) with the boundary condition @xmath15 and the initial condition @xmath16 where @xmath17 is a general nonlinear source .
the general equation is of stronger nonlinearity than those in @xcite and many physical equations are included .
we apply linearized backward euler galerkin method with @xmath18-order finite element approximation ( @xmath19 ) for the general nonlinear system .
we focus our attention on the unconditional convergence ( stability ) and optimal error estimates of the linearized galerkin fems .
a key to our analysis is the a priori estimate of the numerical solution .
we apply the splitting technique proposed in @xcite to bound the numerical solution @xmath6 in @xmath3-norm and @xmath4-norm , such as @xmath20 where @xmath21 .
then with the boundedness of @xmath22 , optimal error estimates can be easily established unconditionally in the routine way of fem error analysis .
the paper is organized as follows . in section 2 ,
we present linearized backward euler galerkin fems for the general nonlinear parabolic equations - and introduce our notations . in section 3 , we prove the boundedness of the numerical solution in the @xmath4 norm in terms of a corresponding time - discrete system , in which a rigorous analysis on the regularity of the solution to the time - discrete pdes is given . due to the boundedness of the numerical solution in the @xmath4 norm , we present unconditionally optimal error estimates in section 4 in a simple and routine way .
numerical examples in two and three - dimensional spaces are presented in section 5 .
numerical results confirm our theoretical analysis and show that no time - step condition is needed .
let @xmath23 be a regular division of @xmath12 into triangles @xmath24 , @xmath25 , in @xmath26 or tetrahedras in @xmath27 , and let @xmath28 denote the mesh size . for a triangle @xmath24 with two corners ( or a tetrahedra with three corners ) on the boundary , we let @xmath29 denote the triangle with one curved side ( or a tetrahedra with one curved face ) . for an interior triangle , we simply set @xmath29 as @xmath24 itself .
finite element spaces on @xmath30 have been well defined , @xmath31 , see @xcite . for a given triangular ( or tetrahedral ) division of @xmath12
, we define the finite element space @xmath32 so that @xmath33 is a subspace of @xmath34 .
let @xmath35 be a coordinate transformation such that both @xmath36 and @xmath37 are lipschitz continuous and , for each triangle @xmath24 , @xmath36 maps @xmath24 one - to - one onto @xmath29 @xcite .
we define an operator @xmath38 by @xmath39 for @xmath40 .
then we set @xmath41 easy to see that @xmath42 is a finite element subspace of @xmath43 and @xmath44 where @xmath45 and @xmath46 is the lagrangian interpolation operator of degree @xmath18 .
let @xmath47 be a uniform partition of the time interval @xmath48 $ ] with @xmath49 and let @xmath50 for @xmath51 . for a sequence of functions @xmath52 , we define @xmath53 a simple linearized backward euler galerkin method for the problem ( [ e - parab-1])-([inic ] ) is to seek @xmath54 , @xmath55 , such that @xmath56 for any @xmath57 , with the initial condition @xmath58 for @xmath59 and @xmath60 for @xmath61 , where @xmath62 is a projection operator defined in section 3.2 . with a linear approximation to the nonlinear source term
, an alternative linearized scheme is defined by @xmath63 where @xmath64 , @xmath65 and @xmath66 , with @xmath67 and @xmath68 denoting the gradient of @xmath69 with respect to the components @xmath70 and @xmath71 , respectively .
the corresponding linearized crank - nicolson schemes can be defined similarly with classical extrapolations @xcite . in this paper
, we only focus our attention on the linearized scheme . the analysis presented in this paper
can be extended to the second linearized scheme and many other schemes .
we assume that @xmath72)$ ] and @xmath73 satisfies the weak ellipticity condition @xmath74
in this section , we assume that the solution to the problem ( [ e - parab-1])-([bc ] ) exists and satisfies that @xmath75 for some positive constant @xmath76 , and we prove the following theorem .
[ errestfemsol ] _ suppose that the system - has a unique solution @xmath7 satisfying the regularity condition . then there exist positive constants @xmath77 and @xmath78 , independent of @xmath79 and @xmath80 , such that the finite element system ( [ fdfem ] ) admits a unique solution @xmath81 when @xmath82 , and @xmath83 _ to prove theorem [ errestfemsol ] , we introduce a corresponding time - discrete equation as proposed in @xcite : @xmath84 with the boundary condition @xmath85 on @xmath86 and the initial condition @xmath87 . in the following two subsections , we estimate the error functions @xmath88 and @xmath89 , respectively , where @xmath10 is the solution of the time - discrete system . for the simplicity of notations ,
we denote by @xmath77 a generic positive constant and by @xmath90 a generic small positive constant , which depend solely upon @xmath76 , @xmath12 , @xmath91 , @xmath92 and @xmath69 , and independent of @xmath2 , @xmath80 and @xmath79 . by the regularity assumption ( [ regularity ] )
, we have @xmath93 .
we set @xmath94)}+\|\nabla u\|_{l^\infty(\omega \times [ 0,t])}+2 \ , .
{ \nonumber}\\ & q_k = [ -k , k]^{d+1 } \times \omega \times [ 0,t ] \ , .\end{aligned}\ ] ] then , by the regularity assumptions on @xmath69 and @xmath92 and the ellipticity condition ( [ strongellip ] ) , there exist positive constants @xmath95 and @xmath96 such that for @xmath97 and @xmath98 , @xmath99 & |\sigma'(s)|+|\sigma''(s)|\leq c_k , \label{g - sigma}\\[-5pt ] & |g(\alpha,\beta , x , t)| + |\partial_{\alpha } g(\alpha , \beta , x , t)| + \sum_{j=1}^d|\partial_{\beta_j}g(\alpha,\beta , x , t)| + \sum_{j=1}^d|\partial_{x_j}g(\alpha,\beta , x , t)|
\leq c_k { \nonumber}.\end{aligned}\ ] ] [ h23](@xmath100-estimate of elliptic equations @xcite ) _ suppose that @xmath101 is a solution of the boundary value problem @xmath102 where @xmath103 , @xmath1 , is a smooth and bounded domain .
then @xmath104 _ in this subsection , we explore the regularity of the solution to the time - discrete system ( [ e - tdparab-1 ] ) and present an error estimate for @xmath88 .
[ errestdissol ] _ suppose that the system - has a unique solution @xmath7 satisfying . then the time - discrete system ( [ e - tdparab-1 ] ) admits a unique solution @xmath105 such that @xmath106 and @xmath107 where @xmath108 is a positive constant independent of @xmath79 , @xmath80 and @xmath2 .
_ _ proof _ for the given @xmath10 , can be viewed as a linear elliptic boundary value problem . with the first inequality in ( [ g - sigma ] ) and classical theory of elliptic pdes , the equation admits a unique solution @xmath109 in @xmath110 .
let @xmath111 .
here we only prove the estimates - .
first , we prove by mathematical induction the inequality @xmath112 under the condition @xmath113 for some positive constant @xmath114 .
since @xmath87 , the inequality holds for @xmath115 .
now we assume that the inequality holds for @xmath116 .
let @xmath111 . from - and
, we see that @xmath117 satisfies the equation @xmath118\nabla u^{n+1 } ) + g(u^n,\nabla u^n , x , t)-g(u^n,\nabla u^n , x , t ) , { \nonumber}\end{aligned}\ ] ] with the boundary condition @xmath119 on @xmath86 and the initial condition @xmath120 , where @xmath121\\ + g(u^n,\nabla u^n , x , t ) -g(u^{n+1},\nabla u^{n+1},x , t)\end{aligned}\ ] ] is the truncation error due to the time discretization . by the regularity assumption ( [ regularity ] )
, we have @xmath122 multiplying the equation ( [ err - tdparab-01 ] ) by @xmath117 , we obtain @xmath123 by , we have further @xmath124 it follows that @xmath125 by choosing @xmath126 and using gronwall s inequality , there exists @xmath127 such that when @xmath128 @xmath129 which implies that @xmath130 applying lemma [ h23 ] for the linear elliptic equation with the induction assumption gives the @xmath131 estimate @xmath132 for @xmath133 . by the sobolev interpolation inequality , @xmath134 moreover , we rewrite the equation ( [ e - tdparab-1 ] ) as @xmath144 by lemma [ h23 ] and , @xmath145 which in turn implies that @xmath146 if we choose @xmath147 . by the sobolev interpolation inequality , @xmath148 which with shows that there exists @xmath149 such that @xmath150 for @xmath151 .
thus ( [ fsdioo123 ] ) holds for @xmath152 when @xmath153 and the induction is closed . from -
, we see that ( [ errestdissol211])-([tderrestlemm ] ) hold when @xmath154 .
secondly , we prove that ( [ errestdissol211])-([tderrestlemm ] ) hold for @xmath155 .
we assume that @xmath156 for some positive constant @xmath157 ( which may depend upon @xmath114 ) since @xmath158 . from
, it is easy to see that @xmath159 then we apply lemma [ h23 ] to . via a similar approach as ( [ sdfjkl23135 ] ) , we can derive that @xmath160 since @xmath161 , we take @xmath162 so that @xmath163 which further shows that @xmath164 thus the induction is complete for @xmath165 . combining the two cases , we complete the proof of theorem [ errestdissol ] . for @xmath166 ,
let @xmath167 be a projection defined by @xmath168 for all @xmath57 , and we set @xmath169 . with the regularity of @xmath10 proved in theorem [ errestdissol ] , we have the following inequalities : @xmath170 where ( [ proj-01])-([proj-02 ] ) are standard error estimates of elliptic equations , ( [ proj-03])-([proj-04 ] ) follow from @xcite and the references therein , ( [ proj-05])-([proj-06 ] ) can be proved in a similar way as in @xcite .
the following inverse inequalities will also be used in our proof .
let @xmath172 by the regularity assumptions for @xmath92 and @xmath69 , there exist @xmath173 and @xmath174 such that @xmath175 & \sigma(s)\geq \sigma_{k_1}^ * , \label{k1asfd}\end{aligned}\ ] ] for all @xmath176 $ ] and @xmath177 .
_ proof of theorem [ errestfemsol]_. we shall prove @xmath178 simultaneously by mathematical induction , where @xmath179 .
it is easy to see that the inequalities - hold for @xmath115 .
so we can assume that -([r>1 ] ) hold for @xmath133 .
by ( [ k122])-([k1asfd ] ) , the coefficient matrix of the linear system ( [ fdfem ] ) is symmetric and positive definite
. therefore , the system ( [ fdfem ] ) admits a unique solution @xmath180 in @xmath42 for @xmath55 . to complete the mathematical induction
, we need to prove - for @xmath152 . for this purpose , we consider two cases .
_ case i : _ @xmath59 .
in this case , @xmath197 and we can apply inverse inequalities for to get @xmath198 which implies that @xmath199 when @xmath200 . this completes the induction for @xmath59 , and - hold for all @xmath201 and @xmath202 .
_ case ii : _ @xmath61 . in this case ,
@xmath203 . to get the boundedness of @xmath204 , we present the @xmath110-estimate with an additional induction assumption : @xmath205 from the initial condition
, we see that holds for @xmath115 and we can assume that it holds for @xmath206 .
we substitute @xmath207 into ( [ errfemfdfem ] ) . with a similar approach to ,
we obtain @xmath208 using ( [ proj-03 ] ) and the sobolev embedding inequalities @xmath209 now by an inverse inequality , we have the estimate @xmath221 and so @xmath222 when @xmath223 .
it suffices to choose @xmath224 and @xmath225 so that the mathematical induction is closed when @xmath226 .
it follows that ( [ uh - bound ] ) , ( [ r>1 ] ) and ( [ r=1 ] ) hold for all @xmath227 when @xmath228 .
when @xmath229 and @xmath61 , we can see from that @xmath230 which together with an inverse inequality implies that @xmath231 if @xmath232 .
therefore , @xmath233 it suffices to choose @xmath224 and @xmath234 so that the mathematical induction is closed for @xmath229 .
it follows that ( [ uh - bound ] ) , ( [ r>1 ] ) and ( [ r=1 ] ) hold for all @xmath227 when @xmath235 .
combining the two cases , we complete the proof of theorem [ errestfemsol ] . 0.1 in * remark 3.1 * we have proved theorem [ errestfemsol ] for any @xmath18-order galerkin fems under the regularity assumption .
based on the classical theory of finite element approximation and interpolation , this assumption is enough to obtain optimal error estimates for linear and quadratic galerkin fems .
in fact , the optimal @xmath236 error bounds for @xmath61 and @xmath237 have been given in .
since the estimates in are @xmath2-independent , by an inverse inequality , @xmath238 by theorem [ errestdissol ] and the projection error estimates in , we have optimal error estimates for the linear and quadratic galerkin fems , which are summarized below .
[ r=2 ] _ under the assumptions of theorem [ errestfemsol ] , there exist positive constants @xmath77 and @xmath78 such that when @xmath82 , @xmath239 for @xmath61 or @xmath237 . _
based on the boundedness of the numerical solution proved in the last section , one can easily obtain optimal error estimates of any @xmath18-order galerkin fems under corresponding regularity assumptions , by following the classical approach of fem analysis .
also it is possible to present the optimal error estimate for @xmath240 as we did in section 3.2 .
however , this requires a rigorous analysis for stronger regularity of the time - discrete system . for simplicity
, we follow the classical fem approach and give a brief proof of optimal error estimates of the fully discrete galerkin fem . in this section , we assume that the solution to the initial - boundary value problem ( [ e - parab-1])-([bc ] ) exists , satisfying ( [ regularity ] ) and the following condition @xmath241 let @xmath242 where @xmath243 is the elliptic projection defined by @xmath244 for all @xmath57 , and we set @xmath245 .
easy to see that ( [ proj-01])-([proj-06 ] ) also hold for the projection operator @xmath246 .
[ asdfa123 ] _ suppose that the problem - has a unique solution @xmath7 satisfying ( [ regularity ] ) and ( [ regularity-2 ] ) for some positive integer @xmath18
. then there exists @xmath247 such that for @xmath248 , @xmath249 _ _ proof _ note that the error function @xmath250 satisfies the following equation : @xmath251 where @xmath252 is the truncation error satisfying ( [ trdfsjhklh ] ) . by the same approach as used in the proof of theorem [ errestfemsol ]
, we can derive that @xmath253 since @xmath254 as implied by theorem [ errestfemsol ] , by taking @xmath255 in ( [ e-1 ] ) , we derive that @xmath256 by gronwall s inequality , there exists a positive constant @xmath257 such that when @xmath258 , we have @xmath259 therefore , ( [ error - r>2 ] ) holds when @xmath258 . for @xmath260 , by theorem [ errestfemsol ] , @xmath261 .
the proof of theorem [ asdfa123 ] is complete .
_ example 5.1 _
first , we consider an artificial example governed by the equation @xmath262 in the domain @xmath263 with @xmath264 .
the function @xmath265 is chosen corresponding to the exact solution @xmath266 which satisfies the homogeneous dirichlet boundary condition .
a uniform triangular partition with @xmath267 nodes in each direction is used in our computation ( with @xmath268 ) .
we solve the system by the proposed method with a linear fe approximation up to the time @xmath269 . to illustrate our error estimates ,
we take @xmath270 and we present numerical results in table [ linear - l2 - 1 - 1 ] , from which we can see that the @xmath236 errors are proportional to @xmath271 . to demenstrate the unconditional convergence ,
we take several different spatial meshes with @xmath272 for each @xmath273 and we present numerical errors in table [ linear - l2 - 2 - 1 ] . based on our theoretical analysis , in this case ,
@xmath274 which tend to @xmath275 as @xmath276
. we can observe from table [ linear - l2 - 2 - 1 ] that for a fixed @xmath2 , numerical errors behave like @xmath275 as @xmath277 , which shows that no time step condition is needed . 0.1 in -0.2 in 0.1 in [ linear - l2 - 1 - 1 ] .@xmath236-norm errors of the linear galerkin fem with refined meshes ( example 5.1 ) .
[ cols="<,<,^,^,^,^ " , ]
we have presented unconditionally optimal error estimates of a class of linearized galerkin fems for general nonlinear parabolic equations , which may cover many physical applications .
the time - step size restriction was always a key issue in previous analysis and practical computation .
our theoretical analysis and numerical results show clearly that no time - step condition is needed for these linearized galerkin fems . our approach is based on a priori estimates of the numerical solution in the @xmath4 norm . with these estimates ,
optimal error estimates can be proved unconditionally from classical fem error analysis . clearly , our approach is applicable to many other time discretization schemes and more general nonlinear equations ( systems ) .
99 y. achdou and j.l .
guermond , convergence analysis of a finite element projection /lagrange - galerkin method for the incompressible navier - stokes equations , _ siam j. numer
_ , 37 ( 2000 ) , 799826 .
s. agmon , a. douglis , and l. nirenberg , estimates near the boundary for solutions of elliptic partial differential equations satisfying general boundary conditions , part i and part ii , _ comm .
pure appl .
_ , 12 ( 1959 ) , 623727 ; 127 ( 1964 ) , 3592 .
akrivis , v.a .
dougalis and o.a .
karakashian , on fully discrete galerkin methods of second - order temporal accuracy for the nonlinear schrdinger equation , _ numer .
_ , 59 ( 1991 ) , 3153 .
w. allegretto , y. lin and s. ma , existence and long time behaviour of solutions to obstacle thermistor equations , _ discrete and continuous dynamical syst . ,
series a _ , 8 ( 2002 ) , 757780 .
w. bao and y. cai , uniform error estimates of finite difference methods for the nonlinear schrdinger equation with wave operator ,
_ siam j. numer .
_ , 50 ( 2012 ) , 492521 .
byun and l. wang , elliptic equations with measurable coefficients in reifenberg domains , _ advances in mathematics _ , 225 ( 2010 ) , 26482673 .
cannon and y. lin , nonclassical @xmath110 projection and galerkin methods for nonlinear parabolic integro - differential equations , _ calcolo _ , 25 ( 1988 ) , 187201 .
z. chen and k. -h .
hoffmann , numerical studies of a non - stationary ginzburg - landau model for superconductivity , adv . math .
appl . , 5 ( 1995 ) , 363389 . ya - zhe chen and lan - cheng wu , _ second order elliptic equations and elliptic systems _ , translations of mathematical monographs 174 , ams 1998 , usa . z. deng and h. ma ,
optimal error estimates of the fourier spectral method for a class of nonlocal , nonlinear dispersive wave equations , _ appl .
_ , 59 ( 2009 ) , 9881010 .
j. douglas , jr . ,
r. ewing and m.f .
wheeler , a time - discretization procedure for a mixed finite element approximation of miscible displacement in porous media , _ rairo anal .
_ , 17 ( 1983 ) , 249265 .
t. dupont , g. fairweather and j.p .
johnson , three - level galerkin methods for parabolic equations , _ siam j. numer .
_ , 11 ( 1974 ) , 392410 .
durn , on the approximation of miscible displacement in porous media by a method of characteristics combined with a mixed method , _ siam j. numer .
_ , 25 ( 1988 ) , 9891001
elliott , and s. larsson , a finite element model for the time - dependent joule heating problem , _ math .
_ , 64 ( 1995 ) , 14331453 . v.j .
ervin , w.w .
miles , approximation of time - dependent viscoelastic fluid flow : supg approximation , _ siam j. numer
_ , 41 ( 2003 ) , 457486 .
ewing and m.f .
wheeler , galerkin methods for miscible displacement problems in porous media , _ siam j. numer .
_ , 17 ( 1980 ) , 351365 .
x. feng , y. he and c. liu , analysis of finite element approximation of a phasse field model for two - phase fluids , math .
comput . , 258 ( 2007 ) , 539571 .
yinnian he , the euler implicit / explicit scheme for the 2d time - dependent navier - stokes equations with smooth or non - smooth initial data , _ math .
_ , 77 ( 2008 ) , 20972124 .
convergence joule heating 50 ( 2010 ) , y. hou , b. li and w. sun , error analysis of splitting galerkin methods for heat and sweat transport in textile materials , siam j. numer .
, to appear .
b. kellogg and b. liu , the analysis of a finite element method for the navier stokes equations with compressibility , _ numer . math .
_ , 87 ( 2000 ) , 153170 .
ladyzenskaja , v.a .
solonnikov , and n.n .
uralceva , _ linear and quasilinear equations of parabolic type _ , translations of mathematical monographs 23 , providence , 1968 .
b. li and w. sun , error analysis of linearized semi - implicit galerkin finite element methods for nonlinear parabolic equations , _ int .
_ @xmath278 _ modeling _ , 10 ( 2013 ) , 622633 .
b. li and w. sun , unconditional convergence and optimal error estimates of a galerkin - mixed fem for incompressible miscible flow in porous media , submitted ( http://arxiv.org/abs/1207.1762v2 ) .
b. li , _ mathematical modelling , analysis and computation for some complex and nonlinear flow problems _ , phd thesis , city university of hong kong , hong kong , july , 2012 . b. liu , the analysis of a finite element method with streamline diffusion for the compressible navier stokes equations , _ siam j. numer
_ , 38 ( 2000 ) , 116 . h. ma and w. sun , optimal error estimates of the legendre - petrov - galerkin method for the korteweg - de vries equation , _ siam j. numer .
_ , 39 ( 2001 ) , 13801394 .
m. mu and y. huang , an alternating crank - nicolson method for decoupling the ginzburg - landau equations , siam j. numer .
, 35 ( 1998 ) , 17401761 .
r. rannacher and r. scott , some optimal error estimates for piecewise linear finite element approximations , _ math . comp .
_ , 38 ( 1982 ) , 437445 .
j. shen , on an unconditionally stable scheme for the unsteady navier stokes equations , _
j. comput .
_ , 8(1990 ) , 276288 .
simader , _ on dirichlet boundary value problem .
an @xmath279 theory based on a generalization of garding s inequality _ , lecture notes in math .
268 , springer , berlin , 1972 .
w. sun and z. sun , finite difference methods for a nonlinear and strongly coupled heat and moisture transport system in textile materials , _ numer math .
_ , 120 ( 2012 ) , 153187
. v. thome , _ galerkin finite element methods for parabolic problems _ , springer - verkag berkub geudekberg 1997 . y. tourigny , optimal @xmath110 estimates for two time - discrete galerkin approximations of a nonlinear schrdinger equation , _
i m a j. numer .
_ , 11 ( 1991 ) , 509 - 523 .
h. wang , an optimal - order error estimate for a family of ellam - mfem approximations to porous medium flow , _
siam j. numer .
_ , 46 ( 2008 ) , 21332152 .
k. wang , y. he and y. shang , fully discrete finite element method for the viscoelastic fluid motion equations , _ discrete contin .
b _ , 13 ( 2010 ) , 665684 .
wheeler , a priori @xmath236 error estimates for galerkin approximations to parabolic partial differential equations , _ siam j. numer .
_ , 10 ( 1973 ) , 723759 . h. wu , h. ma and h. li , optimal error estimates of the chebyshev - legendre spectral method for solving the generalized burgers equation , _ siam j. numer .
_ , 41 ( 2003 ) , 659672 .
yue , numerical analysis of nonstationary thermistor problem , _
j. comput .
_ , 12 ( 1994 ) , 213223 . z.q
. zhang and h. ma , a rational spectral method for the kdv equation on the half line , _ j. comput .
_ , 230 ( 2009 ) , 614625 .
w. zhao , convergence analysis of finite element method for the nonstationary thermistor problem , _
shandong daxue xuebao _ , 29 ( 1994 ) , 361367 .
m. zlmal , curved elements in the finite element method .
i@xmath280 , _ siam j. numer
_ , 10 ( 1973 ) , 229240 .
zouraris , on the convergence of a linear two - step finite element method for the nonlinear schrdinger equation , _
m2an math . model .
_ , 35 ( 2001 ) , 389405 . | the paper focuses on unconditionally optimal error analysis of the fully discrete galerkin finite element methods for a general nonlinear parabolic system in @xmath0 with @xmath1 . in terms of a corresponding time - discrete system of pdes
as proposed in @xcite , we split the error function into two parts , one from the temporal discretization and one the spatial discretization .
we prove that the latter is @xmath2-independent and the numerical solution is bounded in the @xmath3 and @xmath4 norms by the inverse inequalities . with the boundedness of the numerical solution , optimal error estimates can be obtained unconditionally in a routine way .
several numerical examples in two and three dimensional spaces are given to support our theoretical analysis .
* key words : * optimal error estimates , unconditional stability , galerkin , nonlinear parabolic system |
let @xmath3 be the automorphism group of a free group @xmath1 on @xmath4 generators @xmath5 .
a reduced word @xmath6 is called a _ palindrome _ if it is equal to its reverse @xmath7 . in @xcite collins
defines the _ palidromic automorphism group _ @xmath0 as the subgroup of @xmath3 consisting of all automorphisms @xmath8 for which @xmath9 is a palindrome for all @xmath10 .
he showed that the group was generated by three types of automorphisms : * maps @xmath11 , @xmath12 , which send @xmath13 and fix all other generators @xmath14 .
* maps @xmath15 which send @xmath16 and fix all other generators @xmath14 .
* maps corresponding to elements of the symmetric group @xmath17 which permute the @xmath18 among themselves .
the portion of @xmath0 generated by just the @xmath11 is called the _ elementary palindromic automorphism group _ of @xmath1 and denoted @xmath19 .
note that @xmath20 collins showed that a set of defining relators for @xmath19 is given by relations of the form 1 .
@xmath21 2 .
@xmath22 3 .
@xmath23 he remarked how similar this was to the relations for the pure symmetric automorphism group @xmath24 ( see gilbert s work in @xcite ) : 1 .
@xmath25 2 .
@xmath26 3 . @xmath27 where @xmath28 , @xmath12 , sends @xmath29 and fixes all other generators @xmath14 .
on the basis of this , collins conjectured that one could find the virtual cohomological dimension of @xmath0 by employing the methods of @xcite , as he did for @xmath30 in @xcite .
he also speculated that @xmath19 is torsion free , just as @xmath24 is .
we are able to answer both of these questions in this paper , as well as obtaining several interesting facts about the cohomology of @xmath0 .
[ c34 ] let @xmath0 be the palindromic automorphism group of the free group @xmath1 on @xmath4 letters and let @xmath19 be the subgroup of elementary palindromic automorphisms .
then * a ) * the virtual cohomological dimension of @xmath0 is @xmath31 . *
b ) * ( i ) for the prime 2 , the krull dimension of @xmath32 is @xmath4 . for odd primes @xmath33 ,
the krull dimension of @xmath34 is @xmath35 $ ] .
( ii ) in the range where the krull dimension of @xmath34 is 1 , the period is @xmath36 .
* c ) * the group @xmath19 is torsion free . * d ) * the cohomology group @xmath37 .
* e ) * if @xmath33 is an odd prime and @xmath38 , then the farrell cohomology of the palindromic automorphism group is the same as that of the symmetric group on @xmath33 elements : @xmath39 for analogous results concerning @xmath3 , see @xcite and @xcite .
see @xcite for the definition of the farrell cohomology @xmath40 of a group @xmath41 of finite vcd with coefficients in a @xmath41-module @xmath42 and also see @xcite for several useful properties of these cohomology groups .
the remainder of this paper is structured as follows . in section 2
, we discuss @xmath0 and note how it relates to some other groups , while in section 3 we introduce the space @xmath43 which @xmath0 acts on and prove parts a ) and d ) of theorem [ c34 ] .
section 4 is concerned with a realization proposition which allows us to establish parts b ) ( i ) and c ) of the main theorem . finally , section 5 looks in more detail at the cohomology of @xmath0 at odd primes @xmath33 and establishes parts b ) ( ii ) and e ) of the main theorem .
the authors would like to thank john meier for an enlightening conversation about symmetric automorphisms of free groups , aiding the presentation of this paper .
let @xmath44 be the spine of auter space ( see @xcite,@xcite,@xcite ) and @xmath45 .
let @xmath46 be the automorphism which sends @xmath16 for each @xmath10 .
define the @xmath47-graph @xmath48 to be a graph with 2 vertices and @xmath49 edges , where each edge goes from one vertex to the other one .
choose one of the two vertices of @xmath50 to be the basepoint @xmath51 , and define the rose @xmath52 to be the result of wedging together @xmath4 copies of @xmath50 at the basepoint .
the petals of the rose @xmath52 can be identified with the generators @xmath53 of @xmath1 , so that @xmath54 .
there is an action of @xmath55 on @xmath52 given by inverting each petal of the rose .
this action realizes the subgroup @xmath56 in the sense of @xmath57}$ ] ( also cf .
@xcite . ) an action of a group @xmath41 on a graph @xmath58 is _ without inversions _ if @xmath41 does not send any edge @xmath59 to its inverse @xmath60 , and an action is _ reduced _ if there are no @xmath41-invariant subforests in @xmath58 .
the action of @xmath61 on @xmath52 is both without inversions and reduced .
from now on , when we refer to a group action on a graph , it is assumed that the edges of the graph are subdivided as necessary to insure that the group acts without inversions .
note that the palindromic automorphism group @xmath0 is just @xmath62 .
this follows because an easy argument shows that every element of @xmath62 is palindromic , and because the generators of @xmath0 are all in @xmath62 .
for example , the generators @xmath11 are just products of @xmath61-nielsen transformations ( see @xcite where the @xmath41-nielsen transformation @xmath63 of a @xmath41-graph @xmath58 has the same vertex and edge set as @xmath58 but where the terminal point of an edge @xmath64 , @xmath65 , in the new graph is the initial point of the edge @xmath66 in the original graph ; this induces a map @xmath67 from the fundamental groupoid of the first graph to that of the second where @xmath64 is sent to @xmath68 and all other edges are sent to themselves . )
that is , if the petal @xmath53 of the rose consists of the edges @xmath69 , then @xmath11 is the composition @xmath70 . as a note for the curious
, it follows that @xmath0 and @xmath30 are distinct groups for @xmath71 , since a direct argument shows that @xmath30 has no element of order 2 in its center .
in addition , @xmath19 and @xmath24 are also obviously distinct ( for @xmath72 ; for @xmath73 they are the same group with the same presentation , ) since the former abelianizes to an elementary abelian @xmath74-group of rank @xmath75 while the latter abelianizes to a free abelian group of rank @xmath75 .
in addition to its formal palindromic properties , the group @xmath0 arises naturally from looking at hyperelliptic subgroups of mapping class groups ( cf .
gries @xcite for corresponding homological properties . )
we have a commutative diagram @xmath76 where @xmath77 is the mapping class group of an orientable surface of genus @xmath78 , @xmath79 is the mapping class group of an orientable surface of genus @xmath78 with 1 puncture , and @xmath80 is the mapping class group of an orientable surface of genus @xmath78 with two punctures , where each puncture is fixed pointwise . the map from @xmath80 to @xmath81
is obtained first by taking an intersection basis @xmath82 for the fundamental group of the surface @xmath83 with two punctures .
one of the punctures should serve as the basepoint for fundamental group considerations .
the other is treated as an actual puncture , so that the fundamental group of the surface minus this point is a free group @xmath84 on @xmath85 generators .
the map from @xmath80 to @xmath81 is now obtained by sending an element of @xmath80 to the automorphism of @xmath84 that it induces .
the map from @xmath79 to @xmath86 is obtained similarly .
let @xmath87 be a hyperelliptic involution ( see , for example , @xcite . )
then @xmath88 has @xmath89 fixed points , two of which are of course the punctures on the surface @xmath83 .
choose loops @xmath82 , based at one of the punctures , which form an intersection basis for the surface @xmath83 ( say the ones described in @xcite in the section on hyperelliptic riemann surfaces . ) by going along the top row of diagram [ c24 ] and then projecting downward , we see that the image of @xmath88 in @xmath90 is @xmath91 .
let @xmath92 be the image of @xmath88 in @xmath81 .
our goal is to show that @xmath92 is conjugate to @xmath93 in @xmath81 .
[ c25 ] if @xmath94 and the image of @xmath95 in @xmath96 is @xmath91 , then the image of any conjugate @xmath97 of @xmath95 , @xmath98 , is also @xmath91 .
this follows directly since @xmath91 is in the center of @xmath96 .
[ c26 ] if @xmath94 is an involution whose image in @xmath96 is @xmath91 , then @xmath95 can be realized on a marked graph whose underlying graph is the rose .
realize @xmath95 on a reduced marked graph whose underlying graph is @xmath99 .
first , we show that if @xmath59 is an edge of @xmath99 which is not fixed by @xmath95 , then we can assume that one endpoint of @xmath59 is the basepoint @xmath51 .
let @xmath100 .
choose a shortest path @xmath101 from @xmath59 to @xmath51 .
since @xmath102 , @xmath103 for every @xmath104 in the path @xmath101 .
so we can apply a sequence of nielsen transformations ( see @xcite ) and slide @xmath59 along @xmath101 to @xmath51 . note that since @xmath99 is reduced , @xmath105 now forms either a rose @xmath106 based at @xmath51 , or a @xmath50 . proceeding in this manner
, we can slide all of the edges of @xmath99 not fixed by @xmath95 to the basepoint . by way of contradiction , suppose that two of these @xmath50-graphs ( that have been moved so that one vertex of each @xmath50 is the basepoint ) , say @xmath107 and @xmath108 , share another common vertex in addition to the basepoint @xmath51 .
that is , suppose that there is a vertex @xmath109 and each of @xmath110 go from @xmath111 to @xmath51 .
say @xmath112 is the marked graph , and recall that @xmath113 . by replacing @xmath95 by a conjugate
if necessary ( see lemma [ c25 ] ) we can assume that @xmath78 sends the petal @xmath114 of @xmath52 to @xmath115 , the petal @xmath116 to @xmath117 , and the petal @xmath118 of @xmath52 to @xmath119 .
so in @xmath120 we have @xmath121 @xmath122 and @xmath123 hence the first column of @xmath124 in @xmath96 is @xmath125 , the second column is @xmath126 , and the third column is @xmath127 .
this contradicts the fact that @xmath128 .
if the result of sliding to the basepoint @xmath51 all edges of @xmath99 not fixed by @xmath95 yields a rose @xmath52 , then we are done .
otherwise , suppose by way of contradiction that there exist edges @xmath59,@xmath129 , and @xmath104 of @xmath99 such that * both @xmath59 and @xmath129 go from some vertex @xmath109 to @xmath51 .
* @xmath130 . *
@xmath104 goes from @xmath111 to @xmath111 . *
@xmath131 . as before , say @xmath112 is the marked graph . by replacing @xmath95 by a conjugate
if necessary ( see lemma [ c25 ] ) we can assume that @xmath78 sends the petal @xmath114 of @xmath52 to @xmath132 and the petal @xmath116 of @xmath52 to @xmath133 .
so in @xmath120 we have @xmath134 and @xmath135 hence the first column of @xmath124 in @xmath96 is @xmath125 and the second column is @xmath136 .
this contradicts the fact that @xmath128 .
[ c27 ] if @xmath94 is an involution whose image in @xmath96 is @xmath91 , then @xmath95 is conjugate in @xmath3 to @xmath61 .
realize @xmath95 on a marked graph @xmath137 . by replacing @xmath95 by a conjugate if necessary
, we can assume @xmath138 for all @xmath10 .
the involution @xmath95 of the graph @xmath52 must send the petal @xmath114 to some @xmath139 , since it is a graph automorphism .
but since @xmath140 , we see that @xmath139 must be @xmath141 .
similarly , we can see that the graph automorphism @xmath95 sends , for each @xmath10 , the petal @xmath53 to the petal @xmath142 .
this means our current @xmath95 is equal to @xmath61 ( and thus our original @xmath95 , before we replaced it by a conjugate , was conjugate to @xmath61 . )
the following corollary is immediate : [ c28 ] the image @xmath92 of @xmath88 in @xmath81 is conjugate to @xmath93 .
hence the hyperelliptic subgroup @xmath143 is conjugate to a subgroup of @xmath144 .
as remarked in @xcite , the image of @xmath0 in @xmath96 is the subgroup of @xmath96 consisting of invertible matrices where each column has exactly one odd entry ( and the rest are even . )
the subgroup is the semidirect product @xmath145 where @xmath146 is the 2-congruence subgroup defined by the short exact sequence @xmath147 and @xmath17 is standard inclusion of the symmetric group @xmath148
we define a certain contractible space @xmath43 , related to auter space , which @xmath0 acts on with finite stabilizers and finite quotient .
this allows us to obtain some cohomological results .
a graph @xmath58 is a _
@xmath50-tree of rank @xmath4 _ if there exists a pointed tree @xmath149 such that @xmath58 is obtained by `` doubling '' every edge of @xmath149 into a @xmath50-graph .
that is , the vertex set of @xmath58 is the same as the vertex set of @xmath149 and for every edge @xmath59 of @xmath149 going from @xmath111 to @xmath150 , @xmath58 has two edges @xmath151 and @xmath152 , both of which go from @xmath111 to @xmath150 .
there is a natural @xmath153-action on such a graph @xmath58 , which is given by switching the two edges in each @xmath50-graph .
note that the orbit space of @xmath58 under this action is just the tree @xmath149 .
[ c1 ] the reduced graphs @xmath58 which realize the subgroup @xmath56 of @xmath3 are exactly the @xmath50-trees of rank @xmath4 , where @xmath61 acts on the trees via their natural @xmath153-action . we have already mentioned that the rose @xmath52 realizes @xmath61 . from theorem 2 of @xcite , the other reduced graphs
@xmath58 which also realize @xmath154 are those that are nielsen equivalent to @xmath52 ( up to an equivariant isomorphism . )
if @xmath59 is an edge in one of the copies of @xmath50 in @xmath52 and @xmath129 is an edge in a different @xmath50 in @xmath52 , and both @xmath59 and @xmath129 point toward the basepoint , then nielsen transformation @xmath155 has the result of pulling the @xmath50-graph @xmath156 through the @xmath50-graph @xmath157 , so that now @xmath59 terminates at the initial vertex of @xmath129 , rather than at the basepoint @xmath51 . in other words , the result of applying one nielsen transformation to @xmath52
is that of sliding one of the petals of @xmath52 up though another petal .
a basic induction argument now yields that the result of applying a series of nielsen transformations fo @xmath52 will be some @xmath50-tree @xmath58 .
recall from @xcite that an edge @xmath59 of a @xmath41-graph @xmath58 is _ inessential _ if it is in every maximal @xmath41-invariant subforest of @xmath58 .
a @xmath41-graph @xmath58 is _ inessential _ if it has at least one inessential edge and is _ essential _ if it is not inessential .
let @xmath158 be the fixed point subspace of @xmath44 corresponding to some finite subgroup @xmath41 of @xmath3 . from @xcite ( cf .
part iii of @xcite and @xcite ) , both the centralizer @xmath159 and the normalizer @xmath160 act on the contractible space @xmath158 with finite stabilizers and finite quotient .
moreover , the space @xmath158 @xmath41-equivariantly deformation retracts to the space @xmath161 , where @xmath161 is constructed from @xmath158 by considering only essential marked graphs .
hence @xmath161 is a good space to study if one wishes to calculate the cohomology of @xmath159 or @xmath160 .
further recall that a @xmath41-graph @xmath162 is a @xmath41-equivariant _ blowup _ of a @xmath41-graph @xmath58 if some @xmath41-invariant subforest @xmath163 of @xmath162 can be collapsed away to yield @xmath58 .
let @xmath164 be some reduced marked graph realizing @xmath56 . from claim [ c1 ]
, @xmath58 is a @xmath50-tree .
blow up @xmath58 @xmath61-equivariantly to some maximal essential blowup @xmath162 .
[ c2 ] the fixed points / cells of the action of @xmath61 on @xmath162 are exactly the valence 2 vertices of @xmath162 . note that vertices in @xmath162 have valence 2 or 3 and that @xmath162 is obtained from @xmath58 by blowing up an oriented ideal forest ( see @xcite , @xcite , @xcite . )
briefly , ideal edges ( oriented ideal forests ) correspond to subsets of edges ( chains of subsets of edges ) which are pulled away from existing vertices in order to create new graphs which collapse down to the original graph . )
no edge is in @xmath165 because no edge is in @xmath166 and so blowing up ideal edges will not create any new edges that are fixed under the action of @xmath61 ( see @xcite page 229 . ) if a valence 3 vertex is in @xmath165 , then at least one edge of @xmath162 must be fixed by @xmath61 , which is a contradiction .
all of the valence 2 vertices of @xmath58 are in @xmath166 .
new valence 2 vertices created as ideal edges are blown up correspond to either : * old vertices of @xmath58 that used to be valence higher than 2 but have since had edges stripped ( pulled away ) from them .
these are in @xmath165 because all vertices of @xmath58 are in @xmath166 .
* new valence 2 vertices inserted to insure that @xmath61 acts on @xmath162 without inversions .
these are also clearly in @xmath165 .
note that @xmath167 .
cut @xmath162 along each of its valence 2 vertices , yielding a graph @xmath168 with the same number of valence 3 vertices and edges as @xmath162 had , no valence 2 vertices , and twice as many valence 1 vertices as @xmath162 had valence 2 vertices .
[ c3 ] @xmath169 the disjoint union of two trees @xmath170 and @xmath171 , where @xmath172 .
there is a covering map @xmath173 obtained by mapping to the orbit space under the @xmath61-action .
the forest collapse that sends @xmath162 to @xmath58 is @xmath61-equivariant , so it descends to a forest collapse of @xmath174 to @xmath175 .
( that the quotient of the forest upstairs in @xmath162 is also a forest in @xmath174 can be seen by an easy euler characteristic argument . ) but @xmath58 is a @xmath50-tree with a known @xmath61-action on it , and @xmath175 is a tree ( in fact , the underlying tree of the @xmath50-tree . ) hence @xmath174 is a tree .
since @xmath173 is a covering map with fiber two points , @xmath168 is as described .
let @xmath149 be a pointed tree with @xmath176 edges , all vertices valence either 1 or 3 , where @xmath51 is one of the valence 1 vertices .
( then @xmath149 has @xmath177 valence 1 vertices and @xmath31 valence 3 vertices . )
let @xmath178 and @xmath179 be two isomorphic copies of @xmath149 , and let @xmath180 be an isomorphism .
define @xmath181 define a @xmath61-action on @xmath182 by @xmath183 [ c4 ] there is a bijective correspondence between trees @xmath149 as above ( with @xmath176 edges , etc ) and maximal , essential blowups @xmath162 of reduced @xmath61-graphs .
the bijection is given by @xmath184 . from claims [ c2 ] and [ c3 ] , all blowups
@xmath162 have the required form .
finally , any @xmath182 can easily be reduced to a @xmath50-tree by collapsing edges , meaning that it is the blowup of such a graph .
all maximal simplices in @xmath43 have the same dimension , from @xcite .
maximal simplices in @xmath185 are constructed by taking chains of forest collapses from maximal blowups @xmath182 .
alternatively , we can define a _ subforest _ of @xmath149 to be a collection @xmath83 of edges of @xmath149 such that there is no path in @xmath83 from one valence 1 vertex to another .
( if there were such a path , then @xmath186 would be a cycle in @xmath182 . ) in this way , we can think of maximal simplices as coming from chains of subforests of various trees @xmath149 . since @xmath0 acts on the contractible space @xmath43 with finite stabilizers and finite quotient , the vcd of @xmath0 is at most the dimension of a maximal cell from @xmath185 .
such a cell comes from a chain of forest collapses of a tree @xmath149 with @xmath176 edges , @xmath177 valence 1 vertices , and @xmath31 valence 3 vertices .
hence we can collapse at most @xmath31 of the valence 3 vertices into other vertices while doing forest collapses , resulting in maximal simplices of dimension @xmath31 . to show that the vcd of @xmath0 is at least @xmath31
, we note that the subgroup generated by @xmath187 for @xmath188 is isomorphic to @xmath189 .
[ c6 ] let @xmath190 be the underlying graph of a particular marked graph in @xmath43 .
hence @xmath190 comes equipped with a @xmath61-action .
let @xmath191 be a simple closed curve in @xmath190 .
then @xmath192 contains exactly two points , and @xmath193 is the curve @xmath194 , or @xmath191 with the opposite of its original orientation .
@xmath190 can be blown up ( not necessarily uniquely ) to some maximal essential graph @xmath195 .
@xmath58 is the union of two isomorphic copies @xmath178 and @xmath179 of @xmath149 , where @xmath178 and @xmath179 are attached along their corresponding valence 1 vertices . as we collapse from @xmath58 to @xmath190 , the trees @xmath178 and @xmath179 collapse to trees @xmath196 and @xmath197 .
however , the attaching points for @xmath196 and @xmath197 are no longer necessarily just the valence 1 vertices , and could be other vertices as well .
let @xmath198 be a taut path in @xmath196 from one attaching point @xmath199 to some other attaching point @xmath200 , where furthermore there are no attaching points in the interior of @xmath198 .
let @xmath201 be the corresponding path in @xmath197 .
then @xmath202 is a simple closed curve and @xmath203 ) = @xmath204 , or the original curve oriented in the other direction .
hence the curve @xmath202 satisfies the conclusions of the lemma .
our goal is to show that any simple closed curve @xmath191 takes this form .
let @xmath205 be the portion of @xmath191 that is in @xmath196 and let @xmath206 be the portion of @xmath191 that is in @xmath197 .
since @xmath191 is a cycle and yet both @xmath196 and @xmath197 are trees , both @xmath205 and @xmath206 are nonempty . in fact , there must be a path @xmath198 in @xmath205 from one attaching point @xmath199 to some other attaching point @xmath200 , where there are no other attaching points in the interior of the path .
since @xmath191 is a simple closed curve , @xmath198 must be a taut path .
let @xmath207 , some other taut path in @xmath190 from @xmath199 to @xmath200 .
let @xmath208 be the map given by taking the quotient space under the action of @xmath61 .
note that @xmath209 is a path from @xmath199 to @xmath200 .
since @xmath196 is a tree and @xmath198 is the unique taut path in @xmath196 from @xmath199 to @xmath200 , this gives us that @xmath210 .
hence if @xmath59 is an edge in @xmath198 , then @xmath211 is two edges , @xmath212 and @xmath213 .
it follows that all of the oriented edges of the simple closed curve @xmath214 are in the simple closed curve @xmath191 .
hence @xmath215 [ c7 ] let @xmath190 be the underlying graph of a particular marked graph in @xmath43 . hence @xmath190 comes equipped with a @xmath61-action .
choose an edge @xmath59 in @xmath190 .
among all simple closed curves @xmath216 which pass through @xmath59 , choose one curve @xmath191 for which the distance from the curve to the basepoint @xmath51 is minimal .
then @xmath192 is two points @xmath199 and @xmath200 .
one of these two points in the closest point in @xmath191 to the basepoint @xmath51 and one of them is the farthest point in @xmath191 to the basepoint @xmath51 . using the notation of the proof of lemma [ c6 ]
, @xmath191 is the result of following some path @xmath198 in @xmath196 and then @xmath217 in @xmath197 , where @xmath198 goes from the attaching point @xmath199 to the attaching point @xmath200 .
since @xmath198 is a path in a pointed tree , there is a unique vertex @xmath150 in @xmath198 which is closest to @xmath51 . by way of contradiction , suppose that @xmath218 .
then @xmath150 is not an attaching point .
let @xmath219 be the unique taut path in @xmath196 from @xmath51 to @xmath150 .
let @xmath220 be the unique subpath of @xmath219 which contains @xmath150 and exactly one attaching point @xmath221 .
now let @xmath222 be the path in @xmath196 which starts at @xmath221 , follows @xmath220 along to @xmath150 , and then either follows @xmath191 from @xmath150 to @xmath200 or @xmath194 from @xmath150 to @xmath199 ( where we choose whichever possibility insures that @xmath223 . )
then the simple closed curve @xmath224 is closer to @xmath51 than @xmath191 is , which is a contradiction .
hence @xmath150 is @xmath199 or @xmath200 , and the lemma follows .
[ c8 ] let @xmath190 be a graph which occurs as an underlying graph of marked graphs in @xmath43 .
then there is only one possible @xmath61-action on @xmath190 .
we see that our task is to show that a unique @xmath61-action is determined by the properties about simple closed curves listed in lemmas [ c6 ] and [ c7 ] .
define an action @xmath225 on @xmath190 as follows .
let @xmath59 be an oriented edge of @xmath190 . among all simple closed curves @xmath216 which pass through @xmath59 , choose one path @xmath191 for which the distance from the curve to the basepoint @xmath51 is minimal .
let @xmath199 be a point on @xmath191 which is closest to the basepoint .
let @xmath4 be the edge - path distance in @xmath191 from @xmath199 to @xmath59 .
then there is an orientation @xmath226 such that if you traverse @xmath227 starting at @xmath199 and go @xmath4 edges , you get to @xmath59 . define @xmath228 to be the result of traversing @xmath229 starting at @xmath199 and then going @xmath4 more edges . by lemma [ c6 ] and [ c7 ] , the action @xmath225 is well defined and if any @xmath61 acts on @xmath190 , then the @xmath61-action and the @xmath225-action coincide .
denote by @xmath230 the quotient space @xmath185 .
[ c9 ] if two marked graphs in @xmath43 have the same underlying graph , then they correspond to the same vertex in @xmath230 .
that is , the moduli space @xmath230 can be formed by looking only at the poset structure of the underlying graphs of marked graphs in @xmath43
. from proposition [ c8 ] , any underlying graph of a marked graph in @xmath43 has only one possible @xmath61-action .
but from corollary 10.4 of @xcite , @xmath0 acts transitively on the set of marked @xmath61-graphs based on the same @xmath61-graph . the result follows .
the simplices of @xmath43 group themselves into cubes , as described in 3 of @xcite . in @xcite , hatcher and vogtmann show that the quotients in @xmath231 of cubes in @xmath44 have the rational homology of balls .
they use this to create a cubical chain complex which has the same rational homology as @xmath231 .
our goal here is to establish a similar result for the cubes of maximal dimension in @xmath230 , where this time we want the quotients of cubes to have the @xmath232-cohomology of balls , where @xmath33 is any odd prime .
following @xcite , we consider a maximal cube in @xmath43 .
it is given by considering a maximal essential marked graph @xmath233 and considering some maximal subforest @xmath83 of @xmath149 .
recall that by a subforest of @xmath149 we mean a subset @xmath83 of the edges of @xmath149 where there is no path in @xmath83 from one valence 1 vertex of @xmath149 ( or equivalently , one terminal edge of @xmath149 ) to another . from part
a ) of theorem [ c34 ] , @xmath83 has @xmath31 edges in it .
the cube corresponding to the pair @xmath234 can thus be thought of as imbedded in @xmath235 , where each coordinate vector is an edge of the cube , the graph obtained by collapsing each edge of @xmath236 is at the origin , and @xmath182 is at @xmath237 .
let @xmath238 be the group of all ( pointed ) automorphisms of the tree @xmath149 which take @xmath83 to @xmath83 .
the group @xmath239 acts linearly on the cube by permuting the coordinates of @xmath235 , and fixes the diagonal from @xmath240 to @xmath237 .
( the involution @xmath61 acts trivially on the cube , of course , since all of these cubes are coming from @xmath241 . )
hence , just as in @xcite , the quotient of the cube in @xmath230 is a cone with base @xmath242 , where @xmath243 is the boundary of the cube .
[ c10 ] the finite group @xmath244 ( and hence its subgroup @xmath238 ) is all @xmath74-torsion .
let @xmath245 .
now @xmath246 must take the basepoint to the basepoint , and so it must take the unique edge attached to the basepoint to itself . for each @xmath4
, let @xmath247 be the edges in @xmath149 which are at most distance @xmath4 from @xmath51 .
so @xmath248 is just one edge , and @xmath246 fixes it as already mentioned .
since all nonterminal vertices of @xmath149 have valence 3 , an inductive argument yields that @xmath249 fixes @xmath247 pointwise .
the following is an analog of proposition 3.1 of @xcite : [ c11 ] @xmath242 has the @xmath232-cohomology of an @xmath250-sphere or a ball .
the latter possibility happens when there is an element of @xmath238 which induces an odd permutation of the edges of @xmath83 .
the finite group @xmath238 , which is all 2-torsion , acts cellularly on @xmath243 , where the stabilizer of a cell fixes it pointwise .
we use the spectral sequence for equivariant cohomology ( cf .
@xcite vii 7 ) : @xmath251 \in \delta_n^r } h^s(stab(\delta ) ; { \mathbb{z}}_{(p ) } ) \rightarrow h_{aut(t , s)}^{r+s}(s^{n-2 } ; { \mathbb{z}}_{(p)})\ ] ] where @xmath252 $ ] ranges over the set @xmath253 of orbits of @xmath254-simplices @xmath255 in @xmath243 . since @xmath238 is all @xmath74-torsion and finite ,
so are all of the @xmath256 .
hence if @xmath257 , @xmath258 .
so the above spectral sequence converges to @xmath259 .
but another filtration yields a spectral sequence with @xmath260 it follows that @xmath261 unless @xmath262 is @xmath263 or @xmath264 . hence @xmath265 and @xmath266 .
the latter group of invariants is @xmath232 if the action of @xmath238 on @xmath243 preserves orientation and @xmath267 otherwise .
the last assertion in the proposition follows from corollary 3.2 of @xcite .
[ c14 ] the top dimensional cohomology group of @xmath230 vanishes .
that is , @xmath268 .
we show that the quotient of every maximal cube @xmath234 has a free face , so that the interior of the quotient of the cube can be collapsed away .
if we can do this , then @xmath230 will have the same @xmath232-cohomology as an @xmath250-dimensional complex , and we will be done . in the degenerate case where there is an element of @xmath238 which induces an odd permutation of the edges of @xmath83 , then the quotient of the cube @xmath234 is not itself a cube . in this case
, the diagonal from @xmath269 to @xmath270 is exposed in the quotient , and any @xmath250-dimensional simplex in the quotient which lies next to the diagonal is a free face . in the nondegenerate case ,
the quotient of the cube @xmath234 is itself a cube , although its boundary might be self indentified in various ways .
since the subforest @xmath83 of @xmath149 is maximal , @xmath83 must contain at least one terminal edge @xmath59 .
that is , one of the two vertices of @xmath59 is a valence 1 vertex or attaching point .
let @xmath190 be the graph obtained from @xmath195 by collapsing the subforest @xmath271 .
the graph @xmath190 has a maximal subforest corresponding to collapsing the edges @xmath59 and @xmath272 from the forest @xmath236 of @xmath58 .
hence we see that collapsing @xmath59 gives us a face , which we will denote by @xmath273 , of the cube @xmath234 .
it can be shown that this face corresponds to a ( nondegenerate , cubical ) face of the quotient of the cube @xmath234 because [ c15 ] there is an natural injection of @xmath274 into @xmath275 .
define the lift @xmath276 of an automorphism @xmath277 by sending an edge @xmath129 to @xmath278 if @xmath279 and letting @xmath280 .
denote by @xmath111 the valence 1 vertex of @xmath281 ( the attaching point ) and let @xmath150 be the other vertex of @xmath59 . in @xmath282 , @xmath283 .
we must show that @xmath95 sends @xmath150 to @xmath150 .
this follows automatically , however , as @xmath284 is the only valence 4 vertex of @xmath190 and so any automorphism of the graph must fix it .
let @xmath129 and @xmath78 be the two other edges in @xmath149 which share the vertex @xmath150 .
now if @xmath285 then @xmath95 could possibly exchange @xmath129 and @xmath78 , but this is fine as the lift @xmath276 also can .
if @xmath109 , then one of @xmath129 or @xmath78 must be closer to the basepoint , and so @xmath95 must fix both @xmath129 and @xmath78 .
regardless , @xmath276 can be defined as in the statement of the claim .
warning : note that if @xmath59 is not a terminal edge , the above claim is false .
collapsing an interior edge somtimes allows you to construct automorphisms with 3-torsion , which obviously can not be lifted to @xmath244 .
no automorphism @xmath95 of @xmath273 can induce an odd permutation of the edges in @xmath286 , else the lift @xmath276 of @xmath95 to @xmath149 would induce an odd permutation of the edge of @xmath83 .
since @xmath244 is all 2-torsion , it follows from the above claim that @xmath287 is also all 2-torsion . hence the same spectral sequence argument used in proposition [ c11 ] yields that the quotient of the cube corresponding to @xmath273 actually is a @xmath232-cohomology cube .
it remains to be shown that the cubical face corresponding to @xmath273 is free .
first , if another subforest @xmath288 with an edge @xmath289 of @xmath149 gives a cube with a face isomorphic to @xmath290 , then @xmath289 must also be a terminal edge of @xmath288 .
hence the isomorphism @xmath291 maps the vertex that @xmath59 collapsed into to the the vertex that @xmath289 collapsed into , and so we can lift the isomorphism to one from @xmath292 .
second , we must show that blowing up the vertex @xmath150 in @xmath190 only yields graphs isomorphic to @xmath58 .
this follows by considering the ways that the vertex @xmath150 in @xmath190 can be blown up .
say that the edges @xmath129 , @xmath78 , @xmath293 , and @xmath294 are the ones incident to @xmath150 . if the ideal edge orbit @xmath295 is blown up , we get back @xmath58 exactly , and if @xmath296 is blown up , we get a graph isomorphic to @xmath58 . as these are the only ways to blow up the graph @xmath61-equivariantly into another essential graph , we are done .
[ c16 ] @xmath297 .
that @xmath298 follows immediately from theorem [ c14 ] .
recall that @xmath0 acts with finite stabilizers and finite quotient @xmath230 on the contractible space @xmath43 . since the stabilizers are finite , their rational cohomology vanishes , and the standard equivariant spectral sequence yields that @xmath299 .
note that part d ) of theorem [ c34 ] follows from the above corollary . as a final remark for this section ,
we show that @xmath43 is an @xmath300 ( cf .
@xcite ) ; that is , for finite subgroups @xmath41 of @xmath0 , the fixed point subcomplex @xmath301 is contractible .
this follows directly from the corresponding property of @xmath3 .
the following proposition is unneccesary in the specific case of @xmath43 , since ( proof omitted ) @xmath43 actually equals @xmath302 .
this does not normally happen ( for example , the spaces @xmath303 mentioned later in fact [ c21 ] are not equal to the corresponding fixed point space of @xmath44 ) , however , and thus it seems worth noting the more general fact .
[ ptwo1 ] let @xmath83 be a finite subgroup of @xmath3 and let @xmath304 be either @xmath305 or @xmath306 .
let @xmath307 be the retract , defined by krstic and vogtmann and consisting of essential marked graphs , of the fixed point subcomplex @xmath308 of the spine of auter space @xmath44 .
then @xmath307 is an @xmath309 space .
let @xmath310 be a finite subgroup of @xmath304 and let @xmath41 be the ( finite , because @xmath311 ) subgroup generated by @xmath310 and @xmath83
. then @xmath312 , and @xmath158 is contractible from @xcite .
it remains to be shown that @xmath313 deformation retracts to @xmath314 .
given a marked graph @xmath58 representing a vertex of @xmath315 , we must show ( see proposition 3.3 of @xcite ) that for every edge @xmath59 in @xmath58 and every @xmath316 , @xmath59 is @xmath83-inessential if and only if @xmath317 is @xmath83-inessential .
this follows automatically from corollary 4.5 of @xcite , which characterizes essential edges by looking at the stabilizers ( in @xmath83 ) of paths in @xmath58 . since @xmath318 ,
the stabilizers in @xmath104-translates of such paths are still in @xmath83 and are isomorphic ( conjugate by @xmath104 ) to those of the original path .
let @xmath319 be a finite subgroup of @xmath0 and let @xmath320 be the ( finite ) subgroup generated by @xmath319 and @xmath61 . by zimmerman s @xcite realization theorem , we can realize @xmath320 by an action on an @xmath320-reduced graph @xmath58 . from the proposition below
, @xmath58 is also @xmath56-reduced ; that is , @xmath58 is a @xmath50-tree .
note that the corresponding statement is not true in @xmath321 ( have @xmath322 act on a @xmath47-graph @xmath323 and certainly would not be true in @xmath3 if the @xmath61-action were replaced by some other @xmath153-action .
[ c18 ] let @xmath324 be a finite subgroup of the palindromic automorphism group with @xmath325
. realize @xmath320 by an action on an @xmath320-reduced marked graph @xmath164 .
then @xmath164 is also a @xmath56-reduced marked graph . as before , let @xmath326 and identify the petals of the rose @xmath52 with the generators @xmath53 .
note that @xmath58 has no separating edges , else it would not be @xmath320-reduced . in this proof ,
when we refer to concepts such as the number of times an edge @xmath59 of @xmath58 occurs in some @xmath327 , we mean that we should take the unique taut path in @xmath58 , starting and ending at @xmath51 , which is homotopic to the path @xmath327 in @xmath58 , and then count the number of times @xmath59 occurs in this taut path . by way of contradiction , suppose @xmath58 is not @xmath56-reduced .
let @xmath328 be an edge of minimal distance to the basepoint @xmath51 such that @xmath329 is a forest .
* case 1 : * @xmath330 .
since @xmath151 is not a separating edge of @xmath58 , we can choose a nontrivial cycle @xmath331 , starting and ending at @xmath51 , which has just one occurence of @xmath151 and none of @xmath332 . if for all @xmath333 , the cycles @xmath327 have an even number of occurences of @xmath334 , then we could not write @xmath331 as a product of them and their inverses .
so some @xmath335 has an odd number of occurences of @xmath336 in it .
say that the exponent sum of @xmath151 in @xmath335 is @xmath337 , @xmath337 odd .
then the exponent sum of @xmath151 in @xmath338 is still @xmath337 , but the exponent sum of @xmath339 is @xmath340 .
this contradicts the fact that @xmath341 * case 2 : * @xmath342 .
let @xmath8 be a shortest length path from @xmath51 to @xmath151 .
say without loss of generality that @xmath151 is the oriented edge from @xmath111 to @xmath150 and that @xmath8 goes from @xmath51 to @xmath111 .
let @xmath343 .
then @xmath344 is a shortest length path from @xmath51 to @xmath345 .
now @xmath346 , else we could write @xmath347 and get @xmath348 as a @xmath61-invariant forest closer to @xmath51 than @xmath107 is .
( if @xmath349 , then @xmath285 and so @xmath350 necessarily . )
so we have both @xmath8 and @xmath344 are paths from @xmath51 to @xmath111 moreover , @xmath351 ( else @xmath329 is not a forest . )
now @xmath352 is not a forest , since @xmath58 is @xmath320-reduced .
hence we can choose some simple closed curve @xmath331 in @xmath353 that contains @xmath151 .
there must exist some @xmath354 , @xmath355 , such that @xmath356 .
otherwise we could deformation retract @xmath331 to the set of vertices @xmath357 , which contradicts the fact that @xmath331 is a simple closed curve .
now @xmath358 , as @xmath359 .
hence @xmath360 is a nontrivial cycle starting and ending at @xmath51 which contains exactly one occurence of @xmath151 and none of @xmath345 .
so there must be a @xmath361 , @xmath362 , which contains an odd number of occurences of @xmath334 and an even number of occurences of @xmath363 .
( if we had some even / odd @xmath335 , then we could act by @xmath61 to get odd / even , and this would be a @xmath339 .
otherwise , all @xmath327 are all even / even or odd / odd , and so combine together just to get more even / even or odd / odd loops . )
this is a contradiction , however , because @xmath364 still has an odd number of occurences of @xmath334 while @xmath365 has an even number of occurences of @xmath334 . from the action of @xmath366 on the rose @xmath52
such that the @xmath10th generator inverts the @xmath10th petal and leaves all others fixed , we know that the krull dimension at the prime 2 is at least @xmath4 . similarly , there is an action of @xmath367}$ ] on @xmath52 where the first @xmath368 rotates the first @xmath33 petals , the second @xmath368 rotates the next @xmath33 petals , etc .
hence the krull dimension at the prime @xmath33 is at least @xmath35.$ ] let @xmath320 be a maximal rank elementary abelian subgroup of @xmath0 . from proposition [ c18 ] , we can realize @xmath320 by an action of @xmath320 on a @xmath61-graph @xmath58 which is both @xmath320-reduced and @xmath61-reduced .
that is , we have an action of @xmath320 on a pointed @xmath369-tree @xmath58 . since elements of @xmath320 must preserve basepoints , the action of @xmath320 on the tree @xmath175 does not invert edges .
hence we have inclusions @xmath370 the result ( for 2 or odd primes @xmath33 ) now follows from standard facts about @xmath17 ( cf .
theorem 1.3 in chapter vi of @xcite . )
we sketch the proof , which uses standard methods .
suppose that some @xmath371 lies in @xmath19 . from proposition [ c18
] , we can realize @xmath320 by an action of @xmath320 on a @xmath61-graph @xmath58 which is both @xmath320-reduced and @xmath61-reduced .
let @xmath164 be the corresponding marked graph .
let @xmath372 , a pointed tree with an @xmath320-action on it .
first , suppose that the @xmath320-action on @xmath149 is nontrivial .
( this will always be the case if @xmath33 is odd . )
then there are two edges @xmath151 and @xmath152 , both oriented so that their terminal vertices are closer to the basepoint than their initial vertices , of @xmath58 such that a generator of @xmath320 rotates the edge @xmath373 $ ] into the edge @xmath374 $ ] in @xmath149 .
some generator @xmath375 of @xmath1 must be such that @xmath376 contains an odd number of occurences of @xmath151 in it .
choose @xmath377 similarly .
then @xmath378 is a palindromic word in the edges of @xmath58 with either @xmath379 or @xmath380 in the middle of the palindrome .
the generator of @xmath320 ( thought of as an element of @xmath19 ) must send @xmath375 to a palindrome with either @xmath377 or @xmath381 in the center of it .
this contradicts the fact that all elements of @xmath19 send generators @xmath53 to palidromes with @xmath53 in the center of them .
the only remaining case is where @xmath382 and @xmath320 acts trivally on @xmath149 .
so @xmath320 is a subgroup of the group @xmath366 of graph automorphisms of @xmath58 which act by inverting the @xmath369 s in the @xmath369-tree @xmath58 .
hence the generator of @xmath320 corresponds to an element of @xmath0 which , for at least one @xmath10 , sends @xmath53 to a palindrome with @xmath142 in its center . as none of these automorphisms
are in @xmath19 , we again have a contradiction .
let @xmath33 be an odd prime ( as will always be the case from now on in this paper . )
we wish to calculate the farrell cohomology of @xmath0 using ken brown s @xcite normalizer spectral sequence , which states that @xmath383 where @xmath41 is a group with finite virtual cohomological dimension , @xmath384 is the poset of nontrivial elementary abelian @xmath33-subgroups of @xmath41 , @xmath385 is the poset of conjugacy classes of nontrivial elementary abelian @xmath33-subgroups of @xmath41 , and @xmath386 is the set of @xmath254-simplices in the realization @xmath387 . a first step toward performing such a calculation is calculating @xmath387 .
in other words , we wish to calculate conjugacy classes of elementary abelian subgroups @xmath388 . by proposition [ c18 ]
, we can realize such finite groups @xmath389 by reduced actions on @xmath50-trees . if @xmath390 , define a particular subgroup @xmath391 of @xmath0 by letting @xmath392 act on the rose @xmath52 by rotating its first @xmath33 leaves and leaving the last @xmath393 leaves fixed .
that is , @xmath392 corresponds to automorphisms which rotate the first @xmath33 generators @xmath394 and leave the remaining generators fixed .
we show that @xmath392 is the only conjugacy class of nontrivial elementary abelian @xmath33-subgroups that is in @xmath0 . by proposition [ c18 ]
, we see that an arbitrary nontrivial elementary abelian @xmath33-subgroup @xmath320 comes from some action on a @xmath50-tree with @xmath33-symmetry .
since @xmath395 , the only possibility is that @xmath320 acts on a @xmath50-tree @xmath58 by rotating @xmath33 of the @xmath50-leaves and leaving the other @xmath393 @xmath50-edges in the tree fixed .
but it is clear that a product of @xmath397-nielsen transformations takes the rose @xmath52 to the graph @xmath58 , and hence we see that @xmath320 and @xmath392 are conjugate to each other in @xmath0 .
the @xmath406 in the above decomposition comes from automorphisms of @xmath1 which permute the first @xmath33 generators and leave the remaining @xmath407 fixed .
the @xmath403 being acted upon in the semidirect product structure above has @xmath10th generator @xmath408 .
the @xmath404 is the involution which inverts the first @xmath33 generators of @xmath1 and leaves the remaining @xmath407 fixed .
finally , the @xmath402 comes from automorphisms which fix the first @xmath33 generators of @xmath1 and act on the last @xmath407 generators by identifying the subgroup @xmath409 with @xmath403 .
consider the action of @xmath410 on the rose @xmath52 .
@xmath392 rotates the first @xmath33 petals .
label the first @xmath33 petals of the rose as @xmath394 as before , but label the last @xmath407 petals as @xmath411 . since @xmath412 is an odd prime , @xmath413 and in lemma 5.1 of @xcite , we calculated @xmath414 where the first @xmath403 in @xmath415 is the free group on the @xmath392-nielsen transformations @xmath416 for @xmath417 and the latter @xmath403 is the free group on the @xmath392-nielsen transformations @xmath418 , @xmath417 .
note that @xmath419 acts on @xmath415 via @xmath420 and @xmath421 .
in other words , if @xmath422 then @xmath423 .
let @xmath41 be the subgroup @xmath424 of @xmath425 , where @xmath403 is the free group on the generators @xmath426 for @xmath427 , and @xmath428 is included in @xmath429 in the obvious way .
it follows directly that @xmath430 to prove the proposition , we must show that they are equal .
take an arbitrary @xmath431 say @xmath432 , where @xmath433 , @xmath422 , @xmath434 , and @xmath435 .
since @xmath436 , @xmath437 also .
so @xmath438 and @xmath439 this means that the map @xmath440 sends @xmath53 to @xmath142 for @xmath441 and @xmath442 to @xmath443 for @xmath444 .
now both @xmath61 and @xmath59 restrict to maps in @xmath445 and moreover @xmath446 and @xmath447 both restrict to the identity map in @xmath445 . hence for @xmath444 , we have @xmath448 and we see that @xmath449 restricts to @xmath405 in @xmath429 . as @xmath435 ,
this means @xmath450 .
hence @xmath451 also .
since @xmath452 , this gives @xmath453 .
in other words , we have @xmath454 and @xmath455 it follows that @xmath456 so @xmath457 and @xmath458 . in summary , we have shown that an arbitrary element @xmath459 has @xmath460 and @xmath450 .
thus @xmath461 , as desired .
define a _ @xmath33-admissible tree _
@xmath149 to be a triple @xmath465 where @xmath149 is a pointed tree , @xmath466 is a vertex of @xmath149 ( which may be the basepoint @xmath51 ) , @xmath320 is a subset of the vertices of @xmath149 called the _ set of attaching points _ , @xmath467 , and all valence 1 vertices of @xmath149 are in @xmath320 . for a @xmath33-admissible tree @xmath149 ,
define the _ corresponding graph _ @xmath182 as follows : take two isomorphic copies @xmath178 and @xmath179 of the tree @xmath149 , and let @xmath180 be an isomorphism .
then let @xmath468 be the graph @xmath469 let @xmath470 be a @xmath47-graph with @xmath33 edges and two vertices @xmath199 and @xmath200 .
let @xmath471 be the @xmath466-vertex in @xmath178 and let @xmath472 be the @xmath466-vertex in @xmath179 .
finally , let @xmath473 if @xmath474 , then say @xmath149 is a @xmath33-admissible tree of _ rank _ @xmath4 . if @xmath149 is a @xmath33-admissible tree of rank @xmath4 , define a @xmath56-action on the edges of @xmath182 by @xmath475 since this action inverts the edges of the @xmath47-graph in @xmath182 , we then need to subdivide these edges so that the group acts without inversions .
next , define a @xmath392-action on @xmath182 by having @xmath392 fix @xmath468 and rotate the edges of @xmath470 cyclically . in this way , @xmath182 is a @xmath397-graph . a @xmath33-admissible tree is @xmath149 _ reduced _ if the corresponding @xmath397-graph @xmath182 is reduced ; that is , if all vertices of @xmath149 are attaching points .
similarly , a @xmath33-admissible tree @xmath149 is a _
maximal _ if the attaching points of @xmath149 are exactly its valence 1 vertices , the valence @xmath74 vertices of @xmath149 consist of just the point @xmath466 , and @xmath149 has no vertices with valence 4 or more .
as before , a _ subforest
_ of @xmath149 is a collection of edges @xmath83 of @xmath149 such that there is no path in @xmath83 from one attaching point to another .
lastly , isomorphisms of @xmath33-admissible trees must be graph isomorphisms which take @xmath51 to @xmath51 , @xmath466 to @xmath466 , and @xmath320 to @xmath320 . 1 .
there is a bijective correspondence between reduced @xmath33- admissible trees of rank @xmath4 and the underlying graphs of @xmath397- reduced marked graphs , given by @xmath476 .
2 . there is a bijective correspondence between maximal @xmath33-admissible trees of rank @xmath4 and the underlying graphs of maximal essential marked @xmath397-graphs , given by @xmath476 .
the virtual cohomological dimension of @xmath462 is @xmath477 .
4 . let @xmath58 be a graph which occurs as the underlying graph of a marked graph in @xmath463 .
then there is only one possible @xmath61-action on @xmath58 .
if two marked graphs in @xmath463 have underlying graphs which correspond to the same @xmath33-admissible tree , then they correspond to the same vertex in @xmath478 .
that is , we can form the moduli space @xmath478 by looking only at the poset structure of the @xmath33-admissible trees corresponding to marked graphs in @xmath463 .
the top dimensional cohomology class of @xmath478 , with coefficients in @xmath232 , vanishes .
that is , @xmath479 .
@xmath480 . note that ( 4 ) and ( 5 ) above are a little bit different from their analogs proposition [ c8 ] and corollary [ c9 ] .
basically , the underlying graphs @xmath58 always have just one possible @xmath61-action , as before , but it is conceivable ( for example , if the graph contains two or more copies of @xmath470 inside it and we must decide which one @xmath392 rotates ) that it might have several possible @xmath392-actions . that is why we talk about @xmath33-admissible trees instead in ( 5 ) , since the vertex @xmath466 in the tree determines where the @xmath33 edges that @xmath392 rotates are located .
we use the equivariant cohomology spectral sequence for @xmath462 acting on the contractible space @xmath463 with finite stabilizers and finite quotient @xmath478 .
the equivariant cohomology spectral sequence for this action is @xmath482 \in \delta_{n}^r } \hat h^s(stab_{n_{aut(f_n)}(p_n \times \langle\sigma_n\rangle)}(\delta ) ; { \mathbb{z}}_{(p ) } ) \hfill \cr \hfill \rightarrow \hat h^{r+s}(n_{aut(f_n)}(p_n \times \langle\sigma_n\rangle ) ) ; { \mathbb{z}}_{(p ) } ) \cr}\ ] ] where @xmath252 $ ] ranges over the set @xmath483 of orbits of @xmath254-simplices @xmath255 in @xmath463 . from the decomposition
@xmath484 we see that @xmath485 has @xmath33-rank 0 .
since @xmath406 acts trivially on marked graphs in @xmath463 by permuting the edges of the @xmath47-graph attached at @xmath466 , it follows that for every simplex @xmath255 we have @xmath486 the @xmath487-page of the spectral sequence is @xmath267 in the rows where @xmath488 and a copy of the cellular cochain complex with @xmath368-coefficients of the @xmath489-dimensional complex @xmath478 in rows @xmath490 .
it follows that the @xmath491-page has the form : @xmath492 hence we see that the spectral sequence converges at the @xmath491-page .
* case 2 : * m=1 .
then @xmath478 is a contractible 1-dimensional complex with 3 vertices and two edges .
define the maximal @xmath33-admissible tree @xmath149 of rank @xmath4 to be the tree with three vertices @xmath51 , @xmath466 , @xmath111 and two edges @xmath151 , @xmath152 where @xmath151 goes from @xmath51 to @xmath466 and @xmath152 goes from @xmath466 to @xmath111 .
the middle vertex of the 1-dimensional complex @xmath478 corresponds to the graph @xmath182 .
the other two vertices and two edges @xmath478 correspond to the two possible ways that @xmath182 can be collapsed equivariantly .
* case 3 : * m=2 .
then @xmath478 is a 2-dimensional complex with 13 vertices , 28 edges , and 16 two - simplices .
there are two maximal graphs in @xmath478 .
simplices coming from the first graph are listed in figure [ hyper1 ] and simplices from the second graph are listed in figure [ hyper2 ] . in figures
[ hyper1 ] and [ hyper2 ] , the maximal graphs are listed in the center .
these maximal graphs can be collapsed in various ways , and these are listed around the periphery of the figures . in the graphs , a solid dot indicates the basepoint @xmath51 and the hollow dots represent attaching points @xmath466 for the @xmath47-graph @xmath470 . if there is only one hollow dot in a graph , both ends of the @xmath47-graph should be attached to that one vertex . upon identifying the boundaries of the simplices listed in figures [ hyper1 ] and [ hyper2 ]
, we obtain the complex @xmath478 pictured in figure [ hyper3 ] .
the complex is homeomorphic to the fletching of a dart , three half disks , all identified along a common line in their boundary .
this complex is clearly contractible .
d. j. collins , _ palindromic automorphisms of free groups _ , _ combinatorial and geometric group theory ( edinburgh , 1993 ) _ , london math .
lecture note ser . , 204 , cambridge university press cambridge ( 1995 ) 63 - 72 . | we examine the palindromic automorphism group @xmath0 of a free group @xmath1 , a group first defined by collins in @xcite which is related to hyperelliptic involutions of mapping class groups , congruence subgroups of @xmath2 , and symmetric automorphism groups of free groups .
cohomological properties of the group are explored by looking at a contractible space on which @xmath0 acts properly with finite quotient .
our results answer some conjectures of collins and provide a few striking results about the cohomology of @xmath0 , such as that its rational cohomology is zero at the vcd . _
dedicated to peter hilton _ : + doc , note , i dissent .
a fast never prevents a fatness .
i diet on cod . |
computing maximal ( non - extendable ) repetitions in a word is a classical topic in the area of string algorithms ( see for example @xcite and references therein ) .
maximal repetitions of substrings , also called _ runs _ , give information on the repetitive regions of a word , and are used in many applications , for example in the analysis of genomic sequences .
kolpakov and kucherov @xcite gave the first linear - time algorithm for computing all the runs in a word and conjectured that any word of length @xmath2 contains less than @xmath2 runs .
recently , bannai et al .
@xcite , using the notion of lyndon roots of a run , proved this conjecture and designed a much simpler algorithm computing the runs . here
we deal with a generalization of this problem to the commutative setting .
recall that an abelian power is a concatenation of two or more words that have the same parikh vector , i.e. , that have the same number of occurrences of each letter of the alphabet .
for example , @xmath10 is an abelian square , since @xmath11 and @xmath12 both have two @xmath13 s and one @xmath14 , i.e. , the same parikh vector @xmath15 . when an abelian power occurs within a word
, one can search for its `` maximal '' occurrence by extending it to the left and to the right character by character without violating the condition on the number of occurrences of each letter .
following the approach of constantinescu and ilie @xcite , we say that a parikh vector @xmath1 is an abelian period of a word @xmath0 if @xmath0 can be written as @xmath16 for some @xmath17 where for @xmath18 all the @xmath19 s have the same parikh vector @xmath1 and the parikh vectors of @xmath20 and @xmath21 are contained in @xmath1 .
if @xmath22 , we say that the word @xmath0 is periodic with period @xmath1 .
note that the factorization above is not necessarily unique .
for example , @xmath23 and @xmath24 ( @xmath25 denotes the empty word ) are two factorizations of the word @xmath26 both corresponding to the abelian period @xmath27 .
moreover , the same word can have different abelian periods .
in this paper we define an _ abelian run _ of period @xmath1 in a word @xmath0 as an occurrence of a substring @xmath28 of @xmath0 such that @xmath28 is periodic with abelian period @xmath1 and this occurrence can not be extended to the left nor to the right by one letter into a substring periodic with period @xmath1 . for example
, let @xmath29 . then the prefix @xmath30 $ ] has abelian period @xmath31 but it is not an abelian run since the prefix @xmath32 $ ] also has abelian period @xmath31 .
the latter , on the other hand , is an abelian run of period @xmath31 in @xmath0 .
looking for abelian runs in a word can be useful to detect regions in the word where there is some kind of non - exact repetitiveness , for example regions with several consecutive occurrences of a substring or its reversal .
matsuda et al .
@xcite recently presented an offline algorithm for computing all abelian runs of a word of length @xmath2 in @xmath8 time .
notice that , however , the definition of abelian run in @xcite is slightly different from the one we consider here .
we compare both versions in section [ sect - def ] .
basically , our notion of abelian run is more restrictive than the one of @xcite , for which we use the term `` anchored run '' .
we first present an online algorithm that , given a word @xmath0 of length @xmath2 over an alphabet of cardinality @xmath3 and a parikh vector @xmath1 , returns all the abelian runs of period @xmath1 in @xmath0 in time @xmath4 and space @xmath5 , where @xmath6 is the norm of @xmath1 , that is , the sum of its components .
this algorithm improves upon the one given in @xcite which runs in time @xmath33 .
next , we give an @xmath7-time online algorithm for computing all the abelian runs with periods of norm @xmath6 of a word of length @xmath2 , for any given @xmath6 .
finally , we present an @xmath8 ( resp .
@xmath34 ) -time offline randomized ( resp .
deterministic ) algorithm for computing all the abelian runs of a word of length @xmath2 .
the rest of this article is organized as follows .
[ sect - def ] introduces central concepts and fixes the notation . in sect .
[ sect - prev ] we review the results on abelian runs given in @xcite . sect .
[ sect - new ] is devoted to the presentation of our main result : a new solution for computing the abelian runs for a given parikh vector . in sect .
[ sect - newer ] we apply this algorithm in a procedure for computing the abelian runs with periods of a given norm .
next , in sect .
[ sect - all ] , we design a solution for computing all the abelian runs , which builds upon the result recalled in sect .
[ sect - prev ] .
finally , we conclude in sect .
[ sect - conc ] .
let @xmath35 be a finite ordered alphabet of cardinality @xmath3 , and let @xmath36 be the set of finite words over @xmath37 .
we assume that the mapping between @xmath38 and @xmath39 can be evaluated in constant time for @xmath40 .
we let @xmath41 denote the length of the word @xmath42 . given a word @xmath43 $ ] of length @xmath44 , we write @xmath45 $ ] for the @xmath46-th symbol of @xmath42 and , for @xmath47
, we write @xmath48 $ ] to denote a fragment of @xmath42 from the @xmath46-th symbol to the @xmath49-th symbol , both included .
this fragment is an occurrence of a substring @xmath50\cdots w[j]$ ] . for @xmath51 ,
@xmath52 $ ] denotes the empty fragment .
we let @xmath53 denote the number of occurrences of the symbol @xmath54 in the word @xmath42 .
the parikh vector of @xmath42 , denoted by @xmath55 , counts the occurrences of each letter of @xmath37 in @xmath42 , that is , @xmath56 .
notice that two words have the same parikh vector if and only if one word is a permutation ( i.e. , an anagram ) of the other
. given the parikh vector @xmath55 of a word @xmath42 , we let @xmath57 $ ] denote its @xmath39-th component and @xmath58 its norm , defined as the sum of its components . thus , for @xmath59 and @xmath60
, we have @xmath57=|\s{w}|_{a_i}$ ] and @xmath61=|\s{w}|$ ] .
finally , given two parikh vectors @xmath62 , we write @xmath63 if @xmath64\le \qv[i]$ ] for every @xmath65 .
if additionally @xmath66 , we write @xmath67 and say that @xmath1 is contained in @xmath68 .
[ def - ap ] a factorization @xmath69 satisfying @xmath17 , @xmath70 , and @xmath71 is called a _ periodic factorization _ of @xmath0 with respect to @xmath1 .
if a word @xmath42 admits such a factorization , we say that @xmath1 is an _ ( abelian ) period _ of @xmath42 .
we call fragments @xmath72 and @xmath73 respectively the _ head _ and the _ tail _ of the factorization , while the remaining factors are called
_ cores_. note that the head and the tail are of length strictly smaller than @xmath74 ; in particular they can be empty .
observe that a periodic factorization with respect to a fixed period is not unique .
however , it suffices to specify @xmath75 to indicate a particular factorization ; see @xcite .
dealing with factorizations of fragments of a fixed text , it is more convenient to use a different quantity for this aim .
suppose @xmath76=u_0\cdots u_k$ ] is a factorization with respect to abelian period @xmath1 with @xmath77 .
observe that consecutive starting positions @xmath78 of factors @xmath79 differ by exactly @xmath6 .
hence , they share a common remainder modulo @xmath6 , which we call the _ anchor _ of the factorization .
note that the anchor does not change if we trim a factorization of @xmath76 $ ] to a factorization of a shorter fragment , or if we extend it to a factorization of a longer fragment .
a fragment @xmath76 $ ] has an abelian period @xmath1 _ anchored _ at @xmath80 if it has a periodic factorization with respect to @xmath1 whose anchor is @xmath81 .
if @xmath0 has a factorization with at least two cores , we say that @xmath0 is _ periodic _ with period @xmath1 ( anchored at @xmath80 if @xmath81 is the anchor of the factorization ) .
a fragment @xmath76 $ ] is called an _ abelian run _ with period @xmath1 if it is periodic with period @xmath1 and maximal with respect to this property ( i.e. , each of @xmath82 $ ] and @xmath83 $ ] either does not exist or it is not periodic with period @xmath1 ) .
we shall often represent an abelian run @xmath76 $ ] as a tuple @xmath84 where @xmath85 and @xmath86 are respectively the lengths of the head and the of the tail of a periodic factorization of @xmath76 $ ] with period @xmath1 and at least two cores ( see [ figu - def ] ) . note that @xmath87 is the anchor of the factorization , and that @xmath88 is the number of cores , in particular it is an integer .
observe that an abelian run with period @xmath1 may have several valid factorizations .
for example , @xmath89 and @xmath90 are factorization of a run @xmath91 $ ] with period @xmath92 in @xmath93 .
therefore the run can be represented as @xmath94 and as @xmath95 .
however , in @xmath96 only @xmath97 is a representation of @xmath98 $ ] as an abelian run with period @xmath92 .
this is because @xmath99 corresponds to a factorization @xmath98=a\cdot ba \cdot b$ ] with one core only , and such a factorization does not indicate that @xmath98 $ ] is an abelian run .
( 0,0 ) rectangle ( 2,.5 ) ; ( 2,0 ) rectangle ( 4,.5 ) ; ( 4,0 ) rectangle ( 7,.5 ) ; ( 7,0 ) rectangle ( 10,.5 ) ; ( 10,0 ) rectangle ( 13,.5 ) ; ( 13,0 ) rectangle ( 14,.5 ) ; ( 14,0 ) rectangle ( 15,.5 ) ; at ( 2.25,.75 ) @xmath39 ; at ( 13.75,.75 ) @xmath100 ; ( 2,-.5 ) ( 4,-.5 ) ; at ( 3,-1 ) @xmath85 ; ( 13,-.5 ) ( 14,-.5 ) ; at ( 13.5,-1 ) @xmath86 ; ( 4,-.5 ) ( 7,-.5 ) ; at ( 5.5,-1 ) @xmath74 ; ( 7,-.5 ) ( 10,-.5 ) ; at ( 8.5,-1 ) @xmath74 ; ( 10,-.5 ) ( 13,-.5 ) ; at ( 11.5,-1 ) @xmath74 ; ( 4,1 ) ( 13,1 ) ; at ( 8.5,1.5 ) @xmath101 ; matsuda et al .
@xcite gave a different definition of abelian runs , where maximality is with respect to extending a fixed factorization .
in this paper , we call such fragments anchored ( abelian ) runs .
[ def - anchored ] a fragment @xmath76 $ ] is a _
@xmath80-anchored abelian run _ with period @xmath1 if @xmath76 $ ] is periodic with period @xmath1 anchored at @xmath80 and maximal with respect to this property ( i.e. , each of @xmath82 $ ] and @xmath83 $ ] either does not exist or it is not periodic with period @xmath1 anchored at @xmath80 ) . note that every abelian run is an anchored run with the same period ( for some anchor ) .
the converse is not true , since it might be possible to extend an anchored run preserving the period but not the anchor .
for example , in the word @xmath29 considered in the introduction , the fragment @xmath102={\varepsilon}\cdot ab \cdot ab \cdot a$ ] is a @xmath103-anchored run but not an abelian run , since @xmath104=a \cdot ba \cdot ba \cdot a$ ] is periodic with abelian period @xmath31 .
since a factorization is uniquely determined by the anchor , standard inclusion - maximality is equivalent to the condition in definition [ def - anchored ] .
let @xmath76 $ ] and @xmath105 $ ] be fragments of @xmath0 with abelian period @xmath1 anchored at @xmath80 . if @xmath76 $ ] is properly contained in @xmath105 $ ] ( i.e , @xmath106 and @xmath107 , or @xmath108 and @xmath109 ) , then @xmath76 $ ] is not a @xmath80-anchored abelian run with period @xmath1 .
abelian runs enjoy the same property , but its proof is no longer trivial .
[ lemma - maximality ]
let @xmath76 $ ] and @xmath105 $ ] be fragments of @xmath0 with abelian period @xmath1 .
if @xmath76 $ ] is properly contained in @xmath105 $ ] , then @xmath76 $ ] is not an abelian run with period @xmath1 .
we assume that @xmath110 .
the case of @xmath109 is symmetric . for a proof by contradiction
suppose that @xmath76 $ ] is an abelian run and let @xmath76=u_0\cdots u_k$ ] be a periodic factorization with period @xmath1 and at least two cores ( i.e. , satisfying @xmath111 )
. a periodic factorization of @xmath105 $ ] can be trimmed to a factorization @xmath82={v}_0\cdots { v}_{\ell}$ ] . however , since @xmath76 $ ] is an abelian run , this factorization must have at most one core ( i.e. , @xmath112 ) .
moreover , @xmath113 can not be extended to a factorization of @xmath82=u'_0u_1\cdots u_k$ ] . in other words
@xmath114 , the extension of @xmath20 by one letter to the left , must satisfy @xmath115 .
let @xmath77 .
the conditions on the number of cores imply @xmath116|$ ] and @xmath117\pp j]| < |v_0|+2p$ ] . consequently , @xmath118 , i.e. , @xmath114 is a proper prefix of @xmath119 . this yields @xmath120 , which is in contradiction with @xmath115 .
[ coro - run ] let @xmath0 be a word . for a fixed parikh vector @xmath1 , there is at most one abelian run with abelian period @xmath1 starting at each position of @xmath0 .
matsuda et al .
@xcite presented an algorithm that computes all the anchored runs of a word @xmath0 of length @xmath2 in @xmath8 time and space complexity .
the initial step of the algorithm is to compute maximal abelian powers in @xmath0 . recall that an abelian power is a concatenation of several abelian - equivalent words .
in other words , an abelian power of period @xmath1 is a word admitting a periodic factorization with respect to @xmath1 with an empty head , an empty tail and at least two cores .
a fragment @xmath76 $ ] is a maximal abelian power if it can not be extended to a longer power of period @xmath1 ( preserving the anchor ) .
formally , the maximality conditions are 1 .
@xmath121 } \ne \pv_{\s{w}[i\pp i+p-1]}$ ] or @xmath122 , and 2 .
@xmath123 } \ne \pv_{\s{w}[j+1\pp j+p]}$ ] or @xmath124 , where @xmath77 .
the approach of @xcite is to first compute all the abelian squares using the algorithm by cummings & smyth @xcite .
the next step is to group squares into maximal abelian powers . for this
, it suffices to merge pairs of overlapping abelian squares of the form @xmath125 $ ] and @xmath126 $ ] . this way maximal abelian powers
are computed in @xmath8 time .
observe that there is a natural one - to - one correspondence between maximal abelian powers and anchored runs : it suffices to trim the head and the tail of the factorization of an anchored run to obtain a maximal abelian power .
hence , the last step of the algorithm is to compute the maximal head and tail by which each abelian power can be extended .
this could be done naively in @xmath127 time overall , but a clever computation enables to find all the abelian runs in time and space @xmath8 ( see @xcite for further details ) .
in section [ sect - all ] , we extend this result to compute the abelian runs only rather than all the anchored runs .
both these algorithms work offline : they need to know the whole word before reporting any abelian run . in the following two sections
we give several online algorithms , which are able to report a run ending at position @xmath128 of a word @xmath0 before reading @xmath129 $ ] and the following letters .
clearly , not knowing @xmath50 $ ] one can not decide whether the run could be extended to the right , so this is the optimal delay .
however , these methods are restricted to finding runs of a given period or a given norm of the periods , respectively .
in this section we present our online solution for computing all the abelian runs of a given parikh vector @xmath1 of norm @xmath6 in a given word @xmath0 .
the algorithm works in @xmath4 time and @xmath5 space where @xmath130 .
first , in sect . [ sub - anchored ] , we show how to compute all anchored runs of period @xmath1 .
later , in sect . [ sub - abelian ] , we modify the algorithm to return abelian runs only .
we conclude in sect .
[ sub - example ] with an example course of actions in our solution .
we begin with a description of data maintained while scanning the string @xmath0 . for an integer @xmath80 ,
let @xmath131 $ ] be the starting position of the longest suffix of @xmath132 $ ] which has period @xmath1 anchored at @xmath80 .
if there is no such a suffix , we set @xmath131 = \infty$ ] . since this notion depends on @xmath133 only , we store @xmath134 $ ] for @xmath135 only .
let @xmath136 be the starting position of the longest suffix of @xmath132 $ ] whose parikh vector is contained in or equal to @xmath1 . in other words
, we have @xmath137}\subseteq \pv$ ] and @xmath138}\not\subseteq \pv$ ] ( or @xmath139 ) .
note that @xmath140 if @xmath50 $ ] does not occur in @xmath1 .
observe that the tail of any periodic factorization of a suffix of @xmath132 $ ] must be contained in @xmath141 $ ] .
this leads to the following characterization : [ lemma - char - simple ] let @xmath142 .
we have @xmath134\le k$ ] for @xmath143 and @xmath134=\infty$ ] for @xmath144 . for @xmath143 , the fragment @xmath145 $ ] has abelian period @xmath1 anchored at @xmath80 .
( the underlying factorization has empty head , no cores and tail @xmath145 $ ] , unless @xmath146 , when the factorization has one core , empty head and empty tail ) .
hence , we have @xmath134\le k$ ] directly from the definition . for @xmath144 ,
the tail of the factorization with anchor @xmath133 would need to start at position @xmath80 , which is impossible ( see [ figu - case1 ] ) .
( .5,0 ) rectangle ( 7.5,.5 ) ; ( 7.5,0 ) rectangle ( 9,.5 ) ; ( 9,0 ) rectangle ( 12.5,.5 ) ; at ( .25,.25 ) @xmath0 ; at ( 7,.75 ) @xmath147 ; at ( 9.25,.75 ) @xmath136 ; at ( 12.25,.75 ) @xmath39 ; ( 8.5,-.5 ) rectangle ( 12.5,-1 ) ; at ( 13,-.75 ) @xmath148 ; at ( 10.5,-.75 ) impossible tail ; the values @xmath149 and @xmath136 are actually sufficient to describe @xmath150 based on @xmath151 .
[ lemma - algo ] for @xmath142 the following equalities hold : 1 .
[ case1 ] @xmath134=\infty \ne b_{i-1}[k]$ ] for @xmath152 , 2 . [ case2 ] @xmath134=b_{i-1}[k]$ ] for @xmath153 and for @xmath154 , 3 . [ case3 ] @xmath155 = b_i$ ] if @xmath156 and @xmath155=b_{i-1}[i - p+1]$ ] otherwise .
lemma [ lemma - char - simple ] implies that @xmath131=\infty$ ] for @xmath144 and @xmath157=\infty$ ] for @xmath158 ( hence @xmath157=b_{i}[k]$ ] in this latter case ) . for @xmath159
, we have @xmath160}\subseteq \pv$ ] , so we can extend the factorization of a suffix of @xmath161 $ ] whose tail starts at position @xmath80 ( see [ figu - case2 ] ) .
finally , note that @xmath155 $ ] is the starting position of the maximal suffix of @xmath132 $ ] with an empty - tail periodic factorization .
if @xmath162}\ne \pv$ ] ( i.e. , if @xmath163 ) , this is just @xmath141 $ ] .
otherwise , we can extend the factorization of a suffix of @xmath161 $ ] whose tail starts at position @xmath164 .
( .5,0 ) rectangle ( 8,.5 ) ; ( 8,0 ) rectangle ( 9.5,.5 ) ; ( 9.5,0 ) rectangle ( 12,.5 ) ; ( 12,0 ) rectangle ( 12.5,.5 ) ; at ( .25,.25 ) @xmath0 ; at ( 3.75,.75 ) @xmath157 $ ] ; at ( 8.25,.75 ) @xmath149 ; at ( 10.25,.75 ) @xmath80 ; at ( 9.75,.75 ) @xmath136 ; at ( 12.25,.75 ) @xmath39 ; ( 10,-1.5 ) rectangle ( 12.5,-2 ) ; at ( 13,-1.75 ) @xmath165 ; ( 3.5,-2.5 ) rectangle ( 5,-3 ) ; ( 6,-2.5 ) rectangle ( 10,-3 ) ; ( 10,-2.5 ) rectangle ( 12,-3 ) ; at ( 5.5,-2.75 ) @xmath166 ; ( 3.5,-3.5 ) rectangle ( 5,-4 ) ; ( 6,-3.5 ) rectangle ( 10,-4 ) ; ( 10,-3.5 ) rectangle ( 12.5,-4 ) ; at ( 5.5,-3.75 ) @xmath166 ; at ( 1.75,-2.55 ) fragment ; at ( 1.75,-2.95 ) ending at @xmath128 ; at ( 1.75,-3.55 ) fragment ; at ( 1.75,-3.95 ) ending at @xmath39 ; having read letter @xmath50 $ ] , we need to report anchored runs which end at position @xmath128 . for this , we use the following characterization . [ lem - report - anchored ] let @xmath167 .
a fragment @xmath168 $ ] is a @xmath80-anchored run with period @xmath1 if and only if @xmath157 = b\le k-2p$ ] and @xmath134>b$ ] .
clearly an anchored run ending at position @xmath128 must be a left - maximal suffix of @xmath161 $ ] with a given anchor .
moreover , we must have @xmath169 so that the factorization has at least two cores and @xmath134>b$ ] due to right - maximality .
it is easy to see that these conditions are sufficient . by lemma [ lemma - algo ] ,
most entries of @xmath150 are inherited from @xmath151 , so we use a single array @xmath170 and having read @xmath50 $ ] , we update its entries . as evident from lemma
[ lem - report - anchored ] , each anchored run to be reported corresponds to a modified entry .
the algorithm in [ algo - run0 ] implements our approach . the * while * loop increments @xmath80 from @xmath149 to @xmath136 . for @xmath171 ,
we set @xmath172 $ ] to @xmath173 and possibly report a run . note
that @xmath174 is within the scope of case [ case3 ] rather than case [ case1 ] in lemma [ lemma - algo ] .
however , later we set @xmath175 $ ] to @xmath176 if @xmath177 ( as described in case [ case3 ] ) .
nevertheless , if an @xmath46-anchored run needs to be reported , we have @xmath178<\infty = b_{i}[i - p+1]$ ] , so @xmath179 and thus @xmath174 is considered in the loop .
anchoredrun , p , w , n [ loop - for - begin ] [ loop - while - begin ] [ loop - while - end ] [ loop - for - end ] the algorithm computes all the anchored runs with period @xmath1 of norm @xmath6 in a word @xmath0 of length @xmath2 in time @xmath4 and additional space @xmath5 .
the correctness of the algorithm comes from lemmas [ lemma - algo]-[lem - report - anchored ] and the discussion above .
the external * for * loop in lines [ loop - for - begin][loop - for - end ] runs @xmath180 times . the internal * while * loop in lines [ loop - while - begin][loop - while - end ] can not iterate more than @xmath180 times since it starts with @xmath80 equal to @xmath103 and ends when @xmath80 is equal to @xmath2 and @xmath80 can only be incremented by @xmath181 ( in line [ loop - while - end ] ) .
the test @xmath182 } \not\subseteq \pv$ ] in line [ loop - while - begin ] can be realized in constant time once we store @xmath182}$ ] and a counter of its components for which the value is greater than in @xmath1 .
this data needs to be updated once we increment @xmath39 in the * for * loop and @xmath80 in line [ loop - while - end ] .
we then need to increment the component @xmath50 $ ] or decrement the component @xmath183 $ ] of @xmath182}$ ] , respectively .
the global counter needs to be updated accordingly .
all the other operations run in constant time .
thus the total time complexity of the algorithm is @xmath4 .
the space complexity comes from the number of counters ( @xmath3 ) and the size of the array @xmath170 ( @xmath6 ) .
note that the space consumption can be reduced to @xmath184 at the price of introducing ( monte carlo ) randomization . instead of storing the parikh vectors in a plain form
, we can use dynamic hash tables @xcite so that the size is proportional to the number of non - zero entries .
run , p , w , n in this section we extend our algorithm so that it reports abelian runs only . for an offline solution
, we could simply determine the anchored runs ( using the procedure developed above ) and filter out those which are not maximal .
however , in order to obtain an online algorithm , we need a more subtle approach , which is based on the following characterization .
[ lem - report - abelian ] a fragment @xmath168 $ ] is an abelian run with period @xmath1 if and only if it is an anchored run ( with period @xmath1 ) and for each @xmath185 the inequalities @xmath186\ge b$ ] and @xmath187>b$ ] hold . by lemma [ lemma - maximality ] , an abelian run of period @xmath1 can not be properly contained in a fragment with period @xmath1 ( anchored at some @xmath185 ) .
conditions involving @xmath186 $ ] and @xmath188 $ ] enforce left - maximality and right - maximality , respectively . since each abelian run is an anchored run ( with the same period ) and
since all anchored runs are periodic , the claim follows . to apply lemma [ lem - report - abelian ] , it suffices to find an anchor @xmath80 such that @xmath189=\min_{k'}b_{i-1}[k']<\min_{k'}b_{i}[k']$ ] .
there can be several such anchors and in case of ties we are going to detect the one for which the factorization of @xmath168 $ ] has shortest tail .
this factorization maximizes the number of cores , so if @xmath168 $ ] is an anchored run with any anchor , it is with that one in particular . note that the * while * loop in lines [ loop - while - begin][loop - while - end ] of algorithm anchoredrun processes anchors @xmath190 in the order of decreasing tail lengths and updates the underlying values @xmath172 $ ] from @xmath157 $ ] to @xmath131 $ ] . for the sought anchor @xmath80
this update strictly increases the value @xmath191 $ ] , and moreover this is the first increase of the minimum within a given iteration of the outer * for * loop .
hence , we record the original minimum and check for an abelian run only if @xmath191 $ ] increases from that value . to implement the procedure described above
, we need to efficiently compute the smallest element in the array @xmath170 . for this , recall that @xmath192 $ ]
can only be modified from @xmath173 to @xmath80 ( and the value @xmath80 does not decrease throughout the algorithm ) or from some value back to @xmath173 .
we maintain a doubly - linked list @xmath193 of all indices @xmath100 with finite @xmath192 $ ] such that the order of indices @xmath100 in the list
is consistent with the order of values @xmath192 $ ] . to update the list
, it suffices to insert the index @xmath100 to the end of list while setting @xmath192 $ ] to @xmath80 , and remove it from the list setting @xmath192 $ ] to @xmath173 .
then the smallest value in @xmath170 is attained at an argument stored as the first element of the list @xmath193 ( or @xmath173 , if the list is empty ) .
the algorithm , depicted in [ algo - run ] , implements the approach described above .
it uses the following constant - time functions to operate on lists : * that inserts @xmath194 at the end of the doubly - linked list @xmath193 and returns a pointer to the location of @xmath194 in the list ; * that deletes the element pointed by @xmath195 from the doubly - linked list @xmath193 ; * that returns the first element of the list @xmath193 ( 0 if the list is empty ) .
the algorithm also uses an array @xmath196 which maps any anchor @xmath100 to a pointer of the corresponding location in the list @xmath193 ( or @xmath197 if @xmath192=\infty$ ] ) .
the discussion above proves that correctly computes abelian runs with period @xmath1 in @xmath0 .
its running time is the same as that of since the structure of the computations remains the same while additional instructions run in constant time .
memory consumption is still @xmath198 because both @xmath193 and @xmath196 take @xmath184 space . the algorithm computes all the abelian runs with period @xmath1 of norm @xmath6 in a word @xmath0 of length @xmath2 in time @xmath4 and additional space @xmath5 , which can be reduced to @xmath184 using randomization .
let us see the behaviour of the algorithm on @xmath199 , @xmath200 and @xmath201 : + @xmath202 \quad l= ( 0)$ ] @xmath203}\subseteq \pv
\quad b = [ 0 , 0 , \infty , \infty ] \quad l= ( 0,1)$ ] @xmath204}\subseteq \pv \quad b =
[ 0 , 0 , 0 , \infty ] \quad l= ( 0,1,2)$ ] @xmath205}\subseteq \pv \quad b = [ 0 , 0 , 0 , 0 ] \quad l= ( 0,1,2,3)$ ] @xmath206}\not\subseteq \pv \quad b=0 \quad b=[\infty , 0,0,0 ] \quad l= ( 1,2,3 ) \quad k=1 $ ] @xmath207}\subseteq \pv \quad b = [ 1 , 0 , 0 , 0 ] \quad l= ( 1,2,3,0)$ ] @xmath208}\subseteq \pv$ ] @xmath209}\not\subseteq \pv \quad k=2 $ ] @xmath210}\not\subseteq \pv \quad b=0 \quad b = [ 1 , 0 , \infty , 0 ] \quad l= ( 1,3,0 ) \quad k=3 $ ] @xmath211}\subseteq \pv \quad b = [ 1 , 0 , 3 , 0 ] \quad l= ( 1,3,0,2)$ ] @xmath212}\subseteq \pv$ ] @xmath213}\not\subseteq \pv \quad k=4 $ ] @xmath214}\subseteq \pv$ ] @xmath215}\not\subseteq \pv \quad k=5 $ ] @xmath216}\not\subseteq \pv \quad b=0 \quad b = [ 1 , \infty , 3 , 0 ] \quad l= ( 3,0,2)\quad k=6 $ ] @xmath217}\subseteq \pv \quad b = [ 1 , 6 , 3 , 0 ] \quad l= ( 3,0,2,1)$ ] @xmath218}\subseteq \pv$ ] @xmath219}\not\subseteq \pv \quad k=7 $ ] @xmath220}\subseteq \pv$ ] @xmath221}\not\subseteq \pv \quad k=8 $ ] @xmath222}\not\subseteq \pv \quad b=1 \quad b=[\infty,6,3,0 ] \quad l=(3,2,1 ) \quad k=9 $ ] @xmath223}\not\subseteq \pv \quad b=6 \quad b=[\infty,\infty,3,0 ] \quad l=(3,2)\quad k=10 $ ] @xmath224}\subseteq \pv \quad b=[10,\infty,3,0 ] \quad l=(3,2,0)$ ] @xmath225 @xmath226 \quad l= ( 3,0 ) \quad k = 11 $ ] @xmath227 \quad l= ( 0)$ ] @xmath228 @xmath229 \quad l= ( ) \quad k=13 $ ]
in this section we develop an @xmath7-time algorithm to compute all abelian runs with periods of norm @xmath6 .
first , we describe the algorithm for anchored runs and later generalize it to abelian runs .
let us start with a simple offline algorithm which works in @xmath4 time to compute @xmath80-anchored runs with period of norm @xmath6 for fixed values @xmath6 and @xmath80 .
this method is similar to the algorithm of matsuda et al .
@xcite briefly described in section [ sect - prev ] .
namely , it suffices to compute maximal abelian powers with periods of norm @xmath6 anchored at @xmath80 , and then extend them by a head and a tail .
block _ as any fragment of the form @xmath230 $ ] such that @xmath231 .
note that the cores in decompositions with anchor @xmath133 are blocks .
finding @xmath80-anchored powers with periods of a given norm @xmath6 is very easy if the anchor is fixed .
we consider consecutive blocks , naively check if they are abelian - equivalent and merge any maximal chains of abelian - equivalent blocks . determining the head and
the tail of the @xmath80-anchored runs is also simple . for each @xmath231
we compute the longest suffix of @xmath161 $ ] and the longest prefix of @xmath232 $ ] whose parikh vectors are contained in @xmath233}$ ] .
this approach can be implemented online in @xmath5 space as follows : we scan consecutive blocks and ( naively ) check their abelian equivalence .
whenever we read a full block ( say , starting at position @xmath39 ) , we compute the longest suffix @xmath234 $ ] of @xmath161 $ ] whose parikh vector is contained in @xmath233}$ ] .
this gives a periodic factorization of @xmath235 $ ] anchored at @xmath81 .
we then try to extend it to the right while reading further characters .
once it is impossible to extend the factorization , say by letter @xmath236 $ ] , we declare @xmath237 $ ] as a maximal fragment with period @xmath233}$ ] anchored at @xmath80 .
if the decomposition has at least two cores , we report an anchored run .
if we succeed to extend by a full block ( i.e. , if @xmath233}=\pv_{w[i+p\pp i+2p-1]}$ ] ) , we do not restart the algorithm but instead we continue to extend the factorization .
this way , we guarantee that @xmath238 whenever we start building a new factorization .
clearly , the procedure described above computes all @xmath80-anchored runs with period of norm @xmath6 . to compute all anchored runs , we simply run it in parallel for all @xmath6 possible anchors
. there is an algorithm which computes online all the anchored runs with periods of norm @xmath6 in a word @xmath0 of length @xmath2 over an alphabet of size @xmath3 in time @xmath33 and additional space @xmath239 , which can be reduced to @xmath240 using randomization .
let us first slightly modify the algorithm presented in the previous section .
observe that whenever we start a new phase having just read a block @xmath230 $ ] , instead of performing the computations using a simple procedure described above , we could launch the algorithm of section [ sub - anchored ] for @xmath241 $ ] and @xmath242}$ ] , simulate it until it needs to read @xmath243 $ ] and then feed it with newly read letters until the maximal extension of @xmath230 $ ] anchored at @xmath80 is found ( i.e. , until the respective entry of the @xmath170 array is set to @xmath173 ) .
other anchored runs output by the algorithm should be ignored , of course . as before ,
if such a process is running while we have completed reading a subsequent block , we do not start a new phase .
it is easy to see that such an algorithm is equivalent to the previous one .
however , if we use the algorithm of section [ sub - abelian ] instead , we automatically get a possibility to check whether the maximal extension of @xmath230 $ ] anchored at @xmath80 is a maximal fragment with period @xmath233}$ ] .
note that we start the simulation at a position @xmath244 which is smaller than @xmath149 , unless the latter is @xmath103 .
this guarantees that left - maximality is correctly verified despite the fact the fragment prior to position @xmath147 is ignored in the simulation . as before , we disregard any other abelian run that the algorithm of section [ sub - abelian ] may return .
we run this process in parallel for all possible anchors to guarantee that each abelian run with period of norm @xmath6 is reported exactly once .
more precisely , in ambiguous cases a run is reported for the anchor corresponding to the factorization with shortest tail , just as in section [ sub - abelian ] .
there is an algorithm which computes online all the abelian runs with periods of norm @xmath6 in a word @xmath0 of length @xmath2 over an alphabet of size @xmath3 in time @xmath33 and additional space @xmath239 , which can be reduced to @xmath240 using randomization .
in this section we present an @xmath8-time offline algorithm which computes all the abelian runs . as a starting point
, we use the set of all anchored runs computed by the algorithm by matsuda et al .
( see section [ sect - prev ] ) . recall that all abelian runs are anchored runs with the same period .
hence , it suffices to filter out those anchored runs which are properly contained in another anchored run with the same period .
we also need to make sure that every abelian run is reported once only ( despite possibly being @xmath80-anchored for different anchors @xmath80 ) .
note that this filtering can be performed independently for distinct periods .
if we have a list of anchored runs with a fixed period , sorted by the starting position , it is easy to retrieve the abelian runs of that period with a single scan of the list .
ordering by the starting position can be performed together for all periods so that it takes @xmath245 time where @xmath246 is the number of all anchored runs .
hence , the main difficulty is grouping according to the period .
for this , we shall assign to each fragment of @xmath0 an _ identifier _ , so that two fragments are abelian - equivalent if and only if their identifiers are equal .
the identifiers of periods can be easily retrieved since given a @xmath80-anchored run , we can easily locate one of the cores of the underlying factorization .
thus , in the remaining part of this section we design a naming algorithm which assigns the identifiers . a naive solution would be to generate the parikh vectors of all substrings of @xmath0 , sort these vectors removing duplicates , and give each fragment a rank of its parikh vector in that order . however , already storing the parikh vectors can take prohibitive @xmath247 space . to overcome this issue
, we use the concept of _ diff - representation _ , originally introduced in the context of abelian periods @xcite .
observe that in a sense the parikh vectors of fragments can be generated efficiently : for a fixed @xmath6 , we can first generate @xmath248}$ ] , then update it to @xmath249}$ ] , and so on until we reach @xmath250}$ ] .
in other words , the parikh vectors of all fragments of length @xmath6 can be represented in a sequence so that the total hamming distance of the adjacent vectors is @xmath4 . the diff - representation , designed to manipulate sequences satisfying such a property , is formally defined as a sequence of single - entry changes such that the original sequence of vectors is a subsequence of intermediate results when applying this operations starting from the null vector ( of the fixed dimension @xmath246 ) .
note that the diff - representation of a sequence of parikh vectors of all fragments of @xmath0 can be computed in time @xmath8 proportional to its size .
the following result lets us efficiently assign identifiers to its elements .
given a sequence of vectors of dimension @xmath246 represented using a diff - representation of size @xmath251 , consider the problem of assigning integer identifiers of size @xmath252 so that equality of vectors is equivalent to equality of their identifiers .
it can be solved in @xmath253 time using a deterministic algorithm and in @xmath254 time using a monte carlo algorithm which is correct with high probability ( @xmath255 where @xmath256 can be chosen arbitrarily large ) . in our setting this yields the following result .
there exists an @xmath8-time randomized algorithm ( monte carlo , correct with high - probability ) which computes all abelian runs in a given word of length @xmath2 .
additionally , there exists an @xmath9-time deterministic algorithm solving the same problem .
we gave algorithms that , given a word @xmath0 of length @xmath2 over an alphabet of cardinality @xmath3 , return all the abelian runs of a given period @xmath1 in @xmath0 in time @xmath4 and space @xmath5 , or all the abelian runs with periods of a given norm @xmath6 in time @xmath7 and space @xmath239 .
these algorithms work in an online manner .
we also presented an @xmath8 ( resp .
@xmath34)-time offline randomized ( resp .
deterministic ) algorithm for computing all the abelian runs in a word of length @xmath2
. one may wonder if it is possible to reduce further the complexities of these latter algorithms .
we believe that further combinatorial results on the structure of the abelian runs in a word could lead to novel solutions .
r. kolpakov , g. kucherov , finding maximal repetitions in a word in linear time , in : 40th annual symposium on foundations of computer science , focs 1999 , ieee computer society , new - york , 1999 , pp .
http://dx.doi.org/10.1109/sffcs.1999.814634 [ ] .
h. bannai , t. i , s. inenaga , y. nakashima , m. takeda , k. tsuruta , a new characterization of maximal repetitions by lyndon trees , in : p. indyk ( ed . ) , proceedings of the twenty - sixth annual acm - siam symposium on discrete algorithms , soda 2015 , san diego , ca , usa , january 4 - 6 , 2015 , siam , 2015 , pp . 562571 . http://dx.doi.org/10.1137/1.9781611973730.38 [ ] .
s. matsuda , s. inenaga , h. bannai , m. takeda , computing abelian covers and abelian runs , in : j. holub , j. zdrek ( eds . ) , prague stringology conference , psc 2014 , czech technical university in prague , 2014 , pp . 4351 .
g. fici , t. lecroq , a. lefebvre , .
prieur - gaston , online computation of abelian runs , in : a. h. dediu , e. formenti , c. martn - vide , b. truthe ( eds . ) , language and automata theory and applications , lata 2015 , vol .
8977 of lncs , springer international publishing , 2015 , pp .
http://dx.doi.org/10.1007/978-3-319-15579-1_30 [ ] .
m. dietzfelbinger , a. r. karlin , k. mehlhorn , f. meyer auf der heide , h. rohnert , r. e. tarjan , dynamic perfect hashing : upper and lower bounds , siam journal on computing 23 ( 4 ) ( 1994 ) 738761 .
http://dx.doi.org/10.1137/s0097539791194094 [ ] .
t. kociumaka , j. radoszewski , w. rytter , fast algorithms for abelian periods in words and greatest common divisor queries , in : n. portier , t. wilke ( eds . ) , 30th international symposium on theoretical aspects of computer science , stacs 2013 , vol .
20 of lipics , schloss dagstuhl - leibniz - zentrum fuer informatik , 2013 , pp .
http://dx.doi.org/10.4230/lipics.stacs.2013.245 [ ] . | given a word @xmath0 and a parikh vector @xmath1 , an abelian run of period @xmath1 in @xmath0 is a maximal occurrence of a substring of @xmath0 having abelian period @xmath1 .
our main result is an online algorithm that , given a word @xmath0 of length @xmath2 over an alphabet of cardinality @xmath3 and a parikh vector @xmath1 , returns all the abelian runs of period @xmath1 in @xmath0 in time @xmath4 and space @xmath5 , where @xmath6 is the norm of @xmath1 , i.e. , the sum of its components .
we also present an online algorithm that computes all the abelian runs with periods of norm @xmath6 in @xmath0 in time @xmath7 , for any given norm @xmath6 .
finally , we give an @xmath8-time offline randomized algorithm for computing all the abelian runs of @xmath0 .
its deterministic counterpart runs in @xmath9 time .
combinatorics on words , text algorithms , abelian period , abelian run |
for @xmath0 , @xmath1 , a continuous map @xmath2 \to { \mbox{$\mathbb{r}$}}$ ] is said to be piecewise monotone if there are @xmath3 and points @xmath4 such that @xmath5 is monotone on each @xmath6 $ ] , @xmath7 . a piecewise monotone map @xmath5 has _ constant slope _
@xmath8 if @xmath9 for all @xmath10 .
the following results are well known for piecewise monotone interval maps : @xcite if @xmath11 \to [ 0,1]$ ] is piecewise monotone and @xmath12 then @xmath5 is semiconjugate via a continuous non - decreasing onto map @xmath13\to [ 0,1]$ ] to a map @xmath14 of constant slope @xmath15 .
the map @xmath16 is a conjugacy ( @xmath16 is strictly increasing ) if @xmath5 is transitive .
@xcite[t:8 ] if @xmath5 has a constant slope @xmath8 then @xmath17 . for continuous interval maps with a countably infinite number of pieces of monotonicity
neither theorem is true - for examples , see @xcite and @xcite .
one of the few facts that remains true in the countably piecewise monotone setting is : @xcite[p:5 ] if @xmath5 is @xmath8-lipschitz then @xmath18 .
a continuous interval map @xmath5 has _ constant slope _
@xmath8 if @xmath9 for all but countably many points
. the question we want to address is when a _ continuous countably piecewise monotone _
interval map @xmath5 is conjugate to a map of constant slope @xmath19 .
particular attention will be given to the case when a slope is given by the topological entropy of @xmath5 , which we call _ linearizability _ : [ d:5 ] a continuous map @xmath11
\to [ 0,1]$ ] is said to be _
linearizable _ if it is conjugate to an _ interval map _ of constant slope @xmath20 .
we will confine ourselves to the markov case , and explore what can be said if only the transition matrix of a countably piecewise monotone map is known in terms of the vere - jones classification @xcite , refined by @xcite .
the structure of our paper is as follows . in section [ s:2 ] ( _ @xmath21 : the class of countably piecewise monotone markov maps _ ) we make precise the conditions on continuous interval maps under which we conduct our investigation - the set of all such maps will be denoted by @xmath21 ( for _ countably piecewise monotone markov _ ) . in particular , we introduce a _ slack _ countable markov partition of a map and distinguish between _ operator _ , resp . _ non - operator type_. in section [ s:3 ] ( _ conjugacy of a map from @xmath21 to a map of constant slope _ )
we rephrase the key equivalence from ( * ? ? ?
* theorem 2.5 ) : for the sake of completeness we formulate theorem [ t:5 ] , which relates the existence of a conjugacy to an `` eigenvalue equation '' , using both classical and slack countable markov partitions , see definition [ def : primary ] .
section [ s:4 ] ( _ the vere - jones classification _ ) is devoted to the vere - jones classification @xcite that we use as a crucial tool in the most of our proofs in later sections . in section [ s:5 ] ( _ entropy and the vere - jones classification in @xmath21 _
) we show in proposition [ p:7 ] that the topological entropy of a map in question and ( the logarithm of ) the perron value of its transition matrix coincide . using this fact ,
we are able to verify in proposition [ p:10 ] that all the transition matrices of a map corresponding to all the possible markov partitions of that map belong to the same class in the vere - jones classification ; so we can speak about the vere - jones classification of a map from @xmath21 . in section [ s:6 ]
( _ linearizability _ ) we present the main results of this text .
we start with proposition [ p:3 ] showing two basic properties of a @xmath19-solution of equation and theorem [ t:7 ] on leo maps , see definition [ def : leo ] .
afterwards we describe conditions under which a local window perturbation - theorems [ t:6 ] , [ t:9 ] , resp . a global window perturbation - theorems [ t:3 ] , [ t:10 ] results to a linearizable map .
in section [ s:7 ] ( _ examples _ ) various examples illustrating linearizability / conjugacy to a map of constant slope in the vere - jones classes are presented .
* acknowledgments : * we are grateful for the support of the austrian science fund ( fwf ) : project number p25975 , and also of the erwin schrdinger international institute for mathematical physics at the university of vienna , where the final part of this research was carried out . furthermore
, we would like to thanks the anonymous referees for the careful reading and valuable suggestions .
[ def : primary ] a countable markov partition @xmath22 for a continuous map @xmath23\to [ 0,1]$ ] consists of closed intervals with the following properties : * two elements of @xmath22 have pairwise disjoint interiors and @xmath24 \setminus \bigcup \p$ ] is at most countable . *
the partition @xmath22 is finite or countably infinite ; * @xmath25 is monotone for each @xmath26 ( _ classical markov partition _ ) or piecewise monotone for each @xmath26 ; in the latter case we will speak of _ a slack markov partition_. * for every @xmath27 and every maximal interval @xmath28 of monotonicity of @xmath5 , if @xmath29 , then @xmath30 .
the notion of a slack markov partition will be useful in later sections of this paper where we will work with _
window perturbations_. if @xmath31 , then the ordinal type of @xmath22 need not be @xmath32 or @xmath33 .
[ def : cpmm ] the class @xmath21 is the set of continuous interval maps @xmath11\to [ 0,1]$ ] satisfying * @xmath5 is _ topologically mixing _ , for every open sets
@xmath34 there is an @xmath35 such that @xmath36 for all @xmath37 .
* @xmath5 admits a countably infinite markov partition . *
@xmath38 .
since @xmath39 is topologically mixing by definition , it can not be constant on any subinterval of @xmath24 $ ] .
[ def : leo ] a map @xmath40 is called _ leo _ ( _ locally eventually onto _ ) if for every nonempty open set @xmath41 there is an @xmath42 such that @xmath43 $ ] .
[ r:4 ] let @xmath23\to [ 0,1]$ ] be a piecewise monotone markov map , such that orbits of turning points and endpoints @xmath44 are finite .
those orbits naturally determine a _
finite _ markov partition for @xmath5 .
this partition can be easily be refined , in infinitely ways , to _ countably infinite _ markov partitions .
if @xmath5 is topologically mixing and continuous then we will consider @xmath5 as an element of @xmath21 .
[ p:12 ] let @xmath39 with a markov partition @xmath22 .
for every pair @xmath45 satisfying @xmath46 there exist a maximal @xmath47 and intervals @xmath48 with pairwise disjoint interiors such that @xmath49 is monotone and @xmath50 for each @xmath51 . since @xmath39 is topologically mixing , it is not constant on any subinterval of @xmath24 $ ] . fix a pair @xmath45 with @xmath46 . since @xmath5 is continuous , there has to be at least one but at most a finite number of pairwise disjoint subintervals of @xmath52 satisfying the conclusion .
for a given @xmath39 with a markov partition @xmath22 , applying proposition [ p:12 ] we associate to @xmath22 the transition matrix @xmath53 defined by @xmath54 if @xmath22 is a classical markov partition of some @xmath39 then @xmath55 each @xmath56 .
[ r:5 ] for the sake of clarity we will write @xmath57 when a map @xmath39 , a concrete markov partition @xmath22 for @xmath5 and its transition matrix @xmath58 with respect to @xmath22 are assumed . for an infinite matrix @xmath59 indexed by a countable index set @xmath22 we can consider the powers @xmath60 of @xmath59 : @xmath61 [ p:11 ] let @xmath57 .
* for each @xmath62 and @xmath45 , the entry @xmath63 of @xmath64 is finite . *
the entry @xmath65 if and only if there are exactly @xmath66 subintervals @xmath67 , @xmath68 , @xmath69 of @xmath52 with pairwise disjoint interiors such that @xmath70 , @xmath71 .
\(i ) from the continuity of @xmath5 and the definition of @xmath59 follows that the sum @xmath72 is finite for each @xmath73 , which directly implies ( i ) .
+ ( ii ) for @xmath74 this is given by the relation defining the matrix @xmath59 .
the induction step follows from of the product of the nonnegative matrices @xmath59 and @xmath75 .
a matrix @xmath59 indexed by the elements of @xmath22 represents a bounded linear operator @xmath76 on the banach space @xmath77 of summable sequences indexed by @xmath22 , provided that the supremum of the columnar sums is finite .
then @xmath76 is realized through left multiplication @xmath78 the matrix @xmath64 represents the @xmath35th power @xmath79 of @xmath76 and by gelfand s formula , the spectral radius @xmath80 .
[ r:3 ] if @xmath57 the supremum in is finite if and only if @xmath81\colon~\#t^{-1}(y)\le k.\ ] ] since this condition does not depend on a concrete choice of @xmath22 , we will say the map @xmath5 is of an _ operator type _ when the condition is fulfilled and of a _ non - operator type _ otherwise .
this section is devoted to the fundamental observation regarding a possible conjugacy of an element of @xmath21 to a map of constant slope .
it is presented in theorem [ t:5 ] .
let @xmath57 .
we are interested in positive real numbers @xmath19 and nonzero nonnegative sequences @xmath82 satisfying @xmath83 , or equivalently @xmath84 [ d:3 ] a nonzero nonnegative sequence @xmath85 satisfying will be called a _ @xmath19-solution _ ( for @xmath59 ) .
if in addition @xmath86 , it will be called a _ summable @xmath19-solution _ ( for @xmath59 ) .
since every @xmath39 is topologically mixing , any nonzero nonnegative @xmath19-solution is in fact positive : if @xmath85 solves , @xmath87 and @xmath88 then by proposition [ p:11](ii ) for some sufficiently large @xmath35 , @xmath89 .
let @xmath90 denote the class of all maps from @xmath21 of constant slope @xmath19 , @xmath91 if @xmath92 for all but countably many points .
the core of the following theorem has been proved in ( * ? ? ?
* theorem 2.5 ) .
since we will work with maps from @xmath21 that are topologically mixing , we use topological conjugacies only - see ( * ? ? ?
* proposition 4.6.9 ) .
the theorem will enable us to change freely between classical / slack markov partitions of the map in question .
[ t:5]let @xmath39 .
the following conditions are equivalent . * for some @xmath93 ,
the map @xmath5 is conjugate via a continuous increasing onto map @xmath94\to [ 0,1]$ ] to some map @xmath91 .
* for some classical markov partition @xmath22 for @xmath5 there is a positive summable @xmath19-solution @xmath95 of equation .
* for every classical markov partition @xmath22 for @xmath5 there is a positive summable @xmath19-solution @xmath95 of equation .
* for every slack markov partition @xmath96 for @xmath5 there is a positive summable @xmath19-solution @xmath97 of equation .
* for some slack markov partition @xmath96 for @xmath5 there is a positive summable @xmath19-solution @xmath97 of equation .
recently , misiurewicz and roth @xcite have pointed out that if @xmath98 is a @xmath19-solution of equation that is not summable , then the map @xmath5 is conjugate to a map of constant slope defined on the real line or half - line .
[ r:2]let @xmath39 be piecewise monotone with a finite markov partition .
it is well known ( * ? ? ?
* theorem 0.16 ) , ( * ? ? ? * theorem 4.4.5 ) that the corresponding equation has a positive @xmath15-solution which is trivially summable . by theorem [ t:5 ]
it is also true for any countably infinite markov partition for @xmath5 .
the equivalence of ( i ) , ( ii ) and ( iii ) has been proved in @xcite . since ( iv ) implies ( iii ) and ( v ) , it suffices to show that ( iii ) implies ( iv ) and ( v ) implies ( ii ) .
( iii)@xmath99(iv ) .
let us assume that @xmath96 is a slack markov partition for @xmath5 .
obviously there is a classical partition @xmath22 for @xmath5 which is finer than @xmath96 , every element of @xmath22 is contained in some element of @xmath96 . using ( iii ) we can consider a positive summable @xmath19-solution @xmath95 of equation .
let @xmath97 be defined as @xmath100 clearly the positive sequence @xmath97 is from @xmath101 .
denoting @xmath102 and @xmath103 the transition matrices corresponding to the partitions @xmath22
, @xmath96 , we can write @xmath104 where the equality @xmath105 follows from the markov property of @xmath5 on @xmath22 and @xmath96 : @xmath106 so by , for a given slack markov partition @xmath96 ( for @xmath5 ) we find a positive summable @xmath19-solution @xmath97 of equation . ( v)@xmath99(ii )
assume that for some slack markov partition @xmath96 for @xmath5 there is a positive summable @xmath19-solution @xmath97 of equation .
as in the previous part we can consider a classical markov partition @xmath22 finer than @xmath96 .
using again property let us put @xmath107 then @xmath108 is positive and we will show that it is a summable @xmath19-solution of equation . fix an @xmath109 and using the property for @xmath110 for which @xmath111
. then @xmath112 hence summing through all @xmath113 s from @xmath96 that are @xmath5-covered by @xmath109 , we obtain with the help of , @xmath114 since by our assumption on @xmath97 and @xmath115 so @xmath108 is a summable @xmath19-solution of equation .
maps @xmath39 are continuous , topologically mixing with positive topological entropy .
thus all possible semiconjugacies described in ( * ? ? ?
* theorem 2.5 ) will be in fact conjugacies , see ( * ? ? ?
* proposition 4.6.9 ) .
many properties hold under the assumption of positive entropy or for _ countably piecewise continuous _ maps
. one interesting example of a countably piecewise continuous and countably piecewise monotone ( still topologically mixing ) map will be presented in section [ s:7 ] .
however , since the technical details are much more involved and would obscure the ideas , we confine the proofs to @xmath21 .
let us consider a matrix @xmath116 , where the index set @xmath22 is finite or countably infinite
. the matrix @xmath59 will be called * _ irreducible _ , if for each pair of indices @xmath117 there exists a positive integer @xmath35 such that @xmath118 , and * _ aperiodic _ , if for each index @xmath119 the value @xmath120 .
since @xmath39 is topologically mixing , its transition matrix @xmath59 is irreducible and aperiodic . in the sequel
we follow the approach suggested by vere - jones @xcite . [
p:2](i ) let @xmath116 be a nonnegative irreducible aperiodic matrix indexed by a countable index set @xmath22 .
there exists a common value @xmath121 such that for each @xmath117 @xmath122^{\frac{1}{n}}=\sup_{n\in{\tiny { \mbox{$\mathbb{n}$}}}}[m_{ii}(n)]^{\frac{1}{n}}=\lambda_m.\ ] ] ( ii ) for any value @xmath123 and all @xmath117 * the series @xmath124 are either all convergent or all divergent ; * as @xmath125 , either all or none of the sequences @xmath126 tend to zero .
[ r:6]the number @xmath121 defined by is often called the _ perron value of @xmath59_. in the whole text
we will assume that for a given nonnegative irreducible aperiodic matrix @xmath116 its perron value @xmath121 is finite .
to a given irreducible aperiodic matrix @xmath127 with entries from @xmath128 corresponds a strongly connected directed graph @xmath129 containing @xmath130 edges from @xmath52 to @xmath113 .
the _ gurevich entropy _ of @xmath59 ( or of @xmath131 ) is defined as @xmath132 where @xmath133 is the large eigenvalue of the finite transition matrix @xmath134 .
gurevich proved that @xcite[p:4 ] @xmath135 .
since by proposition [ p:2 ] the value @xmath136 is a common radius of convergence of the power series @xmath137 , we immediately obtain for each pair @xmath45 , @xmath138 it is well known that in @xmath139 * @xmath63 equals to the number of paths of length @xmath35 connecting @xmath52 to @xmath113 . following @xcite , for each @xmath62 we will consider the following coefficients : * first entrance to @xmath113 : @xmath140 equals the number of paths of length @xmath35 connecting @xmath52 to @xmath113 , without appearance of @xmath113 in between .
* last exit of @xmath52 : @xmath141 equals the number of paths of length @xmath35 connecting @xmath52 to @xmath113 , without appearance of @xmath52 in between . clearly @xmath142 for each @xmath119 . also it will be useful to introduce * first entrance to @xmath143 : for a nonempty @xmath144 and @xmath145 , @xmath146 equals the number of paths of length @xmath35 connecting @xmath52 to @xmath113 , without appearance of any element of @xmath147 in between
. the first entrance to @xmath143 will provide us a new type of a generating function used in and its applications .
[ r:1]let us denote by @xmath148 , @xmath149 the radius of convergence of the power series @xmath150 , @xmath151 .
since @xmath152 , @xmath153 for each @xmath62 and each @xmath45 , we always have @xmath154 , @xmath155 . proposition [ p:1 ] has been stated in @xcite . since the argument showing the part ( i ) presented in @xcite is not correct , we offer our own version of its proof .
[ p:1 ] ( * ? ? ?
* proposition 2.6 ) let @xmath57 , consider the graph @xmath131 , @xmath136 . * if there is a vertex @xmath113 such that @xmath156 then there exists a strongly connected subgraph @xmath157 such that @xmath158 .
* if there is a vertex @xmath113 such that @xmath159 then for all proper strongly connected subgraphs @xmath160 one has @xmath161 .
* if there is a vertex @xmath113 such that @xmath159 then @xmath162 for all @xmath52 .
for the proof of part ( ii ) see @xcite .
let us prove ( i ) . fix a vertex @xmath73 for which @xmath156 and choose arbitrary @xmath163 .
we can write @xmath164 where @xmath165 , resp .
@xmath166 denotes the number of @xmath167-paths of length @xmath35 that do not contain @xmath52 , resp .
contain @xmath52.*i . *
if @xmath168^{1/n}=\lambda_m$ ] , then there is nothing to prove . *
assume that @xmath168^{1/n}<\lambda_m$ ] . then by our assumption and @xmath169^{1/n}=\lambda_m.\ ] ] let us denote @xmath170 the number of paths of length @xmath35 connecting @xmath52 to @xmath113 , without appearance of @xmath117 after the initial @xmath52 and before the final @xmath113 .
if we denote @xmath171 the number of @xmath172-paths of length @xmath35 connecting @xmath52 to @xmath52 with exactly one appearance of @xmath113 after the initial @xmath52 and before the final @xmath52 , we can write for @xmath173 ( the coefficients @xmath174 are defined analogously as @xmath175 - compare the proof of theorem [ t:12 ] ) @xmath176 by the formula of ( * ? ? ?
* lemma 4.3.6 ) and our assumption , for arbitrary @xmath177 we obtain from either @xmath178^{1/n}=\lambda_m\ ] ] or @xmath179^{1/n}=\lambda_m.\ ] ] if is fulfilled for some @xmath52 the existence of a strongly connected subgraph @xmath157 such that @xmath158 immediately follows .
otherwise , since @xmath180^{1/n}\le \limsup_{n\to\infty}~[f_{ii}(n)]^{1/n},\ ] ] we get @xmath181 for each @xmath119 and the conclusion follows from ( * ? ? ?
* theorem 2.2 ) .
the assertion ( iii ) immediately follows from ( i ) and ( ii ) .
the behavior of the series @xmath182 , @xmath183 for @xmath184 was used in the vere - jones classification of irreducible aperiodic matrices @xcite .
vere - jones originally distinguished @xmath185-_transient _ , _ null @xmath185-recurrent _ and _ positive @xmath185-recurrent _ case . later on , the classification was refined by ruette in @xcite , who added _ strongly positive @xmath185-recurrent _ case .
all is summarized in table 1 which applies independently of the sites @xmath186 for @xmath59 irreducible - compare the last row of table 1 and proposition [ p:1 ] .
we call corresponding classes of matrices _ transient , null recurrent , weakly recurrent , strongly recurrent_. the last three , resp .
two possibilities will occasionally be summarized by @xmath59 is recurrent , resp .
@xmath59 is positive recurrent. _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ [ cols="^,^,^,^,^,^,^,^,^,^,^,^,^,^ " , ] * table 1 . *
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ there are geometrical criteria - see @xcite and also @xcite - for cases of the vere - jones classification to apply depending on whether the underlying strongly connected directed graph can be enlarged / reduced ( in the class of strongly connected directed graphs ) without changing the entropy .
we will use some of them in section [ s:7 ] .
[ t:1]@xcite the following are true : * a graph @xmath187 is transient if and only if there is a graph @xmath160 such that @xmath188 and @xmath189 .
* @xmath187 is strongly recurrent if and only if @xmath190 for any @xmath191 .
* @xmath187 is recurrent but not strongly recurrent if and only if there exists @xmath191 with @xmath192 , but @xmath193 for every @xmath194 . in the whole paper
we are interested in nonzero nonnegative solutions of equation .
analogously , in the next proposition we consider nonzero nonnegative _ subinvariant _
@xmath19-solutions @xmath85 for a matrix @xmath59 , satisfying the inequality @xmath195 .
[ t:2 ] ( * ? ? ?
* theorem 4.1 ) let @xmath116 be irreducible .
there is no subinvariant @xmath19-solution for @xmath196 .
if @xmath59 is transient there are infinitely many linearly independent subinvariant @xmath121-solutions . if @xmath59 is recurrent there is a unique subinvariant @xmath121-solution which is in fact @xmath121-solution of equation proportional to the vector @xmath197 ( @xmath73 fixed ) , @xmath136 . a general statement ( a slight adaption of ( * ? ? ?
* theorem 2 ) ) on solvability of equation is as follows : [ t:12 ] let @xmath116 be irreducible .
the system @xmath83 has a nonzero nonnegative solution @xmath98 if and only if * @xmath198 and @xmath59 is recurrent , or * when either @xmath199 or@xmath198 and @xmath59 is transient , there is an infinite sequence of indices @xmath200 such that ( @xmath201 ) @xmath202 for each @xmath119 . following chung @xcite we will use the analogues of the taboo probabilities : for @xmath203 define @xmath204 and for @xmath205 , @xmath206 clearly
, @xmath207 equals to the number of paths of length @xmath35 connecting @xmath52 to @xmath113 with no appearance of @xmath208 between .
denote also @xmath209 the number of paths of length @xmath35 connecting @xmath52 to @xmath113 with at least one appearance of @xmath208 between . the usual convention that @xmath210 will be used .
the following identities directly follow from the definitions of the corresponding generating functions - see before table 1 - or are easy to verify : for all @xmath211 and @xmath212 , * @xmath213 , * @xmath214 , * @xmath215 , * @xmath216 , * @xcite for @xmath217 , @xmath218 * @xcite @xmath219 is finite . by (
* theorem 2 ) the double limit can be replaced by @xmath220 using the identities ( i)-(vi ) we can write as @xmath221 since by ( vi ) , @xmath222 , using ( v ) we obtain that @xmath223 if and only if @xmath224 the conclusion follows from ( * ? ? ?
* theorem 2 ) .
[ c:1]if for each @xmath52 , @xmath225 except for a finite set of @xmath113 values , then @xmath83 has a nonzero nonnegative solution if and only if @xmath226 . in order to be able to modify nonnegative matrices in question
we will need the following observation . in some cases it will enable us to produce transition matrices of maps from @xmath21 .
let @xmath227 be the identity matrix , see .
[ p:19]let @xmath116 be irreducible . for arbitrary pair of positive integer @xmath208 and nonnegative integer
@xmath228 consider the matrix @xmath229 .
then * @xmath230 , * if for each @xmath52 , @xmath225 except for a finite set of @xmath113 values , the matrix @xmath231 belongs to the same class of the vere - jones classification as the matrix @xmath59 .
both the conclusions clearly hold if @xmath231 is a multiple of @xmath59 , when @xmath232 .
so to show our statement it is sufficient to verify the case when @xmath233 .
\(i ) since @xmath83 if and only if @xmath234 , property ( i ) follows from corollary [ c:1 ] .
\(ii ) by our assumption , for each @xmath52 , @xmath235 except for a finite set of @xmath113 values , so theorem [ t:12 ] and corollary [ c:1 ] can be applied . notice that for any nonnegative @xmath98 , @xmath236
so by theorem [ t:2 ] , the matrix @xmath59 is transient , resp .
recurrent if and only if @xmath233 is transient , resp . recurrent .
in order to distinguish different recurrent cases we will use table 1 . since by ( i ) @xmath237 , we can write @xmath238 by @xmath239 for each @xmath208 and we can put @xmath240 . for each @xmath241 there exists @xmath242 such that @xmath243 whenever @xmath244 .
then using the fact that @xmath245 we can write for any @xmath246 and sufficiently large @xmath247 : @xmath248 hence @xmath249 . by table 1 , @xmath59 is null , resp .
positive recurrent if and only if @xmath231 is null , resp .
positive recurrent .
finally , let @xmath116 be positive recurrent and assume its irreducible submatrix @xmath250 for some @xmath143 , denote @xmath251 . then similarly as above we obtain that @xmath237 , resp .
if @xmath59 is weakly , resp . strongly recurrent , then for some @xmath253 , resp
for each @xmath253 we obtain @xmath254 , resp .
@xmath255 and theorem [ t:1 ] can be applied .
this finishes the proof for @xmath233 .
now , the case when @xmath256 , @xmath257 , can be verified inductively .
the following statement identifies the topological entropy of a map and the perron value @xmath121 of its transition matrix .
[ p:7 ] let @xmath57 . then @xmath258 .
and if there is a summable @xmath19-solution of equation then @xmath259 . for the first equality , we start by proving @xmath260 .
we use proposition [ p:2](i ) and proposition [ p:11](ii ) . by those statements , @xmath261^{\frac{1}{n}}$ ] for any @xmath73 and and for each sufficiently large @xmath35 , the interval @xmath113 contains @xmath262 intervals @xmath263 with pairwise disjoint interiors such that @xmath264 for all @xmath265 .
clearly , the map @xmath266 has a @xmath262-horseshoe @xcite hence @xmath267 and @xmath268^{\frac{1}{n}}$ ] .
since @xmath35 can be arbitrarily large , the inequality @xmath260 follows .
now we look at the reverse inequality @xmath269 .
a pair @xmath270 is a subsystem of @xmath5 if @xmath271 $ ] is closed and @xmath272 .
it has been showed in ( * ? ? ?
* theorem 3.1 ) that the entropy of @xmath5 can be expressed as the supremum of entropies of _ minimal _ subsystems .
let us fix a minimal subsystem @xmath273 of @xmath5 for which @xmath274 .
* _ there are finitely many elements @xmath275 such that @xmath276 .
_ _ proof of claim .
_ let us denote @xmath277\setminus\bigcup_{i\in\p}i^{\circ}$ ] .
then @xmath278 is closed , at most countable and @xmath279 .
assume that @xmath280 .
then @xmath281 which is impossible for @xmath273 minimal of positive topological entropy .
if @xmath282 intersected infinitely many elements of @xmath22 then , since @xmath282 is closed , it would intersect also @xmath278 , a contradiction . thus , there are finitely many @xmath283 of the required property . @xmath284 + our claim together with proposition [ p:12 ] say that connect - the - dots map of @xmath273 is piecewise monotone and the finite submatrix @xmath134 of @xmath59 corresponding to the elements @xmath285 satisfies @xmath286 .
now the conclusion follows from proposition [ p:4 ] .
the second statement follows from theorem [ t:5 ] , proposition [ p:5 ] and the fact that topological entropy is a conjugacy invariant : @xmath287 .
we would like to transfer the vere - jones classification to @xmath21 .
that is why it is necessary to be sure that a change of markov partition for the map in question does not change the vere - jones type of its transition matrix .
this is guaranteed by the following proposition .
given @xmath40 , consider the family @xmath288 of _ all markov partitions _ for @xmath5 .
write @xmath289\setminus\bigcup_{i\in \p_{\alpha}}i^{\circ}$ ] .
the minimal markov partition @xmath290 for @xmath5 consists of the closures of connected components of @xmath24\setminus\bigcap_{\alpha}q_{\alpha}$ ] .
[ p:10]let @xmath39 with two markov partitions @xmath22 , resp .
@xmath96 and corresponding matrices @xmath291 , resp .
the matrices @xmath293 and @xmath294 belong to the same class of the vere - jones classification .
since the map @xmath5 is topologically mixing , each of the matrices @xmath293 , @xmath294 is irreducible and aperiodic .
moreover , by proposition [ p:7 ] the value guaranteed in proposition [ p:2](i ) equals @xmath15 and so is the same for both the matrices @xmath293 , @xmath294 - denote it @xmath19 .
let @xmath277\setminus\bigcup_{j\in\p}j^{\circ}$ ] and @xmath295\setminus\bigcup_{j\in\q}j^{\circ}$ ] .
first , let us assume that @xmath296 .
fix two elements @xmath73 , resp .
@xmath297 such that @xmath298 .
let us consider a path of the length @xmath35
@xmath299 with respect to @xmath22 ; by proposition [ p:12 ] each interval @xmath300 contains @xmath301 intervals of monotonicity of @xmath5 - denote them @xmath302 - such that @xmath303 whenever @xmath304 .
this implies that @xmath305 is the number of paths with respect to @xmath22 through the same vertices in order given by and , at the same time , it is an upper bound of a number of paths @xmath306 with respect to finer @xmath96 such that @xmath307 for each @xmath52 . considering all possible paths in and summing their numbers given by
, we obtain @xmath308 for each @xmath35 . on the other hand , since @xmath5 is topologically mixing and markov , there is a positive integer @xmath309 such that @xmath310 .
it implies for each @xmath35 , @xmath311 using and , we can write , @xmath312 hence by the third row of table 1 , @xmath293 is recurrent if and only if @xmath294 is recurrent . again from and we can see that @xmath313 is positive if and only if @xmath314 is positive and the fourth row of table 1 for @xmath315 can be applied . in order to distinguish weak , resp .
strong recurrence , for a @xmath144 let @xmath316 be such that @xmath317 using and again we can see that the perron values of the irreducible aperiodic matrices @xmath318 and @xmath319 coincide hence the gurevich entropies @xmath320 , @xmath321 are equal if and only if it the case for @xmath322 , @xmath323 ; now theorem [ t:1](ii),(iii ) applies . second , if @xmath324 and @xmath325 , let us consider the partition @xmath290 , where any element of @xmath290 equals the closure of a connected component of the set @xmath326 . the reader can easily verify that @xmath290 is a markov partition for @xmath5 . by the previous , the pairs of matrices
@xmath293 , @xmath327 , resp .
@xmath327 , @xmath294 belong to the same class of the vere - jones classification .
so it is true also for the pair @xmath293 , @xmath294 .
let @xmath57 .
applying proposition [ p:10 ] in what follows we will call @xmath5 transient , null recurrent , weakly recurrent or strongly recurrent respectively if it is the case for its transition matrix @xmath59 . the last three , resp .
two possibilities will occasionally be summarized by @xmath5 is recurrent , resp .
@xmath5 is positive recurrent. it is well known that if @xmath5 is piecewise monotone then it is strongly recurrent ( * ? ? ?
* theorem 0.16 ) .
in this section we investigate in more details the set of maps from @xmath21 that are conjugate to maps of constant slope ( linearizable , in particular ) .
relying on theorem [ t:5 ] , theorem [ t:2 ] and proposition [ p:10 ] our main tools will be local and global perturbations of maps from @xmath21 resulting to maps from @xmath21 .
some examples illustrating the results achieved in this section will be presented in section [ s:7 ] .
we start with an easy but rather useful observation .
its second part will play the key role in our evaluation using centralized perturbation - formula and its applications .
[ p:3 ] let @xmath57 .
* if @xmath5 is leo then any @xmath19-solution of is summable . *
any @xmath19-solution of satisfies @xmath328 ( i ) since @xmath5 is leo , for a fixed element @xmath52 of @xmath22 , there is an @xmath62 such that @xmath329 $ ] . then by proposition [ p:11](ii ) , @xmath330 for each @xmath331
this implies that any @xmath19-solution @xmath332 of satisfies @xmath333 so @xmath86 .
( ii ) we assume that @xmath5 is topologically mixing - see definition [ def : cpmm ] . for any fixed element @xmath334 there is an @xmath62 such that @xmath335 : since @xmath5 is topologically mixing , there exist positive integers @xmath336 and @xmath337 such that @xmath338 for every @xmath339 , resp .
@xmath340\neq\emptyset$ ] for every @xmath341 .
this implies that the interval @xmath342 contains @xmath343 whenever @xmath344 - fix one such @xmath35 .
then @xmath330 for any element @xmath113 of @xmath22 such that @xmath345 ; hence @xmath346 for any @xmath19-solution @xmath332 of .
the fundamental conclusion regarding linearizability of a map from @xmath21 provided by the vere - jones theory follows .
[ t:7 ] if @xmath39 is leo and recurrent , then @xmath5 is linearizable . by assumption
there exists a markov partition @xmath22 for @xmath5 such that the transition matrix @xmath347 is recurrent .
in such a case equation has a @xmath121-solution described in theorem [ t:2 ] .
since @xmath5 is leo the @xmath121-solution is summable by proposition [ p:3](i ) and the conclusion follows from theorem [ t:5 ] .
[ r:7 ] in section [ s:7 ] we present various examples illustrating theorem [ t:7 ] . in particular , we show a strongly recurrent non - leo map of an operator type that is not conjugate to any map of constant slope . in this subsection
we introduce and study two types of perturbations of a map @xmath5 from @xmath21 : local and global window perturbation .
[ d:2 ] for @xmath348 with markov partition @xmath22 ,
let @xmath73 such that @xmath349 is monotone .
we say that @xmath39 is a _ window perturbation of @xmath14 on @xmath113 ( of order @xmath208 , @xmath350 ) _ , if * @xmath5 equals @xmath14 on @xmath24\setminus j^{\circ}$ ] * there is a nontrivial partition @xmath351 of @xmath113 such that @xmath352 and @xmath353 is monotone for each @xmath52 .
notice that due to definition [ d:2 ] a window perturbation does not change partition @xmath22 ( but renders it slack ) . using a sufficiently fine markov partition for @xmath14 , its window perturbation @xmath5 can be arbitrarily close to @xmath14 with respect to the supremum norm . in the above definition an element of monotonicity of a partition is used .
so , for example we can take @xmath22 classical ( non - slack ) , or to a given partition @xmath147 and a given maximal interval of monotonicity @xmath52 of a map we can consider a partition @xmath354 finer than @xmath147 such that @xmath355 . [ p:14]let @xmath39 be a window perturbation of a map @xmath348 .
the following is true . *
if @xmath14 is recurrent then @xmath5 is strongly recurrent and @xmath356 . *
if @xmath14 is transient then @xmath5 is strongly recurrent for each sufficiently large @xmath208 .
fix a partition @xmath22 for @xmath14 , let @xmath5 be a window perturbation of @xmath14 on @xmath73 . applying proposition [ p:10 ] it is sufficient to specify the vere - jones class of @xmath5 with respect to @xmath22 .
consider generating functions @xmath357 , corresponding to @xmath14 , resp . @xmath5 and with radius of the convergence @xmath358 , resp .
notice that @xmath360 hence @xmath361 .
( i ) if @xmath14 is recurrent then by table 1 and , @xmath362 then , since @xmath363 , @xmath364 hence @xmath365 and @xmath5 is strongly recurrent .
( ii ) if @xmath14 is transient then by table 1 and , @xmath366 if for a sufficiently large @xmath208 , @xmath367 , necessarily @xmath368 and @xmath5 is strongly recurrent by table 1 .
let @xmath59 be a matrix indexed by the elements of some @xmath22 and representing a bounded linear operator @xmath76 on the banach space @xmath77 - see section [ s:2 ] .
it is well known @xcite , ( * ? ? ?
* theorem 3.3 ) that for @xmath369 the formula @xmath370 defines the resolvent operator @xmath371 to the operator @xmath372 we will repeatedly use this fact when proving our main results .
the following theorem implies that in the space of maps from @xmath21 of operator type an arbitrarily small ( with respect to the supremum norm ) local change of a map will result to a linearizable map .
[ t:6]let @xmath39 be a window perturbation of order @xmath208 of a map @xmath348 of operator type .
then @xmath5 is linearizable for every sufficiently large @xmath208 .
we will use the same notation as in the proof of proposition [ p:14 ] .
let us denote @xmath373 the transition matrix of a considered window perturbation @xmath374 of @xmath14 , let @xmath375 be the value ensured for @xmath376 by proposition [ p:2 ] , put @xmath377 . since @xmath14 is of operator type , it is also the case for each @xmath374 .
using proposition [ p:14 ] and theorem [ t:2 ] we obtain that for some @xmath378 the perturbation @xmath379 is recurrent and equation is @xmath380-solvable : @xmath381 where @xmath382 , @xmath383 . since by for each @xmath208 , @xmath384
we can deduce that @xmath385 is decreasing and @xmath386 by our definition of a window perturbation , for each @xmath177 , @xmath387 denote @xmath388 the spectral radius of the operator @xmath389 represented by the matrix @xmath390 . using
we can consider a @xmath391 for which @xmath392 .
then , since the resolvent operator @xmath393 represented by the matrix is defined well for each real @xmath394 as a bounded operator on @xmath395 @xcite , we obtain from , remark [ r:1 ] and @xmath396 now since @xmath374 is recurrent , theorem [ t:2 ] and theorem [ t:5 ] can be applied .
let @xmath57 .
for any pair @xmath45 we define the number @xmath397 in the corresponding strongly connected directed graph @xmath131 , @xmath398 is the length of the shortest path from @xmath52 to @xmath113 . in particular ,
such a path contains neither @xmath52 nor @xmath113 inside , so at the same time @xmath399 and @xmath400 for every @xmath401 . since @xmath402 for every pair @xmath403 ,
the suprema @xmath404 are either all finite or all infinite .
moreover , we have the following . [ p:13]let @xmath39 with two markov partitions @xmath22 , resp . @xmath96 .
then @xmath405 is finite for some @xmath203 if and only if @xmath406 is finite for some @xmath407 .
let @xmath277\setminus\bigcup_{j\in\p}j^{\circ}$ ] and @xmath295\setminus\bigcup_{j\in\q}j^{\circ}$ ] .
first , let us assume that @xmath296 .
fix two elements @xmath73 , @xmath297 such that @xmath298 .
since the map @xmath5 is topologically mixing , there exists a positive integer @xmath66 for which @xmath408 . for an @xmath119 and an @xmath409
satisfying @xmath28 we obtain @xmath410 and @xmath411 ; hence @xmath412 inequality together with property show that if @xmath405 is finite for some @xmath203 then @xmath406 is finite for some @xmath407 . on the other hand
, there has to exist an @xmath413 , @xmath414 such that @xmath415 , @xmath416 .
since also @xmath417 , we can write for @xmath413 @xmath418 inequality together with property show that if @xmath406 is finite for some @xmath407 then @xmath405 is finite for some @xmath203 .
if @xmath324 and @xmath325 , we can consider the partition for @xmath5 @xmath419 clearly , @xmath420\setminus\bigcup_{j\in{\mathbb{r}}}j^{\circ}=p\cup q$ ] and we can use the above arguments for the pairs @xmath421 and @xmath422 hence the conclusion for the pair @xmath423 follows .
so , in , for fixed @xmath57 and @xmath73 , we compare the shortest path from @xmath113 to @xmath52 ( numerator ) to the shortest path from @xmath52 to @xmath113 ( denominator ) and take the supremum with respect to @xmath52 .
for example , for our map from subsection [ ss:2 ] the values are equal to @xmath424 , when @xmath5 is leo , is finite for every @xmath73 .
theorem [ t:6 ] explains the role of a window perturbation in case of maps of operator type . in theorem [ t:9 ]
we obtain an analogous statement for maps of non - operator type under the assumption that the quantities in are finite .
[ t:9 ] let @xmath348 with a markov partition @xmath96 and such that the supremum in is finite for some @xmath297 .
let @xmath39 be a window perturbation of order @xmath208 of @xmath14 .
then @xmath5 is linearizable for every sufficiently large @xmath208 . fix a partition @xmath22 for @xmath14 and @xmath73 .
a perturbation of @xmath14 on @xmath113 of order @xmath350 will be denoted by @xmath374 . by our assumption ,
proposition [ p:13 ] and , the supremum @xmath425 is finite . the numbers @xmath426 , @xmath119 , do not depend on any window perturbation on an element of @xmath22 , because such a perturbation does not change @xmath22 ; we define @xmath427 , @xmath428 , @xmath429 , @xmath430 .
obviously for every @xmath35 , @xmath431 to simplify our notation , using proposition [ p:14 ] we will assume that @xmath14 is strongly recurrent , so this is also true for @xmath374 .
similarly as in the proof of proposition [ p:14 ] we obtain for each @xmath208 , @xmath432 moreover , as in , the sequence @xmath385 is decreasing and @xmath433 , @xmath434 .
let us show that for each sufficiently large @xmath208 there is a summable @xmath375-solution @xmath435 of equation .
using , we can write for any @xmath241 , sufficiently large @xmath436 and some positive constants @xmath437 , @xmath438^n.\end{aligned}\ ] ] since by the value @xmath19 does not depend on @xmath208 and @xmath434 , from follows that @xmath439 for any @xmath440 .
clearly the value @xmath441 given by a finite number of summands is finite , so taking , and together , using @xmath442 we obtain @xmath443 whenever @xmath440 .
this finishes the proof .
let @xmath14 be from @xmath21 . in this part we will consider a perturbation of @xmath14 with a markov partition @xmath22 consisting of infinitely many window perturbations on elements of @xmath22 ( and with independent orders ) done due to definition [ d:2 ] .
[ d:6]a perturbation @xmath5 of @xmath14 on @xmath143 will be called _
centralized _ if there is an interval @xmath444 $ ] , @xmath445 such that @xmath446 $ ] . for technical reasons we consider also an _ empty perturbation _
( @xmath447 ) as centralized .
let @xmath5 be a global ( centralized ) perturbation of @xmath14 on @xmath448 , denote @xmath449 .
we can write for @xmath145 @xmath450 where the coefficients @xmath146 were defined before remark [ r:1 ] .
we use formula to argue in our proofs . in the next theorem
the perturbation @xmath5 need not be of an operator type .
let @xmath451 be recurrent and linearizable .
assume that @xmath5 is a recurrent centralized perturbation of @xmath14 on @xmath147 .
if there are finitely many elements of @xmath147 that are @xmath14-covered by elements of @xmath452 , then @xmath5 is linearizable .
let @xmath453 be all elements of @xmath147 that are @xmath14-covered by elements of @xmath454 .
then @xmath455 here , the last inequality follows from our assumption that the map @xmath14 is recurrent and linearizable together with theorem [ t:2 ] and theorem [ t:5 ] .
using , and proposition [ p:3](ii ) we obtain @xmath456 so by theorem [ t:2 ] and theorem [ t:5 ] the map @xmath5 is linearizable . in the next theorem the perturbation @xmath5 need not be of operator type .
[ t:10]let @xmath348 be of operator type . if the transition matrix @xmath457 represents an operator @xmath76 of the spectral radius @xmath458 then any centralized recurrent perturbation @xmath5 of @xmath14 such that @xmath459 is linearizable .
the entropy assumption is always satisfied when @xmath14 is recurrent .
let @xmath5 be a centralized perturbation of @xmath14 on @xmath143 , denote @xmath449 . from proposition [ p:7 ] and our assumption on the topological entropy of @xmath14 and @xmath5 we obtain @xmath460 .
we can write for @xmath145 @xmath461 where the last inequality follows from the fact that @xmath462 for each @xmath463 and @xmath62 - for the definition of @xmath464 , see before remark [ r:1 ] . by our assumption
, formula represents the resolvent operator @xmath465 for every @xmath466 .
in particular , @xmath467 is a bounded operator on @xmath468 @xcite , hence with the help of remark [ r:1 ] we obtain @xmath469 and , can be rewritten as @xmath470 because @xmath471 for topologically mixing @xmath5 by proposition [ p:3](ii ) and theorem [ t:2 ] .
the conclusion follows from theorem [ t:2 ] and , .
it was shown in proposition [ p:14](i ) that for a recurrent @xmath14 we always have @xmath459 . in order to apply theorem [ t:10 ]
let us consider any map @xmath472 of operator type , fix @xmath241 . by theorem [ t:6 ] there is a strongly recurrent linearizable map @xmath14 of operator type for which @xmath473 .
similarly as in we can conclude that the transition matrix of @xmath14 satisfies the assumption of theorem [ t:10 ] . by that theorem ,
any centralized perturbation @xmath5 ( operator / non - operator ) of @xmath14 is linearizable ( such a centralized perturbation @xmath5 can be taken to satisfy @xmath474 ) .
for some @xmath475 consider the matrix @xmath476 given as @xmath477 clearly @xmath59 is irreducible but not aperiodic .
it has period @xmath478 , so we consider only @xmath479 .
obviously , @xmath480 using stirling s formula , we can write in the following statement we describe a class of maps that are not conjugate to any map of constant slope .
in particular they are not linearizable .
a rich space of such maps ( not only markov ) has been studied by different methods in @xcite .
[ p:20]let @xmath485 , @xmath208 even and @xmath228 odd , consider the matrix @xmath484 defined in . then @xmath486 is a transition matrix of a non - leo map @xmath5 from @xmath21 .
the map @xmath5 is null recurrent and it is not conjugate to any map of constant slope .
the matrix @xmath231 represents an operator @xmath487 on @xmath488 and notice that the entries of @xmath231 away from resp . on the diagonal
are even , resp .
draw a ( countably piecewise affine , for example ) graph of a map @xmath5 from @xmath21 for which @xmath231 is its transition matrix . since @xmath484 is null recurrent , the matrix @xmath231 is also null recurrent by proposition [ p:19 ] . solving the difference equation
one can verify that equation with @xmath491 has a @xmath19-solution if and only if @xmath492 ( this follows also from corollary [ c:1 ] ) and none of these solutions is summable . so by proposition [ p:19 ] and
theorem [ t:5 ] , the map @xmath5 is not conjugate to any map of constant slope . for some @xmath493 let @xmath494 be given by @xmath495 again ,
the matrix @xmath59 is irreducible but not aperiodic .
it has period @xmath478 , so we consider only the coefficients @xmath496 , see subsection [ ss:15 ] . in order to find a @xmath19-solution for @xmath59 we can use the difference equation for @xmath497 with the additional conditions @xmath498 and @xmath499 . using corollary [ c:1 ] and the direct computation one can show : * for any choice of @xmath493 , @xmath500 so that @xmath501 . *
if @xmath502 then @xmath503 and @xmath59 is transient .
there is a summable @xmath121-solution for @xmath59 if and only if @xmath1 . *
if @xmath504 then @xmath503 and @xmath59 is null recurrent .
there is a summable @xmath121-solution for @xmath59 if and only if @xmath1 . *
if @xmath505 then @xmath506 , and @xmath59 is strongly recurrent . there is a summable @xmath121-solution for @xmath59 if and only if @xmath507 .
* @xmath508 if and only if @xmath509 , if @xmath510 then @xmath506 . * for any choice of @xmath493 , @xmath500 hence @xmath511^{1/2n}=2\sqrt{ab}$ ] .
the last equality together with ( a ) show that @xmath59 is strongly recurrent if and only if @xmath510 , see table 1 . * if @xmath509 then there is a summable @xmath121-solution for @xmath59 if and only if @xmath1 . * if @xmath512 then there is a summable @xmath121-solution for @xmath59 if and only if @xmath507 . * if @xmath513 then @xmath59 is transient ; if @xmath514 then @xmath59 is null - recurrent ; @xmath510 then @xmath59 is strongly recurrent . *
the matrix @xmath515 is a transition matrix of a strongly recurrent non - leo map @xmath39 if and only if @xmath510 .
the map @xmath5 is not linearizable for @xmath516 . *
the matrix @xmath517 is a transition matrix of a transient non - leo map @xmath39 .
the map @xmath5 is linearizable if @xmath1 .
we have shown in section 5 that the subset of maps from @xmath21 that _ are linearizable _ is sufficiently rich in the case of _ non - leo maps _ of operator / non - operator type . in order to refine the whole picture , in this paragraph
we show how to detect interesting _ leo _ maps of operator / non - operator type . in the next two collections of examples we will use a simple countably infinite markov partition for the full tent map and test various possibilities of its global window perturbations .
for the full tent map @xmath519 , @xmath520 $ ] , consider the markov partition @xmath521\colon~n=0,1,\dots\}.\ ] ] we will study several global window perturbations of @xmath14 of the following general form : let @xmath522 be a sequence of odd positive integers and consider a global window perturbation @xmath523 of @xmath14 such that then using the notation of section [ s:4 ] and remark [ r:1 ] we can consider generating functions @xmath527 corresponding to the element @xmath528 : @xmath529 for each @xmath35 .
one can easily verify that @xmath530 with the help of proposition [ p:7 ] we denote @xmath531 , resp .
@xmath532 the topological entropy of @xmath533 , resp .
radius of convergence of @xmath534 ; also we put @xmath535 .
_ strongly recurrent : _ first of all , consider the set @xmath536 and the choice @xmath537 then by , @xmath538 for @xmath539 and @xmath540 for each @xmath541 , hence @xmath542^{1/n}=3 , \quad \phi=1/3 , \quad \sum_{n\ge 1}f(n)\phi^n=\infty.\ ] ] therefore by table 1 , @xmath543 , @xmath544 - for the upper bound , see @xcite .
this implies that the map @xmath545 is strongly recurrent hence by theorems [ t:2 ] and [ t:6 ] also linearizable for any @xmath228 . from
we obtain @xmath548^{1/n}=3 $ ] , @xmath549 .
moreover , by direct computation we can verify that @xmath550 since always @xmath551 .
it means that the map @xmath552 define by the choice is transient and by table 1 from section 4 in fact @xmath553 , proposition [ p:7 ] implies @xmath554 . if we consider in any set @xmath555 such that the inequalities are still satisfied , the same is true for a resulting perturbation @xmath556 .
misiurewicz and roth @xcite observed that the map @xmath552 is not conjugate to any map of constant slope .
it can be shown that for each choice of a sequence @xmath522 such that the corresponding @xmath5 has finite topological entropy the following dichotomy is true : either @xmath5 is recurrent and then equation has no @xmath19-solution for @xmath557 , or @xmath5 is transient and then equation does not have any @xmath19-solution .
@xmath560 let us define inductively a new set @xmath561 as follows : put @xmath562 and @xmath563 ; assuming that for some @xmath564 we have already defined @xmath565 and @xmath566 , to obtain @xmath567 we omit from @xmath568 the least number - denoted @xmath569- such that the choice still gives @xmath571 for corresponding window perturbation of @xmath572 .
clearly @xmath573 for each @xmath208 .
let @xmath574 and consider the global perturbation of @xmath14 corresponding to @xmath575 given by formula with @xmath576 replaced by @xmath577 .
the set @xmath577 contains infinitely many units ( by any choice @xmath578 gives @xmath579 ) .
moreover , our definition of @xmath577 implies @xmath580 , expanding on the example of ruette ( * ? ? ?
* example 2.9 ) ( see also @xcite ) , we have the following construction .
let @xmath584 be a non - negative integer sequence with @xmath585 , let @xmath586 a slope determined below in , and let @xmath587 $ ] , @xmath588 , be adjacent intervals converging to @xmath589 .
also let @xmath590 , @xmath591 , be adjacent intervals of length @xmath592 converging to @xmath424 and such that @xmath593 is the left boundary point of @xmath594 .
( 14,13)(-1,0 ) ( 0,0)(1,0)12(0,0)(0,1)12 ( 0,12)(1,0)12 ( 12,0)(0,1)12 ( 0,0)(1,1)12 ( 0,2)(1,0)12 ( -1.2,1.9)@xmath595 ( -0.5,7)@xmath596 ( 0,1.2)(1,0)12 ( -1.2,1.1)@xmath597 ( -0.5,1.6)@xmath598 ( 0,0.8)(1,0)12 ( -1.2,0.6)@xmath599 ( 0,0)(1,6)2 ( 2,12)(1,-6)1.8 ( 3.8,1.2)(1,6)0.13 ( 3.98 , 2)(1,-6)0.13 ( 4.16,1.2)(1,6)0.13 ( 4.34 , 2)(1,-6)0.13 ( 4.52,1.2)(1,6)0.13 ( 4.70 , 2)(1,-6)0.2 ( 4.90,0.8)(1,6)0.03 ( 5.2,0.85)@xmath600 ( 7.2,0.4)@xmath600 we define the interval map @xmath11 \to [ 0,1]$ ] with slope @xmath601 by @xmath602 , \\
2-\lambda x & \text { if } x \in [ \lambda^{-1 } , \lambda^{-2}(2\lambda-1 ) ] , \\
\text{composed of } 1 + 2a_n & \text { branches of slope } \mp \lambda \text { alternatively } \\ \text{mapping into } i_n & \text { if } x \in j_n,~ n \ge 1
. \end{cases}\end{aligned}\ ] ] to make sure that @xmath603 , we need @xmath604 to satisfy @xmath605 so , any sequence @xmath606 such that has a positive finite solution @xmath19 leads to the _ linearizable _ map @xmath607 .
one can easily see that @xmath608 is a markov ( slack ) partition for @xmath5 as defined in section [ s:2 ] .
we use the rome technique from @xcite ( see also ( * ? ? ?
* section 9.3 ) ) to compute the entropy of this graph : it is the leading root of the equation @xmath617 if we divide this equation by @xmath618 , then we get @xmath619 from table 1 follows that the graph @xmath187 ( the matrix @xmath59 , the map @xmath5 ) is recurrent for any choice of a sequence @xmath584 and corresponding finite @xmath620 . proposition [ p:7 ] and comparing equations and we find that @xmath621 . by remark [ r:3 ]
the map @xmath5 is of operator type if and only if @xmath622 . in this case , by table 1 and proposition [ p:7 ] , @xmath623 , so the corresponding map is always strongly recurrent .
for the choice @xmath624 for some fixed integer @xmath625 the map @xmath5 is of non - operator type . in this case , @xmath626 , so @xmath627 , hence by table 1 the map @xmath5 is still strongly recurrent .
we can also take @xmath628 for some sublinearly growing integer sequence @xmath629 chosen such that holds for @xmath630 , @xmath631 . in this case , @xmath632 and @xmath633 , and the system is null - recurrent or weakly recurrent ( not strongly recurrent ) depending on whether @xmath634 is infinite or finite .
although up to now all our main results have been formulated and proved in the context of continuous maps , many statements remain true also for countably piecewise monotone markov interval maps that are countably piecewise continuous
. will be twice as large , and the entropy increases accordingly in that case ] we will present a countably piecewise continuous , countably piecewise monotone example adapted from @xcite , where it is studied in detail for its thermodynamic properties .
let @xmath635 be a strictly decreasing sequence in @xmath24 $ ] with @xmath636 and @xmath637 .
we will consider the partition @xmath638 , where the interval map @xmath5 is designed to be linear increasing on each interval @xmath639 for @xmath640 , @xmath641 $ ] , @xmath642 for @xmath640 and @xmath643 $ ] . with a slight modification of our definition from section [ s:2 ]
, @xmath22 is a markov partition for @xmath5 and @xmath5 is leo .
let @xmath58 be the matrix corresponding to @xmath22 , see below .
in order to have constant slope @xmath19 , we need to solve the recursive relation ( 9,11)(-10,-0.5 ) ( -21,6)@xmath650 ( -19,0 ) @xmath651 if @xmath652 ( 0,0)(1,0)10(0,10)(1,0)10 ( 0,0)(0,1)10 ( 10,0)(0,1)10 ( 0,0)(1,1)10 ( 7.5,0)(1,4)2.5 ( 8.3,-0.9)@xmath653 ( 5.5,0)(1,5)2 ( 6.2,-0.9)@xmath654 ( 4,0)(1,5)1.5 ( 4.5,-0.9)@xmath655 ( 2.9,0)(1,5)1.1 ( 3.0,-0.9)@xmath656 ( 2.1,0)(1,5)0.8 ( 1.2,-0.9)@xmath657 ( 1.52,0)(1,5)0.58 ( 1.095,0)(1,5)0.425 ( 0.791,0)(1,5)0.304 ( 0.572,0)(1,5)0.219 ( 0.4138,0)(1,5)0.1582 ( 0.2974,0)(1,5)0.1164 ( 0.21464,0)(1,5)0.08276 it is known that @xmath658 hence by proposition [ p:7 ] , @xmath659 .
if we remove site @xmath660 ( remove the first row and column ) from @xmath59 , the resulting matrix is @xmath59 again , so strongly connected directed graph @xmath131 contains its copy as a proper subgraph and due to theorem [ t:1](i ) , @xmath59 and also @xmath5 is transient .
writing @xmath661 , we have found , in accordance with theorem [ t:2 ] , a positive summable @xmath19-solution of equation for each @xmath662 . summarizing , the map @xmath5 is conjugate to a map of constant slope @xmath19 whenever @xmath662 .
@xmath5 is also _
linearizable _ , since @xmath663 .
* @xmath685 is a conjugacy class of maps in @xmath21 . *
strongly connected directed graph @xmath131 contains its copy as a proper subgraph ( * ? ? ?
* theorem 4.5 , fig .
3 ) , so due to theorem [ t:1](i ) , @xmath5 is transient . *
the common topological entropy equals @xmath686 .
* equation has a positive summable @xmath19-solution for each @xmath687 .
we can factor out the left - right symmetry of this map by using the semiconjugacy @xmath688 , and the factor map @xmath689 has transition matrix @xmath690 with similar properties as the previous example .
therefore @xmath5 is conjugate to a map of constant slope @xmath19 whenever @xmath691 , and also linearizable , since @xmath692 . using proposition [ p:20 ] let @xmath693 .
we have discussed the fact that @xmath253 is a transition matrix of a non - leo map @xmath39 with corresponding markov partition denoted by @xmath22 . clearly by remark
[ r:3 ] @xmath253 represents a bounded linear operator - denote it by @xmath694 - on @xmath468 , so @xmath5 is of operator type .
we can conclude that : * @xmath695 - propositions [ p:7 ] , [ p:20 ] . * @xmath696 - proposition [ p:20 ] , section [ s:2 ] .
* @xmath5 is not conjugate to a map of constant slope ( is not linearizable ) - proposition [ p:20 ] .
* @xmath5 is null recurrent - proposition [ p:20 ] .
* let @xmath147 be a markov partition for @xmath5 , denote @xmath697 the transition matrix of @xmath5 with respect to @xmath147 representing a bounded linear operator @xmath698 on @xmath699 .
since @xmath700 we have @xmath701 - see ( i ) , section [ s:2 ] and proposition [ p:7 ] .
then by theorem [ t:10 ] any recurrent centralized ( operator / non - operator ) perturbation of @xmath5 is linearizable . in particular
it is true for any local window perturbation of @xmath5 on some element of @xmath147 - proposition [ p:14](i ) .
* let @xmath147 be a markov partition for @xmath5 which equals @xmath22 outside of some interval @xmath444\subset ( 0,1)$ ] .
let @xmath556 be a local window perturbation of @xmath5 on some element of @xmath147 ; from the previous paragraph ( v ) follows that @xmath556 is strongly recurrent and linearizable .
consider a centralized ( operator / non - operator ) perturbation @xmath702 of @xmath556 on some @xmath703
. then if @xmath702 is recurrent it is linearizable by theorem [ t:3 ] .
otherwise we can use either theorem [ t:6 ] ( an operator case ) or theorem [ t:9 ] ( non - operator case , @xmath704 is finite for @xmath145 ) to show that a local window perturbation of @xmath702 of a sufficiently large order is linearizable . | we study continuous countably piecewise monotone interval maps , and formulate conditions under which these are conjugate to maps of constant slope , particularly when this slope is given by the topological entropy of the map .
we confine our investigation to the markov case and phrase our conditions in the terminology of the vere - jones classification of infinite matrices . |
magnetic flux rope ( mfr ) , a volumetric current channel with helical magnetic field winding around its center axial field lines , plays a key role in solar eruptions manifested as coronal mass ejections ( cmes ) and/or flares .
the first direct observational evidence of the presence of an outstanding mfr in the sun - earth system is from the near - earth in - situ solar wind observation of the so called magnetic cloud ( burlaga et al .
according to a statistical study ( vourlidas et al .
2013 ) , at least @xmath0 40 % coronagraphic observations of cmes show an apparent flux - rope geometry in the outer corona . however , the direct detection of mfrs in the lower corona prior to and during the eruption has been elusive .
therefore , there has been a heated debate over whether the mfr exists before the onset of cme eruption or it is formed during the cme eruption ( e.g. , forbes 2000 ; klimchuk 2001 ) .
many cme models require a pre - existing mfr and have suggested a variety of initiation processes , including direct flux injection ( e.g , .
chen 1996 ) , the catastrophic loss of equilibrium ( forbes & isenberg .
1991 ; isenberg et al . 1993 ; hu et al . 2003
; chen et al .
2006 , 2007 ) , kink ( trk et al . 2004 ) or torus ( kliem & trk 2006 ; olmedo & zhang 2010 ) instability .
on the other hand , some theoretical models suggest that an mfr be formed during the cme eruption process through the magnetic reconnection of coronal arcade ( e.g. , miki & linker 1994 ; antiochos et al .
1999 ) .
several lines of observations in the lower corona indicate the presence of mfrs , e.g. , filaments / prominences ( rust & kumar 1994 ) , coronal cavities ( wang & stenborg 2010 ) , sigmoids ( titov & dmoulin 1999 ; mckenzie & canfield 2008 ) , hot channels ( zhang et al .
2012 ; cheng et al .
2012 , 2013b ; li & zhang 2013 ) and/or hot blobs ( cheng et al . 2011 ; su et al .
2012 ; patsourakos et al . 2013 ; song et al .
the term``blob '' originates from a 2-dimension description of plasmoids either in numerical simulations or observations .
however , it could also describe the 2-d appearance projected onto the plane of the sky of an intrinsically 3-d mfr structure when observed along the central axis .
all of the features mentioned above can be observed before the cme onset , thus supporting the pre - existing - mfr scenario . to the best of our knowledge ,
direct observations of the mfr formation from a non - mfr structure have been rare .
cheng et al .
( 2011 ) observed an mfr ( hot blob ) growth process during the impulsive phase of a solar eruption .
the hot blob was observed rising from the source active region , stretching the background magnetic field lines upward .
the stretched field lines then approached toward the center below the blob resulting in magnetic reconnection . above the reconnection site
, poloidal magnetic fluxes were injected into the erupting blob , leading to its growth .
different from cheng et al .
( 2011 ) , we report in this letter a formation process of a hot blob from an apparently non - mfr magnetic arcade , instead of the growth from a pre - existing progenitor blob .
the observation was made by the aia on board _ solar dynamic observatory ( sdo)_. the differential emission measure ( dem ) analysis shows its temperature around 10 mk .
the eruption of the hot blob led to a cme , which was recorded by the large angle spectroscopic coronagraph ( lasco ) ( brueckner et al .
1995 ) on board _ solar heliospheric observatory ( soho ) _ and the sun earth connection coronal and heliospheric investigation ( secchi ) ( howard et al . 2008 ) on board _ solar terrestrial relations observatory ( stereo)_. further , the eruption was driven by magnetic reconnections occurred in two spatially different places , which are found to play different roles : causing the arcade expansion and producing the blob , respectively . in section 2 , the instruments are described .
we present the observations and make discussions in section 3 , which are followed by a summary in section 4 .
the aia on board _ sdo _ can image the solar atmosphere in 10 narrow uv and euv passbands with a cadence of 12 seconds , spatial resolution of 1.2 arcseconds and fov of 1.3 r@xmath1 .
the temperature response functions of these passbands cover a temperature range from 0.6 to 20 mk ( odwyer et al .
2010 ; del zanna et al .
2011 ; lemen et al . 2012 ) . during an eruption ,
the 131 and 94 passbands are sensitive to the hot plasma from flare regions and erupting flux ropes , while other passbands are better for viewing dimming regions and cooler leading fronts ( lfs ) of eruptions ( e.g. , zhang et al .
2012 ; cheng et al . 2011
; song et al .
aia s high cadence and broad temperature coverage provide us the opportunity of observing the mfr formation process during an eruption , and make it possible for constructing dem models of corona plasma .
recently , a dem - based temperature analysis method has been used successfully to analyze hot blob temperatures ( cheng et al .
2012 , 2014 ; song et al . 2014 ) . in this letter
, we continue to use the same method to study the heating process of the arcades and hot blob .
in addition , the secchi on board _ stereo _ and lasco on board _ soho _ provide cme images in the outer corona from different perspectives .
the _ reuven ramaty high energy solar spectroscopic imager ( rhessi _ ; lin et al .
2002 ) provides the hard x - ray ( hxr ) spectrum and imaging of flares .
on 2013 november 21 , an m1.2 class soft x - ray ( sxr ) flare was recorded by _ geostationary operational environmental satellite ( goes ) _ , which started at 10:52 ut and peaked at 11:11 ut . at this time
, the separation angle of _ sdo _ and _ stereo - a _ is @xmath0149@xmath2 .
the flare location was at @xmath0s14w89 ( noaa 11893 ) from the perspective of _ sdo _ as shown in aia 131 image of figure 1(a ) .
the subregions depicted with dashed and solid squares show the fov of figure 1(c ) and figures 3(b)-(i ) , respectively .
a hot blob was observed and depicted with the red dotted line . in the stereo - a euvi 195 image ( figure 1(b ) ) , the flare region was shown on the disk as a solid square ( also the fov of figures 2(c)-(d ) ) . during the process of the hot blob formation and eruption , a cool lf ,
best seen in aia 171 in figure 1(c ) ( indicated by the blue dotted line ) , was formed , likely through the compression of the plasma surrounding the hot blob .
note that the lf only formed at a later time in the fast acceleration phase , similar to the formation sequence of the hot channel and its associated lf reported in zhang et al .
the hot blob is completely absent in the cool 171 image , as demonstrated in figure 1(c ) where the red dotted circle indicates the same location of the hot blob shown in 131 image in figure 1(a ) .
lasco observed the corresponding cme from 2.2 to 24 r@xmath1 with a linear speed of 668 km s@xmath3 ; the cme had a three - part structure : a bright loop - like lf ( blue dotted ) , a dark cavity , and an embedded bright core ( red dotted ) , as shown in figure 1(d ) . through inspecting the aia and lasco movies carefully ,
we conclude that the bright core and lf in lasco correspond to the hot blob in 131 and lf in 171 , respectively . using magnetic field observations from helioseismic and magnetic imager ( hmi ) on board _ sdo
_ , we find that the magnetic configuration of noaa 11983 experienced an obvious emergence process for several days before the eruption ( see supplementary hmi movie online ) . on november 12 , it is classified as hale class @xmath4 ( figure 2(a ) ) , and on november 18 , classified as @xmath4@xmath5 ( figure 2(b ) ) . figures 2(c ) and
( d ) present the euvi observations before and after the eruption , respectively ( also see euvi movie online ) .
the location of the bright post - flare arcade and its two footpoints ( depicted with red and blue dots in figure 2(d ) ) indicate where the eruption took place .
as shown in figure 2(b ) where the two footpoints are mapped onto the magnetogram images , the flare and the eruption occurred along the polarity inversion line ( pil ) .
the evolution of the magnetogram indicates that there existed a slow shearing motion along the pil .
such motion can lead to the arcade expansion upward as discussed later .
the most interesting aspect of this event is that the hot blob is formed from a rising loop arcade at the early stage of the eruption .
the sequence of snapshots of the formation process are presented in figure 3 .
figure 3(a ) shows the goes 1 - 8 sxr flux , with eight vertical lines indicating the times of the images in the panels below .
the aia 131 images are processed with an improved radial filter technique ( ma et al .
2011 ) . at 10:50:08 ut ( panel b ) ,
the lower coronal loops were observed clearly with aia 131 , which likely corresponded to the arcade in figure 2(d ) , but the higher arcade ( denoted with red dotted line ) overlying the loops were not obvious . immediately after the flare onset , the arcade became visible ( panel c , 10:52:08 ut ) .
the arcade should have existed before the flare onset and the appearance is due to the heating through the underneath flare reconnection .
the reconnection induced the arcade to expand outward .
in the mean time , new hot loops ( depicted with yellow arrows in panels c - e ) were observed to ascend continuously ( see aia movie online for a better view ) , further pushing the arcade upward ( panels c - e ) .
the rising arcade then stretched the overlying restraining magnetic field lines ( denoted with yellow dotted lines in panels e and f ) .
underneath the arcade apex , the arcade legs and the stretched field lines started to approach each other ( denoted with red arrows in panels f - h ) , forming a classical x - point configuration ( panel i ) .
we believe that a second magnetic reconnection took place at the x - point , different in the location and time from the earlier flare reconnection that drove the expansion of the arcade .
this second reconnection led to the formation of a hot blob .
the hot blob separated completely from the arcade between 11:06 and 11:07 ut , following which the hot blob faded quickly and became difficult to identify with the aia 131 images . before and during this high - lying blob - forming magnetic reconnection ,
the continuous rise of the arcade stretched itself into to a cusp - like geometry , as shown in panel i with red dotted line near the solar limb . during the reconnection
, we observed the shrinkage of the arcade below the x - point , returning the cusp shape to a loop - shape ( see aia movie online for the whole process ) .
therefore , we suggest that in this event there exist two reconnections taking place at different coronal heights and playing different roles in the development of the blob formation and eruption .
one is the low - lying flare reconnection , inducing the heating and expansion of the arcades ; the other is the high - lying reconnection , forming and releasing the hot blob .
existing magnetohydrodynamic ( mhd ) simulations in spherical coordinates ( miki & linker , 1994 ; antiochos et al .
1999 ) have shown that , subject to the localized shearing motion in the photosphere , the magnetic arcade in the corona expands outward in a process that stretches the field lines and produces a current sheet underneath between the arcade and photospheric pil .
magnetic reconnection occurs in the current sheet , leading to the formation and ejection of a blob ( magnetic flux rope ) .
such scenario shown in these simulations is supported by our observations , at least qualitatively . through carefully inspecting observations of aia and euvi , which provide images from different wavelengths and perspectives ,
we conclude that there are no any pre - existing or dormant mfr features before the eruption , i.e. , filament , sigmoid , blob or hot channel as mentioned in the introduction section .
what we observed is that the arcade evolved into a hot blob ( flux rope ) , which is consistent with the mhd models mentioned above .
the expansion of the arcade is possibly caused by the flare magnetic reconnection and/or new rising loops from below .
nevertheless , we want to point out that the arcade - to - flux rope scenario observed in this particular event may not occur in other events , in which mfr may exist before the eruption ( e.g. , zhang et al .
2012 ; cheng et al .
therefore , there is no single model that can explain the diversity of eruptive flares in real observations at present . with the dem method mentioned above ( see song et al .
2014 for details ) , we further illustrate the blob formation process using temperature maps ( figure 4 ) , which show the details of the temperature distribution and evolution of the arcades and the blob .
the temperature maps show that the arcade was heated from about 5 mk to 7 mk in the first 3 minutes ( panels a - c ) .
this is why the arcade appeared in the aia 131 passband ( in figure 3 ) after the flare onset .
the temperature continued to increase and approached to around 10 mk ( panels e - h ) . following the onset of the second magnetic reconnection at the high - lying x - point , the hot blob was formed ( see dem movie online ) . to further elucidate the relation between the magnetic reconnection and the heating , we investigated the location of the hxr sources using _ rhessi _ data , which had effective observations between 11:00 and 11:15 ut for our event .
_ rhessi _ observations are shown as black contours at 50% , 70% , and 90% of the maximum in the 12 - 25 kev band ( panels e - h ) .
the x - ray source appeared only in the lower loops , and no source was observed in the higher arcade and the blob .
it has been shown for some events that the magnetic reconnection region should be between double x - ray sources ( e.g. , liu et al .
2008 ; song et al .
therefore , we suggest that the flux - rope - forming reconnection was weak in this event and did nt have obvious contributions to the heating
. the flare region and the arcade should be heated mainly by the flare magnetic reconnection , which is consistent with song et al .
( 2014 ) .
was the hot blob a driver of the observed cme ?
for this event , the arcade began to rise around 10:52 ut , and the hot blob started to form at about 10:57 ut , while the appearance of the cool lf was around 11:06 ut as shown in figure 1(c ) .
this time sequence of different features supports that the blob is the driver .
however , we did not compare the velocities between the hot blob and the cool lf , because it became difficult to track the blob after 11:06 ut , at which time the lf started to appear . to further address the issue of the cause of the cme
, we studied the kinematic evolution of the rising arcades and the hot blob .
the heights of the arcade and later the hot blob apex ( red plus symbols shown in the middle and bottom panels of figure 3 ) are tracked in the aia images .
note that all heights refer to the top of the arcades ( or the apex of the hot blob at later times ) from the solar surface .
their velocities are plotted in figure 5(a ) ( red solid line ) , along with the _ goes _ sxr 1 - 8 profile shown by the black solid line .
we calculated these velocities from the numerical differentiation by using the three - point lagrangian interpolation of the height - time data .
note that the velocity uncertainties come mainly from the uncertainties in the height measurement .
the measurement errors are estimated to be four pixels , _
i.e. _ , 1700 km , which are propagated to estimate the velocity errors in the standard way . figure 5(a ) shows that the velocities increased from 40 km s@xmath3 to 180 km s@xmath3 continuously .
the velocity variation trend is tightly coinciding with the emission variation of the associated flare .
the temporal profile of the emission intensity integrated over the flare region ( depicted with the white box in figure 3(i ) ) in aia 131 is also plotted in this panel .
the euv profile is very similar to that of the sxr .
the two vertical dotted lines in figure 5 denote the time range when the dominant hot - blob - forming magnetic reconnection took place .
we further study the relation between the reconnection energy release and blob acceleration . in figure 5(b ) we plot the acceleration , _ rhessi _ 12 - 25 kev hxr count rates , and derivation of sxr versus time .
the close temporal correlation between hxr flux and the blob acceleration supports the idea that flares and cmes are two strongly coupled phenomena , possibly through a mutual positive feedback process between magnetic reconnection and mfr acceleration ( e.g. , zhang et al .
2004 ; temmer et al .
( 2008 ) ) .
it is also interesting to note that the hxr profile ( and the sxr derivative as well ) has double peaks . in particular , the second hxr peak appeared at the same time as the high - lying blob - forming reconnection took place .
this observation further supports the two - magnetic - reconnection scenario discussed earlier , _
i.e. _ , the low - lying flare reconnection and the high - lying flux - rope - forming reconnection .
the flare reconnection , responsible for producing the m1.2 class x - ray flare , was strong and located in a lower altitude , _
e.g. _ , near the magnetic loop footpoints , as indicated by the hxr source .
the flux - rope forming reconnection , occurring at the x - point between the arcade and the hot blob , might be a weaker magnetic reconnection , because of the weaker magnetic field near the arcade top .
nevertheless , this reconnection can also produce energetic electrons and demonstrate its existence through the second hxr peak .
the second magnetic reconnection is essential for the blob / flux rope formation and eruption .
an m1.2 class sxr flare and the associated cme occurred at the southwest limb on 2013 november 21 .
this event provided us the first direct observation of mfr formation process from a loop arcade during the eruption .
we summarize the formation process as follows .
the coronal arcade was heated by the underlying flare magnetic reconnection . in the same time , the arcade expanded upward , possibly due to the same reconnection process . the arcade was further strengthened and pushed upward by the new ascending loops from below .
the rising of the arcade further stretched the surrounding background magnetic field .
the arcade and the stretched magnetic field then curved - in beneath the loop apex , leading to the formation of an x - point .
then , the field lines near the x - point continued to approach each other and a second magnetic reconnection was induced , leading to the formation and eruption of the hot blob ( @xmath0 10 mk ) , presumably a mfr , and producing a cme .
we suggest that in this event there exist two reconnections taking place at different coronal heights and playing different roles in the development of the mfr formation and eruption .
one is the low - lying flare reconnection , inducing the heating and expansion of the arcade , and the other is the high - lying magnetic reconnection , forming and releasing the mfr .
we thank the referee for constructive comments that have greatly improved this manuscript .
we also thank y. m. wang , r. liu , q. hu , and g. l. huang for their valuable discussions and s. l. ma for her help on the radial filter technique .
sdo is a mission of nasa s living with a star program .
this work is supported by the 973 program 2012cb825601 , nnsfc grants 41274177 , 41104113 , 41274175 , and 41331068 .
j. zhang is supported by nsf grant atm-0748003 , ags-1156120 and ags-1249270 . | coronal mass ejections ( cmes ) are the most spectacular eruptive phenomena in the solar atmosphere .
it is generally accepted that cmes are results of eruptions of magnetic flux ropes ( mfrs ) . however , a heated debate is on whether mfrs pre - exist before the eruptions or they are formed during the eruptions .
several coronal signatures , _
e.g. _ , filaments , coronal cavities , sigmoid structures and hot channels ( or hot blobs ) , are proposed as mfrs and observed before the eruption , which support the pre - existing mfr scenario .
there is almost no reported observation about mfr formation during the eruption . in this letter , we present an intriguing observation of a solar eruptive event occurred on 2013 november 21 with the atmospheric imaging assembly on board the _ solar dynamic observatory _ , which shows a detailed formation process of the mfr during the eruption .
the process started with the expansion of a low - lying coronal arcade , possibly caused by the flare magnetic reconnection underneath .
the newly - formed ascending loops from below further pushed the arcade upward , stretching the surrounding magnetic field .
the arcade and stretched magnetic field lines then curved - in just below the arcade vertex , forming an x - point .
the field lines near the x - point continued to approach each other and a second magnetic reconnection was induced .
it is this high - lying magnetic reconnection that led to the formation and eruption of a hot blob ( @xmath0 10 mk ) , presumably a mfr , producing a cme .
we suggest that two spatially - separated magnetic reconnections occurred in this event , responsible for producing the flare and the hot blob ( cme ) , respectively . |
a topic of increasing interest to space weather is how the plasma and magnetic field characteristics of coronal mass ejections ( cmes ) result in geomagnetic storm activity , in particular those intense events .
the southward magnetic field and speed of cmes at the earth have received the most attention , because their cross product , the dawn - dusk electric field , controls the rate of the solar wind energy coupling to the terrestrial magnetosphere @xcite .
however , it is still not clear how the ejecta speed and southward magnetic field work together to achieve a sustained , enhanced dawn - dusk electric field and how they lead to the variability of geomagnetic storms . the southward magnetic field is often found within the ejecta reaching the earth in the form of an interplanetary cme ( icme ) with a preceding shock .
this usually leads to a classic geomagnetic storm sequence : a sudden commencement generated by the shock , a main decrease phase caused by the ejecta s southward magnetic field , and then a recovery phase .
in addition to the driver gas , the sheath region between the shock and icme can also be geo - effective ( e.g. , * ? ? ? * ; * ? ? ? * ) as both the sheath speed and southward magnetic field are amplified by shock compression .
the sheath - ejecta scenario has been invoked to explain the two - step development of geomagnetic storms @xcite .
complex ejecta resulting from interactions between cmes @xcite can be very geo - effective owing to their prolonged durations ( e.g. , * ? ? ?
* ; * ? ? ?
? * ; * ? ? ?
they could also cause two - step geomagnetic storms @xcite .
a special case of complex ejecta is the interaction of a preceding ejecta with an overtaking shock ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the shock enhances the pre - existing southward magnetic field inside the ejecta , an idea for increased geo - effectiveness dating back several decades @xcite .
a statistical analysis indicates that 19 out of 49 shocks propagating inside icmes are associated with an intense geomagnetic storm ( minimum @xmath3 nt ; * ? ? ?
a recent study combining remote - sensing and in situ observations suggests a perfect storm " scenario for the generation of an extreme space weather event @xcite : preconditioning of the upstream solar wind by an earlier cme plus in - transit interaction between later two closely launched cmes , in order to have an exceptionally high solar wind speed and unusually strong ejecta magnetic fields at 1 au .
this , again , emphasizes the crucial importance of cme - cme interactions for space weather . on 2015
march 17 and june 22 the earth underwent an intense geomagnetic storm with the minimum @xmath0 of @xmath1 and @xmath2 nt , respectively .
these are the largest geomagnetic storms of solar cycle 24 so far .
they occurred in the declining phase of the solar cycle , a phenomenon that is not uncommon @xcite .
we provide a timely analysis of the driving cmes , in an attempt to identify the plasma and magnetic field characteristics controlling the geomagnetic storm intensity and variability .
another focus of this letter is to test the
perfect storm " scenario proposed by @xcite : whether it is a rare coincidence or if it happens more frequently than what the phrase suggests .
we examine the solar wind signatures and their connections with the development of the geomagnetic storms , complemented with the modeling of the @xmath0 index using two empirical formulae based on the solar wind measurements @xcite .
we also use a grad - shafranov ( gs ) technique @xcite , which has been validated by well separated multi - spacecraft measurements @xcite , to reconstruct the in situ icme structure . the gs method can give a cross section as well as flux - rope orientation without prescribing the geometry . in conjunction with solar observations
it provides a larger spatial perspective of icmes than one - dimensional in situ measurements @xcite .
these efforts are key to understanding how the plasma and magnetic field characteristics of cmes are connected with the intensity of geomagnetic storms as well as their variability .
tracing back to the sun , the drivers of the 2015 march 17 geomagnetic storm were two interacting cmes on march 15 ( figure 1 , left ) .
the second cme ( cme2 ) had a maximum speed of about 1100 km s@xmath4 and was associated with a long - duration c9.1 flare from ar 12297 ( s22@xmath5w25@xmath5 ) that peaked at 02:13 ut on march 15 .
the first cme ( cme1 ) occurred on march 14 and had a speed of about 350 km s@xmath4 .
it was likely associated with a c2.6 flare from the same active region ( s21@xmath5w20@xmath5 ) that peaked around 11:55 ut on march 14 .
a first impression from the coronagraph images is that cme1 was largely propagating southward while cme2 had a major component heading west .
one may expect that the earth would encounter the flank of the ejecta , so this would not raise the alarm for a major geomagnetic storm . another fact that also makes the occurrence of an intense geomagnetic storm surprising is that the associated flares are relatively weak . in this sense , the occurrence of the 2015 march 17 intense geomagnetic storm is similar to the formation of the 2012 july 23 super solar storm that impacted stereo a @xcite . without white - light observations from stereo accurate cme kinematics can not be obtained .
figure 2 shows the in situ signatures observed at wind .
a shock passed wind at 04:01 ut on march 17 and caused the sudden commencement of the geomagnetic storm .
it is difficult to unambiguously connect the in situ signatures with the coronagraph images without wide - angle imaging observations from stereo .
however , application of an empirical model @xcite with the cme speed of 1100 km s@xmath4 gives a predicted arrival time of 23:59 ut on march 16 at wind ( 0.99 au from the sun ) , which is only 4 hours earlier than the observed shock arrival .
as can be seen from the figure , two icmes ( or flux ropes ) are identified .
our interpretation is different from those of @xcite and @xcite who identify a single , shorter icme interval from the data ( although with different durations ) .
the reason that we believe there are two icmes is as follows .
first , there are multiple rotations in the magnetic field components whose polarities change twice in the shaded data intervals .
obviously these features can not be explained by a single flux rope .
second , our icme intervals are an outcome of the gs reconstruction , which is sensitive to the chosen boundaries . despite the magnetic field fluctuations , both of the icmes can be reconstructed fairly well ( see description below ) .
third , the interpretation of two icmes is consistent with what the solar observations indicate ( see figure 1 and discussions below ) .
one may argue for a single icme interval ( say , from 12:58 ut on march 17 to 03:22 ut on march 18 ) based on the depressed proton temperature , a signature often used to identify icmes .
a reasonable gs reconstruction , however , can not be obtained for the interval and its variations .
given the presence of the compression by a high - speed stream from behind and cme - cme interactions , the criterion of a low proton temperature for identifying icmes may not be valid in the current case . within the icme intervals the maximum magnetic field strength is about 33 nt while the southward component reaches @xmath6 nt .
these are not small magnetic fields considering only the flanks of the cmes are encountered ( see figure 1 ) . interactions between these two icmes may have inhibited their expansion , a mechanism to create strong ejecta magnetic fields as we have seen from the 2012 july 23 event @xcite .
also note the high - speed stream compressing icme2 from behind , which may help maintain a strong ejecta magnetic field and a relatively high speed as well .
the @xmath0 profile indicates a two - step geomagnetic storm sequence with a global minimum of @xmath1 nt .
the first dip is produced by the southward magnetic field component in the sheath region behind the shock , while the second one results from the southward fields within the two icmes that last about @xmath7 hours . given the presence of a preceding shock and two interacting icmes , the two - step development of the geomagnetic storm can be classified as a sheath - ejecta - ejecta scenario .
the modeled @xmath0 index using the @xcite formula ( minimum @xmath8 nt ) generally agrees with actual @xmath0 measurements but underestimates the global minimum .
the @xcite scheme gives a deeper global minimum ( @xmath9 nt ) and a shallower recovery phase than measured .
further information on how the plasma and magnetic field characteristics control the geomagnetic storm activity is obtained from the gs reconstruction , as shown in figure 3 .
the reconstructions give a right - handed flux rope structure for both of the icmes , as can be judged from the transverse fields along the spacecraft trajectory together with the axial direction .
their axis orientations are almost opposite to each other : an elevation angle of about @xmath10 and azimuthal angle of about @xmath11 ( in rtn coordinates ) for icme1 , and an elevation angle of about @xmath12 and azimuthal angle of about @xmath13 for icme2 .
these low inclinations are consistent with the slightly tilted neutral line ( n. gopalswamy , private communication ) and filament channel ( m. temmer , private communication ) associated with the active region .
icme1 has a larger elevation angle ( @xmath10 ) , which may help explain the encounter in spite of the largely southward propagation direction of cme1 ( figure 1 ) .
the impact of icme2 may be accounted for by its lower elevation angle ( @xmath12 ) , although cme2 had a major section propagating westward .
in addition , the angular span of an icme can be @xmath14 or even larger @xcite , so it is not surprising that both of the icmes hit the earth .
it is also likely that the cme - cme interactions , the following high - speed stream and even the surrounding coronal magnetic field structures may have changed the propagation directions of both cmes ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
. given those relatively low inclination angles , the geomagnetic storm was mainly caused by the azimuthal magnetic field components of the flux ropes rather the axial components .
a particularly interesting feature shown by the cross section of icme1 is that the azimuthal field component ( with a maximum value of about 30 nt ) is much larger than the axial component ( maximum value of only 8 nt ) .
therefore , if the flux rope were vertically orientated with the axis pointing southward the geomagnetic storm would be much weaker since the southward field would be less .
this is contrary to the common belief that a southward pointing flux rope favors a strong geomagnetic storm .
icme2 has comparable axial and azimuthal magnetic field components .
another prominent feature is the vortices visible in the cross section of icme2 , reminiscent of the kelvin - helmholtz instability .
this is probably a result of the fast stream interacting with icme2 from behind ( see figure 2 ) .
in 2015 june ar 12371 exhibited elevated activity , somewhat similar to ar 11429 in 2012 march @xcite .
the active region produced a cme of about 1200 km s@xmath4 associated with an m3.0 flare from n13@xmath5e45@xmath5 that peaked at 17:36 ut on june 18 , a cme of about 1300 km s@xmath4 associated with an m2.0 flare from n12@xmath5e13@xmath5 peaking at 01:42 ut on june 21 , a cme of about 1000 km s@xmath4 associated with an m6.5 flare from n13@xmath5w05@xmath5 peaking at 18:23 ut on june 22 , and another one of about 1700 km s@xmath4 associated with an m7.9 flare from n10@xmath5w42@xmath5 peaking at 08:16 ut on june 25 .
all these cmes impacted the earth .
the june 21 cme appeared as a single halo event in the coronagraph images ( see figure 1 , right ) , and a near head - on collision with the earth was expected .
the corresponding in situ signatures at wind are displayed in figure 4 .
a cluster of shocks passed wind at 16:05 ut on june 21 , 05:02 ut and 18:00 ut on june 22 , and 13:12 ut on june 24 , respectively .
the icme boundaries are determined from the magnetic field in conjunction with the proton temperature and density .
the first shock ( s1 ) seemed driven by the june 18 cme , and the second one ( s2 ) was likely associated with a cme from june 19 .
no driver signatures are observed at wind for these two shocks , presumably owing to the largely eastward and southward propagation directions of the june 18 and 19 cmes respectively ( not shown here ) .
the icme and its preceding shock ( s3 ) were produced by the june 21 cme ; again , application of the empirical model @xcite with the cme speed of 1300 km s@xmath4 yields a predicted arrival time of 17:02 ut on june 22 at wind ( 1.02 au from the sun ) , which is only 1 hour earlier than the observed s3 arrival . the fourth shock ( s4 ) that was overtaking the icme at 1 au was associated with the june 22 cme .
a series of dips in the magnetic field strength are observed inside the icme , suggestive of the presence of current sheets .
this signature is possibly due to the heliospheric current sheet cutting through the ejecta , which may lead to a chain of small flux ropes within the icme ( see below ) .
the @xmath0 profile shows a multi - step geomagnetic storm with a global minimum of @xmath2 nt .
the first dip is produced by the fluctuating southward field component upstream of s3 ( likely owing to amplification by the two preceding shocks ) , the second one by the southward field in the sheath downstream of s3 ( further enhanced by s3 ) , and the major dip by the southward field in the first hatched interval inside the ejecta .
the southward field in the second hatched interval only creates a flattening of the @xmath0 value , perhaps because of the extremely low density .
note that the solar wind density upstream of the ejecta is significantly enhanced by the three preceding shocks .
this high density may feed the plasma sheet of the magnetosphere , which in turn helps intensify the ring current @xcite . given the presence of more than one preceding shocks and a single icme , the multi - step development of the geomagnetic storm can be classified as a sheath - sheath - ejecta scenario .
again , the modeled @xmath0 index using the @xcite formula ( minimum @xmath15 nt ) underestimates the global minimum but reproduces the recovery phase fairly well , whereas application of the @xcite scheme gives a larger global minimum ( @xmath16 nt ) than measured .
figure 5 shows the reconstructed cross sections of two small flux ropes identified inside the icme . here
we call them small flux ropes rather than icmes , in order to distinguish from the 2015 march 17 case . specifically
, the june 22 event is a single ejecta instead of multiple icmes .
these small flux ropes may have formed from the interaction between the cme and the heliospheric current sheet .
the reconstructions yield a left - handed structure for both of the flux ropes and similar axis orientations : an elevation angle of about @xmath17 and azimuthal angle of about @xmath18 for the first one ( fr1 ) , and an elevation angle of about @xmath19 and azimuthal angle of about @xmath20 for the second one ( fr2 ) .
the same chirality and similar axis orientations support the interpretation of a single icme .
both flux ropes have strong axial magnetic field components compared with the azimuthal ones .
also note the largely southward orientation of the flux ropes .
it is the strong axial field component on top of the largely southward flux - rope orientation , in conjunction with the relatively high solar wind speed , that may have resulted in the intense geomagnetic storm .
we have examined the sources of the 2015 march 17 and june 22 intense geomagnetic storms , the largest ones of solar cycle 24 up to the time of this writing .
key findings are obtained on how the plasma and magnetic field characteristics of cmes control the geomagnetic storm intensity and variability : \1 . a sheath - ejecta - ejecta mechanism and a sheath - sheath - ejecta scenario
are proposed for the development of multi - step geomagnetic storms , based on the observed and reconstructed solar wind structures associated with the cmes .
the 2015 march 17 geomagnetic storm shows a two - step development , which is produced by the southward magnetic field components behind the preceding shock and those within two interacting cmes .
this falls into the sheath - ejecta - ejecta category .
the 2015 june 22 geomagnetic storm exhibits a multi - step development , which is caused by the southward fields due to amplification by a series of preceding shocks and those within a single ejecta .
this is classified as a sheath - sheath - ejecta scenario .
the multiple preceding shocks and sheaths may precondition the magnetosphere for the growth of an intense geomagnetic storm .
we find two contrasting cases of how the cme flux - rope characteristics generate intense geomagnetic storms .
our gs reconstruction of the ejecta responsible for the 2015 june 22 geomagnetic storm indicates that the geomagnetic storm resulted from the largely southward flux - rope orientation with a strong axial magnetic field component .
however , for the 2015 march 17 geomagnetic storm the gs reconstruction reveals a much larger azimuthal field component than the axial component .
the intense geomagnetic storm occurred despite low flux - rope inclinations .
a southward flux - rope orientation is thus not a necessity for a strong geomagnetic storm to occur .
the perfect storm " scenario proposed by @xcite may not be as rare as the phrase implies .
the 2015 march 17 intense geomagnetic storm occurred in spite of the relatively weak solar flares and an encounter with the cme flank .
what makes it an intense geomagnetic storm is the interaction between two successive cmes plus the compression by a high - speed stream from behind , which helps maintain strong ejecta magnetic fields and a relatively high speed .
this is essentially the perfect storm " scenario a combination of circumstances results in an event of unusual magnitude , although the 2015 march 17 event is not super " in the same sense as the 2012 july 23 solar storm .
note that there are many combinations of circumstances that can occur to make an event more geo - effective , including pileup of events , pre - event rarefactions and field line stretching , shock enhancement of southward fields , and following high - speed streams causing compressions .
this perfect storm " scenario now seems useful and necessary to worry about because complex events with these combinations are common .
the research was supported by the recruitment program of global experts of china , nsfc under grant 41374173 and the specialized research fund for state key laboratories of china .
we acknowledge the use of data from wind and soho and the @xmath0 index from wdc in kyoto . | the largest geomagnetic storms of solar cycle 24 so far occurred on 2015 march 17 and june 22 with @xmath0 minima of @xmath1 and @xmath2 nt , respectively .
both of the geomagnetic storms show a multi - step development .
we examine the plasma and magnetic field characteristics of the driving coronal mass ejections ( cmes ) in connection with the development of the geomagnetic storms .
a particular effort is to reconstruct the in situ structure using a grad - shafranov technique and compare the reconstruction results with solar observations , which gives a larger spatial perspective of the source conditions than one - dimensional in situ measurements .
key results are obtained concerning how the plasma and magnetic field characteristics of cmes control the geomagnetic storm intensity and variability : ( 1 ) a sheath - ejecta - ejecta mechanism and a sheath - sheath - ejecta scenario are proposed for the multi - step development of the 2015 march 17 and june 22 geomagnetic storms , respectively ; ( 2 ) two contrasting cases of how the cme flux - rope characteristics generate intense geomagnetic storms are found , which indicates that a southward flux - rope orientation is not a necessity for a strong geomagnetic storm ; and ( 3 ) the unexpected 2015 march 17 intense geomagnetic storm resulted from the interaction between two successive cmes plus the compression by a high - speed stream from behind , which is essentially the perfect storm " scenario proposed by ( * ? ? ?
* i.e. , a combination of circumstances results in an event of unusual magnitude ) , so the perfect storm " scenario may not be as rare as the phrase implies . |
let @xmath2 be a graph ; we use @xmath3 to denote a subdivision of @xmath2 and call the vertices of the @xmath3 corresponding to the vertices of @xmath2 its _ branch _ vertices .
kuratowski s theorem states that a graph is planar iff it contains neither @xmath4 nor @xmath5 .
graphs containing no @xmath4 can be constructed from planar graphs and copies of @xmath0 by pasting them along cliques of size at most two .
the structure of graphs containing no @xmath5 is not well understood .
kelmans @xcite and , independently , seymour @xcite conjectured that 5-connected nonplanar graphs must contain @xmath5 .
thus , if the kelmans - seymour conjecture is true then graphs containing no @xmath5 is planar or admits a cut of size at most 4 .
note that the requirement on connectivity is best possible ( see , for example , @xmath6 ) .
ma and yu @xcite proved the kelmans - seymour conjecture for graphs containing @xmath1 , and kawarabayashi , ma and yu @xcite proved the kelmans - seymour conjecture for graphs containing @xmath7 .
we refer the reader to @xcite for problems and results ( as well as references ) related to the kelmans - seymour conjecture .
it turns out that @xmath1 is the right intermediate structure for studying the kelmans - seymour structure . by a result of
kawarabayashi @xcite , any 5-connected graph containing no @xmath1 has an edge @xmath8 that is _ contractible _
( i.e. , @xmath9 is also 5-connected ) .
thus , our strategy for proving the kelmans - seymour conjecture is to keep contracting edges incident with a special vertex to produce a smaller 5-connected graph . to avoid trivial components associated with @xmath10-cuts or @xmath11-cuts , we also contract triangles ( but we give preference to edges ) .
we now give a more detailed description of our strategy . for a graph @xmath12 and a connected subgraph @xmath13 of @xmath12 ( respectively , an edge @xmath8 of @xmath12 )
, we use @xmath14 ( respectively , @xmath9 ) to denote the graph obtained from @xmath12 by contracting @xmath13 ( respectively , @xmath8 ) .
let @xmath12 be a 5-connected nonplanar graph containing no @xmath1 .
then @xmath12 contains an edge @xmath8 such that @xmath9 is 5-connected .
if @xmath9 is planar , we can apply a discharging argument .
so assume @xmath9 is not planar .
let @xmath13 be a maximal connected subgraph of @xmath12 such that @xmath14 is 5-connected and nonplanar .
let @xmath15 denote the vertex representing the contraction of @xmath13 , and let @xmath16 .
then one of the following holds .
* @xmath17 contains a @xmath1 in which @xmath15 is of degree 2 . *
@xmath17 contains a @xmath1 in which @xmath15 is of degree 3 .
* @xmath17 does not contain @xmath1 , and there exists @xmath18 such that @xmath19 , @xmath20 or @xmath21 , and @xmath22 is 5-connected and planar .
* @xmath17 does not contain @xmath1 , and for any @xmath18 with @xmath19 and @xmath20 or @xmath21 , @xmath22 is not 5-connected .
we plan a series of five papers to establish the kelmans - seymour conjecture .
the purpose of this paper is to prove several results about 5-separations and 6-separations , which will be used in subsequent papers to reduce ( c ) and ( d ) to ( a ) or ( b ) .
note that 5-separations and 6-separations arise naturally when ( d ) occurs . in the second paper
, we will handle ( a ) .
the third and fourth papers take care of ( b ) . in the final paper
, we will deal with ( c ) and ( d ) . in our arguments throughout this series
, we frequently encounter the case when @xmath23 contains @xmath1 ; so the exclusion of @xmath1 ( in @xcite ) is very useful .
one of the main steps in the proofs in @xcite is to deal with 5-connected nonplanar graphs that admit a 5-separation with a planar side .
separation _ in a graph @xmath12 consists of a pair of subgraphs @xmath24 of @xmath12 , denoted as @xmath25 , such that @xmath26 , @xmath27 , and neither @xmath28 nor @xmath29 is a subgraph of the other .
the _ order _ of this separation is @xmath30 , and @xmath25 is said to be a _
@xmath31-separation _ if its order is @xmath31 .
let @xmath12 be a graph and @xmath32 .
we say that @xmath33 is _ planar _ if @xmath12 has a plane representation in which the vertices in @xmath34 are incident with a common face .
ma and yu @xcite proved that if @xmath12 has a 5-separation @xmath25 such that @xmath35 ( for @xmath36 ) and @xmath37 is planar then @xmath12 contains a @xmath5 with all branch vertices contained in @xmath29 . in order to establish the kelmans - seymour conjecture for graphs containing no @xmath1 , we need to study 5-separations and 6-separations with less restrictive structures .
let @xmath12 be a graph and @xmath38 .
then @xmath39 denotes the subgraph obtained from @xmath12 by deleting @xmath40 and all edges of @xmath12 incident with @xmath40 . for a positive integer @xmath41 , we let @xmath42=\{1,\ldots , n\}$ ] .
we often represent a path ( or cycle ) by a sequence of vertices . the following result deals with one type of 5-separations . [ apexside1 ]
let @xmath12 be _ a _ @xmath10-connected nonplanar graph and let @xmath43 be _ a _ @xmath10-separation in @xmath12 .
suppose @xmath44 for @xmath45 $ ] , @xmath46 , and @xmath47 is planar .
then one of the following holds .
* @xmath12 contains _ a _
@xmath5 in which @xmath40 is not _ a _ branch vertex .
* @xmath39 contains @xmath1 . *
@xmath12 has _ a _
@xmath10-separation @xmath48 such that @xmath49 and @xmath50 is the graph obtained from the edge - disjoint union of the @xmath51-cycle @xmath52 and the @xmath53-cycle @xmath54 by adding @xmath40 and the edges @xmath55 for @xmath56 $ ] .
let @xmath12 be a graph . for @xmath57
, we use @xmath58 $ ] to denote the subgraph of @xmath12 induced by @xmath59 , and let @xmath60 $ ] . for @xmath61 ,
we write @xmath62 $ ] instead of @xmath63 $ ] .
for @xmath64 , @xmath65 denotes the graph obtained from @xmath12 by deleting the edges in @xmath59 .
another type of 5-separations considered in this paper are those @xmath25 with the property that @xmath66 $ ] contains a triangle .
[ 5cut_triangle ] let @xmath12 be _ a _ @xmath10-connected graph and @xmath25 be _ a _ @xmath10-separation in @xmath12 .
suppose that @xmath35 for @xmath45 $ ] and @xmath66 $ ] contains _ a _ triangle @xmath67 .
then one of the following holds .
* @xmath12 contains _ a _
@xmath5 in which @xmath40 is not _ a _ branch vertex .
* @xmath39 contains @xmath1 .
* @xmath12 has _ a _
@xmath10-separation @xmath48 such that @xmath49 and @xmath50 is the graph obtained from the edge - disjoint union of the @xmath51-cycle @xmath52 and the @xmath53-cycle @xmath54 by adding @xmath40 and the edges @xmath55 for @xmath56 $ ] .
* for any distinct @xmath68 , @xmath69 contains @xmath5 . in the applications of theorems [ apexside1 ] and [ 5cut_triangle ]
, the vertex @xmath40 will represent the special vertex resulting from the contraction of a connected subgraph formed by a sequence of edges and triangles .
( allowing the contraction of triangles will ensure that @xmath35 in the applications of theorems [ apexside1 ] and [ 5cut_triangle ] . )
so @xmath70 of theorems [ apexside1 ] and [ 5cut_triangle ] gives a @xmath5 in the original graph , and @xmath71 of theorems [ apexside1 ] and [ 5cut_triangle ] allows us to apply the result of ma and yu @xcite to get a @xmath5 in the original graph .
the @xmath5 given in @xmath72 of theorem [ 5cut_triangle ] can be used to derive a @xmath5 in the original graph .
when @xmath73 of theorems [ apexside1 ] and [ 5cut_triangle ] occurs , we will use proposition [ 8cycle ] below , whose proof is included in this section ( as it is short ) .
let @xmath12 be a graph .
by @xmath61 we mean that @xmath17 is a subgraph of @xmath12 .
when @xmath74 and @xmath75 , we let @xmath76 . for @xmath57 , we may view @xmath59 as a subgraph of @xmath12 with vertex set @xmath59 and edge set @xmath77 . for @xmath61
, @xmath78 denotes the neighborhood of @xmath17 ( not including the vertices in @xmath79 ) . for any @xmath80
, we use @xmath81 to denote the neighborhood of @xmath82 in @xmath12 . when understood , the reference to @xmath12 may be dropped .
we may view paths as sequences of vertices .
the _ ends _ of a path @xmath83 are the vertices of the minimum degree in @xmath83 , and all other vertices of @xmath83 ( if any ) are its _ internal _ vertices .
a collection of paths are said to be _ independent _ if no vertex of any path in this collection is an internal vertex of any other path in the collection .
let @xmath12 be _ a _ 5-connected nonplanar graph , @xmath25 _ a _
5-separation in @xmath12 , @xmath84 such that @xmath29 is the graph obtained from the edge - disjoint union of the @xmath51-cycle @xmath85 and the @xmath53-cycle @xmath54 by adding @xmath40 and the edges @xmath55 for @xmath56 $ ] .
suppose @xmath86 .
then , for any @xmath87 , @xmath88 contains @xmath5 . by symmetry between @xmath89 and @xmath90
, we may assume that @xmath91 .
for convenience , let @xmath92 .
first , suppose @xmath93 .
since @xmath12 is @xmath10-connected , @xmath94 has independent paths @xmath95 from @xmath89 to @xmath96 , respectively .
then @xmath97 \cup ( q_1 \cup a_1b_1 ) \cup ( q_2 \cup a_2b_2 ) \cup
( q_3 \cup a_3b_3 ) \cup b_1b_4b_3 $ ] is a @xmath5 in @xmath98 with branch vertices @xmath99 .
now suppose @xmath100 . by symmetry
, we may assume @xmath101 .
since @xmath12 is @xmath10-connected , @xmath94 contains a path @xmath102 from @xmath103 to @xmath104 .
then @xmath97 \cup ( r \cup a_3b_3 ) \cup b_1b_4b_3 $ ] is a @xmath5 in @xmath98 with branch vertices @xmath99 .
finally , assume @xmath105 .
by symmetry , we may assume @xmath106 .
if @xmath107 has independent paths @xmath108 from @xmath109 to @xmath110 , respectively , then @xmath97 \cup ( r_1 \cup a_2b_2 ) \cup ( r_2 \cup a_3b_3 ) \cup b_1b_4b_3 $ ] is a @xmath5 in @xmath98 with branch vertices @xmath99 .
so we may assume that such @xmath108 do not exist .
then there exists a separation @xmath111 in @xmath112 such that @xmath113 , @xmath114 , and @xmath115 .
since @xmath12 is @xmath10-connected , @xmath116 and @xmath117 . but this implies @xmath118 , a contradiction .
the next result deals with a special type of @xmath11-separations , whose proof makes use of theorems [ apexside1 ] and [ 5cut_triangle ] . [ n(x1)capa ]
let @xmath12 be _ a _ @xmath10-connected graph and @xmath80 , and let @xmath25 be _ a _ @xmath11-separation in @xmath12 such that @xmath119 , @xmath66 $ ] contains a triangle @xmath120 , @xmath35 for @xmath45 $ ] .
moreover , assume that @xmath25 is chosen so that , subject to @xmath121 and @xmath35 for @xmath45 $ ] , @xmath28 is minimal .
let @xmath122 .
then @xmath123 , or one of the following holds .
* @xmath12 contains _ a _
@xmath5 in which @xmath82 is not _ a _ branch vertex . *
@xmath12 contains @xmath1 .
* there exists @xmath124 such that for any distinct @xmath125 , @xmath126 contains @xmath5 . * for some @xmath45 $ ] and
some @xmath127 $ ] , @xmath128 , and any three independent paths in @xmath129 from @xmath130 to @xmath131 , respectively , with two from @xmath132 and one from @xmath133 , must contain _ a _ path from @xmath133 to @xmath134 .
note that in @xmath71 of theorem [ n(x1)capa ] , we ask that @xmath12 contain @xmath1 instead of the stronger statement `` @xmath39 contains @xmath1 '' as in ( ii ) of theorems [ apexside1 ] and [ 5cut_triangle ] .
thus , @xmath73 of theorems [ apexside1 ] and [ 5cut_triangle ] does not occur in theorem [ n(x1)capa ] as it gives @xmath71 of theorem [ n(x1)capa ] .
this paper is organized as follows . in section 2
, we use the discharging technique to prove two lemmas about @xmath1 in apex graphs .
( a graph is apex if it has a vertex whose removal results in a planar graph . ) in section 3 , we collect a number of known results and prove theorem [ apexside1 ] . in section 4 , we prove a result about apex graphs ( from which one can see how ( c ) might be taken care of ) . in section 5 , we prove theorem [ 5cut_triangle ] . in section 6 , we prove theorem [ n(x1)capa ] , using theorems [ apexside1 ] and [ 5cut_triangle ] .
we end this section with additional notation and terminology .
let @xmath12 be a graph .
let @xmath74 , @xmath57 , and @xmath135 a collection of 2-element subsets of @xmath136 ; then @xmath137 denotes the graph with vertex set @xmath136 and edge set @xmath138 . if @xmath139 and @xmath140 , we write @xmath141 instead of @xmath142 . a set @xmath57 is a _ @xmath31-cut _ ( or a _ cut _ of size @xmath31 ) in a graph @xmath12 , where @xmath31 is a positive integer , if @xmath143 and @xmath12 has a separation @xmath25 such that @xmath144 and @xmath145 for @xmath45 $ ] .
if @xmath146 and @xmath147 is a cut of @xmath12 , then @xmath148 is said to be a _ cut vertex _ of @xmath12 . given a path @xmath83 in a graph and @xmath149
, @xmath150 denotes the subpath of @xmath83 between @xmath82 and @xmath151 ( inclusive ) .
a path @xmath83 with ends @xmath152 and @xmath148 ( or an @xmath152-@xmath148 path ) is also said to be _ from @xmath152 to @xmath148
_ or _ between @xmath152 and @xmath148_. a _ plane _ graph is a graph drawn in the plane with no edge - crossings . the unbounded face of a plane graph is usually called its _ outer face_. the boundary of the outer face of a connected plane graph is the _ outer walk _ of the graph ( or _ outer cycle _ when it is a cycle ) .
it is well known that if @xmath12 is a 2-connected plane graph then every facial boundary of @xmath12 is a cycle .
let @xmath153 be a cycle in a plane graph .
given @xmath154 , if @xmath155 then @xmath156 denotes the subpath of @xmath153 between @xmath82 and @xmath151 ( inclusive ) in clockwise order ; and if @xmath157 then @xmath156 is simply the trivial path with the single vertex @xmath157 .
in this section , we prove results about @xmath1 in certain apex graphs , using the discharging technique . first , we give a simple lemma on discharging . for a plane graph @xmath12 ,
let @xmath158 denote the set of all faces of @xmath12 and , for each @xmath159 , let @xmath160 denote the number of edges of @xmath12 incident with @xmath161 ( with each cut edge counted twice ) .
[ discharge ] let @xmath12 be a connected plane graph and let @xmath162 , the set of nonnegative integers , such that @xmath163 for all @xmath164 .
let @xmath165 be obtained from @xmath166 as follows : for each @xmath159 with @xmath167 , choose two vertices incident with @xmath161 and send charge @xmath168 from @xmath161 to each of these two vertices . then @xmath169 and
if @xmath170 then , for @xmath146 , @xmath171 by euler s formula , we have @xmath172 now suppose @xmath170 .
then for each @xmath146 , @xmath148 is contained in at most @xmath173 facial triangles .
so @xmath174 .
by considering @xmath175 modulo 4 , we get the desired bounds on @xmath176 . to state the remaining results in this section , we need a concept on connectivity .
let @xmath12 be a graph and @xmath32 , and let @xmath31 be a positive integer .
we say that @xmath12 is _ @xmath177-connected _ for some positive integer @xmath31 if , for any cut @xmath135 of @xmath12 with @xmath178 , each component of @xmath179 contains a vertex from @xmath34 .
recall that @xmath33 is planar if @xmath12 has a plane representation in which the vertices in @xmath34 are incident with a common face .
[ planarside ] let @xmath12 be _ a _ connected graph and @xmath32 such that @xmath180 , @xmath181 , @xmath12 is @xmath182-connected , and @xmath33 is planar .
then , for any @xmath183 , @xmath39 contains @xmath1 .
it suffices to prove the lemma for the case when @xmath34 is an independent set in @xmath12 .
so we assume that @xmath34 is an independent set in @xmath12 .
first , we show that @xmath39 is connected for any @xmath183 . for , if not
, then we may let @xmath184 be two components of @xmath39 such that @xmath185 .
since @xmath12 is @xmath182-connected , @xmath186 for @xmath45 $ ] . now @xmath187 is a cut in @xmath12 such that @xmath188 and @xmath65 has a component contained in @xmath189 , contradicting the assumption that @xmath12 is @xmath182-connected .
take a plane representation of @xmath12 such that the vertices in @xmath34 are incident with the outer face of @xmath12 .
let @xmath190 such that @xmath191 for all @xmath192 . then by lemma [ discharge
] , the total charge is @xmath193 .
note that for any @xmath192 , if @xmath194 then @xmath195 , or @xmath196 and @xmath197 ( in which case , @xmath198 ) . for each @xmath199 with @xmath200 , choose two vertices of @xmath201 incident with @xmath161 ( which exists as @xmath34 is independent ) , and send a charge of @xmath168 from @xmath161 to each of these two vertices .
let @xmath165 denote the resulting charge function .
then @xmath202 for all @xmath199 .
suppose @xmath203 .
then , for each @xmath204 , @xmath176 has the upper bound in lemma [ discharge ] .
so @xmath205 when @xmath206 , @xmath207 for @xmath208 ( as @xmath34 is independent ) , and @xmath209 for @xmath210 .
let @xmath211 denote the outer face of @xmath12 .
since @xmath34 is independent in @xmath12 , @xmath212 .
if at least two vertices in @xmath213 each have degree 2 or more in @xmath39 , then @xmath214 ; so @xmath215 if at most one vertex in @xmath213 has degree 2 or more in @xmath39 then , since @xmath12 is @xmath182-connected , @xmath216 and @xmath217 ; so @xmath218 therefore , @xmath219 a contradiction .
the next result will not be used in this paper , but will be used in subsequent papers in the series ; we include it here as its proof also uses discharging .
[ 6-cut2 ] let @xmath12 be _ a _ graph , @xmath32 , and @xmath183 such that @xmath220 , @xmath221 , @xmath222 is planar , and @xmath12 is @xmath223-connected .
then one of the following holds .
* @xmath39 contains @xmath1 , or @xmath12 contains _ a _
@xmath1 in which the degree of @xmath40 is @xmath224 .
* @xmath12 has _ a _
@xmath10-separation @xmath25 such that @xmath46 , @xmath225 , @xmath226 , and @xmath227 is planar .
we may assume that * @xmath12 has no 5-separation or 6-separation @xmath228 such that @xmath229 , @xmath230 , @xmath231 , and @xmath232 is planar . for
, suppose such a separation @xmath228 does exist in @xmath12 , then @xmath233 is @xmath234-connected . if @xmath235 then @xmath71 holds .
if @xmath236 we may work with @xmath233 and @xmath237 instead of @xmath12 and @xmath34 . by ( 1 )
, @xmath34 is independent in @xmath12 .
we may further assume that * @xmath238 and each vertex in @xmath213 has at least two neighbors in @xmath201 .
first , suppose @xmath239 .
let @xmath240 denote the two vertices in @xmath241 .
since @xmath12 is @xmath223-connected , @xmath242 for @xmath45 $ ] . therefore , since @xmath222 is planar , @xmath243 and @xmath244 .
thus , @xmath70 holds .
so assume @xmath238 .
then by ( 1 ) , each vertex in @xmath213 has at least two neighbors in @xmath201 , completing the proof of ( 2 ) .
we may also assume that * @xmath201 is connected . for
, otherwise , let @xmath245 be two components of @xmath201 .
since @xmath12 is @xmath223-connected , @xmath246 for @xmath45 $ ] . hence , @xmath247 , contradicting the assumption that @xmath222 is planar .
this proves ( 3 ) .
we now apply a discharging argument to @xmath39 by taking a plan representation of @xmath39 in which the vertices in @xmath213 are incident with its outer face .
let @xmath190 such that @xmath191 for all @xmath192 . by ( 2 ) and ( 3 ) , @xmath39 is connected .
so by lemma [ discharge ] , the total charge is @xmath193 .
note that for any @xmath192 , if @xmath194 then @xmath248 , or @xmath196 and @xmath197 ( in which case , @xmath198 ) . for each @xmath199 with @xmath200
, we may assume that @xmath161 is incident with two vertices in @xmath249 ; for , otherwise , @xmath70 holds .
so for each @xmath199 with @xmath200 , we choose two vertices from @xmath249 incident with @xmath161 , and send a charge of @xmath168 from @xmath161 to each of these two vertices .
let @xmath165 denote the resulting charge function .
then @xmath202 for all @xmath199 .
we may assume that @xmath250 , as otherwise @xmath71 holds .
then for each @xmath204 , @xmath176 has the upper bound in lemma [ discharge ] .
thus , @xmath205 if @xmath251 , and @xmath252 if @xmath210 ( by ( 2 ) ) .
denote by @xmath211 the outer face of @xmath39 . since @xmath34 is independent in @xmath12 , @xmath253 .
therefore , @xmath254 .
hence , the total charge is @xmath255 this is a contradiction .
in this section , we prove theorem [ apexside1 ] . for convenience , we introduce the following terminology .
let @xmath12 be a graph and @xmath32 .
we say that @xmath33 is _ plane _ if @xmath12 is drawn in the plane with no edge crossings such that the vertices in @xmath34 are incident with the outer face .
moreover , for @xmath256 , we say @xmath257 is _ plane _
( respectively , _ planar _ ) if @xmath12 is drawn ( respectively , has a drawing ) in a closed disc in the plane with no edge crossings such that @xmath258 occur on the boundary of the disc in this cyclic order ( clockwise or counterclockwise ) .
we also need a few known results .
the first result is a consequence of a more general result of seymour @xcite ( with equivalent versions proved in @xcite ) .
[ 2path ] let @xmath12 be _ a _ graph and let @xmath259 be distinct such that @xmath12 is @xmath260 @xmath261-connected .
then either @xmath12 contains disjoint paths from @xmath262 to @xmath263 and from @xmath264 to @xmath265 , or @xmath266 is planar .
the next lemma is theorem 4.3 in @xcite , where it is used to prove that if a 5-connected nonplanar graph has a 5-separation with one side planar and nontrivial then it contains @xmath5 .
[ w4 ] let @xmath12 be _ a _ graph and let @xmath267 be distinct such that @xmath268 @xmath269 is plane .
let @xmath270 and suppose @xmath12 is @xmath182-connected and @xmath181 .
then there exist @xmath271 , _ a _ cycle @xmath272 in @xmath273 , and four paths @xmath274 from @xmath275 to @xmath34 such that * @xmath276 for @xmath277 , and @xmath278 for @xmath56 $ ] , and * there exist @xmath279 such that @xmath109 is an end of @xmath280 and @xmath281 is an end of @xmath282 .
we need two results from @xcite , which may be viewed as apex versions of lemma [ w4 ] .
they are used in @xcite to deal with 5-separations with an apex side .
lemma [ apex1 ] is corollary 2.11 in @xcite , and lemma [ apex2 ] is corollary 2.12 in @xcite .
[ apex1 ] let @xmath12 be _ a _ connected graph with @xmath181 , @xmath32 with @xmath180 , and @xmath183 , such that @xmath12 is @xmath182-connected , @xmath283 is plane , and @xmath12 has no 5-separation @xmath25 with @xmath284 and @xmath226 .
let @xmath285 such that @xmath275 is not incident with the outer face of @xmath39 .
then * the vertices of @xmath39 cofacial with @xmath275 induce _ a _ cycle @xmath272 in @xmath39 , and
* @xmath39 contains paths @xmath286 from @xmath275 to @xmath213 such that @xmath276 for @xmath287 , and @xmath288 for @xmath289 $ ] .
[ apex2 ] let @xmath12 be _ a _ connected graph with @xmath181 , @xmath32 with @xmath180 , and @xmath183 , such that @xmath250 , @xmath12 is @xmath182-connected , @xmath290 is plane , and @xmath12 has no 5-separation @xmath25 with @xmath225 and @xmath226
. then @xmath39 is 2-connected .
moreover , either @xmath12 is the graph obtained from the edge - disjoint union of the @xmath51-cycle @xmath85 and the @xmath53-cycle @xmath54 by adding @xmath40 and the edges @xmath55 for @xmath56 $ ] , with @xmath291 , or there exists @xmath292 such that * the vertices of @xmath39 cofacial with @xmath275 induce _ a _ cycle @xmath272 in @xmath39 such that @xmath293 , where @xmath153 denotes the outer cycle of @xmath39 , * there exist paths @xmath274 in @xmath12 from @xmath275 to @xmath34 such that @xmath276 for @xmath277 , and @xmath288 for @xmath56 $ ] , and * either @xmath294 , or @xmath295 and we may write @xmath296 such that @xmath297 , @xmath298 for @xmath299 , and @xmath300 occur on @xmath153 in cyclic order .
_ proof of theorem [ apexside1]_. we choose such separation @xmath25 that @xmath29 is minimal .
then we may assume that @xmath29 has no 5-separation @xmath301 such that @xmath302 , @xmath303 , and @xmath304 .
for , suppose such @xmath301 does exist . then @xmath305 ; otherwise , @xmath306 contradicts the choice of @xmath25 ( that @xmath29 is minimal ) . hence , @xmath307 is planar ; so @xmath71 holds by lemma [ planarside ] .
let @xmath308 such that @xmath309 is plane .
let @xmath153 denote the outer walk of @xmath310 ; so @xmath96 occur on @xmath153 in clockwise order .
we may assume that neither @xmath311 nor @xmath312 is planar ; else @xmath71 holds by lemma [ planarside ] .
suppose there exists some @xmath313 .
then by lemma [ apex1 ] , the vertices of @xmath310 cofacial with @xmath275 induce a cycle @xmath272 in @xmath310 , and @xmath310 contains paths @xmath314 from @xmath275 to @xmath213 such that @xmath276 for @xmath315 , and @xmath288 for @xmath289 $ ] . without loss of generality
, we may assume that @xmath316 for @xmath289 $ ] . let @xmath317
. since @xmath12 is 5-connected , there exist independent paths @xmath318 in @xmath28 from @xmath151 to @xmath319 , respectively .
then @xmath320 is a @xmath5 in @xmath12 in which @xmath40 is not a branch vertex .
so @xmath70 holds .
thus , we may assume that @xmath321 .
by the minimality of @xmath29 , @xmath34 is independent in @xmath29 .
hence @xmath322 is planar .
moreover , we may assume @xmath323 .
for otherwise , let @xmath324 be the unique neighbor of @xmath40 in @xmath325 . then @xmath326 is a 5-separation in @xmath12 and @xmath327 is planar . if @xmath328 then @xmath71 holds by lemma [ planarside ] .
so we may assume @xmath329 and let @xmath330 be the only vertex in @xmath331 .
since @xmath12 is 5-connected , @xmath330 is adjacent to all of @xmath332 . but
this forces @xmath324 to have degree at most 4 in @xmath12 ( as @xmath333 is planar ) , a contradiction .
suppose @xmath71 and @xmath73 of theorem [ apexside1 ] both fail .
then by lemma [ apex2 ] , @xmath310 is @xmath224-connected ( so @xmath153 is a cycle ) and there exists @xmath334 such that * the vertices of @xmath310 cofacial with @xmath275 induce a cycle @xmath272 in @xmath310 such that @xmath293 , * there exist paths @xmath335 in @xmath29 from @xmath275 to @xmath34 such that @xmath276 for @xmath336 , and @xmath288 for @xmath56 $ ] , and * either @xmath337 , or @xmath338 and we may relabel @xmath339 ( if necessary ) such that @xmath297 , @xmath298 for @xmath340 , and @xmath341 occur on @xmath153 in cyclic order . for convenience ,
let @xmath342 and assume , without loss of generality , that @xmath96 occur on @xmath153 in clockwise order . by the planarity of @xmath310
, we may assume that @xmath343 is a path for @xmath56 $ ] .
_ case _ 1 .
@xmath337 . without loss of generality
, we may assume that @xmath298 for @xmath56 $ ] .
if @xmath94 has disjoint paths @xmath344 from @xmath345 to @xmath346 , respectively , then @xmath347 is a @xmath5 in @xmath12 in which @xmath40 is not a branch vertex ; so @xmath70 holds .
hence , we may assume that such @xmath344 do not exist .
then , since @xmath28 is @xmath223-connected , it follows from lemma [ 2path ] that @xmath348 is plane and , hence , @xmath349 is connected . _
subcase _ 1.1 .
@xmath349 is not @xmath224-connected .
if @xmath350 then let @xmath245 denote the two disjoint 1-vertex subgraphs of @xmath349 , and otherwise let @xmath351 denote a 1-separation in @xmath349 with @xmath352 for @xmath45 $ ] . since @xmath12 is 5-connected and @xmath348 is plane , @xmath353 and @xmath354 is connected , for @xmath45 $ ] .
without loss of generality , we may assume that @xmath355 and @xmath356 .
suppose @xmath357 . if @xmath358 for some @xmath359 then @xmath360 is planar ; so @xmath71 holds by lemma [ planarside ] .
thus , we may assume that there exist @xmath361 and @xmath362 .
let @xmath363 be obtained from @xmath364 by replacing @xmath365 with @xmath366 , respectively , and let @xmath102 be a path in @xmath367 from @xmath103 to @xmath368 .
then @xmath369 is a @xmath5 in @xmath12 in which @xmath40 is not a branch vertex .
so @xmath70 holds .
hence , we may assume that there exists @xmath370 such that @xmath371 .
recall that @xmath372 .
then @xmath373 has a path @xmath374 which is either from @xmath324 to @xmath103 ( and disjoint from @xmath375 ) or from @xmath324 to @xmath368 ( and disjoint from @xmath376 ) .
first , assume that @xmath374 is from @xmath324 to @xmath103 .
let @xmath377 be the path in @xmath378 between @xmath275 and @xmath324 , and @xmath363 be obtained from @xmath364 by replacing @xmath376 with @xmath377 .
let @xmath379 be a path in @xmath380 $ ] from @xmath109 to @xmath104 , and @xmath381 be a path in @xmath382 $ ] from @xmath40 to @xmath368 .
now @xmath383 is a @xmath5 in @xmath12 in which @xmath40 is not a branch vertex , and @xmath70 holds .
so we may assume that @xmath374 is from @xmath324 to @xmath368 .
then let @xmath384 be the path in @xmath385 between @xmath275 and @xmath324 , and @xmath363 be obtained from @xmath364 by replacing @xmath375 with @xmath384 .
let @xmath379 be a path in @xmath386 $ ] from @xmath40 to @xmath103 , and @xmath381 be a path in @xmath387 $ ] from @xmath109 to @xmath104 .
now @xmath388 is a @xmath5 in @xmath12 in which @xmath40 is not a branch vertex , and @xmath70 holds . 1.2 .
@xmath349 is @xmath224-connected .
let @xmath389 denote the outer cycle of @xmath349 , and let @xmath390 such that @xmath391 , @xmath392 , @xmath393 , @xmath394 , @xmath395 , @xmath396 , @xmath397 , @xmath398 occur on @xmath389 in counterclockwise order , with @xmath399 maximal for @xmath56 $ ] .
recall that @xmath400 .
let @xmath401 be distinct .
suppose @xmath402 may be chosen such that @xmath153 contains disjoint paths @xmath403 from @xmath404 to @xmath405 ( or @xmath406 ) , respectively , and disjoint from @xmath407 ( or @xmath408 ) . by symmetry
, we may assume that @xmath409 are paths in @xmath153 from @xmath404 to @xmath405 , respectively , and disjoint from @xmath407 .
let @xmath363 be obtained from @xmath364 by replacing @xmath410 with the path in @xmath411 from @xmath275 to @xmath324 and replacing @xmath412 with the path in @xmath413 from @xmath275 to @xmath330 .
let @xmath102 be a path in @xmath367 from @xmath103 to @xmath368 .
now @xmath414 is a @xmath5 in @xmath12 in which @xmath40 is not a branch vertex , and @xmath70 holds .
so we may assume that no such @xmath402 can be chosen .
then , without loss of generality , we may assume that @xmath153 has disjoint paths @xmath403 from @xmath404 to @xmath346 , respectively , and disjoint from @xmath415 .
suppose @xmath40 has a neighbor on @xmath416 or @xmath417 . by symmetry
, we may assume the former .
then @xmath418 has disjoint paths @xmath419 from @xmath420 to @xmath421 , respectively .
we modify @xmath364 to obtain @xmath363 by replacing @xmath412 with the path in @xmath422 from @xmath275 to @xmath324 .
then @xmath423 is a @xmath5 in @xmath12 in which @xmath40 is not a branch vertex , and @xmath70 holds .
thus , we may assume that @xmath424 . since @xmath311 is not planar , @xmath40 has a neighbor in @xmath425 , say @xmath426 .
since @xmath12 is @xmath10-connected , @xmath28 has no @xmath224-cut @xmath427 separating @xmath426 from @xmath109 .
so @xmath428 has a path @xmath102 from @xmath426 to @xmath109 .
let @xmath363 be obtained from @xmath364 by replacing @xmath412 with the path in @xmath422 from @xmath275 to @xmath324 .
then @xmath429 is a @xmath5 in @xmath12 in which @xmath40 is not a branch vertex , and @xmath70 holds . _
@xmath430 .
recall the notation in @xmath431 .
if @xmath432 contains disjoint paths @xmath344 from @xmath433 to @xmath434 , respectively , then @xmath347 is a @xmath5 in @xmath12 in which @xmath40 is not a branch vertex ( so @xmath70 holds ) .
hence , we may assume that such @xmath344 do not exist .
then , since @xmath12 is 5-connected , it follows from lemma [ 2path ] that @xmath435 is plane and , hence , @xmath349 is connected .
we may assume that @xmath436 occur on the outer walk of @xmath432 in clockwise order .
recall that @xmath343 is a path for @xmath56 $ ] .
so let @xmath437 and @xmath438 for @xmath439 .
moreover , we may assume that @xmath324 is the only neighbor of @xmath40 on @xmath440 ( otherwise we can shorten @xmath410 ) .
let @xmath441 such that @xmath442 and @xmath443 and , subject to this , @xmath444 is maximal .
suppose there exists @xmath445 .
let @xmath377 be the path in @xmath446 from @xmath275 to @xmath148 , and let @xmath363 be obtained from @xmath364 by replacing @xmath376 with @xmath377 .
let @xmath374 be a path in @xmath447 from @xmath104 to @xmath368 .
then @xmath448 is a @xmath5 in @xmath12 in which @xmath40 is not a branch vertex , and @xmath70 holds .
so we may assume @xmath449 .
however , since @xmath312 is not planar , @xmath40 must have a neighbor in @xmath450 .
hence , by the maximality of @xmath444 , one of the following holds : * @xmath451 occur on @xmath153 in clockwise order and there exists @xmath452 ; or * @xmath453 occur on @xmath153 in clockwise order and there exists @xmath454 .
we consider two cases according to whether or not @xmath349 is 2-connected .
_ subcase _ 2.1 .
@xmath349 is @xmath224-connected .
let @xmath389 denote the outer cycle of @xmath349 and @xmath455 be neighbors of @xmath456 on @xmath389 , respectively .
we may choose @xmath455 so that they are pairwise distinct and occur on @xmath389 in counterclockwise order .
if @xmath457 holds , let @xmath458 , @xmath459 , and @xmath460 be obtained from @xmath364 by replacing @xmath410 with @xmath461 and replacing @xmath375 with @xmath462 .
then @xmath463 is a @xmath5 in @xmath12 in which @xmath40 is not a branch vertex , and @xmath70 holds .
if @xmath464 holds , then let @xmath465 , @xmath466 , and @xmath467 be obtained from @xmath364 by replacing @xmath410 with @xmath468 and replacing @xmath412 with @xmath469 .
then @xmath470 is a @xmath5 in @xmath12 in which @xmath40 is not a branch vertex , and @xmath70 holds . _
subcase _ 2.2 .
@xmath349 is not @xmath224-connected . if @xmath350 then let @xmath245 denote the two disjoint 1-vertex subgraphs of @xmath349 . if @xmath471 then let @xmath351 be a 1-separation of @xmath349 .
since @xmath28 is @xmath223-connected and @xmath472 is planar , we may assume without loss of generality that @xmath473 and @xmath474 , or @xmath475 and @xmath476 .
moreover , @xmath354 is connected for @xmath45 $ ] .
suppose @xmath473 and @xmath474 .
let @xmath379 be a path in @xmath477\cup a_1da_2\cup p_2 $ ] from @xmath275 to @xmath40 , and @xmath363 be obtained from @xmath364 by replacing @xmath376 with @xmath379 .
let @xmath381 be a path in @xmath478 $ ] from @xmath104 to @xmath368 .
then @xmath479 is a @xmath5 in @xmath12 in which @xmath40 is not a branch vertex , and @xmath70 holds .
thus , we may assume that @xmath475 and @xmath476 .
if @xmath457 holds , let @xmath379 be a path in @xmath480 $ ] from @xmath40 to @xmath104 , @xmath381 be a path in @xmath481 $ ] from @xmath103 to @xmath368 , and @xmath460 be obtained from @xmath364 by replacing @xmath410 with @xmath461 and replacing @xmath375 with @xmath462 .
then @xmath463 is a @xmath5 in @xmath12 in which @xmath40 is not a branch vertex , and @xmath70 holds .
so assume that @xmath464 holds .
let @xmath379 be a path in @xmath482 $ ] from @xmath104 to @xmath103 , @xmath381 be a path in @xmath483 $ ] from @xmath40 to @xmath368 , and @xmath467 be obtained from @xmath364 by replacing @xmath410 with @xmath468 and replacing @xmath412 with @xmath469 .
then @xmath470 is a @xmath5 in @xmath12 in which @xmath40 is not a branch vertex , and @xmath70 holds .
in this section , we consider apex graphs . the purpose is to give an indication how case ( c ) in section 1 will be taken care of later ( by combining with later results on @xmath1 ) .
first , we give a consequence of lemma [ planarside ] .
[ contraction ] let @xmath12 be _ a _ @xmath10-connected nonplanar graph , @xmath38 , @xmath484 such that @xmath485 , @xmath20 or @xmath21 , @xmath486 is @xmath10-connected and planar .
then @xmath487 contains @xmath1 .
let @xmath15 denote the vertex of @xmath486 representing the contraction of @xmath135 .
choose @xmath488 such that @xmath180 and @xmath489 .
since @xmath486 is 5-connected and planar , @xmath490 . hence , by lemma [ planarside ] , @xmath491 contains @xmath1 .
so @xmath487 contains @xmath1 .
next we give a result which strengthens corollary 2.9 in @xcite .
[ apexvertex ] let @xmath12 be _ a _ @xmath10-connected nonplanar graph and @xmath38 such that @xmath39 is planar .
then one of the following holds .
* @xmath12 contains _ a _
@xmath5 in which @xmath40 is not _ a _ branch vertex .
* @xmath39 contains @xmath1 .
* @xmath12 has _ a _
@xmath10-separation @xmath43 such that @xmath492 and @xmath29 is the graph obtained from the edge - disjoint union of the @xmath51-cycle @xmath493 and the @xmath53-cycle @xmath54 by adding @xmath40 and the edges @xmath55 for @xmath56 $ ] .
we may assume that @xmath39 is embedded in the plane with no edge crossings .
we may further assume that @xmath250 ; as otherwise @xmath71 holds . let @xmath494 .
we now show @xmath495 by applying to @xmath39 the same discharging argument used in section 2 .
let @xmath190 such that @xmath191 for all @xmath192
. then by lemma [ discharge ] , the total charge is @xmath193 . note that for any @xmath192 , if @xmath194 then @xmath196 , @xmath197 , and @xmath198 ( i.e. , @xmath496 is a triangle in @xmath39 ) . for each @xmath199 with @xmath200 , pick two of its incident vertices , and send a charge of @xmath168 from @xmath161 to each of these two vertices .
let @xmath165 denote the resulting charge function .
then @xmath202 for all @xmath199 .
since we assume @xmath170 , for @xmath204 , @xmath176 satisfies the upper bound in lemma [ discharge ] .
hence , @xmath497 for @xmath498 , and @xmath207 for @xmath499 .
therefore , @xmath500 as @xmath501 .
let @xmath502 be arbitrary , let @xmath503 , and let @xmath504 denote the facial cycle of @xmath505 containing @xmath506 .
note that @xmath507 is @xmath508-connected and @xmath509 is planar .
suppose there exists @xmath510 .
let @xmath511)$ ] .
we wish to apply lemma [ apex1 ] to @xmath98 and @xmath512 . since @xmath495 , @xmath513 .
suppose @xmath98 has a @xmath10-separation @xmath514 with @xmath515 and @xmath516 .
then @xmath12 has a separation @xmath25 such that @xmath517 , @xmath518 and @xmath519 .
note that @xmath520 as @xmath521 is independent in @xmath98 .
so @xmath86 . if @xmath46 then we may apply theorem [ apexside1 ] to @xmath12 to conclude that @xmath70 or @xmath71 or @xmath73 holds . if @xmath522 then @xmath37 is planar ; so @xmath71 holds by lemma [ planarside ] .
hence , we may assume that no such 5-separation @xmath48 exists in @xmath98 .
then by lemma [ apex1 ] , the vertices of @xmath523 cofacial with @xmath275 induce a cycle @xmath272 ( in @xmath523 ) and @xmath523 contains three paths @xmath314 from @xmath275 to @xmath506 such that @xmath276 for @xmath524 , and @xmath525 for @xmath289 $ ] . without loss of generality , assume that @xmath526 for @xmath289 $ ] .
now @xmath527 is a @xmath5 in @xmath12 in which @xmath40 is not a branch vertex , and @xmath70 holds .
so we may assume that @xmath528 .
without loss of generality , we may assume that @xmath529 , @xmath530 , @xmath531 , @xmath532 occur on @xmath504 in clockwise order .
since @xmath500 , we may further assume that there exists @xmath533 .
let @xmath534 with @xmath535 and @xmath536 .
let @xmath537 be the facial cycle of @xmath538 containing @xmath539 .
since @xmath502 is chosen arbitrary , we also have @xmath540 .
thus , @xmath541 ( since @xmath12 is @xmath10-connected ) .
however , this implies that @xmath12 is planar , a contradiction .
in this section , we prove theorem [ 5cut_triangle ] .
we fix the notation @xmath17 to represent the tree on six vertices , two of which are adjacent and of degree 3 .
we need a result on @xmath542 in the proof of theorem [ 5cut_triangle ] ( and also in the proof of theorem [ n(x1)capa ] in section 6 ) .
let @xmath12 be a graph and @xmath543 be distinct vertices of @xmath12 .
we say that a @xmath542 in @xmath12 is _ rooted _ at @xmath544 if in the @xmath542 , @xmath545 are of degree 3 and @xmath96 are of degree 1 .
for convenience , we use _ quadruple _ to denote @xmath546 where @xmath545 are distinct vertices of the graph @xmath12 , @xmath547 , and @xmath548 .
we say that @xmath546 is _ feasible _ if @xmath12 has a @xmath542 rooted at @xmath549 . in @xcite ,
infeasible quadruples are characterized in terms of obstructions . here
we include the descriptions and figures of obstructions given in @xcite .
a quadruple @xmath550 is an _ obstruction _
if @xmath12 has subgraphs @xmath551 ( called _ sides _ ) and @xmath552 ( called _ middle parts _ ) , where @xmath553 , such that * @xmath554}$ ] , where @xmath555}=\cup_{i\in [ k ] } v(a_i)$ ] , * @xmath556,$ ] are pairwise disjoint , * @xmath557 is the disjoint union of @xmath558 , and @xmath559 for @xmath560 $ ] , * @xmath561}$ ] , @xmath562}$ ] , and @xmath563}$ ] , * for any @xmath560 , v(a_i)\cap a\neq \emptyset$ ] and @xmath564 , * if @xmath565 , then @xmath566 and @xmath567 .
note that @xmath568 iff @xmath569 , in which case @xmath570 and there is no restriction on @xmath571 .
obstructions of type i and type ii ] an obstruction @xmath550 is said to be of _ type i _ if @xmath572 , @xmath573 for @xmath45 $ ] , @xmath574 , @xmath575 for @xmath576 , and @xmath577 .
an obstruction @xmath550 is said to be of _ type ii _ if @xmath578 , @xmath579 , @xmath580 , @xmath581 for @xmath45 $ ] , and @xmath582 for @xmath45 $ ] .
obstructions of types
iii and iv ] an obstruction @xmath550 is said to be of _ type iii _ if @xmath578 , @xmath583 for @xmath45 $ ] , @xmath584 , and @xmath585 .
an obstruction @xmath550 is said to be of _ type iv _ if @xmath586 and , for @xmath56 $ ] and @xmath587 $ ] , @xmath588 .
the following result is proved by ma , xie and yu in @xcite , which gives a characterization of feasible quadruples .
[ h ] let @xmath546 be _ a _ quadruple .
then one of the following holds .
* @xmath546 is feasible .
* @xmath12 has _ a _ separation @xmath111 such that @xmath589 and for some @xmath45 $ ] , @xmath590 and @xmath591 .
* @xmath12 has _ a _ separation @xmath111 such that @xmath592 , @xmath593 , and @xmath594 .
* @xmath546 is an obstruction of type i , or type ii , or type iii , or type iv .
it is easy to verify that if @xmath12 is @xmath595-connected and has no separation as in @xmath73 of lemma [ h ] , then @xmath546 is not feasible iff it is an obstruction of type iv .
moreover , in the case when @xmath12 is @xmath595-connected and @xmath546 is an obstruction of type iv , it is straightforward to verify that for any @xmath45 $ ] and @xmath183 , @xmath12 contains four independent paths : one from @xmath596 to @xmath40 , and three from @xmath597 to the three vertices in @xmath213 , respectively .
_ proof of theorem [ 5cut_triangle]_. let @xmath12 be a 5-connected graph and @xmath25 be a 5-separation in @xmath12 .
let @xmath598 such that @xmath599\cong k_3 $ ] .
we wish to prove @xmath72 ; so let @xmath68 be pairwise distinct and @xmath600 .
without loss of generality , we may assume that @xmath601 .
_ claim _ 1 .
we may assume @xmath602 .
since @xmath12 is 5-connected , @xmath310 has independent paths @xmath603 from @xmath89 to @xmath96 , respectively .
first , suppose @xmath604 ; the case when @xmath605 can be treated in the same way .
since @xmath12 is @xmath10-connected , @xmath94 has independent paths @xmath95 from @xmath90 to @xmath96 , respectively .
then @xmath599 \cup u_1a \cup p_1 \cup p_2 \cup u_2a \cup q_1 \cup q_2 \cup ( p_3\cup q_3)$ ] is a @xmath5 in @xmath98 with branch vertices @xmath606 .
next , assume that @xmath607 ; the case when @xmath608 is the same . without loss of generality
, we may assume that @xmath609 .
let @xmath610 .
since @xmath12 is 5-connected , @xmath94 has independent paths @xmath95 from @xmath426 to @xmath96 , respectively .
now @xmath599 \cup u_1a \cup p_1\cup p_2 \cup q_1\cup q_2 \cup ( q_3\cup a_3a)\cup ( p_4\cup q_4)$ ] is a @xmath5 in @xmath98 with branch vertices @xmath611 .
thus we have claim 1 . before proceeding with the proof ,
we introduce additional notation . if there is a 4-separation @xmath111 in @xmath310 such that @xmath593 and @xmath612 , we choose such a separation that @xmath2 is minimal . if such a 4-separation does not exist , we let @xmath613 , @xmath614 and @xmath615 .
let @xmath616 .
since @xmath12 is 5-connected , we may assume that @xmath364 contains disjoint paths @xmath617 from @xmath96 to @xmath618 , respectively ( by relabeling if necessary ) .
_ claim _ 2 .
we may assume that @xmath619 is not feasible . for
, suppose @xmath2 has a @xmath542 rooted at @xmath620 , say @xmath135 .
first , assume that there exists @xmath45 $ ] such that @xmath621 has independent paths from @xmath596 to @xmath622 , respectively .
if @xmath94 has two disjoint paths @xmath623 from @xmath624 to @xmath625 then @xmath626 \cup t \cup y_1\cup y_2\cup y_3\cup y_4\cup p\cup q$ ] is a @xmath5 in @xmath98 with branch vertices @xmath627 .
so assume such paths @xmath623 do not exist in @xmath94 .
then @xmath28 has a separation @xmath48 such that @xmath628 , @xmath629 , @xmath630 , and @xmath631 .
since @xmath12 is 5-connected , @xmath632 for @xmath45 $ ] .
this forces @xmath633 , a contradiction .
thus , we may assume without loss of generality that @xmath634 has independent paths from @xmath89 to @xmath635 , respectively . if @xmath94 has disjoint paths @xmath636 from @xmath637 to @xmath638 , respectively , then @xmath626 \cup t \cup y_1\cup
y_2\cup y_3\cup y_4\cup p_1 \cup p_2 $ ] is a @xmath5 in @xmath98 , with branch vertices @xmath627 .
hence , we may assume that such @xmath639 do not exist .
then by lemma [ 2path ] , @xmath640 is planar . since @xmath35 for @xmath45 $ ] , it follows from theorem [ apexside1 ] that @xmath70 , or @xmath71 , or @xmath73 holds , completing the proof of claim 2 .
since @xmath12 is 5-connected , it follows from the choice of @xmath111 , claim 2 and lemma [ h ] that @xmath641 is an obstruction of type iv .
thus , @xmath2 has four independent paths @xmath642 , with @xmath643 from @xmath89 to @xmath644 , respectively , and @xmath645 from @xmath90 to @xmath646 .
let @xmath647 .
since @xmath12 is 5-connected , @xmath94 contains independent paths @xmath95 from @xmath426 to @xmath96 , respectively .
therefore , @xmath648 \cup ( r_1\cup y_1)\cup ( r_2\cup y_2)\cup ( r_3\cup y_3\cup q_3)\cup q_1\cup q_2\cup ( au_2\cup r_4\cup y_4\cup q_4)$ ] is a @xmath5 in @xmath98 with branch vertices @xmath649 .
hence , @xmath72 holds .
in this section , we prove theorem [ n(x1)capa ] . we need the following result of perfect @xcite .
[ perfect ] let @xmath12 be _ a _ graph , @xmath650 , and @xmath651 .
suppose there exist @xmath31 independent paths from @xmath152 to distinct @xmath652 , respectively , and otherwise disjoint from @xmath34 .
then for any @xmath653 , if there exist @xmath41 independent paths @xmath654 in @xmath12 from @xmath152 to @xmath41 distinct vertices in @xmath34 and otherwise disjoint from @xmath34 then @xmath654 may be chosen so that @xmath298 for @xmath560 $ ] .
* there exists _ a _
2-cut @xmath59 in @xmath102 and there exist pairwise disjoint subgraphs @xmath657 , @xmath289 $ ] , such that @xmath658 and each @xmath659 is _ a _ union of components of @xmath660 .
* there exist 2-cuts @xmath661 of @xmath102 and pairwise disjoint subgraphs @xmath659 of @xmath102 , @xmath289 $ ] , such that @xmath662 , each @xmath659 is _ a _ union of components of @xmath663 , @xmath664 , and @xmath665 are pairwise disjoint .
* there exist pairwise disjoint @xmath224-cuts @xmath661 in @xmath102 and pairwise disjoint subgraphs @xmath659 of @xmath102 such that @xmath666 , @xmath659 is _ a _ union of components of @xmath667 , and @xmath668 has precisely two components , each containing exactly one vertex from @xmath661 . for
, suppose @xmath676 is a component of @xmath677 such that @xmath678 for some @xmath289 $ ] .
if @xmath679 then @xmath12 contains @xmath1 ( as @xmath12 is 5-connected ) .
so assume @xmath680 .
let @xmath681 be obtained from @xmath676 by adding @xmath682 and all edges of @xmath12 from @xmath676 to @xmath682 .
now we may apply theorem [ 5cut_triangle ] to the separation @xmath683 .
if @xmath70 or @xmath71 of theorem [ 5cut_triangle ] holds then we have @xmath70 or @xmath71 .
if @xmath73 of theorem [ 5cut_triangle ] holds then @xmath12 contains @xmath1 ; so @xmath71 holds .
now assume @xmath72 of theorem [ 5cut_triangle ] holds
. then clearly , @xmath73 holds .
so we may assume ( 4 ) . if @xmath688 , then let @xmath689 .
since @xmath12 is 5-connected , @xmath690 has independent paths @xmath691 from @xmath426 to @xmath687 , respectively . now @xmath692 \cup p_1 \cup p_2 \cup ( p_3\cup v_1x ) \cup
q_1 \cup q_2 \cup ( q_4\cup v_2x ) \cup ( p_5\cup q_5)$ ] is @xmath5 in @xmath98 with branch vertices @xmath693 .
thus @xmath71 holds with @xmath694 , and @xmath695 ( as @xmath696 in this case ) .
so we may assume that first , suppose @xmath706 for @xmath45 $ ] .
then @xmath707 $ ] contains a path @xmath83 from @xmath708 to @xmath709 , and a path from @xmath705 to @xmath710 and internally disjoint from @xmath83 ; so @xmath707 $ ] has independent paths from @xmath426 to @xmath711 , respectively . by lemma [ perfect ] ,
@xmath712 $ ] has independent paths @xmath95 from @xmath426 to @xmath713 , respectively , and internally disjoint from @xmath714 . by symmetry between @xmath715 and @xmath716 ,
we may assume @xmath717 .
let @xmath718 if @xmath719 , and @xmath720 if @xmath721 .
now @xmath692 \cup p_1 \cup p_2 \cup ( p_3\cup v_1x ) \cup q \cup q_1 \cup q_2 \cup ( q_3\cup x_3x ) $ ] is a @xmath5 in @xmath98 with branch vertices @xmath722 . hence , we may assume by symmetry that @xmath723 .
since @xmath12 is 5-connected , @xmath724 $ ] has independent paths @xmath95 from @xmath705 to @xmath725 , respectively .
since @xmath726 , @xmath677 has a component @xmath727 such that @xmath728 .
then by ( 4 ) , @xmath729 $ ] contains a path @xmath374 from @xmath708 to @xmath730 .
so @xmath692 \cup p_1 \cup p_2 \cup ( p_3\cup v_1x ) \cup ( p_4\cup q_3)\cup q_1 \cup ( q_4\cup q)\cup x_3x$ ] is a @xmath5 in @xmath98 with branch vertices @xmath731 . _ claim _ 2 .
we may assume that for any @xmath289 $ ] , there do not exist @xmath735 and two disjoint paths @xmath736 in @xmath677 from @xmath737 to @xmath545 , respectively , such that @xmath738 has a @xmath542 rooted at @xmath739 .
suppose otherwise and , without loss of generality , assume that there exist @xmath735 , paths @xmath736 in @xmath677 , and a @xmath542 in @xmath738 rooted at @xmath740 .
let @xmath2 denote the union of @xmath736 and that @xmath542 .
we may assume that @xmath741 contains disjoint paths @xmath742 from @xmath743 to @xmath744 , respectively , and disjoint paths @xmath745 from @xmath743 to @xmath746 , respectively . for ,
otherwise , @xmath747 is planar by lemma [ 2path ] and , hence by ( 1 ) , @xmath70 or @xmath71 follows from theorem [ apexside1 ] ( as @xmath73 of theorem [ apexside1 ] implies that @xmath12 contains @xmath1 ) .
we now set up some notation .
let @xmath750 . if there exists a separation @xmath111 in @xmath50 such that @xmath751 , @xmath752 , and @xmath753 , we choose such @xmath111 that @xmath754 whenever possible and , subject to this , @xmath2 is minimal .
if such a separation does not exist we let @xmath755 , @xmath756 and @xmath615 .
note that if @xmath757 then @xmath758 ; for , otherwise , @xmath755 by ( 4 ) , a contradiction .
since @xmath759 , if @xmath760 then @xmath761 .
let @xmath762 .
_ claim _ 3 . if @xmath763 then @xmath364 has disjoint paths @xmath764 from @xmath765 , respectively , to @xmath766 ; and if @xmath754 then @xmath364 has disjoint paths @xmath764 from @xmath765 , respectively , to @xmath766 , or @xmath767 , or @xmath768 , and internally disjoint from @xmath714 .
suppose @xmath364 does not contain four disjoint paths from @xmath769 to @xmath766
. then @xmath364 has a separation @xmath770 such that @xmath771 , @xmath772 and @xmath773 .
moreover , @xmath774 and @xmath775 , as otherwise @xmath776 would be a cut in @xmath12 of size at most 4 . hence , since @xmath759 , we have @xmath777 .
first , assume @xmath778 .
then @xmath779 ; but then @xmath780 is a cut in @xmath12 of size 4 and separating @xmath781 from @xmath782 , contradicting the choice of @xmath111 ( that @xmath754 whenever possible ) .
so @xmath783 . let @xmath784 .
since @xmath12 is 5-connected , @xmath785 has four disjoint paths @xmath786 from @xmath765 , respectively , to @xmath787 .
if @xmath788 has disjoint paths @xmath742 from @xmath789 to @xmath130 , then @xmath786 and @xmath742 form four disjoint paths in @xmath364 from @xmath769 to @xmath768 .
so assume that such @xmath742 do not exist .
then @xmath790 has a separation @xmath791 such that @xmath792 , @xmath793 , @xmath794 and @xmath795 . since @xmath796 for @xmath45 $ ] , @xmath797 .
if @xmath798 then @xmath799 $ ] contains @xmath1 , and @xmath71 holds .
so we may assume @xmath800 .
then @xmath801 is a @xmath10-cut in @xmath12 .
hence , @xmath12 has a 5-separation @xmath802 such that @xmath803 , @xmath804 , @xmath805 .
clearly , @xmath806 and @xmath807 .
if @xmath808 then @xmath12 contains @xmath1 ( as @xmath12 is 5-connected ) ; so @xmath71 holds .
if @xmath809 then the assertion follows from theorem [ 5cut_triangle ] .
this completes the proof of claim 3 .
note that a @xmath542 in @xmath2 rooted at @xmath810 and the paths in claim 3 would give a contradiction to claim 2 .
hence , @xmath811 is not feasible .
so by the minimality of @xmath2 and the 5-connectedness of @xmath12 , it follows from lemma [ h ] that @xmath811 is an obstruction of type i , or type ii , or type iii , or type iv .
let @xmath812}$ ] be defined as in section 4 with @xmath769 as @xmath34 .
so assume @xmath811 is an obstruction of type i. then @xmath815}$ ] , as otherwise @xmath816})\cup \{x\}$ ] would be a 4-cut in @xmath12 .
since @xmath759 , we may assume @xmath817 and @xmath818 .
without loss of generality , assume @xmath819 and @xmath820 .
let @xmath821 and @xmath822 .
we may assume that @xmath823 contains two disjoint paths @xmath742 from @xmath824 to @xmath825 .
for , otherwise , @xmath823 has a separation @xmath826 such that @xmath827 , @xmath828 , and @xmath829 . then @xmath830 is a cut in @xmath12 of size at most 4 , a contradiction .
since @xmath12 is 5-connected , @xmath831 has independent paths @xmath832 from @xmath89 to @xmath833 , respectively , and @xmath834 has four independent paths @xmath835 from @xmath90 to @xmath836 , respectively . if both @xmath742 end in @xmath837 then @xmath838 is a @xmath542 in @xmath690 rooted at @xmath839 , contradicting claim 2 .
so @xmath840 and @xmath841 .
then @xmath842 is a @xmath542 in @xmath690 rooted at @xmath839 , contradicting claim 2 and completing the proof of claim 4 .
then @xmath846 , and without loss of generality , assume @xmath847 and @xmath848 .
since @xmath759 , we may assume by symmetry that @xmath849 . then @xmath850 and @xmath851 .
since @xmath12 is 5-connected , @xmath831 has independent paths @xmath642 from @xmath89 to @xmath852 , respectively .
let @xmath853 be a path in @xmath854 from @xmath855 to @xmath262 , and @xmath856 be a path in @xmath857 from @xmath265 to @xmath264 .
suppose @xmath858 .
then @xmath859 .
we may assume that @xmath823 has disjoint paths @xmath860 from @xmath861 , respectively , to @xmath862 .
for otherwise , @xmath823 has a separation @xmath826 such that @xmath827 , @xmath863 and @xmath864 . then @xmath865 is a cut in @xmath2 separating @xmath781 from @xmath769 , contradicting the minimality of @xmath2 .
since @xmath12 is 5-connected , @xmath834 has independent paths @xmath866 from @xmath90 to @xmath867 , respectively .
then @xmath868 is a @xmath542 in @xmath2 rooted at @xmath869 .
this @xmath542 and the paths @xmath870 in claim 3 ( from @xmath871 to @xmath130 ) form a @xmath542 in @xmath690 rooted at @xmath872 , contradicting claim 2 .
thus , @xmath873 . then @xmath874 and @xmath875 .
suppose @xmath876 contains a path from @xmath90 to @xmath730 .
since @xmath12 is 5-connected , @xmath834 has four independent paths from @xmath90 to @xmath877 with only @xmath90 in common .
so by lemma [ perfect ] , @xmath834 contains three independent paths @xmath878 from @xmath90 to @xmath879 , respectively , such that @xmath880 , @xmath881 , and disjoint from @xmath882 . without loss of generality
, we may assume @xmath883 and @xmath884 .
then @xmath885 is a @xmath542 in @xmath2 rooted at @xmath886 .
this @xmath542 and the paths @xmath870 in claim 3 ( from @xmath871 to @xmath130 ) form a @xmath542 in @xmath690 rooted at @xmath839 , contradicting claim 2 .
hence , @xmath876 contains no path from @xmath90 to @xmath730 .
then @xmath834 has a separation @xmath887 such that @xmath888 , @xmath889 and @xmath890 . by the minimality of @xmath2
, @xmath891 must have a path from @xmath730 to @xmath90 .
thus we may assume that @xmath892 and @xmath893-\{t_2,t_3,t_4\}$ ] contains a path @xmath374 from @xmath730 to @xmath263 .
let @xmath894 be independent paths in @xmath895 from @xmath90 to @xmath867 , respectively .
then @xmath896 is a @xmath542 in @xmath2 rooted at @xmath897 .
this @xmath542 and the paths @xmath870 in claim 3 ( from @xmath871 to @xmath130 ) give a @xmath542 in @xmath690 rooted at @xmath839 , contradicting claim 2 .
since @xmath12 is 5-connected , @xmath569 or @xmath899 ; as otherwise @xmath900 would be a cut in @xmath12 of size 4 .
moreover , by the choice of @xmath111 ( the minimality of @xmath2 ) , if @xmath899 then @xmath901 is a path in @xmath12 .
thus , since @xmath12 is 5-connected , @xmath2 contains eight independent paths @xmath642 from @xmath89 to @xmath765 , respectively , and @xmath95 from @xmath902 to @xmath765 , respectively .
let @xmath764 be the paths in claim 3 with @xmath903 for @xmath56 $ ] .
suppose for some @xmath56 $ ] and @xmath587 $ ] , @xmath904 has at least two neighbors in @xmath905 . without loss of generality ,
assume @xmath906 . then , @xmath907 .
since @xmath12 is 5-connected , @xmath908 has a path from @xmath262 to @xmath89 .
moreover , @xmath831 has two independent paths from @xmath262 to @xmath909 with only @xmath262 in common . for
, otherwise , @xmath831 has a separation @xmath910 such that @xmath911 , @xmath912 , and @xmath913 .
now @xmath914 , as @xmath262 has at least two neighbors in @xmath831 .
hence , @xmath915 is a cut in @xmath12 of size at most 3 , a contradiction .
so by lemma [ perfect ] , @xmath831 has two independent paths @xmath916 from @xmath262 to @xmath917 , respectively , with @xmath918 , and internally disjoint from @xmath919 . without loss of generality , we may assume @xmath883 .
now @xmath920 form a @xmath542 in @xmath921 rooted at @xmath922 , or @xmath923 , or @xmath924 , however , this contradicts claim 2 .
suppose there exists some @xmath928 , say @xmath929 .
let @xmath496 be the end of @xmath930 other than @xmath262 . by the similar argument as above
, @xmath364 has two independent paths from @xmath262 to @xmath931 .
then by lemma [ perfect ] , there exist two independent paths @xmath932 from @xmath262 to @xmath933 , respectively , with @xmath934 and internally disjoint from @xmath935 . if @xmath936 then @xmath937 is a @xmath542 in @xmath938 rooted at @xmath939 , or @xmath940 , or @xmath941 , contradicting claim 2 .
hence , without loss of generality , we may assume @xmath942 and let @xmath943 be the end of @xmath944 other than @xmath264 .
now @xmath945 is a @xmath542 in @xmath938 rooted at @xmath939 , or @xmath946 , or @xmath941 .
this contradicts claim 2 . hence , @xmath947 . by ( 4 ) and by symmetry between @xmath708 ad @xmath709
, we may assume that @xmath948 for @xmath289 $ ] and @xmath949 .
since @xmath726 , @xmath950 .
thus , by ( 4 ) , @xmath951 $ ] has a path @xmath135 from @xmath952 to @xmath708 . if @xmath953 has two independent paths @xmath954 from @xmath730 to @xmath743 , respectively then @xmath955\cup t_1\cup t_2\cup ( q_3\cup u_2x)\cup ( r_1\cup t)\cup r_3\cup r_4 $ ] is a @xmath5 in @xmath98 with branch vertices @xmath956 .
so assume such @xmath954 do not exist .
then @xmath129 has a separation @xmath957 such that @xmath958 , @xmath959 , @xmath960 and @xmath961 .
since @xmath962 by ( 3 ) , @xmath963 is a cut in @xmath12 of size at most 4 , a contradiction . for
, otherwise , we may assume @xmath703 for @xmath45 $ ] .
since @xmath12 is 5-connected , @xmath966 contains independent paths @xmath95 from @xmath705 to @xmath967 , respectively .
if @xmath968 has a cycle @xmath969 containing @xmath970 , then @xmath971 \cup c\cup ( q_4\cup x_2x_1)\cup q_1\cup q_2 $ ] is a @xmath5 in @xmath98 with branch vertices @xmath972 .
so we may assume that such a cycle @xmath969 does not exist .
we wish to apply lemma [ watkins ] .
first , we show that @xmath968 is 2-connected .
note that @xmath973 is connected by the minimality of @xmath28 .
suppose @xmath968 has a cut vertex , say @xmath148 .
then @xmath974 , and @xmath968 has a separation @xmath975 such that @xmath976 , @xmath977 and @xmath978 . but
this implies that @xmath979 or @xmath980 ( as @xmath12 is @xmath10-connected and @xmath685 ) . if @xmath979 , then any vertex in @xmath981 has at most one neighbor in @xmath973 , contradicting ( 3 ) .
so @xmath982 .
this implies @xmath983 ( as @xmath12 is @xmath10-connected ) .
but now @xmath984 , which is a contradiction .
so by lemma [ watkins ] , @xmath968 has a 2-cut @xmath985 separating @xmath708 from @xmath986 .
let @xmath987 denote the component of @xmath988 containing @xmath708 .
then @xmath989 ; otherwise @xmath990 is a @xmath11-cut in @xmath12 such that @xmath65 has a component strictly contained in @xmath973 , contradicting the choice of @xmath25 ( that @xmath28 is minimal ) .
suppose @xmath991 , and let @xmath992 .
since @xmath12 is 5-connected and @xmath685 , @xmath993 .
so @xmath994\cong k_4 ^ -$ ] , and @xmath71 holds .
therefore , we may assume @xmath995 .
since @xmath965 , @xmath996 . hence , @xmath997 . since @xmath998
, we have @xmath999 .
therefore , @xmath1000 $ ] contains @xmath1 , and @xmath71 holds .
first , we set up some notation . if @xmath966 has a separation @xmath111 such that @xmath1005 , @xmath593 and @xmath1006 we select @xmath111 so that @xmath2 is minimal ; in this case , since @xmath12 is 5-connected , @xmath1007 , and we let @xmath1008 .
if such a separation @xmath111 does not exist in @xmath966 , we let @xmath1009 , @xmath1010 , and @xmath615 .
since @xmath12 is 5-connected , @xmath364 has four disjoint paths @xmath764 from @xmath765 , respectively , to @xmath1011 .
thus , we may assume that @xmath2 has no @xmath542 rooted at @xmath810 ; otherwise such @xmath542 and @xmath764 form a @xmath542 in @xmath966 rooted at @xmath1012 and , hence , claim 2 holds by letting , for @xmath45 $ ] , @xmath1013 and @xmath1014 be the trivial path consisting of @xmath1015 only .
since @xmath12 is 5-connected and by the choice of @xmath111 and lemma [ h ] , @xmath811 is an obstruction of type iv .
as before , we use the notation @xmath1016 and @xmath1017 from section 4 , with @xmath1018 for @xmath56 $ ] and with @xmath769 as @xmath34 . for @xmath56
$ ] , let @xmath844 and @xmath845 .
since @xmath12 is 5-connected , @xmath569 or @xmath899 , and @xmath901 is a path in @xmath1017 ; so @xmath2 has eight independent paths @xmath1019 from @xmath596 to @xmath765 , respectively .
suppose @xmath1020 . without loss of generality ,
let @xmath948 for @xmath289 $ ] , and @xmath949 . if @xmath1021 $ ] has three independent paths @xmath1022 from @xmath708 to @xmath131 , respectively , then @xmath1023\cup p^1_4\cup p^2_4\cup ( p^1_3\cup p^2_3)\cup ( p_1\cup p^1_1 ) \cup ( p_2\cup p^2_2)$ ] is a @xmath5 in @xmath98 with branch vertices @xmath1024 .
so assume that such paths do not exist .
then @xmath1025 $ ] has a separation @xmath957 such that @xmath1026 , @xmath958 , @xmath1027 and @xmath1028 .
since @xmath965 , @xmath1029 is a cut in @xmath12 of size at most @xmath53 , a contradiction .
thus , we may assume without loss of generality that @xmath1030 .
since @xmath12 is 5-connected , @xmath262 has at least two neighbors in @xmath364 , or at least two neighbors in @xmath1031 for some @xmath45 $ ] .
first , assume that @xmath262 has at least two neighbors in some @xmath1031 , say @xmath831 ( by symmetry ) .
then @xmath1032 and @xmath1033 .
we claim that @xmath831 has two independent paths from @xmath262 to two distinct vertices in @xmath909 . for , otherwise , @xmath831 has a separation @xmath910 such that @xmath911 , @xmath1034 and @xmath913 .
since @xmath262 has at least two neighbors in @xmath831 , we see that @xmath914 .
so @xmath1035 is cut in @xmath12 of size at most 3 , a contradiction .
since @xmath908 has a path from @xmath262 to @xmath89 , it follows from lemma [ perfect ] that @xmath831 contains two independent paths @xmath1036 from @xmath262 to @xmath1037 , respectively , for some @xmath1038 , and disjoint from @xmath1039 . without loss of generality
, we may assume that @xmath1040 ends at @xmath1041 .
now @xmath1042 is a @xmath542 in @xmath1043 rooted at @xmath1044 .
hence claim 2 holds with @xmath1045 , @xmath1046 , @xmath1047 , and @xmath1048 .
so we may assume that @xmath262 has at least two neighbors in @xmath364 .
let @xmath1049 be the end of @xmath930 other than @xmath262 .
since @xmath262 has at least two neighbors in @xmath364 and @xmath12 is 5-connected , @xmath364 has two independent paths @xmath1050 from @xmath262 to @xmath1051 and internally disjoint from @xmath1052 .
thus by lemma [ perfect ] ( and the existence of @xmath930 ) , @xmath364 has independent paths @xmath1050 from @xmath262 to @xmath1053 , respectively , and internally disjoint from @xmath1054 , with @xmath1055 . without loss of generality
, we may further assume that @xmath1056 ends at @xmath1057 and let @xmath496 be the end of @xmath944 in @xmath1011 .
now @xmath1058 is a @xmath542 in @xmath1059 rooted at @xmath1044 .
hence , claim 2 holds with @xmath1045 , @xmath1060 , @xmath1047 , and @xmath1061 .
this completes the proof of claim 2 . by claim 2 ,
let @xmath1062 denote the union of @xmath736 and a @xmath542 in @xmath1003 rooted at @xmath1004 .
without loss of generality , we may assume that @xmath1063 has independent paths from @xmath1064 to @xmath1065 , respectively , and @xmath1066 has independent paths from @xmath1067 to @xmath1068 , respectively .
it remains to show that @xmath72 holds for @xmath906 if we can not find the desired @xmath5 for @xmath73 .
suppose @xmath129 contains three independent paths , with one from @xmath709 to @xmath730 and two from @xmath708 to @xmath744 , respectively , or with one from @xmath709 to @xmath1069 and two from @xmath708 to @xmath1070 , respectively .
it is easy to verify that these paths and @xmath1062 form a @xmath5 in @xmath98 with branch vertices @xmath749 .
so we may assume that any three independent paths from @xmath130 to @xmath131 , respectively , with two from @xmath708 and one from @xmath709 , must contain a path from @xmath709 to @xmath952 .
hence @xmath72 holds . | seymour and , independently , kelmans conjectured in the 1970s that every 5-connected nonplanar graph contains a subdivision of @xmath0 .
this conjecture was proved by ma and yu for graphs containing @xmath1 , and an important step in their proof is to deal with a 5-separation in the graph with a planar side . in order to establish the kelmans - seymour conjecture for all graphs , we need to consider 5-separations and 6-separations with less restrictive structures .
the goal of this paper is to deal with special 5-separations and 6-separations , including those with an apex side .
results will be used in subsequent papers to prove the kelmans - seymour conjecture .
ams subject classification : primary 05c38 , 05c40 , 05c75 ; secondly 05c10 , 05c83 keywords : subdivision , independent paths , separation , planar graph , apex graph |
the precision of 3d stereovision applications is intimately related to the precision of the camera calibration , and especially of the camera distortion correction .
an imprecise distortion model produces residual distortion that will be directly back - projected to the reconstructed 3d scene .
such imprecision can be a serious hindrance in remote sensing applications such as the early warnings of geology disasters , or in the construction of topographic maps from stereographic pairs of aerial photographs .
the fast growing resolution of digital cameras and of their optical quality is transforming them into ( potential ) high precision measurement tools .
thus , it becomes important to measure the calibration precision with ever higher precision . a first step toward high - precision distortion corrections is to perform precise distortion measurements .
this basic tool can then be used to evaluate the precision of a correction method , or can become part of the correction method itself . camera and lens distortion measurement methods usually require a flat pattern containing aligned elements .
the pattern is photographed using the target lens , and the distortion is measured by how much the observed elements deviate from the straight alignment on the pattern .
for example the classic dxo - labs software , a good representative of camera maker practice , ( http://www.dxo.com/ ) uses a pattern with a grid of aligned dots .
distortion is measured by the positional errors associated with the maximal deviation in a row , see fig .
[ fig : dxo_distortion ] . similar methods are proposed by the smia , ebu , ie , and i3a standards .
these measurements are generally manual and require a perfectly flat pattern .
every lens distortion correction method includes , implicitly , a lens distortion measurement .
these methods can be roughly classified into four groups : * re - projection error minimization methods ; * pattern matching methods ; * enlarged epipolar geometry based methods ; * plumb - line methods . re - projection methods usually rely on a planar pattern containing simple geometric shapes . in these methods
, the lens distortion is estimated together with the camera internal and external parameters @xcite , by minimizing the re - projection error between the observed control points on the pattern and the re - projected control points simulated by the pattern model and camera model .
the distortion is measured in terms of re - projection error once all the parameters have been estimated .
both the internal parameter errors and the external parameter errors contribute to the re - projection error .
unfortunately , these errors can compensate each other .
thus , a small re - projection error may be observed while the internal and external parameters are not well estimated .
this compensation effect reduces the precision of the estimation of the lens distortion parameters @xcite .
it hinders accurate distortion estimation from the re - projection error .
notice that this high precision is not always required . in some applications ,
the distortion estimated in these bundle adjustment based re - projection methods , which makes the 3d geometry consistent with the pin - hole model , is precise enough for human observer . a quite different kind of method proceeds by matching a photograph of a flat pattern to its digital model .
these methods estimate the distortion field by interpolating a continuous distortion field from a set of matching points .
several variants exist depending on the kind of pattern , matching and interpolation technique .
a common and crucial assumption for these methods is that the pattern is flat . in practice , however , it is difficult to produce a very flat pattern , and the consequences of a tiny flatness flaw are considerable .
for example it is reported in @xcite that a flatness error of about 100 @xmath1 m for a 40 centimeters broad pattern can lead to an error of about @xmath2 pixels in the distortion field computation for a canon eos 30d camera of focal length @xmath3 mm with the distance between the camera and the object about @xmath4 cm . the only physical method to assess a pattern flatness at a high precision is interferometry , but it requires the pattern to be a mirror , which is not adequate for photography .
furthermore , camera calibration requires large patterns , which are therefore flexible .
deformations of the order of 100 @xmath1 m or more can be caused by temperature changes , and by a mere position change of the pattern , which deforms under its own weight .
recently , more attention has been paid to pattern - free methods ( or self - calibration methods ) where the distortion estimation is obtained without using any specific pattern .
the distortion is estimated from the correspondences between two or several images in absence of any camera information .
the main tool is the so - called enlarged epipolar constraint , which incorporates lens distortion into the epipolar geometry .
some iterative @xcite or non - iterative methods @xcite are used to estimate the distortion and correct it by minimizing an algebraic error .
the estimated distortion can be used as the initialization in bundle adjustment methods to improve the camera calibration precision @xcite .
the so called `` plumb - line '' methods , which correct the distortion by rectifying distorted images of 3d straight lines , date back to the 1970s ( see brown s seminal paper in 1971 @xcite ) .
since then , this idea has been applied to many distortion models : the radial model @xcite , the fov ( field of view ) model @xcite , or the rational function model @xcite .
these methods minimize the straightness error of the corrected lines . according to the fundamental theorem to be introduced in section [ sec : theorem ]
, the plumb - line methods minimize an error directly related to the distortion , without suffering from the above mentioned drawback , namely a numerical error compensation .
on the other hand , in spite of their name , current digital plumb - line methods usually involve flat patterns with alignments on them and not the plumb lines that were originally proposed in analog photogrammetry .
taken literally , these methods should use photographs of 3d straight lines .
when a high precision is required , this setup becomes much easier to build than a flat pattern .
the main purpose of this article is to show that very accurate distortion evaluation and correction can be obtained with a basic plumb line tool which we called `` calibration harp '' .
the calibration harp is nothing but a frame supporting tightly stretched strings .
nevertheless , the photographs of a calibration harp require a new numerical treatment to exploit them .
the strings will have to be detected at high sub - pixel accuracy and their distortion converted into an invariant measurement .
distortion measurements can be used to evaluate the distortion of a camera , but also its residual distortion after correction .
two aspects of the measurement should be clarified here . in this paper
we discuss distortion measurements that apply to the camera conceived as a whole : it is impossible to tell which part the relative position and deformation of the ccd , and the lens distortion itself play in the global camera distortion .
the distortion measurement is therefore not a pure optical lens distortion measurement , but the distortion measurement of the full acquisition system of camera+lens in a given state .
different lenses on different cameras can be compared only when the camera calibration matrix is known . on the other hand ,
the residual error due to different correction algorithms can be compared objectively after applying an appropriate normalization on the corrected images .
this paper is organized as follows : the fundamental theorem characterizing undistorted cameras is introduced in section [ sec : theorem ] .
section [ sec : harp_building ] uncovers the simple fabrication secrets of calibration harps .
the image processing algorithms needed for an automatic measurement are presented in section [ sec : edge_detection ] and section [ sec : measurements ] introduces the two most relevant measures .
section [ sec : applications ] demonstrates two applications , to the measurement of residual distortion after applying a calibration method , and to the transformation of existing manual distortion correction methods into automatic and far more precise ones .
finally , section [ sec : conclusion ] concludes the paper .
in multiple - view geometry , the pinhole camera is the ideal model that all techniques tend to approximate at best by calibrating the real cameras .
this model corresponds to the ideal geometric perspective projection .
the next theorem characterizes perspective projections by the fact that they preserve alignments . instead of restating the simplified version in @xcite
, we prefer to make it more precise .
the proof of the theorem can be found in @xcite .
[ theo : fundamental_prop ] _ let @xmath5 be a continuous map from @xmath6 to @xmath7 ( from 3d projective space to 2d projective plane ) .
if there is a point @xmath8 such that : _ 1 .
the images of any three point belonging to a line in @xmath6 not containing @xmath8 , are aligned points in @xmath7 ; 2 .
the images of any two points belonging to a line in @xmath6 containing @xmath8 , are the same point in @xmath7 ; 3 .
there are at least four points belonging to a plane not containing @xmath8 , such that any group of three of them are non aligned , and their images are non aligned either ; then @xmath5 is a pinhole camera with center @xmath8 .
this theorem provides us with a fundamental tool to verify that a camera follows the pinhole model .
nevertheless , _ rectifying straight lines does not define a unique distortion correction _ : two corrections can differ by any 2d homography that preserves all alignments . more concretely ,
assume that the real camera model is @xmath9 $ ] with @xmath8 the coordinate of camera optic center in a given 3d world frame , @xmath10 the camera @xmath11 orientation matrix , @xmath12 the camera @xmath11 calibration matrix and @xmath13 the camera lens non - linear distortion .
the estimated distortion can be written as @xmath14 with @xmath15 the unknown homography introduced in the distortion correction and can be different from one correction algorithm to another . by applying the inverse of the estimated distortion , the recovered pinhole camera is @xmath16 $ ] .
the homography @xmath17 can enlarge or reduce the straightness error , which makes the comparison of different correction algorithms unfair .
this effect can be compensated by two strategies .
to arrive at a universal measurement , a first strategy that we will consider is to normalize the homography : 1 .
select four points @xmath18 in the distorted image in general position ( not three of them aligned ) .
for example , they can be the four corners of the distorted image .
2 . find their corresponding points @xmath19 in the corrected image , according to the correction model : @xmath20 .
note that @xmath17 is different from one correction algorithm to another .
3 . compute the normalization homography @xmath21 which maps @xmath19 to @xmath22 : @xmath23 .
note that @xmath21 is different from one correction algorithm to another .
4 . apply the normalization homography @xmath21 on the corrected image . with this normalization ,
the final correction model is @xmath24 .
a second possible strategy would be to fix specific parameters in the correction model .
for example , since the zoom factor in the distortion correction is mainly determined by the order-@xmath25 parameters in the correction model , it is sufficient to set all the order-@xmath25 parameters to be @xmath25 to obtain a unique distortion measurement
. unfortunately , this will not be possible for some non - parametric distortion correction methods .
the first strategy therefore is more general .
theorem [ theo : fundamental_prop ] suggests to take a set of physical straight lines , as a calibration pattern .
however , a common practice actually contradicts the plumb - line basic idea : line patterns are printed and pasted over a flat plate .
there are many sources of imprecision in this setup : the printer quality is not perfect ; the paper thickness is not perfectly uniform ; the pasting process can add bubbles or a non uniform glue layer ; the supporting surface is not perfectly flat either .
notwithstanding these defects , if only a pixel precision is required , this setup is quite sufficient . nonetheless ,
when high sub - pixel precision is involved , the flatness errors can not be neglected . for current camera precision , a flatness error of 100 @xmath1 m ( the thickness of current writing paper ) for a @xmath26 cm pattern can lead to errors in the observed image coordinates of about @xmath2 pixels @xcite .
high precision aims at final 3d reconstructions far more precise than this .
so the measuring tool error should be also far smaller , if only possible .
the obvious advantage of `` real '' plumb lines on 2d patterns with straight lines on them , is that it is much easier to ensure a very precise physical straightness for lines than a very precise physical flatness for a physical pattern plate . yet
, the precision of the resulting measurement or correction depends on the straightness of the physical lines . in @xcite , to achieve a high precision , plumb lines were made of very fine white thread and stabilized by immersion of the plumb bobs in oil containers .
illumination was provided by a pair of vertically mounted fluorescent fixtures .
a dead black background was provided for the plumb - line array to highlight the contrast .
the points on the lines were measured with a microscope .
the measuring process required from 5 to 6 hours for generating 324 points .
the digital procedure proposed here will be automatic and faster . for the applications where the precision is not a crucial point , straight lines present in the natural scene can be used . in @xcite ,
the straight lines are obtained by photographing the architectural scenes and points on the lines are detected by canny edge detector . + in order to keep the high precision and simplify the fabrication procedure , we built a simple calibration pattern by tightly stretching strings on a frame as shown fig .
[ fig : harp1 ] .
the pattern looks like the musical instrument , hence its name .
the setup warrants the physical straightness of the lines .
its construction does not require any experimental skill , but only good quality strings .
indeed twisted strings show local width oscillations ; other strings do not have a round section , and a little torsion also results in width variations that can have a large spatial period .
rigid strings may have a remanent curvature .
finally , a ( tiny ) gravity effect can be avoided by using well stretched vertical lines . in our experiments
three different strings were tested : a sewing string , a smooth tennis racket string , and an opaque fishing string , all shown in fig . [ fig : three_types_lines1 ] .
sewing strings have a braid pattern and their thickness oscillates .
tennis racket strings are rigid and require a very strong tension to become straight .
fishing strings are both smooth and flexible , and can therefore be easily tightened and become very straight . the transparent ones , however , behave like a cylindrical lens , producing multiple complex edges . opaque fishing strings
end up being the best choice to build a calibration harp .
[ fig : string_high_freq1 ] shows an evaluation of the obtained straightness .
we took photos of the three types of strings , and corrected their distortion .
the green curves show the signed distance from edge points of a corrected line to its regression line .
the red curve shows the high frequency component of the corresponding distorted line .
the high frequency is computed as follows : * transform the edge points extracted from the distorted line into the intrinsic coordinate system , which is determined by the direction of the regression line computed from these points . in this coordinate system ,
the @xmath27-coordinate is the distance between the edge points and a reference edge point along the regression line and the @xmath28-coordinate is the signed distance from the edge points to the regression line .
this produces a one - dimensional signal ( see fig .
[ fig : intrinsic_coord ] ) . *
apply a big gaussian convolution of standard deviation @xmath29 pixels on this one - dimensional signal to keep only the low frequency component .
* the difference between this convolved signal and the original signal in the intrinsic coordinate system is considered as the high frequency .
the red curves in fig .
[ fig : string_high_freq1 ] show the high frequency ( due to the border effect of gaussian convolution , there is a sharp increase of magnitude at the border ) .
direction is determined by the direction of the regression line .
the @xmath27-coordinate is the distance to the reference point along the regression line and the @xmath28-coordinate is the signed distance from the edge points to the regression line.,scaledwidth=60.0% ] to ensure the precision of the edge detection in the string images , a uniform background whose color contrasts well with the string color must be preferred .
using an opaque background is not a good idea because this requires a direct lighting and the strings project shadows on the background ( fig .
[ fig : harp_opaque_bg1 ] and [ fig : closeup_harp_opaque_bg1 ] ) .
the sky itself is hardly usable : a large open space is needed to avoid buildings and trees entering the camera field of view , and clouds render it inhomogeneous , see fig .
[ fig : background1 ] .
the simplest solution is to place a translucent homogeneous paper or plastic sheet behind the harp and to use back illumination , preferably natural light to make it more uniform ( see fig .
[ fig : harp_translucent_bg1 ] and [ fig : closeup_harp_translucent_bg1 ] ) .
the acquisition aspects are also important for producing high quality measurements : lens blur , motion blur , aliasing , noise , must be as reduced as physically possible . to that aim a tripod and timer
were used to reduce camera motions , but also to avoid out of focus strings while taking photos at different orientations .
of course , changing focus changes the distortion .
thus each distortion calibration must be done for a fixed focus , and is associated with this focus .
in this section we describe the procedure to extract accurately and smooth the aligned edge points , which will be used to measure the distortion . devernay s algorithm @xcite is the classic sub - pixel accurate edge detector .
the implementation of devernay s detector is very simple since it is derived from the well - known non - maxima suppression method @xcite . on good quality images ( snr larger than 100 ) , devernay s detector can attain a precision of about @xmath30 pixels .
straightness measurements require the detection of groups of edge points that belong to the same physical straight line , and the rejection of points that do not belong to any line . to this aim ,
line segments are detected on the image using the lsd algorithm @xcite . when applied to photographs of the calibration harp
, the detection essentially corresponds to the strings . in case of a strong distortion
, one string edge could be cut into several line segments .
lsd works by grouping connected pixels into _ line support regions _ , see fig .
[ fig : line - support - regions ] ; for more details we refer to @xcite .
these regions are then approximated by a rectangle and validated .
the _ line support region _ links a line segment to its support pixels .
thus , devernay s edge points that belong to the same _ line support region _ can be grouped as aligned ; points belonging to none are ignored . for photos of strings ,
almost every pixel along each side of one string is detected as an edge point at sub - pixel precision .
so there are about @xmath31 edge points detected for a line of length of about @xmath31 pixels .
this large number of edge points opens the possibility to further reduce the detection and aliasing noise left by sub - sampling the edge points .
the sub - sampling step must be done carefully .
first the edge points are re - sampled to warrant a uniform sampling step along the edge ; this will facilitate the following steps .
the re - sampling uses a step of @xmath32 where @xmath33 is the length of a line and @xmath34 is the number of extracted edge points on the line .
the interpolation of an edge point @xmath35 between two adjacent edge points @xmath36 and @xmath37 is expressed by @xmath38 where @xmath39 and @xmath40 are the distances between the points , see fig .
[ fig : resampling ] .
then , a gaussian blur with @xmath41 is needed before a sub - sampling of factor @xmath42 to avoid aliasing @xcite .
we have two one - dimensional signals ( @xmath27-coordinate and @xmath28-coordinate of edge points ) along the length of the line .
the gaussian convolution is performed on both one - dimension signals , parameterized by the length along the edge .
finally , the sub - sampling of integer factor @xmath42 keeps one edge point out of @xmath42 .
, @xmath43 , @xmath44 are the edge points extracted by devernay s detector .
they are irregularly sampled along the line . the re - sampling ( in white dots )
is made along the line with the uniform length step @xmath45 .
linear interpolation is used to compute the re - sampled points.,scaledwidth=70.0% ]
this section examines two natural distortion measurements that are somewhat complementary . according to theorem [ theo : fundamental_prop ] , the most direct measure should be the straightness error , defined as the root - mean - square ( rms ) distance from a set of distorted edge points that correspond to the same physical line , to their global linear regression line , see fig .
[ fig : straightness ] .
given @xmath34 edge points @xmath46 of a distorted line , the regression line @xmath47 is computed by @xmath48 where @xmath49 since @xmath50 is a unit vector , the signed distance from point @xmath51 to the line is given by @xmath52 given @xmath33 lines , with @xmath53 points in line @xmath54 , the total sum of squared signed distance is given by @xmath55 thus , the rms straightness error is defined as @xmath56 where @xmath57 is the total number of points .
an alternative measure is the average maximal error defined by @xmath58 in the classic camera maker practice , the maximal error is defined by @xmath59 which would become instable with the calibration harp , some of the strings being potentially distorted by blur or wrong detection .
this measure is traditionally used in manual settings , for example see fig .
[ fig : dxo_distortion ] . while traditionally the measures are made
relatively to the line joining the extremities of the distorted edge , see fig .
[ fig : arc ] , here we use the signed distance to the regression line to make it more comparable to the previous measure .
the use of a signed distance and the difference between the maximal and minimal value is needed to handle correctly the fact that there are values on both sides of the regression line , see fig . [
fig : arc ] .
in this section , real photographs of the calibration harp will be used to evaluate the residual camera distortion when this distortion has been corrected with three state - of - the - art correction methods or two popular software .
in addition , we can feed any plumb - line method with the precise edge points detected on the harp images to improve the correction precision .
the efficient alvarez et al .
algorithm thus becomes an automatic parametric distortion correction method .
our lens distortion measurement algorithm can be tested on the online demo version available at http://bit.ly/lens-distortion .
it is reported in @xcite that devernay edge points have a precision better than @xmath30 pixel under zero - noise condition . as proposed in section [ sec : edge_detection ] , the precision of devernay edge points can be further improved by applying gaussian convolution of standard deviation @xmath60 , followed by a sub - sampling of factor @xmath42 . the only parameter to adjust here is the factor @xmath42 , which corresponds to the assumed regularity of the lens distortion .
we are only interested in realistic lens distortion , which makes a straight line globally convex or concave .
thus local edge oscillations due to noise can be harmlessly removed . in the experiments ,
the value of @xmath61 was chosen , which is enough to remove the local oscillation while keeping the global distortion . as a first main application
, the `` calibration harp '' permits to evaluate the performance of any distortion correction algorithm by measuring its residual distortion in corrected images .
the procedure is as follows : 1 .
a series of photos of the `` calibration harp '' are taken at different orientations .
these photos are processed by a camera distortion correction algorithm .
the corrected images are normalized by a homography as described in section [ sec : theorem ] .
the residual distortion is measured by the proposed method .
three distortion correction algorithms and two software were tested . with the exception of the classic lavest et al .
calibration method , all the others are designed to only correct the lens distortion without estimating the other camera parameters : * the lavest et al .
method @xcite : probably the most advanced pattern - based global camera calibration method , which estimates and corrects for the pattern non - flatness , using a bundle adjustment technique .
various distortion parameter configurations are allowed in this method : 2 radial parameters and 2 tangential parameters for a partial distortion model ; 2 radial parameters for a partial radial distortion model ; 5 radial parameters for a complete radial distortion model ; 5 radial parameters and 2 tangential parameters for a full distortion model . * a non - parametric lens distortion correction method requiring a textured flat pattern @xcite .
the pattern is obtained by printing a textured image and pasting it on an aluminum plate , which is thick and solid . * the dxo - optics - pro software : a program for professional photographers automatically correcting lens distortion ( even from fisheyes ) , color fringing and vignetting , noise and blur .
this software reads the exif of each image to know exactly what camera , lens and settings have been used .
it therefore uses a fixed lens distortion estimation for each supported camera model . *
ptlens : photoshop plug - in that corrects lens pincushion / barrel distortion , vignetting and chromatic aberration .
the distorted photographs to be corrected are shown in fig .
[ fig : distorted_calib_harp_photos ] and table [ tab : mesurements_on_different_algo ] shows the residual distortion measurements obtained by the calibration harp , after applying the corrections specified by the various methods .
+ .the distortion correction performance of three algorithms , measured by rms distance @xmath45 and maximal distance @xmath62 . [ cols="<,^,^,^,^,^,^,^",options="header " , ]
a `` calibration harp '' has been proposed for camera distortion measurement , along with its associated image processing chain .
this harp is both a measurement tool and a correction tool . as a measurement tool
, it can be used independently to measure the residual distortion left by any distortion correction methods or any software . as a correction tool ,
the precise edge points detected on the harp can be used as the input to plumb - line methods and lead to more precise distortion correction result . in the future , we aim at finding a more general distortion model to correct more severe distortion by using the calibration harp
. the ideal case would be a harp - free distortion correction method .
but we think that the harp will always remain useful , as a final measurement tool to validate any other correction precision and or to detect its failures .
this work was supported by agence nationale de la recherche anr-09-cord-003 ( callisto project ) , the miss project of centre national detudes spatiales , the office of naval research under grant n00014 - 97 - 1 - 0839 and the european research council , advanced grant `` twelve labours '' .
r. tsai , `` a versatile camera calibration technique for high - accuracy 3d machine vision metrology using off - the - shelf tv cameras and lenses , '' ieee journal of robotics and automation * 3 * , 323344 ( 1987 ) .
r. grompone von gioi , j. jakubowicz , j. morel , and g. randall , `` lsd : a fast line segment detector with a false detection control , '' ieee transactions on pattern analysis and machine intelligence * 32 * , 722732 ( 2010 ) . | this paper addresses the high precision measurement of the distortion of a digital camera from photographs .
traditionally , this distortion is measured from photographs of a flat pattern which contains aligned elements .
nevertheless , it is nearly impossible to fabricate a very flat pattern and to validate its flatness .
this fact limits the attainable measurable precisions .
in contrast , it is much easier to obtain physically very precise straight lines by tightly stretching good quality strings on a frame .
taking literally `` plumb - line methods '' , we built a `` calibration harp '' instead of the classic flat patterns to obtain a high precision measurement tool , demonstrably reaching @xmath0 pixel precisions .
the harp is complemented with the algorithms computing automatically from harp photographs two different and complementary lens distortion measurements .
the precision of the method is evaluated on images corrected by state - of - the - art distortion correction algorithms , and by popular software .
three applications are shown : first an objective and reliable measurement of the result of any distortion correction .
second , the harp permits to control state - of - the art global camera calibration algorithms : it permits to select the right distortion model , thus avoiding internal compensation errors inherent to these methods .
third , the method replaces manual procedures in other distortion correction methods , makes them fully automatic , and increases their reliability and precision .
image recognition , algorithms and filters ( 100.3008 ) , calibration ( 150.1488 ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ note to referees and editor : our lens distortion measurement algorithm can be tested on the online demo version available at http://bit.ly/lens-distortion . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ |
since the work of benamou and brenier @xcite which showed that the squared @xmath0-wasserstein distance between two probability measures @xmath1 and @xmath2 on @xmath3 can be expressed as the infimum of the kinetic energy @xmath4 among solutions of the continuity equation with prescribed endpoints @xmath5 it is natural to view optimal transport theory from a dynamical perspective and to look for geodesics rather than just for transport maps .
similarly , probability measure valued curves governed by the continuity equation ( with specific dependence of the velocity field @xmath6 on the mass @xmath7 ) play a crucial role in the theory of gradient flows in the wassertsein space ( see @xcite and the references therein ) . in the recent paper @xcite ,
dolbeault , nazaret and savar , introduced a new class of distances between probability measures through the introduction of some concave increasing nonnegative nonlinear mobility function @xmath8 . assuming that @xmath9 and @xmath10 , the corresponding squared distance between @xmath1 and @xmath2 is ( formally ) given by the infimum of @xmath11 subject to @xmath12 from a riemannian - like metric viewpoint , the formal interpretation of this mobility function is a conformal deformation factor of the riemannian tensor on the tangent space of the space of probability measures at a given point @xmath7 . from a modelling viewpoint ,
the concavity of @xmath8 is well suited to capture some congestion effects i.e. the fact that crowded zones of high densities result in higher values of the metric .
since mccann s pioneering work on displacement convexity @xcite , it is well known that displacement convex functionals play a distinguished role in the theory of gradient flows on the space of probability measures equipped with the @xmath0-wasserstein distance . in @xcite , carrillo , lisini , slepcev and savar identified structural assumptions that guarantee the convexity of internal energy functionals along geodesics for the dolbeault , nazaret and savar distances with a general mobility function @xmath8 and this analysis will actually show useful for certain estimates in the present paper .
let us mention that the monotonicity assumption on the mobility @xmath8 can be replaced by assuming that @xmath8 is compactly supported in an interval @xmath13 , corresponding to a hard congestion effect , since here the density can not assume values larger than @xmath14 .
such a case has been treated in @xcite and also considered in @xcite .
the main example that enters into this setting is @xmath15 , and a study of geodesics for the associated distance can be found in the article of brenier and puel @xcite , in the context of optimal multiphase transportation with a momentum constraint . in this paper
, we will focus on the case of a concave power mobility , @xmath16 , @xmath17 , and to avoid both compactness and boundary conditions issues , instead of working on a domain or @xmath3 , we will consider the case of the flat torus @xmath18 . our goal is to characterize minimizing geodesics i.e. minimizers of @xmath19 subject to the constraint @xmath20 it is worth noting as in @xcite that @xmath21 naturally interpolates between the @xmath22 and the @xmath0-wasserstein distance when @xmath23 varies between @xmath24 and @xmath25 respectively .
just as in @xcite taking @xmath26 as new variables , it is easy to see that this problem is a convex minimization problem , that is dual to some variational problem posed over some set of potentials @xmath27 ( which naturally play the role of lagrange multipliers associated to the constraint @xmath28 ) . formally , the optimality conditions for @xmath7 , @xmath27 obtained by convex duality read as the system @xmath29 let us remark that this system presents some similarities with the mean - field - games system of lasry and lions @xcite , @xcite . however , in the dual formulation the energy to be minimized is of the form @xmath30 for some positive constant @xmath31 .
since this somehow nonstandard functional is not obviously coercive on some sobolev space , the existence of a potential by the direct method of the calculus of variations is not immediate at all . in order to obtain estimates on minimizing sequences ,
a key ingredient is an estimate for some geodesic distances on @xmath32 given by lemma [ lem : estica ] . in particular
, we have to assume here that @xmath33 , an assumption that recurrently appears in previous works on this distances .
secondly , the potential we obtain is merely @xmath34 in time so some extra work has to be done to derive and justify rigorously the system of optimality conditions .
this result and its proof suggest a different approach to the problem , by somehow lifting the geodesics problem to a variational problem at the level of measures on curves , which is reminiscent to the work of carlier - jimenez - santambrogio @xcite ( see also @xcite ) .
more precisely , given @xmath35 a periodic measure supported by a suitable set @xmath36 of curves and which connects @xmath1 to @xmath2 ( in the sense that the image of @xmath35 by the evalution maps at initial and terminal time are respectively @xmath1 and @xmath2 ) , define the measure @xmath37 on @xmath38\times { \mathbb{t}}^d$ ] through @xmath39 for any continuous and periodic in space function @xmath40 .
we will investigate the precise links between the geodesics problem above and the minimization of @xmath41 the paper is organized as follows . in section [ existadj ] ,
when @xmath33 , we establish the existence and uniqueness of a potential function ( or adjoint state ) and give necessary and sufficient optimality conditions characterizing geodesics in section [ optimalcondi ] .
section [ sec : pbdualcourbe ] is devoted to the equivalent reformulation as a variational problem posed on the set of probability measures over curves .
in this section , we introduce ( a suitably relaxed version of ) the minimization problem , dual to the problem defining the distance between two probability densities .
the existence of a solution strongly relies of the existence of a minimizing sequence satisfying some a priori estimates , the proof of which is postponed to subsection [ aprioriest ] .
let @xmath17 , for @xmath42 , let us define @xmath43 and for @xmath44 , @xmath45 where @xmath46 by direct computation , one checks that @xmath47 and @xmath48 are convex lsc and conjugates : @xmath49 , @xmath50 and that for @xmath51 , one has @xmath52 recall that we work in the space - periodic framework , setting @xmath18 and @xmath53^d$ ] , this means that we identify spaces of functions on @xmath32 to spaces of @xmath54-periodic functions on @xmath3 .
let us then consider @xmath55\times { \mathbb{t}}^d ) } j(\phi):=\int_0 ^ 1 \int_q l(\partial_t \phi , \nabla \phi ) dx dt+\int_q \phi(0 , . )
\rho_0-\int_q \phi(1 , . ) \rho_1\ ] ] where @xmath1 , @xmath2 are two given probability measures on @xmath54 .
we shall see that in some sense to be made more precise , the variational problem ( [ primal ] ) admits as dual the dolbeault - nazaret - savar problem in @xcite @xmath56 where @xmath57 stands for the absolutely continuous part of the measure @xmath7 with respect to the lebesgue measure and the supremum is performed among pairs of measures @xmath58\times { \mathbb{t}}^d)\times { { \cal m}}([0,1]\times { \mathbb{t}}^d)^d$ ] such that @xmath59 is absolutely continuous with respect to the lebesgue measure and @xmath60 is a weak solution of the continuity equation @xmath61 i.e. satisfies @xmath62 for every @xmath63\times { \mathbb{t}}^d)$ ] .
the fact that ( [ dns ] ) coincides with the problem studied in @xcite has to be a little bit further explained : first remark that if @xmath58\times { \mathbb{t}}^d)\times { { \cal m}}([0,1]\times { \mathbb{t}}^d)^d$ ] solves ( [ contweak ] ) then the time marginal of @xmath7 is the lebesgue measure . by disintegration
we can thus write @xmath64 .
note now that @xmath65 , which implies ( see for instance lemma 4.1 in @xcite ) that @xmath66 is continuous for the weak @xmath67 topology of @xmath68 .
finally @xmath69 so that the functional in ( [ dns ] ) can be rewritten as @xmath70 which is exactly the functional to be minimized in @xcite .
assuming @xmath1 and @xmath2 belong to @xmath71 , we shall also need to suitably relax ( [ primal ] ) by considering @xmath72 where @xmath73 and @xmath74 denotes the absolutely continuous part of the measure @xmath75 with respect to the lebesgue measure .
since elements of @xmath76 are bounded and monotone nonincreasing with respect to time , the second and third term in @xmath77 are well - defined by monotone convergence i.e. @xmath78 and @xmath79 are intended as @xmath80 , and @xmath81 . for further use ,
let us also remark that for @xmath82 , by beppo - levi s monotone convergence theorem , one has @xmath83\times q } \phi(t , x ) \rho_0(x)dx dt , \;\ ] ] @xmath84\times q } \phi(t , x ) \rho_1(x)dx dt.\ ] ] [ dualitydns ] let @xmath1 and @xmath2 be two probability measures on @xmath32 such that @xmath85 , then @xmath86 and the infimum in ( [ dns ] ) is attained . if furthermore @xmath1 and @xmath2 belong to @xmath71 then , in addition , @xmath87 if we consider a general concave and nonnegative mobility fonction @xmath8 , the analogue of the problem writes @xmath88,\ ] ] where @xmath89 stands for the concave legendre transform of @xmath8 , defined by @xmath90 the corresponding duality theorem could be proved for such a general @xmath8 .
however , the main difficulty here is to get the existence of an adjoint state : getting estimates as in proposition [ prop : constrphin ] for a general mobility function @xmath8 is an open problem .
let us rewrite ( [ primal ] ) as @xmath91\times { \mathbb{t}}^d ) } \{f(d \phi)+g(\phi)\}\ ] ] where @xmath92 : @xmath93 \times { \mathbb{t}}^d)\to c([0,1]\times { \mathbb{t}}^d)\times c([0,1]\times { \mathbb{t}}^d)^d$ ] is defined by @xmath94 , @xmath95 @xmath96\times { \mathbb{t}}^d)\times c([0,1]\times { \mathbb{t}}^d)^d$ ] and @xmath97 note that , if @xmath63\times { \mathbb{t}}^d)$ ] is such that @xmath98 , then @xmath99 is continuous ( for the uniform topology ) at @xmath100 .
since @xmath101 is continuous and @xmath85 , fenchel - rockafellar s duality theorem thus implies that @xmath102 where @xmath103\times { \mathbb{t}}^d)$ ] is of course identified to the topological dual of @xmath104\times { \mathbb{t}}^d)$ ] . by direct computation , we have @xmath105 thanks to theorem 5 in @xcite , and the fact that @xmath106 , one has @xmath107 where @xmath108 and @xmath109 denote respectively the absolutely continuous part and singular part of @xmath110 , @xmath111 is any measure with respect to which @xmath109 is absolutely continuous ( for instance @xmath112 ) and @xmath113 is the recession function of @xmath48 : @xmath114 replacing in ( [ fenchel ] ) , we thus deduce that @xmath115 .
next we assume that @xmath1 and @xmath2 belong to @xmath71 and @xmath85
. then , still by the fenchel - rockafellar s duality theorem , the infimum in ( [ dns ] ) is attained .
let us finally prove that @xmath116 .
the fact that @xmath117 is obvious .
let @xmath82 ( extended by @xmath78 for @xmath118 and by @xmath79 for @xmath119 ) , then let @xmath120 where @xmath121 with @xmath122 , @xmath123 , @xmath124 , @xmath23 even , @xmath125 , @xmath126 , @xmath127 , @xmath128 even
. moreover we set @xmath129 . to prove the remaining inequality it is enough to prove that @xmath130 to see this , we assume of course that @xmath131 and first remark that @xmath132 . using the fact that @xmath47 is convex and nondecreasing in its first argument we thus get @xmath133 we next observe that thanks to the monotonicity of @xmath27 in time , setting @xmath134 , we have @xmath135 and similarly @xmath136 so that ( [ approx1 ] ) holds .
the crucial step to establish the existence of an optimal potential in ( [ primalr ] ) is to find a minimizing sequence for ( [ primal ] ) that satisfies the estimate @xmath137 the proof of this fact is postponed to subsection [ aprioriest ] .
these bounds enable us to prove the following result .
[ existdualdns ] if @xmath138 and @xmath1 and @xmath2 belong to @xmath71 .
then ( [ primalr ] ) admits a unique solution @xmath27 up to an additive constant .
moreover @xmath139 .
[ rem : lienint ] using the homogeneities in the functional @xmath140 , one easily checks that , if @xmath27 is the unique minimizer , then @xmath141 according to proposition [ prop : constrphin ] below , there is a a minimizing sequence @xmath142 of ( [ primal ] ) ( hence also of ( [ primalr ] ) ) and a constant @xmath143 , such that for every @xmath144 , @xmath145 , and @xmath146 taking a subsequence if necessary and using the monotonicity of @xmath142 with respect to time , we may assume that there is a @xmath147 such that @xmath148 and @xmath149 to prove that @xmath27 solves ( [ primal ] ) , we first deduce from theorem 5 in @xcite , that the functional @xmath150 is the convex conjugate of @xmath151\times { \mathbb{t}}^d)\times c([0,1]\times { \mathbb{t}}^d)^d \mapsto \int_0 ^ 1\int_q h(\theta(t , x ) , v(t , x))dx dt\ .\ ] ] it is therefore lsc for the weak @xmath67 topology of @xmath68 and therefore @xmath152 we finally deduce from ( [ btsci1 ] ) , ( [ btsci2 ] ) that the second term in @xmath140 is lsc for the weak @xmath67 topology of @xmath153 , thanks to ( [ pierre1 ] ) , we thus get @xmath154 so that @xmath27 solves ( [ primalr ] ) . to prove the uniqueness claim , let us first assume that @xmath155 .
we observe that if @xmath156 and @xmath157 are two distinct points in @xmath158 at which @xmath47 is finite , then @xmath47 is strictly convex on @xmath159 $ ] unless @xmath160 .
assume now that @xmath27 and @xmath161 are two solutions of ( [ primalr ] ) .
by the previous observation , we have , for almost all @xmath162 , @xmath163 @xmath164 in particular @xmath165 a.e .
, so that there is a measurable map @xmath166 with @xmath167 .
note that @xmath168 is bv , because @xmath27 and @xmath161 are bv .
next we note that , for a.e .
@xmath169 such that @xmath170 , we have @xmath171 and therefore @xmath172 .
let us assume for a while that the set @xmath173 of @xmath174 for which @xmath170 has a positive measure . then ,
if @xmath169 is a density point of @xmath173 , @xmath175 a.e . for some constant @xmath176 . since @xmath177 a.e
, we get @xmath178 therefore @xmath179 which is impossible because @xmath155 and @xmath180 is a distance .
so @xmath181 .
next we prove that the singular measure @xmath182 is zero .
let us decompose @xmath182 into its positive and negative part @xmath183 and @xmath184 ( i.e. , @xmath185 , @xmath186 ) and let us set @xmath187)$ ] . recall that @xmath183 is concentrated on a borel set @xmath188 $ ] such that @xmath189 .
we claim that @xmath190 is nonincreasing in time .
indeed , for any borel set @xmath191\times q$ ] , @xmath192 note moreover that @xmath193 , @xmath194 , @xmath195 , @xmath196)$ ] so that @xmath197)\;.\ ] ] since @xmath27 is a minimizer , this implies that @xmath198)=0 $ ] , so that @xmath199 . arguing in the same way with @xmath200)$ ] , one gets that @xmath201 . in conclusion
, @xmath168 is constant . if @xmath202 , then @xmath203 is optimal .
if @xmath161 is another optimal solution , one can show as in the first part of the proof that @xmath204 , where @xmath168 is a nonincreasing map . computing the criterium for @xmath168 shows that @xmath205 , so that again @xmath168 is constant .
a consequence of the proof of theorem [ existdualdns ] is the following technical remark , needed below : [ lem : limphin ] under the assumption of theorem [ existdualdns ] , let @xmath206 be a minimizing sequence of ( [ primal ] ) such that , for any @xmath144 , @xmath145 and @xmath207 and let @xmath27 be a limit of @xmath206 , in the sense that @xmath208 then @xmath27 is optimal for ( [ primal ] ) and @xmath209 @xmath210 moreover , if we set @xmath211 then @xmath212 converges to @xmath213 in @xmath214 .
equalities ( [ lim1 ] ) , ( [ lim2 ] ) are straightforward consequences of the proof of theorem [ existdualdns ] . in view of ( [ lim1 ] ) , @xmath212 is bounded in @xmath214 .
therefore a subsequence , still denoted @xmath212 , converges weakly to some @xmath215 .
let now @xmath216\times { \mathbb{r}}^d\to { \mathbb{r}}$ ] be a continuous , positive and periodic map . applying the argument of the proof of theorem [ existdualdns ] to the convex functional @xmath217
we get that @xmath218 therefore @xmath219 .
furthermore , in view of ( [ lim1 ] ) , @xmath220 so that @xmath221 and @xmath212 strongly converges to @xmath213 in @xmath214 .
here is an elementary property of minimizers .
[ prop : esssup ] if @xmath27 is optimal for ( [ primalr ] ) , then @xmath222 and @xmath223 are constant .
let @xmath224 and let us consider @xmath225 .
then one easily checks that @xmath226 while @xmath227 therefore @xmath228 is also optimal for ( [ primalr ] ) , so that by uniqueness @xmath229 .
in particular , @xmath230 for any @xmath169 .
since @xmath27 is nonincreasing in time , we can conclude that @xmath231 is constant .
the other assertion is proved similarly considering @xmath232 with @xmath233 .
we now prove that there exists a minimizing sequence for ( [ primal ] ) which satisfies the estimates needed in the proof of theorem [ existdualdns ] : [ prop : constrphin ] if @xmath138 , then there is a minimizing sequence @xmath206 for ( [ primal ] ) and a constant @xmath143 such that @xmath137 the proof relies on an integral estimate for geodesic distance .
let @xmath17 and @xmath234\times { \mathbb{r}}^d\to { \mathbb{r}}$ ] be a smooth , positive and @xmath235periodic map .
for @xmath236 and any @xmath237 , we consider the minimization problem @xmath238 where the infimum is taken over the set of all @xmath239 maps @xmath240\to { \mathbb{r}}^d$ ] such that @xmath241 and @xmath242 .
it will be convenient for later use to introduce the set @xmath243 of continuous maps @xmath240\to { \mathbb{r}}^d$ ] such that @xmath241 and @xmath242 and to extend the infimum in ( [ defca ] ) to @xmath243 by setting @xmath244 for simplicity , we denote respectively by @xmath245 and @xmath246 the quantity @xmath247 and the set @xmath248 .
[ lem : estica ] if @xmath138 , then , for any @xmath249 , @xmath250^{\frac{1-\alpha}{2-\alpha } } \frac{|x - y|^\sigma}{(t - s)^{\frac{1}{2-\alpha}}}\ ] ] where @xmath251 .
we start by proving the result for @xmath252 , @xmath253 .
the general case is obtained by scaling .
let @xmath254 . since @xmath138 we can fix @xmath255 sufficiently large such that @xmath256 throughout the proof @xmath257 denotes a constant depending on @xmath258 , @xmath23 and @xmath111 , which may change from line to line . let @xmath259 be the set of borel probability measures on @xmath246 endowed with the @xmath260-distance .
we first note that @xmath261 for any @xmath262 , @xmath263 $ ] and @xmath264 ( @xmath265 being the unit sphere of @xmath3 ) , we set @xmath266\\ x+r\sigma \delta^\theta + \frac{(y - x)(s-\delta)}{1 - 2\delta } & { \rm if } \ ; s\in [ \delta,1-\delta]\\ y+r\sigma ( 1-s)^\theta & { \rm if } \ ; s\in [ 1-\delta,1 ] \end{array}\right.\ ] ] and we define the probability measure @xmath267 by @xmath268 for any borel measurable nonnegative map @xmath269
. then @xmath270 where @xmath271 @xmath272^{\frac{\alpha}{2-\alpha}}\times\\ & \qquad & \times\left|\frac{(x - y)}{1 - 2\delta}\right|^{\frac{2}{2-\alpha}}\frac{d}{|s^d|}r^{d-1 } d\sigma dr ds \end{aligned}\ ] ] and @xmath273 let us first estimate @xmath274 . if we set @xmath275 for @xmath276 $ ] , we have , by change of variables and by hlder inequality , @xmath277^{\frac{1-\alpha}{2-\alpha } } \delta^{\frac{\theta d}{2-\alpha } } } \\ \leq & \ds { c\left|x - y\right|^{\frac{2}{2-\alpha}}\delta^{-\frac{d\theta(1-\alpha)}{2-\alpha } } \left[\int_0 ^ 1\int_{q } ( a(s , y))^{\frac{\alpha}{1-\alpha}}dyds\right]^{\frac{1-\alpha}{2-\alpha } } } \end{array}\ ] ] where @xmath278 } b(x_s,\delta^\theta)\subset [ -m , m]$ ] for some @xmath279 independent of @xmath280 , @xmath281 and @xmath282 ( note then that the last inequality comes from the periodicity of @xmath283 ) .
we now estimate @xmath284 . by change of variable and hlder inequality
, we have : @xmath285^{\frac{(1-\alpha)}{(2-\alpha)}}}\times\\ & \qquad { \ds \times\left [ \int_0^\delta\int_{b(x , s^\theta ) } |z - x|^2 s^{-d\theta(2-\alpha)-2 } dz ds \right]^{\frac{1}{(2-\alpha ) } } } \\ \end{array}\ ] ] where the last term is given by @xmath286 note that @xmath287 thanks to the choice of @xmath111 .
the term @xmath288 can be estimated in the same way.therefore @xmath289^{\frac{(1-\alpha)}{(2-\alpha)}}\times\\ & & \qquad \times\left [ \delta^ { \frac{\theta(2+d ) -d\theta(2-\alpha)-1}{(2-\alpha ) } } + \left|x - y\right|^{\frac{2}{(2-\alpha)}}\delta^{-\frac{d\theta(1-\alpha)}{(2-\alpha ) } } \right]\end{aligned}\ ] ] choosing @xmath290 , we get @xmath291^{\frac{(1-\alpha)}{(2-\alpha ) } } \left|x - y\right|^{\frac{2}{2-\alpha}\left(1-\frac{d\theta(1-\alpha)}{2\theta-1}\right)}\ ] ] where the exponent @xmath292 is positive from the choice of @xmath111 and that its limit as @xmath293 is @xmath294 .
it remains to check the general case .
let us notice that @xmath295 where @xmath296
. therefore @xmath297^{\frac{(1-\alpha)}{(2-\alpha ) } } @xmath298^{\frac{(1-\alpha)}{(2-\alpha ) } } = ( t - s)^{-\frac{1-\alpha}{2-\alpha } } \left[\int_s^t \int_{q } ( \tilde a(\tau , y))^{\frac{\alpha}{1-\alpha}}dyd\tau \right]^{\frac{(1-\alpha)}{(2-\alpha)}}\ ] ] this gives the result .
let @xmath142 be a minimizing sequence for ( [ primal ] ) .
without loss of generality we can assume that @xmath299 let us set @xmath300 then , if we set @xmath301 we have that , for any curve @xmath302\to { \mathbb{r}}^d$ ] , @xmath303 therefore , for any @xmath254 , @xmath304 where the infimum is taken over absolutely continuous curves such that @xmath305 .
in view of lemma [ lem : estica ] , for any @xmath249 , we have @xmath306^{\frac{1-\alpha}{2-\alpha } } |x - y|^\sigma\right)\ ] ] for some constant @xmath251 .
since @xmath307 , we finally get @xmath308^{\frac{1-\alpha}{2-\alpha}}\;.\ ] ] let now @xmath309 be a smooth function such that @xmath310 , @xmath311 for @xmath312 and @xmath313 . from now on we replace @xmath142 by @xmath314 : the new sequence @xmath206 is still minimizing and @xmath315^{\frac{1-\alpha}{2-\alpha}}.\ ] ] next we modify @xmath316 .
let @xmath317 be such that @xmath318 with @xmath319 for @xmath320 and @xmath321 .
replacing @xmath142 by @xmath322 , we get again a minimizing sequence such that @xmath323\times q\;.\ ] ] it remains to prove that @xmath324 is bounded : since @xmath142 is minimizing , we have , for @xmath144 sufficiently large @xmath325^{\frac{1-\alpha}{2-\alpha } } , \end{aligned}\ ] ] thanks to ( [ estiphi1 ] ) , @xmath326 and the fact @xmath2 is a probability measure . since @xmath327 , we obtain that @xmath328 is bounded , so that @xmath324 is also bounded .
replacing finally @xmath142 by @xmath329 gives that @xmath330 in order to conlude , let us now show that @xmath331 indeed , by hlder inequality , we have @xmath332^{1-\alpha } \left [ \int_0 ^ 1\int_q |\partial_t\phi_n| \right]^\alpha\ ] ] where the right - hand side is bounded since @xmath333
our aim now is to use the duality between ( [ primalr ] ) and ( [ dns ] ) to write necessary and sufficient optimality conditions in the form of a ( sort of ) system of pdes . to achieve this goal we have to be able to multiply @xmath7 by @xmath74 and @xmath59 by @xmath334 ,
since a priori @xmath335 is only @xmath336 we shall need a uniform in time @xmath153 estimate on @xmath337 when @xmath1 and @xmath2 are @xmath338 .
this estimate will follow from a generalized displacement convexity argument of carrillo , lisini , savar and slepcev @xcite .
this estimate will also be useful to treat the term @xmath339 if the functional in ( [ dns ] ) is finite , one has @xmath340 and @xmath341 whenever @xmath148 which ensures the summability of @xmath342 .
[ carrillo ] if @xmath1 and @xmath2 belong to @xmath343 , then @xmath21 is finite and the infimum in ( [ dns ] ) is achieved by a pair @xmath60 , disintegrating @xmath7 as @xmath64 we also have that @xmath344 for every @xmath169 with @xmath345 and @xmath346 . furthermore , if @xmath347 , and if there exists @xmath143 such that @xmath348 then @xmath349 a.e . on @xmath32 , for every @xmath169 .
the fact that @xmath21 is finite follows from corollary 5.25 in @xcite , the fact that the infimum is achieved then follows for instance from proposition [ dualitydns ] .
the estimate ( [ linfiniestim ] ) is obtained by proving the convexity of @xmath350 along geodesics with respect to @xmath180 for large @xmath351 , and letting @xmath351 goes to @xmath352 .
this has been done in @xcite , in the case of an open bounded convex set @xmath353 in @xmath3 , and is a consequence from the fact that the functional defined above generates a @xmath260-metric contraction gradient flow in the space of probability measures on @xmath353 endowed with @xmath180 ( in the sense of @xcite ) , given by the solution of the following porous medium equation on @xmath353 with zero flux condition on the boundary @xmath354 for all @xmath351 such that @xmath355 using the translation invariance of the equation , the reader may check that all the arguments of @xcite can be directly adapted to the case of the flat torus . to conclude the proof of the first part ,
let us notice that , since @xmath60 has finite energy , @xmath356 so that @xmath346 . when @xmath347 , the displacement convexity results of @xcite ( again adapted to the case of the flat torus ) imply that @xmath357 is convex for any convex @xmath358 , we thus have , taking arbitrary negative powers , @xmath359 } \vert \rho_t^{-1}\vert_{l^s}^s\leq \max ( \vert \rho_0^{-1 } \vert_{l^{\infty } } , \vert \rho_1^{-1}\vert_{l^{\infty}})\ ] ] for every @xmath360 and the desired claim is obtained by letting @xmath361 . [ ippineg ] assume that @xmath138 and that @xmath1 , @xmath2 belong to @xmath343 .
let @xmath362 be a weak solution of ( [ cont ] ) and @xmath82 be such that @xmath148 then @xmath363 we regularize @xmath27 exactly as in the proof of proposition [ dualitydns ] and thus set @xmath364 , since @xmath365 , thanks to ( [ cont ] ) , defining @xmath366 ( having extended @xmath7 by @xmath24 outside @xmath38 $ ] which is consistent with the fact that we have extended @xmath367 in the same way ) , thanks to the periodicity and @xmath368 , we then have @xmath369 since @xmath59 and @xmath334 are in @xmath370 , @xmath371 converges to @xmath372 .
moreover , since @xmath373 is uniformly bounded and converges to @xmath7 in @xmath336hence a.e . up to a subsequence ,
lebesgue s dominated convergence theorem implies that @xmath374 converges to @xmath375 . recalling ( [ convolin1 ] ) , ( [ convolin2 ] ) , we therefore obtain ( [ ipppositif ] ) by letting @xmath376 .
[ theo : opticond ] assume that @xmath138 and that @xmath1 , @xmath2 belong to @xmath343 .
let @xmath362 be a weak solution of ( [ cont ] )
. then @xmath60 solves ( [ dns ] ) if and only if there exists @xmath377 such that @xmath378 , @xmath379 and @xmath380 @xmath381 + 1 ) let us notice that the optimality system in theorem [ theo : opticond ] admits a unique solution @xmath382 ( with the property @xmath378 , and up to a constant for @xmath27 ) .
indeed , we have uniqueness in the dual problem , up to a constant , and the distance problem admits itself an unique solution @xmath110 .
+ 2 ) corollary 5.18 of @xcite states that the distance problem provides a ( unique ) constant speed geodesics .
this last property entails that @xmath383 is constant on @xmath38 $ ] , and then , we can slightly improve the regularity properties of the potential @xmath27 by @xmath384,l^2({\mathbb{t}}^d))\\ \ds \partial_t\phi^{{\rm ac } } \in l^{\infty}([0,1],l^1({\mathbb{t}}^d ) ) . \end{array } \right.\ ] ] \3 ) if @xmath385 solves ( [ dns ] ) and if , in addition , for every @xmath169 , @xmath386 a.e . on @xmath32 then the optimality conditions of theorem [ theo : opticond ] can be improved . in this case
, @xmath367 has no singular part .
indeed , proceeding as in the proof of lemma [ ippineg ] we obtain @xmath387 which together with the last condition of theorem [ theo : opticond ] gives the desired claim . in this case
, we therefore have @xmath388 and since @xmath7 is bounded away from @xmath24 , @xmath27 satisfies the hamilton - jacobi equation @xmath389 almost everywhere . according to theorem [ carrillo ] , in dimension @xmath25 , if @xmath1 , @xmath2 are bounded away from @xmath24 then
so is @xmath337 ( uniformly in @xmath169 ) so that the previous properties hold .
unfortunately , we do not know whether the same holds in higher dimension .
let @xmath60 solve ( [ dns ] ) ( so that @xmath362 by theorem [ carrillo ] ) .
we deduce from theorem [ existdualdns ] that there is a unique ( up to a constant ) solution @xmath82 of ( [ primalr ] ) and it satisfies @xmath148 . by the duality relation of proposition [ dualitydns ] , the fact that @xmath47 and @xmath48 are convex conjugates and young s inequality , we then have @xmath390 with ( [ ipppositif ] ) , we then have @xmath391 so that we should also have a.e .
an equality in young s inequality above .
this means that @xmath392 for a.e .
@xmath393 . therefore @xmath394 a.e . , for a.e .
@xmath162 for which @xmath395 one has @xmath396 and for a.e .
@xmath162 for which @xmath397 one has @xmath398 this proves the necessity claim . to prove sufficiency , assume that @xmath60 and @xmath27 satisfy the claim of the theorem .
let @xmath399 solve the continuity equation ( [ cont ] ) with @xmath400 ( which is without loss of generality in view of theorem [ carrillo ] ) and @xmath401 .
since , as previously , @xmath392 for a.e .
@xmath393 , we have @xmath402dx dt\\ & = \int_0 ^ 1 \int_q ( \mu \partial_t \phi^{\operatorname{ac}}+v\cdot \nabla \phi)+\int_q ( \phi(0,x ) \rho_0(x)- \phi(1,x ) \rho_1(x))dx \end{split}.\ ] ] by lemma [ ippineg ] , we have @xmath403 which finally enables us to conclude that @xmath60 solves ( [ dns ] ) .
let @xmath36 be the set of continuous curves @xmath302\to { \mathbb{r}}^d$ ] endowed with the topology of the uniform convergence and @xmath404 be the subset of such curves which are absolutely continuous and have a derivative in @xmath405)$ ] . for @xmath406 $ ] ,
we denote by @xmath407 the evaluation map at time @xmath169 : @xmath408 .
we denote by @xmath259 the set of borel probability measures @xmath35 on @xmath36 which are @xmath409 periodic in space : if @xmath410 is the translation in @xmath3 with vector @xmath411 and @xmath412 , then @xmath413 .
let @xmath414 be the subset of such measures with @xmath415 to any measure @xmath416 we associate the measure @xmath37 on @xmath38\times q$ ] defined by the equality @xmath417 for any continuous and periodic in space map @xmath40 .
we denote by @xmath418 the set of measures @xmath416 such that the measure @xmath37 is absolutely continuous with respect to the lebesgue measure . in this case
we identify @xmath37 with its density . finally , given two probability densities
@xmath419 , we denote by @xmath420 the set of @xmath421 such that @xmath422 and @xmath423 .
we set @xmath424 let us note for later use that : let @xmath425 weakly converge to @xmath35 . without loss of generality we can assume that @xmath426 . since @xmath427 is bounded in @xmath428 we can assume without loss of generality that @xmath427 converges weakly in @xmath428 to some @xmath429 .
for any continuous periodic and positive map @xmath40 the map @xmath430 is lower semicontinuous in @xmath36 .
hence @xmath431 therefore , @xmath37 is absolutely continuous , with @xmath432 and @xmath433 in order to show that @xmath436 , we recall that , from proposition [ dualitydns ] @xmath437\times { \mathbb{t}}^d ) } j(\phi ) \;.\ ] ] let @xmath438 and @xmath27 be a @xmath439 periodic map such that @xmath440 .
we set @xmath441 and @xmath442 following the same computations as in the proof of proposition [ prop : constrphin ] , we have @xmath443 therefore @xmath444 let now @xmath445 . integrating ( [ phi1phi0gamma ] ) with respect to @xmath35
then gives : @xmath446 therefore @xmath447 so we have proved that @xmath448\times { \mathbb{t}}^d ) } j(\phi ) \geq - \inf_{\eta\in \pi_{2,ac}(\rho_0,\rho_1 ) } k(\eta)\;.\ ] ] we now show the opposite inequality . for this , let @xmath110 be optimal in the problem @xmath449 let @xmath450 be the heat kernel on @xmath3 . then following @xcite the pair @xmath451 satisfies the continuity equation @xmath452 and is such that @xmath453 and @xmath454 note
also that @xmath455 is periodic , smooth in space and bounded below by a positive constant , while @xmath456 is also periodic and smooth .
therefore the flow @xmath457\\ x^x_0=x \end{array}\right.\ ] ] is well defined and satisfies @xmath458 for any @xmath411 . from standard properties related to the continuity equation
( see @xcite for instance ) , we have @xmath459 let us defined @xmath460 by @xmath461 for any continuous periodic and bounded map @xmath40 on @xmath36 .
let us compute @xmath462 .
we have @xmath463 for any continuous , periodic map @xmath40 on @xmath38\times { \mathbb{r}}^d$ ] .
therefore @xmath464 which shows that @xmath465
. moreover @xmath466 in particular , as @xmath467 , @xmath468 , which conclude the first part of the theorem .
the existence of a minimizer is a straightforward consequence of the next lemma
the proof of which is postponed stating that there is a subsequence @xmath469 which converges weakly to some @xmath421 , together with lemma [ lem : ksci ] . for @xmath472 ,
let @xmath473 , then @xmath474^{\frac{1}{2-\alpha}}\ ; \leq \ ; c^{\frac{1}{2-\alpha } } \end{array}\ ] ] since @xmath475 is relatively compact in @xmath36 , the sequence @xmath470 is tight , which implies that it has a converging subsequence .
assume @xmath476 is optimal for the problem ( [ pb : gamma ] ) .
then , for any @xmath477 , one has @xmath478 and @xmath479 .
therefore one gets as optimality condition for @xmath35 the inequality : @xmath480 since the problem is convex , this necessary condition is also sufficient . recalling the definition of @xmath37 ,
the above inequality heuristically means that any @xmath481 in the support of @xmath35 is optimal for the problem @xmath482 where the infimum is taken over the curves @xmath483 such that @xmath484 and @xmath485 .
[ prop : liens ] assume that @xmath1 and @xmath2 belong to @xmath486 .
let @xmath110 , @xmath27 and @xmath35 be optimal for ( [ dns ] ) , ( [ primalr ] ) and ( [ pb : gamma ] ) respectively .
then @xmath487 with the convention that @xmath488 and @xmath489 if @xmath490 .
moreover , there is a minimizer @xmath35 of ( [ pb : gamma ] ) such that @xmath491 and @xmath492 for any continuous , periodic map @xmath493\times { \mathbb{r}}^d\to { \mathbb{r}}^d$ ] .
the second equality is a straightforward consequence of theorem [ theo : opticond ] .
as for the first one , let us use the strategy of proof of theorem [ prop : equicurve ] .
let @xmath450 be the heat kernel on @xmath3 , set @xmath494 , and consider @xmath460 defined by @xmath461 for any continuous periodic and bounded map @xmath40 on @xmath36 , where @xmath495 is the solution of the differential equation ( [ eq : eqdiff ] ) .
we already know that @xmath496 and that , up to some sequence , @xmath497 converges to some @xmath498 which is optimal for ( [ pb : gamma ] ) . since @xmath499 is bounded in @xmath428 , we can consider a weak limit @xmath282 of the @xmath499 in @xmath428 and we have explained in the proof of lemma [ lem : ksci ] that @xmath500 .
hence @xmath501 this implies that @xmath462 strongly converges in @xmath428 to @xmath502 .
the map @xmath503 being affine and @xmath504 strictly convex , we must have @xmath505 .
furthermore , @xmath506 so that @xmath507 a.e . on @xmath508 . to complete the proof we note that @xmath509 therefore @xmath510 a.e . on @xmath511 .
let us finally check that @xmath35 satisfies ( [ lienetarhow ] ) . by definition of @xmath497 , @xmath512 .
moreover , for any continuous , periodic map @xmath493\times { \mathbb{r}}^d\to { \mathbb{r}}^d$ ] , we have @xmath513 defining @xmath514 , it is therefore enough to prove that @xmath515 as @xmath467 to obtain ( [ lienetarhow ] ) .
let @xmath472 and @xmath516 for @xmath517 and @xmath518 , define @xmath519 by standard uniform continuity arguments , observe that for fixed @xmath472 , @xmath520 tends to @xmath24 as @xmath521 .
now since @xmath99 is bounded and @xmath522 ( see lemma [ lem : etan ] ) we have @xmath523 finally , since @xmath524 is continuous for the @xmath260 topology , @xmath525 as @xmath467 which is enough to conclude .
lasry , p .-
lions , _ jeux champ moyen .
ii . horizon fini et contrle optimal .
( french ) [ mean field games .
finite horizon and optimal control ] _ , c. r. math .
paris , * 343 * , no . 10 , 679684 ( 2006 ) . | in this paper , we study the characterization of geodesics for a class of distances between probability measures introduced by dolbeault , nazaret and savar .
we first prove the existence of a potential function and then give necessary and sufficient optimality conditions that take the form of a coupled system of pdes somehow similar to the mean - field - games system of lasry and lions .
we also consider an equivalent formulation posed in a set of probability measures over curves .
* keywords : * dynamical transport distances , power mobility , geodesics in the space of probability measures , optimality conditions . |
a @xmath0-symmetric quantum theory is described by a hamiltonian that commutes with @xmath0 , where the operators @xmath1 and @xmath2 perform space reflection and time reversal @xcite .
even if a @xmath0-symmetric hamiltonian is not dirac hermitian ( that is , it is not invariant under combined matrix transposition and complex conjugation ) , the eigenvalues of the hamiltonian can still be entirely real .
@xmath0-symmetric hamiltonians are particularly interesting because they often have a parametric region of _ unbroken _ @xmath3 symmetry in which the eigenvalues are all real and a region of _ broken _ @xmath0 symmetry in which some of the eigenvalues are complex @xcite .
these regions are separated by a phase transition that has been repeatedly observed in laboratory experiments @xcite .
a heavily studied class of @xmath0-symmetric hamiltonians is @xcite @xmath12 where @xmath13 is a real parameter .
the eigenvalues of this hamiltonian are all real when @xmath14 and mostly complex when @xmath15 .
thus , the region of unbroken @xmath0 symmetry is @xmath14 and the region of broken @xmath0 symmetry is @xmath15 .
these two regions are separated by a phase transition at @xmath16 @xcite .
a special example of a @xmath0-symmetric hamiltonian whose eigenvalues are all real and positive is the cubic hamiltonian @xmath17 the @xmath18-dimensional , euclidean - space , field - theoretic equivalent of this quantum - mechanical theory is described by the lagrangian density @xmath19 this lagrangian is clearly not hermitian , but if we assume that the field @xmath20 transforms as a pseudoscalar , then it is @xmath0-symmetric .
this is because under this assumption , @xmath20 changes sign under space reflection @xmath1 , and since @xmath21 changes sign under @xmath2 , the interaction term is @xmath0 invariant .
while a conventional @xmath8 theory is interesting from a theoretical point of view , it is , of course , a physically unacceptable theory because the real cubic potential @xmath22 is not bounded below . as a consequence
, there can not be a stable ground state .
perturbation theory provides an easy intuitive explanation for the absence of a stable ground state .
the feynman graphical rules for a conventional @xmath8 quantum field theory follow directly from the lagrangian density @xmath23 the momentum - space amplitudes for a vertex and a line are @xmath24 using these feynman rules , we can in principle calculate the ground - state energy density @xmath25 by summing all connected vacuum graphs . because all such graphs have even numbers of vertices , this sum takes the form of a formal taylor series in powers of @xmath26 : @xmath27 where @xmath28 is the contribution of graphs having @xmath29 vertices .
the key point here is that all graphs contributing to the ground - state energy density have the same sign and add in phase , and thus the coefficients in the series ( [ e6 ] ) all have the same sign .
this series is divergent because the number of graphs having @xmath29 vertices grows like @xmath30 @xcite , but unlike the perturbation series for a @xmath10 field theory , it is not a stieltjes series @xcite because it does not alternate in sign .
consequently , the borel sum @xcite of the perturbation series has a cut on the real - positive axis in the complex-@xmath26 plane .
this perturbative argument shows that the ground - state energy density is complex ; the imaginary part of the energy density is the discontinuity across the cut .
we conclude that the ground - state of the conventional @xmath8 theory is unstable ; that is , it decays ( tunnels out to infinity through the barrier in the potential ) with a lifetime given by the imaginary part of @xmath31 . on the other hand
, perturbation theory also gives a simple intuitive argument that the non - hermitian , @xmath0-symmetric lagrangian density ( [ e3 ] ) defines a theory with a stable ground state .
note that the cubic potential in this theory is complex , and thus we can not ask whether it is unbounded below . the idea of a potential being bounded below applies only if the potential is real ; unlike the real numbers , the complex numbers are not ordered , so the notion of boundedness simply does not apply .
we obtain the @xmath0-symmetric lagrangian in ( [ e3 ] ) from the conventional lagrangian in ( [ e4 ] ) by replacing @xmath32 by @xmath33 .
when we do so , the perturbation series in ( [ e6 ] ) now _ alternates in sign_. as a consequence , it is a series of stieltjes and its borel sum is _ real _ @xcite .
we conclude from this argument that it is likely that the ground - state for this theory is stable .
while this perturbative argument is only heuristic , there is a rigorous proof @xcite that the spectrum of the cubic , quantum - mechanical @xmath3-symmetric hamiltonian in ( [ e2 ] ) is real and bounded below .
it is not yet known at a rigorous level whether the energy levels of the unconventional quantum field theory in ( [ e3 ] ) are real and bounded below because for this theory one can only rely on perturbative calculations . to show that the @xmath0-symmetric quantum field theory in ( [ e3 ] ) is a physically acceptable quantum theory one must ( in addition to proving that the spectrum of the theory is bounded below ) verify that there is a hilbert space with a positive inner product and that time evolution is unitary . to demonstrate this , one would have to show that there exists a linear operator @xmath34 whose square is unity and that @xmath34 commutes with both the hamiltonian and with the @xmath0 operator @xcite . in perturbation
theory the @xmath34 operator for the @xmath7 theory has been calculated to leading order @xcite , but it is not known rigorously whether the lagrangian ( [ e3 ] ) defines a physically acceptable theory .
( there may even be a critical value of @xmath32 at which a @xmath3 phase transition from a physically acceptable theory having real energies to an unphysical theory having complex eigenvalues occurs . )
however , we do know for certain that the conventional @xmath8 lagrangian in ( [ e4 ] ) defines a physically _ unacceptable _ theory ! while the conventional lagrangian in ( [ e4 ] ) is physically unacceptable and the unconventional lagrangian in ( [ e3 ] ) may or may not be physically acceptable , it is certainly interesting to study these lagrangians from a mathematical point of view .
the purpose of this article is to examine and contrast the renormalization - group properties of these two lagrangians .
we will show that while a conventional @xmath8 theory in @xmath9 dimensions is asymptotically free , the @xmath7 theory is like a @xmath10 theory in @xmath11 dimensions ; that is , it is stable , perturbatively renormalizable , and trivial . this paper is organized as follows : in sec .
[ s2 ] we review the standard perturbative renormalization treatment of a conventional @xmath8 theory .
then , in sec
. [ s3 ] we carry out the renormalization - group analysis for the @xmath6 theory . in sec .
[ s4 ] we repeat the analysis of sec . [ s3 ] for a @xmath0-symmetric @xmath7 theory .
we give some concluding remarks in sec .
the vacuum persistence functional in the presence of an external source @xmath35 for a @xmath18-dimensional euclidean - space quantum field theory described by a lagrangian @xmath36 is @xmath37=\int\mathcal{d}\phi\,e^{\int d^dx(-\cl+j\phi)}. \label{e7}\ ] ] let us consider the unrenormalized lagrangian for a conventional hermitian @xmath6 quantum field theory in which we include a linear self - interaction term : @xmath38 we can then rewrite @xmath39 $ ] as @xmath37=\mathcal{n}e^{-\int v(\delta/\delta j)}e^{\frac{1}{2}\int\int j d_b j } , \label{e9}\ ] ] where @xmath40 is a normalization constant , @xmath41 is the usual bosonic propagator in coordinate space , and @xmath42 .
the one - loop one - particle - irreducible unrenormalized vertex functions in momentum space are @xmath43 } , \label{e11}\\ \gamma^{(3)}(q_1,q_2)&= & g+g^3\int\frac{d^d p}{(2\pi)^d}\frac{1}{(p^2+m^2 ) [ ( p+q_1)^2+m^2][(p+q_1+q_2)^2+m^2]}. \label{e12}\end{aligned}\ ] ] to evaluate the above integrals we use the standard integral identities @xmath44 where @xmath45 is the metric matrix .
the upper critical dimension for the hermitian @xmath8 theory is @xmath9 . at @xmath46 the cubic operator
@xmath47 is marginal ( just as @xmath48 is marginal at @xmath49 ) .
the theory turns out to be asymptotically free , as we will see below .
normally , in textbooks the @xmath47 theory at or near @xmath9 dimensions is discussed for pedagogical reasons @xcite .
this is because the perturbative results are easily established and the theory provides a simple example of an asymptotically free theory .
furthermore , unlike the @xmath10 theory in @xmath11 dimensions , a contribution to the wave function renormalization constant @xmath50 is already present at the one - loop level . however , no physical meaning is attached to the conventional @xmath8 theory because , as noted earlier , it is unstable ( that is , the spectrum is unbounded below ) . let us now examine the behavior of this @xmath8 theory near @xmath9 .
let @xmath51 , @xmath52 , and @xmath53 represent the three integrals that appear in @xmath54 , @xmath55 , and @xmath56 above .
with the help of ( [ e13 ] ) , at @xmath57 we get @xmath58 where here and in the following we introduce the t hooft scale @xmath59 and give only the divergent parts of @xmath51 , @xmath52 and @xmath53 .
next , we consider the second integral @xmath60}. \label{e18}\ ] ] to extract its divergent part , we take two derivatives : @xmath61 ^ 2},\nonumber\\ \frac{\partial^2 i_2}{\partial q^{\mu}\partial q^{\nu}}&= & \int\frac{d^d p}{(2\pi)^d}\frac{8(p+q)_{\mu}(p+q)_{\nu}-2g_{\mu\nu}[(p+q)^2 + m^2]}{(p^2+m^2)[(p+q)^2+m^2]^3}. \label{e19}\end{aligned}\ ] ] we then expand @xmath62 around @xmath63 : @xmath64 where we have used the identities ( [ e13 ] ) and ( [ e14 ] ) .
the result is @xmath65 finally , for @xmath53 we use the identity @xmath66[(p+q_1+q_2)^2 ] + m^2}\nonumber\\ & = & \int_0 ^ 1 dx\,dy\,dz\,\delta(x+y+z-1)\,\int\frac{d^d p}{(2\pi)^d}\frac{1}{d^3 } \label{e22}\end{aligned}\ ] ] in which @xmath67 is evaluated at @xmath68 : @xmath69+z[(p+k)^2+m^2]\nonumber\\ & = & ( x+y+z)(p^2+m^2)+2 p\cdot(y q_1+z k)+y q_1 ^ 2+z k^2 . \label{e23}\end{aligned}\ ] ] by performing the shift @xmath70 , @xmath67 becomes @xmath71 we then obtain @xmath72 ^ 3}\nonumber\\ & = & \frac{\mu^{-\epsilon}}{64\pi^3\epsilon}+{\rm o}\left(\epsilon^0\right ) .
\label{e25}\end{aligned}\ ] ] in terms of the standard definitions for the renormalized quantities @xmath73 where the @xmath47 coupling constant @xmath74 is made dimensionless by introducing the t hooft scale @xmath59 , the renormalized vertex functions are @xmath75 where we have omitted the finite one - loop contributions . therefore , by adopting the @xmath76-scheme @xcite , we get @xmath77 finally , we define the _ dimensionless _ renormalized couplings @xmath78 , @xmath79 , and @xmath32 , which should not be confused with the bare parameters in ( [ e26 ] ) : @xmath80 the one - loop renormalization - group ( rg ) functions for the dimensionless renormalized couplings are then given by @xmath81
from ( [ e31 ] ) , ( [ e32 ] ) , and ( [ e33 ] ) , we see that near @xmath9 the theory possesses only a gaussian fixed point ( gfp ) : @xmath82 .
as is well known , the linearization of the rg equations around the gfp shows that near this point the couplings scale according to their scaling dimension .
that is , by defining @xmath83 , we find that @xmath84 note that @xmath85 at the gfp . finally , from @xmath86 we see that at @xmath9 the theory is asymptotically free .
the explicit solution of the rg equation @xmath87 is @xmath88 where @xmath89 is an arbitrary scale and @xmath90 . in ( [ e35 ] )
we immediately recognize the usual features of asymptotic freedom and infrared slavery .
however , we emphasize that despite exhibiting these important physical properties , the conventional @xmath8 theory is unstable .
by making the substitutions @xmath91 and @xmath92 in ( [ e31])([e33 ] ) , we find that @xmath93 unlike the conventional @xmath8 theory , we now have _ nontrivial fixed points _ at @xmath94 in addition to the usual gfp . as in the conventional case ,
the flow near the gfp is dictated by the canonical dimensions of the couplings .
near the non - gaussian fixed points , however , the linearization of the rg equations gives the following _ new scaling behavior _ : @xmath95 where @xmath96 , @xmath97 and @xmath98 are the eigenvalues of the @xmath99 jacobian matrix that defines the linearized rg flow around the non - gaussian fixed points , and @xmath100 , @xmath101 , and @xmath102 are arbitrary coefficients .
this result comes from solving the linearized system of differential rg equations around the non - gaussian fixed points ( see fig . [ f1 ] ) . from these equations we see that @xmath78 , @xmath79 , and @xmath32 are still eigendirections of the jacobian matrix , as was the case for the gfp . finally , the anomalous dimension of the field is @xmath103 it is worth noting that the hyperscaling relation that connects the anomalous dimension of the field with the eigenvalue @xmath104 , namely @xmath105 is satisfied , as expected . here , @xmath106 is the exponent that gives the anomalous scaling of the two - point function . near the critical region ,
the latter behaves as @xmath107
we have shown that the @xmath0-symmetric @xmath7 quantum field theory near @xmath9 dimensions possesses three fixed points , the gfp and two nontrivial ones in ( [ e40 ] ) . at @xmath9 ( @xmath16 )
the three fixed points merge in a unique fixed point , which is the gaussian one . from the @xmath86 function ( [ e39 ] )
, we can see that when @xmath16 , the theory is trivial : @xmath108 this allows us to conclude that the @xmath7 theory is energetically _
stable _ , _ perturbatively renormalizable _ , and _
trivial_. this triviality property is the same as for the conventional hermitian @xmath10 theory in @xmath11 dimensions . if we consider this @xmath7 theory in @xmath9 dimensions from an effective - field - theory standpoint ( as is the case for the higgs sector of the standard model ) , it can be treated as a perfectly sensible physical theory . from the rg point of view
, however , what seems to us to be more interesting is what happens when @xmath109 ( @xmath110 ) . in this case , if we consider the @xmath111 plane , we have a situation that closely parallels the ferromagnetic case as described in @xmath112 dimensions , where we have the _ gaussian _ and the _ wilson - fisher _ fixed points . in fig .
[ f2 ] the ( @xmath113 ) plane for the ordinary @xmath10 theory in @xmath112 dimensions is shown and the rg flows on this plane are plotted .
the gfp is at the origin , while the wilson - fisher fixed point is on the left of the @xmath114 axis .
the dashed lines are the eigendirections and the wilson - fisher fixed point is at the crossing of the two eigendirections ( one of which is the relevant direction , the other the irrelevant one ) .
the two fixed points , the gfp and the wffp determine the rg flows on this plane . in the case of a @xmath0-symmetric @xmath7 theory in @xmath115 dimensions
the situation in the @xmath116 plane is essentially the same .
however , the role of the @xmath117 term of the ferromagnetic model is played by @xmath79 , while the role of @xmath32 ( in the @xmath10 term ) is played by @xmath26 ( compare figs .
[ f1 ] and [ f2 ] ) . in the @xmath116 plane
the two eigendirections are the @xmath118 axis and the dashed line of fig .
the non - gaussian fixed point is at the crossing of the two eigendirections .
it is evident from figs .
[ f1 ] and [ f2 ] that the rg flow in the ( @xmath79 , @xmath26 ) plane is the _ same _ as the rg flow in the ferromagnetic case ; that is , it is the same as the flow in the ( @xmath113 ) plane . in both cases these flows
are governed by the two fixed points ( the gaussian one and the non - gaussian one ) . as is clear from figs .
[ f1 ] and [ f2 ] , the gaussian fixed point of the ferromagnetic case corresponds to the gaussian fixed point of the @xmath7 theory ; the wilson - fisher fixed point of the ferromagnetic case corresponds to our non - gaussian fixed point : @xmath118 , @xmath119 .
regarding the two fixed points in ( [ e40 ] ) , @xmath120 , it should be noted that in establishing this parallel , we refer to the square of the coupling constant @xmath26 rather than to the coupling constant @xmath32 itself .
we do this because it is convenient to treat the two fixed points in a unified manner because the physics around either fixed point is the same .
we note that the additional non - gaussian fixed points of the @xmath0-symmetric theory are also present in the conventional @xmath8 theory although they are purely imaginary [ see @xmath86 in ( [ e33 ] ) ] .
therefore , by considering also the purely imaginary solutions to the equation @xmath121 in the conventional @xmath8 theory , in a sense we recover the results obtained by stating from the beginning that the @xmath8 coupling in the lagrangian is purely imaginary ( which is the case for the @xmath0-symmetric theory ) . in summary , while the equation @xmath122 in the conventional @xmath8 theory has one real and two imaginary conjugate solutions , in the @xmath0-symmetric theory all of the three solutions are real . finally , we point out that in both the conventional and the @xmath0-symmetric theories the rg equations for @xmath32 and @xmath79 with @xmath123 and @xmath86 given by ( [ e32])([e33 ] ) and ( [ e38])([e39 ] ) , respectively , can be solved exactly . having defined @xmath124 as before , for the conventional theory we get @xmath125^{-5/9 } , \label{e47}\end{aligned}\ ] ] and
for the @xmath0-symmetric theory we get @xmath126^{-5/9}. \label{e49}\end{aligned}\ ] ] c. m. bender , s. boettcher , and p. n. meisinger , _ j. math
_ * 40 * , 2201 ( 1999 ) .
a. guo , g. j. salamo , d. duchesne , r. morandotti , m. volatier - ravat , v. aimez , g. a. siviloglou , and d. n. christodoulides , _ phys .
lett . _ * 103 * 093902 ( 2009 ) . c. e. rter , k. g. makris , r. el - ganainy , d. n. christodoulides , m. segev , and d. kip , _ nat .
* 6 * , 192 ( 2010 ) .
j. rubinstein , p. sternberg , and q. ma , _ phys .
lett . _ * 99 * , 167003 ( 2007 ) .
k. f. zhao , m. schaden , and z. wu , _ phys .
rev . _ a * 81 * , 042903 ( 2010 ) .
y. d. chong , l. ge , and a. d. stone _ phys .
* 106 * , 093902 ( 2011 ) .
z. lin , h. ramezani , t. eichelkraut , t. kottos , h. cao , and d. n. christodoulides _ phys .
lett . _ * 106 * , 213901 ( 2011 ) . c. zheng , l. hao , and g. l. long , arxiv:1105.6157 [ quant - ph ] .
s. bittner , b. dietz , u. guenther , h. l. harney , m. miski - oglu , a. richter , and f. schaefer , phys . rev . lett .
( in press , 2012 ) .
j. schindler , a. li , m. c. zheng , f. m. ellis , t. kottos , _ phys . rev . _ a * 84 * , 040101 ( 2011 ) .
optical pt lattices : a. szameit , m. c. rechtsman , o. bahat - treidel , and m. segev , phys .
a * 84 * , 021806(r ) ( 2011 ) . c. m. bender and t. t. wu _ phys .
_ * 27 * , 461 ( 1971 ) . c. m. bender and t. t. wu , _ phys . rev .
_ d * 7 * , 1620 ( 1973 ) ; reprinted in j. c. le guillou and j. zinn - justin ( eds . ) , _ large - order behaviour of perturbation theory _
( north - holland , amsterdam , 1990 ) , p. 41 - 57 . c. m. bender and s. a. orszag , _ advanced mathematical methods for scientists and engineers _ ( mcgraw hill , new york , 1978 ) . c. m. bender and g. v. dunne , j. math
. phys . * 40 * , 4616 ( 1999 ) . c. m. bender and e. j. weniger , j. math .
* 42 * , 2167 ( 2001 ) .
p. dorey , c. dunning , and r. tateo , _ j. phys . a : math .
* 34 * , l391 and * 34 * , 5679 ( 2001 ) .
p. dorey , c. dunning , and r. tateo , _ j. phys . a : math .
_ * 40 * , r205 ( 2007 ) . c. m. bender ,
d. c. brody , and h. f. jones , phys .
lett . * 93 * , 251601 ( 2004 ) .
see for instance m. peskin and d. schroeder , _ an introduction to quantum field theory _
( addison - wesley , new york , 1995 ) .
see for instance j. collins , _ renormalization : an introduction to renormalization _
( cambridge university , cambridge , 1984 ) and j. zinn - justin , _ quantum field theory and critical phenomena _
( claredon , new york , 1996 ) . | a quantum - mechanical theory is @xmath0-symmetric if it is described by a hamiltonian that commutes with @xmath0 , where the operator @xmath1 performs space reflection and the operator @xmath2 performs time reversal .
a @xmath0-symmetric hamiltonian often has a parametric region of unbroken @xmath0 symmetry in which the energy eigenvalues are all real .
there may also be a region of broken @xmath3 symmetry in which some of the eigenvalues are complex .
these regions are separated by a phase transition that has been repeatedly observed in laboratory experiments .
this paper focuses on the properties of a @xmath0-symmetric @xmath4 quantum field theory .
this quantum field theory is the analog of the @xmath0-symmetric quantum - mechanical theory described by the hamiltonian @xmath5 , whose eigenvalues have been rigorously shown to be all real .
this paper compares the renormalization - group properties of a conventional hermitian @xmath6 quantum field theory with those of the @xmath0-symmetric @xmath7 quantum field theory .
it is shown that while the conventional @xmath8 theory in @xmath9 dimensions is asymptotically free , the @xmath7 theory is like a @xmath10 theory in @xmath11 dimensions ; it is energetically stable , perturbatively renormalizable , and trivial . |
let @xmath0 be a finite alphabet of size @xmath1 viewed as an additive group mod @xmath1 .
consider the set @xmath2 of vectors of dimension @xmath3 over @xmath0 .
a vector @xmath4 will be written as a concatenation of @xmath5 blocks of length @xmath6 each , @xmath7 for a given vector @xmath4 let @xmath8 be the number of @xmath6-blocks of @xmath4 whose rightmost nonzero entry is in the @xmath9th position counting from the beginning of the block .
the @xmath6-vector @xmath10 will be called the _ shape _ of @xmath4 . for two vectors
@xmath11 let us write @xmath12 if @xmath13 .
a shape vector @xmath10 defines a partition of a number @xmath14 into a sum of @xmath6 parts .
let @xmath15 be the set of all such partitions . for brevity
we write @xmath16 let @xmath17 be a vector of shape @xmath18 . define a weight function ( norm ) on @xmath2 by setting @xmath19 and let @xmath20 denote the metric induced by this norm .
we call the function @xmath21 the _ ordered weight_. the ordered weight was first introduced by niederreiter @xcite in his study of low - discrepancy point sets . later , rosenbloom and tsfasman @xcite independently defined the weight @xmath22 ( more precisely , the weight @xmath23 defined below ) , calling it the @xmath24-metric , and studied codes in @xmath2 with respect to it .
the set @xmath2 together with this metric will be called the _ ordered hamming space _ ( the _ nrt space _ ) and denoted by @xmath25 . unless specified otherwise , below by distance ( weight ) we mean the ordered distance ( weight ) for some fixed value of @xmath6 .
note that the case @xmath26 corresponds to the usual hamming distance on @xmath27 .
an @xmath28 _ ordered code _
@xmath29 is an arbitrary subset of @xmath30 vectors in @xmath2 such that the ordered distance between any two distinct vectors in @xmath31 is @xmath32 or more .
if @xmath0 is a finite field and @xmath31 is a linear code of dimension @xmath33 , we refer to it as an @xmath34 $ ] code .
to define ordered orthogonal arrays , let us call a subset of coordinates @xmath35 _ left - adjusted _ if with any coordinate @xmath36 in the @xmath9th block it also contains all the coordinates @xmath37 of the same block .
a subset @xmath38 is called a @xmath39 _ ordered orthogonal array _ ( ooa ) of strength @xmath40 if its projection on any left - adjusted set of @xmath40 coordinates contains all the @xmath41 rows an equal number , say @xmath42 , of times .
the parameter @xmath42 is called the _ index _ of @xmath43 .
it follows that @xmath44 .
if @xmath31 is a linear @xmath34 $ ] code then the code orthogonal to it with respect to the dot product is a @xmath45 linear ooa of index @xmath46 if @xmath0 is equipped with the structure of an additive group , then one can construct _ additive _ ooas . if @xmath0 is a finite field , it is possible to construct linear ooas .
ooas ( also called hypercubic designs ) were introduced in lawrence @xcite and mullen and schmid @xcite as a combinatorial equivalent of point sets suitable for numerical integration over the cube .
informally this link can be described as follows .
let @xmath47^n$ ] be a unit cube , let @xmath48 be a continuous function of bounded variation and let @xmath49 be a set of @xmath50 points in @xmath51 called a net .
it is known that the error of quasi - monte carlo integration @xmath52 can be bounded above by @xmath53 where @xmath54 is the total variation of @xmath48 on @xmath55 and @xmath56 is the discrepancy factor of the net .
the parameter @xmath56 measures the deviation of the net from a uniformly distributed set of points .
the study of uniformly distributed point sets was initiated by h. weyl and e. hlawka .
@xcite developed the notions described above and gave the first constructions of nets with bounded factor @xmath58 the study of nearly uniform point sets was taken up in niederreiter @xcite which put forward the notion of @xmath59-nets and derived a bound on @xmath56 via its parameters .
we refer to @xcite for detailed background and more references for @xmath59-nets and in particular , to the literature on their constructions .
the following theorem relates ooas and @xmath59-nets .
below we use its statement as a definition of a @xmath59-net .
[ thm : tms ] ( lawrence @xcite , mullen and schmid @xcite ) .
there exists a @xmath59-net if and only if there exists an @xmath60 ooa of index @xmath41 and size @xmath61 independently of this line of work , rosenbloom and tsfasman @xcite considered codes in the ordered hamming space and derived bounds on their size .
it became clear shortly after their work that ooas and ordered codes are dual types of objects in the sense of delsarte s algebraic theory of coding .
this link opened up an avenue for applications of coding - theoretic methods to the study of @xmath59-nets and motivated the study of ordered codes and ooas independently of these applications .
in particular , martin @xcite has constructed an association scheme that describes the combinatorics of the space @xmath62 , formulated the linear programming bound ( lp bound ) , outlined a construction of orthogonal polynomials that describe the eigenvalues of the scheme and derived plotkin and rao bounds on ooas via linear programming .
much attention was also devoted to relations between the weight enumerators of linear codes and their duals in the ordered hamming space and its generalizations .
in particular , a macwilliams theorem for the nrt space was derived by martin @xcite , see also skriganov @xcite , dougherty and skriganov @xcite .
bierbrauer @xcite studied coding constructions of @xmath59-nets and bounds for them including the lp bound .
apart from the combinatorial context , ordered codes arise in a number of applied problems such as recent algebraic list decoding algorithms of reed - solomon codes @xcite , a study of linear complexity of sequences @xcite , and in a problem in communication theory @xcite . in this paper , we derive several new bounds on ooas and ordered codes .
we begin with a bassalygo - elias bound on codes which improves the known upper bounds on their size .
however , the bulk of our results are devoted to code bounds using the approach via association schemes and linear programming .
we begin with a study of the eigenvalues of the ordered hamming scheme , an association scheme that describes the combinatorics of the space @xmath63 the approach that we follow relies on the orthogonality relation of the eigenvalues .
this enables us to identify the eigenvalues as multivariate generalizations of the well - known krawtchouk polynomials , i.e. , a family of real polynomials of @xmath6 discrete variables orthogonal on @xmath64 with respect to the weight given by the multinomial probability distribution .
this is the subject of section [ sect : ac ] . turning to bounds , in sect .
[ sect : lpb ] we derive a new universal estimate of the size of ordered codes and ooas with a given distance ( strength ) .
the asymptotic version of this estimate improves the other results in a certain range of rates .
the final section is devoted to the case @xmath65 for which the bounds can be further improved relying on a direct approach . together with the space @xmath62
we will consider the space @xmath66 which differs from it in that the vectors are read `` from right to left . ''
namely , for @xmath17 let @xmath67 where @xmath68 is the number of blocks @xmath69 such that @xmath70 and @xmath71 let @xmath72 where @xmath73 and let @xmath74 these metric spaces are identical to each other ; the reason for considering them both is that if we equip @xmath62 with the structure of an association scheme then its dual scheme in delsarte s sense gives rise to the space @xmath75 in particular , if @xmath0 is a finite field and @xmath31 is a linear code in @xmath62 then its dual code @xmath76 lives in @xmath75 we elaborate on this below .
an easy combinatorial calculation shows that the number of vectors of shape @xmath18 in @xmath62 is given by @xmath77 and the number of vectors of weight @xmath32 equals to @xmath78 let @xmath79 and let @xmath80 satisfy the equation @xmath81 define the function @xmath82 in the case @xmath26 we write @xmath83 instead of @xmath84 , where @xmath85 let @xmath86 the asymptotic volume of the sphere in @xmath62 is given in the next lemma .
[ lemma : vol ] @xcite ( a ) for @xmath87 , equation ( [ eq : z0 ] ) has a unique positive root @xmath88.$ ] + ( b ) let @xmath89 be fixed , @xmath90 then @xmath91
in general , given the value of the distance or the strength , our goal is to construct as large codes and as small ooas as possible . the latter will also account for small - size @xmath59-nets with bounded discrepancy . in this section
we recall the known bounds on ordered codes and ooas and derive a new bound on the size of codes .
a gilbert - type bound on ordered codes was derived in @xcite .
there exists an @xmath28 code in the space @xmath62 whose parameters satisfy @xmath92 if @xmath0 is a finite field , then there exists a linear code with the same parameters .
a bound that applies specifically to linear codes was proved in @xcite .
it is analogous to the varshamov bound for the hamming space .
[ thm : var ] suppose that @xmath24 and @xmath40 satisfy the conditions @xmath93 then there exists an @xmath94 $ ] linear code in @xmath62 of distance @xmath95 , and a @xmath39 linear ooa of dimension @xmath24 .
while in general bounds on codes do not imply lower bounds on ooas , there are two special cases when these two types of results are equivalent .
first , if @xmath31 is an @xmath34 $ ] linear code over @xmath96 then the code @xmath97 is a @xmath45 linear ooa .
next , if an upper ( lower ) bound on codes ( ooas ) is obtained by linear programming as explained in the next section then the same solution of the lp problem gives a lower ( upper ) bound on ooas ( codes ) .
we next mention some upper bounds on codes and ooas .
the parameters of any @xmath28 code satisfy @xmath98 a plotkin bound on codes was established in @xcite .
namely , the following result holds true .
[ thm : plotkin ] let @xmath99 be a code of size @xmath30 and distance @xmath100 .
then @xmath101 a dual plotkin bound on ooas was proved by martin and visentin @xcite .
@xcite let @xmath31 be a @xmath102 $ ] ooa . if @xmath103 then @xmath104 according to the hamming bound , the parameters of any @xmath105 code satisfy @xmath106 a dual bound in this case is the rao bound which for the nrt space was established by martin and stinson @xcite : the size @xmath30 of any @xmath107 ooa satisfies @xmath108 the next result is new .
[ thm : be ] let @xmath31 be an @xmath28 code . then @xmath109 for any @xmath110 we will rely upon the next lemma .
[ lemma : johnson ] let @xmath111 be a code all of whose vectors have weight @xmath112 and are at least distance @xmath32 apart .
then for @xmath113 , @xmath114 let @xmath115 be a projection of @xmath31 on the @xmath9th block of coordinates .
for a vector @xmath116 let @xmath117 be its suffix of length @xmath118 .
given @xmath119 , we denote by @xmath120 its @xmath9th block and write @xmath121 to refer to the @xmath118-suffix of @xmath122 for @xmath123 let @xmath124 be the number of vectors in the @xmath9th block whose @xmath118-suffix equals @xmath125 we have @xmath126 compute the sum of all distances in the code as follows : @xmath127 to bound above the right - hand side , we need to find the minimum of the quadratic form @xmath128 under the constraints @xmath129 critical points of @xmath130 in the intersection of these hyperplanes , together with ( [ eq : hyp ] ) , satisfy the equations @xmath131 2\lambda_{i,0}^h+\alpha+\beta_{i , h}=0,&1\le i\le n;1\le h\le r\end{array } \quad\alpha , \beta_{i , h}\in\mathbb{r}.\ ] ] the system ( [ eq : hyp])-([eq : lag ] ) has a unique solution for the variables @xmath132 ; in particular , @xmath133 , \quad h=1,\dots , r , i=1,\dots , n,\ ] ] @xmath134 to verify that this critical point is in fact a minimum , observe that the form @xmath130 is convex because its hessian matrix is @xmath135 and is positive definite ( both globally and restricted to the intersection of the hyperplanes ( [ eq : hyp ] ) ) . substituting these values of the @xmath136s and taking account of the fact that @xmath137 we get @xmath138 ^ 2\ ] ] @xmath139 then from ( [ eq : sum1 ] ) we obtain @xmath140 which gives the result .
the proof of the theorem is completed as follows .
let @xmath141 be a sphere of radius @xmath112 around zero .
clearly , @xmath142 where @xmath143 is the maximum size of a distance-@xmath32 code in @xmath144 . with the previous lemma , this finishes the proof .
_ remarks .
_ 1 . this theorem implies a lower bound on the size @xmath30 of a linear ooa @xmath145 : for any @xmath146 @xmath147 and in particular , a lower bound on linear @xmath148-nets , @xmath149 \2 .
caution should be exercised in dealing with codes of a constant weight in the nrt space , i.e. , codes on the sphere @xmath144 in @xmath62 .
indeed , the sphere @xmath144 together with the metric @xmath21 is not homogeneous : in particular , the number of points in @xmath144 located at a given distance from a point @xmath150 depends on @xmath151 .
however , this does not cause problems in the previous theorem .
\3 . the argument used in the proof of lemma [ lemma : johnson ]
can be also used to give a proof of the plotkin bound , theorem [ thm : plotkin ] , that is simpler than the ones known in the literature .
indeed , let @xmath99 be a distance-@xmath32 code .
consider again expression ( [ eq : sum1 ] ) and note that this time there is no restriction on the weight of the codewords .
using the cauchy - schwarz inequality and the fact that @xmath152 , we obtain @xmath153 solving for @xmath30 concludes the proof . in this section
we assume that @xmath154 and @xmath6 is a constant . for a code of size
@xmath30 let @xmath155 be the code _
rate_. given a sequence of @xmath156 codes we will say that its _
asymptotic rate _ is @xmath157 and the asymptotic relative distance is @xmath158 if @xmath159 the plotkin bound implies that the asymptotic rate and distance of any sequence of codes satisfy @xmath160 to state the `` sphere packing '' or `` volume '' bounds on ordered codes we rely upon lemma [ lemma : vol ] .
namely @xcite , there exists a sequence of @xmath161 $ ] linear codes @xmath162 such that @xmath163 , @xmath164 @xmath165 and @xmath166 on the other hand , for any such sequence of codes , @xmath167 the asymptotic version of theorem [ thm : be ] is as follows : ( asymptotic bassalygo - elias bound ) . for @xmath168 the asymptotic rate and distance of any sequence of codes
satisfy @xmath169 this bound is better than the hamming bound for all @xmath170.$ ] it is also often better than the plotkin bound .
for instance , for @xmath171 the bound ( [ eq : be ] ) is better than the plotkin bound for all @xmath172 . for larger @xmath173 the improvement
is attained only for low values of @xmath158 since the right - hand side of ( [ eq : be ] ) becomes @xmath174-convex close to @xmath175 for instance , for @xmath176 this range is @xmath177 etc .
a @xmath59-net is called digital if the ooa that corresponds to it forms a linear subspace of @xmath178 therefore , bounds on linear ooas apply to the special case of digital @xmath59-nets .
however , studying asymptotics for this case requires a different normalization since the strength @xmath179 of the ooa that corresponds to the net equals @xmath180 and both approach infinity independently of @xmath181 . therefore , let @xmath182 denote the rate and @xmath183 denote the relative strength of the ooa that corresponds to the net . to state the bounds , we need to compute the asymptotic behavior of the volume of the sphere , which is different from ( [ eq : sphere ] ) .
the next result is due to bierbrauer and schmid @xcite .
[ thm : gvtms ] there exist families of digital @xmath59-nets with @xmath184 for which @xmath185 satisfy the bound @xmath186 where @xmath187 and @xmath188 is defined by @xmath189 on the other hand , by the rao bound , any family of @xmath59-nets satisfies @xmath190 observe that theorem [ thm : be ] in this case gives the same result as the rao bound because the increase of the packing radius in ( [ eq : beooa ] ) over @xmath191 vanishes asymptotically . indeed ,
taking @xmath192 and replacing @xmath40 with @xmath179 , we obtain from ( [ eq : beooa ] ) @xmath193 the tightest bound is obtained if we take @xmath194 in this inequality . _
remark : _ we note that in the case that both @xmath154 and @xmath195 while @xmath196 tends to a constant bounded away from 0 and 1 , the lower and upper bounds on codes coincide @xcite ( the gilbert - varshamov bound converges to the singleton bound ) .
the coding - theoretic notion of duality prompted a study of the association scheme that describes the combinatorics of the nrt space .
we briefly recall some elements of delsarte s theory of association schemes @xcite .
a symmetric association scheme with @xmath197 classes is a finite set @xmath198 equipped with a set @xmath199 of symmetric binary relations on @xmath200 such that \(i ) @xmath201 \(ii ) for each @xmath202 the number @xmath203 depends only on @xmath204 moreover , @xmath205 the parameters @xmath206 are called the intersection numbers of the scheme @xmath207 the numbers @xmath208 are called the valencies of @xmath209 let @xmath210 be the adjacency matrix of the relation @xmath211 it is clear that @xmath212 and for all @xmath213 the product @xmath214 is contained in the linear span of @xmath215\dots , a_d\}.$ ] the matrices @xmath210 form a commutative algebra over @xmath216 with respect to matrix multiplication , called the bose - mesner algebra .
it is clear that the bose - mesner algebra is also closed under the hadamard ( elementwise ) multiplication @xmath217 , viz .
@xmath218 with respect to @xmath219 this algebra has a basis of primitive idempotents @xmath220 that satisfy @xmath221 the numbers @xmath222 are nonnegative .
they are called the krein parameters of the association scheme .
the quantities @xmath223 are called multiplicities of the scheme @xmath209 the matrices @xmath224 and @xmath225 defined by @xmath226 and @xmath227 are called the _ first _ and _ second eigenvalues _ of @xmath228 , respectively .
the eigenvalue matrices satisfy @xmath229 further , for @xmath230 the eigenvalues satisfy @xmath231 @xmath232 @xmath233 two @xmath197-class association schemes are called delsarte duals of each other if the adjacency matrices @xmath210 , the first eigenvalues @xmath224 , and the intersection numbers @xmath206 of one scheme are , respectively , the idempotents @xmath234 , the second eigenvalues @xmath225 and the krein numbers @xmath222 of the other .
the duality also exchanges the role of the matrix and hadamard multiplication .
a scheme is called _ self - dual _ if it equals its dual .
for instance , the hamming scheme @xmath235 is self - dual .
one of the manifestations of self - duality in this case , obvious from the definition , is the macwilliams theorem that relates the weight distribution of an additive code to that of its dual .
a scheme is called _ formally self - dual _
@xcite if there exists some ordering of primitive idempotents under which @xmath236 in a formally self - dual scheme @xmath237 and @xmath238 following delsarte @xcite a scheme @xmath239 is called an _ extension _ of an @xmath6-class scheme @xmath240 if its vertex set is the @xmath5-fold cartesian product of @xmath241 and the relations @xmath242 are given by @xmath243 apart from @xcite we refer to @xcite for the proofs of these results and more information on association schemes .
the association scheme for the nrt space was constructed and studied by martin and stinson @xcite .
define an @xmath6-class `` kernel scheme '' @xmath244 with the relations @xmath245 [ thm : ms]@xcite the space @xmath246 together with the relations @xmath247 forms a formally self - dual association scheme @xmath248 , called the @xmath6-hamming scheme .
it can be constructed as an @xmath5-fold delsarte extension of @xmath249 the dual scheme of @xmath248 is @xmath250 whose point set is @xmath246 and the set of relations is given by @xmath251
in the conventional case of @xmath26 , eigenvalues of the hamming scheme are given by the krawtchouk polynomials @xmath252 which form a family of polynomials of one discrete variable orthogonal on the set @xmath253 with weight @xmath254 i.e. , the binomial probability distribution .
here we are interested in their generalization for the @xmath6-hamming scheme .
observe that the valencies @xmath255 of the scheme are given by ( [ eq : v ] ) . by self - duality and ( [ eq : ort1 ] ) , the eigenvalues are orthogonal on the space of partitions @xmath64 with weight @xmath256 below it will be convenient to normalize the weight .
let @xmath257 be the space of real polynomials of @xmath6 discrete variables @xmath258 defined on @xmath259 let us define a bilinear form acting on the space @xmath260 by @xmath261 where @xmath262 letting @xmath263 we observe that @xmath264 forms a multinomial probability distribution on @xmath64 .
therefore , @xmath6-variate polynomials orthogonal with respect to this weight form a particular case of multivariate krawtchouk polynomials . for a partition @xmath265 denote by @xmath266 the krawtchouk polynomial that corresponds to it .
let @xmath267 be the degree of @xmath268 our goal in this section is to derive properties of the polynomials @xmath268 in their large part , these properties are obtained by specializing to the current case general relations of the previous section .
however , some work is needed to transform them to a concrete form which will be used in later calculations .
the following relations are useful below .
[ lemma : xixj ] @xmath269 to prove ( [ eq : x1 ] ) , compute @xmath270 @xmath271 the sum on @xmath18 on the last line equals @xmath272 which finishes the proof .
the remaining two identities are proved in a similar way .
\(i ) @xmath274 is a polynomial in the variables @xmath275 of degree @xmath276 there are @xmath277 different polynomials of the same degree , each corresponding to a partition of @xmath278 .
\(ii ) ( _ orthogonality _ ) equation ( [ eq : ort1 ] ) is rewritten as @xmath279 in particular , let @xmath280 be a partition with one part .
we have @xmath281 indeed , equality ( [ eq : ort ] ) is simply ( [ eq : ort1 ] ) specialized to the case at hand and ( [ eq : ortl ] ) is obtained from ( [ eq : v ] ) .
\(iii ) ( _ linear polynomials _ ) for @xmath282 @xmath283 _ proof : _ this is shown by orthogonalizing the set of linear polynomials @xmath284x_r)$ ] .
we take @xmath285 use lemma [ lemma : xixj ] to compute @xmath286 for some constant @xmath287 .
to find @xmath287 , use ( [ eq : ortl ] ) : @xmath288 hence @xmath289 we take @xmath290 choosing @xmath291 so that @xmath292 next let us perform the induction step to compute @xmath293 : @xmath294 where the polynomials @xmath295 have the form ( [ eq : lin ] ) by the induction hypothesis .
straightforward calculations using ( [ eq : x1])-([eq : xixi ] ) show that @xmath296 again using ( [ eq : ortl ] ) , we find that @xmath297 as above , we will choose the minus . ' '' ''
\(iv ) the next property is a special case of ( [ eq : flip ] ) .
@xmath298 in particular , @xmath299 \(v ) for any @xmath300 @xmath301 where the linearization coefficients @xmath302 are the intersection numbers of the scheme .
in particular , @xmath303 this is a special case of property ( [ eq : pijk ] ) .
\(vi ) ( _ three - term relation _ ) let @xmath304 be a column vector of the polynomials @xmath305 ordered lexicographically with respect to all @xmath48 that satisfy @xmath306 the three - term relation is obtained by expanding a product @xmath307 in the basis @xmath308 , where @xmath309 is a first - degree polynomial . by orthogonality ,
the only nonzero terms in this expansion will be polynomials of degrees @xmath310 @xcite .
we will establish an explicit form of the three - term relation for @xmath311 we have @xmath312 where @xmath313 are matrices of order @xmath314 and @xmath315 respectively .
the nonzero elements of these matrices have the following form : @xmath316=l_i ( f_i+1 ) \quad&\text{if } h=(f_1,\dots , f_i+1,\dots , f_r),\\[2 mm ] c_{\kappa}[f , h]=l_i(n-\kappa+1)q^{i-1}(q-1 ) & \text{if } h=(f_1,\dots , f_i-1,\dots , f_r ) , \end{array}\ ] ] @xmath317=\left\{\begin{array}{ll } l_i f_iq^{i-1}(q-2 ) & \text{if } h = f,\\ \displaystyle l_i(f_k+1)q^{i-1}(q-1 ) & \text{if } h=(f_1,\dots , f_k+1,\dots , f_i-1,\dots , f_r),\\ & \quad 1\le k < i,\\ \displaystyle l_i(f_i+1)q^{k-1}(q-1 ) & \text{if } h=(f_1,\dots , f_k-1,\dots , f_i+1,\dots , f_r),\\ & \quad 1\le k < i , \end{array}\right.\ ] ] where @xmath318 according to property ( v ) , the coefficients of the expansion of the product @xmath319 into the basis @xmath320 are given by the intersection numbers of the scheme : @xmath321 because the ordered metric is translation - invariant , we can assume that @xmath322 so @xmath323 is the number of vectors @xmath324 that satisfy @xmath325 for a fixed vector @xmath326 .
in other words , @xmath327 where the nonzero block is located in any of the @xmath5 possible blocks , @xmath328 the numbers @xmath323 are nonzero only in the three following cases . 1 .
@xmath329 by the above we have that @xmath330 for @xmath331 and @xmath332 hence @xmath333 can be chosen so that its @xmath334 blocks of weight @xmath9 annihilate the corresponding blocks of @xmath151 , leaving one such block in any of the @xmath335 locations .
thus , @xmath336 2 . @xmath337 the following numbers are easily verified by ( [ eq : one ] ) : @xmath338 other than these three cases , no other possibilities for @xmath118 arise .
@xmath339 now we should add to @xmath151 one block of weight @xmath9 in any of the @xmath340 all - zero blocks .
thus we obtain @xmath341 and @xmath342 for all other @xmath118 . to prove ( [ eq:3term ] ) we now need to represent @xmath309 as a linear combination of the @xmath343s . using ( [ eq : lin ] ) we find that @xmath344 hence @xmath345 the proof is now concluded by using ( [ eq : pfih ] ) together with the intersection numbers computed above .
' '' '' along with the polynomials @xmath273 below we use their normalized version @xmath346 the polynomials @xmath347 form an orthonormal basis of @xmath260 .
denote by @xmath348 the coefficient matrices of the normalized form of relation ( [ eq:3term ] ) .
the new matrix elements are given by @xmath349=l_i \sqrt{(f_i+1)(n-\kappa)q^{i-1}(q-1 ) } \quad&\text{if } h=(f_1,\dots , f_i+1,\dots , f_r),\\[3 mm ] c_{\kappa}[f , h]=l_i\sqrt{(n-\kappa+1)f_iq^{i-1}(q-1 ) } & \text{if } h=(f_1,\dots , f_i-1,\dots , f_r ) , \end{array}\ ] ] @xmath350=\left\{\begin{array}{ll } l_i f_iq^{i-1}(q-2 ) & \text{if } h = f,\\[3 mm ] \displaystyle l_i\frac{q-1}q\sqrt{(f_k+1)f_iq^{k+i } } & \text{if } h=(f_1,\dots , f_k+1,\dots , f_i-1,\dots , f_r),\\ & \quad 1\le k < i,\\ \displaystyle l_i\frac{q-1}q\sqrt{f_k(f_i+1)q^{k+i } } & \text{if } h=(f_1,\dots , f_k-1,\dots , f_i+1,\dots , f_r),\\ & \quad 1\le k < i .
\end{array}\right.\ ] ] let @xmath351 be the set of polynomials of total degree @xmath352 let @xmath353 be the orthogonal projection of @xmath260 on @xmath354 . define the operator @xmath355 its matrix in the orthonormal basis has the form @xmath356\ ] ] where the @xmath357s are symmetric and @xmath358 on account of property ( v ) and ( [ eq : pe ] ) , the matrix elements of @xmath359 are nonnegative .
the matrix of @xmath360 in the basis @xmath308 has the property @xmath361=v_f \bfs_{\kappa}[h , f ] \quad(f , h\in\delta_{r , n}).\ ] ] \(vii ) ( _ explicit expression _ )
@xmath362 where @xmath363 is a univariate krawtchouk polynomial ( [ eq : k ] ) , @xmath364 and @xmath365 this form of the polynomial @xmath366 was obtained in @xcite ( various other forms were found in @xcite ) .
we remark that ( [ eq : expl ] ) can be proved by performing the gram - schmidt procedure ( [ eq : gram ] ) for monomials of higher degrees .
it is known that the resulting system of polynomials is unique up to a constant factor once the polynomials of degrees 0 and 1 together with the three - term relation ( [ eq:3term ] ) have been fixed , see ( * ? ? ?
* theorem 3.4.9 ) .
\(viii ) the matrix elements of the eigenvalue matrices of the association scheme @xmath62 are given by @xmath367 this follows from the previous property and ( [ eq : ort1 ] ) because the polynomials @xmath308 form a unique orthogonal family on @xmath64 with respect to the weight @xmath368 \(ix ) ( _ fourier transform representation _ ) let @xmath369 be a @xmath1th degree primitive root of unity , @xmath370 then @xmath371 where @xmath372 in @xcite this relation is taken as a definition of the polynomials @xmath305 . under our approach , it follows from the well - known fourier transform representation of the krawtchouk polynomials @xmath373 in the case @xmath26 and theorem [ thm : ms ] .
\(x ) ( _ christoffel - darboux _ ) . let @xmath374 and define @xmath375 let @xmath376 be the matrix of the operator @xmath377 given by @xmath378 written in the basis @xmath379 the action of @xmath309 on @xmath380 is described as follows : @xmath381(k_h(e)k_f(a)-k_h(a)k_f(e))\ ] ] @xmath382(k_h(e)k_f(a)-k_h(a)k_f(e)),\ ] ] the last equality justified by ( [ eq : nonsymm ] ) as follows : @xmath383(k_h(e)k_f(a)-k_h(a)k_f(e))\\ = & \sum_{f , h } \bfs[f , h]\sqrt{\frac{v_h}{v_f } } ( { \widetilde}k_h(e){\widetilde}k_f(a)-{\widetilde}k_f(e){\widetilde}k_h(a))\\=&0 . \end{aligned}\ ] ] a particular case of the above is obtained when @xmath384 .
the kernel @xmath380 , denoted in this case by @xmath385 , equals @xmath386 and we obtain @xmath387 @xmath388 where @xmath389.$ ] this relation is called the christoffel - darboux formula .
\(xi ) the generating function of the polynomials @xmath305 is given by @xmath390 in particular , @xmath391 _ remarks : _ 1 .
the polynomials @xmath273 were considered in @xcite .
however none of these papers constructed them from their definition as eigenvalues of the @xmath6-hamming scheme ( to be more precise , martin and stinson @xcite mention this approach but pursue the path suggested in theorem [ thm : ms ] which makes explicit calculations difficult ) . under the approach taken above , many properties of the polynomials @xmath273 follow as special cases of the general combinatorial results of the previous section .
other generalizations of univariate krawtchouk polynomials were considered earlier in @xcite .
these papers study biorthogonal polynomials for the weight given by the multinomial probability distribution , resulting in polynomial families different from the one considered above .
property ( xi ) implies a macwilliams theorem for nrt codes .
it was previously proved in @xcite using different means .
[ thm : mw ] ( macwilliams theorem in the nrt space ) .
let @xmath99 and @xmath392 be two linear codes that satisfy @xmath393 for every @xmath394 let @xmath395 be the weight enumerator of @xmath31 and let @xmath396 be the same for @xmath397 then @xmath398 where @xmath399
in this section we prove one of our main results , an lp bound on the rate of codes .
let @xmath400 be the eigenvalue matrix of the @xmath6-hamming scheme , where for definiteness we are assuming the lexicographic order on the partitions .
let @xmath401 be a vector of nonnegative real variables , with the same ordering .
define two linear programs , @xmath402 and let lp(i ) and lp(ii ) be their solutions .
let @xmath403 be a code and @xmath404 be an ooa of size @xmath405 then it follows from delsarte s work @xcite that @xmath406 replacing the lp problems by their duals , we obtain the following theorem . [ thm : lp ] let @xmath407 be a polynomial that satisfies @xmath408 then any @xmath28 code satisfies @xmath409 any ooa of strength @xmath410 and size @xmath411 satisfies @xmath412 this result is essentially due to p. delsarte .
however , for the nrt space it was first stated by martin @xcite and later rederived by bierbrauer @xcite .
the fact that the same polynomial gives a bound both on codes and orthogonal arrays is an easy consequence of delsarte s theory , first mentioned in levenshtein s work @xcite . in this section
we use theorem [ thm : lp ] to derive a bound on ordered codes and arrays .
its proof uses a `` spectral method '' first employed in @xcite for the grassmannian space and later used in @xcite to prove classical asymptotic bounds of coding theory .
the gist of the method can be explained as follows .
the polynomial @xmath413 is sought in the form @xmath414 where @xmath415 is a linear function that assures that @xmath416 in ( [ eq : limits ] ) and @xmath417 is a function that maximizes the fourier transform @xmath418 in the univariate case it turns out that a good choice for @xmath419 is a delta - function at ( or near ) @xmath32 .
an approximation of the delta - function is given by the ( dirichlet ) kernel @xmath385 which is its projection on @xmath420 .
we therefore seek to modify the operator @xmath360 so that @xmath385 becomes its eigenfunction with eigenvalue @xmath421 , express the bound of theorem [ thm : lp ] as a function of @xmath421 and optimize on @xmath278 within the limits ( [ eq : limits ] ) .
the reader is referred to @xcite for a more detailed discussion of these ideas .
[ thm : bound ] let @xmath278 be any degree such that @xmath422 for all shapes @xmath18 with @xmath423 where @xmath424 is the maximum eigenvalue of @xmath425 and @xmath426 is an integer .
let @xmath99 be an @xmath28 code . then @xmath427 let @xmath31 be a @xmath428 ooa of size @xmath30
. then @xmath429 consider the operator @xmath430 that equals @xmath360 on @xmath431 and acts on a function @xmath432 by @xmath433 where @xmath434 are some constants indexed by the partitions of weight @xmath278 ( their values will be chosen later ) .
the matrix of @xmath430 in the orthonormal basis equals @xmath435\ ] ] where @xmath436 is a matrix of order @xmath437 let @xmath24 be such that @xmath438 by perron - frobenius @xcite , the spectral radius @xmath439 is well defined and is an eigenvalue of ( algebraic and geometric ) multiplicity one of @xmath440 .
moreover , again using perron - frobenius , @xmath441 then @xmath442 where @xmath443 let @xmath444 be the eigenfunction of @xmath430 with eigenvalue @xmath445 let us write out the product @xmath446 in the orthonormal basis : @xmath447 where @xmath448 is a projection of the vector @xmath419 on the space @xmath449 this implies the equality @xmath450 where @xmath451 is defined after ( [ eq : cd ] ) . now take @xmath452 let us verify ( [ eq : limits ] ) . since multiplication by a function is a self - adjoint operator , we obtain @xmath453 using ( [ eq : pfgh ] ) one can easily check that @xmath454 for all @xmath455 the assumption of the theorem together with ( [ eq : lambda ] ) implies that @xmath456 for @xmath457 hence @xmath458 where in the last step we used the cauchy - schwarz inequality . computing the minimum on @xmath459 we obtain @xmath460 next , @xmath461\sqrt{v_h}.\ ] ]
let @xmath462 for some @xmath463 then using ( [ eq : v ] ) we find @xmath464 \sqrt{v_h}&= l_i\sqrt{(f_i+1)(n-\kappa)q^{i-1}(q-1)}\sqrt{v_h}\\ & = l_i\sqrt{(f_i+1)(n-\kappa)q^{i-1}(q-1 ) } \sqrt{v_f\frac{(n-\kappa)q^{i-1}(q-1)}{f_i+1}}\\ & = \big(1-\frac1 { q^{r - i+1}}\big)(n-\kappa)\sqrt{v_f}. \end{aligned}\ ] ] thus we have @xmath465 @xmath466 substitution of this expression into ( [ eq : f0 ] ) concludes the proof of ( [ eq : bound - r ] ) .
the bound ( [ eq : bound - ooa ] ) follows by ( [ eq : arraybounds ] ) . in this section
we derive an asymptotic lower bound on the spectral radius of @xmath468 this estimate will be later used to optimize the bound ( [ eq : bound - r ] ) on the choice of @xmath469 [ lemma : lambda ] @xmath470 where @xmath471 & + ( q-2)\tau_i ( q^{r}-q^{i-1 } ) + 2\frac{(q-1)}{q } \sum_{k=1}^{i-1}\sqrt{\tau_k\tau_i q^{i+k}}\big).\label{eq : g } \end{aligned}\ ] ] to prove this theorem , we will bound below the largest eigenvalue @xmath472 of the matrix @xmath473 for any real vector @xmath474 we have @xmath475 we will construct a suitable @xmath476-vector @xmath477 its coordinates are indexed by the partitions arranged in the increasing order of their length @xmath478 and lexicographically within a block of coordinates for each value of @xmath479 .
let @xmath480 where @xmath481 let @xmath482 be a shape vector .
for an integer @xmath483 consider the set @xmath484 and denote @xmath485 next , let @xmath486 for @xmath487 where @xmath181 will be chosen later , and @xmath488 otherwise . in the next two lemmas
we derive a lower bound on the part of the product @xmath489 that involves only the rows of @xmath490 that correspond to the shapes @xmath48 of length @xmath491 let @xmath492 be the index set of the nonzero off - diagonal elements in the row in @xmath493 which is indexed by @xmath494 .
let @xmath495)$ ] and let @xmath496.$ ] then @xmath497 i. since @xmath498 for every @xmath499 the quantity @xmath500 equals the number of ordered partitions of 0 into at most @xmath6 parts , each part bounded between @xmath501 and @xmath483 , or the number of ordered partitions @xmath502 the number of such partitions is given by @xcite : @xmath503 writing this expression as a polynomial in @xmath483 , we find the coefficient of @xmath504 to be @xmath505 since this is always positive satisfy the recurrence @xmath506 and then use induction to prove that @xmath507 @xmath508 according as @xmath509 or 3 mod 4 . ] , we conclude that @xmath24 is a degree-@xmath510 polynomial in @xmath483 ; in particular , if @xmath511 then also @xmath512 . \ii .
we have @xmath513.\ ] ] to bound @xmath514 below we estimate the difference between the above sum and the sum of all the nonzero elements of @xmath493 in the rows @xmath515 which is obtained by replacing the range of column indices @xmath516 above with @xmath517 therefore , for a given @xmath515 let us estimate the number @xmath518 of nonzero entries in @xmath519 $ ] not included in the sum .
let @xmath520 and let @xmath118 be of the form @xmath521 for some @xmath522 .
consider the column indices @xmath523 given by @xmath524 for any @xmath525 for any pair @xmath526 of this form , @xmath527\ne 0 $ ] but @xmath528 the number of shapes @xmath118 that result in shapes @xmath516 of the form ( [ eq:+ ] ) equals the number of ordered partitions of @xmath501 into at most @xmath529 parts of magnitude @xmath530 ; equivalently , this is the number of ordered partitions @xmath531 which equals @xmath532 next consider the row indices @xmath533 and column indices @xmath523 given by @xmath534 which again account for @xmath527\ne 0 $ ] and @xmath528 the number of such shapes @xmath118 equals the number of ordered partitions of @xmath535 into @xmath529 or fewer parts @xmath536 denote this number by @xmath537 note that as @xmath511 both @xmath538 and @xmath539 grow proportionally to @xmath540 it is easy to verify that @xmath541 if and only if @xmath118 and @xmath516 are of the described form .
observe that by ( [ eq : b_h ] ) , @xmath542 we then obtain @xmath543\ge \psi_\mu(m - r(r^2-r)(\pi_++\pi_- ) ) = \psi_\mu m(1-o_m(1 ) ) .
\end{aligned}\ ] ] the lemma is proved .
we now consider the part of the product @xmath544 that involves the matrix @xmath545 for a shape @xmath118 let @xmath546 the proof of the next lemma is very similar to the above proof and will therefore be omitted .
let @xmath547)$ ] and let @xmath548.$ ] @xmath549 to complete the proof of theorem [ lemma : lambda ] , compute @xmath550 where @xmath551 ( @xmath552 ) is the smallest of the numbers @xmath553 ( @xmath554 ) above .
note that both @xmath551 and @xmath552 are nonzero .
now let @xmath555 and let us choose @xmath48 in the definition of @xmath556 to be of the form @xmath557 we assume that none of the @xmath558 s approach @xmath559 as @xmath5 grows .
take @xmath560 using ( [ eq : b_h ] ) , and letting @xmath561 we get @xmath562 @xmath563 then since @xmath564 @xmath565 the theorem is proved .
theorems [ thm : bound ] and [ lemma : lambda ] together enable us to prove one of the main results of the paper .
let @xmath566 be the function defined parametrically by the relations @xmath567 then the asymptotic rate of any code family of relative distance @xmath158 satisfies @xmath568 and the rate of any family of ooas of relative strength @xmath158 satisfies @xmath569 to prove this theorem , take the logarithms in ( [ eq : bound - r ] ) and pass to the limit as @xmath570 using the standard asymptotics for the binomial coefficient , we find that the code rate is bounded above by the right - hand side of ( [ eq : boundr ] ) . the condition @xmath422 of the theorem [ thm : bound ] will be satisfied for large @xmath5 if @xmath571 this defines the function in ( [ eq : boundd ] ) .
thus , the proof is complete .
_ for @xmath26 this bound reduces to the linear programming bound on the rate of codes in @xcite . just as that result , the bound of this theorem improves upon the asymptotic plotkin bound for large values of the code distance .
in this section we prove a bound for codes in @xmath572 which improves upon the general result of the previous section .
the improvement is due to the fact that in the case @xmath65 it is possible to work with the polynomials @xmath573 in their explicit form , and base the bound on the behavior of their zeros instead of the spectral radius of the operator @xmath574 namely , let @xmath575 from ( [ eq : expl ] ) we have @xmath576 we also have @xmath577 we will use the following properties of the polynomials @xmath578 whose proofs are found for instance in @xcite .
let @xmath579 be the roots of @xmath578 in the ascending order .
then @xmath580 let @xmath581 then @xmath582 the krawtchouk polynomials satisfy the recurrence @xmath583 and a christoffel - darboux formula of the form ( [ eq : cd ] ) @xmath584 _ remark : _ properties ( [ eq : roots])-([eq : cdk ] ) are usually stated for integer @xmath5 .
this is related to the fact that the polynomials @xmath585 represent the eigenvalues of the hamming association scheme . as pointed out to us by m. aaltonen @xcite , it is possible to prove these properties for any @xmath586 relying on the generating function of the krawtchouk polynomials .
[ thm:2 ] the asymptotic rate of any family of codes of relative distance @xmath158 satisfies @xmath587 , where @xmath588 where the minimum is taken over all @xmath589 that satisfy @xmath590 the asymptotic rate of any family of ooas of relative strength @xmath158 satisfies @xmath591 the remainder of the section is devoted to the proof of this result .
we note that the polynomials @xmath573 are formed as products of two krawtchouk polynomials .
a similar situation arose in @xcite which dealt with the johnson association scheme whose ( second ) eigenvalues are equal to a product of a krawtchouk and a hahn polynomial .
therefore , we adopt some elements of the analysis in @xcite in our proof below . in quest of an lp bound , we require a polynomial @xmath592 that satisfies conditions ( [ eq : limits ] ) . consider the polynomial of the form @xmath593 for some @xmath594 and a subset @xmath595 .
for brevity below we write @xmath596 instead of @xmath597 $ ] and denote @xmath598 we find @xmath599 in order to ensure that @xmath600 we will choose @xmath595 and @xmath601 so that @xmath602 let @xmath603 be a shape that satisfies @xmath604 let @xmath594 satisfy @xmath605 for any @xmath606 denote by @xmath607 the degree such that @xmath608 by ( [ eq : roots ] ) , @xmath609 is well defined and implies the following : @xmath610 , \quad [ ( x_1(n - u , w)\le\beta ) \ ; \rightarrow \;(w\ge \phi(u)+1)].\ ] ] we choose the region @xmath595 to be given by @xmath611 for the moment this choice is not unique because there are many possibilities for @xmath181 .
this ambiguity will be later removed by optimizing the bound on the choice of @xmath181 . to argue about the sign of @xmath612 we need to establish some properties of the region @xmath595 .
first , we claim that for a fixed @xmath613 , @xmath614 indeed , by ( [ eq : roots ] ) , @xmath615 which implies the left - hand side of ( [ eq : ff ] ) . on the other hand ,
@xmath616 which implies the right - hand side . the values of @xmath617 for which @xmath618 are given in ( [ eq : b_h ] ) .
in particular , if @xmath619 , then the set @xmath620 of the shape vectors @xmath118 that index the nonzero matrix elements of @xmath376 and that lie outside the region @xmath595 is as follows : @xmath621 the region @xmath595 and the corresponding set @xmath620 are shown in fig .
[ fig : region_l ] . by our choice of the parameters , @xmath622
let us show that @xmath628 for all @xmath629 for this rewrite @xmath130 as follows : @xmath630 by ( [ eq : pfgh ] ) , the products @xmath631 and @xmath632 are expanded in the basis @xmath308 with nonnegative coefficients .
moreover , the other terms in the above formula also have the needed signs on account of ( [ eq : sgn1])-([eq : sgn3 ] ) .
this establishes our claim .
let us compute @xmath641 denote @xmath642 by ( [ eq : sgn ] ) and ( [ eq : mel ] ) we have @xmath643 let us now evaluate @xmath644 .
@xmath645 let us bound above the last sum .
we shall prove that @xmath646 indeed , using ( [ eq : k3term ] ) , we obtain @xmath647 recall that @xmath648 or @xmath649 in the former case , @xmath650 in the latter , @xmath651 on account of ( [ eq : beta ] ) and ( [ eq : roots ] ) . repeating this procedure @xmath652 times , we arrive at ( [ eq : in ] ) .
note that @xmath653 therefore @xmath654 by ( [ eq : cdk ] ) , ( [ eq : a ] ) , and ( [ eq : k0 ] ) we have @xmath655 @xmath656 where @xmath657 and @xmath658 using these expressions , we can bound @xmath644 as @xmath659 hence using ( [ eq : codebound ] ) , ( [ eq : f ] ) , and ( [ eq : f0 ] ) we can write @xmath660 since @xmath661 as @xmath188 ranges in between the bounds in ( [ eq : a ] ) , it is possible to find @xmath188 such that @xmath662 . with this choice and ( [ eq : v ] ) the last expression turns into ( [ eq : fb2 ] ) .
the estimate ( [ eq : ab2 ] ) follows from ( [ eq : arraybounds ] ) .
the proof of theorem [ thm:2 ] is obtained by passing to asymptotics in ( [ eq : fb2 ] ) .
namely , let @xmath663 by ( [ eq : limrt ] ) , @xmath664 computing the logarithm on the right - hand side of ( [ eq : fb2 ] ) , we observe that the only term of exponential growth arises from @xmath665 . using standard estimates we obtain @xmath666 the tightest bound is obtained by computing the minimum of this expression on @xmath589 .
the range of the variables @xmath667 is obtained on observing that @xmath668 and @xmath669 are the maximizing values of @xmath670 for large @xmath5 ( by a direct calculation from the above expression ; or , specializing from a general result in @xcite ) .
the third restriction in the statement of the theorem is implied by @xmath671 this completes the proof .
the nrt metric is an example of a wide class of metrics on the set @xmath672 termed poset metrics following brualdi et al .
@xcite . to define a poset metric ,
consider a partial order @xmath673 on the set @xmath674 $ ] .
ideal _ in the order is a subset closed under the @xmath675 relation .
the @xmath673-weight of a vector @xmath676 is the size of the smallest ideal that contains the nonzero entries of @xmath677 for instance , for the nrt weight , the relation @xmath675 can be defined as @xmath678 iff @xmath679 where @xmath680 are the indices of the block .
a _ dual order _
@xmath681 is formed of the same set of chains as the order @xmath673 but with the signs reversed within each chain . in @xcite , and in our paper
, the poset duality is realized as the @xmath682 convention .
one of the main questions that arises in this context is to characterise the association scheme that arises from the order and in particular , to derive the macwilliams - type relations .
partial orders that give rise to a univariate macwilliams relation have been described by kim and oh @xcite .
on the other hand , rather little is known about the multivariate case of which the nrt space is an instance . , a functional view of upper bounds on codes , in y. li et al . , eds .
, _ coding and cryptology _ , singapore : world scientific ( 2008 ) , pp .
j. bierbrauer , y. edel , and w. ch .
schmid , _ coding - theoretic constructions of @xmath59-nets and ordered orthogonal arrays _ , j. combin
10 * ( 2002 ) , no . 6 , 403418 . j. l. massey and s. serconek , _ linear complexity of periodic sequences : a general theory _
, `` advances in cryptology '' ( crypto 96 ) , lecture notes in comput .
, vol . *
1109 * , springer , berlin , 1996 , pp
. 358371 . | we derive new estimates of the size of codes and orthogonal arrays in the ordered hamming space ( the niederreiter - rosenbloom - tsfasman space ) .
we also show that the eigenvalues of the ordered hamming scheme , the association scheme that describes the combinatorics of the space , are given by the multivariate krawtchouk polynomials , and establish some of their properties . dept . of electrical and computer engineering and institute for systems research ,
university of maryland , college park , md 20742 , usa , and institute for problems of information transmission , russian academy of sciences , moscow , russia .
email : [email protected] . ] dept . of electrical and computer engineering ,
university of maryland , college park , md 20742 .
email : [email protected] . ] |
given a very large dataset of time series ( presumably with some underlying , non - stationary time dependence between each other ) up to time @xmath0 , predict whether a time series @xmath1 will have an anomaly ( defined formally in section [ sec : detecting ] ) at time @xmath2 , moreover when the data stream arrives at time @xmath2 detect anomalies conclusively using a low latency model .
[ sec : introduction ] the underlying motivation of this piece is the usefulness of anomaly prediction in mission critical components .
a fast anomaly detection platform can by itself be extremely useful for ensuring the reliability of a system , as a result , this problem has been studied extensively by academics and industry experts .
we defer to the authors in @xcite for a literature survey of anomaly detection techniques . in this work ,
we extend the scope of our approach and goal to tackle anomaly prediction .
most work in the literature has narrowed their focus on anomaly detection because it is in and of itself a very challenging problem .
however , the benefits of an even slight temporal advantage in prediction can have huge impacts on the performance of a system .
the contribution of this work can be categorized in two components : * a highly accurate , low - latency anomaly detection system(described in detail in section [ sec : model ] ) where we seek to improve on some of the concepts and techniques introduced in @xcite . *
an novel approach to anomaly prediction by modeling a bayesian network structured based on the coefficients of the lag regressors in the anomaly detection system , coupled with bayesian parameter learning to model the conditional dependency structure between the time series .
our ( non - public ) data set consists of @xmath3 time series @xmath4 , sampled every minute for the past year with some underlying non - stationary dependence between each other and human labeled anomaly events by domain experts ( i.e. , the owner of a time series flagged the @xmath5 time series @xmath6 at time point @xmath7 as an anomaly .
although the dataset has some time series source information , our anomaly detection model makes no assumptions about the inter - independence of the time series to allow us to solve the more general problem of having no apriori knowledge . to illustrate the difficulty of the problem we are trying to solve , in figure [ fig : sample ] we show a subsample of 300 points of one of the time series with labels anomalies denoted by red dots .
clearly the problem involves latent relationships between time series ; observing one time series in isolation is not enough , even for a human , to determine whether a time point is anomalous .
another challenge that we face is that the time series have vastly different probability distributions . to show this
, we normalize the time series and approximate the probability distribution of each time series using kernel density estimation ( with a gaussian kernel and hyperparameter search on the bandwidth ) .
figure [ fig : kde ] shows the estimated probability densities of the 19 most important time series .
a foundational tenet of our model is that our time series have a latent dependence structure between each other . to validate this assumption , for each pair of time series @xmath8 and @xmath9 ( from a selected set of 19 ) , we use kernel density estimation to approximate the marginal distributions of the times series , @xmath10 and @xmath11 , as well as the joint probability distribution @xmath12 .
then we compute the mutual information between the approximate distributions : @xmath13 note that mutual information is a measure of the inherent dependence expressed in the joint distribution of @xmath14 and @xmath15 relative to the joint distribution of @xmath14 and @xmath15 under the assumption of independence , in a concrete sense @xmath16 if and only if x and y are independent random variables .
figure [ fig : mi ] shows a grid plot of the mutual information between 19 selected time series , the @xmath17 coordinate of the figure denotes @xmath18 . as can be seen from the figure below , the time series are highly dependent between each other , moreover , the dependence structure is not uniform across pairs . although we could use a model like this to try to determine latent dependence structure , this method is not scalable as approximating joint probability distributions can be computationally intensive . in section [ sec : model ]
we propose a method that will capture latent dependence structures in a computationally feasible way .
combining previous techniques for anomaly detection @xcite and time series modeling @xcite , we propose an auto regressive distributed lag model on @xmath19 occurs over time rather than all at once . ] with @xmath20 regularization , that is horizontally scalable .
we first define the following : * @xmath21 : number of time series * @xmath22 : value of time series @xmath23 at timestamp @xmath24 * @xmath25 : the window size used in the auto regression * @xmath26 : the length of the time series our model is @xmath27 where @xmath28 is the prediction of model @xmath23 at time @xmath0 , @xmath29 is the least squares coefficient vector , and @xmath30 is the model s error on time series @xmath23 at time @xmath0 . in other words , we use all past information from all time series @xmath31 where @xmath32 , @xmath33 for the prediction of the current value @xmath28 in each time series @xmath34 .
now consider concatenating all the predictions for time series @xmath23 ( there are @xmath35 predictions ) . for each time series @xmath23 , we can write the @xmath35 regression predictions as @xmath0 varies in the matrix form : @xmath36 where @xmath37 are the vector of predictions and error term , @xmath38 and @xmath39 .
x is the matrix of concatenated lag regressors .
the model is trained using ordinary least squares method with @xmath20-regularization . in other words , for each time series @xmath34 , we find the parameter vector @xmath29 of length @xmath40 to minimize @xmath41 ^ 2 + \lambda \|\beta^{(i)}\|_1 \\ & = \|p^{(i ) } - x \beta^{(i)}\|_{2}^{2 } + \lambda \|\beta^{(i)}\|_1 \end{split } \label{eq : loss}\ ] ] here , @xmath42 is the error function and @xmath43 is the hyper - parameter which controls the severity of the penalty on complex models .
we do this for each of @xmath21 time series , so we will end up applying ordinary least squares @xmath21 times .
this model offers a sparse regression @xcite that selects only the time series that have predictive value and identifies the temporal correlation between these time series .
the bias and size of the resulting model can be set by modifying the @xmath20 regularization parameter @xmath43 .
this is crucial since the smaller the model is , the less coefficients it has , and thus , the less data it needs to make predictions .
since retrieving data over a network has latency , and low latency when detecting anomalies is a priority in this algorithm , the size of the data to be retrieved should be as small as possible without greatly affecting the model s accuracy .
finally , we store the estimated standard error corresponding to each of the @xmath21 time series to be used later for anomaly detection : @xmath44 here , @xmath45 is the number of non - zero coefficients .
this estimator is motivated by the unbiased estimator for the error variance in a simple linear regression without regularization .
the performance of its can be seen in reid , tibshirani , and friedman @xcite .
the model described in section [ sec : model ] makes real time predictions of all time series in the system as it streams new data . in order to detect anomalies , we compared the prediction the model does at current time @xmath46 , @xmath47 , with its real value coming from the data stream , @xmath48 by running a t - test with the t - statistic defined in equation ( [ eq : t_one_example ] ) : @xmath49 this test gives us a p - value we use to compare against a given threshold p - value - threshold , which is a parameter of the model .
if the p - value we get with t - statistic ( [ eq : t_one_example ] ) is smaller than p - value - threshold , then we are in presence of an anomaly and an alert should be raised .
the parameter p - value - threshold is probably the most important parameter in the model because it directly specifies how sensitive the model is in detecting anomalies .
usually its value is very low .
a value of around 1e-5 would roughly detect an anomaly every 100,000 minutes .
it s important to note that this detection procedure has very low latency because the model is sparse and it only needs to make few calculations in the linear combination .
this is a key feature that puts this model ahead of lower latency , but accurate , models like dpca @xcite .
there is another aspect of real life systems that has not been covered so far . since in real life systems
, random and short spikes can be normal and not anomalies , we use a smoothed anomaly definition as opposed to the naive comparison .
concretely , since these spikes may result in false positive classification of anomalies instead of testing for an anomaly with just the current data - point @xmath48 , we test for an anomaly with a sample of data - points .
this sample of size @xmath50 for time series @xmath23 is defined as @xmath51\right\}.\ ] ] our model checks the probability that @xmath52 comes from a normal distribution with mean @xmath53 by running a t - test with the t - statistic defined by @xmath54 here , @xmath55 is the sample @xmath52 standard error .
as a baseline to highlight the difficulty of the problem we are trying to solve , we implement and test a model that characterizes the time series as multivariate gaussian . more concretely , we build a generative model where time series are generated from an @xmath45-dimensional multivariate gaussian distribution and then set a threshold of the probability of a time series deviating from its mean .
we can visualize this method in figure [ gaussian ] . the gaussian detector . ] again , highlighting the difficulty of the problem that we are trying to solve , we present an oracle where we use the test set as the training set and calculate the test error , " since we know that the test error should be bounded by the training error , this oracle will allow us to understand the difficulty of our problem definition .
the results shown in section iii , compare the error of the oracle formulation above .
we compare our model s performance with our baseline model ( multivariate gaussian @xcite ) and with the state of the art dynamic principal components model ( dpca ) @xcite .
we run the models on a year worth of server logs data , aggregated as @xmath3 time series . as noted in section [ sec : data ] , the data are labeled by humans ( with domain knowledge ) as anomaly or not anomaly .
we use an auto regression window size @xmath25 of one week .
the output of each model by minute ( anomaly or not anomaly class ) is compared against the real label of the data .
finally , we calculate the @xmath56 score to evaluate the performance of each model .
.accuracy [ cols="<,>",options="header " , ] table [ table : latency ] shows that our model is faster than the very accurate dpca .
it is also faster than the gaussian model , which is explained by the fact that our model is sparse and does not need to load all the data in the auto - regression window , just the data points that are highly correlated .
since low latency is a crucial component in real time applications , we believe our model s combination between good accuracy and extremely low latency is very powerful .
the second part of this piece deals with building a bayesian network to characterize the probability of an anomaly . recall that for every time series
@xmath23 we run a distributed lag regression ( section [ sec : model ] ) where the regressors are lag variables of @xmath23 and every other time series @xmath32 . the model for time series
@xmath23 is parametrized by a vector @xmath57 of coefficients on the regressors ( the @xmath20 regularization term ensures that the majority of these weights are set to zero ) .
we denote by @xmath58 the vector of non - zero weights for time series @xmath23 , and by @xmath59 the non - zero weight component for time series @xmath23 that corresponds to lag regressors of time series @xmath7 with a lag of @xmath24 ( as in section [ sec : model ] ) similarly , let @xmath60 correspond to the lag _ variables _ with non - zero weights for time series @xmath23 regressing on lag regressors of time series @xmath7 where the lag for that regressor is @xmath24 . let the length of this vector of variables be @xmath61 .
the random variables in our network include : @xmath62 for @xmath63 ; @xmath64 for @xmath63 ; and @xmath60 for @xmath63 , @xmath32 , and @xmath65 .
note that although there may be overlap between @xmath66 and @xmath67 for @xmath68 , our bayesian network considers the set of all these variables .
the lag variables @xmath60 represent the value of the lagged time series , their domain is continuous .
the @xmath69 variables take values @xmath70 representing the presence of an anomaly at time @xmath2 and lastly the @xmath62 variables denote the how much our prediction deviates from the true value of the time series ( see section 6.4 ) .
we can visualize the bayes net in figure [ bayesnet ] .
/[count= ] in 1,2,3 ( input- ) at ( 0 , 7 - 2 * ) @xmath71 ; /in missing ( input- ) at ( 0 , 7 - 2 * 4 ) ; /in 4 ( input- ) at ( 0 , 7 - 2 * 5 ) @xmath72 ; /[count= ] in 1,2,3 ( hidden1- ) at ( 3 , 7 - 2 * ) @xmath73 ; /in missing ( hidden1- ) at ( 3 , 7 - 2 * 4 ) ; /in 4 ( hidden1- ) at ( 3 , 7 - 2 * 5 ) @xmath74 ; in cont , cont , cont , missing , cont ( hidden2- ) at ( 5 , 7 - 2 ) ; /[count= ] in 1,2,3 ( hidden3- ) at ( 7 , 7 - 2 * ) @xmath75 ; /in missing ( hidden3- ) at ( 7 , 7 - 2 * 4 ) ; /in 4 ( hidden3- ) at ( 7 , 7 - 2 * 5 ) @xmath76 ; in 1 ( output- ) at ( 11 , 7 - 2 ) @xmath62 ; in 1 ( anomaly- ) at ( 13 , 7 - 2 ) @xmath64 ; ( output-1 ) ( anomaly-1 ) ; iin 1, ... ,4 ( hidden3-i ) ( output-1 ) ; iin 1, ... ,4 ( hidden1-i ) to[out=-15,in=-90 ] ( output-1 ) ; iin 1, ... ,4
( input - i ) to[out=40,in=120 ] ( output-1 ) ; the structure of the bayesian network is defined by adding , for each @xmath63 , an edge from node @xmath60 to node @xmath62 for @xmath32 , @xmath65 , and an edge from @xmath62 .
fan and li @xcite show that under the approximation : @xmath77 the unconstrained formulation of the convex optimization problem as equation ( [ eq : loss ] ) can be solved by iterating : @xmath78 where @xmath79 is the estimate of @xmath29 .
moreover , fan and li @xcite draw from the theory of k - step estimators in @xcite which proves that , with a good starting parameter , a on - step iteration can be as efficient as the penalized maximum likelihood estimator , when the newton - rapshon algorithm is used .
we have @xmath80 \left(x \beta^{(i ) } + \epsilon^{(i)}\right).\ ] ] thus , there is a direct relation between @xmath81 and @xmath62 .
as @xmath81 determines the possibility of anomaly , the derived relation gives rise to the appearance of the node @xmath62 s in the bayesian network . the generative model presented above proved to have many challenges . for concreteness , we highlight some of them .
our initial attempt was discretizing the continuous random variables and using maximum likelihood method , an approach which is sometimes used in literature ( @xcite p. 186 ) .
note however that discretization provides a trade - off between accuracy of the approximation and cost of the computation ( @xcite p. 607 ) .
work done using this approach proved to be nonviable . for concreteness , note that the average number of parents ( @xmath60 s ) for each time series @xmath23 are @xmath82 . even for
a coarse discretization scheme of 10 bins , each local cpd table would have @xmath83 entries , even with the decomposibility of factors in bayesian networks , the amount of data needed to accurately estimate the maximum likelihood ( not to mention the computational complexity of collecting the sufficient statistics ) would be intractable .
even if we could estimate the local maximum likelihood , the complexity of many inference algorithms is exponential in the size of the domain of the conditional probability distributions ( e.g. variable elimination ) .
one possible approach that we are exploring is tuning the hyperparameter of the cost of regularization term to restrict the number of variables in each time series to be used in the bayesian network .
another approach we are exploring is using a hybrid model , which contains both continuous and discrete random variables .
this structure in and of itself presents many challenges ( see , e.g. , @xcite , chapter 14 ) . in parcticular
, we can show that inference on this class of networks is _ np - hard _ , even when the network structure is a polytree ( @xcite p. 615 ) . solving the full bayesian network model is super complicated as discussed above .
thus , we solve a simplified version of this problem .
specifically , we ignore the prior distribution and use the conditional distribution as the posterior distribution .
also , the relation between @xmath62 and @xmath64 is linear .
the detailed procedure for characterizing anomalies is provided below .
we focus on detecting anomalies for one time series only ; other time series in the bayesian network should work in the similar way .
* step 1 : we normalize @xmath62 to be in the @xmath84 interval .
+ @xmath85 + where @xmath86 is the minimum over all observed values of @xmath62 and @xmath87 is the maximum over those values .
the constants @xmath88 and @xmath89 make the normalized @xmath62 fall strictly in @xmath84 .
* step 2 : we specify the relation between @xmath62 and @xmath60 s as a linear gaussian model .
more precisely , we determine the relation @xmath90 this relation can be rewritten as @xmath91 which can then be solved by least square method .
* step 3 : we specify the probability that @xmath92 given @xmath62 that @xmath93 then @xmath64 is characterized as @xmath88 , which is anomaly , if and only if this probability is greater than @xmath94 and @xmath95 otherwise .
the constant @xmath96 can be chosen using cross - validation so as to maximize the @xmath97 score . in this section
we use four time series as our data set in order to characterize anomalies for one of these four time series .
the reason we drastically reduce the size of the data is because matrix @xmath38 ( as defined in equation ( [ eq : loss ] ) ) and limiting the size of our data set was the only way to be able to compute this in a single machine .
our data set has the following size : * @xmath98 ( 30 days of minute data ) * @xmath99 ( 5 days of minute data ) * @xmath100 ( four time series ) with this data set we generate a matrix @xmath14 of float elements and dimensions @xmath101 rows by @xmath102 columns .
this matrix has a size of roughly 17 gb ( depending on the platform used ) .
finally , we obtain @xmath103 and the corresponding @xmath97 score is @xmath104 .
this value is certainly not as good as that of dpca or our original discriminative anomaly detection approach , but the comparison is nuanced since above we only processed one time series with a simplified bayesian network ( because of run time complexity ) .
notwithstanding , this result is expected ; discriminative models will typically outperform generative models when the relationships expressed by the generative model only approximates the true underlying generative process ( particularly for our case where we follow some approximations to simplify the model complexity ) in terms of classification error rate .
however , the advantage of our novel approach is that our model is richer . as a generative model
, we can make explicit claims about the process that underlies the time series .
future work may included running inference queries on the generative model ( after tuning it for better performance ) to better understand the underlying process where the data is coming from .
equation ( [ eq : loss ] ) proposes a loss function @xmath42 composed by an @xmath105 error and an @xmath20 regularization .
we think it is worth to test if an @xmath20 error and an @xmath20 regularization performs better .
the reason behind this is that an @xmath20 error is more biased against outliers .
if our regression @xmath106 is less affected by the anomalies in the training data set , then it will predict anomalies more accurately in the testing data .
also , the model proposed in section [ sec : model ] may show false positives when the predictors have an anomaly .
this is because after an anomaly exists , it still stays in the data that will be later use as an input in out regression .
we did not see this in our experiment , but it is something we would like to test and evaluate . | we develop a supervised machine learning model that detects anomalies in systems in real time .
our model processes unbounded streams of data into time series which then form the basis of a low - latency anomaly detection model . moreover , we extend our preliminary goal of just anomaly detection to simultaneous anomaly prediction .
we approach this very challenging problem by developing a bayesian network framework that captures the information about the parameters of the lagged regressors calibrated in the first part of our approach and use this structure to learn local conditional probability distributions . |
much effort has been devoted to confirm the presence of black holes in x - ray binary systems @xcite as well as in galactic centers @xcite .
the analysis of the radiation emitted from accretion disks swirling around black hole candidates may play a crucial role in the experimental confirmation of the existence of event horizons ( see , e.g. , ref .
it is interesting , thus , to compute the amount of the emitted radiation which is able to reach asymptotic observers rather than be absorbed by the hole . in a recent work , higuchi and
two of the authors analyzed the radiation emitted by a scalar source rotating around a schwarzschild black hole @xcite . in the present work we study
the more realistic case where the _ scalar source _ is replaced by an _ electric charge_. in order to capture the full influence of the spacetime curvature on the emitted radiation , we work in the context of quantum field theory in schwarzschild spacetime ( for a comprehensive account on quantum field theory in curved spacetimes see , e.g. , ref .
@xcite ) . because of the difficulty to express the solution of some differential equations which we deal with in terms of known special functions ( see e.g. ref .
@xcite for a discussion on this issue ) our computations are performed ( i ) numerically but without further approximations and ( ii ) analytically but restricted to the low - frequency regime ( in which case the radial part of the normal modes can be written in terms of legendre functions @xcite ) .
we organize the paper as follows . in section [ sec : quantization ]
we review the gupta - bleuler quantization of the electromagnetic field in a modified feynman gauge in the spacetime of a static chargeless black hole @xcite .
we compute the radiated power from an electric charge swirling around a schwarzschild black hole in section [ sec : rotating ] . in section [ sec : comparison ]
we compare this result with the one obtained considering the charge as orbiting a newtonian object in flat spacetime .
finally we use the previous results to compute in section [ sec : absorption ] what is the amount of the emitted radiation which is able to reach asymptotic observers .
our final remarks are made in section [ sec : final ] .
we assume natural units @xmath0 and metric signature @xmath1 .
in this section we review the quantization of the massless vector field in schwarzschild spacetime following closely ref . @xcite .
we write the line element of a static chargeless black hole as @xmath2 where @xmath3 .
we then consider a massless vector field in this geometry with classical action given by @xmath4 where the lagrangian density in the modified feynman gauge is @xmath5 \label{lagdens}%\ ] ] with @xmath6 , @xmath7 and @xmath8 the corresponding euler - lagrange equations are , thus , @xmath9 which can be cast in the form @xmath10 + \dfrac { 1}{r^{2}}\tilde{\nabla}^{2}a_{r } \nonumber \\ & & + \dfrac{1}{f}\partial_{r}\left ( \dfrac { f}{r^{2}}\right ) \tilde{\nabla}^{a}a_{a}=0\ , , \label{eqr } \\ & & \dfrac{1}{f}\partial_{t}^{2}a_{a}-\partial_{r}\left ( f~\partial_{r}% a_{a}\right ) + \dfrac{1}{r^{2}}\left [ \tilde{\nabla}^{b}\left ( \tilde { \nabla}_{b}a_{a}-\tilde{\nabla}_{a}a_{b}\right ) \right . \nonumber \\ & & \left .
+ \partial_{a}\tilde{\nabla}^{b}a_{b}\right ] + r^{2}\partial_{r}\left ( \dfrac{f}{r^{2}}\right ) \partial_{a}a_{r}=0\ , .
\label{eqang}\end{aligned}\ ] ] here @xmath11 and @xmath12 denote angular variables on the unit @xmath13-sphere @xmath14 with metric @xmath15 and inverse metric @xmath16 [ with signature @xmath17 $ ] , @xmath18 is the associated covariant derivative on @xmath14 , @xmath19 and @xmath20 .
we write the complete set of positive - frequency solutions of eq .
( [ eqmo ] ) with respect to the killing field @xmath21 in the form @xmath22 the index @xmath23 stands for the four different polarizations .
the pure gauge modes , @xmath24 , are the ones which satisfy the gauge condition @xmath25 and can be written as @xmath26 , where @xmath27 is a scalar field .
the physical modes , @xmath28 , satisfy the gauge condition and are not pure gauge .
the nonphysical modes , @xmath29 , do not satisfy the gauge condition .
the modes incoming from the past null infinity @xmath30 are denoted by @xmath31 and the modes incoming from the past event horizon @xmath32 are denoted by @xmath33 .
the @xmath34 and @xmath35 are the angular momentum quantum numbers .
the physical modes can be written as @xmath36 \partial_{\theta}y_{lm } , \dfrac{f}{l\left ( l+1\right ) } \dfrac{d}{dr}\left [ r{\varphi}_{\omega l}% ^{{\rm i}n}\left ( r\right ) \right ] \partial_{\phi}y_{lm } \right)e^{-i\omega t}% \label{modo i}\ ] ] and @xmath37 with @xmath38 ( since the gauge condition @xmath25 is not satisfied for @xmath39 ) .
the radial part of the physical modes satisfies the differential equation @xmath40 + f\dfrac{d}{dr}\left ( f\dfrac{d}{dr}\left [ r{\varphi}_{\omega l}^{\lambda n}\left ( r\right)\right ] \right ) = 0\text { , } \label{equacao em q}%\ ] ] where @xmath41 and @xmath42 is the schwarzschild scattering potential ( see solid line in fig . [ figure1 ] ) . @xmath43 and @xmath44 are scalar and vector spherical harmonics @xcite , respectively .
the remaining modes can be written as @xmath45 and @xmath46 with @xmath47 , where @xmath48 and @xmath49 satisfies @xmath50 the conjugate momenta associated with the field modes are defined by @xmath51 } \right| _ { a_\mu = a^{(i ) } _ { \mu } } \nonumber \\ & = & -\left.\left [ f^{\mu\nu}+g^{\mu\nu } g\right]\right| _ { a_\mu = a^{(i ) } _ { \mu } } \text { , } \label{momento s}\end{aligned}\ ] ] where @xmath52 represents @xmath53 . by writing the conserved current @xmath54 \equiv i\left [ \overline{a_{\nu}^{(i ) } }
\pi^{(j)\mu\nu}-\overline{\pi^{(i)\mu\nu}}a_{\nu}^{(j)}\right ] \ , , \label{corrente w}\ ] ] where the overline denotes complex conjugation , we normalize the field modes through the generalized klein - gordon inner product @xcite defined by @xmath55 \text{. } \label{produto interno de kg generalizado}\ ] ] here @xmath56 , where @xmath57 is the invariant 3-volume element of the cauchy surface @xmath58 and @xmath59 is the future pointing unit vector orthogonal to @xmath58 .
the modes are then normalized such that @xmath60 where the matrix @xmath61 is given by @xmath62{cccc}% 1 & 0 & 0 & 0\\ 0 & 1 & 0 & 0\\ 0 & 0 & 0 & -1\\ 0 & 0 & -1 & -1 \end{array } \right ) \label{matriz}\ ] ] with @xmath63 .
-2 truecm -2 truecm in order to quantize the electromagnetic field , we demand the equal time commutation relations @xmath64 = \left [ \hat{\pi}_{\mu t}\left ( \mathbf{x},t\right ) , \hat{\pi}_{\nu t}\left ( \mathbf{x}^{\prime},t\right ) \right ] = 0\text { , } \label{comutador i}%\ ] ] @xmath65 = \dfrac{i\delta_{\mu}^{\nu}}{\sqrt{-g}% } \delta^{3}\left ( \mathbf{x}-\mathbf{x}^{\prime}\right ) \text{. } \label{comutador ii}%\ ] ] the electromagnetic field operator can be expanded in terms of the normal modes as @xmath66 \
, , \label{expansao dos modos s}%\ ] ] where @xmath67 and @xmath68 are the annihilation and creation operators , respectively , satisfying @xmath69\!\ ! = \ ! ( m^{-1})_{\varepsilon\varepsilon^{\prime}}% \delta_{nn^{\prime}}\delta_{ll^{\prime}}\delta_{mm^{\prime}}\delta\left ( \omega-\omega^{\prime}\right).\ ] ] the fock space of the physical states @xmath70 is obtained by imposing the gupta - bleuler condition @xcite . in our case
, this corresponds to impose @xmath71 where @xmath72 is the positive - frequency part of the operator @xmath73 .
condition ( [ gb ] ) corresponds to @xmath74 the physical states are obtained by applying any number of creation operators @xmath75 , @xmath76 and @xmath77 to the boulware vacuum @xmath78 @xcite defined by @xmath79 the creation operators associated with pure gauge modes take physical states into nonphysical ones .
moreover physical states of the form @xmath80 have zero norm .
therefore we can take as the representative elements of the fock space those states obtained by applying the creation operators associated with the two physical modes to the boulware vacuum .
for this reason we will be concerned only with the two physical modes , @xmath41 , in the rest of the paper .
( a more detailed discussion of the gupta - bleuler quantization of the electromagnetic field in spherically symmetric and static spacetimes can be found in ref .
@xcite . ) the solutions of eq .
( [ equacao em q ] ) are functions whose properties are not well known .
( see ref .
@xcite for some properties . )
we can , however , obtain their analytic form ( i ) in the asymptotic regions for any frequency and ( ii ) everywhere if we keep restricted to the low - frequency regime . in order to study the asymptotic behavior of the physical modes we use the wheeler coordinate @xmath81 and
rewrite eq .
( [ equacao em q ] ) as @xmath82 + \dfrac{d^{2}}{dx^{2}}\left [ r{\varphi}_{\omega l}^{\lambda n}\left ( x\right ) \right ] = 0\text{. } \label{eq tipo schrodinger}%\ ] ] since the schwarzschild potential ( [ potencial em s ] ) vanishes for @xmath83 and decreases as @xmath84 for @xmath85 ( see fig .
[ figure1 ] ) , the solutions of eq .
( [ eq tipo schrodinger ] ) can be approximated in the asymptotic regions by @xmath86{c } b_{\omega l}^{\lambda\rightarrow}\left ( e^{i\omega x}+r_{\omega l } ^{\lambda\rightarrow}e^{-i\omega x}\right ) \\
b_{\omega l}^{\lambda\rightarrow}t_{\omega l}^{\lambda\rightarrow } i^{l+1}\omega xh_{l}^{\left ( 1\right ) } \left
( \omega x\right ) \end{array } \right .
\begin{array } [ c]{c } \left ( x\ll -1\right ) \text{,}\\ \left ( x\gg 1\right ) \text { , } \end{array } \label{solucao assintotica vai}\ ] ] and @xmath87{c } b_{\omega l}^{\lambda\leftarrow}t_{\omega l}^{\lambda\leftarrow}e^{-i\omega x}\\ b_{\omega l}^{\lambda\leftarrow}\left ( \left ( -i\right ) ^{l+1}\omega xh_{l}^{\left ( 1\right ) \ast}\left ( \omega x\right ) \right .
\\ \left . + r_{\omega l } ^{\lambda\leftarrow}i^{l+1}\omega xh_{l}^{\left ( 1\right ) } \left
( \omega x\right ) \right ) \end{array } \right .
\begin{array } [ c]{c } \left ( x\ll -1\right ) \text{,}\\ \ , \\ \ , \\ \left ( x\gg 1\right ) \text { , } \end{array } \label{solucao assintotica vem}\ ] ] where @xmath88 and @xmath89 are solutions incoming from @xmath32 and @xmath30 , respectively . here
@xmath90 is a spherical bessel function of the third kind @xcite , @xmath91 are normalization constants , and @xmath92 are the reflexion and transmission coefficients , respectively , satisfying the usual probability conservation equation @xmath93 . using the generalized klein - gordon inner product defined above
we obtain @xmath94 and @xmath95 let us now find the analytic expressions of the physical modes in the low - frequency approximation .
for this purpose we rewrite eq .
( [ equacao em q ] ) as @xmath96 \label{qz } % \nonumber \\ & & + \left [ l\left ( l+1\right ) -\dfrac { 2}{z+1}-\omega^{2}m^{2}\dfrac{\left ( z+1\right ) ^{3}}{z-1}\right ] { \varphi}_{\omega l}^{\lambda n}\left ( z\right ) = 0\text{,}% \nonumber\end{aligned}\ ] ] where @xmath97 . in the low frequency regime
, we write the two independent solutions of eq .
( [ qz ] ) for @xmath38 as @xmath98 \label{q q}\ ] ] and @xmath99 \text { , } \label{q p}\ ] ] where @xmath100 and @xmath101 are legendre functions of the first and second kind @xcite , respectively , and @xmath102 are normalization constants .
we note that since @xmath103 and @xmath104 for @xmath105 and @xmath106 and @xmath107 for @xmath108 , we obtain from eqs .
( [ q q ] ) and ( [ q p ] ) that @xmath109 diverges in @xmath32 and remains finite in @xmath30 , whereas @xmath110 diverges in @xmath30 and remains finite in @xmath32 .
this is the reason why we have associated @xmath111 and @xmath100 with modes incoming from @xmath32 and @xmath30 , respectively .
now , by fitting asymptotically eqs .
( [ q q ] ) and ( [ q p ] ) with eqs .
( [ solucao assintotica vai ] ) and ( [ solucao assintotica vem ] ) , respectively , we obtain that the normalization constants are ( up to arbitrary phases ) @xmath112 @xmath113 @xmath114 and @xmath115
now let us consider an electric charge with @xmath116 , @xmath117 and angular velocity @xmath118 ( as defined by asymptotic static observers ) , in uniform circular motion around a schwarzschild black hole , described by the current density @xmath119 here @xmath120 is the coupling constant and @xmath121 is the charge s 4-velocity .
we note that @xmath122 is conserved , @xmath123 , and thus @xmath124 for any cauchy surface @xmath58 .
next let us minimally couple the charge to the field through the action @xmath125 then the emission amplitude at the tree level of one photon with polarization @xmath23 and quantum numbers @xmath126 into the boulware vacuum is given by @xmath127 it can be shown that @xmath128 .
this implies that only photons with frequency @xmath129 are emitted once the charge has some fixed @xmath130 .
one can also verify that the pure gauge and nonphysical modes have vanishing emission amplitudes .
this is so for the pure gauge modes because @xmath131 and for the nonphysical modes because they have zero norm .
the total emitted power is @xmath132 where @xmath133 is the total time as measured by the asymptotic static observers .
using now eqs .
( [ modo i])-([modo ii ] ) and ( [ js])-([us ] ) we rewrite eq .
( [ pot s ] ) as @xmath134 \label{pot sch}\ ] ] with @xmath135 ^{2 } } \left ( 1-\dfrac{2m}{r_{s}}\right ) ^{2 } \nonumber \\ & & \times\left [ \dfrac{d}{dr_{s}}\left [ r_{s}~{\varphi}_{\omega_{0 } l}^{{\rm i}n}\left ( r_{s}\right ) \right ] \right ] ^{2}\left\vert y_{lm}\left ( \pi/2,0\right ) \right\vert ^{2 } \label{potencia is}\end{aligned}\ ] ] and @xmath136 ^{2 } \nonumber \\ & & \times \left\vert y^{lm}_\phi \left ( \pi/2,0\right ) \right\vert ^{2}\text{. } \label{potencia iis}%\end{aligned}\ ] ] -2 truecm -2 truecm let us now relate the radial coordinate @xmath137 of the rotating charge with its angular velocity @xmath138 .
according to general relativity for a stable circular orbit around a schwarzschild black hole we have @xcite @xmath139 we use this relation to compute numerically the emitted power given by eqs .
( [ pot sch])-([potencia iis ] ) as a function of @xmath138 .
the numerical method used here is analogous to the one described in ref .
the result is plotted as the solid line in fig .
[ figure2 ] .
the main contribution to the emitted power comes from modes with angular momentum @xmath140 .
the larger the @xmath34 , the less is the contribution to the total radiated power . for a fixed value of @xmath34 ,
the dominant contribution comes from @xmath141 . performing the summation up to @xmath142 in eq .
( [ pot sch ] ) , modes with @xmath140 give almost all the contribution in the asymptotic region , while they contribute with about 65% at @xmath143 ( the innermost stable circular orbit according to general relativity ) .
in this case , the modes with @xmath142 contribute with less than 0.3% of the total radiated power for any position of the rotating charge .
-2 truecm -2 truecm it is interesting to note that the magnitude of the total radiated power in the electromagnetic case is approximately twice the numerical result found previously for a scalar source coupled to a massless klein - gordon field @xcite . in principle , this is not surprising because of the fact that photons have two physical polarizations .
notwithstanding , it should be emphasized that the two polarizations contribute quite differently to the emitted power . for our rotating charge ,
the contribution from mode @xmath144 is negligible when compared with the one from mode @xmath145 for every choice of @xmath146 .
considering angular momentum contributions up to @xmath142 , the ratio between the emitted power associated with modes @xmath144 and @xmath145 is always less than 0.1% .
next , we use our low - frequency expressions for the physical modes , eqs .
( [ q q])-([ctt ii vem ] ) , [ and eq .
( [ rsw ] ) ] in eqs .
( [ pot sch])-([potencia iis ] ) to obtain an analytic approximation for the emitted power .
the result is plotted as the dashed line in fig .
[ figure2 ] .
we see from it that the numerical and analytical results differ sensibly as the charge approaches the black hole but coincide asymptotically , since far away from the hole only low frequency modes contribute to the emitted power .
-2 truecm -2 truecm
in order to exhibit how better a full curved spacetime calculation can be in comparison with a flat spacetime one , let us show how the results found in the previous session differ from the ones obtained in minkowski spacetime . in the latter case ,
the rotating charge is represented by the conserved current density @xmath147 this is formally identical to eq .
( [ js ] ) but it is important to keep in mind that @xmath137 and @xmath148 are associated with schwarzschild and minkowski radial coordinates , respectively , which can not be identified . the charge is regarded now as moving in a circular orbit due to a newtonian gravitational force around a central object with the same mass @xmath149 as the black hole . in order to relate the radial coordinate @xmath150 with the angular velocity @xmath138 ( which is assumed to be measured by the same asymptotic static observers as before ) , we use the keplerian relation : @xmath151 .
the quantization of the electromagnetic field can be performed analogously to the procedure exhibited in section [ sec : quantization ] by making @xmath152 . as a consequence , the scattering potential @xmath153 in eq .
( [ potencial em s ] ) is replaced by @xmath154 . in fig .
[ figure1 ] we plot the minkowski scattering potential for @xmath155 ( see dashed line ) .
assuming the same minimal coupling between the charge and the electromagnetic field as before , we obtain that the emitted power at the tree level is given by @xmath156 \right ] ^{2}\left\vert y_{lm}\left ( \pi/2,0\right ) \right\vert ^{2 } \nonumber \\ & + & 2 q^2 \omega^{8/3 } m^{2/3 } \sum_{l=1}^{\infty}\sum_{m=1}^{l } m^{2 } \left [ j_{l}\left ( m\left ( \omega m\right ) ^{1/3}\right ) \right ] ^{2}\left\vert y_{\phi}^{lm}\left ( \pi/2,0\right ) \right\vert ^{2 } \ , .
\label{potencia final m}%\end{aligned}\ ] ] the modes responsible for the dominant contributions to the radiated power follow the same pattern as in the schwarzschild case . in particular , the contribution from the physical modes @xmath144 is negligible when compared with the contribution from the physical modes @xmath145 . as a consistency check for the flat spacetime results
, we compare our quantum - oriented calculations with classical - oriented ones which lead to the larmor formula for the total emitted power . applying it to the case of a keplerian circular orbit , we obtain @xmath157 with @xmath158 in fig . [ figure3 ] we plot the ratios between @xmath159 and @xmath160 , where the summations in eq .
( [ potencia final m ] ) are performed up to increasing values of @xmath161 .
we see that for @xmath162 the difference between @xmath159 and @xmath160 is less than 0.1% for @xmath163 .
this is in agreement with the fact that the contributions associated with higher values of @xmath34 are negligible to the emitted power .
next we compare our curved and flat spacetime results using our previous expressions for @xmath164 and @xmath165 as functions of the physical observables @xmath149 and @xmath138 as measured by asymptotic static observers .
we plot the ratio between @xmath164 and @xmath165 in fig .
[ figure4 ] obtained from our numerical computations ( solid line ) and from our low - frequency analytic approximation ( dashed line ) . in both cases the ratio tends to the unity as the charge rotates far away from the attractive center , as a consequence of the fact that the schwarzschild spacetime is asymptotically flat . as the rotating charge approaches the central object ,
curved and flat spacetime results differ more significantly . in the innermost relativistic stable circular orbit
, the numerical computation gives that @xmath164 is 30% smaller than @xmath165 .
we emphasize that this is not a simple consequence of the red - shift effect , since the mode functions representing the quanta of the emitted radiation are quite different in curved and flat spacetimes .
now , this is interesting to use our quantum field theory in schwarzschild spacetime approach to compute what is the amount of emitted radiation which can be asymptotically observed .
this is given by @xmath166\ , .
\label{pot obs}%\end{aligned}\ ] ] our numerical result is shown as the solid line in fig .
[ figure5 ] . in order to compute @xmath167 in the low - frequency approximation ( see dashed line in fig .
[ figure5 ] ) we calculate the expressions for the transmission and reflection coefficients . for this purpose
we use @xmath168 in eq .
( [ q q ] ) to write @xmath169 comparing the above expression with eq . ( [ solucao assintotica vai ] ) and using that @xmath170 for @xmath171 , we obtain @xmath172 with the normalization constants given by eqs .
( [ bi ] ) , ( [ bii ] ) , ( [ ctt i vai ] ) and ( [ ctt ii vai ] ) .
the reflexion coefficients are determined using that @xmath173 .
-2 truecm -2 truecm we see from fig .
[ figure5 ] that the black hole absorbs only a small amount of the emitted radiation . even for the innermost stable circular orbit
the black hole absorbs only @xmath174 of the total radiated power .
these results are consistent with the fact that the absorption cross section of a schwarzschild black hole is proportional to @xmath175 for small frequency photons ( see , e.g. , @xcite ) .
in this paper we have considered the radiation emitted by an electric charge rotating around a chargeless static black hole in the context of quantum field theory in curved spacetimes .
we have obtained that the two physical photon polarizations give very different contributions to the total emitted power .
indeed , the contribution of one of the physical modes is negligible as compared with the other one . as a consistency check of our procedure
we have computed , using a similar approach , the emitted power from a charge in minkowski spacetime rotating around a massive object due to a newtonian force and showed that this is in agreement with larmor s classical result .
then we compared the radiation emitted ( as measured by asymptotic static observers ) considering the attractive central object with mass @xmath149 as ( i ) a schwarzschild black hole and ( ii ) a newtonian massive object in flat spacetime .
we have obtained that curved and flat spacetime results coincide when the charge orbits far away from the massive object but differ considerably when the charge orbits close to it .
the difference reaches 30% for the innermost stable circular orbit .
this result corroborates the importance of considering the curvature of the spacetime in astrophysical phenomena occurring in the vicinity of black holes when they involve particles with wavelengths of the order of the event horizon radius .
finally , we have computed the amount of the emitted radiation which is absorbed by the black hole .
we have shown that most of the emitted radiation can be asymptotically observed .
for the case of the innermost stable circular orbit at @xmath176 , about 97% of the emitted power can be in principle detected at infinity .
the authors are grateful to conselho nacional de desenvolvimento cientfico e tecnolgico ( cnpq ) for partial financial support .
r. m. and g. m. would like to acknowledge also partial financial support from coordenao de aperfeioamento de pessoal de nvel superior ( capes ) and fundao de amparo pesquisa do estado de so paulo ( fapesp ) , respectively . r. genzel _ et al _ , nature * 425 * , 934 ( 2003 ) ; b. c. bromley , w. a. miller and v. i. pariev , nature * 391 * , 54 ( 1998 ) ; y. tanaka _ et al _ , nature * 375 * , 659 ( 1995 ) ; r. narayan , i. yi and r. mahadevan , nature * 374 * , 623 ( 1995 ) ; m. j. rees _ et al _ , nature * 295 * , 17 ( 1982 ) . | we consider an electric charge , minimally coupled to the maxwell field , rotating around a schwarzschild black hole . we investigate how much of the radiation emitted from the swirling charge is absorbed by the black hole and
show that most of the photons escape to infinity . for this purpose
we use the gupta - bleuler quantization of the electromagnetic field in the modified feynman gauge developed in the context of quantum field theory in schwarzschild spacetime .
we obtain that the two photon polarizations contribute quite differently to the emitted power .
in addition , we discuss the accurateness of the results obtained in a full general relativistic approach in comparison with the ones obtained when the electric charge is assumed to be orbiting a massive object due to a newtonian force . |
a _ @xmath0triangulation _ of a @xmath1manifold is a triangulation in which every edge has _ degree _
@xmath2 or @xmath3 ( i.e. there are @xmath2 or @xmath3 tetrahedra around each edge ) , and every triangle contains only one edge of degree @xmath2 .
thurston conjectured that every closed @xmath1manifold that admits a @xmath0triangulation has gromov hyperbolic fundamental group .
elder , mccammond and meier @xcite established this statement , via a computer - assisted proof .
in this paper we place the question on hyperbolic triangulations of @xmath1manifolds into the frame of a combinatorial nonpositive curvature . we introduce a new local combinatorial condition of _ @xmath4location _ ( where @xmath5 ) for flag simplicial complexes . when @xmath6 this condition is an analogue of the negative curvature , as the following main result of the paper shows .
let @xmath7 be a simply connected @xmath8located locally @xmath2large simplicial complex .
then the @xmath9skeleton of @xmath7 , equipped with the standard path metric , is gromov hyperbolic .
there are many @xmath8located simplicial complexes and groups acting on them geometrically resulting from various notions of combinatorial nonpositive curvature appearing in the literature ( see section [ s : fin ] for a discussion ) . however , with the notion of @xmath4location we enter for the first time the world of triangulations of manifolds in dimensions above @xmath10 .
every @xmath0triangulation of a @xmath1manifold is an @xmath8located locally @xmath2large simplicial complex .
this provides a new proof of thurston s conjecture .
our methods are quite elementary and we do not use computer computations . moreover , theorem b shows that theorem a is a generalization of the original statement .
there are @xmath8located triangulations of @xmath1manifolds that are not @xmath0triangulations .
furthermore , our proof provides a uniform bound on the hyperbolicity constant and describes the structure of combinatorial balls in @xmath0triangulations .
clearly , theorem a applies in a much broader context than the one of @xmath1manifolds .
it concerns general simplicial complexes . in section [ s : fin ]
we indicate other applications .
we also relate @xmath8location to the well known notion of combinatorial nonpositive curvature " _ systolicity_. note that systolic complexes are in a sense very ( asymptotically ) far from triangulations of manifolds above dimension @xmath10 .
this makes the @xmath8location very interesting , since it may be applied , unlike systolicity , to some classical spaces and groups . in the same section [ s : fin ]
we briefly discuss @xmath4location , for @xmath11 , as an analogue of a nonpositive , but not negative , curvature .
the proof of theorem a is presented in section [ s : hyp ] .
it uses a local - to - global technique developed by the author in @xcite .
the main global property of simply connected @xmath8located complexes the property sd is a variation of the main feature of _ weakly systolic complexes _ the _ simple descent property_. in section [ s:568 ] we prove theorem b. * acknowledgment . * i thank tadeusz januszkiewicz and misha kapovich for discussions on triangulations of manifolds .
this research was supported by narodowe centrum nauki , decision no dec-2012/06/a / st1/00259 , and by the erc grant analytic no .
let @xmath7 be a simplicial complex .
the @xmath12skeleton of @xmath7 is denoted by @xmath13 .
a subcomplex @xmath14 of @xmath7 is _ full _ if every subset @xmath15 of vertices of @xmath14 contained in a simplex of @xmath7 , is contained in a simplex of @xmath14 . for a finite set @xmath16 of vertices of @xmath7 , by @xmath17 or by @xmath18
we denote the _ span _ of @xmath15 , i.e. the smallest full subcomplex of @xmath7 containing @xmath15 . thus
@xmath19 " or @xmath20 " mean that the corresponding sets span a simplex in @xmath7 . we write @xmath21 ( respectively , @xmath22 ) if @xmath23 ( respectively , @xmath24 ) .
moreover , we write @xmath25 ( respectively , @xmath26 ) when @xmath27 ( respectively , @xmath28 ) for @xmath29 .
a simplicial complex @xmath7 is _ flag _ whenever every finite set of vertices of @xmath7 joined pairwise by edges in @xmath7 , is contained in a simplex of @xmath7 .
link _ of a simplex @xmath30 of @xmath7 is a simplicial complex @xmath31 .
let @xmath32 .
a _ @xmath33cycle _
@xmath34 is a triangulation of a circle consisting of @xmath33 vertices : @xmath35 , and @xmath33 edges : @xmath36 and @xmath37 .
a _ @xmath33wheel ( in @xmath7 ) _
@xmath38 ( where @xmath39 s are vertices of @xmath7 ) is a subcomplex of @xmath7 such that @xmath40 is a full cycle , and @xmath41 .
a flag simplicial complex @xmath7 is _ @xmath33large _ if there are no full @xmath42cycles in @xmath7 , for @xmath43 .
@xmath7 is _ locally @xmath33large _ if all its links are @xmath33large .
observe that local @xmath2largeness means that there are no @xmath44wheels .
if not stated otherwise , speaking about a simplicial complex @xmath7 , we always consider the metric on the @xmath45skeleton @xmath46 , defined as the number of edges in the shortest @xmath9skeleton path joining two given vertices .
we denote this metric by @xmath47 . given a nonnegative integer @xmath12 and a vertex @xmath48 , a ( combinatorial ) _ ball _ @xmath49 ( respectively , _ sphere _
@xmath50 ) _ of radius @xmath12 around @xmath51 _ ( or _ @xmath12ball _ around @xmath51 ) is a full subcomplex of @xmath7 spanned by vertices at distance at most @xmath12 ( respectively , at distance @xmath12 ) from @xmath51 .
we use the notation @xmath52 and @xmath53 for , respectively , @xmath49 and @xmath50 , when it is clear what is the underlying complex @xmath7 . : a @xmath54dwheel ( left ) , and a @xmath55dwheel ( right).,scaledwidth=70.0% ] a _ @xmath56dwheel _ ( from double wheel " ) @xmath57 is the union of a @xmath33wheel @xmath58 and an @xmath59wheel @xmath60 , with @xmath61 or @xmath62 see figure [ f : dwheels ] .
the _ boundary _ of the dwheel @xmath63 is the cycle @xmath64 , and the _ boundary length _ is @xmath65 if @xmath61 or @xmath66 otherwise .
[ d : loc ] a flag simplicial complex is _ @xmath4located _ if every dwheel with the boundary length at most @xmath4 is contained in a @xmath9ball see figure [ f : locsd ] .
location ( left ) , and the vertex condition ( v ) for the property sd ( right).,scaledwidth=60.0% ] in this article by a _ covering _ we mean a _ simplicial covering _ , that is a simplicial map restricting to isomorphisms from @xmath9balls onto spans of their images . in particular , it follows that wheels are mapped isomorphically onto wheels , and thus the following result holds .
[ p : cov ] a covering of a locally @xmath33large complex is itself locally @xmath33large .
a covering of an @xmath4located complex is itself @xmath4located .
the goal of this section is to prove theorem a. in subsection [ s : glo ] we introduce a global combinatorial property @xmath67 that , together with @xmath8location and local @xmath2largeness , implies hyperbolicity theorem [ t : hyp ] . in the subsequent subsection [ s : lotoglo ]
, we prove a local - to - global theorem [ t : logl ] implying directly theorem a. [ d : sd ] let @xmath7 be a flag simplicial complex , let @xmath68 be its vertex , and let @xmath69 .
we say that @xmath7 _ satisfies the property @xmath70 _ if for every @xmath71 the following two conditions hold . for every edge @xmath72
the intersection @xmath73 is non - empty . for every vertex @xmath74 , and for every two vertices @xmath75
, there exists a vertex @xmath76 such that @xmath77 see figure [ f : locsd ] .
@xmath7 satisfies the property
_ @xmath78 _
( respectively , the property _ @xmath67 _ ) if @xmath79 holds for every @xmath80 ( respectively , for every @xmath80 and @xmath68 ) .
[ p : sconn ] let @xmath7 be a simplicial complex satisfying the property @xmath78 , for some vertex @xmath68 . then @xmath7 is simply connected .
the proof follows closely the proof of ( * ? ? ?
* lemma 5.5 ) . here , instead of the quadrangle condition one uses the vertex condition ( v ) from definition [ d : loc ] .
[ l:21 ] let @xmath7 be an @xmath8located locally @xmath2large simplicial complex satisfying the property @xmath81 , for some vertex @xmath68 .
let @xmath82 and let @xmath83 be such that @xmath84 .
let @xmath85 be a vertex adjacent to @xmath86 , given by the vertex condition ( v ) .
let @xmath87 be vertices with @xmath88 , given by the triangle condition ( t ) . then @xmath89 ; @xmath90 ; @xmath91 ; and @xmath92 . by @xmath2largeness we have immediately that @xmath89 ; @xmath90 ; and @xmath91 .
for the rest we proceed by contradiction .
assume that @xmath93 .
then , by the vertex condition ( v ) , there is a vertex @xmath94 adjacent to @xmath95 .
note that , by @xmath2largeness , it is not possible that @xmath96 , thus there is a @xmath3wheel @xmath97 , or a @xmath2wheel @xmath98 , or @xmath99 . by the triangle ( t ) and the vertex ( v ) condition
there are vertices @xmath100 with @xmath101 , and @xmath102 .
possibly @xmath103 , but @xmath104 , by @xmath2largeness . in any case
we obtain an @xmath56dwheel @xmath63 , spanned by vertices in @xmath105 , with the boundary length at most @xmath8 .
thus , by @xmath8location , @xmath63 is contained in a @xmath9ball , which contradicts the fact that @xmath106 .
[ t : hyp ] let @xmath7 be an @xmath8located locally @xmath2large simplicial complex satisfying the property @xmath67
. then @xmath46 equipped with a path metric induced from @xmath107 is @xmath108hyperbolic , for a universal constant @xmath108 .
we use a criterion by papasoglu @xcite , i.e. we reduce the proof to showing that intervals are uniformly thin . let @xmath109 be two vertices and let @xmath110 be the interval between them , i.e. the set of vertices lying on geodesics between @xmath68 and @xmath111 .
by @xmath112 we denote the intersection @xmath113 , where @xmath114 .
we show that for every @xmath115 , for every two vertices @xmath116 , we have @xmath117 .
this shows also that the hyperbolicity constant is universal . by contradiction
let @xmath33 be the maximal number such that there are vertices @xmath116 with @xmath118 .
then in @xmath119 there exist vertices @xmath120 , and @xmath121 with @xmath122 . by the vertex condition ( v )
, there is a vertex @xmath123 in @xmath119 , possibly with @xmath124 . by the triangle condition ( t )
there are vertices @xmath125 , with @xmath126 if @xmath124 , such that @xmath127 , and @xmath128 . by the vertex condition ( v ) ,
in @xmath112 there exist vertices @xmath129 ; @xmath130 ; @xmath131 , possibly , with @xmath132 , @xmath133 , @xmath134 . among the vertices
@xmath135 we choose the first one ( in the given order ) , that is at distance @xmath1 from @xmath51 .
denote this vertex by @xmath136 . using the vertex condition ( v )
, we obtain a full path @xmath137 in @xmath112 of diameter @xmath1 , with @xmath138 and @xmath139 . in the remaining part of the proof
we show that the existence of such path is impossible .
this leads to a contradiction yielding the theorem . for @xmath140 ,
let @xmath141 be a vertex in @xmath142 given by the triangle condition ( t ) .
all the possible cases ( up to renaming vertices ) of mutual relations between vertices @xmath143 are shown in figure [ f : hyp ] cases : i , ii , and iii .
* case i. * in this case we assume that for all triples of vertices @xmath143 as above we are not in case ii or iii , that is @xmath144 , for @xmath145 .
first , suppose that @xmath146 .
we may assume that @xmath147 . by ( t )
, there are vertices @xmath148 with @xmath149 . if @xmath150 then we are in case i(a ) ( see figure [ f : hyp ] ) .
further we assume that there are no such @xmath151 .
if @xmath152 then , by ( t ) , there is a vertex @xmath153 adjacent to @xmath154 case i(b ) ( see figure [ f : hyp ] ) .
if not then , by ( v ) , there is a vertex @xmath155 adjacent to @xmath154 . if @xmath156 then we are in the previous case taking @xmath157 instead of @xmath158 .
similarly for @xmath159 , thus further we assume that @xmath160 . by ( t ) , in @xmath161 there exist @xmath162 with @xmath163 . by lemma [ l:21 ]
, we have @xmath164 and @xmath165 .
it follows that there is a @xmath54dwheel @xmath166 @xmath63 . by @xmath8location , @xmath63
is contained in a @xmath9ball contradiction , since @xmath167 .
.,scaledwidth=70.0% ] now we suppose that @xmath168 , and @xmath169 .
then , by ( v ) , in @xmath142 there is @xmath170 .
we may assume that @xmath171 . by ( t ) ,
in @xmath172 there are @xmath173 , with @xmath174 . by lemma [ l:21 ] , @xmath175 ; @xmath176 ; @xmath177 ; and @xmath178 .
therefore , @xmath179 are contained in a @xmath180dwheel @xmath63 , with the boundary length at most @xmath8 . by @xmath8location , @xmath63
is contained in a @xmath9ball @xmath181 .
replacing @xmath182 by @xmath183 we are in case i(a ) or i(b ) .
if @xmath184 then , by ( v ) , in @xmath142 there are @xmath185 with @xmath186 . by ( t ) , in @xmath172 there exist @xmath187 with @xmath188 , and @xmath189 . by lemma [ l:21 ] @xmath190
; @xmath191 ; @xmath192 ; and @xmath193 .
thus the vertices @xmath194 are contained in a @xmath180dwheel @xmath63 , with the boundary length at most @xmath8 . by @xmath8location , @xmath63
is contained in @xmath195 , and hence , replacing @xmath182 by @xmath196 , we are in the previous case .
* case ii .
* proceeding similarly as in case i above ( i.e. using the conditions ( v ) and ( t ) , and lemma [ l:21 ] ) , we come to the two possible cases : ii(a ) and ii(b ) , as in figure [ f : hyp ] with @xmath197 and @xmath153 . in the remaining part of the proof
we analyze separately all the cases : i(a ) , i(b ) , ii(a ) , ii(b ) , and iii , showing that they lead to contradiction .
to do this we project " edges and vertices towards @xmath111 now : by ( t ) , in @xmath119 there exist vertices @xmath198 ( possibly some of them coinciding ) , with @xmath199 . by ( v )
, we have that @xmath200 . _
ad case iii : _ this leads to a contradiction , since @xmath201 . _
ad case ii(a ) : _ by lemma [ l:21 ] , we have @xmath202 ; @xmath203 ; @xmath204 ; and @xmath205 .
therefore , the vertices @xmath206 are then contained in a @xmath180dwheel @xmath63 with the boundary length at most @xmath8 .
thus , by @xmath8location , @xmath63 is contained in a @xmath9ball contradiction , since @xmath201 .
_ ad case ii(b ) : _ by lemma [ l:21 ] , we have @xmath202 ; @xmath203 ; @xmath204 ; and @xmath205 . by ( v ) , there is @xmath207 with @xmath208 , possibly with @xmath209 .
we may assume that @xmath210 . by ( t ) ,
in @xmath211 there are vertices @xmath212 and @xmath213 , with @xmath214 , by ( v ) .
therefore the vertices @xmath215 are contained in a @xmath180dwheel @xmath63 with the boundary length at most @xmath8 .
thus , by @xmath8location , @xmath63 is contained in a @xmath9ball contradiction , since @xmath216 . _
ad case i(a ) : _ the vertices @xmath217 are contained in a @xmath180dwheel @xmath63 with the boundary length at most @xmath8 .
thus , by @xmath8location , @xmath63 is contained in a @xmath9ball contradiction , since @xmath218 .
_ ad case i(b ) : _ if @xmath219 or @xmath205 then we obtain a @xmath220dwheel containing @xmath221 , leading to contradiction , by @xmath8location .
similarly , when @xmath222 or @xmath223 .
thus , further we assume that @xmath224 . by ( v ) , in @xmath119 there is a vertex @xmath225 .
we may assume that @xmath226 . by ( t ) ,
in @xmath211 there exist vertices @xmath227 , such that @xmath228 . by lemma [ l:21 ] , @xmath229 ; @xmath230 ; @xmath231 ; and @xmath232 .
thus there is a @xmath54dwheel @xmath233 , which lies in a @xmath9ball , by @xmath8location .
this however contradicts the fact that @xmath216 .
in this section we prove the following result implying theorem a from introduction .
[ t : logl ] let @xmath7 be an @xmath8located locally @xmath2large simplicial complex .
then its universal cover @xmath234 is an @xmath8located locally @xmath2large simplicial complex satisfying the property @xmath67 . in particular
, @xmath235 is @xmath108hyperbolic .
the proof follows closely up to much of the notations the proof of the analogous theorem 4.5 from @xcite ( compare also the proof of ( * ? ? ?
* theorem 1 ) ) .
we construct the universal cover @xmath236 of @xmath7 as an increasing union @xmath237 of combinatorial balls .
the covering map is then the union @xmath238 where @xmath239 is locally injective and @xmath240 , for @xmath241 .
we proceed by induction .
choose a vertex @xmath68 of @xmath7 .
define @xmath242 , @xmath243 and @xmath244 .
assume that we have constructed the balls @xmath245 and the corresponding maps @xmath246 to @xmath7 so that the following conditions are satisfied : ( @xmath247 ) : : @xmath248 for @xmath249 ; ( @xmath250 ) : : @xmath251 satisfies the property @xmath252 ; ( @xmath253 ) : : @xmath254 is an isomorphism onto the span of the image for @xmath255 and it is an isomorphism for @xmath256 .
observe that those conditions are satisfied for @xmath257 and @xmath258 , i.e. that conditions ( @xmath259 ) , ( @xmath260 ) and ( @xmath261 ) hold .
now we construct @xmath262 and the map @xmath263 . for a simplex @xmath264 of @xmath251 ,
we denote by @xmath30 its image @xmath265 in @xmath7 .
let @xmath266 and let @xmath267 define a relation @xmath268 on @xmath269 as follows : @xmath270 the transitive closure @xmath271 of the relation @xmath272 will be further used to define @xmath273 .
the following lemma shows that @xmath271 is not too far " from @xmath272 .
[ l : equiv ] if @xmath274 then there is @xmath275 with @xmath276 . if @xmath277 , for @xmath278 , or if @xmath279 , or @xmath280 , or @xmath281 , then the assertion trivially holds .
thus further we assume this is not the case . by ( @xmath247 ) and ( @xmath250 ) , in @xmath282 there are vertices @xmath283 with @xmath284 .
* claim . * for @xmath278 , we have @xmath285 .
furthermore , for @xmath286 , @xmath287 and @xmath288 .
_ proof of claim .
_ we show only that @xmath289 .
other assertions follow in the same way . by contradiction
assume that @xmath290 .
observe that @xmath291 , by ( @xmath253 ) and by the definition of the set @xmath269 , and that @xmath292 , by ( @xmath253 ) and our initial assumptions .
therefore , by ( @xmath253 ) and by our definition of @xmath269 there is a @xmath44wheel @xmath293 in @xmath7 contradiction with @xmath2largeness .
this finishes the proof of claim . @xmath294 by ( @xmath250 ) , in @xmath282 there are vertices @xmath295 with @xmath296 , and @xmath297 , by claim .
when @xmath298 then @xmath299 . we may assume that @xmath300 if not then we take @xmath301 instead of @xmath302 .
further , we may assume that if @xmath303 then @xmath304 if not than we choose @xmath301 instead of @xmath305 . by the definition of the set @xmath269 and by ( @xmath253 )
, it follows that in @xmath7 there is a wheel spanned by vertices @xmath306 .
furthermore , if @xmath307 then we may assume that @xmath308 otherwise we replace @xmath309 by @xmath310 .
it may however happen that @xmath311 . then , by the definition of the set @xmath269 and by ( @xmath253 ) , in @xmath7 there is a @xmath2wheel @xmath312 .
if @xmath313 then we have a wheel spanned by vertices @xmath314 .
since @xmath315 , in any case we obtain in @xmath7 a dwheel @xmath63 with the boundary length at most @xmath8 .
therefore , by @xmath8location , there is a vertex @xmath316 with @xmath317 . by ( @xmath253 ) applied to the vertex @xmath305 there is a vertex @xmath318 , with @xmath319 .
again by ( @xmath253 ) , we have that all the vertices @xmath320 are adjacent to @xmath321 in @xmath322 . hence , to prove the lemma it is enough to show that @xmath323 .
if not then , by ( @xmath253 ) , there is @xmath324 such that @xmath325 . by ( @xmath253 )
, we have then that @xmath326 , which is a contradiction , since @xmath327 .
this proves the lemma .
observe that , by lemma [ l : equiv ] , if @xmath328 then there is a vertex @xmath329 with @xmath323 and @xmath330 .
we define the flag simplicial complex @xmath331 in the following way .
its @xmath45skeleton is by definition the set @xmath332 .
now we define the @xmath9skeleton @xmath333 of @xmath331 as follows .
edges between vertices of @xmath251 are the same as in @xmath251 . moreover ,
for every @xmath334 , there are edges joining @xmath335 with @xmath336 \in z/{\stackrel{\bar e}{\sim}}$ ] ( here @xmath336 $ ] denotes the equivalence class of @xmath337 ) and the edges joining @xmath336 $ ] with @xmath338 $ ] , for @xmath339 . having defined @xmath333 the higher dimensional skeleta
are determined by the flagness property .
definition of the map @xmath340 is clear : @xmath341 and @xmath342)=z$ ] .
we show that it can be simplicially extended .
it is enough to do it for simplices in @xmath343 .
let @xmath344,\ldots,[{\widetilde}w_{l},z_l],{\widetilde}w_1',\ldots,$ ] @xmath345 be a simplex .
then , by definition of edges in @xmath331 , we have that @xmath346 and @xmath347 , for @xmath348 and @xmath349 .
since @xmath350)=z_p$ ] , @xmath351 and since @xmath352 was simplicial , it follows that @xmath353 .
hence , by the simplicial extension , we can define the map @xmath354 .
now we check that @xmath331 and @xmath355 satisfy conditions ( @xmath356 ) , ( @xmath357 ) and ( @xmath358 ) .
the proof of the conditions ( @xmath356 ) and ( @xmath358 ) is exactly the same as the one in the proof of theorem 4.5 in @xcite ( even up to the notations ; see page 12 there ) .
* condition * ( @xmath357 ) . by the condition ( @xmath250 )
it is enough to verify the triangle condition ( t ) and the vertex condition ( v ) from definition [ d : sd ] only for , respectively , edges and vertices in @xmath359 . by the definition of edges in @xmath360
it is clear that @xmath361 is non - empty , for an edge @xmath362 .
therefore , the condition ( t ) is satisfied . the vertex condition ( v ) , for a vertex @xmath363\in { \widetilde s}_{i+1}$ ]
follows immediately from the definition of edges in @xmath273 , and from lemma [ l : equiv ] . having established conditions
( @xmath356 ) , ( @xmath357 ) and ( @xmath358 ) we are able to construct a complex @xmath364 and a map @xmath365 with the following properties .
the complex @xmath234 satisfies the property @xmath79 for every @xmath80 and the map @xmath366 is a covering map .
thus , by proposition [ p : cov ] , the cover @xmath367 is @xmath8located and locally @xmath2large . from proposition
[ p : sconn ] it follows that @xmath367 is simply connected and therefore @xmath367 is the universal cover of @xmath7 .
since the vertex @xmath68 was chosen arbitrarily in our construction and since the universal cover of @xmath7 is unique it follows that @xmath367 satisfies the property @xmath79 for every vertex @xmath68 and for every natural number @xmath80 .
this finishes the proof of the theorem .
the goal of this section is to show that @xmath0triangulations are @xmath8located locally @xmath2large simplicial complexes , i.e. to prove theorem b from introduction .
a link of a vertex in a @xmath0triangulation of a @xmath1-manifold is a _ @xmath0triangulation of a @xmath10sphere _ , that is , each vertex has degree ( the number of edges adjacent to ) @xmath2 or @xmath3 and no two vertices of degree @xmath2 are connected by an edge .
a _ soccer tiling of a @xmath10sphere _ is the cellulation dual to a @xmath0triangulation of a @xmath10sphere see @xcite .
its cells pentagons and hexagons correspond to vertices of the triangulation , its vertices correspond to triangles . a _ soccer diagram _ is a subcomplex of a soccer tiling , which is homeomorphic to a disc .
a path in the boundary of a soccer diagram which is contained in a single cell is called an _ exposed path_. throughout this subsection we assume that @xmath7 is a @xmath0triangulation of a @xmath1manifold . [
l : wheels ] @xmath7 is locally @xmath2large and , for @xmath368 , every @xmath33wheel @xmath369 in @xmath7 is contained in a link of a vertex @xmath370 .
the lemma above follows immediately from lemma [
l : cycles ] below .
the latter is an elementary result concerning @xmath0triangulations of a @xmath10sphere . for brevity ,
in the proof we use results from ( * ? ? ?
* section 3 ) .
[ l : cycles ] let @xmath14 be a @xmath0triangulation of a @xmath10sphere .
there are no full @xmath44cycles in @xmath14 , and , for @xmath368 , every full @xmath33cycle in @xmath14 is contained in a wheel .
let @xmath371 and assume there is a full @xmath33cycle @xmath372 in @xmath14 .
in the soccer tiling @xmath373 corresponding to @xmath14 , we have then the chain of cells @xmath374 , with @xmath375 and @xmath376 sharing an edge ( here and further we use the convention @xmath377 ) . by @xmath378
we denote the subcomplex of @xmath373 being the union of @xmath374 .
observe that if , for @xmath379 , @xmath375 and @xmath380 have an edge in common then @xmath378 ( in @xmath14 ) is not full , thus further we assume this is not the case .
consider now the two _ boundary paths _ @xmath381 and @xmath382 of @xmath378 in @xmath373 see figure [ f:6cycle ] . since @xmath383 and cells are pentagons or hexagons , the sum of lengths @xmath384 is at most @xmath385 .
therefore , without loss of generality we assume that @xmath386 . for @xmath387 : the cycle @xmath378 and the paths @xmath388 ( left ) ; the exposed path in @xmath389 ( right).,scaledwidth=70.0% ]
consider a soccer diagram @xmath389 with boundary @xmath381 that has at most @xmath3 pentagons such a diagram always exists by ( * ? ? ?
* remark 2.9 ) . by (
* theorem 3.10 ) , either @xmath389 contains an exposed path of length @xmath44 , or @xmath389 is one of the two diagrams in ( * ? ? ?
* figure 2 ) .
it is easy to see that the latter is impossible .
therefore , for the rest of the proof we suppose that there is an exposed path of length @xmath44 in @xmath389 .
this is impossible when @xmath390 , hence further we consider @xmath368 .
if , for some @xmath391 , the cell @xmath375 belongs to @xmath389 ( i.e. @xmath389 is outside @xmath381 " in figure [ f:6cycle ] ) then @xmath392 ( or @xmath393 if @xmath394 ) and @xmath376 share an edge , contrary to our assumptions .
therefore @xmath389 does not contain any @xmath375 ( i.e. @xmath389 is inside @xmath381 " in figure [ f:6cycle ] ) . without loss of generality
we assume that the exposed path @xmath395 of length @xmath44 consists of edges @xmath396 such that @xmath397 .
let @xmath398 be a cell of @xmath389 containing @xmath395 .
if @xmath399 then the lemma is proved , since then the cycle @xmath378 ( in @xmath14 ) is contained in the @xmath33wheel @xmath400 in @xmath14 .
therefore , further we assume that @xmath401 , where @xmath402 is a new ( nonempty ) soccer diagram .
assume that among the cells @xmath375 only the cells @xmath403 intersect @xmath398 , for @xmath404 .
then we obtain a new cycle @xmath405 of cells : @xmath406 , of length strictly smaller than @xmath33 .
this corresponds to a full cycle @xmath405 in @xmath14 . if @xmath407 then this is impossible as shown above . if @xmath387 then @xmath405 has to have length @xmath2 and , as we have just proved @xmath405 is contained in a wheel @xmath408
this means that both @xmath398 and @xmath409 are pentagons , which contradicts the definition of the soccer diagram , and thus yields the lemma .
[ l:7wheels ] let @xmath410 be a @xmath411wheel in @xmath7 .
then there exist vertices @xmath412 such that @xmath413 , and , up to renaming cyclically vertices @xmath375 , the vertex @xmath316 is adjacent to @xmath414 , and @xmath415 is adjacent to @xmath416 . the first part of the proof is the same as in the proof of lemma [ l : cycles ] above
: we consider the cycle " of cells @xmath417 in the soccer tiling corresponding to @xmath418 , and we find a path @xmath419 of length at most @xmath420 . considering a soccer diagram @xmath389 with the boundary @xmath419 and with at most @xmath3 pentagons , we conclude that @xmath389 does not contain any @xmath375 , and that there is an exposed path of length @xmath44 .
it is not possible that @xmath389 consists of a single cell , so that the cell @xmath398 containing the exposed path forms a new cycle of less than @xmath411 cells . for the new cycle we apply lemma [ l : cycles ] , to obtain a single filling " cell @xmath366 .
the pair of cells @xmath421 corresponds to the pair of required vertices : @xmath422 or @xmath423 .
_ proof of theorem b. _ assume that @xmath424 is a @xmath56dwheel in @xmath7 with the boundary length at most @xmath8 .
without loss of generality we assume that @xmath425 . by lemma [ l : wheels ]
the only possibilities for the pair @xmath56 are : @xmath426 . by lemma [ l : wheels ]
, there is a vertex @xmath427 such that @xmath428 .
first we consider the case when @xmath383 . by lemma [ l : wheels ]
, there is a vertex @xmath429 such that the wheel @xmath430 .
if @xmath431 then @xmath432 and we are done . hence further we assume that @xmath433 . if @xmath61 then @xmath434 is a @xmath44cycle in @xmath435 . by lemma
[ l : wheels ] , this cycle is not full and thus @xmath436 . hence , the vertices @xmath437 span a @xmath44simplex contradiction . if @xmath438 then , by the definition of a dwheel @xmath63 , we have @xmath62 and , by our assumptions , @xmath439 . in this case the cycle @xmath440 is a @xmath2cycle in @xmath435 .
if this cycle is not full then we have a contradiction as before .
if it is full then the edge @xmath441 has degree @xmath2 .
however , also the edge @xmath442 has degree @xmath2 contradicting the definition of the @xmath443triangulation .
it remains to treat the case @xmath444 .
observe that in this case we have @xmath61 .
let @xmath412 be vertices as in lemma [ l:7wheels ] for the @xmath411wheel @xmath58 .
either @xmath445 are all adjacent to @xmath316 or to @xmath415 , or , without loss of generality , we assume that @xmath446 , and @xmath447 . in the former case we proceed as above for @xmath383 .
thus further we assume that @xmath448 .
then there is a @xmath2cycle @xmath449 in the link of the edge @xmath450 .
if this cycle is not full then we obtain a simplex of dimension above @xmath1 in @xmath7 contradiction .
therefore the cycle is full and hence the degree of @xmath450 is @xmath2 .
this means however that there are two degree @xmath2 edges in the triangle @xmath451 , contradicting the definition of the @xmath443triangulation . @xmath294
there is another ,
more natural " , version of @xmath4location , that originated in fact our studies : a flag ( now , not necessarily @xmath2large ) simplicial complex is _ @xmath4located _ if every homotopically trivial loop of length at most @xmath4 has a filling diagram with at most one internal vertex ( see eg .
@xcite for basics on filling diagrams ) .
the new property does not imply the one considered in this paper , neither vice versa .
the following analogue of theorem a holds . a more general scheme for local conditions implying gromov hyperbolicity relies on the requirement that short loops allow small filling diagrams " .
this leads to particular instances of the local - to - global principle for the linear dehn function .
the notion of @xmath8location can be applied not only in the case of ( closed ) manifolds or manifolds with boundaries .
having , a @xmath443triangulation of a @xmath1manifold one can construct a branched cover that will still be an @xmath8located locally @xmath2large complex .
there are also various amalgamation procedures that can be applied to obtain singular " spaces beginning with manifolds .
further , it is possible to construct @xmath8located pseudomanifolds using constructions from @xcite and similar ( cf .
e.g. @xcite for further references ) .
objects obtained in this way may be asymptotically different from the existing ones compare the next subsection .
clearly , @xmath8location applies to higher dimensional manifolds as well .
nevertheless , it is an interesting question , whether for @xmath44manifolds ( and in higher dimensions ) one may formulate conditions on a triangulation in the spirit of thurston s condition , implying @xmath8location , and thus hyperbolicity .
there are various notions of combinatorial nonpositive curvature " appearing in the literature see e.g. @xcite for more details on this . in particular , for @xmath452 , a _
@xmath33systolic complex _
@xcite is a simply connected locally @xmath33large simplicial complex . in @xcite
the notion of a _ weakly systolic _ complex , being a generalization of @xmath33systolicity , has been introduced .
the following immediate corollary of theorem a provides a local condition on a weakly systolic complex implying its hyperbolicity .
the new condition is more general than the well studied conditions for hyperbolicity : local @xmath411largeness for @xmath33systolic complexes ( see ( * ? ? ?
* section 2 ) ) , and @xmath453 for weakly systolic complexes ( see ( * ? ? ?
* section 7 ) ) .
[ c:67syst ] the following condition implies hyperbolicity of a weakly systolic complex @xmath7 : for every vertex @xmath51 such that @xmath418 is not @xmath3large ( respectively , is @xmath3large but not @xmath411large ) , every its neighbor @xmath196 has an @xmath8large ( respectively , @xmath411large ) link @xmath454 .
note however , that ( weakly ) systolic complexes do not fit well the world of manifolds of dimension above @xmath10 e.g. there are no @xmath33systolic triangulations of such manifolds @xcite .
more strikingly , groups acting geometrically on systolic complexes are far from acting in the same way on aspherical manifolds see e.g. @xcite .
thus the @xmath8location is a ( asymptotically ) far going generalization of @xmath411systolicity , that allows still to obtain strong results on the structure of the complex as e.g. in the description of combinatorial balls provided within the proof of theorem [ t : logl ] . in the current paper we focus on @xmath8located complexes .
nevertheless , the @xmath411location seems to be an interesting nonpositive - curvature - like property , as argued below .
it is not so with @xmath3location there exist @xmath3located locally @xmath2large triangulations of the @xmath10sphere .
note that all @xmath3systolic complexes are @xmath411located .
in particular , the @xmath411location does not imply hyperbolicity : the tiling of the euclidean plane by equilateral triangles is @xmath411located .
however , with much more effort than for the @xmath8location , one can prove a local - to - global result similar to theorem [ t : logl ] also for @xmath411located complexes .
we believe that there are other nonpositive - curvature - like properties satisfied by such complexes ( cf .
e.g. @xcite ) .
this , together with the fact that @xmath411location goes far beyond systolicity , make us believe that this notion deserves further studies . | we introduce and study local combinatorial conditions on a simplicial complex , implying gromov hyperbolicity of its universal cover .
we apply the theory to thurston s problem on @xmath0triangulations of @xmath1manifolds , providing a new proof and generalizing the original result .
we indicate further applications . |
the utility of weak- and strong - coupling expansions and of dualities in nearly all branches of physics can hardly be overestimated .
this article is devoted to several inter - related fundamental questions .
mainly : ( 1 ) what information does the existence of finite order complementary weak- and strong - coupling series expansion of given physical quantities ( e.g. , partition functions , matrix elements , etc . ) provide ?
( 2 ) to what extent can dualities be employed to _ partially solve _ those various problems ? by _ partial solvability _ , we mean the ability to compute a specific physical quantity with complexity polynomial in the size of the system , given partial information that is determined by other means .
as we will demonstrate in this work , a universal problem deeply binds to the above two inquiries , and raises the critical question ( 3 ) why do numerous dualities in very different fields always turn out to be _ conformal transformations _ ? to set the stage , we briefly recall general notions concerning dualities .
consider a theory of ( dimensionless ) coupling strength @xmath3 for which weak- and strong - coupling expansions may , respectively , be performed in powers of @xmath3 and @xmath4 or in other monotonically increasing / decreasing functions @xmath5 .
common wisdom asserts that as ordinary expansion parameters ( e.g. , @xmath3 and @xmath4 ) behave very differently , weak- and strong - coupling series can not , generally , be simply compared . on a deeper level , if these expansions describe different phases ( as they generally do ) then the series must become non - analytic ( in the thermodynamic limit ) at finite values of @xmath3 ( where transitions occur ) and thus render any equality between them void .
a duality may offer insightful _
information _ on a strong coupling theory by relating it to a system at weak coupling that may be perturbatively examined .
as is well known , when they are present , self - dualities are manifest as an equivalence of the coefficients in the two different series ; this leads to an invariance under an inversion that is qualitatively ( and in standard field theories , e.g. , qed / electroweak / qcd is exactly ) of the canonical form `` @xmath6 '' ( or , more generally , @xmath7 ) .
for example , in vacuum qed with lagrangian density @xmath8 $ ] , the ratio @xmath9 of the couplings in front of the @xmath10 and @xmath11 terms relates to a @xmath12 reciprocity .
this reciprocity is evident from the invariance of maxwell s equations in vacuum under the exchange of electric and magnetic fields @xcite , @xmath13 and the lagrangian density that results . in yang - mills ( ym ) theories , such an exchange between dual fields has led to profound insights from analogies between the meissner effect and the behavior of vortices in superconductors to confinement and flux tubes a hallmark of qcd @xcite .
abstractions of dualities in electromagnetism and in ym theories produced powerful tools such as those in hodge and donaldson theories @xcite . in both classical and quantum models , dualities ( and the @xmath7 inversion ) are generated by _ linear transformations _ ( appearing , e.g. , as unitary transformations or more general isometries relating one local theory to another in fundamental `` bond - algebraic '' @xcite incarnations or , in the standard case , fourier transformations @xcite ) .
such linear transformations lead to an _ effective _ inversion of the coupling constant @xmath3 .
dual models share , for instance , their partition functions ( and thus the same series expansion ) .
as realized by kramers and wannier ( kw ) @xcite , self - dualities provide structure that enables additional information allowing , for instance , the exact computation of phase transition points .
this does not imply that the full partition function is determined with complexity polynomial in the size of the system , that is , it is _ solvable _ via self - dualities _ alone _ ( and indeed as we illustrate in this work , self - dualities do not suffice ) .
now here is a main point that concerning question ( 3 ) which we wish to highlight in this article . in diverse arenas , the weak- and strong - coupling expansion parameters @xmath14 and @xmath15 _
are related to one another via conformal transformations _ that are of the fractional linear type . amongst many others ,
prevalent examples are afforded by @xmath16 dualities in ym theories as well as those in ising models and ising lattice gauge theories . in all of these examples ,
the transformations linking @xmath17 to @xmath18 are particular special cases of conformal ( or fractional linear ( mbius ) ) transformations .
that is , in these , @xmath19 with @xmath20 and @xmath21 complex coefficients , and determinant @xmath22 a well known mathematical property of fractional linear maps is their composition property : given any two fractional linear functions @xmath23 ( with @xmath24 ) , direct substitution demonstrates that @xmath25 ( i.e. , yet another fractional linear transformation ) where @xmath26 this group multiplication property will be of great utility in our analysis of dualities .
fractional linear maps , as is commonly known by virtue of the trivial equality ( valid when @xmath27 ) @xmath28 which may be expressed as compositions of transformations of the ( formal ) forms : translation ( @xmath29 ) , scaling / rotation ( @xmath30 ) , and inversion ( @xmath31 ) . as each of these individual operations generally map circles and lines onto themselves so do the general transformations of eq .
( [ circle ] ) .
this may be understood as a consequence of a projective transformation from the riemann sphere onto the complex plane .
relating lorentz transformations to mbius transformations is one of the principal ideas underlying twistor theory @xcite . envisioning _
standard dualities _
@xcite as particular induced maps on the euclidean @xmath32 sphere will be an outcome of the current work .
the set of all conformal self - mappings of the upper half complex plane forms a group , with @xmath16 a subgroup ( `` full modular group '' ) that consists of all the fractional linear transformations with @xmath20 and @xmath21 integers , and determinant @xmath33 . in the aforementioned ym theories , e.g. , @xcite
, an @xmath16 structure follows from a canonical invariance of the form @xmath34 ( stemming from charge quantization ) . as we will detail in the current work , in ising models and ising gauge theories , a canonical form of the duality
is given by @xmath35 the transformation of eq .
( [ mobius ] ) may trivially be associated to one with @xmath36 @xcite by a uniform scaling @xmath37 which does not change the ratio in eq .
( [ abcd ] ) .
more widely , any fractional linear transformation of the form of eq .
( [ abcd ] ) with a finite determinant may similarly be related to one with @xmath36 by a uniform scaling of all four elements of the matrix . in general , we are interested in duality mappings as applied to matrix elements , partition functions or path integrals , while the typical scenario in ym theories focuses on mappings of the action ( or hamiltonian ) . in what will follow
, we will first address question ( 3 ) and illustrate that disparate duality transformations must be of the form of eq .
( [ abcd ] ) . when applied to the expansion parameters
, we will then demonstrate that these _ fractional linear maps lead to linear constraints between the strong- and weak - coupling series coefficients_. a main message of this work is that these conformal transformations of eq .
( [ abcd ] ) , leading to linear relations among series coefficients , will allow a broad investigation of questions ( 1 ) and ( 2 ) above .
specifically , we will examine arbitrarily large yet _ finite size _
systems for which _ no _ phase transitions appear . as is well known , _
analyticity enables a full determination of functions _ over entire domains given their values in only a far more restricted regime ( even if only of vanishing measure ) . for a finite size system , the weak - coupling ( w - c ) and strong - coupling ( s - c ) expansions describe the same analytic function and are everywhere convergent and
may thus be equated to one another .
thus , a trivial yet practical consequence is , contrary to some lore , that _ the naturally perturbative w - c and the seemingly more involved s - c expansions are equally hard_. we will apply this approach to the largest ising model systems for which the exact expansions are known to data on both finite size cubic and square lattices .
we further test other aspects of our methods on ising and generalized wegner models .
the substitution of eq .
( [ abcd ] ) relates the w - c and s - c expansion parameters in general dual models .
we will more generally : ( 1 ) equate the w - c and s - c expansions to find _ linear constraints _ on the expansion coefficients , and ( 2 ) when possible , invoke self - duality to obtain yet further linear equations that those coefficients need to satisfy .
this analysis will lead to the concept of _ partial solvability _ : the linear equations that we will obtain will enable us to _ localize np hardness _ of finding the exact partition function coefficients ( or other quantities ) to that of evaluating only a fraction of these coefficients . the remainder of these coefficients can be then trivially found by the linear relations that are derived from the duality of eq .
( [ abcd ] ) .
a highly non - trivial consequence of our work is that of _ relating mathematical identities to dualities _ such as those broadly generated by eq .
( [ abcd ] ) .
specifically , as a concrete example in this work , we will illustrate how the relations that we obtain connecting the w - c and s - c expansions lead to _ new combinatorial geometry equalities _ in general dimensions .
as a particular example we will do this by noting that , in ising and generalized wegner models , the expansion coefficients are equal to the number of geometrical shapes of a given magnitude of the @xmath21-dimensional surface areas .
the equality between the w - c and s - c expansions then lead to identities connecting these numbers .
for the ising , ising gauge , and several other theories that we study in this work , the mapping between the w - c and s - c coupling expansion parameters is afforded by the particular mbius transformation @xmath38 associated with eq .
( [ mobius ] ) .
this transformation trivially satisfies babbage s equation @xmath39 for all @xmath40 . for self - dual models , such as the @xmath41 ising model or @xmath42
ising gauge theories , we can easily find the critical ( self - dual ) point , @xmath43 , by solving the equation @xmath44 .
we will term theories obeying eq .
( [ ffz ] ) as those that exhibit a `` one- '' duality . in general
, one may find such transformations , represented by a function @xmath45 , in terms of some parameter @xmath40 ( a coupling constant which can be complex - valued ) .
richer transformations appear in diverse arenas including renormalization group ( rg ) calculations . based on these considerations we may have @xmath46 more general transformations @xmath47 may yield linear equations in a manner identical to those appearing for the ising theories studied in the current work .
expansion parameters @xmath40 in self - dual theories satisfy @xmath48 ; this yields a _
constraint on all possible self - dualities_. solutions are afforded by fractional linear ( conformal ) maps @xmath49 with the determinant of eq .
( [ deteq ] ) being non - zero , @xmath50 . as we will further expand on elsewhere , another related duality appearing in ising and all potts models is given by @xmath51 with determinant @xmath52 such that @xmath53 is satisfied .
in fact , as we will next establish in section [ proof - section ] , all two- " dualities satisfying eq .
( [ f1f2 ] ) must be of the form of eqs .
( [ f1f2-conformal ] ) .
specifically , all duality mappings that can be made meromorphic by a change of variables , _ can only be of the fractional linear type_. this _ uniqueness _ may rationalize the appearance of fractional linear ( dual ) maps in disparate arenas ranging from statistical mechanics models , such as the ones that we study here , to s - dualities in , e.g. , ym theories .
thus far , we focused on `` one- '' and `` two - dualities '' for which the coupling constants satisfy either eq .
( [ ffz ] ) or eq .
( [ f1f2 ] ) , respectively .
our calculations may be extended to `` @xmath54-duality '' transformations for which @xmath55 as the reader may verify , replicating the considerations invoked in the next section leads to the conclusion that if they are meromorphic each of the functions @xmath56 ( with @xmath57 ) in eq .
( [ ff ... f ] ) must be of the fractional linear ( conformal ) form @xmath58 with @xmath59 and @xmath60 being constants . in general
, whether a function @xmath1 solving eq .
( [ ffz ] ) for all @xmath40 is meromorphic in appropriate coordinates or not , it is impossible that any such function @xmath45 obeying eq .
( [ ffz ] ) will map the entire complex plane ( or riemann sphere ) onto a subset @xmath61 of the complex plane ( or riemann sphere ) .
this subset @xmath61 could be a disk or strip or any other subset of the complex plane .
that is , it is impossible that a solution to eq .
( [ ffz ] ) will be afforded by a function @xmath1 which for all complex @xmath40 , will map @xmath62 . the proof of this latter assertion is trivial and will be performed by contradiction : consider a point @xmath63 , then a single application of @xmath1 on @xmath64 leads to an image @xmath65 . as for all points @xmath40 ( including those that lie in @xmath61 ) the image @xmath45 is in @xmath61 , we have @xmath66 .
however , as stated in the beginning of our proof , @xmath63 .
this thus shows that @xmath67 . in other words ,
( [ ffz ] ) can not be satisfied by such a function .
thus , if we regard the map @xmath68 as a finite `` time evolution '' ( or `` flow '' in the parlance of rg ) , the function @xmath45 must `` evolve '' @xmath40 as an `` _ incompressible fluid _ '' with area preserving dynamics in the complex plane ( or riemann sphere ) .
this flow must be of period two in order to satisfy eq .
( [ ffz ] ) .
charles babbage , `` the father of the computer '' , @xcite and others since , e.g , @xcite , have shown that the functional equation problem of eq .
( [ ffz ] ) enjoys an infinite number of solutions .
this observation can be summarized as follows : given a particular solution @xmath69 to babbage s equation , @xmath70 , a very general class of solutions can be written as @xmath71 where @xmath72 is an _ arbitrary _ ( or in a physics type nomenclature,``gauge like '' ) function with a well defined inverse @xmath73 . in other words , if we have a particular solution we can find other solutions using a function @xmath72 with and inverse defined in a specific domain .
that is , @xmath74 to make babbage s observation clear , we note that if , as an example , we examine the mbius transformation ( figure [ mobius.fig ] ) of eq .
( [ centralm * ] ) , @xmath75 , and consider @xmath76 and a particular branch @xmath77 for complex @xmath78 ( or the standard @xmath79 function for real @xmath80 ) then it is clearly seen that @xmath81 is also a solution to the equation @xmath82 . similarly ,
if we choose @xmath83 then @xmath84 which the astute reader will recognize as the transformation of eq .
( [ central * ] ) .
we now turn to a rather trivial yet as far as we are aware _ new result
_ concerning this old equation that we establish here . we assert that if there exists a transformation @xmath72 that maps complex numbers @xmath40 on the riemann sphere , @xmath85 , such that the resulting function @xmath1 is meromorphic then _ any such function @xmath1 solving eq .
( [ ffz ] ) must be of the fractional linear form ( a particular conformal map ) _ of eq . .
of course , a broad class of functions of the form of eq .
( [ bab ] ) may be generated by choosing arbitrary @xmath72 that have an inverse yet all possible rational functions will be of the fractional linear form .
for instance , the function @xmath81 discussed in the example above is , obviously , not of a fractional linear form .
* proof : * the proof below is done by contradiction . a general meromorphic function on the riemann sphere is a rational function , i.e. , @xmath86 with @xmath87 and @xmath88 relatively prime polynomials .
( if the polynomials @xmath89 and @xmath90 are not relatively prime then we can obviously divide both by any common factors that they share to make them relatively prime in the ratio appearing in eq .
( [ pq ] ) ) . as a first step
, we may find the solution(s ) @xmath91 to the equation @xmath92 unless both @xmath93 and @xmath94 are linear in @xmath91 , there generally will be ( by the fundamental theorem of algebra ) more than one solution to this equation ( or , alternatively , a single solution may be multiply degenerate ) . that is ,
unless @xmath89 and @xmath90 are both linear in @xmath91 , the polynomial @xmath95 will be of order higher than one ( @xmath96 ) in @xmath91 and will , for general @xmath40 , have more than one different ( non - degenerate ) zero . when varying @xmath40 over all possible complex values , it is impossible that the polynomial @xmath97 will always have only degenerate zero(s ) for the relatively prime @xmath93 and @xmath94 ( we prove this in the rather simple _ ( multiplicity ) lemma _ below ) .
we denote the general zeros of the polynomial @xmath97 by @xmath98 .
that is , @xmath99 now if @xmath48 , then all solutions @xmath100 to the equations @xmath101 ( for which the polynomial ( in @xmath40 ) , @xmath102 vanishes ) will , for all @xmath103 , solve the equation @xmath104 in the last equation above , on the righthand side there is a single ( arbitrary ) complex number @xmath40 whereas on the lefthand side there are _ multiple _ ( see , again , the ( multiplicity ) lemma ) viable different solutions @xmath105 thus , at least one of the solutions in this set @xmath106 .
we denote one such solution by @xmath107 .
putting all of the pieces together , the equation @xmath0 can not be satisfied for all complex @xmath40 ( in particular , it is not satisfied for @xmath108 ) .
thus , both @xmath87 and @xmath88 must be linear in @xmath40 , and the fractional linear form of eq .
( [ eqbaba ] ) follows once it is restricted to this class . replicating the above steps _ mutatis mutandis _ for `` two - dualities '' satisfying eq .
( [ f1f2 ] ) similarly leads to the conclusion that if the _ transformations are meromorphic they must be given by ratios of linear functions ( and thus conformal)_. in this case , @xmath109 can be a general fractional linear transformation with a finite determinant and further constraints on @xmath110 are afforded by the requirement that eq .
( [ f1f2 ] ) is indeed obeyed .
the calculation then leads to the result of eq .
( [ f1f2-conformal ] ) .
we will elaborate on this restriction in section [ section - n ] . * ( multiplicity ) lemma : * we prove ( by contradiction ) that it is impossible for @xmath97 ( eq . ( [ wpz ] ) ) to have an @xmath111-th order ( @xmath96 ) degenerate root for all @xmath40 .
assume , on the contrary , that @xmath112 with @xmath113 and @xmath114 functions of @xmath40 , @xmath96 , and @xmath93 , @xmath94 , relatively prime polynomials of @xmath91 .
at @xmath115 ( with infinitesimal @xmath116 ) , the degenerate root is given by @xmath117 that is , by definition , @xmath118 we next use the taylor expansion @xmath119 given the above form of @xmath97 , its partial derivative @xmath120 at @xmath121 , for @xmath96 .
similarly , @xmath122 .
lastly , from eq .
@xmath123 putting all of the pieces together , @xmath124 therefore , @xmath121 is a root of @xmath94 .
as the root of @xmath94 is independent of @xmath40 , this implies that the assumed multiply degenerate root ( i.e. , @xmath114 ) of @xmath97 is independent of @xmath40 , i.e. @xmath125 . recall ( eq .
( [ wpz ] ) ) that @xmath126 .
as @xmath127 is ( for all @xmath40 ) a root of both @xmath128 and @xmath94 , it follows that @xmath129 is also a root of @xmath93 .
it follows that both @xmath93 and @xmath94 share a root ( and a factor of ( @xmath130 ) when factorized to their zeros ) , e.g. , when written as @xmath131 with @xmath132 and @xmath133 constants and with @xmath134 and @xmath135 the roots of @xmath93 and @xmath94 respectively , at least one of the zeros ( @xmath134 ) of @xmath93 must be equal to one of the zeros ( @xmath135 ) of @xmath94 .
thus , @xmath93 and @xmath94 are not relatively prime if @xmath96 .
this , however , is a contradiction and therefore establishes our assertion and proves this lemma .
thus far , we largely focused on `` two-''dualities satisfying eq .
( [ ffz ] ) .
the ideas underlying our proof in section [ proof - section ] illustrated that all meromorphic dualities must be of the fractional linear form , eq .
( [ abcd ] ) . as elaborated , when applied to `` two-''dualities satisfying eq .
( [ ffz ] ) , the most general meromorphic solution is that of eq .
( [ eqbaba ] ) . similarly , more general dualities for which eq .
( [ f1f2 ] ) is obeyed enjoy more solutions ( such as those afforded by eq .
( [ f1f2-conformal ] ) ) .
we now explicitly solve the general case of eq .
( [ ff ... f ] ) . as proven , the fractional linear transformations , eq .
( [ fkz ] ) , are the only possible meromorphic solutions .
we thus confine our attention to these . in
what follows , we will invoke the composition property of eq . ( [ group ] ) . on the right hand side of eq .
( [ ff ... f ] ) , the function @xmath40 may be expressed in matrix form as @xmath136 with @xmath137 an arbitrary complex number .
this is so as the matrix elements @xmath138 are such that , rather trivially , the associated fractional linear function of eq .
( [ abcd ] ) is @xmath139 . if all functions @xmath140 , in eq . (
[ ff ... f ] ) are of the same form of eq .
( [ abcd ] ) , then when the representation of eq .
( [ vi ] ) is inserted we will trivially have @xmath141 whose solutions are straightforward .
when diagonalized by a unitary transformation , the matrix @xmath142 must only have @xmath54-th roots of @xmath137 .
thus , @xmath143 with @xmath144 arbitrary integers and @xmath145 any @xmath146 unitary matrix .
the latter may , of course , most generally be written as @xmath147 $ ] with @xmath148 , the triad of pauli matrices , @xmath149 an arbitrary real number and @xmath150 a unit vector .
the factorization of @xmath151 was performed in eq .
( [ matrix ] ) because , as we briefly remarked earlier , a uniform scaling of all four elements of the general @xmath146 matrix does not alter the fractional linear transformation of eq .
( [ abcd ] ) .
all possible dualities are exhausted by _ the space spanned by all of the matrices _ @xmath152 of the form of eq .
( [ matrix ] ) , and a duality with real @xmath153 can then be interpreted as an induced map on the euclidean @xmath32 sphere ( or , more precisely , one of its hemispheres as we will explain shortly ) . in the case of @xmath154
( i.e. , that of eq .
( [ ffz ] ) ) , the only non - trivial solution ( i.e. , non - identity matrix ) solution of the form of eq .
( [ matrix ] ) is formed by having @xmath155 ( @xmath156 ) . when this occurs , eq .
( [ matrix ] ) becomes @xmath157 the solution of eq .
( [ mform ] ) is , of course , identical to that of eq .
( [ eqbaba ] ) once we set @xmath158 .
for example , the ising model duality of eq .
( [ centralm * ] ) is associated with the unit vector @xmath159 .
we thus see how the particular solutions that we obtained earlier are a particular case of this more general approach . for `` two-''dualities with real @xmath153 , any point on the southern hemisphere ( i.e. , one with @xmath160 )
is associated with a different transformation .
this is so as scaling the global multiplication of the matrix by @xmath161 ( associated with @xmath162 ) does not alter the fractional linear transformation of eq .
( [ abcd ] ) .
this space spanned by the hemisphere is , of course , identical to that of the @xmath163 group associated with nematic liquid crystals having a two - fold homotopy group , @xmath164 and two associated possible defect charges .
geometrically , we may thus understand dualities by thinking of the space spanned by the group elements . in a similar vein , in the @xmath54-duality solution of eq .
( [ matrix ] ) , the eigenvalues of @xmath142 are any two roots of the identity ( or stated equivalently , any two elements of the cyclic group @xmath165 ( which , on its own , form the center of the group @xmath166 ) multiplying @xmath151 .
we now return to the general problem posed by eq .
( [ ff ... f ] ) . repeating our arguments
thus far , it is readily seen that the most general meromorphic solution is afforded by the fractional linear maps of eq .
( [ fkz ] ) with the @xmath54-th @xmath146 matrix ( associated with the fractional linear map @xmath167 ) set by the inverse of all others .
that is , rather explicitly , @xmath168^{-1}.\end{aligned}\ ] ]
the considerations of sections [ proof - section ] and [ section - n ] can be extended to not only two but also to many coupling constants @xmath169 , @xmath170 . in the particular case of two coupling constants ,
the duality mapping will be of the form @xmath171 , where the functions @xmath172 and @xmath173 must be fractional linear maps of two complex variables @xcite . to obtain the proper fractional linear map in several variables
, one has to remember that it is important to preserve the composition property of these maps , that is , the application of two of these maps should generate another fractional linear map . consider a fractional linear map @xmath174 involving two complex variables @xmath175 where all the coefficients @xmath176 ( @xmath177 ) are complex numbers .
then , it is straightforward to verify that the composition of these generalized fractional linear maps , @xmath178 , generates another fractional linear map and induces a , non - abelian in general , group structure .
that is , we may associate with each fractional linear map a @xmath179 matrix @xmath180 given by @xmath181 with a determinant @xmath182 . as can be explicitly verified , the composition of maps corresponds to matrix multiplication .
moreover , we can re - scale all coefficients by the ( in general , complex ) factor @xmath183{\delta}$ ] without affecting the map , so that the re - scaled ( associated ) matrix has a determinant equal to unity .
the subset of @xmath184 complex matrices with determinant @xmath185 forms a group denoted @xmath186 .
the fixed points of the transformation , @xmath187 , solve the equations @xmath188 when these are satisfied @xmath189 and @xmath190 is the solution of a cubic equation ( there are obviously three such cubic equation solutions ) .
armed with the above , we now investigate the extension babbage s equation of eq .
( [ ffz ] ) with two variables @xmath191 . that is
, we now explicitly solve the equation @xmath192 there are several solutions to this equation .
an important class , characterized by non - zero values of @xmath193 and @xmath194 is given by @xmath195 this solution constitutes a generalization of eq .
( [ eqbaba ] ) to the case of @xmath196 complex variables for `` two-''dualities .
the generalization of these ideas to @xmath197 is formally straightforward leading to the @xmath198 group structure .
the geometry of these mappings is a very interesting mathematical problem beyond the scope of the current paper .
we will now examine _ constraints that stem from the fractional linear maps _ that we found , i.e. , a particular set of conformal transformations .
a highlight of the remainder of this work is that the results of eqs .
( [ eqbaba ] , [ f1f2-conformal ] , [ fkz ] , [ matrix ] , [ group+ ] ) allow for the _ partial solvability _ of many different theories . how this is done in practice
will be best illustrated by detailed calculations . to make the concepts concrete and relatively simple to follow , we will employ , in sections [ series - sec ] and thereafter , as lucid examples some of the best studied statistical mechanics models , ising models and generalized ising - type lattice gauge theories and focus on @xmath154 dualities with a single coupling constant ( @xmath199 ) . in this section
, we wish to sketch the central idea behind this technique .
let us consider an arbitrarily large yet _ finite size _
system for which no phase transition occurs and thus the partition function @xmath200 ( or any other function ) is an
_ analytic _ function of all couplings and/or temperature .
for such a finite size system , the w - c and s - c expansions ( or , correspondingly , high- and low - temperature expansions ) of @xmath200 , can often be written as finite order series ( i.e. , polynomials ) in the respective expansion parameters @xmath17 and @xmath201 .
that is , we consider the general finite order w - c and s - c series for the partition function @xmath200 ( or any other analytic function ) @xmath202 where @xmath203 are analytic functions and @xmath204 ( for which , according to eq .
( [ f1f2 ] ) , @xmath205 ) . as in eq .
( [ two_expansions ] ) , the two expansions converge to the very same function @xmath200 , we trivially have , by the transitive axiom of algebra , two equivalent relations , @xmath206 for the finite number of series coefficients @xmath207 and @xmath208 . according to the simple results of section [ proof - section ] , the functions @xmath209 appearing in the arguments of @xmath203 and in the expansion itself are of the fractional linear type , i.e. , functions of the form of eq .
( [ f1f2-conformal ] ) .
now , here is the crux of our argument : when the functions of eq .
( [ f1f2-conformal ] ) are inserted , eqs .
( [ main_point ] ) may give rise to constraints amongst the coefficients @xmath207 and @xmath210 and thus _ partially solve _ for the function @xmath200 with _ no additional input_. similarly to the `` @xmath54-duality '' mappings of section [ sec_dualconst ] , the general methods of partial solvability introduced above may be trivially extended to this more general case .
this , in particular , may also enable the examination of not only w - c and s - c series but also the matching of partition functions on finite size systems which in the thermodynamic limit will have multiple phases ( and associated series for thermodynamic quantities and partition functions ) .
( [ ff ... f ] ) applies in systems having a certain number of such regimes in each of which the partition function may be expressed as a different finite order series of the form of eq .
( [ two_expansions ] ) , i.e. , @xmath211 with @xmath212 , where @xmath111 is the number of finite order representations of the partition function @xmath200 , then we will be able to find analogs of eqs .
( [ main_point ] ) .
these , as before , will lead to partial solvability .
as the discussion above is admittedly abstract , we will now turn to concrete examples in the next few sections .
one of the most pragmatic consequences of our approach , detailed in section [ practical : sec ] and @xcite is that the complexity of determining the w - c and s - c series expansions may be trivially identical .
this lies diametrically opposite to the maxim that s - c series expansions are in many instances far harder to determine than perturbative w - c expansions @xcite .
to demonstrate our concept , we will first use standard expansions @xcite of the ising models of eq .
( [ ising_model ] ) and their generalizations . the hamiltonian @xmath213 @xmath214 . in the remainder of this work
, we will consider this and various other models on hypercubic lattices @xmath215 of @xmath216 sites in @xmath133 dimensions ( with even length @xmath217 ) , endowed with periodic boundary conditions .
unless stated otherwise , we will focus on uniform ferromagnetic systems ( @xmath218 for all lattice links @xmath219 ) .
in @xcite we consider other boundary conditions , system sizes and lattice aspect ratios , and show that our results are essentially unchanged for large systems with random @xmath220 . in the notation of earlier sections ,
the coupling constant is @xmath221 with @xmath222 the inverse temperature .
defining @xmath223 , the identity @xmath224= \cosh k [ 1 + ( s_{\sf a } s_{\sf b } ) \tilde{t}]$ ] leads to a high - temperature ( h - t ) , or w - c , expansion for the partition function @xmath225}.\end{aligned}\ ] ] the sum @xmath226 if @xmath227 at each site @xmath228 appears an even number of times and vanishes otherwise .
thus , @xmath229 where @xmath230 is the number of ( not necessarily connected ) _ loops _ of total perimeter @xmath231 ( @xmath232 ) that can be drawn on the lattice and @xmath233 . for each such loop , i.e. , @xmath234 formed by the bonds ( nearest neighbor pair products @xmath235 ) appearing in eq .
( [ htlt * ] ) , there is a complementary loop @xmath236 for which the sum of eq .
( [ htlt * ] ) remains unchanged .
consequently , the h - t series coefficients are trivially symmetric , @xmath237 .
we next briefly review the low - temperature ( l - t ) , or s - c , expansion .
there are two degenerate ground states ( with @xmath238 for all sites @xmath239 or @xmath240 ) of energy @xmath241 .
all excited states can be obtained by drawing closed surfaces marking domain wall boundaries .
the domain walls have a total @xmath242 dimensional surface area @xmath243 , the energy of which is @xmath244 .
taking into account the two - fold degeneracy , the l - t expansion of the partition function in powers of @xmath245 is @xmath246 with @xmath247 the number of ( not necessarily connected ) closed _ surfaces _ of total area @xmath248 ( @xmath249 ) .
that is , the l - t expansion is in terms of @xmath242-dimensional `` surface areas '' enclosing @xmath133-dimensional droplets .
geometrically , there are no closed surfaces of too low areas @xmath243 .
thus , in the l - t expansion of ising ferromagnets , @xmath250 where @xmath251 . the l - t coefficients exhibit a trivial _ complementarity symmetry _
akin to that in the h - t series . given any spin configuration @xmath252
, there is a unique correspondence with a staggered spin configuration @xmath253 where @xmath254 are the ( integer ) cartesian components of the hypercubic lattice site @xmath239 ( i.e. , @xmath255 ) .
domain walls associated with such staggered configuration are inverted relative to those in the original spin configuration @xmath256 .
that is , if a particular domain wall appears in @xmath256 then it will not appear in @xmath257 and vice versa . as a result , @xmath258 ( for the even @xmath217 hypercubic lattices that we consider )
we will now follow the program outlined in section [ partial : sec ] . our approach is to compare h - t and l - t series expansions of the ising ( and other arbitrary ) models by means of a duality mapping . in the ising model , the mbius transformation ( that satisfies the `` one- '' duality condition of eq .
( [ ffz ] ) ) @xmath259 relates expansions in @xmath260 to those in @xmath261 . in either of the expansion parameters @xmath262 ( i.e. , @xmath260 or @xmath261 ) , eqs .
( [ central * ] ) are examples of the fractional linear transformations discussed above .
@xmath263 is the magnetization of a single ising spin immersed in an external magnetic field of strength @xmath264 when there is a minimal coupling ( a zeeman coupling ) between the dual fields : the ising spin and the external field .
this transformation may be applied to ising models in _ all dimensions _
@xmath133 not only to the @xmath41 model for which the kw correspondence holds .
these transformations emulate , yet are importantly _ different from _ , a @xmath12 correspondence ( the latter never enables an equality of two finite order polynomials in the respective expansion parameters ) . employing the second of eqs .
( [ central * ] ) , @xmath265.\end{aligned}\ ] ] by virtue of eq .
( [ htlt * ] ) , this can be cast as a finite order series in @xmath260 multiplying @xmath266 .
indeed , by invoking @xmath267 and the binomial theorem , @xmath268%\big],\end{aligned}\ ] ] where @xmath269 analogously , @xmath270% \big ] .
\end{aligned}\ ] ] equating eqs .
( [ htlt * ] ) and ( [ ltt+ * ] ) and eqs.([ltt * ] ) and ( [ htt+ ] ) and invoking eq .
( [ zero ] ) leads to a linear relation among expansion coefficients , @xmath271 where @xmath272 and @xmath89 are , respectively , @xmath273component and @xmath274component vectors defined by @xmath275 in eq .
( [ wvp ] ) , the rectangular matrix @xmath276 where the @xmath277 matrix @xmath278 is equal to @xmath279 with a square matrix @xmath280 whose elements @xmath281 ( @xmath282 ) are given by eq .
( [ ad+ ] ) . constraints ( [ zero ] ) are captured by @xmath283 in eq .
( [ wmd ] ) , @xmath284 , where the matrix elements @xmath285 , and @xmath286 otherwise ; @xmath287 is a @xmath288 null matrix .
apart from the direct relations captured by eq .
( [ wvp ] ) that relate the h - t and l - t series coefficients to each other , there are additional constraints including those ( i ) originating from equating coefficients of odd powers of @xmath260 and @xmath261 to zero and ( ii ) of trivial symmetry related to complimentary loops / surfaces in the h - t and l - t expansion that we discussed earlier , @xmath289 and @xmath290 .
it may be verified that these restrictions are already implicit in eq .
( [ wvp ] ) .
notably , as the substitutions @xmath291 , @xmath292 in eq .
( [ ad+ ] ) show , eqs .
( [ ltt+ * ] ) and ( [ htt+ ] ) are , respectively , invariant under the two independent symmetries @xmath293 and @xmath294 and thus the linear relations of eq .
( [ wvp ] ) adhere to these symmetries .
thus , the equalities between the lowest ( small @xmath295 ) and highest ( i.e. , ( @xmath296 ) ) order coefficients are a consequence of the duality given by eq .
( [ central * ] ) that relates expansions in the w - c and s - c parameters . the total number of unknowns ( series coefficients ) in eq .
( [ wvp ] ) is @xmath297 with 1/2 of these unknowns being the h - t expansion coefficients and the other 1/2 being the l - t coefficients ( the components @xmath298 ) . in @xcite
( in particular , table 1 therein ) , we list the rank ( @xmath299 ) of the matrix @xmath300 appearing in eq .
( [ wvp ] ) for different dimensions @xmath133 and number of sites @xmath301 .
as seen therein , for the largest systems examined the ratio @xmath302 tends to 3/4 suggesting that in _ all _ @xmath133 only @xmath303 of the combined l - t and h - t coupling series coefficients need to be computed by combinatorial means .
the remaining @xmath304 are determined by eq .
( [ wvp ] ) .
this fraction might seem trivial at first sight .
if , for instance , the first 1/2 of the h - t coefficients @xmath305 are known ( i.e. , those with @xmath306 ) then the remaining h - t coefficients @xmath305 ( with @xmath307 ) can be determined by the symmetry relation @xmath308 and once all of the h - t series coefficients are known ( and thus the partition function fully determined ) , the partition function may be written in the form of eq .
( [ ltt * ] ) and the l - t coefficients @xmath309 extracted . thus by the symmetry relations alone knowing a @xmath310 of the coefficients alone suffices .
the symmetry relations are a rigorous consequence of the duality relations for any value of @xmath301 .
as the duality relations may include additional information apart from symmetries , it is clear that @xmath311 for finite @xmath301 ( i.e. , knowing more than a 1/4 of the coefficients is not necessary in order to evaluate all of the remaining h - t and l - t coefficients with the use of duality ) . for a given aspect ratio ,
the smaller @xmath301 is ( and the smaller the number of unknowns @xmath312 ) , the additional relations of eqs .
( [ zero ] ) carry larger relative weight and the ratio @xmath302 may become larger .
thus , 3/4 is its lower bound .
indeed , this is what we found numerically for all ( non self - dual ) systems that we examined @xcite . as @xmath133 increases ,
the lowest non - vanishing orders in the l - t expansion become more separated and eqs .
( [ wvp ] ) become more restrictive for small @xmath301 systems @xcite .
the h - t and l - t series are of the form of eqs .
( [ htlt * ] ) and ( [ ltt * ] ) for all geometries that share the same minimal @xmath133 dimensional hypercube ( i.e. , of minimal size @xmath313 ) of @xmath314 sites .
thus , equating the series gives rise to linear relations of the same form for both a hypercube of size @xmath216 ( with general even @xmath217 ) as well as a tube of @xmath315 hypercubes stacked along one cartesian direction .
however , although the derived linear equations are the same , the partition functions for systems of different global lattice geometries are generally dissimilar ( indicating that the linear equations can never fully specify the series ) .
thus , the set of coefficients not fixed by the linear relations depends on the global geometry .
parity and boundary effects may influence the rank @xmath299 of the matrix @xmath300 in eq .
( [ wvp ] ) . as demonstrated in @xcite for @xmath41 lattices in which ( at least ) one of the cartesian dimensions @xmath217 is _ odd _ , as well as systems with non - periodic boundary conditions , @xmath316 .
that is , in such cases @xmath317 of the coefficients need to be known before eq .
( [ wvp ] ) can be used to compute the rest . as explained in @xcite , symmetry and
duality arguments can be enacted to show that in these cases , @xmath318 for finite @xmath301 , i.e. , its lower bound is 2/3 .
a further restriction is that of discreteness the coefficients @xmath319 ( counting the number of loops / surfaces of given perimeter / surface area ) must be non - negative integers for the ferromagnetic ising model .
let us illustrate the concepts above with a minimal periodic @xmath146 ferromagnetic ( @xmath320 ) system with hamiltonian @xmath321 $ ] . from eqs .
( [ htlt * ] , [ ltt * ] ) @xmath322 , \nonumber \\ { \cal z}_{\sf l - t}&= & 2 e^{8k } [ 1 + c'_{2 } e^{-4k } + c'_{4 } e^{-8k } \nonumber \\ & & + c'_{6 } e^{-12k } + c'_{8 } e^{-16k } ] .
\end{aligned}\ ] ] invoking eqs .
( [ vp],[wmd ] ) , @xmath323 , @xmath324 , and @xmath325 there are @xmath326 unknown coefficients in eq . ( [ wvp ] ) ;
the rank ( @xmath299 ) of the matrix @xmath327 is eight .
thus , in this minimal finite system , the eqs .
( [ wvp ] ) are linearly independent ( @xmath328 ) and all coefficients may be determined @xmath329 generally , not all coefficients may be determined by duality alone . as we discussed , in the large system limit , @xmath330 .
a @xmath331 example appears in @xcite .
if @xmath332 independently on each lattice link @xmath333 , then eqs .
( [ zero ] ) need not hold . instead of eq .
( [ wvp ] ) , we have @xcite @xmath334 this less restrictive equation ( by comparison to eq .
( [ wvp ] ) ) , valid for all @xmath332 , is of course still satisfied by the ferromagnetic system . for the matrix @xmath335
, a large system value of @xmath336 is still obtained @xcite ( see table 2 therein ) .
the partition functions for different @xmath332 realizations will be obviously different . nevertheless , all of these systems will share these linear relations @xcite . unlike the ferromagnetic system , the integers
@xmath337 may be negative .
computing the partition function of general binary spin glass @xmath41 ising models is a problem of _ polynomial _ complexity in the system size .
when @xmath338 , _ the complexity becomes that _ of an np complete problem @xcite .
therefore , our equations partially solve and `` _ localize _ '' np - hardness to only a fraction of these coefficients ; the remaining coefficients are determined by linear equations .
the complexity of computing @xmath339 , required for each element of @xmath335 , is @xmath340 .
our equations enable a polynomial ( in @xmath301 ) consistency checks of partition functions . in performing the expansions of eqs .
( [ htlt * ] ) and ( [ ltt * ] ) , the complexity of determining the number of loops ( or surfaces ) of given size @xmath341 ( or @xmath342 ) ( i.e. , the coefficients @xmath230 or @xmath343 ) increases rapidly with @xmath341 ( or @xmath342 ) .
our relations may be applied to systematically simplify the calculation of these coefficients . as we now explain , the situation becomes exceedingly transparent in the ising models discussed thus far . for these theories ,
the coefficients are symmetric : @xmath344 , @xmath345 . by virtue of these symmetries that are embodied in the duality relations of eq .
( [ svq ] ) , it is clear that if the lower 1/2 of the h - t coefficients @xmath346 ( or , similarly , the lower 1/2 of l - t coefficients .
i.e. , @xmath347 ) , i.e. , a 1/4 of the combined h - t and l - t series coefficients , were known then the remaining h - t ( or l - t ) coefficients are trivially determined .
then , armed with either the full h - t ( or l - t ) series , the exact partition functions can be equated @xmath348 and written in the form of eqs .
( [ htlt * ] ) and ( [ ltt * ] ) to determine the remaining unknown l - t ( or h - t ) coefficients .
that is , once the partition functions are known , the series expansions ( and thus coefficients ) are uniquely determined . by construction ,
( [ svq ] ) incorporates , of course , the relation @xmath349 which forms the core of our analysis .
thus , as the symmetry is a consequence of the duality relations , it is clear that knowing a 1/4 of the combined h - t and l - t coefficients suffices to determine all of them via eq .
( [ svq ] ) , i.e. , that the required fraction of coefficients to find all of the others via duality satisfies the inequalty @xmath350 .
as the asymptotic ratio of @xmath336 suggests , and as we have verified , knowing the first 1/4 of both the h - t and l - t coefficients ( i.e. , those with @xmath351 and @xmath352 ) instead of 1/2 of the h - t ( or l - t ) coefficients discussed above , suffices to completely determine all other coefficients . as the difficulty of evaluating coefficients increases rapidly with their order , systematically computing this 1/4 lowest order coefficients ( @xmath353 ) is less numerically demanding than computing the first 1/2 of all the h - t coefficients ( @xmath346 ) , or calculating the first 1/2 all of the l - t coefficients ( @xmath354 ) .
the central idea underlying our approach is that , for finite size systems , the h - t and l - t series expansions are different representations of the very same partition function , eq . ( [ too_trivial ] ) .
this equality followed from the analyticity of the partition function on any ( arbitrary size yet ) finite size system .
as the astute reader noted throughout all earlier sections , this relation forms the nub of the current study .
it is worthwhile to step back and ask what the practical implications of our results are for disparate h - t and l - t series expansions ( or other w - c and s - c series ) .
first and foremost , eq . ( [ too_trivial ] ) implies , of course , that the generation of the h - t and l - t series on finite size lattice are equally hard , as obtaining one immediately yields the other . as stated by certain insightful textbooks , e.g. , @xcite ,
the h - t and l - t expansions differ in their conceptual premise .
for instance , as @xcite notes , `` the derivation of a high - temperature expansion is , in principle , straightforward '' , since it just amounts to counting the number of closed loops , while , as befits the more meticulous examination of the ground states and myriad possible excitations about them , it may seem that the generation of lengthy low - temperature series is a highly specialized art " .
much work has been devoted to a finite lattice method that improves the bare h - t and l - t series ( as in , e.g. , the h - t loop tallying briefly reviewed in section [ series - sec ] ) @xcite .
many specialized texts @xcite laud the simplifying features of general h - t expansions vis a vis their l - t counterparts , including commending their features such as `` smoothness '' @xcite , the uniform sign of the h - t coefficients in disparate theories , and their applicability to gapless systems @xcite . in a more recent detailed exposition @xcite
, it was noted that while the high - temperature series are well - behaved the situation at low temperatures is less satisfactory , in particular above two dimensions " . in a related vein
, we remark that the h - t series are well known to naturally relate to one of the oldest and simplest expansions that of the virial coefficients @xcite as well as large-@xmath54 expansions @xcite .
thus , with all of the above , it would generally seem that h - t and l - t qualitatively differ .
however , as seen by eq .
( [ too_trivial ] ) and the linear equations that we derived in earlier sections connecting the h - t and l - t expansions , the complexity of deriving either expansion on all general finite size lattices is the same .
thus for finite size lattices with finite order h - t and l - t series related by a transformation of their expansion parameter , the general maxim concerning the different intrinsic complexity of the h - t and l - t expansions does not hold . concretely , we may derive h - t coefficients from l - t coefficients and vice versa from the simple relation of eq .
( [ too_trivial ] ) . in the case of the ising model that formed much of the focus of the current study , from eq .
( [ wvp ] ) we have that @xmath355 . \label{lh}\end{aligned}\ ] ] in @xcite , we apply our method to derive the h - t expansions from their l - t counterparts on finite size periodic two- and three - dimensional lattices @xcite .
it is notable that our method _ applies to non - trivial systems such as the three - dimensional ising model_. our relations enable _ a consistency check of proposed series solutions _ and the derivation of the entire series from a knowledge of only a fraction of coefficients .
indeed , we verified that the l - t series provided in @xcite satisfy the linear equations of eq .
( [ wvp ] ) ( and our derived h - t series adhere to the same relations ) .
as we explained in section [ equal ] for regular uniform coupling systems , and in section [ psolve ] for less constrained non - uniform systems , a partial knowledge of both the l - t and/or h - t series may enable a construction of the full partition function . as we have reiterated earlier and do
so once again here , our approach applies to arbitrarily large yet finite size lattices .
_ mathematical identities _ are system independent and enable the general transformation of one set of objects into another . as such
, they are reminiscent of dualities .
symbolically , let us consider particular partition functions ( or generating functions " ) @xmath356 that encode all quantities that we wish to determine in a particular set of systems .
if certain identities universally apply , we may invoke these relations to transform each function into an equivalent dual , and formally rewrite @xmath357 for the two sets of functions . in eq .
( [ z1z2 ] ) , @xmath358 can be interpreted as the set of generating functions of very different problems or physical systems . as such , dualities and ,
in particular , the universal relations that we obtained from conformal transformations linking dual systems may encode very general mathematical relations . in what follows ,
we concretely demonstrate that dualities may lead to an extensive number of ( new ) mathematical relations such as those connecting the number of surfaces and volumes of a particular size .
these relations are already contained in our previously derived eqs .
( [ svq ] ) .
the key conceptual point is that dualities between different types of partition functions ( irrespective of the general coupling constants associated with a large set of such functions ) can hold generally by virtue of mathematical identities .
wegner s duality @xcite relates interactions between @xmath359 on the boundaries of `` @xmath21 dimensional cells '' to generalized ising gauge type models with interactions between @xmath360 on the boundaries of `` @xmath361 dimensional cells '' .
these generalized ising lattice gauge theories are given by the hamiltonian @xmath362 with @xmath214 and @xmath363 general coupling constants . here
, a `` @xmath364 dimensional cell '' corresponds to a ( one - dimensional ) nearest neighbor edge ( i.e. , one whose boundary is @xmath365 ) associated with standard @xmath366 interactions that we largely focused on thus far ( i.e. , the ising model hamiltonian of eq . ( [ ising_model ] ) .
the case of @xmath367 corresponds to a product of four @xmath256 s at the centers of the four edges which form the boundary of a two - dimensional plaquette ( as in standard hypercubic lattice gauge theories ) .
that is , @xmath367 corresponds to the lattice gauge hamiltonian @xmath368 where the link variables @xmath369 , and with @xmath370 being the standard `` @xmath367 dimensional '' cells ( i.e. , square plaquettes ) whose one - dimensional boundary @xmath371 is the formed by the nearest neighbor one - dimensional links @xmath372 , @xmath373 , @xmath374 , and @xmath375 .
the case of @xmath376 corresponds to the product of six @xmath256 s at the center of the six two - dimensional faces which form the boundary of a three - dimensional cube , etc .
the hamiltonian is the sum of products of @xmath377 @xmath256 s on the boundaries of all of the @xmath21 dimensional hypercubes in the lattice ( in a lattice of @xmath378 sites there are @xmath379 such hypercubes and @xmath380 ising variables @xmath256 at the centers of their faces ) .
if the dimensionless interaction strength for a @xmath21 dimensional cell is @xmath363 then the couplings in the two dual models will be related by eqs .
( [ central * ] ) or , equivalently , @xmath381 .
the @xmath382 duality corresponds to the duality between the @xmath383 ising model and the @xmath383 ising gauge theory .
the @xmath384 case is that of the kw self - duality . for general @xmath21
, wenger derived his duality from an equivalence between the h - t and l - t coefficients .
we now turn to _ new , and rather universal , geometrical results _ obtained by our approach that hold in general dimensions . if the ground state degeneracy is @xmath385 ( e.g. , @xmath386 for the standard ( @xmath364 ) ising models , @xmath387 in @xmath388 ising gauge theories with periodic boundary conditions )
, then we find @xcite that , _ irrespective of the coupling constants _ , the h - t and l - t series for these models are given by eqs .
( [ htlt * ] ) and ( [ ltt * ] ) with the following substitutions @xmath389 thus , eq . ( [ svq ] ) obtained for standard ( @xmath364 ) ising models also holds for general @xmath21 following this substitution .
_ in systems with @xmath21 dimensional cells , @xmath390 and @xmath391 denote , respectively , the number of closed surfaces of total @xmath21 and @xmath361 dimensional surface areas equal to @xmath295 . _ by building on our earlier results , we observe that , when the hypercubic lattice length @xmath217 is even , _ eq . _
( [ svq ] ) _ universally relates , in any dimension @xmath133 ( and for any @xmath21 ) _ , these numbers to each other leaving only @xmath392 of these undetermined . by comparison to eq .
( [ svq ] ) , additional geometrical conditions that hold for @xmath364 ( eqs .
( [ zero ] ) ) produce the slightly more restrictive eq .
( [ wvp ] ) .
similar additional constraints appear for @xmath393 .
a kw type self - duality present for @xmath394 leads to linear equations that relate @xmath395 ( the number of surfaces of total @xmath396-dimensional surface area @xmath397 ) to themselves .
we next explicitly discuss the @xmath41 , @xmath364 case ( i.e , the standard @xmath41 ising model ) .
similar considerations hold for any @xmath394 system .
more information can be gleaned for self - dual systems , e.g. , the kw self - duality of the @xmath41 ising model . in this model , @xmath398 ( as @xmath305 and @xmath399 are both the number of closed @xmath364 dimensional loops of length @xmath295 ) when eqs .
( [ wvp ] ) are applied to large systems ( @xmath400 ) , see @xcite .
consequently , the number of coefficients that need to be explicitly evaluated is nearly 1/2 of those obtained by matching the h - t and l - t expansions without invoking self - duality @xcite : @xmath401 of the coefficients are determined by self - duality once @xmath402 of the coefficients are provided .
we caution that the relation @xmath398 is only asymptotically correct in the limit of large system sizes .
consequently , we find @xcite that @xmath302 asymptotically approaches @xmath403 from below ( and not from above as it would have if this relation were exact for finite size systems @xcite ) as @xmath301 becomes larger .
we demonstrated that _ all meromorphic duality transformations on the riemann sphere ( satisfying a generalized form of babbage s equation ) must be a conformal map of the fractional linear type _ ( and simple generalizations in the case of multiple coupling constants ) , in the appropriate coupling constants .
the bulk of our analysis was focused on investigating the consequences of such general duality maps . as we demonstrated in this work
, these maps may lead to linear constraints relating finite order series expansions of two dual models .
we speculate that in models with numerous isometries ( e.g. , @xmath404 supersymmetric ym theories @xcite ) , much of the theory might become encoded in relations analogous to the linear equations studied here . employing cramer s rule and noting that the determinants of the matrices appearing therein are volumes of polytopes spanned by vectors comprising the columns of these matrices , relates series amplitudes to _ polyhedral volumes _ @xcite . in @xmath404
supersymmetric ym theories , polyhedral volume correspondences for scattering amplitudes led to a flurry of recent activity @xcite .
a main theme of our approach is that the analyticity of any quantity ensures that its different series expansions must match for all values of the coupling constants .
consequently , a main outcome of our study is that the _ mere existence _ of two or more such finite order series expansions of a theory , related by dualities ( of the form of eq .
( [ abcd ] ) ) , may `` partially solve '' that theory . by _ partial solvability
_ we allude to the ability to compute , with complexity polynomial in the system size , the full partition function @xmath200 , for instance , given partial information ( e.g. , a finite fraction @xmath405 of all series coefficients in the examples discussed in this work ) .
stated equivalently , we saw how to systematically exhaust all of the information that duality relations between disparate systems provides .
this yields restrictive linear equations on the combined set of series coefficients of the dual systems .
these equations allow for more than the computation of one set of ( e.g. , low - temperature ( l - t ) or strong - coupling ( s - c ) ) coefficients in terms of the other half ( e.g. , high - temperature ( h - t ) or weak - coupling ( w - c ) ) . in ising models and generalized ising gauge ( i.e. , wegner type ) theories on even length hypercubic lattices in general dimensions @xmath133 , only @xmath303 of the coefficients were needed as an input to fully determine the partition functions ; in the self - dual planar ising model only @xmath406 of the coefficients were needed as an input the self - duality determined all of the remaining coefficients by linear relations . for an ising chain ,
the h - t series expansion contains only one ( two ) term(s ) for open ( periodic ) boundary conditions , i.e. , @xmath407)$ ] , thus trivially all coefficients are determined . as ising models on varied @xmath408 lattices and random ising spin glass systems all
solve a common set of linear equations , our analysis demonstrates that properties such as _ critical exponents can not _ be determined by dualities alone . to avoid confusion ,
we briefly elaborate on this point .
although all of the properties may , of course , be determined by the series coefficients ( especially when investigated via powerful tools such as pad approximants @xcite and numerous others ) , the information supplied by the duality relations _ on their own _ does not suffice to establish the exact critical exponents some direct calculations of the coefficients must be invoked .
our linear relations might nevertheless prove useful in evaluating critical exponents more efficiently as they allow for a double pincer approach in which the h - t and l - t series inform about each other .
for the even size hypercubic lattices with periodic boundary conditions studied in this work there are no closed loops ( surfaces ) of an odd length .
consequently , @xmath409 for odd @xmath341 or odd @xmath342 as we have invoked .
if we were to formally allow for additional odd @xmath341 or @xmath342 coefficients then the ratio @xmath410 instead of the values of @xmath302 that we derived ( see table [ table0 ] ) .
however , when the conditions @xmath411 for odd @xmath341 are imposed for the h - t coefficients these lead ( via duality ) to non - trivial constraints on the l - t series coefficients @xmath412 with @xmath413 linear functions .
( similarly , a vanishing of the l - t series coefficients leads to non - trivial relations amongst the h - t coefficients . )
these constraints lead to @xmath414 and to the universal geometric equalities discussed earlier .
we earlier obtained lower bounds on @xmath302 using a complementarity symmetry ; the linear constraints may relate to the complementarity of the coefficients . from _ a practical point of view _ , we explained and showed how _
s - c series expansions may be generated from their w - c counterparts _ ( and vice versa ) .
thus , we saw that seemingly easily perturbative w - c ( or h - t ) and more nontrivial s - c ( or l - t ) expansions are actually identically equally hard to generate .
we applied these ideas @xcite to concrete test cases for some of the _ largest exactly known series for both two- and three - dimensional ising models on finite size lattices _
it is worth reiterating this and underscoring that this construct may be thus applied to general non - integrable systems ( such as the three - dimensional ising model , the general @xmath415 models in table [ table0 ] ) , and numerous other theories .
ising hypercubic & @xmath416 & @xmath417 + ising hypercubic spin - glass & @xmath416 & @xmath417 + wegner models & @xmath416 & @xmath417 + spin - glass wegner models & @xmath418 & @xmath417 + self - dual ising & @xmath419 & @xmath403 + honeycomb and triangular ising & @xmath419 & @xmath417 + potts hypercubic ( @xmath420 ) & @xmath416 & @xmath421 + self - dual ising gauge & @xmath422 & @xmath403 + table [ table0 ] summarizes our findings for numerous models on even size lattices in @xmath133 dimensions endowed with periodic boundary conditions @xcite . in @xcite ,
we discuss other lattice sizes and boundary conditions . with the aid of our linear equations , the np hardness of models such as the ising spin glass in finite dimensions @xmath415
is confined to a fraction @xmath405 of determining all @xmath423 coefficients in these models .
as we underscored , once these are computed , the remaining fraction @xmath302 of the coefficients are given by rather trivial linear equations .
a similar matching of series , performed in this work for the partition function , may be replicated for _ any physical quantity _ , such as matrix elements of operators , admitting a finite series expansion .
although the illustrative models shown in table [ table0 ] are all classical , all of our proofs concerning the conformal character of general dualities and the restrictions that these imply are completely general and _ hold for both classical and quantum systems_. a highly nontrivial consequence of our work is the systematic derivation of new mathematical relations via dualities . in the test case of the ising , ising gauge , and generalized wegner models explored in detail in this work ,
_ we found an extensive set of previously unknown equalities in combinatorial geometry _ given by substituting eqs .
( [ geometry_sub ] ) into eq .
( [ svq ] ) .
this work was partially supported by nsf cmmt 1106293 .
zn gratefully acknowledges insightful discussions with c. m. bender , s. kachru , p. h. lundow , and m. ogilvie .
we are extremely appreciative to per hakan lundow for further correspondence and for providing tables of series coefficients that were very instrumental for a practical application of our method . in the notation of what will follow in the current article , by `` standard dualities ''
we are alluding to the typical two - dualities " with one coupling constant @xmath3 .
for these , @xmath0 ( the parameter @xmath40 defining the theory , @xmath424 , is mapped back onto itself following two consecutive applications of @xmath1 ) . as a curiosity
, we remark that up to a trivial permutation and rescaling , the matrix of eq .
( [ mobius ] ) embodying a duality transformation , diagonalizes the transfer matrix of the one - dimensional ising chain of eq .
( [ ising_model ] ) with nearest neighbor @xmath425 ( and for which we define @xmath426 with @xmath222 the inverse temperature ) .
that is , @xmath427 that all meromorphic duality transformations must generally be of the form of eq .
( [ ww12 ] ) is seen by extending the arguments of section [ proof - section ] . here
, we briefly provide simple details explaining this assertion . if each of the functions @xmath172 and @xmath173 is meromorphic on riemann spheres associated with both @xmath428 and @xmath429 , then ( similar to eq . ( [ pq ] ) ) they must be rational , @xmath430 . for any fixed @xmath431 ,
the proof of section [ proof - section ] demonstrates that the two functions @xmath432 are both functions of the form of eq .
( [ abcd ] ) in the variable @xmath428 .
similarly , if @xmath428 is held constant , @xmath209 must become fractional linear functions in @xmath429 .
thus the four binomials @xmath433 and @xmath434 are of order no higher than linear in both @xmath428 and @xmath429 .
similarly , for any number ( @xmath435 ) of variables @xmath436 , the most general functions @xmath437 will be found to be ratios of two @xmath438-th order multinomials which are _ linear _ in each of the variables @xmath436 . in what follows ,
we return to the case of two complex variables and show how this general form is further restricted . if bilinear ( i.e. , @xmath439 ) terms are present in any of the functions @xmath440 and @xmath441 then a recursive application of the duality transformations ( i.e. , @xmath442 ) ) generally leads to the appearance of higher powers of @xmath428 and @xmath429
. thus , closure under applications by @xmath443 generally appears only when all four functions @xmath440 and @xmath441 are linear in both @xmath428 and @xmath429 with _ no bilinear terms _ allowed .
this then _ only allows _ for the generalized fractional linear transformations of eq .
( [ ww12 ] ) . if , hypothetically , in equating the h - t and l - t expansions in similar systems , the non - vanishing coefficients in the l - t expansion remain far separated and only appeared at order @xmath444 for @xmath445 then replicating our calculations leads to @xmath446 .
information regarding even a single coefficient might differentiate amongst different @xmath447 realizations ( all adhering to eqs .
( [ svq ] ) ) .
for instance , @xmath448 if and only if @xmath449 for all @xmath333 . r. haggkvist and p.h .
lundow , j. stat .
phys . * 108 * , 429 ( 2002 ) ; r. haggkvist , a. rosengren , d. andren , p. kundrotas , p.h .
lundow , and k. markstrom , phys .
e * 69 * , 046104 ( 2004 ) ; http://www.theophys.kth.se/~phl/data/index.html if the h - t and l - t coefficients were exactly equal to each other for finite size systems @xmath301 ( i.e. , if @xmath452 ) then by virtue of the symmetries @xmath453 and @xmath454 ( that are valid for all @xmath301 ) , we would have @xmath455 . in the below , we provide explicit details and examples of our method .
we will first illustrate how some of the entries in table i of the main text were determined , study spin glass systems , examine various boundary conditions and system sizes , relate dualities derived from dualities to polytope volume ratios , and then demonstrate how our approach can readily relate weak- to strong - coupling ( or h - t to l - t ) series coefficients to yield such expansions generally from one another .
this latter result shows that strong and weak coupling series expansions on finite size systems are equally hard . the equations and figures in this online section are consecutively numbered so as to follow the equations and figures of the article ( these equations and figures will be referred to throughout this online supplement ) .
[ [ rank - of - ferromagnetic - ising - models - in - general - d - dimensions ] ] rank of ferromagnetic ising models in general @xmath133 dimensions ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ we start by explicitly examining the consequences of the conformal transformation of eq .
( [ centralm * ] ) ( see fig . [ mobius.fig ] ) in the main text , that implemented dualities in the ising model . towards that end , below , in table [ table1 ] , we display the rank of the matrix @xmath300 of eq .
( [ wmd ] ) . as is seen , for the larger systems examined , in all spatial dimensions @xmath133 , the ratio @xmath458 between the rank of the matrix ( @xmath299 ) formed by the linear relations implied by the dualities and the total number of unknown ( @xmath312 ) series coefficients ( for the combined h - t and l - t expansions ) tends to @xmath459 . specifically , for any finite system size @xmath301 , the rank @xmath460 while the number of unknowns , @xmath461 the additive contribution of @xmath462 to the rank originates from the number of conditions of eqs .
( [ zero ] ) of the main text .
we note that although we tabulate here results for symmetric hypercubes , similar values appear in @xmath463 lattices for which all sides @xmath464 are even ( and consequently all possible loop lengths @xmath341 and surface areas @xmath243 are even and lead to eqs .
( [ htlt * ] ) and ( [ ltt * ] ) in the main text ) .
c|c|c|c|c|c|l & @xmath301 & @xmath312 & @xmath299 & @xmath465 & @xmath466 & + & & 8 & 8 & 1.00000 & + & & 32 & 26 & 0.81250 & + & & 72 & 56 & 0.77778 & + & & 128 & 98 & 0.76563 & + & & 200 & 152 & 0.76000 & + & & 288 & 218 & 0.75694 & + & & 392 & 296 & 0.75510 & + & & 512 & 386 & 0.75391 & + & & 648 & 488 & 0.75309 & + & & 800 & 602 & 0.75250 & + & & 1800 & 1352 & 0.75111 & 0.75000 + & & 24 & 21 & 0.87500 & + & & 192 & 147 & 0.76563 & + & & 648 & 489 & 0.75463 & + & & 1536 & 1155 & 0.75195 & + & & 64 & 52 & 0.81250 & + & & 1024 & 772 & 0.75391 & + & & 160 & 125 & 0.78125 & + & & 384 & 294 & 0.76563 & + & & 896 & 679 & 0.75781 & + & & 2048 & 1544 & 0.75391 & + as discussed in the main text , for small system size @xmath301 , the conditions of eqs .
( [ zero ] ) therein are more restrictive .
this further leads to the _ monotonic decrease _ of @xmath302 as @xmath301 is increased . for the lowest terms ,
when @xmath133 is large , these additional conditions play a more prominent role and the problem becomes more solvable . in table
[ table1 ] , we examined the uniform ( ferromagnetic ) ising model .
we now briefly turn to ising models with general ( random ) signs , @xmath220 .
in such a case , the restrictions of eq .
( [ zero ] ) of the main text no longer apply . furthermore , while @xmath233 , in the l - t expansion @xmath467 need not be unity ( as , apart from global @xmath468 ( i.e. , @xmath469 at all lattice sites @xmath470 ) , the ground state may have additional degeneracies ) .
thus , the equation to be solved is @xmath471 c|c|c|c|c|c|l & @xmath301 & @xmath312 & @xmath299 & @xmath465 & @xmath466 & + & & 9 & 7 & 0.77778 & + & & 33 & 25 & 0.75758 & + & & 73 & 55 & 0.75342 & + & & 129 & 97 & 0.75194 & + & & 201 & 151 & 0.75124 & + & & 289 & 217 & 0.75087 & + & & 393 & 295 & 0.75064 & + & & 513 & 385 & 0.75049 & + & & 649 & 487 & 0.75039 & + & & 801 & 601 & 0.75031 & + & & 1801 & 1351 & 0.75014 & 0.75000 + & & 25 & 19 & 0.76000 & + & & 193 & 145 & 0.75130 & + & & 649 & 487 & 0.75039 & + & & 1537 & 1153 & 0.75016 & + & & 65 & 49 & 0.75385 & + & & 1025 & 769 & 0.75024 & + & & 161 & 121 & 0.75155 & + & & 385 & 289 & 0.75065 & + & & 897 & 673 & 0.75028 & + & & 2049 & 1537 & 0.75012 & + here , @xmath272 is a @xmath472component vector , and @xmath90 is a @xmath473component vector defined by @xmath474 in eq .
( [ mvpd ] ) , the matrix @xmath475 where the @xmath476 matrix @xmath477 is equal to @xmath478 with matrices @xmath479 , whose elements are given by @xmath480 for @xmath481 and @xmath482 where @xmath483 .
the vector @xmath484 is given by @xmath485 the last row of the matrix @xmath484 and the vector @xmath90 capture the constraint @xmath486 as is seen in table [ table2 ] , the effect of the restrictions of eqs .
( [ zero ] ) of the main text diminishes for large system sizes ; the ratio @xmath487 for @xmath488 also in this case when eqs .
( [ zero ] ) of the main text can no longer be invoked . for any finite @xmath301 , the rank @xmath489 ( in this case eq . ( [ zero ] ) of the main text no longer holds )
while ( as stated above ) , the number of unknowns or components of @xmath272 is @xmath490 we remark that although we tabulate above the results for symmetric hypercubes , i.e. , those in which along all cartesian directions @xmath103 , @xmath491 , similar values appear for general lattices of size @xmath463 with even @xmath464 ( for all @xmath492 ) .
as expected , for finite @xmath301 , the ratio @xmath458 in the more restricted ferromagnetic case ( table [ table1 ] ) is always larger than that in the corresponding random @xmath493 system ( table [ table2 ] ) .
we now turn to finite size ( vector ) self - dualities of the @xmath41 ising model @xcite with both periodic ( @xmath494)/anti - periodic ( @xmath495 ) boundary conditions , @xmath496 where @xmath497 is the coupling dual to @xmath498 ( as determined by @xmath499 , i.e. , @xmath500 ) .
inserting eqs .
( [ htlt * ] ) and ( [ ltt * ] ) of the main text into eq .
( [ self - dual - long ] ) , allowing @xmath501 to be different from unity ( @xmath502 ) , and repeating our earlier steps we obtain ( with @xmath503 ) , @xmath504 where @xmath505 . furthermore , @xmath506 setting @xmath507 in eq .
( [ self - dual - long ] ) , we have @xmath508 , with @xmath272 a @xmath509 dimensional vector whose components are @xmath510 and where @xmath511 is @xmath512 here , @xmath513 are multiples of the identity ( @xmath514 ) matrix , @xmath287 is the null matrix , the elements of @xmath515 are given by @xmath516 ( @xmath505 ) and @xmath517 and @xmath518 are all @xmath519 matrices .
c|c|c|c|c|c|l & @xmath301 & @xmath312 & @xmath299 & @xmath465 & @xmath520 & + & & 30 & 25 & 0.83333 & + & & 102 & 88 & 0.86275 & + & & 222 & 193 & 0.86937 & + & & 390 & 340 & 0.87179 & + & & 606 & 529 & 0.87294 & + & & 870 & 760 & 0.87356 & 0.87500 + & & 1182 & 1033 & 0.87394 & + & & 1542 & 1348 & 0.87419 & + & & 1950 & 1705 & 0.87436 & + & & 2406 & 2104 & 0.87448 & + & & 3462 & 3028 & 0.87464 & + the rank @xmath299 of the matrix @xmath511 is provided in table [ table3 ] . in the @xmath41 self - dual ising model the ratio @xmath521 . in the large system ( @xmath301 )
limit , @xmath398 .
thus , by comparison to the non self - dual ising models , there is indeed a reduction by a factor of two in the number of coefficients needed before the rest are trivially determined by the linear relations @xmath508 .
it is noteworthy that , contrary to the behavior for non - self dual systems , the values of @xmath458 are _ monotonically increasing _ in @xmath301 , approaching their asymptotic value of @xmath403 from below .
we now examine the analog of eq .
( [ wmd ] ) of the main text for planar systems of size @xmath522 in which the parity of @xmath523 can be either even or odd .
systems with only even @xmath464 were examined in the main text ( and table [ table1 ] ) to find a ratio of @xmath336 . as seen below ,
when either ( or both ) @xmath524 are odd , @xmath316 ( i.e. , 1/3 of the coefficients are needed as an input to determine the full series by linear relations ) .
& @xmath312 & @xmath299 & @xmath465 & @xmath525 & + 3 & 3 & 9 & 24 & 19 & 0.79167 & + 3 & 9 & 27 & 75 & 55 & 0.73333 & + 5 & 5 & 25 & 70 & 51 & 0.72857 & 0.66667 + 5 & 9 & 45 & 128 & 91 & 0.71094 & + 9 & 9 & 81 & 234 & 163 & 0.69658 & + @xmath301 & @xmath312 & @xmath299 & @xmath465 & @xmath525 & + 2 & 3 & 6 & 18 & 13 & 0.72222 & + 2 & 9 & 18 & 54 & 37 & 0.68519 & + 4 & 9 & 36 & 108 & 73 & 0.67593 & 0.66667 + 4 & 25 & 100 & 300 & 201 & 0.67000 & + 8 & 9 & 72 & 216 & 145 & 0.67130 & + in a system with periodic boundary conditions , when at least one of the cartesian dimensions @xmath464 of the lattice is odd , closed loops of odd length may appear : the h - t series is no longer constrained to be of the form of eq .
( 2 ) of the main text with even @xmath341 .
by contrast , as the total number of lattice links is even , the difference between low energy or `` good'' ( i.e. , @xmath526 ) and high energy `` bad '' ( @xmath527 ) bonds @xmath366 is even and eq .
( [ ltt * ] ) of the main text still holds . in tables
[ table4 ] and [ table5 ] we provide the results for @xmath41 periodic lattices . as seen , when at least one of the lattice dimensions is odd , @xmath316 .
the reduction in the value of @xmath302 relative to the even size lattice with periodic boundary conditions originates from the fact that the @xmath528 expansion admits both even and odd powers of @xmath260 ; there are more undetermined h - t coefficients .
more potently , a lower bound of @xmath318 can be proven by invoking the symmetries captured by the duality relations .
the h - t expansion is symmetric ( @xmath237 ) . in the h - t expansion
both odd and even powers @xmath341 appear .
by contrast , in the l - t expansion only even powers @xmath342 arise .
the number of non - vanishing h - t series coefficients @xmath529 is @xmath530 ; the number of l - t expansion coefficients @xmath531 in the low temperature form of the partition function , @xmath532 is @xmath533 .
given @xmath534 , by invoking duality ( i.e.,by equating @xmath348 ) we may compute all @xmath529 .
thus if we compute these @xmath533 l - t coefficients ( a 1/3 of the combined h - t and l - t coefficients ) , the partition function and all remaining h - t coefficients can be fully determined .
we next turn to systems with open boundary conditions . in these systems ,
no odd power shows up in the h - t series expansion .
by contrast , odd powers may appear in the l - t expansion ( as , e.g. , when there is a single @xmath535 at the boundary on a @xmath41 lattice in which all other sites @xmath536 have @xmath537 for which there are three `` bad '' bonds ) . in table
[ table6 ] we list the matrix rank for open boundary conditions .
once again , we find that for these @xmath316 . the reduction in the value of @xmath302 by comparison to the even size lattice with periodic boundary conditions in this case has its origins in the fact that here the @xmath538 expansion admits both even and odd powers of @xmath539
; there are , once again , more undetermined coefficients .
we remark that the symmetries @xmath540 and @xmath541 of the h - t and l - t expansions that were present for periodic boundary conditions no longer appear when the system has open boundaries . in the h - t expansion ,
only even powers appear . in the l - t expansion
both odd and even orders @xmath342 are present .
thus , the number of h - t coefficients is essentially 1/2 of that of the l - t coefficients .
thus , if all of the h - t coefficients @xmath230 were known ( i.e. , a 1/3 of all of the combined coefficients ) , it is clear ( even without doing any calculations ) that the remaining l - t coefficients @xmath531 can be determined by duality by setting @xmath542 and computing @xmath534 .
thus , for any finite size system , the fraction of coefficients required to compute the rest via the duality relations of eqs .
( [ wvp ] ) is bounded from above , i.e. , @xmath543 .
asymptotically , for large @xmath301 , the fraction @xmath302 approaches @xmath421 from below ( that the approach is from below and not from above follows from this inequality ) . [ [ random - pm - j - ising - systems - on - a - periodic-2-times-2-plaquette ] ] random @xmath493 ising systems on a periodic @xmath146 plaquette ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ in the main text we examined a periodic @xmath146 ferromagnetic ( @xmath320 ) system with hamiltonian @xmath321 $ ] . if instead of the ferromagnetic hamiltonian , we have an ising model with general ( random ) @xmath332 on each of the links @xmath545 then eq . ( [ zero ] ) of the main text will no longer hold . instead of eq .
( [ wvp ] ) of the main text , the equation satisfied is given by eq .
( [ mvpd ] ) where the @xmath530 dimensional ( or , in this case , eight - dimensional ) matrix @xmath142 differs from @xmath327 by the omission of the matrix @xmath546 ( the single last column of @xmath327 in this @xmath146 example ) .
that is , rather explicitly , @xmath547 is given by as noted in the main text , the @xmath530 dimensional vector @xmath90 in the @xmath493 ising system differs from @xmath89 in the ferromagnetic case in that it does not have an additional @xmath462 ( which equals one in this @xmath41 example ) last entries being equal to zero .
it is clear that the ferromagnetic system investigated above fulfills eq .
( [ mvpd ] ) ( the ferromagnetic system satisfies one further constraint related to the last row of @xmath327 ) .
we wish to underscore that _ any _ @xmath146 ising system with _ general _ couplings @xmath332 on each of the links @xmath549 will comply with eq .
( [ mvpd ] ) . for a periodic @xmath331 ferromagnetic system , there are @xmath550 coefficients and the rank of the matrix @xmath327 is @xmath551 .
thus , the coefficients are linear functions of a subset of six coefficients . choosing these to be , e.g. , @xmath552
the remaining coefficients are given by @xmath553 and @xmath554 [ [ polytope ] ] cramer s rule and amplitudes as polytope volume ratios a @xmath146 test case illustration ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ in this section , we explicitly illustrate how the computation of the remaining series coefficients from the smaller number of requisite ones using our linear equations is , trivially , related to a volume ratio of polytopes ( high dimensional polyhedra ) . in the main text
, we remarked on the two key ingredients of this correspondence : ( 1 ) given known coefficients , we may solve for the remaining ones via our linear equations by applying cramer s rule wherein the coefficients are equal to the ratio of two determinants .
( 2 ) the determinants ( appearing in cramer s rule ) as well as those of any other matrices are equal to volumes of polytopes spanned by the vectors comprising the columns or rows of these matrices . to make this lucid
, we consider the periodic @xmath146 ferromagnetic system of the main text . as it was mentioned earlier , the first six rows and last two rows of matrix @xmath327 are linearly independent .
thus , in this example .
we may choose these rows to construct an @xmath557 matrix @xmath558 .
the corresponding vector @xmath559 that needs to be solved for satisfies the equation @xmath560 . here ,
@xmath561 and @xmath562 we may invoke cramer s rule and find all of the undetermined coefficients , @xmath564 as the denominator in eq .
( [ cramers ] ) is common to all @xmath298 , we see that @xmath298 is essentially given by the determinant @xmath565 . the matrix @xmath556 is obtained by replacing the @xmath103-th column of @xmath558 with @xmath566 .
we summarize the results in table [ table7 ] . putting all of the pieces together ,
we obtain ( as we must ) the exact partition function , @xmath567 it is hardly surprising that cramer s rule can be applied that is obvious given the linear equations .
what we wish to highlight is that each of the determinants appearing in cramer s rule ( eq .
( [ cramers ] ) ) can be interpreted as the volume of a high - dimensional parallelepiped spanned by the vectors comprising the matrix columns . in the case above , the dimension @xmath568 of each of the matrices @xmath556 and @xmath558 is equal to their rank @xmath569 .
the volume of the corresponding high dimensional tetrahedron ( or polytope ) spanned by the vectors forming @xmath556 is given by @xmath570 . in systems
in which cramer s rule may be applied , the dimensionality @xmath568 is given by the rank of @xmath299 of the system of linear equations , @xmath571 . as we emphasized in the main text ( and in section [ practical : sec ] therein in particular ) , we may derive the h - t coefficients from the l - t coefficients and conversely from the simple relation of eq .
( [ too_trivial ] ) . in the case of the ising model that formed much of the focus of the current study , eq .
( [ lh ] ) provides the means to carry out this transformation . in this subsection
, we all too explicitly , apply this method to several finite size lattices .
these are ( i ) a square lattice ( @xmath572 ) and ( ii ) a cubic lattice ( @xmath573 ) with periodic boundary conditions in all directions . in both examples ,
the l - t coefficients were reported by @xcite .
it is with the aid of these that we found the h - t coefficients .
one may , of course , similarly derive the l - t coefficients from the h - t coefficients that we list below ( or verify our results ) . to save some repetition and space , we employed the two reflection simple symmetry properties @xmath574 and @xmath575 , and thus list only half of the coefficients . below are the l - t coefficients for this system @xcite . with the above reported values
@xcite , we find from eq.([lh ] ) , the h - t coefficients to be @size7.5@mathfonts @xmath576 @xmath577 @xmath578 @xmath579 @xmath580 r. haggkvist and p.h .
lundow , j. stat . phys . * 108 * , 429 ( 2002 ) ; r. haggkvist , a. rosengren , d. andren , p. kundrotas , p.h .
lundow , and k. markstrom , phys .
e * 69 * , 046104 ( 2004 ) ; http://www.theophys.kth.se/~phl/data/index.html | we relate duality mappings to the babbage equation " @xmath0 , with @xmath1 a map linking weak- to strong - coupling theories . under fairly general conditions
@xmath1 may only be a specific conformal transformation of the fractional linear type .
this deep general result has enormous practical consequences .
for example , one can establish that weak- and strong - coupling series expansions of arbitrarily large finite size systems are trivially related , i.e. , after generating one of those series the other is automatically determined through a set of linear constraints between the series coefficients . this latter relation _ partially solve _ or , equivalently , localize the computational complexity of evaluating the series expansion to a simple fraction of those coefficients . as a bonus
, those relations also encode non - trivial equalities between different geometric constructions in general dimensions , and connect derived coefficients to polytope volumes .
we illustrate our findings by examining various models including , but not limited to , ferromagnetic and spin - glass ising , and ising gauge type theories on hypercubic lattices in @xmath2 dimensions . |
compact stars have a large number of pulsation modes that have been extensively studied since the seminal work of chandrasekhar on radial oscillations @xcite . in general , these modes are very difficult to observe in the electromagnetic spectrum ; therefore most efforts have concentrated on gravitational wave asteroseismology in order to characterise the frequency and damping times of the modes that emit gravitational radiation . in particular , various works focused on the oscillatory properties of pure hadronic stars , hybrid stars and strange quark stars trying to find signatures of the equation of state of high density neutron star matter ( see @xcite and references therein ) .
more recently , compact star oscillations have attracted the attention in the context of soft gamma ray repeaters ( sgrs ) , which are persistent x - ray emitters that sporadically emit short bursts of soft @xmath3-rays . in the quiescent state ,
sgrs have an x - ray luminosity of @xmath4 erg / s , while during the short @xmath3-bursts they release up to @xmath5 erg / s in episodes of about 0.1 s. exceptionally , some of them have emitted very energetic giant flares which commenced with brief @xmath3-ray spikes of @xmath6 s , followed by tails lasting hundreds of seconds .
hard spectra ( up to 1 mev ) were observed during the spike and the hard x - ray emission of the tail gradually faded modulated at the neutron star ( ns ) rotation period .
the analysis of x - ray data of the tails of the giant flares of sgr 0526 - 66 , sgr 1900 + 14 and sgr 1806 - 20 revealed the presence of quasi - periodic oscillations ( qpos ) with frequencies ranging from @xmath7 18 to 1840 hz @xcite .
there are also candidate qpos at higher frequencies up to @xmath8 khz in other bursts but with lower statistical significance @xcite ; in fact , according to a more recent analysis only one burst shows a marginally significant signal at a frequency of around 3706 hz @xcite .
several characteristics of sgrs are usually explained in terms of the _ magnetar _ model , assuming that the object is a neutron star with an unusually strong magnetic field ( @xmath9 g ) @xcite . in particular ,
giant flares are associated to catastrophic rearrangements of the magnetic field .
such violent phenomena are expected to excite a variety of oscillation modes in the stellar crust and core .
in fact , recent studies have accounted for magnetic coupling between the crust and the core , and associate qpos to global magneto - elastic oscillations of highly magnetized neutron stars @xcite .
there has also been interest in the possible excitation of low order @xmath10-modes because of their strong coupling to potentially detectable gravitational radiation @xcite . in the present paper we focus on radial oscillations of neutron stars permeated by ultra - strong magnetic fields .
these modes might be relevant within the magnetar model because they could be excited during the violent events associated with gamma flares .
since they have higher frequencies than the already known qpos , they can not be directly linked to them at present . however , it is relevant to know all the variety of pulsation modes of strongly magnetized neutron stars because the number of observations is still small and new features could emerge in future flares data . on the other hand ,
in the case of rotating objects we can expect some amount of gravitational radiation from even the lowest ( @xmath11 ) quasi - radial mode @xcite making them potentially relevant for gravitational wave astronomy .
in this section we present an overview of the hadronic equations of state ( eos ) used in this work .
we describe hadronic matter within the framework of the relativistic non - linear walecka ( nlw ) model @xcite . in this model
we employ a field - theoretical approach in which the baryons interact via the exchange of @xmath12 mesons in the presence of a magnetic field @xmath13 along the @xmath14axis .
the total lagrangian density reads : @xmath15 where @xmath16 , @xmath17 , @xmath18 and @xmath19 are the baryons , mesons , leptons and electromagnetic field lagrangians , respectively , and are given by @xmath20 where he @xmath21-sum runs over the baryonic octet @xmath22 , @xmath23 is the corresponding baryon dirac field , whose interactions are mediated by the @xmath24 scalar , @xmath25 isoscalar - vector and @xmath26 isovector - vector meson fields .
the baryon charge , baryon mass and isospin projection are denoted by @xmath27 , @xmath28 and @xmath29 , respectively , and the masses of the mesons are @xmath30 mev , @xmath31mev and @xmath32mev .
the strong interaction couplings of the nucleons with the meson fields are denoted by @xmath33 , @xmath34 and @xmath35 .
we consider that the couplings of the hyperons with the meson fields are fractions of those of the nucleons , defining @xmath36 , where the values of @xmath37 are chosen as @xmath38 and @xmath39 @xcite .
the term @xmath40 denotes the scalar self - interactions @xcite , with @xmath41 and @xmath42 .
the mesonic and electromagnetic field tensors are given by their usual expressions @xmath43 , @xmath44 and @xmath45 .
the @xmath46-sum runs over the two lightest leptons @xmath47 and @xmath48 is the lepton dirac field .
the symmetric nuclear matter properties at saturation density adopted in this work are given by the gm1 parametrization @xcite , with compressibility @xmath49 mev , binding energy @xmath50 mev , symmetry energy @xmath51 mev , slope @xmath52 mev , saturation density @xmath53 @xmath54 and nucleon mass @xmath55 mev .
the following equations present the scalar and vector densities for the charged and uncharged baryons @xcite , respectively : @xmath56 , \\ & \label{densities4}\rho_{b}^{v}=\frac{1}{2\pi^{2}}\sum_{s}\bigg[\frac{1}{3}(k_{f , s}^{\,b})^{3}\bigg ] , \end{aligned}\ ] ] where @xmath57 and @xmath58 .
@xmath59sgn@xmath60 are the landau levels for the fermions with electric charge @xmath27 , @xmath61 is the spin and assumes values @xmath62 for spin up and @xmath63 for spin down cases .
the energy spectra for the baryons are given by @xcite : @xmath64 where @xmath65 .
the fermi momenta @xmath66 of the charged baryons and @xmath67 of the uncharged baryons and their relationship with the fermi energies of the charged baryons @xmath68 and uncharged baryons @xmath69 can be written as : @xmath70 for the leptons , the vector density is given by : @xmath71 where @xmath72 is the lepton fermi momentum , which is related to the fermi energy @xmath73 by : @xmath74 with @xmath75 .
the summation over the landau level runs until @xmath76 ; this is the largest value of @xmath77 for which the square of fermi momenta of the particle is still positive and corresponds to the closest integer , from below to : @xmath78 , \quad\mathrm{leptons}\label{ll1}\\ & & \nu_{\mathrm{max}}=\bigg[\frac{(e_{f}^{\,b})^{2}-\bar{m}_{b}^{2}}{2|q_{b}|b}\bigg ] , \quad\mathrm{charged~baryons}.\label{ll2}\end{aligned}\ ] ] the chemical potentials of baryons and leptons are : @xmath79 from the lagrangian density ( [ lt ] ) , and mean - field approximation , the energy density is given by @xmath80 where the expressions for the energy densities of charged baryons @xmath81 and neutral baryons @xmath82 are , respectively , given by : @xmath83,\label{ea1 } \\ \varepsilon_{b}^{n } & = & \frac{1}{4\pi^{2}}\sum_{s}\bigg[\tfrac{1}{2 } k_{f,\nu , s}^{\,b}(e_{f}^{\,b})^{3 } - \tfrac{1}{4}\bar{m}_{b } \bigg(\bar{m}_{b}k_{f,\nu , s}^{\,b}e_{f}^{\,b } \nonumber \\ & & + \bar{m}_{b}^{3}\ln\bigg|\frac{e_{f}^{\,b}+k_{f,\nu , s}^{\,b}}{\bar{m}_{b}}\bigg|\bigg)\bigg]\ , .
\label{ea2}\end{aligned}\ ] ] the expression for the energy density of leptons @xmath84 reads @xmath85\ , .
\label{ea3}\ ] ] the pressures of baryons and leptons are : @xmath86 where the expression of the vector densities @xmath87 and @xmath88 are given in ( [ densities2 ] ) and ( [ vector_density_leptons ] ) , respectively .
we assume that the magnetic field @xmath13 in the eos depends on the density according to @xcite @xmath89 \right\}\,,\ ] ] where @xmath90 is the baryon density , @xmath91 is the saturation density , @xmath92 is the magnetic field on the surface of a magnetar , taken equal to @xmath0 g in agreement with observational values , and @xmath93 is the magnetic field for larger densities . the parameters @xmath94 and @xmath3 are chosen to reproduce two behaviors of the magnetic field : a fast decay with @xmath95 and @xmath96 and a slow decay with @xmath97 and @xmath98 @xcite . according to the discussion in the previous subsection , we use two values for the magnetic field @xmath93 , namely @xmath99 g and @xmath100 g. notice that in the previous subsections we assumed that the matter pressure ( @xmath101 ) is isotropic in spite of the high values of the magnetic field . as it has been shown in ref .
@xcite , the anisotropic effects around @xmath100 g are small , thus we restrict ourselves to magnetic fields below this value .
however , the purely field - related pressure @xmath102 may become dominant in the core of the star .
in such cases , the total pressure perpendicular to the magnetic field lines would be significantly larger than the pressure parallel to the field lines .
therefore , 3d or at least 2d stellar structure equations should be used in order to incorporate the effect of the pressure anisotropy .
we must stress however , that the magnetic field geometry inside a neutron star can be extremely complex , and depending on its configuration the use of the spherically symmetric stellar structure equations can still be a good enough approximation , as we argue below .
if we consider , for example , a purely dipole field ( purely poloidal field ) , a 1d stellar structure calculation would be certainly inappropriate for some very high field objects .
but such configuration is unstable and can not be realised in nature ; in fact , any neutron star with a purely poloidal or a purely toroidal magnetic field is unstable in general relativity ( see @xcite and references therein ) .
this strongly supports the idea that any long - lived magnetic field configuration in a ns has to consist of a mixture of poloidal and toroidal field components . on the other hand , while stellar shapes have long been considered to be oblate due to the effects of centrifugal and/or magnetic forces , this is not necessarily true .
for example , it has been found recently that the shape of our sun is perfectly round , against the common expectation of an oblate shape due to its rotation @xcite . in the specific case of neutron stars
, the results found thus far are that purely poloidal magnetic fields make stars oblate ( equatorial radius larger than polar radius ) , while purely toroidal magnetic fields lead stars to become prolate ( polar radius larger than equatorial radius ) .
very recently , it has been shown that equilibrium states of magnetized stars with mixed poloidal@xmath103toroidal magnetic fields are possible ; in particular , it was possible to build neutron stars with twisted - torus configurations in equilibria with any toroidal and poloidal field energy content @xcite .
such poloidal and toroidal magnetic fields act as increasing and decreasing mechanisms for stellar oblateness , respectively @xcite .
an additional finding of ref .
@xcite is that for a fixed polar magnetic field strength , a higher relative content of toroidal field energy ( @xmath104 ) implies in general a much higher total ( poloidal and toroidal ) magnetic energy inside the star .
this means that a highly magnetized neutron star can harbor internal magnetic fields that are significantly stronger than commonly expected .
in addition , a very complex field is expected from the formation process of magnetized neutron stars , since they are born hot , highly convective and differentially rotating @xcite . during its early evolution ,
neutron star s magnetic fields can be significantly amplified and redistributed by several mechanisms including possibly dynamo action and shear instabilities . in this context
, a distribution of magnetic energy with poloidal and toroidal fields close to equipartition appears as a very reasonable candidate for the internal magnetic field configuration .
in such a case , since both the toroidal and poloidal components are of the same order , we may expect that oblateness and prolateness cancel out approximately , leading to stars close to the spherical symmetry .
since a study of stellar pulsations in 2d or 3d is numerically involving , we shall adopt spherical symmetry as a first approximation , and use the much more simple 1d equations . within this approach
, we must add to the matter pressure ( @xmath101 ) given in sec .
[ sec : a ] , an effective mean magnetic pressure @xmath105 representing the effective isotropic contribution arising from the combined effect of poloidal and toroidal field components of the same order .
thus , the total energy density and the total pressure of the system read : @xmath106^{2 } , \\ p & = & p_{m}+p_{l } + p_b .\label{et222}\end{aligned}\ ] ] for finding the effective mean magnetic pressure @xmath105 we shall follow here an averaging procedure similar to the one employed in works that focus on the study of magnetic fluctuations in turbulent fluids @xcite .
the magnetic stress tensor @xmath107 is given by @xmath108 where @xmath109 is the kronecker tensor and the brackets @xmath110 denote the averaging . for a completely isotropic distribution of the magnetic field we have @xmath111 , and
the magnetic stress tensor reads @xmath112 the effective mean magnetic pressure @xmath105 is related to the magnetic stress tensor by @xmath113 .
therefore , we find : @xmath114^{2 } .
\label{wa1}\end{aligned}\ ] ] a similar expression has been used in @xcite for studying the influence of asymmetry on a magnetized proto - neutron star .
in order to study the radial oscillations of a compact star , we must know first its equilibrium configuration .
such configuration is perturbed in such a way that the spherical symmetry of the body is not violated . the space - time and fluid perturbations
are inserted into the einstein equations and into the energy , momentum and baryon number conservation equations , assuming a sinusoidal time dependence @xmath115 and retaining only the first - order terms .
the result of this procedure is a second order ordinary differential equation for the perturbations @xcite or alternatively a set of two first order ordinary differential equations , as shown below . in the following
we present the explicit form of the equilibrium and oscillation equations employed in the present work .
we consider that the unperturbed compact star is totally composed of a perfect fluid . in this case
the stress - energy momentum tensor can be expressed as @xmath116 where @xmath117 and @xmath118 are the energy density and pressure respectively .
the background spacetime of the spherically symmetric star , can be expressed through the line element @xmath119 where @xmath120 are the set of schwarzschild - like coordinates , and the metric potentials @xmath121 and @xmath122 are functions of the radial coordinate @xmath123 only .
the einstein equations in such a spacetime lead to the following set of stellar structure equations ( tolman - oppenheimer - volkoff equations ) @xmath124 where @xmath125 is the gravitational mass inside the radius @xmath123 .
the metric function @xmath77 has the following boundary condition @xmath126 where @xmath127 is the radius of the star and @xmath128 its mass . with this condition
the metric function @xmath77 will match smoothly to the schwarzschild metric outside the star .
the boundary conditions for @xmath125 and @xmath118 are @xmath129 and @xmath130 . , ( b ) the gravitational redshift @xmath131 , and ( c ) the central energy density @xmath132 .
in all the figures we set three different values for the magnetic field : b@xmath133 g and @xmath134 g . we also show the effect of slow and fast decay in the magnetic field profile .
, title="fig : " ] , ( b ) the gravitational redshift @xmath131 , and ( c ) the central energy density @xmath132 .
in all the figures we set three different values for the magnetic field : b@xmath133 g and @xmath134 g . we also show the effect of slow and fast decay in the magnetic field profile .
, title="fig : " ] , ( b ) the gravitational redshift @xmath131 , and ( c ) the central energy density @xmath132 . in all the figures we set three different values for the magnetic field : b@xmath133 g and @xmath134 g . we also show the effect of slow and fast decay in the magnetic field profile . ,
title="fig : " ] ) as a function of the neutron star mass .
the values for the magnetic field as well as the slow and fast decays in the magnetic field profile are the same as given in fig .
[ fig1 ] . ]
several forms of the oscillation equation have been presented in the literature ( for more details the reader is referred to @xcite ) . in this work ,
we use the set of equations of gondek et al .
@xcite and adopt @xmath135 .
the equations read @xmath136 where @xmath137 is the eigenfrequency and the quantities @xmath138 and @xmath139 are assumed to have a harmonic time dependence @xmath140 . to solve equations ( [ ecuacionparaxi ] ) and ( [ ecuacionparap ] )
one needs two boundary conditions .
the condition of regularity at the centre ( @xmath141 ) can be written as @xcite @xmath142 where the eigenfunctions are normalized in order to have @xmath143 .
the second boundary condition , expresses the fact that the lagrangian perturbation in the pressure at the stellar surface is zero , thus : @xmath144 to solve numerically the oscillation equations we employ a shooting method in order to fulfil the required boundary conditions . for more details on the method see ref . @xcite .
in this section we analyse the effect that a strong magnetic field could produce on the fundamental mode of the radial oscillations of hadronic stars . as mentioned before , we consider that the magnetic field decays with the density following the fast and slow profiles presented in section [ b - density - depndence ] .
all the models for hadronic stars investigated in the present work have a maximum mass in agreement with the recent observation of the pulsars psr j1614 - 2230 with @xmath145 @xcite and psr j0348 - 0432 with @xmath146 @xcite . in fig .
[ fig1 ] we see that a magnetic field profile with @xmath147 g , produces very small changes on the oscillation period of the fundamental mode with respect to an unmagnetized star , for both slow and fast decays .
this can be explained by the small effect that such magnetic field intensity has on the equation of state .
in contrast , when @xmath148 g is selected , there is a clear change in the oscillation period . as a function of the stellar mass ,
the curves fall below and to the right of the curves for weaker fields . notice that for large mass objects the period changes because of the shift of the curves due to the increase of the maximum stellar mass . for smaller masses ,
the curves for @xmath148 g are also significantly different with respect to the unmagnetized case ; e.g. for a neutron star with @xmath149 the period is around @xmath150 smaller , and the difference increases for less massive stars .
for completeness we present also the oscillation period as a function of the gravitational redshift ( see middle panel of fig . [ fig1 ] ) and as a function of the central mass - energy density ( lower panel of fig .
[ fig1 ] ) .
the effect of strong magnetic fields is more apparent in the frequency of the fundamental mode as can be seen in fig .
the oscillation frequency for @xmath147 g is slightly above the one of an unmagnetized object of the same mass , and there is almost no difference between the fast and the slow decaying profiles of the magnetic field .
however , if @xmath148 g the oscillation frequencies are clearly larger than for an unmagnetized star of the same mass .
for example , for a star with @xmath149 the frequency is around @xmath150 larger and for a @xmath151 star it is around @xmath152 larger .
the difference between the fast and the slow decaying profiles of @xmath13 is very small . as stated before
, purely radial modes do not emit gravitational waves and consequently they are essentially damped by the bulk viscosity , originated from the re - establishment of chemical equilibrium when a fluid element of the star is compressed and rarified during pulsations .
unfortunately , there is great uncertainty about the amount of viscosity inside neutron stars since it depends sensitively on the composition of matter which is uncertain beyond few times the nuclear saturation density @xcite .
if the damping time due to viscous forces is long enough , radial pulsations in magnetars can leave an imprint in the microstucture of magnetar flare lightcurves opening a new window for the study of highly magnetized ultradense matter .
l. b. castro thanks cnpq , brazil , grants no 455719/2014 - 4 ( universal ) and no 304105/2014 - 7 ( pq ) for partial support . c. vsquez flores acknowledges the financial support of capes , brazil .
g. lugones acknowledges fapesp and cnpq for financial support .
we acknowledge an unknown referee for valuable comments . | we investigate the effect of strong magnetic fields on the adiabatic radial oscillations of hadronic stars .
we describe magnetized hadronic matter within the framework of the relativistic nonlinear walecka model and integrate the equations of relativistic radial oscillations to determine the fundamental pulsation mode .
we consider that the magnetic field increases , in a density dependent way , from the surface , where it has a typical magnetar value of @xmath0 g , to the interior of the star where it can be as large as @xmath1 g. we show that magnetic fields of the order of @xmath2 g at the stellar core produce a significant change in the frequency of neutron star pulsations with respect to unmagnetized objects .
if radial pulsations are excited in magnetar flares , they can leave an imprint in the flare lightcurves and open a new window for the study of highly magnetized ultradense matter . |
the most widely considered theory of 2d melting is the so - called kthny theory of kosterlitz and thouless @xcite , halperin and nelson @xcite , and young @xcite , which predicts that melting in two dimensions occurs via two continuous phase transitions , first from solid to hexatic fluid , and then from hexatic fluid to isotropic fluid .
this theory begins from the fact that true translational order can not exist at any non - zero temperature in 2d because of the infrared divergence caused by the zero point motion of long - wave - length density fluctuations . according to the kthny theory , another form of true long - range order exists below some non - zero temperature @xmath6 where only the directions of the nearest - neighbor bonds order
this long - range bond order disappears above @xmath6 because of dislocation unbinding which leads to an intermediate phase , the hexatic phase , characterized by topological order , where while dislocations are unbound , disclinations with opposite topological charge remain bound .
these disclinations become unbound at a higher temperature @xmath7 where the system becomes an isotropic disordered fluid .
simulation of melting in classical two - dimensional ( 2d ) systems has been tackled by means of a variety of computational studies@xcite for several decades without reaching a definite conclusion regarding its nature . in particular for hard disks in 2d , a large number of computer simulation studies have been applied to understand 2d melting , since this is the toy model on which the metropolis monte carlo method itself was first introduced@xcite and soon afterward , the 2d melting of hard disks was studied@xcite .
one of the reasons for the difficulty to reach an unequivocal conclusion is that in 2d a conventional solid with true translational order can not exist , and , instead the correlations decay very slowly over long distance .
this requires large size systems where the relaxation time scales become very long for these types of phenomena .
in particular for hard disk systems , when using a local updating algorithm or even molecular dynamics , particles remain stuck in their local `` cage '' for large computational time scales , precisely because of the hard disk constraint .
one might think that monte carlo simulation of soft - core potentials , such as the lennard - jones system in 2d , might not be plagued by the same level of computational severity as the hard - disk systems , because of the softening of the hard - core constraint . as a matter of fact
there are a number of studies of the lennard - jones solid@xcite by computer simulation where also a general consensus about the nature of melting has not been established .
some studies have favored a first - order transition from solid to liquid @xcite , as predicted by the grain boundary melting suggestion@xcite , while other studies @xcite have leaned toward the kthny theory .
the most thorough of these studies , however , are at least one decade old and because of the fact that the computational resource constraints of today are significantly better , a more thorough study should be possible . in the present paper ,
we study the nature of melting of a two dimensional ( 2d ) lennard - jones solid using large scale monte carlo simulation .
we use systems of up to 102,400 particles to capture the decay of the correlation functions associated with translational as well as the bond - orientational order .
we find that to carry out thorough investigations beyond these sizes , calculations using the metropolis local update become impractical using today s high performance computing because of the long relaxation time scales .
further technical details of our simulation are described in the next section , and the remainder of the paper is organized as follows . in sec . [ defects ]
we discuss the role of defects in the kthny theory of melting and present the results of a geometric defect analysis . in sec .
[ orders ] we show the temperature dependence of both order parameters , @xmath10 and @xmath11 , as well as their second moments , @xmath12 and @xmath13 .
the system - size dependence of @xmath12 and @xmath13 can be used to determine the critical exponents @xmath8 and @xmath14 , as shown in sec .
[ exponents ] . in the same section ,
the kthny values of the critical exponents at melting , @xmath15 and @xmath16 , are used to estimate the transition temperatures @xmath17 and @xmath18 .
next , in sec .
[ correlations ] , we present our results on the correlation function associated with bond orientational order above @xmath7 and determine the temperature dependent correlation length @xmath5 . in the same section ,
we demonstrate finite - size scaling of @xmath13 . a similar presentation is given in sec .
[ distributions ] for the pair distribution function and the correlation length of translational order , @xmath4 .
in addition , we present our findings for the scaling behavior of the second moment of @xmath12 in this same section .
[ binder ] presents an analysis of the melting transition using binder s cumulant ratio @xcite for each order parameter , and also includes a discussion of finite - size scaling theory in the presence of multiple correlation lengths . finally , in sec .
[ conclusions ] , we give a brief summary of our main findings and conclusions .
in the lennard - jones potential , for two particles separated by a distance @xmath19 , @xmath20 an attractive inverse sixth power tail is combined with a repulsive inverse twelfth power hard core , such that there are only two parameters : @xmath21 , the potential well depth , and @xmath22 , the hard - core diameter . however
, our results can be inferred for any particular value of these parameters ( for a specific real system ) , since in our calculations , distance is measured in units of @xmath22 and temperature in units of @xmath23 .
in our calculations we have truncated the lennard - jones potential at a distance of 3@xmath22 , and shifted the value of the potential within this cutoff distance by a constant so that the resulting potential approaches zero at 3@xmath22 ( matching the values beyond 3@xmath22 ) .
this truncation is justified because the lennard - jones potential is already quite small ( -0.005@xmath21 ) at this distance , and is not expected to significantly affect the accuracy of our simulations . additionally , by using a cutoff distance , we are able to use a cell list structure in our algorithms so that our computations scale as @xmath24 instead of the @xmath25 scaling without a cutoff distance ( @xmath26 is the number of particles in our simulation cell ) .
we have collected data for systems of 1600 , 6400 , and 25600 particles over a wide temperature range at a density of 0.873 . additionally , we have simulated a system of 102400 particles for three temperature values at the same density in order to verify our results for the smaller system sizes . to accommodate the expected low temperature triangular solid phase , a periodic simulation cell of proportion 2:@xmath27 is used .
we have performed our calculations on the florida state university shared high - performance computing facility , which contains several thousand compute nodes .
the processors on these nodes range in speed from 2.3 ghz to 2.8 ghz , and it takes about 34 hours to perform 1,000,000 monte carlo sweeps for @xmath28 particles , including calculating observables every 100 monte carlo sweeps ( mcs ) .
we have found that , except for the @xmath29 particle system , a million mcs are sufficient to reach equilibrium , even near the critical points .
the data presented here is obtained over one million mcs , after a period of one ( for @xmath29 ) or two ( for @xmath30 , and 25600 ) million mcs of equilibration . to take advantage of our computational resources , we utilized a trivially parallel monte carlo implementation of 100 threads , each with a unique random number seed and initial configuration
. simulations begin from an initial near - ordered configuration ( particles are placed in a triangular lattice , with 5% lattice spacing random fluctuations ) .
statistics for thermodynamic variables are collected by generating averages on each of the 100 parallel threads , then using the central limit theorem we obtain the total average , as we have 100 independent means .
although in our preliminary studies we have computed thermodynamic quantities for a range of densities and temperatures , the effects of critical slowing down near the melting transition and our desire to study the largest possible systems have led us to focus on a single density , 0.873 ( all densities are in units of particles per @xmath31 ) .
this density was chosen for several reasons .
this is a density that could be readily compared to prior numerical simulations of lennard - jones melting @xcite . also , we wanted a density that is relatively low , but large enough to avoid the solid - vapor coexistence phase at low temperatures . strictly speaking ,
there is a solid phase in the zero temperature limit only at densities of 0.9165 ( the density at which the spacing of the triangular lattice is the same as the position of the lennard - jones potential minimum ) and above . below this density
there is a solid - vapor coexistence phase .
however , the triple point density is roughly 0.82 , so at higher densities the system will in general become solid before the onset of melting occurs @xcite .
in two dimensions , the densest packing of particles of uniform size is achieved in a triangular lattice . in such a configuration , each particle has exactly six nearest neighbors .
thermal fluctuations will lead to distortions in the lattice , or even destroy it completely . to quantify this
, we use the delaunay triangulation to determine the nearest neighbor network of our particle configurations .
the nearest neighbor network tells us the number of nearest neighbors , or coordination number , of each particle . for a system of particles in a periodic plane , the average coordination number is always six @xcite .
particles in a triangularly ordered region will be six - coordinated , while disruptions in the lattice will lead to particles with coordination numbers greater than or less than six .
a defect is defined as any coordination number other than six .
these non - six - coordinated atoms may be thought of as disclinations of charge @xmath32 , their coordination number being @xmath33 .
the most common type of disruption , or defect , is a five- or seven - coordinated particle .
these may be interpreted as disclinations of charge plus or minus one .
two oppositely charged disclinations may be thought of as a dislocation .
more complex arrangements of disclinations are possible , such as dislocation pairs and grain boundary loops , but in our analysis we have only considered individual defects .
the defect fraction , @xmath34 , is defined as the fraction of particles that do not have six neighbors , where @xmath26 is the number of particles in the system , and @xmath35 is the number of six - coordinated particles in the system . remembering that dislocations are made of two bound disclinations of opposite charge , and that dislocations become unbound above the melting point
, we can expect the defect fraction to experience a jump at the melting point @xcite .
additionally , at low temperatures we can expect an energy gap to occur , which is the energy cost to create a dislocation pair . because the overall disclinicity of the system must be zero , as well as the net burgers vector of any dislocations , the lowest - energy defect excitation is a dislocation pair of opposite burgers vectors . in practice
this is usually two pairs of 5- and 7-coordinated particles .
this leads to an exponential behavior in the defect fraction , @xmath36 , where @xmath37 is the lowest - energy for a defect type excitation of the system . 0.3 in particles at t=0.70 .
defects are shown in red . ] 0.3 in particles at t=0.90 .
defects are shown in red . ] 0.3 in particles at t=1.10 .
defects are shown in red . ] 0.3 in in figures [ data_net070],[data_net090],[data_net110 ] the delaunay triangulated configuration of a 1600 particle system is shown at temperatures 0.7 , 0.9 and 1.1 respectively .
the defects are shown in red . at low temperature
as demonstrated in figure [ data_net070 ] , we see that defects occur in quadruplets consisting of two 5-coordinated and two 7-coordinated particles . as the temperature
is raised to 0.9 ( figure [ data_net090 ] ) we can see isolated dislocations ( one 5-fold coordinated atom bound to a 7-fold coordinated atom ) . at yet higher temperature , such as 1.1 ( figure [ data_net110 ] )
we can observe isolated disclinations .
this can also be seen in the pair distribution functions @xmath38 , @xmath39 , and @xmath40 , for pairs of 7-coordinated particles , pairs of 5-fold coordinated atoms and for 5-fold-7-fold coordinated atoms respectively . in figure
[ data_g77 ] , a sharp peak in @xmath38 is observed at low temperatures ( @xmath41 ) , indicating that dislocations are tightly bound . at higher temperatures ( @xmath42 and @xmath43 ) , the peak in @xmath38
is greatly diminished , and dislocations become first weakly bound ( @xmath42 ) and then completely unbound ( @xmath43 ) .
@xmath39 , while not shown , behaves qualitatively similar to @xmath38 , as both are representative of the pair distribution of dislocations . 0.3 in .
the peak for t=0.70 extends to @xmath44.,title="fig : " ] 0.3 in 0.3 in .
the peak for t=0.70 extends to @xmath45.,title="fig : " ] 0.3 in the pair distribution function for disclinations , @xmath40 , is shown in figure [ data_g57 ] . while the sharp peak at low ( @xmath41 ) and intermediate ( @xmath42 ) temperature is expected , the peak at @xmath43 , while quite lower , is still very substantial .
this indicates that disclinations have not become completely unbound , and indeed it is difficult to find isolated disclinations in the snapshot configurations presented in figure [ data_net110 ] .
when isolated disclinations do occur , they are still next - nearest neighbors with at least one other disclination of opposite charge . 0.3 in , as defined by the fraction of non - six - coordinated particles in the delaunay triangulation , @xmath46 .
the rapid rise in @xmath47 from near zero to almost 25% is a possible sign that dislocation and/or disclination unbinding is occurring.,title="fig : " ] 0.3 in 0.3 in appears to lessen in magnitude as the system size is increased.,title="fig : " ] 0.3 in 0.3 in , while at higher temperatures , such as t=2.00 shown above , there is a loss of order over moderate length scales.,title="fig : " ] 0.3 in according to the kthny theory , disclinations remain very tightly bound below @xmath6 . above @xmath6 ,
the disclinations are screened from one another by the presence of free dislocations yet remain bound , albeit by a weaker logarithmic binding@xcite .
thus , we expect a proliferation of defects to occur around @xmath6 , and to continue growing until somewhere above @xmath7 , where a saturation should occur . in figure [ data_defects ]
we show the average defect fraction as a function of temperature . at low temperature , there are very few defects , while at high temperature there is a considerable fraction of the system that is defected . in between
, there is a region of rapidly increasing defect fraction , from @xmath48 to @xmath49 .
this can be quantitatively verified by calculating the temperature derivative of the defect fraction , which is indeed found to have a broad peak in this temperature region .
the overall shape of @xmath50 is very similar to that of the specific heat capacity , to be shown next .
additionally , we can see some size dependence in the region @xmath51 , although this seems to be an issue mostly for comparisons of the smallest system size ( @xmath52 ) to the larger system sizes .
the specific heat per particle at constant volume , @xmath53 , can be calculated from the energy fluctuations , @xmath54 where @xmath55 is the total energy of an @xmath26 particle system .
we have calculated the specific heat and show it as a function of temperature in figure [ data_specific_heat ] .
one can see a broad peak in the specific heat per particle . according to the kthny theory
, there should be an essential singularity in the specific heat at both @xmath6 and @xmath7@xcite . however , it is not clear whether this will be visible above background contributions to the specific heat .
either way , the peak in specific heat points to a rearrangement of order in the systems studied .
also , if we look at the distribution function ( figure [ data_distrib ] ) , we see ordering at low temperatures , and fluid behavior at high temperatures .
overall , it is clear that there is a phase transition occurring , with a disordered fluid state at high temperatures and an ordered state at low temperatures . in the kthny theory ,
dislocations are bound at low temperatures , and there is a defect core energy associated with their creation .
this leads to an energy gap , and thus using the arrhenius law , we expect @xmath56 , where we have used @xmath57 because dislocation pairs are the lowest energy excitation ( isolated dislocations are forbidden ) . in table [ table1 ] we show the defect activation energy as calculated by the arrhenius law at low temperatures . taking the low temperature limit , we find @xmath58 . .defect activation energy for various system sizes and temperatures , as computed using the arrhenius law .
the numbers in parentheses are the uncertainty of the trailing digits . [
cols="<,<,<,<",options="header " , ]
let us define a global order parameter of translational order , @xmath59 where @xmath60 is a reciprocal lattice vector , and @xmath61 is the position vector of particle @xmath62 .
if there is translational ordering in a system , then clearly @xmath63 will be non - zero if @xmath60 is a reciprocal lattice vector of the appropriate lattice geometry .
due to the shape of our simulational cell , at low temperatures this will be a triangular lattice with nearest - neighbors in the x - direction . at high temperatures ,
no translational ordering is present , and all possible values of @xmath60 should give the same ( qualitative ) result .
however , at intermediate temperatures , it may be possible for there to be some degree of translational ordering that is not strictly commensurate with our simulation cell .
indeed , we have observed `` canted '' solid phases at intermediate temperatures , where we find partial triangular order with nearest neighbors in a direction titled from the x - axis by a small angle . in this case , if @xmath60 for the triangular order commensurate with our simulation cell is used , @xmath63 will be found to be zero .
however , if we use an appropriate @xmath60 for the order present , @xmath63 will be found to be non - zero .
for this reason , we define the true translational order to be the maximum value of @xmath63 for all @xmath60 . in practice
, it is not possible to perform this optimization for each monte carlo configuration , so we make the following assumptions .
first , due to the nature of ordering in two dimensions , we assume any lattice will be triangular . second , because the density of particles is fixed , we assume the lattice spacing in said triangular solid to be the same as that for the commensurate cell .
thus , we keep the magnitude of @xmath60 constant , and simply determine the direction of solid ordering for each configuration by looking at the average bond direction between nearest neighbor particles .
this turns out to be a good estimate of the true translational order for a system , but it must be remembered that it is strictly speaking a _
lower bound_. the local order parameter which measures the degree of six - fold orientational ordering is defined as @xmath64 where @xmath65 is the angle of the bond between particles @xmath66 and @xmath62 and the sum over @xmath62 extends over all @xmath67 nearest neighboring atoms found by the delaunay triangulation .
the global order parameter associated with bond - orientational order is obtained as an average over all particles .
@xmath68 in a perfectly bond - ordered triangular solid , we have that @xmath69 and @xmath70 for all @xmath71 .
in such case @xmath72 . in the low temperature phase , there is bond - orientational order , so @xmath73 should be a point on the perimeter of a circle with a radius approaching unity as @xmath74 . in the hexatic phase there is quasi - long - range bond - orientational order , which implies that the distribution of @xmath73 should become a ring in the imaginary plane . in the isotropic phase
both @xmath73 and @xmath75 should be distributed around zero value . 0.3 in 0.4 in 0.3 in in the top panel of figure [ data_psi2 ] we show the second moment of the translational order parameter , @xmath76 .
there appears to be a transition from a translationally ordered phase at low temperatures to a disordered phase at higher temperatures . in the ordered phase there is a clear relation between @xmath77 and system size .
we will explore this relation in a later section , but for now let us point out that this finite - size scaling relation begins to break down above @xmath78 .
this is expected within the kthny theory of melting due to the unbinding of dislocations . however , on closer inspection , the behavior of the curves for @xmath79 and @xmath28 in the region @xmath80 is not a smooth connection of the data at higher and lower temperature .
this is due to our measured quantity @xmath10 being a lower bound of translational order .
also shown in figure [ data_psi2 ] is the second moment of the bond orientational order parameter , @xmath81 ( bottom panel ) . at low temperatures
there is substantial bond orientational order .
below @xmath82 there is very little dependence of @xmath81 on system size . as the temperature
is increased , @xmath81 begins to show a marked dependence on system size as well as a steep decline in value as we approach the high temperature disordered phase .
the main prediction of halperin and nelson@xcite and young@xcite is that if two dimensional melting is the result of dislocation unbinding , as proposed by kosterlitz and thouless@xcite , then a second unbinding transition ( of disclinations ) is required to reach an isotropic fluid state .
this implies the presence of a novel hexatic fluid phase . in figure [ data_psimaginary ]
we show an intensity plot of the distribution of @xmath83 and of @xmath1 on the complex plane for three different temperatures . at @xmath41 ( top row )
, our calculation of the distribution of the order parameters finds a ring of values for @xmath83 , while @xmath1 is localized in a small region away from the origin ( a very narrow peak showing as a `` star '' along the positive real axis ) .
this is consistent with the presence of long - range bond orientational order ( @xmath84 ) , while the ring of @xmath83 values is expected for quasi - long - range translational order . at @xmath42 ( middle row )
, we that @xmath85 is clustered about the origin , indicating a lack of translational order .
interestingly , @xmath1 now shows a ring of values about the origin , indicating quasi - long - range order .
this is exactly what is expected of the hexatic fluid phase .
finally , at @xmath43 ( bottom row ) we see that both order parameters are distributed about the origin , indicating an isotropic fluid phase of no order . 0.3 in ( left column ) and @xmath11 ( right column ) on the complex plane for t=0.70 ( top row ) , t=0.90 ( middle row ) , and t=1.10 ( bottom row ) .
] 0.3 in
in the topological solid phase , the scaling form for the second moment of the translational order parameter is @xmath86 , where @xmath87 is the ( linear ) system size and @xmath8 is a critical exponent . in the hexatic fluid phase , a similar relation holds for bond orientational order , @xmath88 . according to the kthny theory of melting , the critical exponents @xmath8 and @xmath9 will have specific values at melting .
the translational critical exponent is bounded at lower melting temperature : @xmath89 .
additionally , the bond orientational critical exponent grows from zero at @xmath6 to 1/4 at @xmath7 , and is related to the translational correlation length : @xmath90 @xcite . by plotting @xmath91 ( or @xmath92 ) versus @xmath87 on a log - log plot , we can find @xmath8 ( or @xmath9 ) . to demonstrate the validity of this scaling law and that our results are not limited by system size , in figure [ data_logeta6 ] we plot @xmath93 versus @xmath94 for all system sizes at the two temperatures where we have results for the @xmath95 system .
results of linear least squares fits to the three smallest system sizes ( used to generate the data for figure [ data_etas ] ) are shown as a dotted blue line ( for data at t=0.80 ) , a dashed red line ( for data at t=0.84 ) , and a long - dashed green line ( for data at t=0.92 ) .
for the higher temperature , the @xmath95 data falls directly on this line , within error bars . at t=0.84
, however , the @xmath95 data indicates that a smaller value for @xmath14 at this temperature may be necessary .
this could either be due to the ( presumably ) large translational correlation lengths at this temperature , which would invalidate results for small system sizes , or perhaps a very long relaxation time . either way , from our data it is clear that by t=0.92 the kthny value of @xmath9 at @xmath7 has been well passed . 0.3 in with system size @xmath87 , shown on a logarithmic plot
results for t=0.84 are shown as blue circles , while data collected at t=0.92 is represented by red squares . in both cases ,
the data for the three smaller systems was fit to the equation @xmath96 , and the result is plotted as the dotted and dashed lines ( the solid line is the kthny value of @xmath9 at @xmath7 ) . in both cases ,
the value of @xmath97 of the largest system size ( @xmath95 ) is reasonably close to the value expected from scaling . ] 0.3 in 0.3 in at @xmath6 , according to kthny theory ; in the bottom figure , the dashed line represents the predicted value of @xmath9 at @xmath7 . ] 0.3 in in figure [ data_etas ] we show the extracted values of @xmath8 and @xmath9 , the critical exponents of translational and bond orientational order . in both panels ,
we show our results as red circles . in the top panel
, we can see that @xmath8 crosses the kthny melting value in the temperature range @xmath98 .
in the bottom panel we show the critical exponent of bond orientational order , @xmath9 .
this exponent crosses the kthny melting value ( see dashed line ) at a temperature near 0.89 , in close agreement with the value for @xmath7 derived from the divergence of the correlation length @xmath99 obtained in the next section .
this value for @xmath7 is also in good agreement with the value reported in ref . .
however , in figure [ data_etas ] we also show the algebraic exponents reported by udink and van der elsken@xcite . in both panels , we can see that their values cross the kthny melting zone at higher temperatures than our values .
we believe this disagreement may be due to insufficient thermalization time in their study , as this could lead to artificially low values of the critical exponents .
we should sound a note of caution here in regards to the scaling of @xmath91 . because our measurements for @xmath12 are lower bounds
, it is possible that the extracted exponents @xmath2 are not correct in the temperature regime where @xmath60 is no longer commensurate with the simulation cell , as is the case for @xmath100 .
the correlation function for bond orientational order is given by @xmath101 where @xmath102 is the local bond - orientational order parameter defined in sec .
[ orders ] . in the isotropic fluid phase ,
the asymptotic form of @xmath103 is @xmath104 @xcite . at shorter distances , however , a power law decay comes into play , such that as @xmath99 diverges as @xmath7 is approached from above , then the asymptotic form becomes @xmath105 at @xmath7 and below , with @xmath106 .
additionally , we observe oscillations in @xmath103 that seem to decay with an exponential envelope .
thus , we used the following fitting form for the bond orientational correlation function for distances @xmath19 much less than the system size @xmath87 : @xmath107 we have used a particle - centric definition of the bond orientational correlation function , so in our calculations of @xmath103 there will be an influence from @xmath108 , the pair distribution function . in the limit of perfect bond orientational ordering ( @xmath109 everywhere ) , @xmath103 and @xmath108 will be equivalent .
we approximate the oscillatory portion of @xmath103 which is due to the translational atomic arrangement using a damped oscillator .
the periodic form is captured by using @xmath110 , where k is expected to be near the first reciprocal lattice vector in magnitude ( @xmath111 ) and @xmath112 is just a phase - shift parameter .
the size of the oscillations are expected to decay exponentially in the fluid phase , and algebraically in the hexatic phase , so we add also a power law , ending up with a term @xmath113 .
an example fit is shown in figure [ data_g6_fit097 ] .
note that the fitting procedure returns parameters much more precise than the error bars in figure [ data_g6_fit097 ] would indicate are possible .
this is due to the high degree of correlation between neighboring points of @xmath103 .
in fact , up to a separation of @xmath114 the values of @xmath103 are still 99% correlated !
this simply means that the relative form ( including the rate of decay ) of @xmath103 is consistent between our various calculations , remembering that we average the values of 100 independent parallel monte carlo simulations .
we wish to note that we observe an upturn in @xmath103 as @xmath19 approaches @xmath115 . at temperatures closer to melting ( larger correlation lengths ) ,
the upturn occurs further from @xmath115 .
next , we would like to determine a characteristic distance @xmath116 for a given finite - system of linear dimension @xmath87 so as to stay away from this upturn due to finite - size effects .
namely , we wish to limit the range of @xmath19 in our fit of the correlation function to the form given by eq .
[ data_eq_g6_fit ] in the range @xmath117 .
let us assume a periodic form for the correlation function : @xmath118 neglecting the power - law term , the upturn is expected to occur when the @xmath119 terms are a significant fraction of the @xmath19 terms .
thus , @xmath120 where @xmath116 is the distance at which the @xmath119 terms are a fraction @xmath121 of the @xmath19 terms .
using @xmath122 or 5% , this leads to @xmath123 0.3 in , to the form shown in equation [ data_eq_g6_fit ] .
the data is for @xmath28 particles at @xmath124 .
the critical exponents are fixed at their maximum values , @xmath125 and @xmath126 .
the extracted correlation lengths are @xmath127 and @xmath128.,title="fig : " ] 0.3 in 0.3 in 0.3 in in figure [ data_xi6 ] we show @xmath5 as determined by fitting @xmath103 in the range @xmath129 .
these values were fit to the kthny form of the expected divergence of @xmath99 as @xmath7 is approached from above : @xmath130 , where @xmath131 and @xmath132 .
this fit gives a value for @xmath7 near 0.89 .
0.4 in versus @xmath133 , i.e. , the logarithm of the ratio of the finite - system - size to the correlation length , for various size systems.,title="fig : " ] 0.4 in using the calculated correlation length and critical exponent @xmath9 , in figure [ fss_bond ] we plot the dimensionless quantity @xmath134 as a function of the dimensionless ratio @xmath133 for all size - lattice considered here .
notice that the data collapse onto the same scaling function using the same values of the parameters for our fit to @xmath5 shown in figure [ data_xi6 ] , and also setting @xmath135 .
this provides additional support for the theory .
0.4 in , to the form shown in equation [ data_eq_g_fit ] .
the data is for @xmath28 particles at @xmath124 .
the critical exponent @xmath8 is fixed at its maximum value , 1/3 .
the extracted correlation length is @xmath136.,title="fig : " ] 0.4 in in the disordered phase ( @xmath137 ) the distribution function can be obtained as an angular average of the bond - orientational correlation function @xmath138 as @xmath139 the integration of @xmath140 will give us a zeroth - order bessel function of the first kind , @xmath141 , and using the kthny form of the translational correlation function , @xmath142 , we wind up with the following form for the radial pair distribution function ( in the high temperature limit ) : @xmath143 where @xmath144 is some amplitude .
the @xmath145 that we use here is the same as in the definition of the translational order parameter , namely we use the first reciprocal lattice vector of the idealized triangular lattice that is commensurate with our simulation cell . for the density considered ( @xmath146 ) , this means @xmath147 .
thus @xmath148 is quite large for moderate values of @xmath19 , and we can use the asymptotic expansion of @xmath149 , namely @xmath150 . thus in practice we fit @xmath108 in the disordered phase to the following form , @xmath151 an example fit is shown in figure [ data_g_fit097 ] . in figure [ data_xi ]
we show the correlation length of translational order as calculated by fitting @xmath108 to the above form .
results are shown for the @xmath28 and @xmath95 particle systems .
clearly , @xmath152 remains finite even as the orientational correlation length diverges .
however , there are some discrepancies in our values of @xmath152 . at @xmath153
, the value of @xmath152 extracted from the @xmath28 particle system does not agree with the value for @xmath95 particles .
we believe that some of this difference may be attributable to finite size effects .
additionally , there is also the possibility that the @xmath95 particle system has not fully thermalized . while we have tried to ensure that the data for this largest system is completely thermalized , it can be very difficult to distinguish between stable and metastable states . in either case
, we can not consistently fit all the data to the kthny form , @xmath154 , so instead we have made the fit for only the @xmath28 data .
the result of a fit with @xmath155 , @xmath156 and @xmath157 using @xmath158 is shown in figure [ data_xi ] as the red curve .
in addition , a few other curves are also shown for different values of these parameters with the same value of @xmath159 whose significance is discussed next . 0.3 in to @xmath160 , with @xmath8 fixed at its maximum value of 1/3 .
at @xmath161 , the fitting range is from @xmath152 to @xmath162 .
the solid lines are obtained from the kthny form , @xmath163 , using @xmath158 and various values of the other parameters .
our best fit corresponds to the red curve .
, title="fig : " ] 0.3 in 0.3 in versus @xmath164 , i.e. , the logarithm of the ratio of the finite - system - size to the measured correlation length , for the two size systems , using @xmath165 and for various parameters . due to our calculation of @xmath166 being a lower bound translational order , only data for @xmath167 are shown.,title="fig : " ] 0.3 in versus @xmath164 , i.e. , the logarithm of the ratio of the finite - system - size to the measured correlation length , for the two size systems , using @xmath165 and for various parameters . due to our calculation of @xmath166 being a lower bound translational order , only data for @xmath167 are shown.,title="fig : " ] 0.4 in in figure [ fss_trans ] we show the approximate validity of finite - size scaling by plotting the dimensionless quantity @xmath168 versus @xmath164 , i.e. , the logarithm of the ratio of the finite - system - size to the measured correlation length , for two different size systems using the lower bound of @xmath169 according to the kthny theory , namely @xmath170 .
the best collapse is obtained for the parameters @xmath171 , @xmath172 and @xmath159 shown as the top curve in figure [ fss_trans ] .
( note that we have used the constraint @xmath173 as indicated by the behavior of the critical exponent @xmath8 ) . using the values of the parameters obtained for this `` best '' collapse
we obtain the curve for @xmath4 shown as a green line in fig .
[ data_xi ] . the collapse obtained by using the parameters obtained by the best fit to the correlation length ( corresponding to the red curve in fig .
[ data_xi ] ) is shown as the graph at the bottom .
we have also included two more fits of both types of data , obtained by using parameter values between the above two extremes .
we can observe that while we do not obtain the best fit of both sets of data ( i.e. , collapse of @xmath168 versus @xmath164 ( fig . [ fss_trans ] ) , and the temperature dependence of @xmath4 ( fig .
[ data_xi ] ) for the same values of these parameters we see that the values of @xmath6 and @xmath174 are close , only the prefactor @xmath175 can not be accurately determined .
we feel that the overall quality of fit is reasonable given the fact that we had the difficulty discussed above in determining the correlation length associated with translational order .
several experimental investigations @xcite have used the decay of the envelope of @xmath108 to extract @xmath152 .
the resulting values of @xmath152 appear not to diverge across the melting transition , so perhaps there is some shortfall in using @xmath108 to get @xmath152 at low temperature .
for instance , murray and van winkle observe a finite peak in @xmath152 , while for @xmath99 a divergence is seen to occur @xcite .
regardless of these differences , if we plot the results for @xmath4 on the same plot with the results for @xmath5 as shown in figure [ compare ] , we see clearly that these two correlation length are very different and the differences between these various fitting forms for @xmath152 are not significant on this scale . 0.3 in .,title="fig : " ] 0.3 in
a central concept in finite - size scaling theory is that any dimensionless quantity should be a function of dimensionless ratios of the finite - size length ( @xmath87 ) of the system to the correlation length @xmath4 which emerges naturally and it diverges near the critical point @xcite .
therefore , close enough to the critical point a dimensionless quantity becomes a scaling function @xmath176 . at precisely the critical point , where the correlation length diverges
, all dimensionless quantities are expected to be independent of the system size .
a straightforward way to construct a dimensionless variable is to take the ratio of cumulants .
a simple non - trivial ratio is the so - called binder ratio @xcite of the fourth and second cumulants , @xmath177 as mentioned above , this ( dimensionless ) variable is expected to be system - size independent at a critical point .
hence , if the values of @xmath178 for several system sizes are plotted across a continuous phase transition , they should cross at the critical point .
this is the standard way of estimating for example the critical temperature of a thermal phase transition using the method of binder ratios . in the case of melting in two dimensions ,
however , we have seen that there are two correlation lengths : one for translational order , and another for bond orientational order .
clearly , if we approach very close to either @xmath7 or @xmath7 only one of these two correlation lengths dominates .
for example if we approach @xmath6 sufficiently close from above , @xmath152 becomes very large and @xmath99 is infinite .
thus , there is only one finite correlation length .
when we approach @xmath7 from above , both @xmath99 and @xmath152 are finite , but if we are sufficiently close to @xmath7 , @xmath179 and so we can neglect the influence of @xmath152 . in practice , however , because @xmath99 grows very rapidly as the temperature @xmath7 is approached and we can only study finite - size size systems , the size of @xmath152 is not necessarily negligible as compared to the size of @xmath99 .
this implies that the scaling function becomes @xmath180 . as we have shown in the previous section ,
@xmath152 is still finite when @xmath181 diverges at the upper critical temperature , @xmath18 .
thus , the binder ratio would only be expected to have a crossing at @xmath18 if @xmath182 , which is not the case for the system sizes we have considered ( @xmath183 , half the length of the smallest system size ) . however , depending on the exact form of the scaling function , there may still be a crossing in the vicinity of @xmath18 . 0.3 in near the crossing temperature , @xmath184.,title="fig : " ] 0.3 in looking at the binder ratio in figure [ abind ]
, we can see that there is an apparent crossing of @xmath185 at @xmath186 .
although our statistical uncertainty is too great to identify the system - size dependence of the binder ratio crossing ( see inset ) , the finite - size scaling theory outlined above indicates that @xmath184 should approach @xmath18 when @xmath182 .
thus , while we could use the value of @xmath184 as an estimate of @xmath18 , the method obtained earlier for critical exponents is expected to yield more accurate results to the system sizes considered here . while we have also calculated @xmath187 , the shortfalls of our estimator for translational order in the temperature region @xmath188 lead to an inconclusive analysis of the binder crossing .
lastly , let us point out that the finite - size scaling theory discussed in this section should be applicable to any dimensionless parameter . in sec .
[ correlations ] we demonstrated the scaling collapse of the quantity @xmath189 when plotted as a function of @xmath190 . in light of the analysis above , it is clear that we have neglected the @xmath152 dependence of this dimensionless quantity . in figure [ data_xi ] we can see that for @xmath191 @xmath152 is more or less constant ( @xmath192 ) . but as @xmath18 is approached , @xmath152 increases more rapidly , such that @xmath193 .
this could explain the scatter seen in the scaling collapse of @xmath189 shown in figure [ fss_bond ] .
we have shown that several key predictions from the kthny theory of two - stage continuous melting are seen in the classical system of lennard - jones ( lj ) particles in two dimensions .
first , using delaunay triangulation we can define disclinations and dislocations and this allows us to investigate the role of defects in the 2d melting process .
we can clearly observe at low temperature that disclinations of 5-fold coordinates atoms and disclinations of 7-fold coordinated atoms are bound into dislocations which themselves are bound into dislocation pairs . near @xmath6 we begin to see unbound dislocations and at a higher temperature we begin to observe unbinding of disclinations .
the derivative with respect to temperature of the total defect fraction exhibits a broad peak near @xmath194 very similar to the specific heat peak .
near this temperature we find that the short - range peak ( main peak ) of the pair distribution function of the 5-fold coordinated atoms and that of the 7-fold coordinated atoms greatly diminishes .
the pair distribution function of 5-fold-7-fold coordinated particles also decreases greatly at roughly the same temperature .
we calculated the distribution of the order parameters @xmath0 and @xmath1 on the complex plane .
below @xmath6 , we see the characteristic `` mexican hat''-like circularly symmetric distribution for @xmath0 , i.e. , while the magnitude of @xmath0 is finite below @xmath6 , its phase fluctuates causing the system to lose its translational order .
the orientational order parameter , @xmath1 , however , remains frozen to a particular direction below @xmath6 because the system is large enough to allow , for all practical purposes , for such a spontaneous symmetry breaking .
in the temperature range @xmath195 , the `` mexican - hat''-like distribution of @xmath0 collapses to a distribution around zero , while the distribution of @xmath1 becomes `` mexican - hat''-like .
this could serve as a textbook description of the hexatic order .
for @xmath137 the distribution of both @xmath0 and @xmath1 are centered around zero value .
we also calculated the temperature dependence of the second moment of the above two order parameters for various size systems and from the size - dependence of the results we have extracted the anomalous dimensions , i.e. , the critical exponents @xmath8 and @xmath9 .
furthermore , we calculated the correlation functions @xmath196 and @xmath197 of the order parameter @xmath0 and @xmath1 respectively .
we find that both are controlled by two characteristic correlation lengths , one is @xmath4 , which characterizes the decay of the correlation of the atomic positions and the other is @xmath5 , which provides the decay of the bond - orientation correlations .
we demonstrate that we can accurately extract both @xmath4 and @xmath5 .
we find that the two correlation lengths @xmath5 and @xmath4 have very different temperature dependence , each diverging as we lower the temperature at two different characteristic critical temperatures @xmath7 and @xmath6 respectively , obtained by fitting the calculated correlation length to the forms suggested by kthny theory .
furthermore , using the calculated correlation length and critical exponent @xmath9 , we find that the dimensionless quantity @xmath134 as a function of the dimensionless ratio @xmath133 for all size - lattice considered here collapse onto the same scaling function . a similar conclusion is also reached for the finite - size scaling of the corresponding quantities related to the translational order , i.e. , @xmath198 versus @xmath164 .
this provides additional support for the kthny theory .
99 j. m. kosterlitz and d. j. thouless , j. phys .
c , * 6 * , 1181 ( 1973 ) . b. i halperin and d. r. nelson , phys . rev
* 41 * , 121 ( 1978 ) .
d. r. nelson and b. i halperin , phys . rev .
b * 19 * , 2459 ( 1979 ) .
young , phys .
b * 19 * , 1855 ( 1979 ) .
w. krauth , _ statistical mechanics : algorithms and computations _ , oxford university press oxford , u.k . , (
n. metropolis et al . ,
j. chem . phys .
21 , 1087 ( 1953 ) .
b. j. alder and t. e. wainwright , phys . rev .
127 , 359 ( 1962 ) .
k. j. strandburg , rev .
* 60 * , 161 ( 1988 ) .
s. toxvaerd , j. chem .
phys . * 69 * , 4750 ( 1979 ) ; phys .
lett . * 44 * , 1002 ( 1980 ) ; phys .
a * 24 * , 2735 ( 1981 ) .
f. f. abraham , phys .
* 44 * , 463 ( 1980 ) .
a. f. bakker , c. bruin , and h. j. hilhorst , phys .
52 * , 449 ( 1984 ) .
k. j. strandburg , j. a. zollweg , and g. v. chester , phys .
b * 30 * , 2755 ( 1984 ) .
s. t. chui , phys .
* 48 * , 933 ( 1982 ) ; phys .
b * 28 * , 178 ( 1983 ) .
d. frenkel and j. p. mctague , phys .
lett . * 42 * , 1632 ( 1979 ) .
c. udink and j. van der elsken , phys .
b * 35 * , 279 ( 1987 ) .
f. l. somer , jr .
, et al . , phys .
lett . * 79 * , 3431 ( 1997 ) . f. l. somer , jr . , g. s. canright , and t. kaplan , phys .
e * 58 * , 5748 ( 1998 ) .
k. binder , z. phys .
b * 43 * , 119 ( 1981 ) .
m. p. allen , d. frenkel , w. gignac , and j. p. mctague , j. chem .
* 78 * , 4206 ( 1983 ) .
k. wierschem and e. manousakis , physics procedia * 3 * , 1515 ( 2010 ) . c. a. murray and d. h. van winkle , phys .
58 * , 1200 ( 1987 ) .
y. tang , a. j. armstrong , r. c. mockler , and w. j. osullivan , phase transitions * 21 * , 75 ( 1990 ) . v. privman , _ finite size scaling and numerical simulation of statistical systems _ ( world scientific , singapore,1990 ) . | we study the nature of melting of a two dimensional ( 2d ) lennard - jones solid using large scale monte carlo simulation .
we use systems of up to 102,400 particles to capture the decay of the correlation functions associated with translational order ( to ) as well as the bond - orientational ( bo ) order .
we study the role of dislocations and disclinations and their distribution functions .
we computed the temperature dependence of the second moment of the to order parameter ( @xmath0 ) as well as of the order parameter @xmath1 associated with bo order . applying finite - size scaling of these second moments we determined the anomalous dimension critical exponents @xmath2 and @xmath3 associated with power - law decay of the @xmath0 and @xmath1 correlation functions .
we also computed the temperature dependent distribution of the order parameters @xmath0 and @xmath1 on the complex plane which support a two stage melting with a hexatic phase as an intermediate phase . from the correlation functions of @xmath0 and @xmath1 we extracted the corresponding temperature dependent correlation lengths @xmath4 and @xmath5 .
the analysis of our results leads to a consistent picture strongly supporting a two stage melting scenario as predicted by the kosterlitz , thouless , halperin , nelson , and young ( kthny ) theory where melting occurs via two continuous phase transitions , first from solid to a hexatic fluid at temperature @xmath6 , and then from the hexatic fluid to an isotropic fluid at a critical temperature @xmath7 .
we find that @xmath4 and @xmath5 have a distinctly different temperature dependence each diverging at different temperature and that their finite size scaling properties are consistent with the kthny theory .
we also used the temperature dependence of @xmath8 and @xmath9 and their theoretical bounds to provide estimates for the critical temperatures @xmath6 and @xmath7 , which can also be estimated using the binder ratio .
our results are within error bars the same as those extracted from the divergence of the correlation lengths . |
the initial mass function ( imf ) is a fundamental property of stellar populations , hence one of the most crucial ingredients in models of galaxy formation and evolution .
it determines several key properties of stellar populations and galaxies , such as the yield of heavy element production , the luminosity evolution over time , the mass - to - light ratio , the total star formation rate at low and high redshifts as inferred from empirical estimators , and the energetic feedback into the interstellar medium . yet , the imf is usually taken as a free parameter , particularly at the low - mass end ( for recent reviews on the imf see larson 1998 ; scalo 1998 , 1999 ) .
observational constraints on the imf are therefore of the greatest astrophysical importance . knowing the imf at @xmath16 in spiral bulges and elliptical galaxies is of special interest because these spheroids contain a large fraction , perhaps a majority , of all the stellar mass of the universe ( e.g. , fugugita , hogan & peebles , 1998 ) .
however , there is presently no way to directly determine the imf of spheroids except by measuring the luminosity function ( lf ) of our own bulge as the only surrogate for the unresolvable population in other galaxies .
although the low mass end of the stellar imf has been determined for the solar neighborhood ( kroupa , tout & gilmore 1993 ; gould , bahcall & flynn 1997 ; reid & gizis 1997 ) and in young open clusters ( hillenbrand 1997 ; bouvier et al . 1998 ; luhman et al .
1998 ) it is only in the galactic bulge that one can be confident that the stellar population is old , largely coeval , and metal rich ( whitford 1978 ; ortolani et al . 1995 ; mc william & rich 1994 ) , i.e. , the closest we can come in a nearby , resolved stellar population to what prevails in other spiral bulges and elliptical galaxies ( renzini 1999 ) . the recent discovery of a high rate of microlensing events towards the bulge ( udalski et al .
1994 ; alcock et al .
1997 ) has made the determination of the faint end of the imf a yet more urgent problem . in brief , if the bulge imf is close to that of the solar - neighborhood ( gould et al . 1997 ) then the bulk of the short ( @xmath17 day ) microlensing events would remain unexplained , perhaps requiring a large population of brown dwarfs ( han 1997 ) .
however , an imf extending to the h - burning limit with a salpeter law can account for both the total mass of the bulge and the frequency of microlensing events ( zhao , spergel , & rich 1995 ) .
it is therefore tempting to suspect that the bulge and solar - neighborhood imfs are different .
however , the interpretation of the microlensing events relies on assumptions about the phase - space distribution of both the lenses and sources , and some events may be caused by collapsed stars , brown dwarfs or even non - stellar objects .
therefore , the most reliable way to resolve these ambiguities ( and thus maximize the information from microlensing itself ) is to obtain a representative stellar inventory of the bulge from star counts , and to incorporate this into the microlensing analysis .
a recent determination of the bulge imf down to @xmath18 has been provided by holtzman et al .
( 1998 ) , based on hubble space telescope ( hst ) + wide field and planetary camera ( wfpc2 ) observations of baade s window .
globular clusters offer another approach towards determining the imf for @xmath19 , and deep hst observations are indeed providing important information on their present - day mass functions ( de marchi et al .
1999 ; piotto & zoccali 1999 , and references therein ) .
however , clusters suffer from dynamical evolution and evaporation of low - mass stars , and therefore there is no model - independent way to infer their imfs from their observed present day mass function ( mf ) . because faint ,
low - mass stars have such low temperatures , infrared observations give a crucial advantage over optical data .
moreover , in the near - ir the effects of extinction and differential reddening are considerably reduced , and the bolometric luminosities of m dwarfs ( the vast majority of the sampled stars ) are best determined in the near - ir both because of their cool temperatures and severe molecular blanketing in the optical .
finally , the relatively low ir background of hst , combined with diffraction limited resolution , gives a fundamental advantage when dealing with very faint sources in a crowded field .
therefore , the nicmos near - ir cameras offer a unique opportunity to reach the faintest stars possible in the galactic bulge , thus extending to lower masses the range over which the imf is observationally constrained . in order to ensure the success of the project we paid special attention to the selection of the bulge field to be observed .
the most widely studied field in the galactic bulge is the @xmath20 field known as baade s window .
however , for the nicmos observations we did not choose to point hst at this field .
a priori , in baade s window , crowding might have been too severe to confidently undertake this experiment that aims at counting the faintest bulge stars in the frame . at @xmath20
the average surface brightness ( corrected for @xmath21 mag extinction ) is 18.7 @xmath22 mag@xmath23arcsec@xmath24 ( terndrup 1988 ) , and with a true modulus of 14.5 mag one samples @xmath25 mag@xmath23arcsec@xmath24 , corresponding to a bolometric luminosity of @xmath26arcsec@xmath24 ( using population synthesis models , e.g. , by maraston 1998 ) .
hence , the nic2 camera samples a total bolometric luminosity @xmath27 .
this allows one to estimate the number of main sequence stars in a hst nic2 frame , knowing that for a @xmath28 gyr old population the scale factor in the imf
@xmath29 is given by @xmath30 ( renzini 1998 ) .
integrating the imf from 0.1 to 0.9 @xmath6 , with @xmath8 ( the salpeter imf slope ) , and @xmath31 , we get that a nic2 frame will contain @xmath32 stars .
since the nic2 camera has @xmath33 pixels , while selecting the target field we therefore concluded that accurate photometry would have been difficult towards the faint end of the lf , if the imf were to follow the salpeter s slope all the way to the hydrogen burning limit . for our observations we selected instead the field at @xmath34 , where the surface brightness is @xmath35 mag lower , and then we expect @xmath36 times fewer stars in a nic2 frame compared to baade s window , significantly improving the stars / pixel number ratio .
although more distant from the nucleus than baade s window , the field population is still dominated by the metal rich stars characteristic of the bulge , as shown by the strongly descending red giant branch in the @xmath37 diagram ( rich et al .
1998 ) , which makes sure that we are properly studying the metal rich bulge in this location .
moreover , photometry down to the hydrogen burning limit should not be compromised by crowding , especially if the imf were to flatten out below the salpeter s slope as in the solar neighborhood ( gould et al .
1997 ) , implying a smaller number of low - mass stars .
the selected field ( ra=18:11:05 , dec=@xmath3831:45:49 ( j2000 ) ; @xmath39=0.277 , @[email protected] ) was observed with the nic2 camera of nicmos , on board hst , through the filters f110w and f160w .
parallel observations with nic1 were collected through the f110w filter .
fourteen orbits were allocated , for total nic2 integration times of 10240 s and 25600 s in f110w and f160w , respectively . only the f110w filter was used for the nic1 parallel observations , for a total integration time of 35850 s. all exposures were obtained using the multiaccum readout mode and the step64 time sequence through an eight - position spiral dithering with size of @xmath41 .
the pixel size of the nic2 detector is @xmath42 , giving a field of view of @xmath43 for each frame .
small offsets and rotations among the frames gave us a slightly larger total field ( @xmath44 ) .
figure [ field ] shows the observed bulge region as it appears in a combination of all the frames . the images were bias subtracted ,
dark corrected and flat - fielded by the standard nicmos pipeline calnica .
this routine also combines the multiple readouts of the multiaccum mode , giving an output image which is expressed in counts per second per pixel .
we therefore multiplied each of these images by its total exposure time , so that the photometry software would measure the correct signal - to - noise ratio ( s / n ) .
0.4 cm the data quality file corresponding to each image was used to mask out the saturated and bad pixels by setting them to a very high value , that was discarded in the photometry . following the nic2 manual ,
the read - out noise of each frame was assumed to be 32 electrons , corresponding to 6.1 adu , with a conversion factor of 5.4 @xmath45adu .
the mean sky level of each 640 s exposure was @xmath46 and @xmath47 adu in f110w and f160w respectively .
that is , the noise is dominated by read noise rather than the sky . preliminary star finding and aperture photometry was carried out on each frame using the daophotii photometry package ( stetson 1987 ) .
we then used all the stars identified in each frame to obtain the coordinate transformations among all the frames .
these transformations were used to register the frames and obtain a median image .
the latter , having the highest s / n , was used to create the most complete star list , by means of two complete runs of daophotii and allstar .
the final star list , together with the coordinate transformations , was finally used as input for allframe ( stetson 1994 ) , for the simultaneous reduction of all the frames .
particular attention was devoted to modeling the nic2 point spread function ( psf ) in the two filters .
this was performed using specific software ( multipsf ) , provided by p.b .
stetson , that allows measurement of a unique psf from the brightest and most isolated stars in a set of different frames .
assuming that the psf profile does not change from frame to frame , we were able to measure the same @xmath48 stars in all the frames of each filter .
the spatial dithering allowed us to measure the selected stars in different locations on the chip , centered in different positions inside a pixel , so the final psf was of considerably better quality than the one we could obtain for each frame taken individually .
the stellar full width at half maximum is @xmath49 pixels while the adopted model psf was defined up to a 14 pixels radius .
aperture corrections were empirically determined on the most isolated stars , and applied to the allframe measures in order to obtain the stellar magnitudes in a @xmath50 aperture .
the magnitudes were then converted to count rates , and multiplied by 1.15 to correct to an infinite aperture .
the inverse sensitivity , given as the keyword photflam in the header of the images , together with the zero points photzpt given in table 2 of stephens et al .
( 1999 ) , were used to convert the count rates into hst @xmath51 and @xmath52 magnitudes .
the latter were then transformed in the cit / ctio system according to the calibration equations determined by stephens et al .
( 1999 ) from the comparison between nicmos and ground - based observations of the same 14 bright stars . as discussed by stephens et al .
( 1999 ) , this calibration is consistent ( over the common color range ) with the one described in the nicmos calibration documentation .
a second , independent reduction of the data was carried out with the same software ( daophotii / allstar ) but with somewhat different procedures .
nearly identical results were obtained as in the first reduction . in this reduction
we performed simple star finding and psf fitting on each individual frame but without using the median image , or allframe .
the resulting photometry is somewhat shallower but it provides a useful consistency check both in terms of magnitudes and numbers of identified stars , in the common magnitude range .
figure [ andy ] shows the comparison between the ( calibrated ) output photometry of the two procedures .
the two bottom panels show that , for the stars identified in both cases ( i.e. , except for the fainter ones , identified only by allframe ) the measured magnitudes are in very good agreement , with a very small offset @xmath53 , due to some systematic error in one ( or both ) of the aperture corrections .
the @xmath54 magnitudes of the brightest stars ( @xmath55 ) also differ by @xmath56 .
the top panel shows the two lfs ( before completeness correction ) which are almost identical down to @xmath57 , where the allframe reductions go significantly deeper
. 0.4 cm
the observed color magnitude diagram ( cmd ) for the 780 stars measured in our @xmath44 field is shown in figure [ cmd_iso ] .
only the stars identified in at least 5 independent frames per filter are plotted . a further selection on the magnitude error and on the _ sharp _
parameter was imposed to discard spurious detections due to noise and intersecting diffraction spikes that may remain around the brightest stars .
the bulge main sequence ( ms ) is well defined from the turnoff ( @xmath58 ) down to magnitude @xmath59 where the sequence starts to broaden and the density of stars falls abruptly .
a prominent feature in this cmd is the sharp bend that is clearly visible at @xmath60 .
fainter than this point , the ms is almost vertical . as predicted by stellar models ( cassisi et al .
1999 ; baraffe et al .
1997 ) , this behavior is due to the competition between the tendency towards redder colors due to both the decreasing effective temperature and the increasing molecular absorption at optical wavelengths , and the increasing collision - induced absorption of molecular hydrogen at infrared wavelengths ( cia mechanism , saumon et al .
1994 ) .
the two brightest stars in the cmd of figure [ cmd_iso ] , located in the left side of our field ( fig . [ field ] ) , were saturated ; their magnitudes have been measured independently by extrapolating their psf profiles into the central region .
0.4 cm also shown in figure [ cmd_iso ] is the theoretical isochrone by cassisi et al .
these models have been constructed by adopting the most updated input physics , such as stellar opacities , equation of state , and outer boundary conditions ( see cassisi et al .
1999 , for more details ) .
we adopt here the absolute distance modulus and reddening of this region of the galactic bulge , as measured by rich et al .
( 1998 ) : @xmath61 and @xmath62 . by assuming @xmath63
the extinction is @xmath64 , which can be converted to the corresponding @xmath65 and @xmath66 by means of the relations given by cardelli et al .
( 1989 ) : @xmath67 and @xmath68 .
the isochrone shown in figure [ cmd_iso ] refers to solar metallicity ( [ fe / h]=[@xmath69/fe]=0 ) and an age of 10 gyrs . the model is a satisfactory match to the general shape of the observed ms , in particular the position of the bend at @xmath4 ( @xmath70 ) is well reproduced , even if its strength seems to be a little overestimated .
this feature also provides a good check of the zero point of the photometric calibration and the adopted distance and reddening .
the present nicmos data provide a too sparse sampling of the turnoff area for properly addressing the issue of the age of the bulge stellar populations .
this will be attempted in a future paper , combining our nicmos data with deep wfpc2 observations of the same field , as well as wide field @xmath22 and @xmath71 observations taken at the eso / mpia 2.2 m telescope ( zoccali et al .
in order to obtain the stellar lf of our field , particular attention was devoted to estimating the completeness of our sample .
standard artificial - star tests were carried out on the nic2 field , in the same way as described in detail by piotto & zoccali ( 1999 ) .
we performed ten independent tests , by adding about seventy stars each time , with magnitudes in the range @xmath72 .
visual inspection of the stars - subtracted image insured that our photometry was complete for brighter magnitudes .
the artificial stars were arranged in a spatial grid such that the separation between the centers of each star pair was two psf radii plus one pixel .
this allowed us to add the maximum number of stars , without creating overcrowding .
in addition , the position of each star in the grid was randomly located inside one pixel , so as to prevent the centers of all the artificial stars from falling on the same position within a pixel , which would have biased their probability of being detected .
the artificial stars were added on each individual @xmath9 and @xmath54 image .
it should be noticed that the stars must be added in the same position _ on the sky _
, therefore their coordinates must be different in different frames , following the frame - to - frame coordinate transformations calculated from the original photometry .
a high precision is required in this process , in order to be able to measure the artificial stars with the same photometric accuracy of the original ones .
we then ran the same photometry procedure used for the original photometry : star finding was performed on the median of all the star
added images , and then allframe was used for the simultaneous photometry of all the frames .
the same selection criteria used for the original stars were applied to the output list of the artificial star tests .
0.4 cm the completeness correction obtained in this way was applied to the lf obtained from the cmd of figure [ cmd_iso ] .
this procedure also automatically compensates for the differences of the total integration time across the @xmath73 field .
it is worth noting that the scatter in the color of the stars on the right of the main sequence , for @xmath74 , is also present in the cmd for the artificial stars , which indicates that the effect is spurious .
visual inspection of these stars on the image revealed that they are all located on the left side of the field , where scattered light from a few very bright objects is also present . some of them could be residual noise spikes , but some are likely to be real stars whose magnitude has been enhanced due to the proximity of brighter stars .
the fact that these objects are present only on the right side of the main sequence indicates that such effect is stronger for the @xmath54 magnitudes , a likely result of the poorer psf in the @xmath54-band .
the way in which we applied the completeness correction ( i.e. determining the completeness fraction as a function of the recovered magnitude of the artificial stars , instead of the input magnitude ) automatically takes into account the effect of the migration of the stars towards brighter magnitudes , therefore we did nt impose any further selection on the cmd of figure [ cmd_iso ] .
the resulting @xmath9-band lf is shown in figure [ lf ] ; it is very smooth over the whole range from the bin at @xmath75 ( turnoff region ) to the faint limit at @xmath76 . also shown as
a dotted histogram is the raw lf , without completeness correction . in the determination of the imf
, we did not use the first two bins that , according to our model , correspond to evolved stars , nor the very last bin ( @xmath77 ) because its completeness is @xmath78 .
the second to last bin , at @xmath76 , is complete at @xmath79 .
since the field is located at low galactic latitude , contamination by disk stars can not be neglected .
we offer an estimate of this contamination using kent s ( 1992 ) model for the @xmath80 band luminosity distributions of the disk and bulge .
if the lfs of the disk and bulge have the same form as the observed lf in our field , scaled for distance and stellar density , then we find that about 11% of the stars in our field are disk stars , with a modest trend from about 9% at the faint end ( @xmath81 ) , to about 14% at the bright end ( @xmath82 ) of the lf in figure [ lf_fro ] .
we then adopt an overall reduction of the lf by 11% for @xmath83 .
we note , however , that this is only a rough estimate of the disk contamination .
available data do not allow a more accurate correction , and further optical and ir data would be required to address this problem more properly .
the lf of figure [ lf ] is listed in table [ tab_lf ] .
column 1 gives the @xmath9 magnitude , column 2 and 3 give the raw and completeness corrected counts , respectively , column 4 is the error , and column 5 the estimated contribution from disk stars .
contamination by extragalactic objects is estimated using the nicmos @xmath54-band galaxy counts ( yan et al .
this gives less than one galaxy for @xmath84 , and between 1.2 and 2.9 galaxies in the last two bins of our lf , corresponding to @xmath85 and @xmath76 , respectively .
we therefore conclude that this source of contamination can be neglected in our analysis .
0.4 cm an additional set of data on a nearby bulge field is provided by our parallel observations with the nic1 camera through the f110w filter .
the field of view of nic1 is significantly smaller ( @xmath86 ) than that of nic2 , but thanks to its smaller pixel size ( 0@xmath87043 ) , it allows more accurate sampling of the psf and therefore it yields more accurate photometry for stars with good photon statistics . in our case , due to the rotation of the nic2 field in different visits , the nic1 camera actually mapped a larger region ( @xmath88 ) .
the use of only one filter has the disadvantage that it is not possible to construct a cmd , but allowed a longer exposure time ( 35850 s ) .
we expect nic1 to be more complete than nic2 at intermediate magnitudes because of its better resolution and longer exposure time ( and hence higher s / n ) . at faint magnitudes
the higher nic1 read noise implies that nic1 data should have only slightly better s / n despite the longer exposures .
nevertheless , we expected to be able to push the lf to somewhat fainter magnitudes by incorporating the nic1 data .
we reduced the nic1 frames with the same algorithm adopted for nic2 , and were able to measure about 800 stars .
unfortunately , the selection criteria adopted for nic2 were not sufficient to ensure `` clean '' photometry in this case because in all the nic1 frames there was a shaded region , apparently due to some flat - field problems .
many faint , possibly spurious stars were identified in this region , and we were not able to find a suitable selection criterion to discard them without also losing what seemed to be real stars . due to this problem , and since we had no cmd to guide the selection between real stars and spurious detections , we decided to check each of the 800 stars by eye on the image .
this certainly introduced a brighter magnitude limit , because it was hard to make a selection in the very last magnitude bin , and also prevented us from making any artificial star test , as the selection criterion was not automatic .
thus , despite our expectations , the lf extracted from the nic1 photometry could not be used to extend the nic2 lf to fainter magnitudes .
however , it is useful as a cross check on the nic2 results .
figure [ lfnic1 ] shows the lfs extracted from the nic1 and nic2 data , with no correction for incompleteness .
the two lfs were normalized according to the relative areas of the two fields .
the lf from nic1 falls abruptly below @xmath57 , due primarily to the visual selection to eliminate spurious detections .
however , it is reassuring to note that the two lfs track each other very closely , down to magnitude @xmath57 .
0.4 cm the lf for low - mass stars ( @xmath89 ) can be converted into a mf which is the same as the imf , since the stars are unevolved and their number is unaffected by dynamical processes . in order to transform the lf into the imf a mass luminosity relation ( mlr ) is required .
an empirical mlr in the infrared bands has been determined for solar metallicity stars by henry & mccarthy ( 1993 ) from a sample of visual and eclipsing binaries in the solar neighborhood .
this mlr is shown as a dashed line in figure [ mlr ] ( top panel ) together with the individual data points .
this relation was obtained from a series of quadratic fits in different mass intervals , and would introduce features in the imf at each of the abrupt changes in the slope of the mlr .
also shown in figure [ mlr ] are the mlrs for two sets of theoretical models ( cassisi et al .
the empirical and theoretical mlrs are in very good agreement , apart from a discrepancy by few hundredths of a solar mass near the faint limit . given the large spread and errorbars of the data points , this small discrepancy appears to be completely negligible . from this comparison , and the good fit of the cmd of figure [ cmd_iso ] ,
we feel confident in adopting the theoretical mlr to convert the observed lf into an imf .
the mlr for solar metallicity is adopted in the present work . however , as shown in figure [ mlr ] the metallicity dependence of the mlr is so small that we would expect no appreciable effect even if the average metallicity of the stars in our bulge sample were very different from solar , which is not the case ( mcwilliam & rich 1994 ) .
this is illustrated in the bottom panel of figure [ mlr ] , where the imf obtained from the adopted model mlr ( [ m / h]=0 ) is compared with the one obtained using the mlr for [ m / h]@xmath90 .
the resulting imf for the galactic bulge is shown in figure [ mf_bulge ] . within the errors ,
the imf can be represented over the entire mass range by a single power - law of the form @xmath91 , having a slope @xmath7 ( where salpeter has @xmath8 ) .
it is worth noting that adopting the widely used distance modulus @xmath92 to the galactic center ( reid 1993 ) , instead of the value of 14.38 adopted for this paper , the resulting slope is @xmath93 over the whole range @xmath94 .
hence the result is fairly insensitive to errors in the distance .
as it appears from figure [ mf_bulge ] , there is a hint for the imf to flatten slightly below @xmath95 .
a two - slope imf gives indeed a marginally better fit , with @xmath96 for @xmath97 , and @xmath98 for @xmath99 .
to some extent the presence of binaries can introduce a bias in the derived imf slope .
however , the frequency of binaries in the bulge and the distribution of their mass ratios remains unconstrained by the present data , and therefore we do not simulate the effect of binaries in this work .
holtzman et al . (
1998 ) assumed various binary fractions ( defined as the fraction of systems that are binaries ) up to 50% .
they find that the slope at the faint end steepens by 0.4 for a binary fraction of 50% ( 2/3 of all stars in a binary system , with both primaries and secondaries following the same imf ) .
we conclude that adopting the same procedure of holtzman et al . would bring @xmath69 from @xmath100 to @xmath101 for such an extreme fraction of binaries .
0.4 cm 0.4 cm figure [ mf_bulge ] shows the comparison between the imf derived above and the imf from holtzman et al . ( 1998 ) for the stars in baade s window . for a more consistent comparison
, we used the @xmath71-band lf from their figure [ lf_fro ] , and converted it into an imf by means of the same theoretical mlr that we have used for our derivation ( except for the color transformation to the @xmath71 band instead of the @xmath9 band ) , and without applying a correction for binaries .
the derived imf is very similar to the one that holtzman et al .
( 1998 ) originally obtained down to @xmath102 with an empirical mlr which is in good agreement with that of henry & mccarthy ( 1993 ) .
the imf so derived turns out to be very similar to our bulge imf for the @xmath103 field .
holtzman et al.s baade s window imf has a slope @xmath104 for @xmath97 , and @xmath105 for @xmath106 , virtually identical to our result over the wider mass range that was accessible for the @xmath103 field .
it is also interesting to compare the bulge imf with the imf of the galactic disk .
the comparison is shown in figure [ mf_disk ] , which displays our bulge imf together with the disk imf as recently derived by reid & gizis ( 1997 ) and by gould et al .
reid & gizis ( 1997 ) extract their lf from the study of a volume - complete sample of low mass stars with @xmath107 and within 8 pc from the sun .
their imf , shown in figure [ mf_disk ] , is not corrected for binary stars , and therefore it can be compared with the imf we derive for the galactic bulge . note that this disk imf has also been derived using the henry & mccarthy ( 1993 ) empirical mlr .
the reid & gizis ( 1997 ) disk imf is well represented by a power law , but its slope @xmath108 differs by about @xmath109 from that of the bulge .
the disk imf by gould et al .
( 1997 ) is also shown in figure [ mf_disk ] . for their sample of disk
m dwarfs they found an imf with a slope @xmath110 , in @xmath111 agreement with reid & gizis ( 1997 ) disk imf , but definitely flatter than the bulge imf . on the contrary ,
the imf for the @xmath103 field is in very good agreement with the disk imf obtained by kroupa et al .
( 1993 ) , for stars within @xmath112 pc from the sun , which has a slope @xmath113 for @xmath114 and @xmath115 for @xmath116 .
similar results come from the comparison of the bulge imf with the mf of young open clusters .
recent work in this field has been done by hillenbrand ( 1997 ) , luhman et al .
( 1998 ) and bouvier et al .
( 1998 ) . from an extensive optical study of the orion nebula cluster , hillenbrand ( 1997 ) found an imf slope of @xmath117 for @xmath118 , but also a sharp peak at @xmath119 and a turnover for lower masses .
in contrast , both luhman et al .
( 1998 ) , for the young cluster ic 348 , and bouvier et al .
( 1998 ) , for the pleiades , found a log - normal imf that matches that of miller & scalo ( 1979 ) for @xmath120 and a flatter imf , with slope @xmath121 , for lower masses . 0.4 cm finally , it is interesting to compare the bulge imf with the mf measured in some galactic globular clusters ( gcs ) .
gcs are strongly affected by dynamical evolution which modifies their stellar mf .
several gcs have short relaxation times with respect to their age , and therefore their observed mf changes with radius due to mass segregation and evaporation .
they are also affected by tidal shocks caused by passage through the galactic disk and bulge that preferentially strip lower mass stars . according to dynamical models ( vesperini & heggie 1997 )
the only way to measure a mf unaffected by these dynamical processes is to observe gcs with high mass ( i.e. , long relaxation time ) and very wide orbits ( i.e. , that do not cross the galactic plane frequently ) . in figure [ ggc ] the imf of the galactic bulge
is compared with the mf of a few gcs .
clusters with extreme mfs were chosen to make the clearest plot .
the mf of ngc 7078 ( piotto et al .
1997 ) , a massive ( @xmath122=6.3 , djorgovski 1993 ) very metal poor cluster with very wide orbit ( dauphole et al . 1996 ) , is very similar to the imf of the galactic bulge .
the mf of @xmath123 cen ( pulone et al .
1998 ) , very massive ( @xmath122=6.6 ) but with a smaller orbit , is only slightly flatter , while the mf of ngc 6397 ( king et al .
1998 ) is significantly flatter , this cluster having a smaller mass ( @xmath124 ) and a very tight orbit .
most extreme is the case of ngc 6712 ( de marchi et al .
1999 ) that has a mf with an inverted slope @xmath125 .
this is a low mass cluster ( @xmath126 ) , with an orbit that brings it to within @xmath127 pc of the galactic center .
our finding that the bulge imf is similar to that of the less dynamically affected clusters suggests that the gcs and the bulge may have the same imf .
the similarity of the imf of the solar metallicity bulge with that of ngc 7078 at @xmath128=-2 $ ] suggests that the slope of the imf is relatively independent of metallicity ( see also grillmair et al .
in this section we briefly discuss a few implications and applications of the bulge imf derived in the previous sections .
the bulge lf extending from near the ms turnoff down to the lower ms can be combined with an appropriate lf for bright , evolved stars in the bulge .
this approach permits us to construct a complete bulge lf that extends from the tip of the red giant branch ( rgb ) to nearly the bottom of the ms , which can be compared with theoretical stellar evolution models , and which can be used as a template in a variety of applications . to this end , we have combined our lf with the lf of tiede et al .
( 1995 ) , appropriately scaled by its 8.01 times greater area ( 4056 arcsec@xmath129 versus 506 arcsec@xmath129 ) .
the tiede et al .
field is located only 10 arcmin from our bulge field , and thus should have essentially the same stellar population .
the brightest part of the lf @xmath130 is adapted from the wide - area survey of bulge m giants in baade s window @xmath131 ( frogel & whitford 1987 ) , properly normalized to the 506 arcsec@xmath129 nic2 field both for the area , and for the lower surface brightness of our field .
these lfs have been corrected for the disk contamination .
the disk contribution to the ms was evaluated as described in 4 .
for the stars brighter than the ms turnoff the fraction of disk stars was estimated in a more direct way .
we used @xmath22 and @xmath71 observations of a wide region including the small nic2 field , taken with the 2.2 m telescope + the wide field imager ( wfi ) at eso la silla , on 24 march 1999 ( zoccali et al . 1999 ) .
the cmd derived from these images is very well populated from the tip of the rgb down to about 2 magnitudes below the turnoff , and allows one to separate very clearly the extended disk ms from the evolved population of the bulge : rgb + horizontal branch ( hb ) + asymptotic giant branch ( agb ) . lines of constant @xmath9 magnitude drawn on the @xmath37 cmd using color transformations ( allard et al .
1999 ) allow one to count the number of disk and bulge stars in each @xmath9 bin .
the decontaminated lf , renormalized to our nic2 field is reported in table [ tab_disk ] along with the value of the decontamination correction .
therefore , the lfs in table [ tab_disk ] and table [ tab_lf ] have the same normalization , and the resulting global lf is shown in figure [ lf_fro ] . superimposed on this empirical lf
is the theoretical lf from models by cassisi et al .
( 1999 ) , extended to the tip of the rgb using models by bono et al .
note that the theoretical lf does not include either the hb clump ( clearly visible in the empirical lf ) or the agb .
the sharp peak at @xmath132 is the rgb bump , which is produced by the pause in evolution along the rgb when the hydrogen shell burns through the hydrogen discontinuity left by the deepest penetration of the convective envelope .
this feature is not very clear in the observed lf because it is smeared by distance dispersion and differential reddening , and partially merged with the hb red clump .
both the slope of the rgb lf and the sharp drop between ms and rgb stars are well reproduced by the model .
we note that the apparent overabundance of stars in the brightest bins is due to the observed lf including agb stars , while the model lf does not .
as is apparent , this theoretical lf is in excellent agreement with the empirical one , when allowance is made for the hb and agb contributions .
the sharp drop near @xmath75 corresponds to the beginning of the post - ms evolution , and its location is age dependent .
however , no tight limits on the age can be placed here , as this drop would be displaced towards fainter luminosities by only @xmath133 mag per 1 gyr increase in age .
the age of the bulge stellar population is discussed by ortolani et al .
it will be examined again using a variety of data now available for this @xmath103 field ( zoccali et al .
1999 ) . 0.4 cm having determined the actual slope of the imf
, we are now in a position to check the theoretical prediction concerning the number of ms stars in the observed field : @xmath134 where @xmath30 ( renzini 1998 ) . from an optical cmd ( zoccali et al .
1999 ) referring to a field of @xmath135 arcmin@xmath129 and using the distance and reddening adopted in the present paper , we determine an average surface brightness of the @xmath103 field of 0.55 @xmath136 . correcting for the disk contamination
leaves an average surface brightness of the bulge alone of 0.35 @xmath136 .
hence , the total luminosity sampled by our 506 arcsec@xmath129 field is @xmath137 , or @xmath138 .
correspondingly , the number of stars in the @xmath139 range is given by : @xmath140 which compares well to the 820 stars observed .
by integrating the lf shown in figure [ lf_fro ] we determine the total @xmath9 band luminosity sampled by a 506 arcsec@xmath129 field to be @xmath142 .
note that the stellar population sampled by our nic2 field would not be representative of the entire bulge , being very small , and chosen to be in a region lacking very bright stars .
however , the bright part of the lf was derived using stars in a field 8 times wider , and therefore we can trust the total luminosity calculated above as representative of the average surface brightness of the bulge at @xmath34 . the total bulge mass in stars included in our 506 arcsec@xmath129 field corresponds to the sum of the mass of the detected stars , plus the mass of m dwarfs and brown dwarfs with @xmath143 , plus the mass of white dwarfs , neutron stars , and black hole remnants , the end products of now defunct stars with @xmath144 .
we estimate the total mass in our field as follows .
first , we simply sum the masses of the stars actually observed in the field ( corrected for incompleteness and disk contamination ) and obtain @xmath145 . by extrapolating the @xmath146 imf from @xmath147 all the way down to zero mass
, we obtain @xmath148 of unseen dwarfs , thus totaling @xmath149 in living stars and brown dwarfs . to account for the remnants we need to adopt an initial mass - final mass relation .
we used the semi - empirical relation proposed by renzini & ciotti ( 1993 ) , with white dwarf remnants of mass @xmath150 for initial masses @xmath151 , neutron star remnants of @xmath152 for @xmath153 , and black holes remnants of mass @xmath154 for @xmath155 .
since the present data do not give any constraint on the slope of the imf for @xmath156 , we explore the effect on the total mass of various plausible assumptions : * imf # 1 : an imf with slope @xmath146 , like the one we observed , all the way to @xmath157 .
this is perhaps an extreme possibility , since all the determinations of the imf in this mass range give steeper values ( see scalo 1998 for a recent review ) , and even our own imf may steepen for @xmath97 .
* imf # 2 : an imf with slope @xmath146 up to @xmath158 and @xmath159 for @xmath160 .
this is the most conservative assumption , since the imf we observed is best fit with a slope @xmath159 for @xmath97 .
* imf # 3 : an imf with @xmath146 up to @xmath158 and @xmath8 ( salpeter s value ) for @xmath160 .
* imf # 4 : finally , we consider an imf with @xmath146 up to @xmath158 , salpeter slope for @xmath161 and @xmath162 ( scalo 1986 ) for @xmath163 . for each of these four choices , table 3 gives the total mass in the 506 arcsec@xmath129 field , as well as the contribution of white dwarfs , neutron stars , and black holes .
of course , the mass of the unseen dwarfs ( @xmath164 ) and detected ms dwarfs is the same for all the imf options .
finally , the last column gives the corresponding @xmath165 ratio .
option 1 is clearly top heavy , with most of the bulge baryonic mass in 20 - 50 @xmath166 black holes . with option 2 one
gets rid of most of black holes , and the mass - to - light ratio drops to near unity .
further steepening the imf , such as in options 3 and 4 , ceases to have a major effect on the mass - to - light ratio , while reducing to just a trace contribution the mass of relativistic remnants .
as the microlensing statistics improves , microlensing experiments may eventually allow us to select the best among these ( or other ) options .
in this section we consider the implications of our bulge imf for the interpretation of the microlensing events that have been observed in the direction of the galactic bulge .
only about 50 of these events have been published so far ( udalski et al .
1994 ; alcock et al.1997 ) , but by now at least ten times more events should have been detected .
the initial results have generated two somewhat orthogonal puzzles .
first , the distribution of event timescales @xmath167 is peaked toward much lower values ( @xmath168days ) than would be expected if the bulge imf were as flat as @xmath169 quoted in 7 , because the latter was obtained on the restricted mass interval in common with our nicmos data ] , as reported by gould et al .
( 1997 ) for the disk mf , but would be well explained by a power - law imf with @xmath170 and cut off near the hydrogen - burning limit ( zhao et al .
1995 ; han & gould 1996 ) .
lower - mass lenses produce events that on average are shorter ( @xmath171 ) , so a steeper imf gives rise to a @xmath167 distribution skewed toward shorter timescales .
the slope reported here ( @xmath172 ) is apparently not quite steep enough , though correcting for binaries may steepen the slope by a few tenths .
moreover , it has been shown that many of the shorter events seen toward the bulge are `` amplification biased '' events of faint sources that are below the threshold of detection han ( 1997 ) .
these are mistaken for events of much brighter sources in the same seeing disk in which they are detected , and so the observed timescale for the period of significant apparent magnification is much shorter than the actual event timescale .
thus , the combination of our steeper imf and the amplification bias may well allow the bulge microlensing events to be explained by ordinary stars ( perhaps with a smooth extension into the brown dwarf regime ) .
it is not worth trying here to expand further on the implications of the bulge imf for the interpretation of microlensing experiments , given the very small number of published events compared to the huge number that will soon become available .
it will then be possible to make a detailed comparison between the observed timescale distribution from a large , very clean sample and that predicted on the basis of the imf reported here .
as for the @xmath141 ratio , the @xmath173 distribution will depend not only on the imf of still living stars , but also on the number and mass of the dead remnants , white dwarfs , neutron stars , and black holes .
such a distribution could provide constraints on the bulge imf at masses greater than the present turnoff ( @xmath174 ) , even to @xmath175 and beyond ( gould 1999 )
we have presented the results of stellar photometry on deep images obtained with nicmos on board of hst .
the data refer to a field in the galactic bulge , at a projected distance from the galactic center of @xmath176 . from the @xmath9-band lf of the stars in the field we derive the imf of the galactic bulge with the aid of a theoretical mass - luminosity relation which provides an excellent fit to the empirical mlr .
the imf so obtained refers to the mass range from @xmath177 down to @xmath178 , being therefore the deepest imf so far obtained for a galactic bulge .
nevertheless , this low - mass limit is still nearly a factor of @xmath179 above the hydrogen burning limit .
the imf is well fit by a single slope power law with @xmath180 , therefore much flatter than salpeter s imf with @xmath8 .
a two - slope imf with @xmath181 above @xmath182 and @xmath183 below gives a better fit , formally at the @xmath184 level .
however , in view of the larger error bars in the upper mass range , and the evolutionary effect away from the zero - age ms , we prefer to quote the single - slope power law as our main conclusion .
this result is robust within current uncertainties in the reddening , distance modulus of the galactic center , disk and binary stars contamination , and average metallicity of the bulge stars . for the mass range in common ( @xmath185 )
, the derived imf is in very good agreement with the bulge imf obtained from optical observations with wfpc2 by holtzman et al .
our bulge imf , however , is appreciably steeper than the low mass imf for the solar neighborhood found in two recent determinations , which give slopes of @xmath11 ( reid & gizis 1997 ) and @xmath12 ( gould et al .
however , the present bulge imf is virtually identical to yet other determinations of the solar neighborhood imf ( kroupa et al .
1993 , reid et al .
1999 ) , and an assessment as to whether bulge and disk imfs are the same or not will require an understanding the origin of the large discrepancies among the various determinations of the disk imf .
we have also compared the bulge imf with the present day mf of some galactic globular clusters with different metallicities and affected to various degrees by dynamical processes . in all clusters
the mf is flatter than that of the bulge , but it appears to be closer to the bulge imf in those clusters that are less affected by dynamical processes .
this suggests little or no dependence of the imf on metallicity for old systems .
one major issue concerns the amount of bulge mass that is locked in unseen dwarfs .
there is no hint for the imf slope to change towards the lower mass limit ( @xmath186 ) of the explored range . assuming the slope can be extrapolated all the way to mass zero , gives a total mass of brown dwarfs ( @xmath187 ) in the nic2 field of @xmath188 , i.e. , @xmath189 of the total stellar mass in the field ( c.f .
table [ tab_ml ] ) .
we note that some support for this extrapolation comes from the local density of l dwarfs ( reid et al .
1999 ) . with all stars making up to @xmath190 of the total baryonic mass of the universe
, this result suggests that brown dwarfs may represent not more than @xmath191 , i.e. a minor fraction , of the baryonic mass of the universe .
we are expecially grateful to claudia maraston for help in constructing table [ tab_ml ] .
we also thank peter stetson for providing us the software to calculate the stellar psf .
we finally thank the anonymous referee for the many constructive comments that have significantly improved the paper .
support for this work was provided by nasa through grant number go-7891 from the space telescope science institute , which is operated by the association of universities for research in astronomy , incorporated , under nasa contract nas5 - 26555 .
alcock , c. , et al .
1997 , apj , 479 , 119 baraffe , i. , chabrier , g. , allard , f. , & hauschildt , p. h. 1997 , a&a , 327 , 1054 bono , g. , caputo , f. , cassisi , s. , castellani , v. , & marconi , m. 1997 , apj , 479 , 279 bouvier , j. , stauffer , j. r. , martn , e. l. , barrado y navascus , d. , wallace , b. , & bjar , v. j. s. 1998 , a&a , 336 , 490 cardelli , j. a. , clayton , g. c. , & mathis , j. s. 1989 , apj , 345 , 245 cassisi , s. , et al . 1999 , in preparation dauphole , b. , geffert , m. , colin , j. , ducourant , c. , odenkirchen , m. , & tucholke , h .- j .
1996 , a&a , 313 , 119 de marchi , g. , leibundgut , b. , paresce , f. , & pulone , l. 1999 , a&a , 343 , l9 djorgovski , s. g. 1993 , in : asp conf.ser.50 , structure & dynamics of globular clusters , djorgovski s.g .
, meylan g. ( eds . ) , asp , san francisco , 373 frogel , j. a. , & whitford , a. e. 1987 , apj , 320 , 199 frogel , j. a. , terndrup , d. m. , blanco , v. m. , & whitford , a. e. 1990 , apj , 353 , 494 .
fugugita , m. , hogan , c. j. , & peebles , p. j. e. 1998 , apj , 503 , 518 gould , a. 1999 , apj , submitted , ( astro - ph/9906472 ) gould , a. , bahcall , j. n. , & flynn , c. 1997 , apj , 482 , 913 grillmair , c. j. , forbes , d. a. , brodie , j. p. , & elson , r. a. w. 1998 , aj , 117 , 167 han , c. 1997 , apj , 484 , 555 han , c. , & gould , a. 1996 , apj , 467 , 540 henry , t. j. , & mccarthy , d. w.jr .
1993 , aj , 106 , 773 hillenbrand , l. a. 1997 , aj , 113 , 1733 holtzman , j. a. , watson , a. m. , baum , w. a. , grillmair , c. j. , groth , e. j. , light , r. m. , lynds , r. , & oneil , e. j.jr .
1998 , aj , 115 , 1946 kent , s. m. 1992 , apj , 387 , 181 king , i. r. , & anderson , j. , cool , a. m. , & piotto , g. 1998 , apj , 492 , l37 kroupa , p. , tout , c. a. , & gilmore , g. 1993 , mnras , 262 , 545 larson , r. b. 1998 , mnras , 301 , 569 luhman , k. l. , rieke , g. h. , lada , c. j. , & lada , e. a. 1998 , apj , 508 , 347 maraston , c. 1998 , mnras , 300 , 872 mc william , a. , & rich , r. m. 1994 , apjs , 91 , 749 miller , g. e. , & scalo , j. m. 1979 , apjs , 41 , 513 ortolani , s. , renzini , a. , gilmozzi , r. , marconi , g. , barbuy , b. , bica , e. , & rich , r. m. 1995 , nature , 377 , 701 piotto , g. , cool a. m. , & king i. r. 1997 , aj , 113 , 1345 piotto , g. , & zoccali , m. 1999 , a&a , 345 , 485 pulone , l. , de marchi , g. , paresce , f. , & allard , f. 1998 , apj , 492 , l41 reid , i. n. , & gizis , j. e. 1997 , aj , 113 , 2246 reid , i. n. , et al . 1999 ,
apj , in press , ( astro - ph/9905170 ) reid , m. 1993 , ara&a , 31 , 345 renzini , a. , & ciotti , l. 1993 , apj , 416 , l49 renzini , a. 1998 , aj , 115 , 2459 renzini , a. 1999 , in : when and how do bulges form and evolve , carollo c. m. , ferguson , h. c. , wyse , r. f. g. ( eds . ) , cambridge university press , in press , ( astro - ph/9902108 ) rich , r. m. , ortolani , s. , bica , e. , & barbuy , b. 1998 , aj , 116 , 1295 saumon , d. , bergeron , p. , lunine , j. i. , hubbard , w. b. , & burrows , a. 1994 , apj , 424 , 333 scalo , j. 1998 , in : asp conf.ser.142 , the stellar initial mass function , gilmore g. , howell d. ( eds . ) , san francisco , 201 scalo , j. 1999 , in : the birth of galaxies , guiderdoni b. , et al .
( eds . ) , gif - sur - yvette , in press stetson , p. b. 1987 , pasp 99 , 191 stetson , p. b. 1994 , pasp 106 , 250 stephens , a. w. , frogel , j. a. , renzini , a. , ortolani , s. , jablonka , p. , rich , r. m. , & davies , r. 1999 , aj , submitted tiede , g. p. , frogel , j. a. , & terndrup , d. m. 1995 , aj , 110 , 2788 udalski , a. , szymanski , m. , stanek , k. z. , kaluzny , j. , kubiak , m. , mateo , m. , krzeminski , w. , paczynski , b. , & venkat , r. 1994 , acta astron . , 44 , 165 vesperini , e. , & heggie , d. c. 1997 , mnras , 289 , 898 whitford , a. e. 1978 , apj , 226 , 777 yan , l. , mc carthy , p. j. , storrie - lombardi , l. j. , & weymann , r. 1998 , apj , 503 , l19 zhao , h. s. , spergel , d. , & rich , r. m. 1995 , apj,440,l13 zoccali , m. , et al .
1999 , in preparation rrrrc 16.5 & 5 & 5 & 2 & 0.20 + 17.0 & 7 & 7 & 3 & 0.18 + 17.5 & 6 & 6 & 2 & 0.18 + 18.0 & 20 & 18 & 4 & 0.14 + 18.5 & 18 & 19 & 4 & 0.13 + 19.0 & 25 & 26 & 5 & 0.13 + 19.5 & 32 & 32 & 6 & 0.13 + 20.0 & 38 & 40 & 6 & 0.13 + 20.5 & 45 & 50 & 10 & 0.12 + 21.0 & 54 & 60 & 10 & 0.12 + 21.5 & 69 & 82 & 15 & 0.11 + 22.0 & 66 & 81 & 14 & 0.11 + 22.5 & 86 & 112 & 19 & 0.10 + 23.0 & 81 & 119 & 22 & 0.10 + 23.5 & 73 & 119 & 21 & 0.10 + 24.0 & 65 & 139 & 31 & 0.09 + 24.5 & 57 & 162 & 27 & 0.08 + rrrc 8.750 & 0.010 & 0.002 & 0.00 + 9.250 & 0.009 & 0.003 & 0.00 + 9.750 & 0.009 & 0.003 & 0.00 + 10.250 & 0.026 & 0.004 & 0.00 + 10.750 & 0.033 & 0.004 & 0.00 + 11.250 & 0.052 & 0.005 & 0.00 + 11.750 & 0.032 & 0.005 & 0.00 + 12.250 & 0.112 & 0.116 & 0.00 + 12.750 & 0.112 & 0.113 & 0.05 + 13.250 & 0.661 & 0.268 & 0.10 + 13.750 & 0.447 & 0.223 & 0.08 + 14.250 & 0.661 & 0.263 & 0.13 + 14.750 & 0.661 & 0.246 & 0.24 + 15.250 & 1.000 & 0.290 & 0.30 + 15.750 & 1.230 & 0.310 & 0.35 + 16.250 & 1.660 & 0.343 & 0.41 + 16.750 & 2.754 & 0.464 & 0.35
+ cccccccc @xmath192 & @xmath192 & @xmath192 & 3207 & 326 & 247 & 2213 & 4.7 + @xmath192 & @xmath193 & @xmath193 & 744 & 150 & 38 & 135 & 1.1 + @xmath192 & @xmath194 & @xmath194 & 601 & 133 & 15 & 32 & 0.9 + @xmath192 & @xmath194 & @xmath195 & 562 & 124 & 8 & 10 & 0.8 + | we present a luminosity function ( lf ) for lower main sequence stars in the galactic bulge near @xmath0 to @xmath1 , corresponding to @xmath2 .
this lf is derived from hubble space telescope ( hst ) + near infrared camera and multi object spectrometer ( nicmos ) observations of a region of @xmath3 , with the f110w and f160w filters .
the main sequence locus in the infrared shows a strong change in slope at @xmath4 ( @xmath5 which is well fit by new low mass models that include water and molecular hydrogen opacity .
our derived mass function ( which is not corrected for binary companions ) is the deepest measured to date in the bulge , and extends to 0.15 @xmath6 with a power law slope of @xmath7 ; a salpeter mass function would have @xmath8 .
we also combine our @xmath9 band lf with previously published data for the evolved stars to produce a bulge lf spanning @xmath10 15 magnitudes .
we show that this mass function has negligible dependence on the adopted bulge metallicity and distance modulus . although shallower than the salpeter slope , the slope of the bulge imf is steeper than that recently found for the galactic disk ( @xmath11 and @xmath12 from the data of reid & gizis , 1997 , and gould et al .
1997 , respectively , in the same mass interval ) , but is virtually identical to the disk imf derived by kroupa et al .
( 1993 ) .
the bulge imf is also quite similar to the mass functions derived for those globular clusters which are believed to have experienced little or no dynamical evolution .
finally , we derive the @xmath13 ratio of the bulge to be @xmath14 @xmath15 , and briefly discuss the implications of this bulge imf for the interpretation of the microlensing events observed in the direction of the galactic bulge .
= 11 versim#1#22.9truept = 12 |
the data on the glitch healing parameter , @xmath0 , provide the best tool for testing the starquake ( ruderman 1972 ; alpar et al 1996 ) and vortex unpinning ( alpar et al 1993 ) models of glitch generation in pulsars .
both of these mechanisms of glitch generation , in fact , consider nss , in general , a two component structure : a superfluid interior core surrounded by a rigid crust ( in the present study we shall use the term ` envelope ' which includes the solid crust and other interior portion of the star right up to the superfluid core ) . in the starquake model , @xmath0 is defined as the fractional moment of inertia , i.e. the ratio of the moment of inertia of the superfluid core , @xmath21 , to the moment of inertia of the entire configuration , @xmath22 , as ( pines et al 1974 ) @xmath23 recently , crawford & demiaski ( 2003 ) have collected the all measured values of the glitch healing parameter @xmath0 for crab and vela pulsars available in the literature and found that for 21 measured values of @xmath0 for crab glitches , a weighted mean of the values yields @xmath24 , and the range of @xmath25 encompasses the observed distribution for the crab pulsar . in order to test the starquake model for the crab pulsar ,
they have computed @xmath0 ( as given by eq.(1 ) ) values for seven representative eoss of dense nuclear matter , covering a range of neutron star masses .
their study shows that the much larger values of @xmath26 for the crab pulsar is fulfilled by all the six eoss ( out of seven considered in the study ) corresponding to a ` realistic ' neutron star mass range @xmath27 .
by contrast , a weighted mean value of the 11 measurements for vela yields a much smaller value of @xmath28 and the all estimates for vela agree with the likely range of @xmath2 .
thus , their results are found to be consistent with the starquake model predictions for the crab pulsar .
they have also concluded that the much smaller values of @xmath2 for the vela pulsar are inconsistent with the starquake model predictions , since the implied vela mass based upon their models corresponds to a value @xmath29 for @xmath2 , which is too low as compared to the ` realistic ' ns mass range .
thus , in the literature , the starquake is considered as a viable mechanism for glitch generation in the crab and the vortex unpinning , the another mechanism , is considered suitable for the vela pulsar , since it can avoid some other problems associated with the starquake explanation of the vela glitches ( see , e.g. crawford & demiaski ( 2003 ) ; and references therein ) .
however , it seems surprising that if the internal structure of nss are described by the same two component conventional models ( as mentioned above ) , different kinds of glitch mechanisms are required for the explanation of a glitch !
furthermore , it also follows from the above discussion that the main reason for not considering the starquake , the feasible mechanism for glitch generation in the vela , lies in the fact that there exists none of the sequence of ns models in the literature which could explain simultaneously , on the basis of starquake model , both the extreme limiting cases of glitch healing parameter , @xmath0 , corresponding to the vela ( @xmath2 ) and the crab ( @xmath1 ) pulsars in the range @xmath30 for the ` realistic ' ns mass values for both the pulsars . the present study , therefore ,
deals with the construction of such models we assume that _ all _ the nss belong to the same family of ns sequence which terminates at the _ maximum _ value of mass .
certainly , this _ maxima _ should correspond to an _ upper bound _ on ns masses . in order to construct such a sequence
, we have to set the extreme causal eos ( in geometrized units ) , @xmath31 ( where @xmath32 is the pressure and @xmath33 the energy - density ) to describe the core .
firstly , because various observational studies like - the gamma - ray burst data , x - ray burst data and the glitch data etc . , and their explanation ( see , e.g. lindblom 1984 ; cottam et al 2002 ; datta & alpar 1993 ) favour the stiffest eoss .
the latest estimate of the moment of inertia for the crab pulsar ( based upon the ` newest ' observational data on the crab nebula mass ) rules out most of the existing eoss of the dense nuclear matter , leaving only the stiffest ones ( bejger & haensel 2002 ; haensel et al 2006 ) . secondly ,
because of the fact that the ` real ' eos of the dense nuclear matter beyond the density range @xmath34 are largely unknown due to the lack of knowledge of nuclear interactions ( see , e.g. dolan 1992 ; and references therein ; haensel et al 2006 ) , and the various eoss available in the literature ( see , e.g. arnett & bowers 1977 ) for ns matter represent only an extrapolation of the results far beyond this density range .
though , the status of the ` real ' eos for ns matter is not certain , one could impose some well - known physical principle , independent of the eos , such as the ` causality condition ' ( @xmath31 ) throughout the core of the star beyond a fiduciary density , @xmath35 , at the core - envelope boundary to ascertain a definite upper bound on ns masses ( see , e.g. , rhoades & ruffini 1974 ; hartle 1978 ; lindblom 1984 ; friedman & ipser 1987 ; kalogera & baym 1996 ) . in this connection
this is also to be pointed out here that the maximum mass for _ any _ eos describing the core , beyond the density @xmath35 , with a subluminal sound velocity turns out to be less than that of the upper bound obtained by using the extreme causal eos ( see , e.g. , haensel et al 2006 ) .
the envelope of our models ( below the density @xmath35 at the core - envelope boundary ) may be characterized by the well - known eos of classical polytrope @xmath36ln@xmath37ln@xmath38 ( where @xmath39 denotes the density of the rest - mass and @xmath40 is a constant known as the adiabatic index ) for different values of the constant @xmath41 and 2 respectively .
the reason for considering the polytropic eos for the entire envelope lies in the fact that with this eos , our models yield an upper bound on ns masses _ independent _ of the value of @xmath40 , and this upper bound ( for a fiduciary choice of @xmath35 ) is found fully consistent with those of the values cited in the literature ( kalogera & baym 1996 ; friedman & ipser 1987 ) .
thus , the choice of the said polytropic eos for the entire envelope may be regarded entirely equivalent to the choice of the various eoss like wff ( wiringa , fiks & fabrocini 1988 ) , fps ( lorenz , ravenhall & pethick , 1993 ) , nv ( negele & vautherin 1973 ) , or bps ( baym , pethick & sutherland 1971 ) in an appropriate sequence below the density range @xmath35 , adopted by various authors in the conventional models of nss ( see , e.g. , kalogera & baym 1996 ; friedman & ipser 1987 ) , so that the constant @xmath40 appearing in the polytropic eos may be looked upon as an ` average ' @xmath40 for the density range below @xmath35 , specified by the sequence of various eoss in the conventional models of nss .
the choice of the constant @xmath42 and 2 thus become obvious , since this choice can cover almost the entire range of density discussed in the literature for ns matter which is also applicable for the envelope region - the polytropic eos with @xmath43 represents the eos of extreme relativistic degenerate electrons and non - relativistic nuclei ( chandrasekhar 1935 ) , @xmath44 represents the well - known eos of non - relativistic degenerate ` neutron gas ' ( oppenheimer & volkoff 1939 ) , and @xmath45 represents the case of extreme relativistic baryons interacting through a vector meson field ( zeldovich 1962 ) ( the value of @xmath46 is also possible for some eos describing the ns matter , e.g. , malone , johnson & bethe 1975 ; clark , heintzmann & grewing 1971 , however , the results obtained in this paper remain unaffected for the choice of @xmath46 ) , and the outcome of this study ( in terms of explaining the glitch healing parameter for various pulsars and predicting the upper bound on the compactness of nss ( since the upper bound on mass is independent of the value of @xmath40 ) ) would finally decide , among the chosen values , the ` appropriate ' value of @xmath40 for the ns envelope .
the validity of assuming the extreme causal eos in the core and a polytropic eos in the envelope of the present models , in view of the various modern eos of dense nuclear matter , is also discussed in the last section of the present paper .
we have noted that in all conservative models of nss , the choice of the core - envelope boundary , @xmath47 ( corresponding to a density denoted by @xmath35 ) , is somewhat _ arbitrary _ in the sense that there are no criteria available for the choice of a particular matching density , @xmath35 , below which the eos of the ns matter is assumed to be known and unique .
one can freely choose somewhat lower values of @xmath35 ( which will increase the core size ) to obtain higher values of @xmath0 ( see , e.g. shapiro & teukolsky 1983 ; datta & alpar 1993 ) .
to avoid such a procedure , we choose the core - envelope boundary of our models on the basis of the ` compatibility criterion ' which asserts that for an assigned value of the ratio ( @xmath48 ) of central pressure , @xmath49 , to central energy - density , @xmath50 , the compactness parameter @xmath51 ; total mass to radius ratio in geometrized units ) of any _ regular _ configuration should not exceed the compactness parameter @xmath52 of the homogeneous density sphere , in order to assure the compatibility with the hydrostatic equilibrium ( negi & durgapal 2001 ; negi 2004a ) .
this criterion is capable of constraining the core - envelope boundary of any physically realistic ns model . a combination of this criterion with those of the observational data on the glitch healing parameter and the recently estimated minimum value of the moment of inertia for the crab pulsar ( based on the newly estimated ` central value ' of the crab nebula mass @xmath53 in the time - dependent acceleration model ) , @xmath54 ;
where @xmath55 g@xmath56 ( bejger & haensel 2003 ) can provide the desired ns models discussed above , since both the theory ( criterion ) and the observations ( stated above ) are being used to construct the ns models .
since we are using geometrized units ( i.e. @xmath57 , where @xmath58 represent the universal constant of gravitation and @xmath59 is the speed of light in vacuum ) , the metric for spherically symmetric and static configurations can be written in the following form @xmath60 where @xmath61 and @xmath62 are functions of @xmath63 alone .
the oppenheimer - volkoff ( o - v ) equations ( oppenheimer & volkoff 1939 ) , resulting from the einstein field equations for systems with isotropic pressure @xmath32 and energy - density @xmath33 can be written as @xmath64/r(r - 2 m ) \\
\nu'/2 & = & - p'/(p + e ) \\
m'(r ) & = & 4\pi e r^2 \,;\end{aligned}\ ] ] where the prime denotes the radial derivative and @xmath65 represents the mass contained within the radius @xmath63 @xmath66 the coupled eqs.(3 - 5 ) are solved for the model ( supplemented by the boundary conditions : @xmath67 , @xmath68 , @xmath69 at @xmath70 ) by considering the eos , @xmath31 , in the core and choosing various values of @xmath40 in the polytropic envelope for various assigned values of @xmath48 such that for each value of @xmath48 , the compactness ratio of the whole configuration always turns out to be less than or equal to the compactness ratio of the corresponding sphere ( with the same @xmath48 ) of the homogeneous density distribution and should not exceed the _ exact _
absolute upper bound on compactness ratio of nss compatible with causality and pulsational stability ( see , e.g. negi 2004b ) .
the fulfillment of both of the conditions stated in the last sentence is usually called the ` appropriate ' fulfillment of ` compatibility criterion ' .
we find that this condition is uniquely fulfilled by all the models corresponding to an envelope with @xmath41 and 2 respectively , if the _
minimum _ value of the ratio of pressure to energy - density , @xmath71 , at the core - envelope boundary reaches about @xmath72 .
the results of this study are presented in tables 1 - 2 and fig.1 for the value of matching density , @xmath73 , i.e. about 3.55 times the nuclear saturation density ( note that the particular choice of @xmath35 used here turns out to be a consequence of the constraints ( i ) and ( ii ) mentioned in the abstract of the present study , and therefore , does not represent a fiduciary quantity as discussed in the next section ) .
it is seen that the models become pulsationally stable up to the maximum value of mass @xmath74 and radius , @xmath75 .
the minimum radius results for the model with a @xmath45 envelope thus maximizes the compactness ratio for the stable configuration , @xmath76 , as shown in tables 1 - 2 .
this behaviour indicates that among various values of @xmath40 chosen in the envelope for the density range right from 3.55 times the nuclear saturation density up to zero at the surface , the ` average ' value of @xmath40 is appropriately described by a polytropic index @xmath77 .
this fact will also follow from some special features of the model with @xmath45 envelope discussed in the next section .
the binding energy per unit mass @xmath78 ; where @xmath79 is the rest - mass ( see , e.g. shapiro & teukolsky 1983 ) ] also approaches a maximum for about 0.3201 for the maximum value of mass up to which the configurations remain pulsationally stable .
however , for the @xmath80 envelope model the binding energy reaches a maximum beyond the maximum value of mass .
for slowly rotating configurations like the crab and the vela pulsars ( rotation velocity about 188 and 70 rad sec@xmath81 respectively ) the moment of inertia may be calculated in the first order approximation that appears in the form of the lense - thirring frame dragging - effect . for crab and vela pulsars ,
the first - order effects turn out to be about 1 - 2% ( other effects like mass shift and deformation from spherical symmetry due to rotation represent second - order effects which are significant for the case of millisecond pulsars . for crab and vela like pulsars ,
the second - order effects turn out to be about @xmath82 or even lower ; see , e.g. arnett & bowers 1977 ; crawford & demiaski 2003 .
therefore , these effects can be safely ignored when studying the macroscopic parameters of the slowly rotating pulsars as carried out in the present paper ) .
these effects are reproduced by an empirical formula that is based on the numerical results obtained for thirty theoretical eoss of dense nuclear matter . for nss , the formula yields ( bejger & haensel 2002 ) @xmath83 where @xmath84 is the compactness ratio measured in units of @xmath85 , i.e. @xmath86 only static ( non - rotating ) parameters of the spherical configuration appear in the formula .
equation ( 6 ) is used , together with coupled eqs.(3 - 5 ) , to calculate the fractional moment of inertia given by eq.(1 ) and the moment of inertia of the entire configuration for an assigned value of the matching density , @xmath35 .
the results of the calculations are presented in table 3 and figs . 2 - 3 .
it is seen from table 3 that among all models corresponding to different values of @xmath40 in the envelope , the lower limits of both the observational constraints for the crab pulsar , @xmath1 and @xmath87 , are appropriately satisfied by the @xmath45 envelope model . since this particular model yields a unique value of matching density , @xmath88 g @xmath89 , for the last two constraints in such a manner that the constraint @xmath90 is appropriately fulfilled by all stable ns models corresponding to the @xmath44 and 2 envelopes for another constraint @xmath1 as shown in table 3 and figs.2 - 3 respectively ( the model with a @xmath80 envelope
is ruled out since it always yields stable models with @xmath91 ) .
another special feature of the @xmath45 envelope model lies in the fact that it yields somewhat wider range of glitch healing parameter , @xmath92 , for the stable models as compared to the @xmath44 envelope model which yields this parameter in the range @xmath93 for stable models as shown in fig.2 and tables 2 - 3 .
furthermore , the @xmath45 envelope model yields somewhat higher value of maximum surface redshift , @xmath94 , for the vela pulsar ( @xmath13 ) and somewhat lower value of minimum surface redshift , @xmath7 , for the crab pulsar ( @xmath95 ) simultaneously , as compared to the @xmath44 envelope model as shown in tables 2 - 3 .
thus provides somewhat wider range of lower and upper bounds on the energy of a gravitationally redshifted radiation from the surface of the vela and the crab pulsars respectively . and above all
, the compactness maximizes , @xmath76 , for the @xmath45 envelope model in the stable sequence of ns models as shown in tables 1 - 2 and fig.1 .
this special feature of the model is also discussed in section 2 .
the above - mentioned features indicate the appropriateness of the @xmath45 envelope model and thus emerges an important consequence of this study which states that the envelope of ` real ' nss may be well approximated by an eos of classical polytrope with ` average ' value of polytropic index , @xmath96 , closer to 1 . for the minimum value of @xmath97 , table 3 yields @xmath98 for the @xmath45 envelope model and from table 2 , on the basis of @xmath45 envelope model
, we obtain the minimum values of mass , @xmath8 , and surface redshift , @xmath99 , for the crab pulsar .
on the other hand , for the maximum value of @xmath13 which belongs to the vela pulsar , table 3 and fig.3 yield @xmath100 corresponding to the @xmath45 envelope model and table 2 and fig.2 yield the maximum values of mass , @xmath101 , and surface redshift , @xmath94 , for the vela pulsar ( note that the upper weighted mean value of @xmath102 for the vela pulsar , table 2 and fig.2(for the @xmath45 envelope model ) yield the maximum value of mass , @xmath103 , and surface redshift , @xmath104 , for the vela pulsar .
obviously , these values are much closer to the values corresponding to the maximum value of @xmath105 for the vela pulsar ) .
however , for the observational constraint of the ` central ' weighted mean values of @xmath106 for the crab and @xmath18 for the vela pulsar , table 2 and fig.2 ( for the @xmath45 envelope model ) yield the slightly increased values of mass , @xmath17 , and the surface redshift , @xmath16 , for the crab and somewhat decreased values of mass , @xmath107 , and surface redshift , @xmath108 , for the vela pulsar respectively . for these central weighted mean values of @xmath0 , tables 2 - 3 and fig.3 yield for the @xmath45 envelope model , the moment of inertia @xmath109 for the crab and @xmath110 for the vela pulsar respectively . for the lower weighted mean value of @xmath111 corresponding to the vela pulsar , tables 2 - 3 and fig.2 yield ( for the @xmath45 envelope model ) the minimum values of mass , @xmath112 , and surface redshift , @xmath113 , whereas table 3 and fig.3 yield a minimum value of moment of inertia , @xmath114 , for the vela pulsar .
apart from the models of the crab and vela pulsars , let us consider two of the main findings related to the extraordinary gamma - ray burst of 5 march 1979 : ( i ) it gives the only reliable estimate of the surface redshift , @xmath115 ( after taking due account of thermal blueshift ) , associated with the supernova remnant n49 in the large magellanic cloud ( see , e.g. higdon & lingenfelder 1990 ; douchin & haensel 2001 ; and references therein ) , and ( ii ) the implied peak luminosities of the repeating burst grb790305b correspond to an energy of @xmath116 ergs , which is possible only when a starquake releases at least @xmath117 of the ns gravitational binding - energy of @xmath118 ergs ( higdon & lingenfelder 1990 ) .
in order to reproduce both of these findings , one would require a ns model compatible with starquake model predictions that could also account for a surface redshift @xmath119 .
both of these requirements are , in fact , fulfilled by our models . for the case of the @xmath45 envelope model
, we get from tables 2- 3 and fig . 2 - 3 the central value of surface redshift @xmath120 for a mass @xmath121 with @xmath122 .
the binding energy corresponding to this case is obtained as @xmath123 ergs which is capable of releasing @xmath124 ergs of energy required for the latter burst .
this study constructs the stable sequences of ns models terminate at the value of maximum mass , @xmath125 , independent of the eoss of the envelope , for the matching density , @xmath88 g @xmath89 , at the core - envelope boundary .
this value of ` matching density ' is a consequence of the observational constraints @xmath97 and @xmath126 ( associated with the crab pulsar ) imposed together on the @xmath45 envelope model and in this sense does not represent a fiduciary quantity .
the upper bound of the surface redshift , @xmath127 ( corresponding to a @xmath128 value @xmath129 ) , however , belongs to the model with a @xmath45 envelope which is consistent with the absolute upper bound on the surface redshift of ns models compatible with causality and pulsational stability ( negi 2004b ) .
this special feature , together with some other remarkable ones , discussed in the present study underline the appropriateness of the @xmath45 envelope model . since among the variety of modern eoss discussed in the literature , the upper bound on ns mass compatible with causality and dynamical stability can reach a value up to @xmath130 ( in this category , the sly ( douchin & haensel 2001 ) eos yields a maximum mass of @xmath131 , whereas the bgn1 ( balberg & gal 1997 ) and the apr ( akmal et .
al . 1998 ) eoss yield the maximum masses of @xmath132 and @xmath133 respectively ( see , e.g. haensel et al 2006 ) ) . in view of this result ,
the model - independent maximum mass , @xmath134 , obtained in this study may be regarded as good as those obtained on the basis of modern nuclear theory .
in addition to this result , the appropriate sequences of stable ns models obtained in this study can explain the glitch healing parameter , @xmath0 , of any glitching pulsar , provided the weighted mean values of @xmath0 lie in the range @xmath135 .
this finding also reveals that if the starquake is considered to be a viable mechanism for glitch generation in all pulsars , then the envelope of ` real ' nss may be well approximated by a polytropic eos corresponding to a polytropic index , @xmath96 , closer to 1 .
for the value of matching density , @xmath88 g @xmath89 , the @xmath45 envelope model yields the minimum values of mass @xmath136 and surface redshift @xmath7 for the crab ( @xmath97 ) and the maximum values of mass @xmath101 and surface redshift @xmath94 for the vela pulsar ( @xmath137 ) .
the minimum mass and surface redshift for the crab pulsar are slightly increased up to the values @xmath17 and @xmath16 respectively , if the ` central ' weighted mean value of @xmath106 and the moment of inertia @xmath138 are also imposed on these models . however , for the ` central ' weighted mean value of @xmath139 corresponding to the vela pulsar , the maximum mass and surface redshift are somewhat decreased to the values @xmath140 and @xmath108 respectively .
this value of mass and surface redshift for the vela pulsar can further decrease up to the values @xmath112 and @xmath113 respectively , if the lower weighted mean value of @xmath141 for the vela pulsar is imposed .
these results predict the upper and lower bounds on the energy of a gravitationally redshifted electron - positron annihilation line in the range of about 0.309 - 0.315 mev from the crab and in the range of about 0.425 - 0.439 mev from the vela pulsar respectively . for a comparison , if the observational constraint of the minimum value of @xmath142 ( the value of moment of inertia for the crab pulsar obtained earlier by bejger & haensel ( 2002 ) , on the basis of the constant - acceleration model for the crab nebula ) together with @xmath97 is imposed on the models studied in the present paper , the @xmath45 envelope model yields the value of matching density , @xmath143 g @xmath89 .
this value of @xmath35 yields a model - independent upper bound on ns mass @xmath144 .
this value of maximum mass , however , represents an ` average ' of the maximum ns masses in the range @xmath145 obtained by kalogera & baym 1996 ( and references therein ) on the basis of other eoss for ns matter , fitted to experimental nucleon - nucleon scattering data and the properties of light nuclei . for this lower value of matching density ,
the @xmath45 envelope models yield the minimum value of mass @xmath146 for the crab ( @xmath97 ) and the maximum value of mass @xmath147 for the vela pulsar ( @xmath13 ) .
the minimum mass for the crab pulsar is slightly increased up to the value @xmath148 , if the ` central ' weighted mean value of @xmath106 and the moment of inertia @xmath149 are also imposed on these models .
however , corresponding to the ` central ' weighted mean value of @xmath139 , the maximum mass of the vela pulsar is somewhat decreased to the value @xmath150 .
this value of mass for the vela pulsar can further decrease up to the value @xmath151 , if the lower weighted mean value of @xmath141 for the vela pulsar is imposed .
furthermore , the study can also explain some special features associated with the extraordinary gamma - ray burst of 5 march 1979 .
the author acknowledges the aryabhatta research institute of observational sciences ( aries ) , nainital for providing library and computer - centre facilities .
g @xmath89 at the core - envelope boundary .
the labels a , b and c represent the models for an envelope with @xmath152 and 2 respectively .
the minimum value of the ratio of pressure to energy - density , @xmath153 , at the core envelope boundary is obtained as @xmath72 , such that for an assigned value of @xmath154 , the inequality @xmath155 is always satisfied for _ all _ the models as shown in table 1 .
, width=359,height=359 ] vs. total mass @xmath156 ( in units of @xmath157 ) for the configurations presented in tables 2 - 3 and fig.1 . the labels a
, b and c represent the models for an envelope with @xmath152 and 2 respectively .
, width=359,height=359 ] shapiro , s. l. , & teukolsky , s. a. 1983 , black holes , white dwarfs , and neutron stars : the physics of compact objects ( new york : wiley ) wiringa , r. b. , fiks , v. , & fabrocini , a. 1988 , phys .
c 38 , 1010 vs. moment of inertia of the entire structure @xmath158 for the configurations presented in tables 2 - 3 and figs.1 - 2 . the labels a , b and c represent the models for an envelope with @xmath152 and 2 respectively .
@xmath158 is defined as @xmath159 .
, width=359,height=359 ] | we construct for the first time , the sequences of stable neutron star ( ns ) models capable of explaining simultaneously , the glitch healing parameters , @xmath0 , of both the pulsars , the crab ( @xmath1 ) and the vela ( @xmath2 ) , on the basis of starquake mechanism of glitch generation , whereas the conventional ns models can not give such consistent explanation .
furthermore , our models also yield an upper bound on ns masses similar to those obtained in the literature for a variety of modern equations of state ( eoss ) compatible with causality and dynamical stability . if the lower limit of the observational constraint of ( i ) @xmath1 for the crab pulsar and ( ii ) the recent value of the moment of inertia for the crab pulsar ( evaluated on the basis of time - dependent acceleration model of the crab nebula ) , @xmath3 ( where @xmath4 ) , both are imposed together on our models , the models yield the value of matching density , @xmath5 at the core - envelope boundary .
this value of matching density yields a model - independent upper bound on neutron star masses , @xmath6 , and the strong lower bounds on surface redshift @xmath7 and mass @xmath8 for the crab ( @xmath9 ) and the strong upper bound on surface redshift @xmath10 , mass @xmath11 and the moment of inertia @xmath12 for the vela ( @xmath13 ) pulsar .
however , for the observational constraint of the ` central ' weighted mean value @xmath14 , and @xmath15 , for the crab pulsar , the minimum surface redshift and mass of the crab pulsar are slightly increased to the values @xmath16 and @xmath17 respectively , whereas corresponding to the ` central ' weighted mean value @xmath18 for the vela pulsar , the maximum surface redshift , mass and the moment of inertia for the vela pulsar are slightly decreased to the values @xmath19 and @xmath20 respectively .
these results set an upper and lower bound on the energy of a gravitationally redshifted electron - positron annihilation line in the range of about 0.309 - 0.315 mev from the crab and in the range of about 0.425 - 0.439 mev from the vela pulsar . |
the unprecedent convergence of observational results that we are currently witnessing has narrowed down the region of the cosmological parameter space which is compatible with all the different and independent probes of cosmology : supernovae @xcite @xcite , cmb @xcite and large scale structures @xcite .
robustly growing evidence suggests that _
i ) _ ordinary matter is a minority ( @xmath0 ) of all the matter content of the universe , _ ii ) _ matter mostly dark , non - baryonic matter is a minority ( @xmath1 ) of all the cosmological mass energy density , mostly contributed by an obscure form of energy referred to as ` dark energy ' , _ iii ) _ the 3d spatial geometry of the universe is flat and _ iv ) _ the expansion of the cosmic metric has been accelerating for the last @xmath2 gyrs of our universe lifetime . even though the big picture is in place , the two dominant contributions to the stress - energy tensor i.e. dark energy and dark matter
still have a hypothetical nature and they have not been discovered yet . while there is widespread evidence for the existence of the non - baryonic dark matter component producing the potential wells of large - scale structures @xcite , as yet no persuasive theoretical explanation has been able to elucidate the physical nature of the dark energy component @xcite . as a matter of fact , unveiling the nature of dark energy and its role in cosmology and gravitation is a difficult and subtle challenge . in such a context , it should not be overlooked that the large roaming from model to model , and the abundance of weakly predictive theories , might eventually limit the possibility of discriminating between different competitors proposed so far for explaining the observed dynamics of the accelerating universe . in the absence of a compelling theoretical explanation for the dark energy component , and in a minimal , zero - order approach , we explore here the possibility of preserving the standard metric interpretation of the accelerated expansion against possible alternative physical scenarios . to this end ,
we first evaluate and then discuss the observable consequences of local , non - gravitational mechanisms which could in principle accelerate matter in our hubble patch of the universe .
we assume here that the universe is described by general relativity , that it is dominated by components which satisfy the usual energy conditions ( according to which the universe can only decelerate ) and that the onset of recent accelerated expansion is the result of the presence of a hypothetical non - gravitational force field . such an alternative explanation is rather conservative , since it assumes neither a cosmological constant ( or negative - pressure fluid ) nor a modification of general relativity .
accordingly , we first work out a self - consistent , non - geometric model for the cosmic acceleration that is able to reproduce the current observations of standard candles ( i.e. snia ) and then we discuss a falsifiability procedure aimed at testing its observational predictions . the motivation behind this work is to put strong limits on a hypothetical ( or non usually considered ) physics that is possibly missing in our picture of the universe , and , in turn , to strengthen the evidence supporting the standard paradigm with which we are currently explaining its past history , its present stage and its future fate .
our goal is to work out testable predictions that allow us to reject non - metric acceleration models . to this end , we construct here a general , phenomenological model in which the role of dark energy is mimicked by an alternative mechanism of non - gravitational origin .
we consider here a background universe with a metric expansion as predicted by general relativity .
we assume that in such a universe a hypothetical large
scale , non - gravitational force field influences the overall dynamics of large scale structures in a patch of the universe with typical dimensions of the local hubble volume . in this scenario
every object which at time @xmath3 sits on the shell of a sphere of proper radius @xmath4 centered on the observer feels a peculiar acceleration field @xmath5 that is radially directed and time dependent .
+ we further speculate that the only cosmological component contributing to the stress - energy tensor is dark matter . in other words , we assume that there is no dark energy at all in the universe and that what we interpret as apparent isotropic acceleration of the metric is indeed the combination of the decelerated cosmological expansion predicted by general relativity plus the doppler effect sourced by matter which is accelerating outward under the effect of the radial force field we have added to this cosmological scenario .
in such a case , by appropriately tuning the non - cosmological contribution to the observed redshift , one can reconstruct a functional form for the luminosity distance @xmath6 ( see eq .
[ eq.dlgeneral ] below ) which reproduces the standard one derived within a model with cosmological constant ( or dark energy ) : i.e. @xmath7 in other words , the standard @xmath8cdm cosmological parameter set ( represented by the vector @xmath9 ) observationally inferred by simply plugging in the observed redshift @xmath10 into the luminosity distance formula are biased with respect to the _ true cosmological _ values ( represented by the vector @xmath11 ) that would be naturally inferred by recognizing the presence of a physical mechanism responsible for the acceleration of matter _ in situ_. having outlined the phenomenological model we are pursuing , let us now elucidate its finer details , i.e. the dependence of the luminosity distance ( eq .
1 ) on the peculiar acceleration term @xmath12 . +
the luminosity distance of an object standing at the observed redshift @xmath10 is given , in a pure matter scenario , by @xmath13 \ ; , \label{eq.dlgeneral}\ ] ] where @xmath14 , @xmath15 ( if @xmath16 ) , @xmath17 ( if @xmath18 ) , @xmath19 ( if @xmath20 ) , and @xmath21^{1/2}$ ] .
+ while in the standard model the measured redshift @xmath10 has a pure cosmological interpretation , in the accelerated model it additionally includes the contribution of the peculiar velocity @xmath22 induced by the radial non gravitational field and it can be written as @xmath23 clearly , there are an infinite variety of cosmological models that , when combined with an equally arbitrary variety of radial acceleration fields @xmath12 , could in principle reproduce the observational features of the standard cosmological model .
so , for the sake of simplicity , and in order to capture the essential physics of the problem , we assume in the following that the universe is flat , that its total density is contributed only by matter @xmath24 , and its expansion rate is @xmath25 .
+ under these hypotheses , the luminosity distance can be written as @xmath26^{-\frac{1}{2}}\big\ } \ ; , \label{eq.dl_om1}\ ] ] nonetheless , we stress that these working assumptions do not influence the generality of the conclusions presented in @xmath27 : analogous considerations hold , in fact , for a low density universe , open universe ( with , e.g. , @xmath28 ) .
we can now compute how a peculiar velocity @xmath22 in this scenario depends on the peculiar acceleration field : @xmath29 $ ] at a given cosmological redshift @xmath30 .
+ using the expression of the proper distance to a given coordinate , @xmath31 , we obtain the expression of the acceleration @xmath32 by means of which a perturbed metric expands radially : @xmath33 where @xmath34 and the peculiar acceleration term is given by @xmath35 after transforming this equation from the time to the redshift domain , we find the general solution @xmath36 we model the peculiar acceleration term in this equation by means of a general , non - divergent polynomial function @xmath37 where the @xmath38 are free parameters to be adjusted with the constraint given by eq .
1 . note that in the absence of a peculiar acceleration acting on cosmic matter , we recover the standard result that any primordial peculiar velocity is damped by the background expansion ( i.e. , @xmath39 ) . we show here that by appropriately tuning the parameters @xmath38 we can match the luminosity distance @xmath6 of standard candles like snia as inferred within the standard model of cosmology .
+ to this end , we first set the integration constant @xmath40 by imposing that , in the limit @xmath41 , the peculiar velocity @xmath42 .
we also set @xmath43 in order to avoid any primordial acceleration field .
then , we fit eq .
[ eq.dlgeneral ] to reproduce the currently available supernovae data .
we find that even a flat , matter dominated ( @xmath44 ) cosmological model , when supplied with an appropriate local acceleration field , would be able to reproduce the luminosity distance derived by fitting the observed data assuming ( _ wrongly _ ) a cosmological origin of the source redshift ( see fig.[fig2 ] ) .
+ we obtain an acceptable fit ( @xmath45 ) by expanding up to order @xmath46 in eq .
the best fitting coefficients in the redshift range @xmath47 ( for a background cosmological model with @xmath44 ) are : @xmath48m / s@xmath49 , @xmath50m / s@xmath49 , @xmath51m / s@xmath49 , @xmath52m / s@xmath49 . _
( left ) _ and peculiar velocity @xmath22 _ ( right ) _ needed to reproduce current snia data ( in the redshift range @xmath47 ) are shown as a function of redshift in a flat , decelerating @xmath44 background cosmology.,title="fig:",width=283 ] _ ( left ) _ and peculiar velocity @xmath22 _ ( right ) _ needed to reproduce current snia data ( in the redshift range @xmath47 ) are shown as a function of redshift in a flat , decelerating @xmath44 background cosmology.,title="fig:",width=257 ] cdm model obtained by using distant supernovae @xcite ( solid ) , a decelerating einstein desitter model ( dotted ) and a model in which a large scale peculiar acceleration field acts on top of the eds world model ( short - dashed curves ) .
the shaded area shows the state - of - the - art dispersion in the observed snia samples from the snls collaboration @xcite and from the hst gold sample @xcite . _
right : _ we show the convergence , at high redshift , of the peculiar acceleration model to the underlying background cosmological model ( an einstein de sitter model in this case).,title="fig:",width=283 ] cdm model obtained by using distant supernovae @xcite ( solid ) , a decelerating einstein desitter model ( dotted ) and a model in which a large scale peculiar acceleration field acts on top of the eds world model ( short - dashed curves ) .
the shaded area shows the state - of - the - art dispersion in the observed snia samples from the snls collaboration @xcite and from the hst gold sample @xcite .
_ we show the convergence , at high redshift , of the peculiar acceleration model to the underlying background cosmological model ( an einstein de sitter model in this case).,title="fig:",width=283 ] we find that the large scale peculiar acceleration required to fit the snia data ( see fig.[fig2 ] ) is negligible with respect to other local acceleration fields . given the smallness of its amplitude , it modifies almost undetectably the inertial nature of the reference systems to which it is applied at high redshift .
note that , in general , this is the order of magnitude of any large - scale , peculiar acceleration field one needs to introduce in order to reinterpret snia data over the redshift range @xmath53 .
for @xmath54 , the distance modulus predicted within this paradigm converges to the standard behavior expected in the underlying , decelerated , cosmological model because the peculiar perturbation vanishes at early epochs .
we stress here that , by definition , the acceleration induced by such an hypothetical non - gravitational force should act differentially on the baryonic and non - baryonic matter components .
therefore , any possible non - gravitational accelerating mechanisms are physically viable only if it does not destroy ( e.g. , by segregating it into parts ) the astronomical object on which they act ( for instance , by segregating the dark matter halo from the luminous baryonic component hosted within it ) . in other words ,
the acceleration time scale @xmath55 in a cosmological frame has to be much larger than the typical dynamical time scale @xmath56 of a system trapped inside a dark matter halo .
+ by inserting the time - averaged value of eq .
8 ( @xmath57 , if @xmath44 ) into eq . 5 and solving by using @xmath34
, we obtain hence @xmath58^{3/4 } \approx ( g\rho)^{-1/2 } \ ; , \ ] ] where @xmath59 is the comoving radius of the dark matter halo assumed in virial equilibrium , and @xmath60 is its mean density .
this simple physical argument puts stringent constraints on the viability of such a class of models . unless one is willing to consider non standard model of gravity sourced only by baryons [ e.g. , mond ( milgrom 1983 ) or teves theories ( bekenstein 2004 ) ] , or as yet unexplored interactions mediated by the dark matter particles
, one can use the previous argument to doubt the physical reliability of such non - gravitational acceleration mechanism .
moreover , models of non - cosmological acceleration that are able to reproduce the metric acceleration without invoking the effect of dark energy are not positively defined over all the epoch of cosmological interest . specifically , to explain the acceleration of distant standard candles without violating local constraints
, one would need to invoke an acceleration of opposite sign at low and high redshift ( see the left panel of fig .
1 ) . additionally , these models yield also quite high peculiar velocities , which can even be of the order of a tenth of the speed of light at a characteristic epoch of @xmath61 ( see the right panel of fig .
1 ) . in principle
, the strong sensitivity of the value of the large - scale peculiar velocity in such scenarios can also be used to put constraints on the viability of the non - gravitational acceleration models .
on one hand , one must mention the ability of these models to handle the age - problem in frw cosmologies with no dark energy . a recurring argument for a dark energy component
has come from considerations that the observed stellar ages are too long to be consistent with the observed hubble parameter @xmath62 km s@xmath63 mpc@xmath63 for an universe which has always been decelerating . since our results have been presented , for the sake of illustration , for a flat cdm model , they are consistent with the stellar estimates of the age of the universe only if low values of @xmath64 km s@xmath63 mpc@xmath63 are viable @xcite .
although there are authors claiming given the still large systematic uncertainties affecting @xmath65 measurements that at present the age argument can not be used to definitively exclude the einstein de sitter model and , therefore , arguing for a cosmological constant @xcite @xcite , we note here that the non gravitational acceleration models can naturally reconcile decelerated background cosmological models with the age of the oldest stars in the universe . as a matter of fact the observed hubble constant evaluated by comparing the observed redshift ( @xmath10 ) with redshift - independent distance indicators ( d ) would overestimate the true universal value @xmath65 obtained by using , as prescribed by theory , the cosmological redshift ( z ) .
the snia hubble diagram is constructed in such a way to be independent of the values of @xmath65 .
anyway , one can directly estimate the amplitude of the @xmath65 bias . in the accelerated model the expansion rate in the local universe
is overestimated by a factor @xmath66 .
therefore , at @xmath67 , the isotropic peculiar expansion needed to recover a hubble parameter of @xmath68 km s@xmath63 mpc@xmath63 is of order of @xmath69 .
we stress here the analogy between this _reconciliation mechanism_ and the one invoked within the ` hubble bubble ' scenarios .
in fact , in such models @xcite we are supposed to live inside an underdense region ( hubble bubble ) that is expanding faster than the average .
measurements of the hubble constant within the underdense region would therefore overestimate the universal value by @xmath70 thus offering a different way to bypass the age problem within a decelerating background metric .
we will see in the next session , however , that a crucial observational test can be devised to reject such non - gravitational acceleration mechanisms .
we now discuss the observable effects that a hypothetical non - gravitational force field , say a _ dark force field _ , should have on the global expansion of the universe , as traced by a set of distant standard candles ( e.g. , snia ) .
we show that the physical imprints of such a _ ` dark force ' _ can be unambiguously contrasted with those of a _ ` dark energy ' _ component .
in particular , it is possible to discriminate between them directly in the hubble diagram rather than , as usually done , in the equation of state parameter space . indeed ,
if such a non - gravitational , radial force field exists ( see , e.g. , the preliminary suggestion that a large - scale magnetic tension might mimic the effects of dark energy @xcite , or the equivalent proposal of a large local void @xcite ) , we can show that it accelerates cosmological objects ( test particles ) at the same radial distance in a different way .
in fact , since only the gravitational force has the property to depend on the gravitational mass which via the equivalence principle is equal to the inertial mass it follows that , in the proposed ` _ dark force _ ' scenario sne are not inertial systems in free - fall along geodesics of the space - time . the acceleration of sne will thus have a unique signature : it will depend on the mass of the hosting system .
since in this _ dark force _ scenario the luminosity distance @xmath71 of a cosmic object depends on its acceleration ( see eq .
[ eq.dl_om1 ] and [ eq.wz ] ) , the different inertial masses of the test particles will cause predictable deviations from the mean , best fitting curve shown in the left panel of fig .
2 . one can parameterize the mass dependence of the peculiar acceleration experienced by a given system @xmath72 in terms of the best fitting mean acceleration @xmath73 using a simple scaling law @xmath74 where the exponent @xmath75 characterizes the specific physical mechanism responsible for the acceleration of supernovae .
the limiting case @xmath76 describes the action of the gravitational field showing that , in this case , the acceleration of test particles is mass - independent .
in contrast , the acceleration generated by non - gravitational force fields is generically described by setting @xmath77 . in particular
, the case @xmath78 illustrates the large class of force fields in which the acceleration is inversely proportional to the mass of the test particle ( for example models in which the strength of the force is independent of the mass of the object experiencing the field ) . in eq .
[ acmo ] and by letting the minimum and maximum mass of the test particles ( sne host systems ) vary over five orders of magnitude i.e. , from normal isolated galaxies to galaxy clusters.,width=529 ] for the sake of simplicity let us discuss first the case @xmath78 .
a direct consequence of this assumption is that , at fixed redshift , sne hosted in isolated normal galaxies ( with typical mass of @xmath79 ) , when subject to the force field we are challenging here , will experience an acceleration of order @xmath80 stronger than sne hosted in rich galaxy clusters ( with typical mass of @xmath81 ) .
( we assume here that the masses of the host systems remain constant with time ) .
this effect systematically shifts the distance modulus of sne in low ( high ) mass hosts towards higher ( lower ) values with respect to the reference , best fitting curve characterizing supernovae hosted in some fiducial system of mass @xmath82 ( see fig . [ fig3 ] ) .
similarly , and without loss of generality , we find that for all the values @xmath83 , ( i.e. even in the extreme cases of an acceleration field which has a weak dependence on the host mass ) , the family of distance moduli associated to sne hosted in systems whose mass differs by up to five orders of magnitude still spans the whole dispersion region characterizing current data . the systematic deviation from the best fitting relation of the distance moduli of supernovae hosted in a large variety of systems , going from small to large masses , is therefore practically insensitive to the exponent @xmath75 parameterizing the mass dependence of the acceleration felt by supernovae .
this general result follows from the fact that a ) the distance modulus calculated by including the peculiar acceleration term can not be smaller than the one computed in the associated background cosmological model , and b ) to each value of the exponent @xmath75 one can always associate an appropriately tuned value of the fiducial mass in the range @xmath84 in such a way that a change of five orders of magnitude in the mass of the host gives distance moduli which are always confined between the upper and lower envelopes of the dispersion in the snia hubble diagram .
therefore , we conclude that in a _ dark force _ scenario , part of the scatter in the snia hubble diagram has a physical origin : for a fixed redshift @xmath30 and @xmath85 ( @xmath86 ) , snia with larger values of @xmath87 in fig .
[ fig3 ] should be hosted in small ( big ) systems , while snia with smaller values of @xmath87 should be hosted in rich clusters ( isolated normal galaxies ) . to summarize , a null - test of the metric nature of the accelerated expansion can be easily performed by means of an environmental analysis of the cosmic structures in which snia are found .
such a study could be optimally performed by future large - area sky surveys of distant sne such as snap @xcite .
the strong sensitivity to the host system mass of the test strategy that we suggest here will allow to exclude any hypothetical non - metric large - scale interaction which could be in principle responsible for the observed kinematics of the universe .
such a result would hence shed a much clearer light on the nature of the assumed dark energy component dominating the late stages of the cosmic evolution .
1.truecm * acknowldgements*.
we thank the referee for useful comments .
we acknowledge stimulating and useful discussions with d. fouchez , p. taxil and j.m .
thanks the centre de physique thorique de marseille for hospitality . | we discuss the testable predictions of a phenomenological model in which the accelerated expansion of the universe is the result of the action of a non - gravitational force field , rather than the effect of a negative - pressure dark - energy fluid or a modification of general relativity .
we show , through the equivalence principle , that in such a scenario the cosmic acceleration felt by distant standard candles like snia ( type ia supernovae ( sne ) ) depends on the mass of the host system , being larger in isolated galaxies than in rich clusters . as a consequence ,
the scatter in the observed snia hubble diagram has mostly a physical origin in this scenario : in fact , the snia distance modulus is increasing , at fixed redshift , for sne that are hosted in isolated galaxies with respect to the case of sne hosted in rich galaxy clusters .
due to its strong dependence on the astrophysical environments of standard candles , we conclude that alternative non - gravitational mechanisms for the observed accelerated expansion of the universe can be interestingly contrasted against the standard metric interpretation of the cosmological acceleration by means of an environmental analysis of the cosmic structures in which snia are found .
the possible absence of such environmental effects would definitely exclude non - gravitational mechanisms being responsible for the accelerated cosmological expansion and will therefore reinforce a metric interpretation . |
and non - perturbatively by the bielefeld group @xcite . in a series of works
@xcite we have pointed out that , instead of pqcd features , unexpected inverse power corrections in temperature dominate above the phase transition , a non - perturbative feature which finds a natural explanation in terms of dimension-2 condensates of the dimensionally reduced theory . on the other extreme
we have also found a hadronic representation in the confined phase both for the polyakov loop @xcite and its correlators @xcite ( see also @xcite ) . for a pedagogical review
see e.g. @xcite .
perturbative calculations of the renormalized polyakov loop have been pursued to nnlo @xcite .
the mysterious power corrections have been confirmed on the lattice for @xmath0 @xcite .
quite recently , the tum lattice calculation for physical quark masses in 2 + 1 flavours has been carried out in the temperature range @xmath1 @xcite concluding that at earliest pqcd to nnlo @xcite might set in at @xmath2 . here
we display the power correction pattern over a huge range of temperatures and argue on the missing singly heavy hadronic states below the phase transition .
the polyakov loop is a local operator which in the static gauge reads @xmath3 ( @xmath4 ) .
its expectation value is a ratio of two partition functions ( we take conventionally @xmath5 ) @xmath6 where the qcd lagrangian for @xmath7 flavours @xmath8 reads , in terms of the re - scaled gluon field @xmath9 with @xmath10 , @xmath11 the renormalized polyakov loop is uniquely defined up to a constant factor which corresponds to @xmath12 .
there is no natural way to fix the ambiguity , but it can be removed by using the corresponding entropy @xmath13 \ , .
\end{aligned}\ ] ] being dimensionless , one should have @xmath14 , with @xmath15 the renormalization scale .
renormalization group invariance requires @xmath16 where the beta function and the mass anomalous dimension are , @xmath17 respectively .
direct evaluation yields the shift in the specific heat @xmath18 \right\ } \equiv \frac{\partial \delta u_q}{\partial t } , \end{aligned}\ ] ] where @xmath19 is the trace of the energy momentum tensor , the argument of ref .
the entropy shift is not a true entropy .
for instance , the true entropy must be a monotonous function of the temperature since from @xmath20 it follows @xmath21 .
thus _ both _ @xmath22 and @xmath23 but the sign of @xmath24 is not fixed .
the exact relations for thermodynamics of heavy quarks @xcite are still subjected to ambiguities which are removed in the specific heat . ]
@xmath25 the entropy shift can be obtained by integrating with suitable boundary conditions featuring the dimensions of the hilbert space with and without polyakov loop . at low temperatures @xmath26 for @xmath27 degenerate flavours and @xmath28 , whereas at high temperatures @xmath29 , thus @xmath30 the first condition is the third principle of thermodynamics for degenerate states .
the recent tum lattice calculations directly provide the entropy in the range @xmath1 @xcite taking the convention @xmath31 in harmony with their normalization @xmath32 which , unlike ours , _ is not _ a partition function at low temperatures .
the critical temperature was found to be @xmath33 .
in the confined phase we expect a hadronic representation of the polyakov loop @xcite .
there the ambiguity comes from the heavy quark mass which is subtracted from the hadron total mass . in @xcite
we explicitly reconstructed the entropy @xmath34 although the available lattice data were noisier than the recent ones @xcite . in fig .
[ fig : entropy - models ] we show the lattice entropy results @xcite and compare them with the hadron resonance gas using either the bag model ( centered at the heavy quark source ) , the pdg or the rqm of isgur , godfrey and capstick for mesons and baryons @xcite taking either the charm or the bottom as the putative heavy quark .
we have noted in previous works that these rqm singly heavy states follow a hagedorn - like pattern with a hagedorn - polyakov temperature of about @xmath35 for @xmath36-quarks .
results from a simple constituent quark model ( cqm ) are also shown @xmath37 with @xmath38 and @xmath39 and spin degeneracies @xmath40 , @xmath41 .
at high temperatures the polyakov loop can be expanded in powers of @xmath42 @xcite @xmath43 .\end{aligned}\ ] ] @xmath44 has both perturbative ( pert ) and non - perturbative ( np ) contributions , whence the entropy reads @xmath45 where @xmath46 is a slowly varying function with temperature . in fig .
[ fig : entropy - inv ] we display the recent tum data @xcite as a function of @xmath47 .
a clear straight line behaviour emerges over a huge range of temperatures .
a fit for @xmath48 , neglecting the @xmath49 dependence in @xmath46 , gives @xmath50 with @xmath51 for @xmath52 .
this gives @xmath53 which is within the expected @xmath54 .
when a heavy colour source in the fundamental representation of the @xmath55 group is placed into the hadronic vacuum there arises an entropy shift as a measurable and unambiguous observable .
we noted long ago that lattice data for the corresponding free energy display corrections to the perturbative result in the unequivocal form of an inverse second power of the temperature .
this behaviour has been corroborated by subsequent lattice studies .
the present analysis improves on previous ones thanks to the quality of the tum lattice data , and the fact that the unambiguous entropy is used for the comparison . at low temperatures the quick increase in the number of active states , @xmath56 ,
suggests that there are missing states in the singly heavy hadronic spectrum . at high temperatures power corrections
dominate over the perturbative contributions , in harmony with our previous findings .
taken at face value , the impressive agreement calls for a deeper understanding on the nature of these power corrections above the phase transition .
o. kaczmarek et al .
b543 ( 2002 ) 41 .
e. megias , e. ruiz arriola and l.l .
salcedo , jhep 01 ( 2006 ) 073 .
e. megias , e. ruiz arriola and l.l .
salcedo , phys .
d75 ( 2007 ) 105019 .
e. megias , e. ruiz arriola and l.l .
salcedo , phys .
d80 ( 2009 ) 056005 .
e. megias , e. ruiz arriola and l.l .
salcedo , phys .
d81 ( 2010 ) 096009 .
e. megias , e. ruiz arriola and l.l .
salcedo , phys .
( 2012 ) 151601 .
e. megias , e. ruiz arriola and l.l .
salcedo , phys .
d89 ( 2014 ) 076006 .
e. megias , e. ruiz arriola and l.l .
salcedo , ( 2016 ) , 1603.04642 .
n. brambilla et al .
d82 ( 2010 ) 074019 .
m. berwein et al .
d93 ( 2016 ) 034010 .
a. mykkanen , m. panero and k. rummukainen , jhep 05 ( 2012 ) 069 .
a. bazavov et al . , (
2016 ) , 1603.06637 . | a heavy quark placed in the medium modifies its specific heat . using a renormalization group argument
we show a low energy theorem in terms of the defect in the trace of the energy - momentum tensor which allows the unambiguous determination of the corresponding entropy shift after imposing the third principle of thermodynamics for degenerate states .
we show how recent lattice qcd data can be understood in the confined phase in terms of a single - heavy hadronic spectrum and above the phase transition through power corrections which are analyzed by means of a dimension 2 gluon condensate of the dimensionally reduced theory . |
let @xmath0 be a graph with vertex set @xmath1 and edge set @xmath2 .
let @xmath3 ( where @xmath4 denotes the nonnegative real numbers ) .
we will call @xmath5 a length function .
we will call such a pair , @xmath6 , a linkage . note that this is not standard terminology
. however , it seems appropriate in the given context to formalise the intuition that a weighted graph is the mathematical model for a mechanical linkage consisting of hinges and bars that are constrained to move in a plane ( we ignore the issue of self intersections ) .
given such a linkage @xmath7 we define the space of planar configurations of @xmath8 as follows : @xmath9 where @xmath10 denotes the standard euclidean distance between @xmath11 and @xmath12 . by definition
, @xmath13 is a subset of @xmath14 and thus inherits a natural metric space structure .
observe that there is a canonical action of the group of orientation preserving isometries of the plane on @xmath13 .
we define the the moduli space of the linkage , denoted by @xmath15 or @xmath16 , to be the orbit space of this action .
it is easy to see that if @xmath0 is connected then @xmath15 is a compact real algebraic variety . in general , it is difficult to decide whether or not @xmath15 is even nonempty .
an element of @xmath15 is called a realisation of the linkage @xmath8 .
the problem of finding a realisation of @xmath8 is known as the molecule problem - see @xcite . in the case where @xmath15 is nonempty , it is difficult to say much about the topology of this space without imposing some restrictions on the structure of the underlying graph @xmath0 .
the case where @xmath0 is a polygonal graph ( i.e. connected with every vertex of degree two ) is quite well understood and much is known about the topology of @xmath15 in this case .
for example , we have the following ( see theorem 1.6 of @xcite , for example ) . [ polygonal ]
if @xmath0 is a polygonal graph , then @xmath15 is nonempty if and only if the longest edge has length at most half the total length of all the edges .
moreover @xmath15 is connected if and only if the sum of the lengths of the second and third longest edges is at most half of the total length of all the edges .
indeed , much more detailed information about the topology of @xmath15 is available when @xmath0 is polygonal .
the homotopical and homological properties of these spaces are well understood - see @xcite , @xcite or @xcite , for example . for an overview of some of the theory of polygonal linkages ,
we refer the reader to @xcite .
our purpose in this paper is to study @xmath15 where @xmath0 is a series parallel graph ( see below for definitions ) .
we will show that it is possible to easily determine , for a given series parallel graph @xmath0 and weight function @xmath5 , whether or not @xmath15 is nonempty and , in the case when it is nonempty , whether or not it is connected .
the problem of deciding whether the space is connected or not is connected with the motion planning problem in robotics .
the motion planning problem is concerned with the existence of a path betweeb two configurations of a robot .
if we think of our linkages as a model for mechanical linkages , then the motion planning problem for this particular type of robot " is equivalent to finding a continuous path in @xmath15 with specified endpoints . we will show that for the class of series parallel graphs these problems can be answered by considering a finite system of linear inequalities in the edge lengths .
in contrast , we note that even for the complete graph on four edges , the smallest 2-connected graph which is not series parallel , it is necessary to solve a polynomial equation of total degree 6 ( quartic in each variable ) in the edge lengths to determine whether or not a realisation exists . throughout this paper
we adopt the convention that @xmath7 and that @xmath17 .
in this section , we review the basic constructions and facts concerning the class of series parallel graphs .
first , we fix some conventions regarding some standard graph theory .
a graph is a pair @xmath18 where @xmath2 is a multiset of unordered pairs of distinct elements of @xmath1 .
thus , in particular multiple edges with the same endpoints are allowed . however loops are not allowed .
a path graph is a graph isomorphic to a graph with vertex set @xmath19 and edge set @xmath20 .
a path linkage is a linkage @xmath21 where @xmath22 is a path graph .
a polygonal graph is a graph isomorphic to a graph with vertex set @xmath19 and edge set @xmath23 .
a polygonal linkage is a linkage @xmath6 where @xmath0 is a polygonal graph .
a two terminal graph ( ttg ) is an ordered triple @xmath24 where @xmath25 and @xmath26 are distinct vertices of @xmath0 called the source and the sink , respectively .
collectively @xmath25 and @xmath26 are called the terminal vertices of the ttg .
given ttgs @xmath27 and @xmath28 we can define the series composition @xmath29 to be the ttg @xmath30 where @xmath31 denotes the graph obtained by identifying the vertices @xmath32 and @xmath33 .
also we define the parallel composition @xmath34 to be the ttg @xmath35 see figure [ spcomposition ] for an illustration of these constructions .
observe that the operation of parallel composition is a commutative associative operation on the class of ttgs .
thus , in particular , given ttgs @xmath36 for @xmath37 we can unambiguously refer to the parallel composition @xmath38 parallel and series composition of two terminal graphs . ]
let @xmath39 denote the complete graph with vertex set @xmath40 .
we define the class of two terminal series parallel graphs ( ttspgs ) to be the smallest class of ttgs that contains @xmath41 and that is closed under the operations of series and parallel composition .
a series parallel graph is a graph @xmath0 such that @xmath24 is a ttspg for some choice of vertices @xmath25 and @xmath26 .
thus , for example , path graphs are series parallel .
also polygonal graphs are series parallel , since a polygon is the parallel composition of two paths .
a series parallel linkage is a linkage @xmath6 such that @xmath0 is a series parallel graph .
we note that the operations of parallel composition and series composition extend in an obvious way to linkages - so it makes sense to refer to the parallel composition @xmath42 or the series composition @xmath43 , where @xmath44 and @xmath45 are linkages rather than graphs .
observe that for a given series parallel graph , there may be many possible choices of terminal vertices
. however , the choice is not completely arbitrary - some pairs of vertices can not be the terminal vertices of a given series parallel graph .
for example , the existence of a subgraph of @xmath0 homeomorphic to @xmath46 implies that @xmath47 is not a ttspg .
there are other possible obstructions . for a more detailed discussion of the possible choices of terminal vertices ,
. the following lemma will prove useful for our analysis of the connectedness of the moduli space of a series parallel linkage .
recall that a graph @xmath0 is 2-connected if the complement of any vertex is connected .
observe that a series parallel graph is 2-connected if and only if it can not be expressed as a series composition of proper subgraphs .
[ twoconnspglemma ] let @xmath0 be a 2-connected series parallel graph .
there are vertices @xmath25 and @xmath26 in @xmath0 such that @xmath24 is a ttspg and such that @xmath48 where @xmath49 and @xmath50 are paths joining @xmath25 and @xmath26 and @xmath51 is a ( possibly empty ) subgraph of @xmath0 such that @xmath52 is a ttspg .
let @xmath22 be a subgraph of @xmath0 such that @xmath22 is a path and such that every interior vertex of @xmath22 has degree 2 in @xmath0 ( i.e. no other edges of @xmath0 are incident to the interior of @xmath22 ) .
let @xmath25 and @xmath26 be the endpoints of @xmath22 . by an easy modification of the proof of lemma 9 in @xcite , we see that @xmath24 is a ttspg ( note that the hypothesis of 2-connectedness is necessary at this point ) and thus @xmath53 . here
@xmath54 is the subgraph of @xmath0 spanned by all the edges that are not in @xmath22 .
now it also easy to show ( for example , by induction on the number of edges ) that in any series parallel graph that is not itself a path graph , it is possible to find two distinct path subgraphs @xmath49 and @xmath50 with common endpoints and such that no other edges of @xmath0 are incident with any of the internal vertices of @xmath49 and @xmath50 .
applying our previous observation to @xmath49 and @xmath50 completes the proof of the lemma .
series parallel graphs are a well studied class of graphs ( see @xcite and @xcite for example ) .
of particular interest to us is the following result of belk and connelly ( see @xcite ) .
we say that a graph is @xmath55-realisable ( where @xmath55 is a positive integer ) if given any positive integer @xmath56 and any function @xmath57 , there exists a function @xmath58 such that @xmath59 for all edges @xmath60 .
intuitively , this means that any embedding of @xmath0 into some ( possibly high dimensional ) euclidean space can be squashed into @xmath61 so that the edge lengths are preserved .
[ connellybeck ] ( belk , connelly ) a graph is 2-realisable if and only if it is a series parallel graph . of course , knowing that a given graph @xmath0 is 2-realisable does not tell us whether or not @xmath6 is realisable for a particular length function @xmath5 , nor does it tell us anything about the topology of @xmath16 .
however theorem [ connellybeck ] does suggest that the class of series parallel graphs is an interesting class for which to study the space @xmath16 .
now suppose that @xmath24 is a ttspg graph and that @xmath5 is a length function on @xmath0 .
let @xmath8 be the linkage @xmath6 .
let @xmath62 = \ { |p(s)-p(t)| : p \in m(l)\}.\ ] ] here we are abusing notation somewhat by writing @xmath63 for an element of @xmath64 , but also using @xmath63 to denote a particular representative in @xmath65 of the orbit under the action of orientation preserving isometries of @xmath66 .
however , this clearly does not cause any problems with this definition as the quantity @xmath67 is preserved by this action .
we will consistently abuse notation in this way throughout the remainder of the paper .
in other words @xmath68 $ ] is the set of all possible values of the distance between @xmath69 and @xmath70 as @xmath63 varies over all realisations in @xmath15 . in the case where @xmath8 is a path linkage ,
there is only one possible choice for the set of terminal vertices , so we will write @xmath71 $ ] for @xmath68 $ ] in this case . note , that @xmath68 $ ] could be empty . indeed , the linkage is realisable ( i.e. @xmath15 is nonempty ) if and only if @xmath68 $ ] is nonempty . we will show that it is possible to easily compute @xmath68 $ ] for a given ttspg .
observe that in general it is difficult to compute the set of possible distances between a pair of points as we vary over all realisations of a ( possibly non series parallel ) graph .
however for the special situation that we consider , it is possible
. let @xmath72 and let @xmath73 and let @xmath74 .
then @xmath62 = [ l_1,s_1,t_1 ] \cap [ l_2,s_2,t_2].\ ] ] in particular , @xmath8 is realisable if and only if @xmath75 \cap [ l_2,s_2,t_2]$ ] is nonempty .
if @xmath76 $ ] then there is some @xmath77 such that @xmath78 . for @xmath79 , let @xmath80 .
now @xmath81 , so @xmath82 $ ] .
this shows that @xmath68 \subseteq [ l_1,s , t]\cap [ l_2,s , t]$ ] . for the other inclusion ,
suppose that @xmath83\cap [ l_2,s , t]$ ] .
so , for @xmath79 , there are realisations @xmath84 of @xmath85 such that @xmath86 .
clearly , @xmath87 and @xmath88 together induce a realisation @xmath63 of @xmath8 such that @xmath67 . for series compositions ,
we make the following definition .
given intervals @xmath89 $ ] and @xmath90 $ ] with @xmath91 and @xmath92 , define the composition @xmath89 \circ [ c , d]$ ] to be the interval @xmath93\ ] ] if we write @xmath94 to denote the empty interval , then we define @xmath89\circ \phi : = \phi$ ] .
observe , for example , that @xmath89 \circ [ c , d ] = [ 0,b+d]$ ] if and only if @xmath89\cap[c , d]$ ] is nonempty .
let @xmath95 and @xmath96 be linkages such that @xmath75 $ ] and @xmath97 $ ] are both closed intervals .
let @xmath98 .
then @xmath99 = [ l_1,s_1,t_1]\circ [ l_2,s_2,t_2]\ ] ] this follows immediately from the observation that @xmath100 $ ] if and only if there is @xmath101 $ ] and @xmath102 $ ] such that @xmath103 , @xmath104 and @xmath105 are the lengths of the sides of a triangle .
let @xmath24 be a ttspg and let @xmath7 .
then @xmath68 $ ] is either empty or is a closed bounded interval of @xmath106 .
this follows from a simple induction on the number of edges in @xmath0 .
we note that there are efficient algorithms available for recognizing series parallel graphs and for finding a series parallel decomposition of a given series parallel graph ( see @xcite and @xcite ) .
now , it is clear how to compute @xmath68 $ ] when @xmath0 is a ttspg . in particular
, this allows us to easily determine whether or not a given series parallel linkage is realisable .
[ linkageexample ] let @xmath8 be the series parallel linkage whose combinatorial structure is indicated in the diagram below @xmath107 the label on an edge is the length of that edge .
consider the following five sublinkages of @xmath8 @xmath108 clearly , @xmath109 now @xmath110 = [ 7,13]$ ] , @xmath111=[3,5]$ ] , @xmath112=[0,4]$ ] , @xmath113 = [ 1,13]$ ] and @xmath114 = [ 0,5]$ ] .
therefore , @xmath115 & = & ( ( [ 3,5]\cap [ 0,4])\circ[7,13])\cap[1,13 ] \cap[0,5 ] \\ & = & ( [ 3,4]\circ[7,13])\cap[1,13]\cap[0,5 ] \\ & = & [ 3,17]\cap[1,13]\cap[0,5 ] \\ & = & [ 3,5 ] \end{array}\ ] ] in particular , the linkage @xmath8 is realisable .
we conclude this section by showing that the realisability problem for a given series parallel linkage can be answered by looking only at the polygonal sublinkages of the given linkage .
this is not necessarily true for linkages whose underlying graph is not series parallel .
for example , consider the complete graph on four vertices where each edge is given length 1 .
every polygonal sublinkage of this linkage is realisable in the plane but the complete linkage is not .
however , for series parallel graphs , we have the following .
let @xmath116 be a series parallel linkage
. then @xmath8 is realisable if and only if , for every polygonal subgraph @xmath51 of @xmath0 , the linkage @xmath117 is realisable .
it is obvious that if @xmath8 is realisable then every sublinkage of @xmath8 is also realisable . for the other implication , we argue by contradiction .
suppose that @xmath8 is a counterexample to the statement with the minimal number of edges .
so @xmath8 is not realisable but every polygonal sublinkage of @xmath8 is realisable .
note that the minimality of @xmath8 ensures that every proper sublinkage of @xmath8 is realisable .
in particular @xmath8 can not be decomposed as a series composition of proper sublinkages .
so there is some pair of vertices @xmath118 in @xmath0 such that @xmath119 , and such that @xmath120 $ ] and @xmath121 $ ] are nonempty but @xmath122 \cap [ l_2,s , t]$ ] is empty .
assume without loss of generality that @xmath122=[a_1,b_1]$ ] lies to the left of @xmath121=[a_2,b_2]$ ] ( i.e. @xmath123 ) .
now we observe that there is some path graph @xmath49 joining @xmath25 to @xmath26 contained in @xmath124 such that @xmath110 = [ \alpha , b_1]$ ] and there is some path graph @xmath50 joining @xmath25 to @xmath26 contained in @xmath125 such that @xmath111 = [ a_2,\beta]$ ] .
it is clear that the polygonal linkage @xmath126 is not realisable which contradicts our assumption that all polygonal sublinkages of @xmath8 are realisable .
to understand the connectedness of @xmath15 for a series parallel linkage @xmath8 , we need to more precisely understand the relationship between configurations of a path linkage and the corresponding distances between the images of the terminal vertices . throughout this section
let @xmath22 be a path linkage with @xmath127 edges and suppose that @xmath128 .
we suppose that all the edges of the linkage @xmath22 have nonzero length .
let @xmath25 and @xmath26 be the terminal vertices of @xmath22 .
let @xmath129 $ ] , @xmath130 . in this section we will show that @xmath131 has a certain lifting property .
the basic idea is to use morse theory to analyse the fibrewise structure of @xmath131 .
we remark that the differential properties of the map @xmath131 are well understood ( see @xcite ) .
it is differentiable at all points not in @xmath132 .
also the points where the derivative of @xmath131 vanishes are precisely the straight line configurations of @xmath22 ( i.e those points @xmath63 for which the set @xmath133 lies in an affine line in @xmath66 ) .
we will say that @xmath134 is a critical point of @xmath131 if either @xmath135 or @xmath136 .
the basic question that we now consider is this .
suppose that @xmath63 and @xmath137 are two configurations of a path linkage @xmath22 . clearly , since @xmath64 is pathwise connected ( it is homeomorphic to @xmath138 ) , it is possible to find a path ( in the sense of topological spaces ) in @xmath64 that connects @xmath63 to @xmath137 .
however , suppose that the motion of the endpoints of @xmath22 is specified .
is it possible to find a path in @xmath64 connecting @xmath63 and @xmath137 so that the endpoints of @xmath22 move in a specified way ? theorems [ liftingtheorem1 ] and [ liftingtheorem2 ] below will be the key to constructing paths in @xmath15 when @xmath8 is a series parallel linkage . first , we need some notation to describe a particular subset of @xmath139 $ ] .
we define @xmath140 to be the following subset of @xmath139 $ ] .
@xmath141 : \theta^{-1}(x ) \text { is connected}\}.\ ] ] note that if @xmath22 has just two edges of length @xmath142 and @xmath143 , then @xmath144 ( i.e. @xmath140 consists of two points ) . when @xmath22 has more than two edges , @xmath140 is union of at most two closed intervals , as the following analysis shows .
let @xmath145 be the lengths of the edges in @xmath22 , with @xmath146 .
we suppose for the moment that @xmath147 .
( note that permuting the edge lengths does not affect the homeomorphism type of @xmath148 . ) in the following lemma we are using the convention that for @xmath149 , @xmath89 $ ] is the empty set .
[ connectivityrange ] let @xmath150
. then @xmath140 is @xmath151\cap \biggl([2(l_2+l_3)-s , l_3]\cup [ l_3,\min\{l_1,s-2l_2\}]\cup[\max\{l_1,s-2(l_1+l_2)\},s]\biggr).\ ] ] this is a straightforward application of theorem [ polygonal ] to the polygonal linkage obtained by adjoining the edge @xmath40 to @xmath22 and extending the length function @xmath5 by defining @xmath152 , where @xmath103 is and arbitrary element of @xmath139 $ ] .
in particular , lemma [ connectivityrange ] shows that for a given path linkage , @xmath22 , it is very straightforward to calculate @xmath140 .
suppose @xmath22 has 3 edges and that @xmath153 . then @xmath154 $ ] which is a proper subset of @xmath139 = [ 0,3]$ ] .
[ liftingtheorem1 ] let @xmath155 and suppose that neither @xmath156 nor @xmath157 are critical values of @xmath131 .
let @xmath158 \rightarrow [ p]$ ] be continuous and suppose that @xmath159 and @xmath160 is non empty then there exists a continuous lift @xmath161 \rightarrow m(p)$ ] such that @xmath162 , @xmath163 and @xmath164 .
note that theorem [ liftingtheorem1 ] requires that neither @xmath63 nor @xmath137 lie in the preimage of a critical value . in order to remove this hypothesis
, we must tighten the requirements on @xmath165 .
in particular , we may require @xmath165 remains stationary for some positive amount of time near 0 or near 1 .
more precisely , we have the following . [ liftingtheorem2 ] let @xmath155 .
let @xmath158 \rightarrow [ p]$ ] be continuous and suppose that @xmath166 for @xmath167 $ ] for some @xmath168 , and that @xmath169 for @xmath170 $ ] for some @xmath171 .
if @xmath172 is non empty then there exists a continuous lift @xmath161 \rightarrow m(p)$ ] such that @xmath162 , @xmath163 and @xmath173 . in order to prove theorems [ liftingtheorem1 ] and [ liftingtheorem2 ] we will need to understand the fibrewise structure of the map @xmath129 $ ] .
this map has been studied by previous authors using the techniques of morse theory .
in particular , shinamoto and vanderwaart have given a very clear account of this theory in @xcite ( their notation is somewhat different to ours ) .
we can summarise the situation as follows .
let @xmath174 .
then @xmath175 $ ] is a differentiable function .
moreover , in this restricted domain , @xmath131 has finitely many critical points , all of which are nondegenerate .
if we also include @xmath176 as a critical value of @xmath177 $ ] , then there are finitely many critical values @xmath178 . for @xmath179 ,
let @xmath180)$ ] . by standard results of morse theory ( see @xcite ) , we know that for each @xmath181 there is a smooth closed @xmath182 dimensional manifold @xmath183 such that @xmath184 \times \sigma_i}{\sim}\ ] ] where @xmath185 collapses some subsets of @xmath186 to points and also some subsets of @xmath187 to points . in other words @xmath188 is obtained by taking @xmath189\times \sigma_i$ ] and making some identifications over the endpoints @xmath190 and @xmath191 .
indeed , we can be more explicit over the non zero critical points . in those cases
the identifications are obtained by collapsing some finite number of embedded spheres in @xmath192 . in the case
where @xmath193 , the collapsing over @xmath194 can be a little more complicated , but that does not affect the validity of our arguments below .
so , for each @xmath195 , we have a commutative diagram @xmath196^{\equiv}\ar[dr]_{\theta } & & \frac{[a_i , a_{i+1 } ] \times \sigma_i}{\sim}\ar[dl]^{\text{projection}}\\ & [ a_i , a_i+1 ] & } \ ] ] clearly all of the non zero critical values of @xmath131 are contained in @xmath140 ( this is an easy exercise for the reader ! ) .
moreover , it is known ( see @xcite ) that if @xmath197 , then @xmath148 is the disjoint union of two copies of @xmath198 . in other words , @xmath192 is disconnected if and only if @xmath199 .
see figure [ fibrewisefig ] for an illustration of the structure of @xmath131 . for the purposes of illustration
we have represented pieces of @xmath64 as two dimensional surfaces , even though @xmath64 is actually a torus of dimension @xmath200 .
however , figure [ fibrewisefig ] does give a reasonably faithful picture of how the fibres of @xmath131 behave .
in the example illustrated in the figure the open interval @xmath201 lies in the complement of @xmath140 .
the curves drawn in the interior of @xmath202 , @xmath188 and @xmath203 are meant to represent the fibres of @xmath131 over the points @xmath103 , @xmath104 and @xmath105 respectively .
the fibrewise structure of the map @xmath131 . ]
now , it is clear how we should prove theorems [ liftingtheorem1 ] and [ liftingtheorem2 ] .
_ proof of theorem [ liftingtheorem1 ] . _
we make the following observations .
the projection @xmath204\times \sigma_i}{\sim } \rightarrow [ a_i , a_{i+1}]\ ] ] has sections .
indeed given points @xmath205 and @xmath206 that lie in the same path component of @xmath192 and given distinct points @xmath207 and @xmath208 in @xmath189 $ ] there exists a continuous section @xmath209\rightarrow \frac{[a_i , a_{i+1}]\times \sigma_i}{\sim}\ ] ] such that @xmath210 and @xmath211 . by combining @xmath165 with a judicious use of these sections ,
it is clear that we can find the required lift of @xmath165 . for the sake of completeness
, we have included the details of this argument below .
however these details are rather tedious and not particularly enlightening .
thus , if the reader if sufficiently convinced by the arguments already presented , he may , at this point , skip the rest of this proof .
we must consider several different cases .
first , let us deal with the case where @xmath63 and @xmath212 happen to lie in same fibre of @xmath131 .
if @xmath213 for all @xmath214 $ ] then the conclusion is clearly true , as by assumption we must have @xmath215 , and therefore we can lift @xmath165 by choosing a path within @xmath216 that connects @xmath63 and @xmath137 . if @xmath165 is not a constant function then we can choose some @xmath217 $ ] such that @xmath218 but so that @xmath219)$ ] is contained within one of the open intervals @xmath220 .
now it is clear that we can lift @xmath221}:[0,b]\rightarrow [ p]$ ] since @xmath131 restricts to a trivial fibre bundle over @xmath219)$ ] .
thus we are left the problem of lifting @xmath222}:[b,1]\rightarrow [ p]$ ] with specified lifts of @xmath223 and of 1 .
in other words , we have reduced to case where @xmath63 and @xmath137 lie in different fibres of @xmath131 .
suppose now that @xmath63 and @xmath224 lie in different fibres of @xmath131 .
also suppose that @xmath225 ( similar arguments apply if @xmath226 ) .
since we have assumed that @xmath156 is not a critical value , @xmath156 must in fact lie in the interior of @xmath140 .
therefore , by concatenating local sections over @xmath189 $ ] of the type described above , we can find a ( global ) section @xmath227 \rightarrow m(p)$ ] such that @xmath228 and @xmath229 .
let @xmath230 .
clearly @xmath231 is the required lift in this case . finally , we consider the case where @xmath232 and @xmath233 .
choose some @xmath234 $ ] such that @xmath235 ( our hypotheses guarantee the existence of at least one such @xmath236 ) .
now we choose a point @xmath237 in the fibre @xmath238 .
if @xmath239 is in the interior of @xmath140 , we can choose @xmath237 arbitrarily within the fibre .
however , if @xmath239 happens to be on the boundary of @xmath140 ( and is therefore also a critical value of @xmath131 ) , we must be more selective in our choice of @xmath237 . in this case
we choose @xmath237 to a critical point of @xmath240 .
now , once we have chosen @xmath237 in this way , we can find two global sections @xmath241 and @xmath242 of @xmath131 , such that @xmath243 , @xmath244 and @xmath245 .
now , for @xmath246 $ ] , let @xmath247 and for @xmath248 $ ] , let @xmath249 .
one readily checks that @xmath250 is the required lift of @xmath165 in this case .
_ proof of theorem [ liftingtheorem2 ] .
_ in the case that one of @xmath63 or @xmath137 lies in a critical fibre ( i.e the preimage of one of the @xmath190s ) , then it may be necessary to first move to a different point in that fibre to ensure that when we lift along sections , we end up in the right path component of subsequent fibres .
the hypotheses of theorem [ liftingtheorem2 ] allow the intervals @xmath251 $ ] and @xmath252 $ ] to carry out this adjustment within the fibre .
we will also need the following lifting result later .
its proof is again an straightforward consequence of the fibrewise structure of @xmath253 described above , so we shall leave the reader to fill in the details in this case .
[ liftingtheorem3 ] let @xmath134 and suppose that @xmath254\rightarrow [ p]$ ] is a continuous function such that @xmath255 .
there is some continuous lift @xmath256 \rightarrow m(p)$ ] such that @xmath162 and @xmath257 .
now we show that any path in @xmath64 that connects two different path components of a fibre @xmath148 must pass through @xmath258 .
[ connlemma ] suppose that @xmath197 and let @xmath63 and @xmath137 be two realisations of @xmath22 that lie in different components of @xmath148 .
suppose that @xmath259\rightarrow m(p)$ ] is a continuous function such that @xmath260 and @xmath261 .
then there is some @xmath234 $ ] such that @xmath262 .
it is clear from the above description of @xmath131 that there must exist @xmath236 such that @xmath263 for some critical value @xmath264 of @xmath131 .
however , as remarked above , @xmath265 .
we conclude this section by observing that if @xmath266 , if @xmath267 and if @xmath268 is any orientation reversing isometry of @xmath66 , then @xmath269 and @xmath63 lie in different path components of @xmath148 .
now we present a method for checking the connectedness of @xmath15 when @xmath116 is a series parallel linkage .
first observe that we may as well restrict our attention to the case where the graph @xmath0 is a 2-connected series parallel graph .
if @xmath0 is not 2-connected , then it can be decomposed into a series composition of 2-connected series parallel graphs .
it is clear that @xmath15 is connected if and only the moduli space of each of the series components of @xmath8 is connected .
let @xmath8 be a 2-connected series parallel linkage such that @xmath15 is not empty .
recall that , by lemma [ twoconnspglemma ] , we can find vertices @xmath270 and @xmath271 such that @xmath47 is a ttspg and such that @xmath272 where @xmath273 is a sub ttspg of @xmath47 and where each @xmath274 is a path joining @xmath270 and @xmath271 , and @xmath275 .
see figure [ parallelpaths ] for an illustration of this situation . ]
we will write @xmath276 to denote the sublinkage @xmath277 of @xmath8 .
[ spconone ] with notation as above , if @xmath278 $ ] is empty for some @xmath279 then @xmath15 is disconnected .
. we can construct another realisation of @xmath8 by reflecting the vertices @xmath274 in a line through @xmath11 and @xmath12 . call this realisation @xmath137 .
now if @xmath158\rightarrow m(l)$ ] is any path , then by assumption @xmath281 for all @xmath214 $ ] .
therefore , by lemma [ connlemma ] and the observation that immediately follows that lemma , @xmath165 can not be a path that connects @xmath63 and @xmath137 .
what can we say about the connectedness of @xmath15 if the hypothesis of theorem [ spconone ] is not satisfied , in other words if @xmath278 $ ] is non empty for all @xmath279 ? in this case
, we construct a path linkage @xmath282 as follows .
suppose that @xmath283 = [ a_i , b_i]$ ] for @xmath284 .
let @xmath285 .
let @xmath286 ( note that @xmath287 since @xmath6 is realisable )
. now let @xmath282 be a path linkage with four edges @xmath288 and assign length @xmath289 to edge @xmath290 as follows ; @xmath291 and @xmath292 . [ qproperty ]
@xmath293=[a , b]$ ] and @xmath294 $ ] .
the first statement is obvious and the second statement follows immediately from lemma [ connectivityrange ] let @xmath25 and @xmath26 be the terminal vertices of @xmath282 and define a linkage @xmath44 by @xmath295 in other words @xmath296 is obtained from @xmath8 by replacing all the @xmath274s by the single path linkage @xmath282 . note that @xmath44 has a strictly smaller series parallel decomposition in terms of path linkages than @xmath8 does .
[ spcontwo ] with the notation as above , suppose that @xmath297 $ ] is nonempty for each @xmath37
. then @xmath15 is connected if and only if @xmath298 is connected .
we first observe that , by construction , @xmath293 = [ p_1]\cap \dots \cap[p_n]$ ] .
it follows that @xmath299 where @xmath300 and @xmath301 are the canonical maps induced by restriction .
now , suppose that @xmath298 is connected and let @xmath63 and @xmath137 be elements of @xmath15 .
we must show that there is a path in @xmath15 joining @xmath63 and @xmath137 .
we construct this path in several stages .
first , by our observations above , we can choose some realisations @xmath87 and @xmath302 of @xmath44 that agree with @xmath63 and @xmath137 on @xmath276 .
now since @xmath298 is connected , there is some path @xmath158 \rightarrow m(l_1)$ ] such that @xmath303 and @xmath304 .
now we can apply theorem [ liftingtheorem3 ] to construct a path @xmath305 \rightarrow m(l ) $ ] such that @xmath306 and @xmath250 agrees with @xmath307 on vertices of @xmath54 .
we just define @xmath308 for all vertices @xmath309 .
to lift @xmath250 to @xmath15 , we use theorem [ liftingtheorem3 ] ( once for each @xmath274 ) .
in particular @xmath310 . of course , it may happen that for some or all of the @xmath274s , @xmath311 .
so we have to concatenate other paths onto the end of @xmath250 to correct " it on the @xmath274s .
we can do this one @xmath274 at as time as follows .
let @xmath312\rightarrow [ l , u , v]$ ] such that @xmath313 and such that @xmath314 for some @xmath315 $ ] .
moreover , we can certainly choose @xmath316 so that it is stationary in a neighbourhood of @xmath176 and in a neighbourhood of @xmath317 .
now , it is clear that by applying theorem [ liftingtheorem2 ] to @xmath316 we can find some @xmath318 \rightarrow m(l)$ ] such that @xmath319 , @xmath320 and @xmath321 agrees with @xmath322 for vertices that are not in @xmath274 .
concatenating @xmath323 and @xmath324 corrects " the final position of vertices of @xmath274 .
we can repeat this process for all the @xmath274s , if necessary , and we eventually end up with the required path in @xmath15 connecting @xmath63 and @xmath137 . the converse can be proved in much the same way .
suppose that @xmath15 is connected and let @xmath87 and @xmath302 be points in @xmath298 .
we can find @xmath63 and @xmath137 in @xmath15 that agree with @xmath87 and @xmath302 on @xmath276 .
since @xmath15 is connected , we can find a path @xmath254 \rightarrow m(l_k)$ ] such that @xmath165 connects @xmath325 and @xmath326 and such that @xmath327 is contained in the image of the natural map @xmath328 . now , using theorem [ liftingtheorem3 ] , we can lift @xmath165 to a path @xmath329\rightarrow m(l_1)$ ] and using theorem [ liftingtheorem2 ] we can correct @xmath330 so that it agrees with @xmath302 on @xmath282 as necessary . note that lemma [ qproperty ] ensures that the hypotheses of theorem [ liftingtheorem2 ] are satisfied in this situation .
theorems [ spconone ] and [ spcontwo ] form the basis of a simple recursive algorithm for deciding whether or not @xmath15 is connected for a 2-connected series parallel linkage .
we informally describe this algorithm by the following sequence of steps .
we assume that @xmath15 is nonempty . 1 .
find a parallel decomposition of the form @xmath331 = ( p_1,u , v)\|(p_2,u , v)\|\dots \|(p_n , u , v ) \| ( k , u , v)\ ] ] where @xmath275 .
2 . compute @xmath332 $ ] using the methods described in section [ sec : realisability ] .
compute @xmath333 for each @xmath279 using lemma [ connectivityrange ] .
if @xmath278 $ ] is empty for any @xmath279 , then @xmath15 is not connected and we can stop .
4 . if @xmath278 $ ] is non empty for all @xmath279 , and @xmath54 is empty then @xmath15 is connected and we can stop .
if @xmath278 $ ] is non empty for all @xmath279 , and @xmath54 is non empty then construct the linkage @xmath44 as described above and go back to step ( 1 ) with linkage @xmath44 as the input .
let @xmath8 be the same linkage that we considered in example [ linkageexample ] and let @xmath49 , @xmath50 , @xmath334 , @xmath335 and @xmath336 be the sublinkages described earlier
. now observe that @xmath337 it is easy to check ( as in example [ linkageexample ] ) that @xmath338 = [ 3,4]\ ] ] moreover , @xmath339 and @xmath340 .
thus , in this case , the hypotheses of theorem [ spcontwo ] are satisfied .
the linkage @xmath44 looks like @xmath341 now , one computes that @xmath342 = [ 3,5]$ ] .
however , @xmath343 which does not meet @xmath344 $ ] .
therefore , by theorem [ spconone ] , @xmath345 is disconnected .
therefore by theorem [ spcontwo ] , @xmath15 is disconnected .
observe that the linkage @xmath44 in this example has the property that every polygonal sublinkage has connected moduli space , but that @xmath298 is disconnected .
we observe that the path lifting results described in section [ sec : lifting ] do not in general hold for linkages that are not series parallel . in @xcite
, examples are given to demonstrate this .
this is one of the reasons why series parallel linkages are easier to understand .
we also remark that it is sometimes possible to adapt our methods to understand linkages that are not series parallel
. it may be that a linkage can be series parallel decomposed into smaller linkages , which while not themselves series parallel , are amenable to analysis by other methods . in this case
our results may still have some value .
again , see @xcite for examples .
we would like to thank javier aramayona for many helpful comments and suggestions . | we study spaces of realisations of linkages ( weighted graphs ) whose underlying graph is a series parallel graph . in particular , we describe an algorithm for determining whether or not such spaces are connected . |
99 , ed . r.p . behringer and j. jenkins , balkema , rotterdam ( 1997 ) ; _ powders and grains 2001 : proc . of the fourth international conference on powders and grains _ , ed .
t. kishino , swets and zeitlinger , lisse ( 2001 ) .
* figure 1 : * evolution pictures for the coarse - grained density and velocity fields in the ics .
the upper rhs frame shows the density field at @xmath115 for @xmath116 .
regions with density greater than the average are marked in grey .
the black circles denote the vortex centers of the velocity field .
the upper lhs frame shows the direction vectors of the corresponding velocity field . for clarity ,
we only show a 32@xmath117 corner ( as depicted in the rhs frame ) of the 256@xmath117 lattice .
the middle frames are the density and velocity fields at @xmath115 for @xmath118 .
the lower rhs frame shows an evolution picture ( at @xmath119 ) for the gc - tdgl equation with @xmath120 , corresponding to average density 0.46 .
the lower lhs frame shows an evolution picture ( at @xmath119 ) for the gc - xy model with @xmath121 .
+ + * figure 2 : * dependence of @xmath67 on system parameters . the crossover time is obtained from energy - decay plots ( not shown here ) as the point of deviation from haff s law .
the error bars are defined by the symbol sizes .
( a ) plot of @xmath76 vs. @xmath67 for @xmath77 ; and @xmath18 ranging from @xmath122 to @xmath123 .
the solid line denotes the best linear fit to the data .
( b ) plot of @xmath78 vs. @xmath79 for @xmath19 ; and @xmath80 ranging from @xmath124 to @xmath125 .
+ + * figure 3 : * ( a ) scaled structure factors for the coarse - grained density field for @xmath116 .
data is plotted for @xmath126 .
( b ) analogous to ( a ) , but for @xmath118 .
+ + * figure 4 : * analogous to fig . 3 , but for the coarse - grained velocity field . | we present results from comprehensive event - driven ( ed ) simulations of nonlinear pattern formation in freely - evolving granular gases . in particular , we focus on the the morphologies of density and velocity fields in the inhomogeneous cooling state ( ics ) .
we emphasize the strong analogy between the ics morphologies and pattern formation in phase ordering systems with a globally conserved order parameter .
there has been much recent interest in the properties of powders or granular materials @xcite .
these materials are of great scientific and technological relevance and display properties intermediate to those of solids and fluids .
perhaps the most relevant feature of powders is that the grains dissipate energy on collision .
therefore , granular materials exhibit interesting dynamical properties only when the collisional energy loss is compensated by a continuous input of energy @xcite . it is also relevant to investigate the dynamical evolution of a freely - evolving homogeneous system of inelastic granular particles . in the initial stages , the system continuously loses energy in a homogeneous cooling state ( hcs ) , where the density field is approximately uniform @xcite .
however , the hcs is unstable to density fluctuations , and the system evolves into an inhomogeneous cooling state ( ics ) , where particle - rich clusters are formed and grow @xcite .
there is a good understanding of the hcs and the instabilities which result in the ics @xcite .
however , there is only a limited understanding of the nonlinear evolution of the ics @xcite . in this letter , we study nonlinear domain growth processes for the granular density and velocity fields in the asymptotic time - regime .
we present results from comprehensive event - driven ( ed ) simulations for a wide range of inelasticity and density parameters .
in particular , we invoke analogies from studies of phase ordering dynamics @xcite to obtain a _ quantitative _ characterization of the evolving morphologies in the ics .
the similarities between freely - evolving granular gases and phase ordering systems have been briefly discussed by van noije , ernst and co - authors @xcite , and baldassari _ et al .
_ @xcite .
of course , we stress that the mechanisms driving pattern evolution are very different in both cases .
we consider a system of identical inelastic hard spheres ( with mass @xmath0 and diameter @xmath1 ) in a @xmath2 box of size @xmath3 , with periodic boundary conditions .
there is no external input of energy in our simulations .
we use an event - driven algorithm @xcite , which only keeps track of particle collisions , to evolve the system . after a collision between the @xmath4 and @xmath5 particles , having velocities @xmath6 and @xmath7 respectively , the new velocities are @xmath8{\hat n},~~ \vec v_j^{\prime } = \vec v_j+\frac{1+e}{2 } [ { \hat n}\cdot ( \vec v_i-\vec v_j)]{\hat n}$ ] . here , @xmath9 is the unit vector parallel to the relative position of the particles . for elastic collisions
, we have the coefficient of restitution @xmath10 . for granular materials , @xmath11 in general .
the initial condition for each run consisted of @xmath12 particles with a homogeneous spatial distribution and a maxwellian velocity distribution .
for circular particles , the average packing fraction is @xmath13 .
we always fix @xmath14 and vary @xmath12 , as specified below .
we will label our results using the number fraction , @xmath15 .
numerical data are obtained for the following parameter sets : ( i ) @xmath16 ( or @xmath17 ) , and @xmath18 ranging from 0.80 to 0.975 in steps of 0.025 ; ( ii ) @xmath19 , and @xmath12 ranging from 10000 to 40000 in steps of 5000 , i.e. , @xmath20 to 0.61 .
we characterize the dynamical evolution of the granular gas using various statistical quantities , which are calculated as an average over 5 independent runs , i.e. , correlation functions , structure factors , and domain growth laws for the density and velocity fields . for early times
, inelastic collisions result in cooling as @xmath21 , where @xmath22 is the temperature at time @xmath23 ; @xmath24 ; and @xmath25 is the collision frequency at temperature @xmath26 @xcite .
the form of @xmath25 is approximately determined from the enskog theory for elastic hard spheres as @xmath27 , where @xmath28 is the granular temperature at @xmath29 . this yields haff s cooling law for the hcs @xcite as @xmath30^{-2 } = t_0 e^{-\frac{\epsilon}{d } \tau}$ ] , where @xmath31 is the average number of collisions per particle upto time @xmath23 .
our subsequent results are presented in terms of @xmath32 , as this constitutes a reasonable measure of time in the present context .
the values of @xmath32 are obtained directly from the ed simulations .
we should stress that @xmath31 vs. @xmath23 fluctuates considerably for individual runs in the ics regime . in fig . 1
, we show evolution snapshots for the @xmath2 granular gas with @xmath33 .
the lhs and rhs frames refer to the coarse - grained velocity field @xmath34 and density field @xmath35 , respectively .
the coarse - grained fields at a lattice point are obtained by averaging over boxes of size @xmath36 centered at that point . to clarify the nature of pattern formation
, we have `` hardened '' the velocity field in fig . 1 , i.e.
, the length of all vectors has been set to unity .
points where the velocity field is zero , due to the absence of particles in the coarse - graining box , are unmarked in fig .
1 . the density field on the rhs of fig . 1 is depicted in a binary representation .
we introduce the order parameter @xmath37 with values @xmath38 at points where the number density is larger ( less ) than the average number density . the upper frames in fig .
1 correspond to @xmath39 and @xmath40 .
the velocity field is characterized by the emergence and diffusive coarsening of vortices .
there is a progressive parallelization of the local velocity field due to inelastic dissipation of the normal velocity components .
however , as the total momentum is fixed at zero and the system is doubly - periodic , defects will be present in the velocity field for all times .
typically , pattern evolution in the density field is slower than that for the velocity field @xcite .
nevertheless , well - defined clusters corresponding to the ics are already seen at @xmath41 ( not shown here ) .
these clusters grow with time in a manner reminiscent of phase ordering systems @xcite .
the vortex centers in the velocity field are also shown ( as black circles ) in the frame on the rhs .
the vortex density diminishes with time , and vortices are primarily confined to regions of low density , as there is a rapid parallelization of velocities in the high - density ( solid - like ) region due to multiple collision processes .
there is no strong correlation between the location of defects in the velocity field ( i.e. , vortices ) and defects in the density field ( i.e. , interfaces or domain boundaries ) .
the middle frames in fig .
1 correspond to @xmath19 .
the broad features are the same as those for @xmath39 , though the time - scales are slower . all time - scales for pattern formation diverge as @xmath42 , i.e. , the clustering instability is delayed for more elastic systems .
the perfectly elastic case ( @xmath10 ) is singular , and does not exhibit clustering . a coarse - grained description of the granular fluid is provided by nonlinear hydrodynamic equations for the density , velocity and temperature fields , in conjunction with an energy loss term @xcite .
wakou _
et al .
_ @xcite have shown that fluctuations in these quantities obey tdgl - like equations of phase ordering dynamics with a nonconserved order parameter @xcite .
typically , these equations are obtained as the overdamped limit of a hamiltonian formulation . in this letter
, we clarify the tdgl - description of granular dynamics .
the evolution of the density field is analogous to phase ordering dynamics in two - component ( ab ) mixtures , which is described by the equation @xcite : @xmath43,\ ] ] where @xmath44 differentiates between @xmath45-rich ( @xmath46 ) and @xmath47-rich ( @xmath48 ) regions .
eq .
( [ che ] ) with @xmath49 is the tdgl equation , and describes nonconserved systems . in this case , the coarsening system is characterized by a diffusive growth law , @xmath50 @xcite , where @xmath51 is the characteristic domain size at time @xmath23 . on the other hand , eq .
( [ che ] ) with @xmath0 is the cahn - hilliard ( ch ) equation , and describes conserved systems . in this case , the domain growth process obeys the lifshitz - slyozov ( ls ) law , @xmath52 @xcite .
finally , there have also been studies of eq .
( [ che ] ) with @xmath53 , which is referred to as the globally - conserved ( gc ) tdgl equation @xcite .
this model is known to be in the same dynamical universality class as the nonconserved tdgl equation @xcite , with the difference that the initial composition is preserved in the gc - tdgl equation .
similarly , the evolution of the velocity field in the ics is comparable to ordering dynamics in the xy model , which is the vector version of eq .
( [ che ] ) : @xmath54,\ ] ] where @xmath55 .
the evolution in eq .
( [ xym ] ) parallelizes @xmath56 locally via the annihilation of vortices and anti - vortices , driven by a surface - tension reduction mechanism .
eq .
( [ xym ] ) with @xmath49 corresponds to the case with nonconserved order parameter .
the corresponding domain growth law @xcite is @xmath57 . on the other hand , eq .
( [ xym ] ) with @xmath0 corresponds to the conserved xy model , where the dynamics locally conserves @xmath58 .
the evolution of the granular fluid conserves both the density and velocity fields .
intuitively , one may expect that the analogous phase ordering models would be the conserved ( @xmath0 ) versions of eqs .
( [ che])-([xym ] ) . however , the evolution morphologies in the upper and middle frames of fig .
1 are comparable to those for the case with nonconserved order parameter @xcite .
this is a consequence of the non - diffusive dynamics of granular particles , which move in straight lines until they collide with other particles . in this collision ,
the density and momentum are conserved quantities , of course . however , depending on the density of the fluid , the distance traveled by particles prior to collision may be considerable .
therefore , the variables are conserved on the macroscopic length - scale of the mean - free path , and not on the microscopic length - scale of the lattice spacing .
now , the ics consists of regions of high density and low density .
granular particles stream relatively unhindered through the low - density regions and deposit on distant clusters . typically , the conservation length scale is comparable to the length scale of the coarsening clusters , which diverges in time .
therefore , the density and velocity fields are globally conserved ( gc ) , rather than locally conserved . to elucidate this analogy , the lower lhs frame in fig .
1 shows the velocity field from the gc - xy model with @xmath59 .
the lower rhs frame in fig .
1 shows the density field for the gc - tdgl model with @xmath60 , corresponding to an average density of @xmath61 . before we quantify the morphologies in fig . 1
, we would like to study the hcs @xmath62 ics crossover time .
brito and ernst @xcite use mode - coupling techniques to obtain the asymptotic energy decay ( in the ics ) as @xmath63 where @xmath64 and @xmath65 , with @xmath66 being the time - independent mean - free path .
a comparison of haff s law and eq .
( [ mode ] ) yields the crossover time @xmath67 as the solution of @xmath68 where @xmath69 , and @xmath70 is the @xmath71-dimensional solid angle .
we consider the case @xmath72 , so that @xmath73 . for @xmath2 ,
eq .
( [ tauc ] ) simplifies as @xmath74 , where we ignore prefactors .
we use the verlet - levesque approximation for the @xmath2 hard - sphere correlation function at contact , viz . , @xmath75 . in fig .
2(a ) , we plot @xmath76 vs. @xmath67 for @xmath77 and a range of @xmath18-values . in fig .
2(b ) , we plot @xmath78 vs. @xmath79 for @xmath19 and a range of @xmath80-values .
these approximately linear plots confirm the validity of the scaling behavior of @xmath81 , as expected from eq .
( [ tauc ] ) .
next , we examine the time - dependent structure factors of the density and velocity fields in the ics . the evolving morphologies are characterized by unique length scales , and we expect the structure factors to exhibit dynamical scaling , i.e. , @xmath82 , where @xmath83 is the relevant length scale and @xmath84 is a scaling function @xcite . in fig .
3(a ) , we plot the scaled structure factors of the density field for @xmath85 , i.e. , @xmath86 $ ] vs. @xmath87 from three different times ( @xmath88 ) .
the quantity @xmath89 is the first moment of the structure factor , and is related to the length scale as @xmath90 .
the reasonable data collapse confirms the validity of dynamical scaling . the dashed line in fig .
3(a ) denotes the fourier transform of the analytic result due to ohta _
et al . _ ( ojk ) @xcite for the correlation function of the nonconserved tdgl equation : @xmath91 , where @xmath92 .
the ojk function provides an excellent description of our numerical data . the dot - dashed line in fig .
3(a ) is obtained from a numerical simulation of the ch equation with @xmath93 .
the ch result is at variance with our numerical results except for the tail region .
in particular , the local conservation law dictates that @xmath94 as @xmath95 @xcite .
however , the ics structure factor appears to decay almost monotonically from @xmath96 , as is usual for structure factors in ordering problems characterized by a nonconserved order parameter .
( of course , global density conservation dictates that @xmath97 . )
fig .
3(b ) is analogous to 3(a ) , but corresponds to the case with @xmath98 .
we have confirmed numerically that the scaled structure factor for the density field in the ics has only a weak dependence on the off - criticality @xmath99 , which measures the average density .
furthermore , it is numerically comparable to the ojk function for a broad range of @xmath99-values .
the nature of defects in the ordering field dictates some general properties of the structure factor .
for example , scattering off interfaces in the density field gives rise to a power - law ( or porod ) decay in the structure factor tail , @xmath100 for large @xmath101 @xcite , which is seen clearly in figs .
3(a)-(b ) .
in figs .
4(a)-(b ) , we plot scaled structure factors , @xmath102 $ ] vs. @xmath87 , for the velocity field in the ics .
the dashed line denotes the fourier transform of the bray - puri - toyoki ( bpt ) function for ordering in the nonconserved xy model @xcite : @xmath103 ^ 2 f\left(\frac{1}{2},\frac{1}{2};\frac{n+2}{2};\gamma^2\right),\ ] ] where @xmath104 ; @xmath105 is the beta function ; and @xmath106 is the hypergeometric function .
the bpt result corresponds to defects with @xmath107 symmetry , and the case with @xmath108 is relevant here .
again , we see that our numerical data is described well by the bpt function , and is completely different from the result for the conserved xy model ( obtained numerically , and denoted as a dot - dashed line in figs .
4(a)-(b ) ) .
we have seen that vortex defects characterize the morphology of the ics velocity field .
these vortices give rise to a power - law decay for the structure factor tail , @xmath109 for large @xmath101 .
the general bpt result in eq .
( [ bpt ] ) exhibits a power - law tail , @xmath110 , and is relevant for the ics in granular systems with @xmath111 . for the time - regimes considered here ,
a description in terms of two uncoupled order parameters is reasonable .
a more complete description of the asymptotic ordering dynamics should account for _ holes _ in the velocity field due to absence of particles in some regions of space .
this is provided by a model with spin - vacancy phase separation in conjunction with xy - like ordering of the spin variable . in the current context
, the relevant model has coupled nonconserved ( or globally conserved ) dynamics for the two ordering fields . at a later stage
, we will discuss in detail the applicability of this model to the ics .
finally , we would like to briefly discuss domain growth laws in the ics . we should stress that late - stage dynamics is affected by inelastic collapse , where a group of particles undergo an infinite number of collisions in a finite time @xcite .
the results discussed here correspond to the regime prior to inelastic collapse .
subsequent to the hcs @xmath62 ics crossover , our numerical results ( not shown here ) are consistent with diffusive growth , @xmath112 @xcite .
furthermore , our numerical data is consistent with an asymptotic power - law behavior , @xmath113 , for a wide range of parameter values .
therefore , in real time , the length scales behave as @xmath114 @xcite .
the domain growth law for the ics density field is the same as the ls growth law for phase separation of binary mixtures . in conclusion
, we have undertaken comprehensive ed simulations of nonlinear pattern formation in the density and velocity fields of an inelastic granular gas .
we find that there is a close analogy between ics morphologies and phase ordering systems with globally conserved order parameters .
the nature of defects in the ordering fields dictates general properties of the relevant structure factors and correlation functions .
furthermore , our numerical results for the structure factors of the ics density and velocity fields are described well by analytic results for nonconserved phase ordering systems . clearly , the general formalism of phase ordering dynamics is of great utility in diverse problems of pattern formation as many features of evolution morphologies are determined by general principles , e.g. , defect structures , conservation laws , etc .
+ + + * acknowledgements * sp is grateful to m.h .
ernst , i. goldhirsch , h. hayakawa , s. luding , u.m .- b .
marconi and h. nakanishi for helpful discussions and critical inputs on this problem .
the authors are grateful to i. goldhirsch and s. luding for their critical reading of this manuscript . |
since the discovery of the light exotic hypernuclei in the early sixties [ 1 - 2 ] , peoples are giving much attention to study the structure of such multistrange hypernuclei [ 3 - 4 ] .
some of the experimentally observed one- and two-@xmath0 hypernuclei are @xmath4he , @xmath5he , @xmath6be , @xmath7be , @xmath8c and @xmath9b [ 5 - 11 ] .
information about the @xmath2 and @xmath3 forces can be extracted from the bound state properties of such nuclei .
these informations on @xmath2 and @xmath3 interaction together with the well known @xmath10 interaction may give better understanding of the strong quark - quark ( qq ) interaction , since the hyperons also have qqq structure as that of nucleons .
as for example proton has three quarks namely up , up , down ( @xmath11 ) , neutron has quarks up , down , down ( @xmath12 ) , @xmath13 has quarks up , down , strange ( @xmath14 ) etc . in the earlier stage of their discovery , emulsion experiments provided some information on the binding energies of @xmath0-particle in the light exotic hypernuclei and their decay rates ( life times@xmath15 ) [ 2 ] . from
the available binding energy data physicists gathered some qualitative informations about the @xmath0-nucleon ( @xmath3 ) interaction and single particle potential strength for the @xmath0-particle in hypernuclei [ 12 ] .
the hyperon - nucleon scattering experiments have also been performed but these are not so profound and are still in the primary stages and do not give detailed phase shifts to construct the potential reliably .
some @xmath3 and @xmath16 total cross - sections and very few angular distribution at low energies have been measured [ 13 - 18 ] , but these are not sufficient to allow the phase shift analysis . under the circumstances , the bound state properties of single- and double- @xmath0 hypernuclei can only give useful indirect information about @xmath3 and @xmath2 interactions .
for example , one can take phenomenological forms of @xmath3 and @xmath2 interactions and see if they reproduce the experimentally observed observables of the hypernuclei .
alternatively one can adjust the parameters of the empirical potential to reproduce the bound state properties and thus predict the effective @xmath3 and @xmath2 interactions .
in some earlier works [ 19 - 22 ] physicists have used variational and approximate few body methods for the hypernuclei treating them a few body system . in the present work
, we investigate the structure and properties of the ground and excited states of double-@xmath0 hypernuclei with mass number @xmath17 and ground state of their binary subsystems consisting the core and one outer core @xmath0 hyperon .
a simple two - body model has been adopted here for the single-@xmath0 hypernucleus ( @xmath18 ) consisting of a @xmath19 core and one valence @xmath0 particle . and for the double-@xmath0 hypernuclei ( @xmath20 )
, we adopt a three - body model consisting core of mass @xmath21 ( = 4,6,30,40,90 ) and two valence @xmath0 particles . for @xmath2 potential
a phenomenological form has been chosen .
the core-@xmath0 potential has been obtained by folding an effective @xmath3 potential by the density of the core nucleus .
the strength of the effective @xmath3 potential for a given core is adjusted to reproduce the experimentally known be of the single @xmath0 hypernuclei @xmath4he , @xmath22he [ 6 ] , @xmath22be [ 23 ] , @xmath23s [ 24 ] , @xmath24ca [ 24 ] and @xmath25zr [ 25 ] .
the same folded core-@xmath0 potential is then used for the double-@xmath0 hypernucleus for the same core .
+ we used hyperspherical harmonics expansion ( hhe ) method to solve the three - body system .
this method is a powerful tool for the @xmath26 solution of the few body schrdinger equation for a given set of binary interaction potentials among the constituent particles .
this method has been used for bound states in atomic [ 26 - 33 ] , nuclear [ 34 - 40 ] and particle physics [ 41 - 43 ] .
attempts have been made to use it in scattering problems as well [ 44 ] . in this method
, the wave function is expanded in a complete set of hyperspherical harmonics ( hh ) , which are , for a three - body system , the six - dimensional analogue of ordinary spherical harmonics .
the schrdinger equation reduces to a set of coupled differential equations which can be solved numerically .
the hhe method is essentially exact and involves no approximation other than an eventual truncation of the expansion basis .
any desired precision in the binding energy can , in principle , be achieved by gradually expanding the expansion basis and checking the rate of convergence . however
the number of coupled differential equations and hence , the complexity in the numerical solution increases rapidly as the expansion basis is increased by introducing larger hyper angular momentum quantum numbers .
computer limitations set an ultimate limit to the precision attainable .
+ we have calculated the two-@xmath0 separation energy ( @xmath27 ) , @xmath2 bond energy ( @xmath28 ) and some size parameters for all the above double @xmath0 hypernuclei . here
@xmath27 and @xmath2 bond energy @xmath29 are defined as @xmath30 c^{2}\ ] ] and @xmath31 as the ground state be is quite large , one may expect the excited states to be observable experimentally , although they have not been reported so far . however , the hhe method is known for its rather slow convergence , which is especially manifested in the excited states .
the excited state wave function is more spread out than the ground state and therefore needs larger @xmath32 for the same degree of convergence .
we circumvent this difficulty by invoking supersymmetric quantum mechanics ( ssqm ) [ 45 ] .
in ssqm , we construct the supersymmetric partner potential ( @xmath33 ) of the original potential ( @xmath34 ) [ 1 ] . according to ssqm ,
the spectra of @xmath34 and @xmath33 are identical except for the absence of the ground state of @xmath34 in the spectrum of @xmath33 [ 45 ] .
hence the ground state of @xmath33 corresponds to the first excited state of @xmath34 .
thus we can solve for the ground state of @xmath33 , whose convergence is fast ( being a ground state ) .
its be is the be of the first excited state of the original potential .
the wave function of the first excited state of @xmath34 can be obtained by applying an appropriate operator on the ground state wave function of @xmath33 .
+ the paper is organized as follows : in section ii , we review the hhe method for a three - body system consisting of non identical particles with a brief review of ssqm .
results of calculation and discussion are presented in section iii . finally in section
iv we draw our conclusions .
we treat each of the six double-@xmath0 hypernuclei @xmath35he , @xmath36be , @xmath36he , @xmath37s , @xmath38ca , @xmath39zr as a three body system , where each of the core @xmath40he , @xmath41be , @xmath41he , @xmath42s , @xmath43ca , @xmath44zr is labeled as particle 1 and the two valence @xmath0 hyperons as particles 2 and 3 respectively ( see fig .
1 ) . for a given partition @xmath45 ( in which @xmath45 is the spectator )
, a set of jacobi co - ordinates is defined as : @xmath46^{\frac{1}{4 } } ( \vec{r_{j } } - \vec{r_{k } } ) \\
\vec{y_{i } } & = & \left[\frac{m_{i}(m_{j}+m_{k})^{2}}{m_{j}m_{k}m } \right]^{\frac{1}{4 } } \left(\vec{r_{i } } - \frac{m_{j } \vec{r_{j } } + m_{k } \vec{r_{k } } } { m_{j } + m_{k } } \right ) \\
\vec{r } & = & \frac{1}{m } \left(m_{i } \vec{r_{i } } + m_{j } \vec{r_{j } } + m_{k } \vec{r_{k } } \right ) .
\end{array } \right\}\ ] ] for ( i , j , k)=(1,2,3 cyclic ) .
here @xmath47 , @xmath48 are the mass and position of the @xmath49 particle and @xmath50 .
@xmath51 is the coordinate of the centre of mass of the system .
since the interactions among the constituent particles depend only on their relative separations , the centre of mass motion separates out automatically .
thus the relative motion of the three - body system is described by the schrdinger equation @xmath52 \psi ( \vec{x_{i } } , \vec{y_{i } } ) \:= \ : 0\ ] ] where @xmath53^{\frac{1}{2}}}$ ] is an effective mass parameter and @xmath54 is the interaction potential between particles @xmath45 and @xmath55 . in terms of the hyperspherical variables [ 40 ]
@xmath56 the schrdinger equation becomes @xmath57 \psi ( \rho , \omega_{i } ) \:=\ : 0,\ ] ] where the hyper - radius @xmath58 is invariant under three - dimensional rotations and permutations of the particle indices .
the five hyper angles constituted by polar angles ( @xmath59 of @xmath60 ) , polar angles ( @xmath61 ) of @xmath62 and the angle @xmath63 , are collectively denoted by @xmath64 .
thus ( @xmath65 ) constitutes the six hyperspherical variables of which the five hyper angles @xmath66 depend on the choice of partition @xmath45 . in the above equation @xmath67 = @xmath68
is the total interaction potential expressed in terms of the hyperspherical variables in the partition @xmath45 and @xmath69 is the square of hyper angular momentum operator given by [ 40 ] @xmath70 where @xmath71 and @xmath72 are the squares of ordinary orbital angular momentum operators associated with @xmath73 and @xmath74 motions respectively .
the operator @xmath75 satisfies an eigenvalue equation [ 40 ] @xmath76 here k ( @xmath77 being a non - negative integer ) represents the hyper angular momentum quantum number and @xmath78 is a short hand notation , in which l and m denote the total orbital angular momentum and its projection , respectively .
the normalized eigen functions @xmath79 are called the hyperspherical harmonics ( hh ) and these depend on the choice of partition .
a detailed analytic expression for @xmath79 can be found in [ 40 ] . in the hhe method , @xmath80 is expanded in the complete set of hh corresponding to the partition @xmath45 : @xmath81 substitution of eq .
( 9 ) in eq .
( 6 ) and the use of orthonormality of hh leads to a set of coupled differential equations ( cde ) in @xmath58 @xmath82 u_{k \alpha_{i}}(\rho ) \\ + & \sum_{k^{\prime } \alpha_{i}^{~\prime } } < k \alpha_{i } \mid v(\rho , \omega_{i } ) \mid k^{\prime } \alpha_{i}^{~\prime } > u_{k^{\prime } \alpha_{i}^{~\prime}}(\rho ) \ : = \ : 0 , \end{array}\ ] ] where @xmath83 and @xmath84 the number of coupled differential equations to be solved , eq .
( 10 ) is in principle , an infinite one , which arises out of an infinite number of basis states eq .
( 9 ) . for practical computation , the hyperspherical harmonics expansion basis in eq .
( 9 ) is truncated to a finite set by retaining all values of k up to a maximum @xmath32 . for a given k , all allowed values of the set of quantum numbers @xmath85
are included .
the basis states is further reduced by imposing the constraints arising out of symmetry requirements and associated conserved quantum numbers .
thus the infinite set of cde is now reduced to a finite set . this finite set of cde
is solved numerically by hyperspherical adiabatic approximation [ 31,46 ] which is based on the idea of adiabatic separation of hyper - angular motions from the hyper - radial one . in this procedure
the potential matrix @xmath86 together with the diagonal hyper - centrifugal term is diagonalized for a fixed value of @xmath58 .
the lowest eigenvalue @xmath87 and the corresponding eigenvector @xmath88 are obtained as parametric function of @xmath58 .
then the set of cde , eq .
( 10 ) is approximately decoupled into a single ordinary differential equation @xmath89 \psi(\rho ) = 0\ ] ] numerical solution of eq .
( 12 ) , subject to appropriate boundary conditions gives the ground state energy e and partial waves of the three - body system .
the partial waves are given by @xmath90 we now present a brief review of properties of supersymmetric quantum mechanics ( ssqm ) [ 45 ] used here to calculate the observables for the excited states . let us consider a one dimensional schrdinger equation of the form @xmath91 \psi(\rho ) = 0\ ] ] for the single-@xmath0 hypernuclei , @xmath92 is just the @xmath0-core potential plus the centrifugal term and @xmath58 is the core-@xmath0 separation . for double-@xmath0 hypernuclei
@xmath92 is the lowest eigen potential @xmath87 and @xmath58 is the hyper - radius .
suppose the ground state energy is @xmath93 .
we shift the energy scale by @xmath93 , such that the ground state is at zero energy and the corresponding potential is renamed @xmath94 ( = @xmath95 ) .
then for the ground state @xmath96 \psi_{0}(\rho ) = 0\ ] ] from this we get @xmath97 let us now define a super potential , @xmath98 as @xmath99 and two operators @xmath100 then one can easily check that @xmath101 and @xmath102 here @xmath103 is the original hamiltonian in the shifted energy scale .
we next define a partner hamiltonian @xmath104 in which the partner potential is given by @xmath105 let us denote the @xmath106 eigenvalue and corresponding eigen function of @xmath107 by @xmath108 and @xmath109 respectively , while the @xmath106 eigenvalue and corresponding eigenfunction of @xmath103 are denoted by @xmath110 and @xmath111 respectively .
then it is easily verified that @xmath112 which means that @xmath113 is an eigenfunction of @xmath107 corresponding to energy @xmath110 .
similarly @xmath114 thus @xmath115 is an eigenfunction of @xmath103 corresponding to energy @xmath108 . now from eq .
( 17 ) and ( 18 ) @xmath116 thus @xmath117 is a trivial solution of eq .
the ground state of @xmath107 will thus be @xmath118 corresponding to energy @xmath119 ( note that @xmath120 ) .
hence @xmath121 , and from eq . ( 24 ) , @xmath122.thus if we solve for the ground state of @xmath123 , we get the energy @xmath119 ( in shifted energy scale ) , which is the energy of the first excited state of @xmath94 .
it is easily seen that the corresponding normalized eigenfunction is @xmath124 hence we numerically solve eq .
( 14 ) for the ground state and calculate the super - potential @xmath98 and the partner potential @xmath123 according to eq .
( 17 ) and ( 22 ) respectively
. then we solve the same schrdinger equation with @xmath123 to obtain @xmath125 and @xmath126 . from @xmath126 , we get @xmath127 using eq(26 ) . finally the energy of the first excited state of @xmath94 is obtained by back shifting @xmath125 .
in the present calculation we have taken the core to be structureless . since the core ( @xmath40he , @xmath41be , @xmath128he , @xmath42s , @xmath43ca , @xmath44zr ) contains only nucleons and no @xmath0-particles , there is no symmetry requirements under exchange of the valence @xmath0 particles with the core nucleons .
the only symmetry requirements are ( i ) anti - symmetrization of the core wave function under exchange of the nucleons and ( ii ) anti - symmetrization of the three body wave function under exchange of the two @xmath0 particles .
the former is implicitly taken care of by the choice of the core as a building block . for double-@xmath0 hypernuclei ,
the latter is correctly incorporated by restricting the @xmath129 values , explained in the following .
thus , within the three - body model , the symmetry requirements are correctly satisfied without any approximation . for the single-@xmath0 hypernuclei
, there are no symmetry requirements .
the ground state of all experimentally known double-@xmath0 hypernuclei have a total angular momentum quantum number @xmath130 and even ( or positive ) parity .
we assume this to be true for all double-@xmath0 hypernuclei with core having @xmath131 .
the possible total spin ( s ) of the three - body system @xmath132 can take two values 0 or 1 since the spin of the core in all the above cases has a value 0 .
thus the total orbital angular momentum l can be either 0 or 1 corresponding to s=0 or 1 respectively .
hence the set of quantum numbers ( ls)j for the ground state of all even - even core double-@xmath0 hypernuclei is ( 00)0 and ( 11)0 which corresponds to @xmath133 and @xmath134 states .
since the ground state of these nuclei have definite parity , odd and even parity states will not mix . in our case ,
the ground state of all the six double-@xmath0 hypernuclei is a pure @xmath133 state .
since the core is spin less , the spin singlet state ( s=0 ) corresponds to zero total spin of the valence @xmath0- particles ( i.e. @xmath135=0 ) .
hence the spin part of the wave function is antisymmetric under the exchange of the spins of the two @xmath0-particles .
thus the spatial part must be symmetric under the exchange of the two @xmath0-hyperons .
the symmetry of the spatial part is determined by the hyper spherical harmonics , since the hyper radius @xmath58 and hence the hyper radial partial waves ( @xmath136 ) are invariant under permutation of the particles . under the pair exchange operator @xmath137 which interchanges particles 2 and 3 , @xmath138 and @xmath139 remains unchanged ( see eq .
consequently @xmath137 acts like the parity operator for ( 23 ) pair only .
choosing the two valence @xmath0-hyperons to be in spin singlet state ( spin antisymmetric ) , the space wave function must be symmetric under @xmath137 .
this then requires @xmath129 to be even .
for the spin singlet state total orbital angular momentum , l=0 , hence we must have @xmath140 = even integer .
since @xmath141 , where @xmath142 is a non - negative integer , @xmath143 must be even and @xmath144 again for the triplet state ( s=1 ) , the two valence @xmath0-hyperons will be in spin triplet state ( @xmath145 , spin symmetric ) .
hence the space wave function must be antisymmetric under @xmath137 .
this then requires @xmath129 to be odd . for the spin triplet state , the total orbital angular momentum , l=1 , hence @xmath146 may take values @xmath129 and @xmath147 ( and only 1 if @xmath148 ) but the parity conservation allows @xmath146=@xmath129 only ( except @xmath148 ) . again since @xmath141 , where @xmath142 is a non - negative integer
, k must be even ( @xmath149 ) and @xmath150 for a practical calculation , the hh expansion basis ( eq(9 ) ) is truncated to a maximum value ( @xmath32 ) of k. for each allowed k @xmath151 with k = even integers , all allowed values of @xmath129 are included .
the even values of @xmath129 correspond to l=0 , s=0 and odd values of @xmath129 correspond to l=1 , s=1 .
this truncates eq(10 ) to a set of n coupled differential equations , where @xmath152 which will be solved by the hyperspherical adiabatic approximation ( haa ) [ 46 ] . + + since realistic @xmath2 potential is not available at this stage .
we used phenomenological three term gaussian @xmath2 potential with parameters adjusted to reproduce the experimental data for known double-@xmath0 hypernuclei .
the @xmath2 potential used here is given by @xmath153 the parameters of this potential are presented in table i. a short range repulsive term in the @xmath2 potential is included to simulate pauli principle between the valence @xmath0 particles .
the core-@xmath0 potential is obtained by folding phenomenological @xmath0-nucleon potential into the density distribution of the core nucleus .
the chosen density distribution function of the core is given by @xmath154 with @xmath155 fm , @xmath156 , @xmath157 ( where @xmath158 is termed the half density radius and a , the skin thickness ) .
the the density constant @xmath159 is determined by the condition @xmath160 where @xmath21 is the mass of the core in units of nucleon mass .
the value of @xmath161 and @xmath162 are chosen following suggestions in the literature [ 10 ] .
the phenomenological @xmath3 potential is given by @xmath163 with @xmath164 and @xmath165 adjusted to reproduce the experimental binding energy ( @xmath166 ) of single-@xmath0 hypernuclei ( see table ii ) in the core-@xmath0 subsystem .
then the core-@xmath0 potential is given by @xmath167 a two dimensional plot of the @xmath2 , effective @xmath3 and core-@xmath0 folded potential are shown in fig .
the lowest eigen values ( @xmath168 ) of the three body effective potential is plotted against the hyper - radial distance @xmath58 as shown in fig .
+ + the strength of @xmath3 potential is expected to be weakened with the increase in mass of the core due to the screening or shielding effect by neighboring nucleons within the core when the interacting nucleon is embedded in the core .
the @xmath169-mesic decay of @xmath0 hyperon ( @xmath170 ) is predominant in the free space but tends to be suppressed in hypernucleus by the pauli - exclusion principle and instead nonmesic weak process ( @xmath171 ) becomes dominant with increasing mass number [ 7,47 - 51 ] .
thus we actually get an effective @xmath3 interaction by the folding process .
the parameters of this effective @xmath3 potential are listed in table ii .
the variation of the depth of the effective @xmath3 interaction against the core mass ( @xmath172 ) is displayed in fig .
+ + the truncated set of cde for a given @xmath32 takes the form @xmath173
u_{k l_{x_{1}}ls}(\rho ) \\ + & \sum_{k^{\prime}=0 , 2 , .... } ^{k_{max } } \sum_{l_{x_{1}}^{\prime}(allowed ) } < k l_{x_{1 } } \mid v(\rho , \omega_{1 } ) \mid k^{\prime } l_{x_{1}}^{~\prime } > u_{k^{\prime } l_{x_{1}}^{~\prime } l s}(\rho ) \ : = \ : 0 \end{array}\ ] ] ( allowed @xmath174=0 , 2 , ... only for s=0 , l=0 ) .
note that the subscripts @xmath175 ( = @xmath129 ) or @xmath176 ( = @xmath174 ) have been suppressed for brevity .
( 12 ) is solved to get the ground state energy @xmath177 and wave functions @xmath178 , by the hyperspherical adiabatic expansion ( haa ) .
for the single-@xmath0 hypernuclei , we solve eq .
( 14 ) with @xmath92 replaced by @xmath179 plus centrifugal potential .
the results are presented in the last two columns of table ii . for double-@xmath0 hypernuclei , we use the same @xmath180 and solve eq .
( 12 ) for the ground state . then following the prescription of sec .
ii , we obtain the energy and wavefunction of the higher excited states . the ground and excited state wave functions for @xmath35he , @xmath38ca , and @xmath39zr are respectively shown in fig . 5 , 6 and 7 . the calculated values of two-@xmath0 separation energy ( @xmath27 ) at @xmath181 for the ground state ( ie .
@xmath182 ) , different excited states ( ie .
@xmath183 , etc ) and @xmath2 bond energy ( @xmath184 ) for the ground state of @xmath5he , @xmath185be , @xmath185he , @xmath37s , @xmath38ca , @xmath39zr are presented in table iii .
the variation of the single-@xmath0 separation energy @xmath166 with the mass of the core @xmath172 is shown in fig .
8 . on the same graph ,
the variation of the two-@xmath0 separation energy @xmath166 with the mass @xmath186 is also plotted .
the calculated two-@xmath0 separation energy ( @xmath27 ) and @xmath2 bond energy ( @xmath28 ) of @xmath5he are in agreement with the experimental value [email protected] @xmath188 [ 7 ] and 1.01 @xmath188 [ 7 ] respectively .
the predicted two-@xmath0 separation energies for the ground and excited states of all the six double@xmath0 hypernuclei considered here are depicted in the energy level diagram ( see fig .
9 ) . after obtaining the three body wave function by the hhe approach , root mean
square ( rms ) matter radius ( @xmath189 ) of the three - body system has been calculated following the equation @xmath190^{1/2},\ ] ] where@xmath21 , @xmath191 are the masses of the core and the @xmath0 hyperon ( in units of nucleon mass ) and @xmath192 is the matter radius of the core determined by the relation @xmath193 with @xmath194 .
the rms core-@xmath0 separation is defined as @xmath195^{1/2}.\ ] ] the expectation value of the observable @xmath196 is given by the expression @xmath197
cos^2\phi sin^2\phi d\phi .
\end{array}\ ] ] the rms separation between the valence @xmath0-hyperons @xmath198 is given by the expression @xmath199^{1/2},\ ] ] where @xmath200 the result of these calculations are recorded in table iv .
the variation of the ground state rms matter radius ( @xmath189 ) with mass number ( a ) of the double-@xmath0 hyper nuclei is displayed in fig .
finally we have calculated the contribution of various orbital angular momenta @xmath129 at @xmath181 to the ground state probability density distribution to identify the dominating component of the partial waves , and the results are presented in table v.
since hyperons and nucleons both have three quark ( qqq ) structure ( eg .
@xmath11 , @xmath12 , @xmath201 etc . )
interactions among hyperons as well as with nucleons should give valuable inputs to the knowledge of strong interactions .
but not much attention has so far been directed to the study of hyperon - hyperon and hyperon - nucleon interaction through the critical investigation of hypernuclei .
we have undertaken a systematic study of the bound state properties of few light and medium mass hypernuclei to gather some idea about nature of hyperon - hyperon and hyperon - nucleon interactions and the internal structure of the three - body systems .
the hyperspherical harmonics expansion ( hhe ) method adopted here is an essentially exact method , where calculations can be carried out up to any desired precision by gradually increasing the expansion basis . for @xmath5he , the calculated two-@xmath0 separation energy @xmath27 and @xmath2 bond energy @xmath184 ( see table iii )
both at @xmath181 are well within the experimental error limits @xmath202 [ 7 ] and @xmath203 [ 7 ] .
both values are less than our previous calculation [ 52 ] done by using old data of [ 5 ] .
the strength of the core-@xmath0 potential for a fixed range parameter ( @xmath165 ) , is found to become weaker as the mass of the core increases . and this trend is found to fit a logarithmic function of the form @xmath204 with @xmath205 and @xmath206 . in the low mass region
@xmath207 decreases faster than in the mid mass range before achieving saturation in the heavier mass region .
this phenomenon may be viewed as the effect of screening of the interacting nucleon embedded in the core , by the surrounding nucleons . and
this effect of screening increases as the number of neighboring nucleons increases .
one may expect such a result from physical intuition . following the above empirical formula one can estimate the depth of the effective @xmath3 potential for the single-@xmath0 hypernuclei in any mass range .
the single @xmath0 separation energy @xmath166 is found to increase with the increase in core mass approximately following the equation @xmath208 with @xmath209 and @xmath210 .
similarly , the trend of variation of the two-@xmath0 separation is found to follow the equation @xmath211 with @xmath212 and @xmath213 .
( 42 ) and ( 43 ) may respectively be useful in determining @xmath166 and @xmath27 for any one- and two @xmath0 hypernuclei .
also the approximate size of any double @xmath0 hypernuclei of mass number ( a ) can be estimated by our empirical formula @xmath214 with @xmath215 and @xmath216 .
it can be seen from energy level diagram ( see fig .
9 ) that the energy levels for the exotic isobars @xmath217he and @xmath217be are almost identical . however
the slight difference in their one- and two-@xmath0 separation energies ( see column 4 of table ii and iii ) is due to the relatively stronger coulomb repulsion in the core of the latter .
the calculated values of the partial probability ( see table v ) shows that larger contribution to the ground state total wave function comes from the @xmath148 partial wave and the rest cones from the higher even @xmath129 components . the contribution from the odd
@xmath129 values are zero . from the calculated values of the single- and double- @xmath0 separation energies ( see table ii & iii ) and the rms matter radii ( see table iv ) it may be concluded that the addition of one- or two-@xmath0 particle to the core makes the system strongly bound and more compact .
this is quite true as the weakly bound @xmath218 or unbound @xmath219 core becomes strongly bound on addition of one or two @xmath0 hyeperon(s ) .
hence , it is quite possible that our predictions for the binding energies of the ground and excited states of experimentally unobserved double @xmath0 hypernuclei are valid .
thus we may say that that the @xmath3 force is much stronger and much attractive than the @xmath10 force .
perhaps this may be due to the presence of the strange quark in the @xmath0 hyperon quark constituents .
+ * acknowledgments * the author gratefully acknowledges computer facilities provided by the department of physics , aliah university
. references : m. danysz et al .
, nucl . phys . *
49 * , 121(1963 ) .
d. j. prowse , phys . rev . lett .
* 17 * , 782(1966 ) .
r. h. dalitz , d. h. davis , p. h. fowler , a. montwill , j. poriewski , and j. a. zakrzewski , proc .
london , ser . *
a426 * , 1(1989 ) .
d. h. davis , nucl .
a754 * 3c(2005 ) ; r. h. dalitz , nucl . phys . *
a754 * 4c(2005 ) .
h. himeno et al . , prog .
89 * , no .
1 , 109(1993 ) . l. majling , nucl . phys . *
a585 * , 211c(1995 ) .
takahashi et .
al . , phys . rev .
lett . * 87 * , 212502(2001 ) .
s. aoki et al . , prog .
* 85 * , 1287(1991 ) .
h. nemura , y. suzuki , y. fuziwara and c. nakamoto , prog .
105 * , no . 5 , 929(2000 ) .
r. r. roy and b. p. nigam , nuclear physics theory and experiment , fifth wiley eastern reprint , february ( 1993)pp-389 ; j. caro et al . , nucl . phys . *
a646 * , 299(1999 ) . y. yamamoto et al .
, nucl . phys . *
a547 * , 233c(1992 ) .
a. gal , advances in nucl .
phys . , eds . m. baranger and e. vogt .
* vol.8 * , 1(1975 , plenum ) .
g. alexander et al .
, phys . rev . *
173 * , 1452(1968 ) .
b. sechi - zorn et al .
, phys . rev . *
175 * , 1735(1968 ) .
j. a. kadyk et al . , nucl . phys . * b27 * , 13(1971 )
. j. m. hauptman , lbl report no .
* lbl-3608*(1974 ) .
f. eisele et al . , phys . lett .
* 37b * , 204(1971 ) .
r. engelmann et al . , phys . lett . * 21 * , 487(1966 ) .
y. c. tang and r. c. herndon , nuovo cimento * xlvib * , n.1 . , 117(1966 ) .
r. h. dalitz and g. rajasekaran , nucl .
phys . * 50 * , 450(1964 ) .
s. ali and a. r. bodmer , phys .
lett . * 24b * , 343(1967 ) .
a. r. bodmer , q. n. usmani , and j. carlson , nucl .
phys . * a422 * , 510(1984 ) .
h. bando , t. motoba and z. j. zofka , int . j. mod .
a5 * , 4021(1990 ) ; see references therein .
g. a. lalazissis , m. e. grypeos , and s. e. massew , phys . rev . *
c37 * , 2098(1988 ) . y. yamamoto , t. motoba , h. himeno , k. ikeda , and s. nagata , prog .
suppl . * 117 * , 361(1994 ) . c. d. lin , phys
. rep . * 257 * , 1(1995 ) .
t. h. gronwall , phys . rev . * 51 * , 655(1937 ) .
j. macek , j. phys .
* b 1 * , 831(1968 ) . j. l. ballot and j. navarro , j. phys.*b 8 * , 172(1975 ) . v. d. efros , a. m. frolov and m. i. mukhtarova , j. phys.*b 15 * , 1819(1982 ) .
t. k. das , r. chattopadhyay and p. k. mukherjee , phys .
, 3521(1994 ) .
r. chattopadhyay and t. k. das , phys . rev . * a56 * , 1281(1997 ) .
a. khan , s. k. datta , and t. k. das , fizika b ( zagreb)*8 * , 4 , 469(1999 ) .
t. k. das , h. t. coelho and m. fabre de la ripelle , phys .
c26 * , 2288(1982 ) .
khan m. a. , dutta s. k. , das t. k. and pal m. k. , j. phys .
g : nucl . part .
* 24 * , 1519(1998 ) .
a. simonov , yad . fiz .
* 3 * , 630(1960 ) [ sov .
. phys . * 3 * , 461(1960 ) ] ; in proced . of the international symposium on the present status and
novel developments in the nuclear many body problem .
rome , 1972 , edited by f. calogena and c. ciofi degli atti ( editrice composition , bologna , 1973 ) , p527 ; sov .
* 7 * , 722(1968 ) .
f. zernike , h. c. brinkman , proc .
wtensch , * 33 * , 3(1975 ) . m. fabre de la ripelle , proc . int
phys . , predeal 1969
. m. fabre de la ripelle , comp .
, * 269b * , 80(1970 ) ; * 273a * , 1007(1971 ) . j. l. ballot and m. fabre de la ripelle , ann .
( n. y. ) * 127 * , 62(1980 ) .
j. m. richard , phys . rep . * 212 * , 1(1992 ) .
h. leeb , h. fiedeldey , e. g. o. gavin , s. a. sofianos and r. lipperheide , few body systems * 12 * , 55(1992 ) .
n. barnea and a. novoselsky , ann .
( n. y. ) * 256 * , 192(1997 ) .
s. watanabe , y. hosoda and d. kato , j. phys .
* b26 * , l495(1993 ) . f. cooper , a. khare and u. sukhatme , phys . rep .
* 251 * , 267(1995 ) .
t. k. das , h. t. coelho and m. fabre de la ripelle , phys . rev . * c26 * , 2281(1982 ) .
r. h. dalitz and l. liu , phys .
rev * 116 * , 1312(1959 ) .
m. m. block and r. h. dalitz , phys .
. lett . * 11 * , 96(1963 ) .
j. b. adams , phys . rev . * 156 * , 1611(1967 ) . c. y. chung , d. p. heddle and l. s. kisslinger , phys
. rev . * c27 * , 335(1983 ) .
b. h. mc kellar and b. f. gibson proc .
int . conf . on hypernuclear and kaon physics , 1982 ,
heidelberg , ed .
b. povh , p-156 ; phys .
* c30 * , 322(1984 ) . m. a. khan and t. k. das , pramana- j. phys , * 56 * no.1 , 57(2001 ) .
1 . * fig . 1 * choice of jacobi coordinates for the partition 1 . 2 .
2 * plot of : ( i ) @xmath2 potential ; ( iia ) effective @xmath3 potential for @xmath5he ; ( iib ) effective @xmath3 potential for @xmath39zr ; ( iiia ) core-@xmath0 folded potential for @xmath5he ; ( iiib ) core-@xmath0 folded potential for @xmath39zr ; ( v is in mev and r is the relative separation ) 3 . *
3 * plot of lowest eigen potential @xmath220 against hyperradial distance @xmath58 for double-@xmath0 hypernuclei ( a ) @xmath5he , ( b ) @xmath38ca and ( c ) @xmath39zr 4 . * fig .
4 * plot of core-@xmath0 potential depth @xmath221(mev ) against mass of core @xmath222 5 . * fig . 5 * plot of ground and excited state wavefunctions for @xmath5he 6 . *
6 * plot of ground and excited state wavefunctions for @xmath38ca 7 .
* fig . 7 * plot of ground and excited state wavefunctions for @xmath39zr 8 . *
8 * plot of ground state- ( i ) single-@xmath0 separation energy @xmath166(mev ) and ( ii ) two-@xmath0 separation energy @xmath27 ( mev ) , against mass of the core @xmath172 9 . * fig . 9 * energy level diagram of double-@xmath0 hypernuclei [ @xmath27 is in mev ] 10 . * fig .
10 * variation of ground state rms matter radius @xmath189 ( fm ) with mass a ( in nucleon mass unit ) of the doule-@xmath0 hypernuclei
* table i. parameters of @xmath223 interaction . * + | bound state properties of few single and double-@xmath0 hypernuclei is critically examined in the framework of core-@xmath0 and core+@xmath1 few - body model applying hyperspherical harmonics expansion method ( hhem ) .
the @xmath2 potential is chosen phenomenologically while the core-@xmath0 potential is obtained by folding a phenomenological @xmath3 interaction into the density distribution of the core .
the depth of the effective @xmath3 potential is adjusted to reproduce the experimental data for the core-@xmath0 subsystem .
the three - body schrdinger equation is solved by hyperspherical adiabatic approximation ( haa ) to get the ground state energy and wave function .
the ground state wavefunction is used to construct the supersymmetric partner potential following prescription of supersymmetric quantum mechanics ( ssqm ) algebra .
the newly constructed supersymmetric partner potential is used to solve the three - body schrdinger equation to get the energy and wavefunction for the first excited state of the original potential .
the method is repeated to predict energy and wavefunction of the next higher excited states .
the possible number of bound states is found to increase with the increase in mass of the core of the hypernuclei .
the root mean squared ( rms ) matter radius and some other relevant geometrical observables are also predicted .
+ _ md . abdul khan + department of physics , aliah university , + aa - ii , plot no .
iia/27 , newtown , kolkata-700156 , west bengal , india . +
email : [email protected] ; [email protected] + _ pacs 21.80.+a ; 21.60.jz ; 21.30.fe + * key words : * + hypernuclei , few body system ( fbs ) , hyperspherical harmonics ( hh ) , hyperspherical harmonics expansion ( hhe ) , super symmetric quantum mechanics ( ssqm ) . |
stars are detectable as x - ray sources at several important stages of their lives .
pre - main sequence stars are x - ray sources because of their enhanced magnetic activity @xcite .
massive ob and wolf - rayet stars produce x - rays through shocks in their stellar winds @xcite , and possibly from magnetically - confined plasma close to their stellar surfaces @xcite .
neutron stars are bright x - ray sources if they are young and still have latent heat from what was once the stellar core @xcite , if they accelerate particles in rotating , moderate - strength ( @xmath7@xmath8@xmath9 g ) fields @xcite , or if they have extremely strong fields ( @xmath7@xmath8@xmath10 g ) that decay and accelerate particles @xcite .
white dwarfs , neutron stars , and black holes are bright x - ray sources if they are accreting matter from a binary companion @xcite , or in principle from the interstellar medium ( see , e.g. , * ? ? ?
therefore , x - ray surveys can be used to study the life cycles of stars , particularly their start and end points . here
we present a catalog of x - ray sources detected in _
chandra _
observations toward the inner 2@xmath1 by 0.8@xmath1 of the galaxy .
the region encompasses about 1% of the galactic stellar mass @xcite , and possibly up to 10% of the galactic population of young , massive stars @xcite .
therefore , these data provide a statistically meaningful sample of the galactic stellar population .
previous catalogs based on _ chandra _ data on the galactic center have been published by @xcite using 630 ks of data taken through 2002 june on the central 17@xmath017 around , and by @xcite using observations taken through 2005 june on the inner 2@[email protected]@xmath1 of the galaxy . however , since the publication of these catalogs , a large amount of new data have been obtained .
these data increase the number of point sources identified by a factor of 2.5 .
they also provide much better astrometry for individual x - ray sources .
the improvement in astrometry enables the identification of rare objects such as wolf - rayet stars , x - ray binaries , and rotation - powered pulsars , through comparisons of our x - ray catalog with radio and infrared data sets ( e.g. , mauerhan , j. , in prep ) .
therefore , we provide here an updated catalog of point sources , which incorporates and supercedes the previous catalogs .
we also describe the spatial and luminosity distributions of the x - ray sources . throughout this paper ,
we adopt a distance to the galactic center of @xmath11=8 kpc @xcite , and an average absorption column of @xmath12=@xmath13 @xmath14 @xcite .
as of 2007 august , the central 2@[email protected]@xmath1 of the milky way has been observed with the imaging array of the _ chandra_advanced ccd imaging spectrometer ( acis - i ; * ? ? ?
* ) on numerous occasions .
the majority of the new sources in this catalog come from 600 ks of exposure that we obtained in fifteen 40 ks pointings covering @xmath61@xmath1 of the galactic center .
the new data since 2005 also includes 370 ks on the central 20 pc around , 100 ks on the arches cluster @xcite , and 100 ks on sgr c. we also include sources previously identified in 630 ks of data on the inner 20 pc around @xcite ; thirty 12 ks exposures of the 2@[email protected]@xmath1 survey obtained by ( * ? ? ?
* see also muno 2006a ) ; and deep pointings toward the radio arches ( 50 ks ; * ? ? ?
* ) and sgr b2 ( 100 ks ; * ? ? ?
the dates , observation identifiers ( obsids ) , durations , locations , roll angles , and some values relevant for the astrometry ( 2.1 ) for each exposure are listed in table [ tab : obs ] .
the observations in the table are sorted by right ascension and declination , so that observations near the same point are grouped .
the acis - i is a set of four , 1024-by-1024 pixel ccds , covering a field of view of 17 by 17 .
when placed on - axis at the focal plane of the grazing - incidence x - ray mirrors , the imaging resolution is determined primarily by the pixel size of the ccds , 0492 .
the ccd frames are read out every 3.2 s , which provides the nominal time resolution of the data .
the ccds also measure the energies of incident photons within a calibrated energy band of 0.58 kev , with a resolution of 50300 ev ( depending on photon energy and distance from the read - out node ) .
however in some of the earlier , shallow exposures ( obsids 2267 through 2296 ) , an event filter was employed on the satellite that removed x - rays with energies below 1 kev before the data were sent to the ground .
the lack of 0.51.0 kev photons had a minor impact on our results , because there were only 76 sources that were detected below 2 kev for which the photometry was derived entirely from the obsids 2267 through 2296.25% .
the soft color is also systematically high .
for instance , sources with @xmath15 @xmath14 will have @xmath16@xmath6@xmath17 using a 0.52.0 kev soft band , and @xmath16@xmath6@xmath18 using a 1.02.0 kev soft band .
] we omitted obsid 242 from our analysis , because it was taken with the detector at a cooler temperature ( 110 k , versus 120 k ) . a flux image and composite exposure map
is displayed in figure [ fig : obs ] , and an adaptively - smoothed three - color image is displayed in figure [ fig : rgb ] .
the data were processed using the _ chandra _ interactive analysis of observations ( ciao ) package .
the data were processed as they arrived , so we used ciao versions 3.3 and 3.4 .
information on the detectors was taken from the calibration database ( caldb ) versions 3.2.1 and 3.3.0 .
the differences between the two versions of the software were too minor to justify re - processing the older portions of the dataset .
we only used data from acis - i ; data from the s array was omitted because was offset far from the aim point , and the large point spread function on the detector resulted in bad stellar confusion .
we started with the level 1 event files provided by the _ chandra_x - ray center ( cxc ) , and reprocessed the data using the tool acis_process_events in order to remove pixel randomization and apply more recent energy calibration .
we then removed events associated with bad pixels , and applied the standard grade filters to the events and good - time filters supplied by the cxc .
we applied acis_run_hotpix to flag events associated with cosmic - rays , and removed them from the event list ( we did not run acis_detect_afterglow , because it sometimes removes genuine x - ray events .
we did , however , later remove sources that were cosmic ray afterglows ; see 2.2 ) .
we then searched each observation for time intervals when particles encountering the detector caused the background event rate to flare to @xmath193@xmath4 above the mean level , and removed them .
these background flares were found in 12 observations , and lasted @xmath35% of the duration of each observation .
next , we applied the sub - pixel event repositioning algorithm of @xcite .
finally , if an astrometric correction was available from the cxc for any observation , we applied it at this point , by modifying the header keywords for the event file , and by correcting the columns for the right ascension and declination in the aspect solution provided with the observation . before proceeding to explain our algorithms for source detection
, we would like to explain some minor weaknesses of our approach .
unfortuantely , because the data were searched for sources as they arrived , and because the exposures were highly non - uniform across the field , some parameters of our detection algorithm , particularly the detection threshholds , were not kept consistent . to compensate for this , we did two things .
first , we verified the reality of each source as part of our photometric algorithm ( 2.2 ) .
this should eliminate most spurious sources on the faint end , in a uniform manner .
second , we determined the completeness limits of our survey using monte carlo simulations that mimicked our source detection algorithms ( 2.3 ) .
this is the best way to establish what portion of our sample is complete . with the experience we have gained , in principle we could develop a more streamlined and straightforward approach to building the inital catalog .
however , it would take several months of computer time to reprocess the data , or a similar amount of time re - writing our software to be more efficient .
the improvements in the final catalog would be slight , so we decided not to delay releasing our catalog any further .
we are making the data products from the following sections available in fits format from a web site .
the catalog itself will also be available with the electronic version of the paper .
source detection and localization was approached iteratively .
we first searched for point sources in each observation separately .
the locations of the point sources found in the first stage were used to refine the astrometry .
second , the astrometrically - corrected images were combined to search for fainter point sources .
finally , the source lists from the individual observations were merged with those from the combined images .
we searched each observation individually for point sources using the wavelet decomposition algorithm wavdetect @xcite .
we employed the default `` mexican hat '' wavelet , and used a sensitivity threshold of @xmath20 .
this threshold roughly corresponds to the chance of detecting a spurious source in an area corresponding to the point spread function ( psf ) , if the local background is spatially uniform . for the earlier data , taken before 2006 , we used images at three different resolutions : one at 05 resolution covering the inner 1024@xmath01024 pixels , one at 1 resolution covering the inner 2048@xmath02048 pixels , and one at 2 resolution covering the entire field . for later observations
, we simplified the process and used only two resolutions : 05 covering the inner 2048@xmath02048 pixels , and 2 covering the entire field . using a test field
, we confirmed that there was no difference in the number of sources detected using the two techniques ; the only difference is that the technique that used three , smaller images was computationally faster .
we used wavelet scales that increased by a factor of @xmath21 , over the range of 14 pixels for the 05 image , 18 pixels for the 1 image , and 116 pixels for the 2 image . for each resolution
, we made images in three energy bands : 0.58.0 kev to cover the full bandpass , 0.52.0 kev to provide sensitivity to foreground sources , and 48 kev to provide sensitivity to highly - absorbed sources . for each image ,
a matching exposure map was generated for photons with an energy of 4 kev , so that the wavelet algorithm could keep track of regions with rapidly - varying exposure , such as bad columns and the edges of the ccds .
the lists derived from each image resolution were combined to form master source lists for each energy band .
we found that the positions would be most accurate from the sources identified in the image with the finest resolution .
therefore , we discarded sources from the lower - resolution images if their separations from sources identified at high resolution were smaller than the radii of the 90% contour of the psf .
in this way , we produced three lists for each observation , one for each energy band .
next , we used the point sources detected so far to register the absolute astrometry to the two micron all - sky survey ( 2mass ; * ? ? ?
the 2mass frame is consistent with the international celestrial reference system to within 15 mas .
we compared the positions of 2mass sources to those of x - ray sources detected in the 0.52.0 kev band that were identified by wavdetect as having @xmath223@xmath4 significance , and identified matches as those with offsets @xmath31 .
the offsets between the 2mass and _ chandra _ frames were computed using a least - chi - squared algorithm .
the x - ray sources that we used for astrometry are flagged in table [ tab : positions ] ( see [ sec : tdesc ] ) . for observations longer than 20 ks
, we found between 3 and 36 x - ray sources in the soft band that could be associated unambiguously with stars in the 2mass catalog , and so we used the average offsets between the x - ray and infrared sources to correct the astrometry of the x - ray observations .
we evaluated the accuracy of the registration based on the standard deviation in the mean of the offsets of the individual stars .
the registraton was accurate to 006 for the deepest exposures , and to 02 for the shallower ones ( 1@xmath4 ) .
unfortunately , for exposures shorter than 20 ks , too few x - ray sources were found with 2mass counterparts to correct the astrometry to better than the default value , 05 ( 1@xmath4 ) .
once the deepest observation at each point was registered to the 2mass frame , the shallower observations were registered using the offsets of x - ray sources detected in the 0.58.0 kev band in pairs of observations . between 2 and 759 x - ray sources matched between the deepest and shallower observations , depending upon the exposure time of the shallower observation .
the uncertainty in the astrometry of each observation is listed in the last column of table [ tab : obs ] .
the composite image and exposure map for our survey are displayed in figure [ fig : obs ] .
having corrected the astrometry for fields that included deep observations , we then combined subsets of the images in order to perform a deeper search for point sources .
two wavelet algorithms were used , on the series of images listed below .
first , the tool wavdetect @xcite was used to identify point sources in : * composite images made from all of the observations of .
twelve , 1024@xmath01024 images were made , in four resolutions ( 025 , 05 , 1 , and 2 ) and using three energy bands for each resolution ( 0.58.0 kev , 0.52.0 kev , and 48 kev ) . *
three sets of composite images made from observations of in 2002 , 2004 , and 2005 .
these images were designed to be sensitive to faint , variable sources .
the same image resolutions and energy bands were used as for the composite image of all of the data .
* composite images made from three pointings that were taken with same same roll angle , because the original 40 ks exposure had to be split up to accomodate scheduling constraints ( obsids 7038 , 7041 , and 7042 ) .
three images were made for each aimpoint , one for each of the 0.58.0 kev , 0.52.0 kev , and 48 kev energy bands .
each image was made at 05 resolution , and had 2048@xmath02048 pixels .
the parameters used with wavdetect were the same as for the individual observations .
second , we used the tools wvdecomp and findpeak in the _ zhtools _ package written by a. vikhlinin to search for faint sources that fell below the wavdetect threshold .
we searched on wavelet scales of 13 pixels , and required that a candidate source be identified with a minimum signal - to - noise of 4.5 , corresponding to 16 spurious sources per 2048 by 2048 pixel image .
five iterations of the search procedure were performed .
the tool wvdecomp iteratively cleans the image of point sources identified in previous passes through the data , so it is more efficient at separating close pairs of sources .
moreover , unlike wavdetect , wvdecomp does not use any information about the shape of the point spread function in searching for sources , so it is better at identifying point sources when observations with very different aimpoints have been combined .
therefore , we used wvdecomp on composite images generated from all data covering the positions at which deep observations were obtained ( i.e. , obsids 3392 , 4500 , 5892 , 70347048 , and 944 ) . each image was produced with 2048@xmath02048 pixels at 05 resolution for the 0.58.0 kev band .
we then generated a list containing the unique sources , by merging the lists generated by wavdetect from individual observations and from combined images , and from the lists generated by wvdecomp .
we found that almost all the duplicates could be removed by identifying sources with separations smaller than the prescription for positional uncertainties in @xcite : for sources offset from the center of each image by @xmath23@xmath35 , the separation was cut at 06 , whereas for larger offsets it was cut at @xmath24 ( @xmath23 is in arcmin ) .
for each image , we gave preference to positions from the full - band sources , then from the soft sources , and finally from the hard sources . across observations , the priority was given to the deeper observations , and to sources detected with wavdetect over those detected with wvdecomp .
we examined the final list visually by comparing it to images of the survey fields , and we removed several hundred sources that were portions of extended , diffuse features ( from * ? ? ?
* ) , and a couple dozen duplicates that were not identified automatically .
finally , two sources were not picked up by the detection algorithms because they were blended with nearby , brighter sources .
we added these to our catalog by hand ( cxougc j174502.8282505 and j174617.4281246 ) . at this stage , we considered our source lists to be provisional , both because the search algorithms used non - uniform parameters , and because the large , spatially - variable background was likely to cause our wavelet algorithms to generate a significant number of spurious sources . in order to confirm their validity , we next computed photometry for each provisional source .
we computed aperture photometry for each source using the package , versions 3.96 , 3.101 , and 3.128 @xcite , along with some custom code .
the algorithm proceeded in several steps .
first , for each source and each observation , we obtained a model psf with the ciao tool mkpsf . for most sources
, we used a psf for a fiducial energy of 4.5 kev .
however , if a source was only detected with wavdetect in the soft band , we used a psf for an energy of 1.5 kev . to determine a region from which we would extract source counts , we then constructed a polygon enclosing 90% of the psf .
if the polygons for two sources overlapped in the observations in which the sources were closest to the aim - point , we generated a smaller polygon .
the final extraction regions enclosed between 70% and 90% of the psf . sources for which the psf fraction was @xmath390% were considered to be confused . moreover , because the psf grows rapidly beyond 7 from the aim - point , we also considered sources to be confused if they were located beyond 7 from the aim - point and their psfs overlapped .
photometry was not computed for observations in which confused sources fell @xmath227off - axis .
fortunately , these second type of confused sources were always located on - axis in another observation , or else they would not have been identified .
finally , for similar reasons , we only computed photometry for sources that lay within 7 of if the relevant observations had as the aim point .
second , we extracted source event lists , source spectra , effective area functions , and response matrices for each source in each observation .
the detector responses and effective areas were obtained using the ciao tools mkacisrmf and mkarf , respectively . for each source , the spectra from all of the relevant observations were summed .
the responses and effective areas were averaged , weighted by the exposures in each observation .
third , we extracted background events from circular regions surrounding each point source in each observation , omitting events that fell within circles that circumscribed @xmath690% of the psfs around any point sources .
the background regions were chosen to contain @xmath6100 total counts for the wide survey , and @xmath61000 total counts for the deeper field .
fewer than 1% of the counts in the background region originate from known point sources . for each source ,
the background spectra from all of the relevant observations were scaled by the integrals of the exposure maps ( in units of @xmath25 s ) over the source and background regions , and then summed to create composite background spectra .
fourth , we eliminated spurious sources .
we compared the number of source and background counts to estimate the probability that no source was present , based on poisson statistics @xcite . if a source had a @xmath2210% chance of being spurious , we eliminated it from our catalog .
we eliminated 1962 sources in this way .
we also eliminated sources in which the majority of events were cosmic ray afterglows . specifically , we removed 46 sources because the events associated with the candidate source fell in a single pixel during 510 consecutive frames .
our final catalog contains 9017 x - ray sources , and is listed in table [ tab : positions ] .
the majority of sources , 7152 , were found with wavdetect . of the sources detected with wavdetect ,
4823 were detected in the full band , 948 in the soft band , and 1381 in the hard band .
another 1865 sources were only detected with wvdecomp . in the field alone , we found 3441 sources with wavdetect , of which 2715 were detected in the full composite image , 275 in 2002 , 48 in 2004 , 90 in 2005 , and 313 in individual observations .
an additional 364 were found in the field with wvdecomp .
fifth , we compared the source and background spectra using the kolmogorov - smirnov ( ks ) statistic , in order to flag potentially - spurious objects that could be variations in the background .
caution should be used when studying sources that resemble the background .
for instance , in the central parsec around , there is an over - abundance of faint ( @xmath26@xmath27 ) , soft point sources that have spectra consistent with that of the background warm plasma ( note that almost all of the excess bright sources that we discuss in 3.1 do have spectra that are distinct from the background ) .
therefore , we suspect that most of these are @xmath80.1 pc scale variations in the density of that plasma . unfortunately , we can not be certain .
indeed , the spectrum of the bright x - ray source associated with irs 13 ( cxougc j174539.7290029 ) resembles the background according to the ks test .
if the diffuse background is merely unresolved point sources @xcite , then most faint point sources should have spectra that resemble the background .
sixth , we computed the net counts in the 0.52.0 kev , 2.08.0 kev , 2.03.3 kev , 3.34.7 kev , and 4.78.0 kev bands .
we estimated the photon flux from each source , by dividing the net counts by the average of the effective area function in each band .
table [ tab : phot ] lists the 0.52.0 kev and the 28 kev flux , the latter of which is the sum of the fluxes in the three sub - bands .
figure [ fig : lumdist ] displays histograms of the net counts and fluxes in the 0.52.0 kev and 28 kev bands .
histograms of upper limits are also plotted , for sources that were detected in one band but not the other .
finally , using custom code that was not part of acis extract , we computed 90% uncertainties on the net counts in each band , through a bayesian analysis of the poisson statistics , with the simplifying assumption that the uncertainty on the background is negligible @xcite .
we used the net counts to compute the hardness ratios @xmath28 , where @xmath29 and @xmath30 are the numbers of counts in the higher and lower energy bands , respectively .
the resulting hardness ratios are bounded by @xmath31 and @xmath32 .
we defined a soft color using counts in the 2.03.3 kev and 0.52.0 kev bands ( @xmath16 ) , a medium color using counts in the 3.34.7 kev bands and 2.03.3 kev ( @xmath33 ) , and a hard color using counts in the 4.78.0 kev and 3.34.7 kev bands ( @xmath34 ) .
we calculated uncertainties on the ratios using the 90% uncertainties on the net counts and equation 1.31 in lyons ( 1991 ; page 26 ) .
the hardness ratios are listed in table [ tab : phot ] , and histograms showing their distributions are displayed in figure [ fig : hrdist ] . the soft color , @xmath16 , was used to distinguish foreground sources from objects that were likely to lie near or beyond the galactic center .
we select foreground x - ray sources as those with soft colors in the range @xmath35@xmath36@xmath16@xmath3@xmath37 , which corresponds to absorption columns equivalent to @xmath12@xmath26@xmath38 @xmath14 .
most of these should lie within 4 kpc of earth ( e.g. , * ? ? ?
we selected x - ray sources that were located near or beyond the galactic center as those either that had soft colors @xmath16@xmath19@xmath37 , or that were not detected in either of the 0.52.0 and 2.03.3 kev bands .
this x - ray selection corresponds to absorption columns equivalent to @xmath12@xmath39@xmath38 @xmath14 .
we find 2257 foreground x - ray sources , and 6760 sources near or beyond the galactic center .
foreground and absorbed sources are plotted separately in figure [ fig : hrdist ] . most absorbed sources do not have measured soft colors .
we searched for variability using the arrival times of the events .
we searched for three kinds of variations : long - term variations that occurred between observations , short - term variability within individual observations , and periodic variability within individual observations .
we searched for variations that occurred between observations by comparing the event arrival times from all of the observations to a constant flux model using the ks statistic . any source with a @xmath30.1% chance of being described by a constant flux model
was considered to vary on long time scales .
there were 856 sources that exhibited long - term variability , 137 of which also exhibited short - term variability .
therefore , about 10% of sources vary on the day to month time scales between observations .
we characterized these long - term variations by computing the mean photon flux during each observation of a variable source .
table [ tab : longterm ] lists the source name , observations in which the largest and smallest fluxes were observed , the values of the largest and smallest fluxes , and the ratios of those values .
figure [ fig : longterm ] compares the amplitude of the variations to the maximum flux . in order to exclude measurements with poor signal - to - noise
, the largest flux was defined as the measurement with the largest lower limit , and the smallest flux was defined as the measurement with the smallest upper limit . in most ( 740 ) cases ,
the smallest flux was consistent with zero , and the lower limit to the flux ratios was provided . in 224 cases , the uncertainties in the largest and
smallest fluxes overlapped , and the formal lower limit to the ratio was less than 1 .
the statistics on faint sources with low - amplitude variability tended to be poor , so table [ tab : longterm ] would be best used to identify highly - variable sources for further study .
we searched for variability within each observation by comparing the light curves to constant count rate models using the ks statistic .
if the arrival times of events had a @xmath30.1% chance of being described by a uniform distribution , we considered a source to have short - term variability .
we identified 294 sources , or 3% of our sample , as having clear short - term variations .
we roughly characterized the nature of the variability by dividing each time series into intervals that were consistent with having constant count rates , using the bayesian blocks algorithm of @xcite . in brief , the algorithm compared the probability that an interval could be described by two different count rates to the null hypothesis that the photons arrived with a single rate .
if the ratio of the two probabilities exceeded a user - specified prior odds ratio , then the interval was divided at the time that produced two intervals with the largest calculated likelihood .
this process was iterated until no sub - intervals were divided any further .
we chose to apply the algorithm in order to describe each variable light curve with the fewest intervals with distinct rates ( blocks ) .
we applied three progressively looser odds ratios , successively demanding that the probability for the two - rate model exceed the null hypothesis by factors of 1000 , 100 , and 10 if the larger odds ratio failed to identify a change point . in this way
, large flares were described with a few `` blocks '' using a large odds ratio , whereas small - amplitude variations were still characterized using a smaller odds ratio .
this approach was deemed necessary in part because the bayesian interpretation of the odds ratios does not have a good frequentist analog that could be compared to the probabilities returned by the ks test , and in part because the ks test and the bayesian blocks tests are most sensitive to slightly different forms of variability .
ultimately , only 60% of the variable sources identified with the ks test were characterized with more than one block in the bayesian block algorithm . despite the mismatch between the two tests , the characteristics of the variable sources identified by both the ks and bayesian blocks tests
are illustrative . in table
[ tab : shortterm ] , we list some properties of the variable sources : their names , the obsids in which variability occurred , the odds ratio at which the bayesian blocks algorithm identified a source as variable , the number of blocks used to describe the events , the durations of the brightest portions of the light curves , the minimum and maximum fluxes , and the ratios of the maximum to minimum fluxes . figure [ fig : shortterm ] compares the duration and amplitude of the variability . for 40% of the variable sources ,
the minimum flux was consistent with zero , so the ratio represents a lower limit to the variability amplitude .
we find that all variations had time scales of @xmath2210 minutes .
the amplitudes ranged from barely - detectable 30% variations in the flux ( cxougc j174534.8290851 ) , to one flare in which the flux increased by a factor of 250 ( cxougc j174700.7283205 ) .
foreground sources are over - represented among variable sources they compose only 25% of our entire catalog , but 50% of the short - term variables which is consistent with the expectation that they are nearby k and m dwarf flare stars ( e.g. , * ? ? ?
we searched for periodic variability in the brightest sources by adjusting the arrival times of their photons to the solar system barycenter and computing fourier periodograms using the rayleigh statistic @xcite .
the individual x - ray events were recorded with a time resolution of 3.2 s , so the nyquist frequency was @xmath60.15 hz , which represents the limit above which our sensitivity could not be well - characterized .
however , we computed the periodogram using a maximum frequency of @xmath60.2 hz , to take advantage of the limited sensitivity to higher frequency signals , and to ensure that any observed signal was not an alias .
we considered sources that , in individual observations , produced a large enough number of counts ( @xmath40 ) that a fully - modulated signal could be detected with 99% confidence .
the power @xmath41 required to ensure that a source had a chance @xmath3@xmath42 of being produced by white noise can be computed if one knows the number of trials in a search , ( @xmath43 ) , and is given by inverting @xmath44 here , @xmath41 is normalized to have a mean value of 1 , and the approximation is valid for @xmath41@xmath451 @xcite . a count threshold can be determined by noting that , if background photons are negligible , the fractional root - mean - squared amplitude of a sinusoidal signal ( @xmath46 ) is given by @xmath47 .
a fully - modulated signal has @xmath46=0.71 . after iterating to determine the number of trials corresponding to each count limit
, we found that a source with @xmath48=86 could be identified with @xmath49=0.99 if it produced a fully - modulated signal . in total , we searched for pulsations in 717 event lists from 256 different sources , which required @xmath50 trials .
a single signal that had @[email protected] given this number of trials must have had @[email protected] . however , multiple observations were searched for many sources , so we recorded signals with lower powers and checked whether they also appeared at the same frequency in other observations .
we identified two sources with periodic variability at @xmath2299% confidence , cxougc j174532.7290550 and cxougc j174543.4285841 .
we had previously identified both of these by combining 500 ks of exposure over the course of two weeks @xcite .
the other sources in @xcite were too faint for their periodic variability to be identified in individual observations .
we also identified a third source as a good candidate for having periodic variability , cxougc j174622.7285218 .
signals from this source were identified with periods of @xmath61745 s in observation 4500 with @xmath41=10.9 from 763 photons , and in observation 7048 with @xmath41=13.4 from 310 photons ( figure [ fig : fft ] ) .
the joint probability that these signals were produced at the same frequency by noise @xcite , given @xmath43=@xmath50 , was only 1.4% .
periodic signals were not detected from this source in observation 945 because the source fell on a chip edge , nor in observations 2273 and 2276 because their exposures were too short .
we refined our initial estimates of the period for cxougc j174622.7285218 for each observation by computing pulse profiles from non - overlapping @xmath51 s intervals , and modeling the differences between the assumed and measured phases using a first - order polynomial .
the reference epochs of the pulse maxima for the two observations were 53165.3781(6 ) and 54145.1644(7 ) ( mjd , barycentric dynamical time ) .
the best - fit periods were 1745@xmath523 s and 1734@xmath5216 s for observations 4500 and 7048 , respectively .
the pulse profiles for each observation are displayed in figure [ fig : prof ] .
the fractional root - mean - squared amplitudes of the pulsations were 21% and 32% , respectively . given the long period for this source , it is most likely a magnetically - accreting white dwarf @xcite .
we calculated the sensitivity of our observations using synthetic - star tests , following the basic methods described in @xcite and muno ( 2006a ; see wang 2004 for another approach ) .
we generated maps of our sensitivity both for each of the stacked observations ( i.e. , centered on obsids 3392 , 4500 , 5892 , 70347048 , and 944 ) , and for a fiducial field with an exposure time of 12 ks for those regions only covered by the shallow exposures of @xcite . in brief , for each pointing , we generated a background map by ( 1 ) removing events from within a circle circumscribing @xmath690% of the energy of the psf around each detected source , and then ( 2 ) filling the `` holes '' in the image with numbers of counts drawn from poisson distributions with means equal to those of surrounding annuli .
we then simulated 100 star fields per pointing .
we placed @xmath65000 point sources at random positions in each background image , with fluxes distributed as @xmath53 with a slope @xmath54 , and minimum fluxes that would produce 3 counts in a 100 ks exposure .
we converted these fluxes to expected values for the numbers of counts using an exposure map .
the exposure map was normalized to produce the mean flux - to - counts conversion for x - ray sources located at or beyond the galactic center ( @xmath55 ; * ? ? ? * ) .
then , to account for the eddington bias , we drew observed numbers of counts from poisson distributions with mean values equal to the expected counts .
next , we obtained model images of the psf from the routine mkpsf , averaged them when appropriate , and used the composite psf as the probability distribution to simulate the 2-dimensional image of the counts .
these were added to the synthetic exposure . finally , we searched the synthetic image for point sources using for the 12 ks exposures , and wvdecomp for the stacked observations . by comparing the input and output lists , we estimated the minimum flux at which a source would be detected in 50% and 90% of trials over a grid of points covering our survey .
we interpolated between these points to make a map of the sensitivity for each image .
none of our observations are formally confusion - limited at our completeness limits ( hogg 2001 ; see also muno 2003 ) .
if the background diffuse x - ray emission is unresolved stellar sources , then confusion caused by undetected sources is accounted for naturally by our background maps . in order to produce a global sensitivity map , we combined the sensitivity maps from the above simulations by recording the best sensitivity at each point in the image .
the map of 90% confidence limits is displayed in figure [ fig : sens ] .
the effective area of the survey as a function of limiting photon flux and luminosity is displayed in figure [ fig : area ] .
we are sensitive to @xmath6@xmath2 ( 0.5 - 8.0 kev , assuming @xmath11=8 kpc ) at 90% confidence over one square degree , and to @xmath6@xmath56 over 0.1 square degrees .
this is a factor of @xmath62 improvement over @xcite .
however , we still find that the majority of x - ray sources are detected at fluxes below our completeness limits . of 6760 sources that are likely to lie at or beyond the galactic center ( @xmath55 ) ,
only 15% are brighter than the 90% completeness limit at the point at which they were detected , and only 40% are brighter than the 50% completeness limit .
this is caused by two effects .
first , 20% of the sources are detected only in the hard band , whereas our completeness limits are for the full band .
second , the number - flux distribution is steep ( 3.3 ) , such that many of the faint sources are only detected because of positive poisson fluctuations in their count rates .
these maps are used in selecting complete samples of sources for measuring the spatial ( 3.2 ) and flux ( 3.3 ) distributions .
sources below our completeness limits are still securely detected , although other sources with similar intrinsic fluxes have been missed .
experience with matching x - ray and optical sources as part of the _ chandra _ deep fields and orion ultradeep projects suggests that the positions of the x - ray sources can be refined with respect to those provided by the wavelet algorithms @xcite .
therefore , we used the implementations of their techniques in acis extract to refine the positions of our sources . for each source
, we made a composite image by combining the event lists from each relevant observation , and then made a matching composite psf image that we weighted by the values of the exposure maps at the source positions . from this image , we computed two additional estimates for the source position : the mean positions of the events within each source region , and a centroid determined by cross - correlating the psf and source images .
following @xcite , if a source lay within 5 of the aim point , we used the mean position of the events within the source extraction region .
if the source lay beyond 5 , we used the position determined by cross - correlating the source image and psf image .
however , if the offset of the refined position from the wavelet position was larger than the smallest source extraction radius that we used , we assumed that a nearby source had caused confusion , and retained the wavelet position .
unfortunately , we could not empirically calibrate the uncertainties on our source positions , because even the foreground infrared sources had such a high density that @xmath650% of those that fell within 3 of an x - ray source were chance alignments .
therefore , we computed 95% positional uncertainties using equation 5 in @xcite , which is based on the positions of sources reported by wavdetect in simulated observations : @xmath57 \nonumber \\ & + & 0\farcs03 \left [ \frac{\theta}{\log_{10}(c + 2)}\right]^2 + 0\farcs0006 \left[\frac{\theta}{\log_{10}(c + 3)}\right]^4 \label{eq : posunc}. \end{aligned}\ ] ] here , @xmath23 is the offset in arcminutes of the source from the nominal aim point , and @xmath58 is the net number of counts . for sources detected in composite images , we defined @xmath58 to be the net counts summed over all observations , and @xmath23 to be the exposure - weighted averages of the sources offsets from the aim points of their respective observations . for sources detected in individual observations
, we defined @xmath58 and @xmath23 to be the values for the observations in which the sources were identified .
if @xmath59 is larger than the smallest radius of the region used to extract photometry for the source ( the `` source radius '' ) , the uncertainty was set equal to the radius of the extraction region .
these sources are marginally detected , and the high background in the galactic center produces a large tail in the distribution of possible positions .
we retained them because they passed all of our other selection criteria .
@xcite established equation [ eq : posunc ] by running wavdetect on simulated , single observations that were generated using a ray - tracing code .
unfortunately , our observations are more complicated . on the one hand ,
most of the positions are determined from composite images generated from observations with very different aim points .
the inclusion of data with large @xmath23 could add uncertainty to our measurements .
on the other hand , our positions have been refined compared to the wavdetect values , so the uncertainty on some sources could be smaller .
therefore , we view equation [ eq : posunc ] as a compromise .
nonetheless , a comparison of the offsets between 500 foreground x - ray sources and the blue 2mass sources that are their counterparts ( as described in detail in j. mauerhan , in prep ) reveals that the positions in the new catalog are @xmath660% better than in @xcite and @xcite .
we also note that because of the way we averaged the psf , the positions and uncertainties for the of sources that vary in flux between observations ( 10% of our sample ) could be mis - estimated .
for example , a variable source that was only bright in an off - axis observation would have a larger uncertainty than might be expected if it were also bright during an on - axis observation .
we have not evaluated whether systematic offsets in the positions are expected .
table [ tab : positions ] contains the locations of the point sources , parameters related to the observations of each source , and information on the data quality .
its columns are as follows : \(1 ) record locators that can be used to cross - correlate with other tables .
\(2 ) the source names , which are derived from the coordinates of the source based on the iau format , in which least - significant figures are _ truncated _ ( as opposed to rounded ) .
the names should not be used as the locations of the sources .
( 34 ) the right ascensions and declinations of the sources , in degrees ( j2000 ) .
\(5 ) the 95% uncertainties in the positions ( the error circles ) .
there are 5810 sources with uncertainties @xmath361 ( half of which are within 7 of ) , and 1950 with uncertainties @xmath3605 ( 85% of which are within 7 of ) .
\(6 ) flag indicating how the positions were derived .
a `` d '' indicates the position is from the mean position of events , a `` c '' indicates it was derived by cross - correlating the image and the psf , and a `` w '' indicates it was derived from a wavelet algorithm .
sources marked with a `` w '' are likely to be confused with a nearby source , or in a region of high background .
\(7 ) the images in which the sources were identified .
the tags `` full '' , `` 2002 '' , `` 2004 '' , and `` 2005 '' indicate a source was found in composite images of the field .
all other values are the observations in which a source was detected .
two sources added manually are tagged with `` hand '' .
\(8 ) additional information about how the sources were detected . the tag `` full '' refers to any source detected with wavdetect in the 0.58.0 kev band ; `` soft '' sources were detected in the 0.52.0 kev but not the full band ; `` hard '' sources were detected in the 48 kev band but neither of the other two bands .
the tag `` tile '' indicates that the source was detected in a composite 0.58.0 kev image with wvdecomp .
\(9 ) the offsets ( @xmath23 ) from the aim point , in arcmin . if a source position was estimated from a composite image , @xmath23 is the mean offset weighted by the exposure . if a position was taken from a single observation , @xmath23 is the offset for that observation .
\(10 ) the number of observations used to compute the photometry for each source .
\(11 ) the exposure times in seconds .
\(12 ) the fractions of the psf enclosed by the source extraction regions .
\(13 ) the fiducial energies of the psfs used to construct the source extraction regions .
\(14 ) the smallest radius for the extraction region that was used for a source , in arcseconds .
this is determined from the observation in which the source was closest to the aim point .
it is also an absolute upper bound to the positional uncertainty for a source .
\(15 ) the 50% completeness limit at the position of the source .
sources brighter than these completeness limits can be used to compute spatial and flux distributions , although the sensitivity map ( fig .
[ fig : sens ] ) is needed to compute the corresponding survey area .
\(16 ) flags denoting quality , and other information : `` a '' for sources used to register the astrometry of fields ; `` s '' for sources variable on short time scales , as indicated by probabilities of @xmath30.1% that the event arrival times for at least one observation were consistent with a uniform distribution according to the ks test ; `` l '' for sources that were variable on long time scales , as indicated by a probability of @xmath30.1% that the fluxes for all observations were consistent with a uniform distribution according to the ks test ; `` e '' for sources that may be part of an extended , diffuse feature @xcite ; `` c '' for sources confused with another nearby source ; `` g '' for sources that fell near the edge of a detector in one or more observations ; `` b '' for sources for which the source and background spectra have a @xmath2210% chance of being drawn from the same distribution according to a ks test ; `` x '' for sources for which the 0.52.0 kev band photometry is inaccurate because the satellite was programmed to omit photons below 1 kev from the telemetry ; and `` p '' for sources that suffered from photon pile - up .
table [ tab : phot ] contains the x - ray photometry for each source .
it contains the following columns \(1 ) the record locators .
\(2 ) the source names .
\(3 ) the log of the probabilities that the source and background spectra are derived from the same distribution , according to a ks test .
large negative values indicate that the source and background spectra are distinct , and therefore that the source is most likely real .
\(4 ) the total numbers of counts in the 0.52.0 kev band .
\(5 ) the estimated numbers of background counts in the 0.52.0 kev band .
\(6 ) the net numbers of counts in the 0.52.0 kev band , and the 90% lower and upper uncertainties . in the case of non - detections , an upper limit is provided .
\(7 ) the total numbers of counts in the 28 kev band .
\(8 ) the estimated numbers of background counts in the 28 kev band .
\(9 ) the net numbers of counts in the 28 kev band , and the 90% lower and upper uncertainties . in the case of non - detections , an upper limit is provided .
\(10 ) the fluxes in the 0.52.0 kev band , in units of .
\(11 ) the fluxes in the 28 kev band , in units of .
\(12 ) the mean energy of photons in the source region , statistically corrected for the background .
\(13 ) the soft colors and 90% upper and lower uncertainties .
\(14 ) the medium colors and 90% upper and lower uncertainties .
\(15 ) the hard colors and 90% upper and lower uncertainties .
these tables were designed to be inclusive , so sources of questionable quality are included .
for instance , 134 sources have net numbers of counts in the 0.58.0 kev band that are consistent with 0 at the 90% confidence level .
these sources are only detected in a single band and are presumably either very hard or very soft , detected in single observations because they were transients , or detected in stacked observations with wvdecomp at marginal significance .
we have chosen to include them because they passed the test based on poisson statistics from @xcite .
with a catalog of x - ray sources and associated maps of our sensitivity , it is straightforward to examine the flux and spatial distributions of our sources .
we have previously reported these quantities based on the catalogs produced for the central 20 pc around @xcite and on the wide , shallow survey data that was in the archive as of 2005 june @xcite . here
, we derive these quantities for the new catalog , and briefly compare the distributions to recent results from @xcite on the distribution of diffuse iron emission . in figure
[ fig : hit ] , we plot the hard color versus the flux from each source .
foreground sources are indicated with open red circles , and sources at or beyond the galactic center with filled blue circles .
there are 6381 galactic center sources and 1091 foreground sources with measured hard colors .
we have calculated the hardness ratios and photon fluxes that we would expect to get from these energy bands for a variety of spectra and 0.58.0 kev luminosities using and . in figure
[ fig : hit ] , we plot the colors and fluxes expected for power - law spectra with the dotted lines , and for a optically - thin thermal plasma with the solid lines . we have assumed a distance of 8 kpc and @xmath13 @xmath14 of absorption from interstellar gas and dust .
the median hard color for the galactic center sources is 0.17 . for interstellar absorption
, this corresponds to a @[email protected] power law . using a simulated spectrum ,
we have determined that the photon fluxes can be converted to energy fluxes according to 1 @xmath61 ( 0.58.0 kev ) .
the de - absorbed 0.58.0 kev flux is approximately 1.7 times larger , so that for a distance @xmath11=8 kpc , @xmath62 equals @xmath63 .
the large median value of the hard color is inconsistent with that expected from a thermal plasma ( of any temperature ) attenuated by interstellar gas and dust .
however , our earlier study of the spectra of brighter sources suggest that intrinsic absorption is present , and that the underlying spectrum is consistent with a @xmath64=79 kev thermal plasma @xcite . for sources that are intrinsically absorbed , the luminosities will be significantly higher than implied by figure [ fig : hit ] .
we present the spatial distribution of x - ray sources located near or beyond the galactic center ( @xmath16@xmath22@xmath37 ) in figure [ fig : londist ] .
we examined only sources brighter than @xmath65 , and only included a source if the 50%-confidence flux limit at its position was less than or equal to @xmath65 .
this flux limit was chosen as a compromise between the area over which the distribution is derived , which decreases for lower flux limits ( fig .
[ fig : sens ] ) , and the number of sources used in the distribution , which tends to increase for lower flux limits ( fig .
[ fig : lumdist ] ) . in the top panel of figure [ fig : londist ]
, we display the locations of each of the 479 sources that met the flux criteria .
the area over which the flux limit is @xmath3@xmath65 is displayed in white , and the greyed areas indicate regions of poorer sensitivity .
a concentration of x - ray sources is evident near the position of . in the bottom panels of figure [ fig : londist ] , we display histograms of the numbers of sources per unit area , as functions of galactic longitude and latitude . only regions of good sensitivity are used .
we then compared the spatial distributions to that of the stellar mass that has been inferred from infrared observations .
our mass model consists of the young nuclear bulge and cusp and the old galactic bulge from @xcite , and the model for the galactic disk from ( * ? ? ?
* see muno 2006a for further details ) . to make a direct comparison with our unevenly - sampled spatial distributions , we integrated the model for the stellar mass from 6 to 14 kpc along the line of sight at points on a 1 grid covering our survey region , and interpolated the resulting values onto the image .
we then summed the values of the integrated mass over areas of good sensitivity , to match the longitude and latitude bins of the observed histogram .
finally , we minimized chi - squared over one parameter to scale the binned mass model to the observed distributions of x - ray sources .
we find a best - fit scaling factor of @xmath66 x - ray sources per solar mass for sources brighter than @xmath65 for both the latitude and longitude distributions .
for the longitude distribution , @xmath67=22.4/19 , and for the latitude distribution @xmath67=20.6/17 .
the best - fit models are displayed with solid lines in the bottom panels of figure [ fig : londist ] .
the models are acceptable descriptions of the data . however , in the plot as a function of longitude , at the inner few arcminutes around , and just to the east toward the arches and quintuplet regions , there is an @xmath62.8@xmath4 excess in the number of observed x - ray sources .
we find that this excess is also present with similar significance if we chose tighter or looser flux limits between 1 and @xmath68 .
the model also predicts more sources than observed at @xmath69=0.50.6@xmath1 at the 1@xmath4 level , but this is probably because the sgr b molecular complex attenuates x - rays from sources behind it .
we computed the number - flux distribution based on the maximum - likelihood algorithm described in @xcite , which we modified to use poisson statistics in the manner described in appendix b of @xcite .
we examined three regions that had well - defined flux limits and effective exposure times : the inner 8 around , the 8 around the arches cluster ( excluding the overlap with the field ) , and the portions of the survey covered by the 40 ks pointings taken between 2006 and 2007 .
we assumed that the number - flux distribution was a single power law over the ranges of fluxes that we measured , @xmath70 .
we display the resulting cumulative number - flux distributions in figure [ fig : lognlogs ] , and list the best - fit parameters in table [ tab : lognlogs ] .
our distributions extend a factor of @xmath62 deeper than in @xcite .
the fit to the distribution from the region is formally poor , because the distribution steepens at low fluxes @xcite .
however , we do find that the arches region has a flatter flux distribution ( @[email protected] ) than either the inner 8 around sgr a ( ( @[email protected] ) or the wide survey field ( @[email protected] ) .
the difference is only significant at the 1.4@xmath4 level .
nonetheless , given that there are also excess sources coincident with the arches region in the spatial distribution , we suggest that there is a genuine over - abundance of bright x - ray sources in this region of recent star formation .
lcccccc[htp ] sgr a * & 0.5 & 323 & 44 & [email protected] & 0.41 & 0.00 + arches & 1 & 17 & 22 & [email protected] & 0.08 & 0.88 + field & 3 & 92 & 813 & [email protected] & 0.02 & 0.88 a similar asymmetry has been identified in the flux of diffuse emission from helium - like iron @xcite .
we suggest that both the excess point sources and the excess iron emission are related to the concentration of young stars in this region , the most dramatic manifestations of which are the arches and quintuplet clusters ( e.g. , * ? ? ?
the iron emission is probably diffuse , hot plasma that forms in shocks where the stellar winds from the clusters impact the ism @xcite .
the excess point sources are probably young , ob and wolf - rayet stars in binaries ( e.g. , * ? ? ?
we have presented a catalog of 9017 x - ray sources located in the inner 2@xmath1 by 0.8@xmath1 around the galactic center .
this increases the number of sources known in the region by a factor of 2.5 . for all of the sources , we provide tables listing their positions ( table [ tab : positions ] ) , photometry , and colors ( table [ tab : phot ] ) . of these sources ,
6760 have hard colors that are consistent with high absorptions columns @xmath12@xmath39@xmath38 @xmath14 , which indicates that they lie at or beyond the galactic center .
in addition , the positions of the x - ray sources in this catalog are more accurate than earlier versions .
this catalog contains 2029 sources with @xmath30.5 uncertainties ( 90% confidence ) , and another 3981 with uncertainties between 0.5 and 1 .
this catalog will be excellent for comparisons with multi - wavelength ones , in order to search for young stars , high - mass x - ray binaries , and pulsars ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the luminosity range that we cover , from @xmath74 to @xmath62 ( 0.58.0 kev ; assuming a @xmath60=1.5 power law , @xmath12=@xmath13 @xmath14 , and @xmath11=8 kpc ) , is at least an order of magnitude fainter than studies of local group galaxies ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
consequently , the natures of the sources that we study are also very different . whereas the detectable stellar population of external galaxies in x - rays is dominated by accreting black holes and neutron stars ,
most of our sources are probably cataclysmic variables ( e.g. , * ? ? ?
the hardness of the x - ray colors ( fig .
[ fig : hit ] ) suggests that the sources are specifically magnetically - accreting white dwarfs @xcite .
therefore , the x - ray population probably represents old stars .
indeed , the spatial distribution of sources brighter than @xmath65 ( 28 kev ) traces that of the old stellar population ( fig .
[ fig : londist ] ) .
this makes the population of x - ray sources in the galactic center similar to those seen in globular clusters ( e.g. , * ? ? ?
* ; * ? ? ?
although the distribution of the majority of the x - ray sources traces that of the old stellar population , we have found 2.8@xmath4 evidence for an excess of sources in two regions where young , massive stars are forming : in the inner few arcminutes around , and in the region where the arches and quintuplet star clusters lie .
the excess of sources near these young star clusters also appears in the number of sources as a function of limiting flux , in which relatively more bright x - ray sources are found near the arches and quintuplet ( fig .
[ fig : lognlogs ] and table [ tab : lognlogs ] ) . in total , these two regions contain a couple dozen more bright sources than our stellar mass model predicts .
we suggest that these excess x - ray sources are part of the young stellar population in these region @xcite . in the near future
, we will publish additional ob and wolf - rayet stars that have been identified through infrared spectroscopy of counterparts to x - ray sources ( j. mauerhan , in prep ) .
llcccl[htp ] 174354.8 - 294441 & 1e 1740.7 - 2942 & 265.97864 & @xmath75 & 0.5 & @xcite + 174417.2 - 293943 & ax j1744.3 - 2940 & 266.07190 & @xmath76 & 0.5 & @xcite + 174433.0 - 284427 & bursting pulsar & 266.13788 & @xmath77 & 0.5 & @xcite + 174451.6 - 292042 & ks 1741 - 293 & 266.21515 & @xmath78 & 0.5 & @xcite + 174457.4 - 285021 & xmm j174457 - 2850.3 & 266.23944 & @xmath79 & 0.3 & @xcite + [ 2pt ] 174502.3 - 285449 & granat 1741.9 - 2853 & 266.25983 & @xmath80 & 0.4 & @xcite + 174535.6 - 290133 & ax j1745.6 - 2901 & 266.39853 & @xmath81 & 0.4 & @xcite + 174535.5 - 290124 & & 266.39822 & @xmath82 & 0.3 & @xcite + 174537.1 - 290104 & 1a 1742 - 289 & 266.40494 & @xmath83 & 0.4 & @xcite + 174538.0 - 290022 & & 266.40863 & @xmath84 & 0.3 & @xcite + [ 2pt ] 174540.0 - 290005 & & 266.41699 & @xmath85 & 0.4 & @xcite + 174540.0 - 290030 & & 266.41684 & @xmath86 & 0.3 & @xcite + 174540.9 - 290014 & & 266.42078 & @xmath87 & 0.4 & @xcite + 174553.9 - 290346 & swift j174553.9 - 290347 & 266.47467 & @xmath88 & 0.4 & +
174554.4 - 285455 & xmm j174554.4 - 285456 & 266.47690 & @xmath89 & 0.4 & @xcite + [ 2pt ] 174621.0 - 284342 & 1e 1743.1 - 2843 & 266.58768 & @xmath90 & 0.4 & @xcite + 174702.5 - 285259 & sax j1747.0 - 2853 & 266.76080 & @xmath91 & 0.4 & @xcite + & xte j1748 - 288 & 267.02108 & 28.47383 & 0.6 & @xcite + & xmm j174544 - 2913.0 & 266.43546 & 29.21683 & 4.0 & @xcite + a small fraction of the x - ray sources should be accreting black holes and neutron stars .
around 300 such x - ray binaries are known in the galaxy , about half of which contain low - mass donors that over fill their roche lobe , and half of which contain high - mass ( ob and wolf - rayet ) stars that donate mass through a stellar wind @xcite .
these x - ray binaries are most - easily identified when they are bright and variable @xcite . in total , over the history of x - ray astronomy , 19 x - ray sources in our survey field have been observed to be @xmath22@xmath62 in x - rays , and have varied by at least an order of magnitude in x - ray flux ( table [ tab : transients ] ) .
fifteen of these transient x - ray sources were bright during the time span of our _ chandra _ observations ( 1a 1742 - 289 and xte j1748 - 288 never entered outburst ) .
half of them have been discovered in the last 9 years using _ chandra _ , _ xmm - newton _ or _ swift _
( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
surprisingly , despite having obtained 600 ks of new data in 2006 and 2007 , we did not detect any new , bright ( @xmath22@xmath62 ) , transient x - ray sources .
this suggests that we have identified all of the x - ray binaries that are active on time scales of a decade . as mentioned in
[ ref : obs ] , the tables from this work will be available in the electronic edition of this journal , and additional products will be made available from the authors web site .
the data available from the authors site includes fits images of all of the images presented in this paper , as well as the averaged event lists , snapshot images , spectra , and calibration files for each source in the catalog .
combined with an increasing amount of multi - wavelength data , this data set can be used to better understand the interactions between stars and interstellar media in the galactic center , and the population of x - ray emitting objects in general .
mpm , rmb , wnb , gcb , psb , ac , sdh , jcm , qdw , zw , and fyz received support from nasa thorugh chandra award number g06 - 7135 issued by the chandra x - ray observatory center , which is operated by the smithsonian astrophsyical observatory for and on behalf of the national aeronautics space administration under contract nas8 - 03060 .
tjwl and nek received funding for basic research in astronomy at the naval research laboratory , which is supported by 6.1 base funding .
0 alexander , d. m. 2003 , , 126 , 539 baganoff , f. k. 2003 , , 591 , 891 bauer , f. e. , alexander , d. m. , brandt , w. n. , schneider , d. p. , treister , e. , hornschmeier , a. e. , & garmire , g. p. 2004
, , 128 , 2048 buccheri , r. 1983 , , 128 , 245 berghfer , t. w. , schmitt , j. h. m. m. , danner , r. , & cassinelli , j. p. 1997
, , 322 , 167 brandt , w. n. 2001 , , 122 , 2810 broos , p. s. , townsley , l. k. , getman , k. , & bauer , f. e. 2002 , acis extract , an acis point source extraction package " ( university park : the pennsylvania state university ) , http://www.astro.psu.edu / xray / docs / tara/. chevalier , c. & ilovaisky , s. a. 1998 , , 330 , 201 cox , a. n. 2000 , allen s astrophysical quantities , 4th ed . , new york , aip press , springer davies , r. d. , walsh , d. , browne , i. w. a. , edwards , m. r. , & noble , r. c. 1976 , , 261 , 476 , h. & ishida , m. 1999 , , 120 , 277 figer , d. f. , kim , s. s. , morris , m. , serabyn , e. , rich , r. m. , & mclean , i. s. 1999 , , 525 , 750 figer , d. f. , rich , m. r. , kim , s. s. , morris , m. , & serabyn , e. 2004 , , 601 , 319 freeman , p. e. , kashyap , v. , rosner , r. , & lamb , d. q. 2002 , , 138 , 185 gagn , m. oksala , m. e. , cohen , d. h. , tonnesen , s. k. , ud - doula , a. , owocki , s. p. , townsend , r. h. , & mcfarlane , j. j. 2005 , , 628 , 968 gaensler , b. m. & slane , p. ) .
2006 , , 44 , 17 getman , k. v. 2005 , , 160 , 319 heinke , c. o. , wijnands , r. , cohn , h. n. , lugger , p. m. , grindlay , j. e. , pooley , d. , & lewin , w. h. g. 2006 , , 651 , 1098 hjellming , r. m. , rupen , m. p. , & mioduszewski , a. j. 1998 , iauc 6934 hogg , d. w. 2001 , , 121 , 1207 hong , j. , van den berg , m. , schlegel , e. m. , grindlay , j. e. , koenig , x. , laycock , s. , & zhao , p. 2005a
, , 635 , 907 in t zand , j. , heise , j. , bazzano , a. , ubertini , p. , muller , j. m. , & smith , m. 1997 , iauc 6618 kennea , j. a. , wijnands , r. , burrows , d. n. , nousek , j. , & gehrels , n. 2006 , atel , 920 , 1 kent , s. m. , dame , t. m. , & fazio , g. 1991 , , 378 , 131 kilgard , r. e. 2005 , , 159 , 214 koyama , k. 2007 , , 59 , s245 kraft , r. p. , burrows , d. n. , & nousek , j. a. 1991 , , 374 , 344 launhardt , r. , zylka , r. , & mezger , p. g. 2002 , , 384 , 112 law , c. & yusef - zadeh , f. 2004 , , 611 , 858 laycock , s. , grindlay , j. , van den berg , m. , zhao , p. , hong , j. , koenig , x. , schlegel , e. m. , & persson , s. e. 2005 , , 634 , l53 li , j. , kastner , j. h. , prigozhin , g. y. , schulz , n. s. , feilgelson , e. d. , & getman , k. v. 2004 , , 610 , 1204 liu , q. z. , van paradijs , j. , & van den heuvel , e. p. j. 2006 , , 455 , 1165 liu , q. z. , van paradijs , j. , & van den heuvel , e. p. j. 2007 , , 469 , 807 lu , f. j. , wang , q. d. , & lang , c. c. 2003 , , 126 , 319 lyons , l. 1991 , _ a practical guide to data analysis for physical science students _ , cambridge university press .
manners , j. c. 2003 , , 343 , 293 maeda , y. , koyama , k. , sakano , m. , takeshima , t. , & yamauchi , s. 1996 , , 48 , 417 marshall , d. j. , robin , a. c. , reyl , c. , schultheis , m. , & picaud , s. 2006 , , 453 , 635 mauerhan , j. c. , muno , m. p. , & morris , m. 2007 , , 662 , 574 mcnamara , d. h. , madsen , j. b. , barnes , j. , & ericksen , b. f. 2000 , , 112 , 202 mezger , p. g. & pauls t. 1979 , iaus , 84 , 357 mikles , v. j. , eikenberry , s. s. , muno , m. p. , bandyopadhyay , r. m. , & shannon , p. 2006
, , 651 , 408 muno , m. p. 2003a , , 589 , 225 muno , m. p. , baganoff , f. k. & arabadjis , j. s. 2003b , , 598 , 474 muno , m. p. , baganoff , f. k. , bautz , m. w. , brandt , w. n. , garmire , g. p. , & ricker , g. r. 2003c , , 599 , 465 muno , m. p. 2004a , , 613 , 326 muno , m. p. 2004b , , 613 , 1179 muno , m. p. , pfahl , e. , baganoff , f. k. , brandt , w. n. , ghez , a. , lu , j. , & morris , m. r. 2005 , , 622 , l113 muno , m. p. , bauer , f. e. , bandyopadhyay , r. m. , & wang , q. d. 2006a , , 165 , 173 muno , m. p. , bower , g. c. , burgasser , a. j. , baganoff , f. k. , morris , m. r. , & brandt , w. n. 2006b , , 638 , 183 muno , m. p. , baganoff , f. k. , brandt , w. n. , morris , m. r. , & starck , j. l. 2007 , , 669 , 251 murdoch , h. s. , crawford , d. f. , & jauncey , d. 1973 , , 183 , 1 negueruela , i. 1998 , , 338 , 505 nishiyama , s. 2006 , , 638 , 839 perna , r. , narayan , r. , rybicki , g. , stella , l. & treves , a. 2003 , , 594 , 936 plucinsky , p. p 2008 , , 174 , 366 porquet , d. , rodriguez , j. , corbel , s. , goldoni , p. , warwick , r. s. , goldwurm , a. , & decourchelle , a. 2003 , , 406 , 299 porquet , d. , grosso , n. , blanger , g. , goldwurm , a. , yusef - zadeh , f. , warwick , r. s. , & predehl , p. 2005
, , 443 , 571 predehl , p. & schmitt , j. h. m. m. 1995 , , 293 , 889 preibisch , t. & feigelson , e. d. 2005 , , 160 , 390 psaltis , d. 2006 , in compact stellar x - ray source , eds . w. lewin & m. van der klis , cambridge university press , pp .
138 ramrez , s. v. , arendt , r. g. , sellgren , k. , stolovy , s. r. , cotera , a. , smith , h. a. , & yusef - zadeh , f. 2007 , , in press , arxiv:0709.3113 ransom , s. m. , eikenberry , s. s. , & middleditch , j. 2002 , , 124 , 1788 reid , n. j. 1993 , , 31 , 345 revnivtsev , m. , sazonov , s. , gilfanov , m. , churazov , e. , & sunyaev , r. 2006 , , 452 , 169 revnivtsev , m. & sazonov , s. 2007 , , 471 , 159 reike , g. h. & lebofsky , m. j. 1985 , , 288 , 618 sakano , m. , warwick , r. s. , decourchelle , a. , & wang , q. d. 2005 , , 357 , 1211 scargle , j. d. 1998 , , 504 , 405 sidoli , l. , belloni , t. , & mereghetti , s. 2001 , , 368 , 835 sidoli , l. , mereghetti , s. , israel , g. l. , chiappetti , l. , treves , a. , & orlandini , m. 1999 , , 525 , 215 skrutskie , m. f. 2006 , , 131 , 1163 takagi , s. , murakami , h. , & koyama , k. 2002 , , 573 , 275 townsley , l. k. , feigelson , e. d. , montmerle , t. , broos , p. s. , chu , y .- h . , & garmire , g. p. 2003
, , 593 , 874 trudolyubov , s. & priedhorsky , w. 2004 , , 616 , 821 useda , y. 1999 , , 518 , 656 verbunt , f. , bunk , w. h. , ritter , h. , & pfeffermann , e. 1997 , , 327 , 602 waldorn , w. l. & cassinelli , j. p. 2007
, , 668 , 456 walter , f. m. , wolk , s. j. , & neuhauser , r. 1996 , , 379 , 233 wang , q. d. , lu , f. j. , & lang , c. c. 2002b , , 581 , 1148 wang , q. d. , dong , h. , & lang , c. 2006 , , 371 , 38 wang , q. d. , gotthelf , e. v. , & lang , c. c. 2002 , , 415 , 148 wang , q. d. 2004 , , 612 , 159 , b. 1995 , _ cataclysmic variable stars _ , cambridge university press weisskopf , m. c. , brinkman , b. , canizares , c. , garmire , g. , murray , s. , & van speybroeck , l. p. 2002
, , 114 , 1 weisskopf , m. c. , wu , k. , trimble , v. , odell , s. l. , elsner , r. f. , zavlin , v. e. & kouveliotou , c. 2007 , , 657 , 1026 wijnands , r. 2006 , , 449 , 1117 wijnands , r. , miller , j. m. , & wang , q. d. 2002 , , 579 , 422 wijnands , r. & wang , q. d. , , 568 , l93 woods , p. m. & thompson , c. 2006 , in compact stellar x - ray source , eds . w. lewin & m. van der klis , cambridge university press , pp .
547 - 586 yusef - zadeh , f. , law , c. , wardle , m. , wang , q. d. , fruscione , a. , lang , c. c. , & cotera , a. 2002 , , 570 , 665 lclccccccc[htp ] 2001 jul 21 04:35:09 & 2286 & gcs 28 & 11.6 & 266.01026 & -29.87409 & 283.8 & & & 0.5 + 2001 jul 21 08:03:39 & 2289 & gcs 29 & 11.6 & 265.81895 & -29.77175 & 283.8 & & & 0.5 + 2001 jul 21 11:32:10 & 2290 & gcs 30 & 11.6 & 265.62792 & -29.66920 & 283.8 & & & 0.5 + 2000 aug 30 16:59:32 & 658 & 1e 1740.7 - 2942 & 9.2 & 265.97607 & -29.75041 & 270.8 & & & 0.5 + 2001 jul 20 18:09:40 & 2278 & gcs 25 & 11.6 & 266.12796 & -29.70796 & 283.8 & & & 0.5
+ [ 2pt ] 2006 jun 27 16:38:52 & 7042 & deep gcs 12 & 14.4 & 265.92331 & -29.62836 & 297.2 & 8 & & 0.2 + 2006 jul 01 13:53:08 & 7346 & deep gcs 12 & 15.2 & 265.92332 & -29.62846 & 297.2 & & 8 & 0.3 + 2006 jun 28 16:24:41 & 7345 & deep gcs 12 & 10.0 & 265.92328 & -29.62841 & 297.2 & & 2 & 0.4 + 2001 jul 20 21:38:10 & 2281 & gcs 26 & 11.6 & 265.93676 & -29.60580 & 283.8 & & 8 & 0.2 + 2007 apr 29 06:03:16 & 7043 & deep gcs 13 & 34.2 & 265.72756 & -29.51198 & 82.9 & 9 & & 0.13 + [ 2pt ] 2001 jul 21 01:06:39 & 2283 & gcs 27 & 11.6 & 265.74591 & -29.50336 & 283.8 & & 7 & 0.2 + 2001 jul 20 08:00:49 & 2270 & gcs 22 & 10.6 & 266.24516 & -29.54168 & 283.8 & & & 0.5 + 2005 jul 22 01:58:12 & 5892 & sgr c & 97.9 & 266.08819 & -29.43665 & 278.0 & 25 & & 0.08 + 2001 jul 20 11:12:40 & 2272 & gcs 23 & 11.6 & 266.05414 & -29.43966 & 283.8 & & 15 & 0.2 + 2006 sep 15 14:54:55 & 7041 & deep gcs 10 & 19.9 & 265.84657 & -29.35563 & 269.0 & 4 & & 0.2 + [ 2pt ] 2006 sep 27 01:21:42 & 8214 & deep gcs 10 & 17.7 & 265.84656 & -29.35560 & 269.0 & 4 & & 0.2 + 2001 jul 20 14:41:10 & 2275 & gcs 24 & 11.6 & 265.86352 & -29.33749 & 283.8 & & 4 & 0.2 + 2007 apr 25 14:04:05 & 7040 & deep gcs 8 & 36.7 & 266.35283 & -29.38207 & 84.0 & 15 & & 0.10 + 2001 jul 19 06:41:38 & 2296 & gcs 19 & 11.1 & 266.36200 & -29.37542 & 283.8 & & 6 & 0.2 + 2006 oct 31 03:58:52 & 7038 & deep gcs 6 & 19.7 & 266.15861 & -29.29070 & 268.2 & 3 & & 0.2 + [ 2pt ] 2007 feb 19 01:49:07 & 8459 & deep gcs 6 & 19.3 & 266.15788 & -29.27900 & 91.6 & & 16 & 0.3 + 2001 jul 19 10:01:48 & 2267 & gcs 20 & 8.7 & 266.17124 & -29.27356 & 283.8 & & 11 & 0.2 + 2006 jul 27 05:11:38 & 7039 & deep gcs 7 & 37.8 & 265.96598 & -29.18723 & 276.1 & 7 & & 0.17 + 2001 jul 20 04:37:11 & 2268 & gcs 21 & 10.8 & 265.98088 & -29.17148 & 283.8 & & 7 & 0.2 + 2006 aug 24 20:26:37 & 7037 & deep gcs 5 & 39.4 & 266.47130 & -29.22528 & 275.2 & 12 & & 0.12 + [ 2pt ] 2001 jul 18 20:49:28 & 2291 & gcs 16 & 10.6 & 266.47846 & -29.20889 & 283.8 & & 11 & 0.2 + 2006 jul 26 18:06:34 & 7035 & deep gcs 3 & 38.0 & 266.27807 & -29.12211 & 276.3 & 5 & & 0.12 + 2001 jul
19 00:01:18 & 2293 & gcs 17 & 11.1 & 266.28790 & -29.10730 & 283.8 & & 21 & 0.2 + 2007 jul 19 20:45:11 & 8567 & deep gcs 4 & 19.5 & 266.09028 & -29.02150 & 278.3 & 10 & & 0.18 + 2007 jul 09 16:12:13 & 7036 & deep gcs 4 & 19.9 & 266.09096 & -29.02114 & 283.5 & 7 & & 0.15 + [ 2pt ] 2001 jul 19 03:21:28 & 2295 & gcs 18 & 11.1 & 266.09774 & -29.00538 & 283.8 & & 8 & 0.2 + 2001 jul 14 01:51:10 & 1561b & & 13.5 & 266.41344 & -29.01281 & 264.7 & & 24 & 0.12 + 2000 oct 26 18:15:11 & 1561a & & 35.7 & 266.41344 & -29.01281 & 264.7 & & 68 & 0.06 + 2007 jul 20 02:27:01 & 7557 & & 5.0 & 266.42069 & -29.01498 & 278.4 & & 9 & 0.15 + 2004 aug 28 12:03:59 & 5360 & & 5.1 & 266.41477 & -29.01211 & 271.0 & & 9 & 0.1 + [ 2pt ] 2005 jul 29 19:51:11 & 5952 & & 43.1 & 266.41508 & -29.01219 & 275.5 & & 76 & 0.06 + 2005 jul 30 19:38:31 & 5953 & & 45.4 & 266.41506 & -29.01218 & 275.3 & & 78 & 0.06 + 2005 jul 24 19:58:27 & 5950 & & 48.5 & 266.41519 & -29.01222 & 276.7 & & 64 & 0.06 + 2005 jul 27 19:08:16 & 5951 & & 44.6 & 266.41512 & -29.01219 & 276.0 & & 66 & 0.06 + 2006 jul 17 03:58:28 & 6363 & & 29.8 & 266.41541 & -29.01228 & 279.5 & & 59 & 0.06 + [ 2pt ] 2006 jul 30 14:30:26 & 6643 & & 5.0 & 266.41510 & -29.01218 & 275.4 & & 11 & 0.18 + 2005 aug 01 19:54:13 & 5954 & & 18.1 & 266.41502 & -29.01215 & 274.9 & & 36 & 0.1 + 2006 sep 25 13:50:35 & 6645 & & 5.1 & 266.41448 & -29.01195 & 268.3 & & 7 & 0.18 + 2006 aug 22 05:54:34 & 6644 & & 5.0 & 266.41484 & -29.01202 & 271.7 & & 10 & 0.18 + 2004 jul 06 22:29:57 & 4684 & & 49.5 & 266.41597 & -29.01236 & 285.4 & & 94 & 0.06 + [ 2pt ] 2004 jul 05 22:33:11 & 4683 & & 49.5 & 266.41605 & -29.01238 & 286.2 & & 98 & 0.06 + 2006 oct 29 03:28:20 & 6646 & & 5.1 & 266.41425 & -29.01178 & 264.4 & & 9 & 0.18 + 2006 jul 04 11:01:35 & 6642 & & 5.1 & 266.41633 & -29.01237 & 288.4 & & 7 & 0.18 + 2003 jun 19 18:28:55 & 3549 & & 24.8 & 266.42092 & -29.01052 & 346.8 & & 49 & 0.06 + 2006 jun 01 16:07:52 & 6641 & & 5.1 & 266.42018 & -29.00440 & 69.7 & & 7 & 0.18 + [ 2pt ] 2002 may 22 22:59:15 & 2943 & & 34.7 & 266.41991 & -29.00407 & 75.5 & & 86 & 0.15 + 2002 jun 03 01:24:37 & 3665 & & 89.9 & 266.41992 & -29.00407 & 75.5 & & 162 & 0.06 + 2002 may 25 15:16:03 & 3392 & & 165.8 & 266.41992 & -29.00408 & 75.5 & 23 & & 0.06 + 2002 may 28 05:34:44 & 3393 & & 157.1 & 266.41992 & -29.00407 & 75.5 & & 759 & 0.06 + 2002 may 24 11:50:13 & 3663 & & 38.0 & 266.41993 & -29.00407 & 75.5 & & 82 & 0.10 + [ 2pt ] 2006 apr 11 05:33:20 & 6639 & & 4.5 & 266.41890 & -29.00369 & 86.2 & & 11 & 0.18 + 2006 may 03 22:26:26 & 6640 & & 5.1 & 266.41935 & -29.00383 & 82.8 & & 8 & 0.18 + 2005 feb 27
06:26:04 & 6113 & & 4.9 & 266.41870 & -29.00353 & 90.6 & & 6 & 0.12 + 2002 may 07 09:25:07 & 2954 & & 12.5 & 266.41938 & -29.00374 & 82.1 & & 25 & 0.15 + 2007 feb 11 06:16:55 & 7554 & & 5.1 & 266.41853 & -29.00343 & 92.6 & & 11 & 0.2 + [ 2pt ] 2002 feb 19 14:27:32 & 2951 & & 12.4 & 266.41867 & -29.00335 & 91.5 & & 35 & 0.15 + 2002 mar 23 12:25:04 & 2952 & & 11.9 & 266.41897 & -29.00343 & 88.2 & & 23 & 0.15 + 2002 apr 19 10:39:01 & 2953 & & 11.7 & 266.41923 & -29.00349 & 85.2 & & 28 & 0.15 + 2007 mar 25 22:56:07 & 7555 & & 5.1 & 266.41420 & -29.00013 & 88.0 & & 10 & 0.15 + 2001 jul 18 11:13:58 & 2282 & gcs 13 & 10.6 & 266.59457 & -29.04224 & 283.8 & & & 0.5 + [ 2pt ] 2001 jul 18 14:25:48 & 2284 & gcs 14 & 10.6 & 266.40415 & -28.94090 & 283.8 & & 23 & 0.2 + 2007 feb 22 03:29:33 & 7034 & deep gcs 1 & 39.6 & 266.20175 & -28.83783 & 91.0 & 36 & & 0.03 + 2001 jul 18 17:37:38 & 2287 & gcs 15 & 10.6 & 266.21412 & -28.83905 & 283.8 & & 18 & 0.2 + 2000 jul 07 19:05:19 & 945 & gc arc & 48.8 & 266.58221 & -28.87193 & 284.4 & 12 & & 0.12 + 2004 jun 09 08:50:32 & 4500 & arches cluster & 98.6 & 266.48260 & -28.81691 & 55.2 & 26 & & 0.10 + [ 2pt ] 2007 feb 14 03:59:44 & 7048 & deep gcs 27 & 38.2 & 266.71020 & -28.87828 & 92.3 & 6 & & 0.17 + 2001 jul 18 00:48:28 & 2273 & gcs 10 & 11.2 & 266.71024 & -28.87570 & 283.8 & & 6 & 0.2 + 2001 jul 18 04:16:58 & 2276 & gcs 11 & 11.6 & 266.52002 & -28.77435 & 283.8 & & 12 & 0.2 + 2001 jul 18 07:45:28 & 2279 & gcs 12 & 11.6 & 266.33030 & -28.67280 & 283.8 & & 9 & 0.2 + 2007 feb 24 04:56:00 & 7047 & deep gcs 26 & 36.7 & 266.31970 & -28.66905 & 90.8 & 9 & & 0.17 + [ 2pt ] 2007 jul 19 10:04:05 & 7046 & deep gcs 25 & 36.5 & 266.82943 & -28.72059 & 278.4 & 18 & & 0.10 + 2001 jul 17 14:11:51 & 2288 & gcs 7 & 11.1 & 266.82550 & -28.70883 & 283.8 & & 10 & 0.2 + 2006 nov 01 14:11:44 & 7044 & deep gcs 23 & 37.9 & 266.63030 & -28.61480 & 268.2 & 12 & & 0.11 + 2001 jul 17 17:51:28 & 2292 & gcs 8 & 11.6 & 266.63554 & -28.60780 & 283.8 & & 12 & 0.2 + 2001 jul 17 21:19:58 & 2294 & gcs 9 & 11.6 & 266.44598 & -28.50635 & 283.8 & & 8 & 0.2 + [ 2pt ] 2007 feb 23 07:15:57 & 7045 & deep gcs 24 & 37.0 & 266.43720 & -28.50006 & 90.9 & 9 & & 0.13 + 2001 jul
16 11:52:55 & 2277 & gcs 4 & 10.4 & 266.94060 & -28.54206 & 283.8 & & & 0.5 + 2000 mar 29 09:44:36 & 944 & sgr b2 & 97.5 & 266.78070 & -28.44160 & 87.8 & 23 & & 0.10 + 2001 jul 16 15:01:25 & 2280 & gcs 5 & 10.4 & 266.75080 & -28.44106 & 283.8 & & 9 & 0.2 + 2001 jul 16 18:09:55 & 2285 & gcs 6 & 10.4 & 266.56137 & -28.33985 & 283.4 & & & 0.5 + [ 2pt ] 2001 jul 16 02:15:50 & 2269 & gcs 1 & 10.5 & 267.05519 & -28.37520 & 283.8 & & & 0.5 + 2001 jul 16 05:35:55 & 2271 & gcs 2 & 10.4 & 266.86561 & -28.27427 & 283.8 & & & 0.5 + 2001 jul 16 08:44:25 & 2274 & gcs 3 & 10.4 & 266.67640 & -28.17316 & 283.8 & & & 0.5 lccccccccccccccc 1 & 174457.1285740 & 266.23813 & @xmath92 & 1.8 & c & full & soft & 9.8 & 14 & 372669 & 0.90 & 1.50 & 9.8 & 32.2 & s + 2 & 174457.4285622 & 266.23941 & @xmath93 & 2.0 & c & 2004 & full & 10.2 & 15 & 377775 & 0.90 & 4.51 & 11.2 & 26.8 & lb + 3 & 174459.1290604 & 266.24620 & @xmath94 & 1.3 & c & full & full & 10.8 & 12 & 571227 & 0.87 & 4.51 & 11.8 & 9.0 & sc + 4 & 174459.9290324 & 266.24982 & @xmath95 & 2.0 & c & 2002 & full & 9.3 & 23 & 923358 & 0.90 & 4.51 & 9.6 & 10.0 & g + 5 & 174459.9290538 & 266.24994 & @xmath96 & 1.5 & c & full & hard & 10.3 & 18 & 732385 & 0.90 & 4.51 & 11.5 & 9.6 & b + [ 2pt ] 6 & 174500.2290057 & 266.25113 & @xmath97 & 2.6 & w & 2005 & soft & 8.6 & 18 & 423683 & 0.76 & 1.50 & 6.3 & 22.7 & c + 7 & 174500.6290443 & 266.25244 & @xmath98 & 1.3 & c & full & full & 9.7 & 20 & 810677 & 0.69 & 4.51 & 6.8 & 10.2 & lc + 8 & 174501.3285501 & 266.25580 & @xmath99 & 0.5 & w & 242 & full & 10.1 & 18 & 423686 & 0.77 & 4.51 & 7.9 & 23.3 & lcp + 9 & 174501.4290408 & 266.25598 & @xmath100 & 1.1 & c & full & hard & 9.3 & 26 & 945772 & 0.80 & 4.51 & 7.6 & 10.2 & c + 10 & 174501.7290313 & 266.25732 & @xmath101 & 1.1 & c & full & hard & 8.9 & 27 & 956403 & 0.90 & 4.51 & 9.3 & 11.5 & gb + [ 2pt ] 11 & 174501.8290206 & 266.25760 & @xmath102 & 1.1 & c & full & full & 8.6 & 26 & 943948 & 0.69 & 4.51 & 5.3 & 10.8 & c + 12 & 174501.9285719 & 266.25827 & @xmath103 & 1.8 & c & full & soft & 8.9 & 18 & 423686 & 0.69 & 1.50 & 5.1 & 24.7 & lbc + 13 & 174502.2285749 & 266.25946 & @xmath104 & 1.0 & c & 5951 & full & 8.7 & 18 & 423686 & 0.87 & 4.51 & 7.5 & 23.5 & c + 14 & 174502.4290205 & 266.26007 & @xmath105 & 1.1 & c & 2002 & soft & 8.5 & 27 & 956403 & 0.70 & 1.50 & 4.9 & 9.7 & c + 15 & 174502.4290453 & 266.26039 & @xmath106 & 1.4 & c & full & soft & 9.4 & 29 & 967611 & 0.88 & 1.50 & 8.4 & 11.4 & - + [ 2pt ] 16 & 174502.5290415 & 266.26077 & @xmath107 & 2.0 & w & full & tile & 9.1 & 29 & 967611 & 0.75 & 4.51 & 6.5 & 11.2 & lbc + 17 & 174502.7290127 & 266.26101 & @xmath108 & 1.6 & c & 3392 & full & 8.3 & 27 & 956431 & 0.89 & 4.51 & 7.7 & 9.2 & - + 18 & 174502.8290429 & 266.26198 & @xmath109 & 1.3 & c & full & soft & 9.1 & 30 & 972102 & 0.77 & 1.50 & 6.1 & 11.1 & bc + 19 & 174502.9285920 & 266.26222 & @xmath110 & 1.1 & c & full & full & 8.2 & 18 & 423686 & 0.90 & 4.51 & 7.3 & 19.4 & - + 20 & 174503.7285805 & 266.26531 & @xmath111 & 1.2 & c & 2005 & full & 8.3 & 20 & 440908 & 0.70 & 4.51 & 4.8 & 21.7 & bc + [ 2pt ] 21 & 174503.8290004 & 266.26584 & @xmath112 & 0.9 & c & full & full & 8.0 & 30 & 985035 & 0.70 & 4.51 & 4.5 & 9.5 & gc + 22 & 174504.1285902 & 266.26740 & @xmath113 & 4.7 & w & 6646 & soft & 8.0 & 30 & 955147 & 0.74 & 1.50 & 4.7 & 13.8 & bc + 23 & 174504.2285653 & 266.26758 & @xmath114 & 1.4 & c & 242 & full & 8.6 & 22 & 457849 & 0.90 & 4.51 & 7.8 & 22.7 & l + 24 & 174504.2290410 & 266.26764 & @xmath115 & 1.5 & c & 2002 & soft & 8.7 & 34 & 1011810 & 0.70 & 1.50 & 4.9 & 8.7 & c + 25 & 174504.2290610 & 266.26841 & @xmath116 & 1.9 & c & 2005 & hard & 9.8 & 32 & 988816 & 0.83 & 4.51 & 8.5 & 10.3 & c + lcccccccccccccc 1 & 174457.1285740 & 4.50 & 86 & 26.3 & @xmath117 & 185 & 186.4 & @xmath118 & 6.8 & & 1.0 & @xmath119 & & + 2 & 174457.4285622 & 0.09 & 52 & 53.0 & @xmath120 & 357 & 298.8 & @xmath121 & & 8.2 & 4.6 & & @xmath122 & @xmath123 + 3 & 174459.1290604 & 2.57 & 118 & 88.1 & @xmath124 & 498 & 349.1 & @xmath125 & 1.9 & 13.3 & 3.5 & @xmath126 & @xmath127 & @xmath128 + 4 & 174459.9290324 & 1.09 & 96 & 91.1 & @xmath129 & 369 & 334.6 & @xmath130 & & 2.4 & 3.1 & @xmath131 & @xmath132 & + 5 & 174459.9290538 & 0.97 & 107 & 115.0 & @xmath133 & 550 & 434.9 & @xmath134 & & 10.3 & 4.8 & & @xmath135 & @xmath136 + [ 2pt ] 6 & 174500.2290057 & 1.99 & 41 & 25.1 & @xmath137 & 138 & 133.5 & @xmath138 & 1.5 & & 0.5 & @xmath139 & @xmath140 & + 7 & 174500.6290443 & 1.90 & 50 & 44.5 & @xmath141 & 316 & 200.3 & @xmath142 & & 11.1 & 4.6 & & @xmath143 & @xmath144 + 8 & 174501.3285501 & 40.35 & 175 & 69.8 & @xmath145 & 13585 & 3163.7 & @xmath146 & 9.7 & 1625.1 & 4.7 & @xmath147 & @xmath148 & @xmath149 + 9 & 174501.4290408 & 4.12 & 92 & 100.1 & @xmath150 & 523 & 371.2 & @xmath151 & & 11.9 & 5.4 & & @xmath152 & @xmath153 + 10 & 174501.7290313 & 0.76 & 89 & 104.1 & @xmath154 & 561 & 433.0 & @xmath155 & & 8.7 & 5.1 & @xmath156 & @xmath157 & @xmath158 + [ 2pt ] 11 & 174501.8290206 & 2.37 & 39 & 34.2 & @xmath159 & 248 & 144.9 & @xmath160 & & 8.0 & 4.6 & & @xmath161 & @xmath162 + 12 & 174501.9285719 & 0.88 & 41 & 19.9 & @xmath163 & 163 & 143.6 & @xmath164 & 2.1 & 3.2 & 3.2 & @xmath165 & @xmath166 & @xmath167 + 13 & 174502.2285749 & 5.62 & 60 & 38.7 & @xmath168 & 378 & 281.3 & @xmath169 & 1.6 & 6.6 & 2.6 & @xmath170 & @xmath171 & @xmath172 + 14 & 174502.4290205 & 2.76 & 70 & 31.1 & @xmath173 & 159 & 109.7 & @xmath174 & 1.8 & 3.2 & 2.8 & @xmath175 & @xmath176 & @xmath177 + 15 & 174502.4290453 & 7.28 & 172 & 91.6 & @xmath178 & 349 & 340.4 & @xmath179 & 3.3 & & 0.9 & @xmath180 & @xmath181 & + [ 2pt ] 16 & 174502.5290415 & 0.04 & 69 & 56.4 & @xmath182 & 218 & 194.8 & @xmath183 & 0.6 & & 3.1 & @xmath184 & @xmath185 & + 17 & 174502.7290127 & 1.10 & 80 & 101.9 & @xmath186 & 372 & 312.8 & @xmath187 & & 3.4 & 6.3 & & @xmath188 & @xmath189 + 18 & 174502.8290429 & 0.49 & 139 & 86.3 & @xmath190 & 308 & 279.6 & @xmath191 & 2.2 & & 2.9 & @xmath192 & & @xmath193 + 19 & 174502.9285920 & 1.28 & 43 & 33.7 & @xmath194 & 234 & 164.0 & @xmath195 & & 9.0 & 4.9 & & @xmath196 & @xmath197 + 20 & 174503.7285805 & 0.63 & 17 & 15.3 & @xmath198 & 143 & 75.0 & @xmath199 & & 11.2 & 4.7 & @xmath200 & @xmath201 & @xmath202 + [ 2pt ] 21 & 174503.8290004 & 1.94 & 56 & 51.9 & @xmath203 & 341 & 232.2 & @xmath204 & & 6.8 & 4.3 & @xmath205 & @xmath206 & @xmath207 + 22 & 174504.1285902 & 0.22 & 27 & 19.9 & @xmath198 & 90 & 71.8 & @xmath208 & & 1.2 & 2.7 & @xmath209 & @xmath210 & @xmath211 + 23 & 174504.2285653 & 2.59 & 46 & 56.2 & @xmath212 & 419 & 353.4 & @xmath213 & & 10.3 & 6.5 & & & @xmath214 + 24 & 174504.2290410 & 2.66 & 121 & 66.9 & @xmath215 & 250 & 249.8 & @xmath216 & 2.1 & & 0.9 & @xmath217 & & + 25 & 174504.2290610 & 1.31 & 129 & 124.1 & @xmath218 & 543 & 496.0 & @xmath219 & & 3.1 & 4.6 & & @xmath220 & @xmath221 + llcccccc[htp ] 2 & 174457.4 - 285622 & gc & 5953 & @xmath222 & 4684 & @xmath223@xmath52@xmath224 & @xmath225 + 7 & 174500.6 - 290443 & gc & 7556 & @xmath226 & 3392 & @xmath227@xmath52@xmath228 & @xmath229 + 8 & 174501.3 - 285501 & gc & 4684 & @xmath230@xmath52@xmath231 & 1561a & @xmath232@xmath52@xmath233 & @xmath234 + 12 & 174501.9 - 285719 & f & 2284 & @xmath222 & 5953 & @xmath235@xmath52@xmath224 & @xmath236 + 16 & 174502.5 - 290415 & gc & 2943 & @xmath237 & 7556 & @xmath238@xmath52@xmath239 & @xmath240 + [ 2pt ] 23 & 174504.2 - 285653 & gc & 4683 & @xmath241 & 5953 & @xmath235@xmath52@xmath242 & @xmath243 + 31 & 174504.8 - 285410 & f & 5951 & @xmath244 & 6363 & @xmath245@xmath52@xmath246 & @xmath247 + 35 & 174505.2 - 285713 & f & 2953 & @xmath248 & 4683 & @xmath235@xmath52@xmath249 & @xmath229 + 45 & 174506.1 - 285710 & gc & 5954 & @xmath241 & 2953 & @xmath250@xmath52@xmath251 & @xmath252 + 53 & 174507.0 - 290452 & gc & 6113 & @xmath253 & 2953 & @xmath254@xmath52@xmath255 & @xmath256 + [ 2pt ] 54 & 174507.1 - 285720 & gc & 5950 & @xmath241 & 2951 & @xmath257@xmath52@xmath258 & @xmath259 + 58 & 174507.5 - 285614 & f & 3665 & @xmath260 & 6640 & @xmath261@xmath52@xmath262 & @xmath256 + 71 & 174508.4 - 290033 & f & 3663 & @xmath237 & 5360 & @xmath263@xmath52@xmath264 & @xmath265 + 80 & 174509.2 - 285457 & gc & 2951 & @xmath266 & 5954 & @xmath267@xmath52@xmath268 & @xmath236 + 89 & 174509.5 - 285502 & gc & 3392 & @xmath269 & 6363 & @xmath270@xmath52@xmath271 & @xmath272 + [ 2pt ] 96 & 174510.1 - 285624 & f & 3549 & @xmath241 & 1561a & @xmath235@xmath52@xmath273 & @xmath274 + 98 & 174510.2 - 285505 & gc & 3663 & @xmath237 & 6642 & @xmath275@xmath52@xmath276 & @xmath277 + 100 & 174510.3 - 290642 & gc & 7557 & @xmath278 & 2954 & @xmath279@xmath52@xmath280 & @xmath281 + 101 & 174510.4 - 285433 & gc & 4683 & @xmath237 & 5953 & @xmath270@xmath52@xmath224 & @xmath282 + 102 & 174510.4 - 285544 & gc & 5950 & @xmath237 & 3665 & @xmath283@xmath52@xmath228 & @xmath284 + [ 2pt ] 103 & 174510.4 - 285545 & gc & 5950 & @xmath237 & 3665 & @xmath285@xmath52@xmath228 & @xmath284 + 106 & 174510.6 - 285437 & gc & 3549 & @xmath286 & 3663 & @xmath227@xmath52@xmath224 & @xmath252 + 109 & 174510.8 - 285606 & gc & 6641 & @xmath287 & 3665 & @xmath285@xmath52@xmath228 & @xmath229 + 112 & 174510.9 - 285508 & gc & 5954 & @xmath266 & 4683 & @xmath288@xmath52@xmath249 & @xmath281 + 129 & 174511.6 - 285915 & gc & 5953 & @xmath269 & 3392 & @xmath289@xmath52@xmath290 & @xmath284 + [ 2pt ] 143 & 174512.1 - 290005 & gc & 2952 & @xmath286 & 3393 & @xmath291@xmath52@xmath292 & @xmath284 + 155 & 174512.4 - 285318 & gc & 7555 & @xmath226 & 3393 & @xmath283@xmath52@xmath228 & @xmath256 + 164 & 174512.8 - 285441 & gc & 6644 & @xmath287 & 2953 & @xmath293@xmath52@xmath255 & @xmath294 + 191 & 174514.1 - 285426 & gc & 5953 & @xmath286 & 5954 & @xmath279@xmath52@xmath255 & @xmath295 + 226 & 174515.1 - 290006 & gc & 4684 & @xmath269 & 3392 & @xmath296@xmath52@xmath290 & @xmath297 llcccccccc[htp ] 1 & 174457.1 - 285740 & f & 1561b & 100 & 2 & 3152 & @xmath298 & @xmath299 & @xmath300 + 3 & 174459.1 - 290604 & gc & 3663 & 10 & 2 & 19600 & @xmath301 & @xmath302 & @xmath303 + 71 & 174508.4 - 290033 & f & 5360 & 1000 & 2 & 2736 & @xmath304 & @xmath305 & @xmath306 + 186 & 174513.7 - 285638 & gc & 4684 & 10 & 2 & 21312 & @xmath286 & @xmath307 & @xmath303 + 257 & 174516.6 - 285412 & f & 4684 & 1000 & 2 & 16512 & @xmath308 & @xmath309 & @xmath310 + [ 2pt ] 304 & 174517.8 - 290653 & gc & 3392 & 1000 & 2 & 110160 & @xmath311 & @xmath312 & @xmath313 + 364 & 174519.5 - 285955 & gc & 4683 & 10 & 2 & 6896 & @xmath314 & @xmath315 & @xmath316 + 398 & 174520.3 - 290143 & gc & 2943 & 1000 & 2 & 3472 & @xmath317 & @xmath318 & @xmath319 + 424 & 174520.6 - 290152 & f & 3392 & 1000 & 5 & 33360 & @xmath320 & @xmath321 & @xmath322 + 424 & & & 3393 & 1000 & 4 & 22144 & @xmath323 & @xmath324 & @xmath325 + [ 2pt ] 424 & & & 5951 & 1000 & 2 & 23760 & @xmath326 & @xmath327 & @xmath328 + 470 & 174521.7 - 290151 & gc & 5950 & 100 & 2 & 10400 & @xmath329 & @xmath330 & @xmath331 + 472 & 174521.8 - 285912 & f & 3392 & 1000 & 3 & 18560 & @xmath314 & @xmath332 & @xmath333 + 663 & 174525.1 - 285703 & f & 3665 & 1000 & 2 & 1664 & @xmath334 & @xmath335 & @xmath336 + 811 & 174527.1 - 290730 & gc & 3549 & 10 & 2 & 3296 & @xmath241 & @xmath337 & @xmath338 + [ 2pt ] 1100 & 174530.3 - 290341 & gc & 3392 & 1000 & 3 & 21440 & @xmath339 & @xmath340 & @xmath341 + 1100 & & & 5950 & 1000 & 2 & 17120 & @xmath342 & @xmath343 & @xmath344 + 1183 & 174531.1 - 290219 & gc & 3393 & 10 & 2 & 70704 & @xmath345 & @xmath346 & @xmath347 + 1525 & 174534.2 - 290011 & f & 3392 & 100 & 2 & 49392 & @xmath348 & @xmath349 & @xmath350 + 1569 & 174534.5 - 290236 & gc & 3392 & 1000 & 2 & 73504 & @xmath348 & @xmath351 & @xmath352 + [ 2pt ] 1608 & 174534.8 - 290851 & f & 1561a & 100 & 2 & 21168 & @xmath301 & @xmath353 & @xmath354 + 1676 & 174535.5 - 290124 & gc & 3549 & 1000 & 2 & 15920 & @xmath355 & @xmath356 & @xmath357 + 1676 & & & 5950 & 100 & 2 & 16544 & @xmath358 & @xmath359 & @xmath360 + 1676 & & & 6644 & 1000 & 2 & 2304 & @xmath361 & @xmath362 & @xmath363 + 1686 & 174535.6 - 290133 & gc & 3665 & 1000 & 3 & 61632 & @xmath364 & @xmath365 & @xmath366 + [ 2pt ] 1686 & & & 6641 & 1000 & 2 & 4336 & @xmath367 & @xmath368 & @xmath369 + 1691 & 174535.7 - 285357 & f & 5951 & 100 & 2 & 29264 & @xmath370 & @xmath371 & @xmath372 + 1706 & 174535.8 - 290159 & gc & 3393 & 1000 & 2 & 90464 & @xmath269 & @xmath373 & @xmath374 + 1748 & 174536.1 - 290806 & gc & 3393 & 1000 & 2 & 75104 & @xmath375 & @xmath376 & @xmath377 + 1765 & 174536.3 - 285545 & f & 3392 & 100 & 4 & 7568 & @xmath378 & @xmath379 & @xmath380 | we present a catalog of 9017 x - ray sources identified in _
chandra_observations of a [email protected]@xmath1 field around the galactic center .
this enlarges the number of known x - ray sources in the region by a factor of 2.5 .
the catalog incorporates all of the acis - i observations as of 2007 august , which total 2.25 msec of exposure . at the distance to the galactic center ( 8 kpc )
, we are sensitive to sources with luminosities of @xmath2 ( 0.58.0 kev ; 90% confidence ) over an area of one square degree , and up to an order of magnitude more sensitive in the deepest exposure ( 1.0 msec ) around .
the positions of 60% of our sources are accurate to @xmath31 ( 95% confidence ) , and 20% have positions accurate to @xmath305 .
we search for variable sources , and find that 3% exhibit flux variations within an observation , 10% exhibit variations from observation - to - observation .
we also find one source , cxougc j174622.7285218 , with a periodic 1745 s signal ( 1.4% chance probability ) , which is probably a magnetically - accreting cataclysmic variable . we compare the spatial distribution of x - ray sources to a model for the stellar distribution , and find 2.8@xmath4 evidence for excesses in the numbers of x - ray sources in the region of recent star formation encompassed by the arches , quintuplet , and galactic center star clusters .
these excess sources are also seen in the luminosity distribution of the x - ray sources , which is flatter near the arches and quintuplet than elsewhere in the field .
these excess point sources , along with a similar longitudinal asymmetry in the distribution of diffuse iron emission that has been reported by other authors , probably have their origin in the young stars that are prominent at @[email protected]@xmath1 . |
the minimal supersymmetric standard model ( mssm ) is one of the promising extensions of the standard model @xcite .
the model requires superpartners of the standard model particles ( sparticles ) , and the large hadron collider ( lhc ) at cern might confirm the existence of the new particles @xcite .
the lhc is a @xmath5 collider at @xmath6 tev , whose operation is currently expected to start in 2007 .
the integrated luminosity will be 10 fb@xmath7/year at the beginning ( low luminosity runs ) , and then upgraded to 100 fb@xmath7/year ( high luminosity runs ) .
supersymmetry ( susy ) must be broken and the sparticle mass spectrum depends on the susy breaking mechanism .
measurement of the sparticle masses provides a way to probe the origin of the susy breaking in nature .
the sparticle mass measurement at the lhc has been therefore extensively studied . among the sparticles the third generation squarks , stops ( @xmath8 ) and sbottoms ( @xmath9 ) ( @xmath10 ) , get special imprints from physics at the very high energy scale .
one of the examples is found in a model with universal scaler masses
. although the scalar masses are universal at a high energy scale , the third generation squarks are much lighter than the first and second generation squarks due to the yukawa running effect . on the other hand ,
some susy breaking models , such as the flavor u(2 ) model @xcite or the decoupling solution @xcite , and the superconformal model @xcite for the susy flavor problem , have non - universal boundary conditions for the third generation mass parameters at the gut scale .
in addition , the @xmath11-@xmath12-higgs trilinear coupling @xmath13 is comparable to the gluino mass at the weak scale when the origin of the susy breaking comes from the gut scale or the planck scale physics .
this appears in the stop mass matrix with a large coefficient @xmath14 ; @xmath15 .
the stop left - right mixing is therefore expected to be sizable , leading to an even lighter stop mass compared to those of other squarks @xcite .
it should be stressed that the stop masses and the mixing are very important parameters to predict the light higgs mass @xcite , or the rare @xmath16 decay ratios @xcite .
we may be able to access the nature of the stop and sbottom at the lhc provided that they are lighter than the gluino ( @xmath17 ) .
they may copiously arise from the gluino decay if the decay is kinematically allowed .
the relevant decay modes for @xmath9 ( @xmath10 ) , @xmath18 , to charginos @xmath19 ( @xmath20 ) or neutralinos @xmath21 ( @xmath22 ) are listed below ( indices to distinguish a particle and its anti - particle is suppressed unless otherwise stated ) for this paper . ] , @xmath23 in previous literatures @xcite , the lighter sbottom @xmath24 is often studied through the mode ( i)@xmath25 , namely the @xmath26 channel .
this mode is important when the second lightest neutralino @xmath27 has substantial branching ratios into leptons .
the difference of sparticle masses such as @xmath28 is determined by measuring kinematical end points of invariant mass distributions for signal events . in a previous paper @xcite
we proposed to measure the edge position of the @xmath1 distribution for the modes ( iii)@xmath29 and ( iv)@xmath30 , where @xmath1 is the invariant mass of a top - bottom ( @xmath4 ) system .
the decay modes are expected to be dominant in the minimal supergravity model ( msugra ) , since the branching ratios @xmath31 could be as large as 60% .
we focused on the reconstruction of hadronic decays of the top quark , because the @xmath1 distribution of the decay makes a clear `` edge '' in this case . the parton level @xmath1 distributions for the modes ( iii)@xmath32 and ( iv)@xmath33 are expressed as functions of @xmath34 , @xmath35 , @xmath36 , and the chargino mass @xmath37 : @xmath38 , and the edge position ( end point ) of the @xmath1 distribution @xmath39 for the modes ( iii)@xmath32 and ( iv)@xmath33 are written as follows ; @xmath40 the measurement of the end points is sensitive to both @xmath35 and @xmath41 , provided that @xmath42 and @xmath43 are determined from other measurements .
the edge height is then closely related to @xmath44 . in some model parameters
@xmath45 is very close to @xmath46 .
when they are experimentally indistinguishable , it is convenient to define a weighted mean of the end points ; @xmath47 as the final states @xmath48 from the decay chain @xmath49 ( mode ( iii)@xmath50 ) could have an irreducible contribution to the @xmath4 final state , they are included in the definition of @xmath51 .
in the previous paper we showed that the weighted mean @xmath51 was successfully measured from a fit to the @xmath1 distribution .
we also discussed the interpretation of the measurement . in this paper
we significantly extend our study to the branching ratio measurements , identification of the stop and sbottom decays into heavier charginos , and the interpretation in the msugra model .
this paper is organized as follows . in section ii , we explain our reconstruction and fitting procedure of the edge structure in detail .
we use a sideband method to estimate the background distribution due to misreconstructed events , which plays a key role in the edge reconstruction .
monte carlo simulations show that the distribution of the signal modes ( iii ) and ( iv ) after subtracting the sideband background is very close to the parton level distribution . in section iii ,
we explore the msugra parameter space to check if our fitting procedure reproduces the end point @xmath51 and the number of events going through the decay modes ( iii ) and ( iv ) .
we then turn into more delicate issues such as extraction of branching ratios . in section
iv , we discuss the stop and sbottom decays into the heavier chargino @xmath52 by looking into the @xmath1 distribution of events with additional leptons . in section
v , we study the msugra parameter region where the decay modes ( iii ) and/or ( iv ) are open , and discuss how to extract the fundamental parameters using the @xmath1 distributions .
the lhc s potential to extract the top polarization arising from the decay @xmath53 is discussed in section vi . section vii is devoted to conclusions .
in the msugra model the sparticle spectrum is parameterized by the universal scalar mass @xmath54 , the universal gaugino mass @xmath55 , the trilinear coupling of the scalar fields @xmath56 at the gut scale ( @xmath57 ) , the ratio of the vacuum expectation values @xmath58 , and sign of the higgsino mass parameter @xmath59 . in order to demonstrate the end point reconstruction , we take an msugra point with @xmath60 gev , @xmath61 gev , @xmath62 gev , @xmath63 and @xmath64 .
this corresponds to the sample point a1 in table [ masscross ] .
the masses and mixings of sparticles are calculated by isasusy / isasugra 7.51 @xcite .
two different event generators , pythia 6.1 @xcite and herwig 6.4 @xcite , are utilized to generate monte carlo susy events using the masses and mixings . in this section
we show results with pythia .
the events are then passed through a fast detector simulation program for the atlas experiment , atlfast @xcite .
this program performs jet reconstruction in the calorimeters and momentum / energy smearing for leptons and photons , giving a list of reconstructed jets as well as identified leptons and photons .
we generate a total of @xmath65 susy events , which correspond to an integrated luminosity of 120 fb@xmath7 .
in addition to the susy events , @xmath66 events are generated using pythia as the standard model background .
the number of generated events amounts to @xmath67 , corresponding to an integrated luminosity of 286 fb@xmath7 . in the plots in this paper , the @xmath66 background is not included unless otherwise noted . in the detector simulation ,
a lepton is identified if @xmath68 gev and @xmath69 for an electron , and @xmath70 gev and @xmath69 for a muon , respectively .
a lepton is regarded as isolated if it is separated by @xmath71 from other calorimeter clusters and the transverse calorimeter energy in a cone size @xmath72 around the lepton is less than 10 gev .
the cone size is defined as @xmath73 , where @xmath74 and @xmath75 are the azimuthal angle and the pseudo - rapidity , respectively . by default ,
jets are reconstructed by a cone - based algorithm with a cone size @xmath76 in the detector simulation .
values , @xmath77 and @xmath78 are also tried to check the dependence on @xmath79 of the cone - based algorithm . for comparison , another jet reconstruction algorithm , the @xmath80 algorithm ,
was also tried with @xmath81 , and 0.7 .
the comparison is given in the appendix .
] after applying the algorithm , jets having a transverse energy ( @xmath82 ) more than 10 gev are kept as reconstructed jets .
the @xmath83 and @xmath84 tagging efficiencies are set to be 60% and 50% , respectively .
the energy of the reconstructed jet is recalibrated according to its jet flavor using a parameterization optimized to give a proper scale of the dijet mass .
@xmath34&@xmath85&@xmath86 & @xmath87 & @xmath43&@xmath88a1 & 707 & 427 & 570 & 613 & 220 & 26 a2 & 706 & 496 & 587 & 614&211 & 25t1&707&327&570&613 & 220 & 30t2 & 707&477&570&612&211 & 25b & 609 & 402 & 504 & 534 & 179 & 56c & 931 & 636 & 771 & 805 & 304 & 5 g & 886 & 604 & 714 & 763 & 285 & 7 i & 831 & 571 & 648 & 725 & 265 & 10e1 & 515 & 273 & 521 & 634&153 & 77 e2 & 747 & 524 & 770 & 898&232 & 8 first , we apply the following selection cuts to the simulated events for the @xmath4 signal : 1 .
the missing transverse energy @xmath89 is greater than 200 gev , where @xmath89 is calculated from the reconstructed jets , leptons , photons , and unreconstructed calorimeter energies .
this calculation is performed before the recalibration of jet energies .
2 . the effective mass @xmath90 is greater than 1000 gev , where @xmath90 is the sum of the missing transverse energy and transverse momentum of reconstructed jets , namely @xmath91 .
3 . there are two and only two @xmath83-jets with @xmath92 gev in an event .
4 . excluding the two @xmath83-jets and those identified as tau - jets
, the number of remaining reconstructed jets with @xmath92 gev and @xmath93 should be between four and six , inclusive .
distributions of the cut variables are shown in fig .
[ presel ] .
the first two cuts are to enhance the susy events against the standard model background events .
the other two cuts are to reduce combinatorial background ( wrong combinations of jets ) in the reconstruction of the top quark .
to reconstruct the hadronic decay of the top quark , we take the following steps ( i)-(iv ) : * we first take jet pairs consistent with a hadronic @xmath94 boson decay with a cut on the jet pair invariant mass @xmath95 : @xmath96 gev . the @xmath95 distribution is shown in fig . [ mjj](a ) , where the selected mass region is marked as * w * ( the @xmath94 mass region ) .
although fake @xmath94 pairs dominate the distribution , a small bump due to real @xmath94 bosons can be seen in the mass region . * the invariant mass of the jet pair and one of the @xmath83-jets , @xmath97 ,
is then calculated .
all possible combinations of jet pairs and @xmath83-jets are tried in an event to select the combination which minimizes the difference @xmath98 .
the distribution of the selected @xmath97 is shown in fig .
[ mbjj](a ) . * the energy and momentum of the jet pair
are scaled so that @xmath99 , and the invariant mass @xmath97 is recalculated .
the distribution is shown in fig .
[ mbjj](b ) .
the jet combination is regarded as a top quark candidate if @xmath100 gev . as
jets are supplied from gluino or squark decays and there are several jets in a selected event , events with a fake @xmath94 boson ( a jet pair that accidentally has @xmath101 ) still dominate the @xmath97 distribution .
the contribution of the fake @xmath94 boson in the @xmath94 mass region is estimated from the events that contain jet pairs with the invariant mass in the regions * a * : @xmath102 gev and * b * : @xmath103 gev .
we call them `` the @xmath94 sidebands '' . the energy and momentum of the jet pairs
are then scaled linearly to be in the @xmath94 mass region @xmath104 gev .
the @xmath95 distributions before the linear scaling are shown for the @xmath94 region and the @xmath94 sidebands in fig .
[ mjj](b ) .
the distribution of the fake top quark candidates is estimated by following the same steps ( ii)-(iv ) for the scaled jet pair ( see the hatched histogram in fig . [ mbjj](c ) ) .
the `` true '' distribution is obtained by subtracting the background distribution estimated by the @xmath94 sidebands from the original distribution ( see fig . [
mbjj](d ) ) .
the estimation is based on an assumption that the background comes from the jets without significant correlation with the @xmath83-jets .
the @xmath66 production is the dominant standard model background .
the @xmath97 distributions of the @xmath66 events surviving at this stage are shown in fig .
[ mbjj](c ) and ( d ) before and after the sideband background subtraction , respectively .
the integrated luminosity of the @xmath66 events is normalized to that of the susy events . to reduce the @xmath66 events ,
one of the following two lepton cuts may be used , depending on the signal / background situation . *
loose lepton cut : if there are high-@xmath105 isolated leptons , the invariant mass of any high-@xmath105 lepton and the remaining @xmath83-jet ( @xmath106 ) should be greater than 150 gev .
the distribution of the minimum @xmath106 in events with a top candidate is shown in fig . [ mbl ] . * hard lepton cut : an event should have no isolated leptons .
the cuts reduce the fraction of the events having a @xmath66 pair , where one of the top quarks decays leptonically .
the @xmath97 distributions with the lepton cuts are shown in fig .
[ mbjj1 ] .
the numbers of selected events are listed in table [ lepton ] .
hereafter we use the loose lepton cut .
.numbers of events having a top candidate with or without the lepton cuts .
the integrated luminosity is 120 fb@xmath7 .
the numbers in parentheses are the ratios of the @xmath66 events to the susy events . [ cols= " < , > , > , > , > " , ] for the remaining events , the top candidate is then combined with the other @xmath83-jet , which is not used to reconstruct the top candidate , to calculate the invariant mass of the @xmath4 system @xmath1 . the distribution is shown in fig . [ mtb](a ) .
however , the expected @xmath4 end point is not clearly visible in the @xmath1 distribution due to the fake @xmath94 events .
here we can again utilize the @xmath94 sidebands for the background estimation . the estimated background distribution is shown in fig .
[ mtb](b ) , which is obtained by averaging distributions from the sidebands * a * and * b*. the estimated background distribution is subtracted from the signal distribution in fig . [ mtb](c ) .
the _ corrected _ signal distribution ( c ) shows a better end point structure compared to ( a ) . by using the information from the event generator
, one can show the sideband method is indeed reproduces the @xmath1 distribution for the signal modes ( iii ) and ( iv ) .
[ mtb](d ) is the same distribution as fig .
[ mtb](c ) but for the events which contain either the decay chain ( iii)@xmath107 , the decay chain ( iv)@xmath30 , or the decay chains ( iii)@xmath50 , which are irreducible to the chain ( iii)@xmath107 ; @xmath108 . note that if @xmath109 has an invariant mass consistent to a top quark , the decay is kinematically equivalent to the mode ( iii)@xmath107 . fig .
[ mtb](d ) shows two end points as expected ( @xmath110 gev and @xmath111 gev ) , demonstrating that the sideband method works well .
note that the signals from the modes ( iii ) and ( iv ) in eq .
( [ gluinodecay ] ) are significant in the total selected events .
the total distribution shown in fig .
[ mtb](c ) contains contribution from mis - reconstructed events such as other gluino decay chains with @xmath94 and @xmath83-jets , or stop / sbottom pair productions .
we fit the total distribution shown in fig .
[ mtb](c ) , which is made with a bin size @xmath112 10 gev , by a simple fitting function described with the end point @xmath113 , the edge height @xmath114 per @xmath115 bin , and the smearing parameter @xmath116 originated from the jet energy resolution ; @xmath117 ^2 } \right ) dm .
\cr&&\end{aligned}\ ] ] to reduce the number of free parameters for good convergence capability of the fit , we set the smearing parameter @xmath116 to be 10 % of the end point @xmath113 . this assumption is based on the dijet mass resolution of the atlas detector @xcite .
we assume that the signal distribution is sitting on a linearly - decreasing background expressed by a function @xmath118 .
the parameters @xmath119 and @xmath83 are also determined by the fit .
the fit results , especially the edge height @xmath114 , depend on the mass range used for the fit .
we therefore apply an iterative fitting procedure to obtain stable results .
the initial @xmath120 value is determined by fitting a rather wide mass range @xmath121 gev@xmath122800 gev ( 45 histogram bins ) . in the next step ,
lower 8 bins and higher 25 bins with respect to the @xmath120 value are used to obtain new fit results including the new @xmath120 value .
the relatively small number of lower bins is chosen because the edge height is sensitive to the distribution in the mass region near the end point . including too many lower bins
is found to degrade the sensitivity .
the higher bins determines the linearly - decreasing background .
this step is repeated until the fit results become stable .
the number of iterations is typically three or four .
if the edge height @xmath114 has a large fit error , the number of lower bins is increased to 9 or 10 , until the error becomes reasonably small ( typically less than 15 % ) .
the fit result is shown in fig .
[ tblfit ] .
we obtain @xmath123 gev and @xmath124 /(10
gev ) , where the errors are statistical .
in the previous section we discuss the selection of the @xmath4 events and the extraction of the edge structure in the @xmath1 distribution originated from the gluino decays ( iii)@xmath107 , ( iii)@xmath50 or ( iv)@xmath50 .
two values are obtained from the @xmath1 distribution ; the end point @xmath125 is directly related to the stop , sbottom , and gluino masses , while the edge height @xmath114 is sensitive to the gluino production cross section and the decay branching ratios .
this section is aimed to compare the reconstructed values with expectations .
we will see that the kinematical quantities @xmath126 and the peak value of the effective mass distributions ( @xmath127 ) agree very well to @xmath51 and @xmath128 , respectively . on the other hand , the reconstructed edge height @xmath114 , which is related to the total number of gluinos decaying through the modes ( iii ) or ( iv ) , is dependent on the event generator , and also on the decay patterns of the neutralino , chargino and squarks .
we clarify key issues to reduce the uncertainties in this section .
the gluino decay widths and the decay kinematics depend on the squark masses and mixings .
the charginos and neutralinos in the decay chains ( i)-(iv ) further decay into jets / leptons , and the decay patterns depend on the electroweak susy parameters such as @xmath129 and the slepton masses . the @xmath1 distribution is therefore a function of all susy parameters . to check the validity of our reconstruction and subtraction scheme
, we study the @xmath1 distributions for various sets of susy parameters .
the reference point a1 introduced in the previous section corresponds to the msugra model with a large negative @xmath56 value .
the stop and sbottom masses can be changed by varying the @xmath56 value without changing other sparticle masses drastically ( see section v for the detail ) .
we thus make another msugra point a2 , by changing only the @xmath56 value from @xmath130 gev to @xmath131 gev from the point a1 . in addition , two non - msugra points t1 and t2 are made from the point a1 , where the chargino / neutralino sector and the gluino mass are kept unchanged but the stop masses and mixing are modified by changing the mass parameter @xmath132 .
the stop mass parameters and the mixing angle at the low energy scale are listed in table [ stopsector ] .
@xmath132 & @xmath133 & @xmath13&@xmath134a1 & 427 & 482 & 573 & @xmath135 & 0.99 a2 & 496 & 521 & 591 & @xmath136 & 1.01 t1 & 327 & 366 & 573 & @xmath135 & 1.14 t2 & 477 & 551 & 573 & @xmath135 & 0.84
next we choose msugra points b , c , g and i , which were discussed in the paper @xcite .
they satisfy @xmath137 so that the decay mode ( iii)@xmath107 is open .
the points also satisfy the condition @xmath138 tev , which is required for a statistical reason .
the gluino production is dominated by the process @xmath139 in the case of @xmath140 , and the production cross section is parameterized as @xcite @xmath141 for example the cross section is 14 pb and 3.1 pb for @xmath142 gev and 900 gev , respectively .
assuming that the branching ratios and the reconstruction efficiency are equal to those of the point a1 , the gluino mass must be less than 1 tev to reconstruct more than 2000 susy @xmath4 events after the sideband subtraction for an integrated luminosity of 100 fb@xmath7 .
finally we select two more points e1 and e2 , where only the decay chain ( iii)@xmath107 is kinematically open and the other two body gluino decays into squarks are closed .
this happens when @xmath54 and @xmath56 are large .
as the susy events at these points typically contain four bottom quarks , the combinatorial background for the @xmath4 reconstruction is significantly large .
@xmath143 & @xmath56 & @xmath58 a1&300 & 100 & 100 & @xmath130 & 10 a2&300 & 100 & 100 & 300 & 10b&255 & 102 & 102 & 0 & 10c&408 & 92 & 92 & 0 & 10 g&383 & 125 & 125 & 0 & 20i&358 & 188 & 188 & 0 & 35 e1&200 & 500 & 200 & @xmath144 & 10 e1&300 & 700 & 500 & @xmath144 & 10 for the points we study in this paper , we list the relevant sparticle masses and the universal parameters at the gut scale in table [ masscross ] and table [ gut ] , respectively . at each susy point
we generate two monte carlo data samples , each having @xmath145 events , where one is generated by pythia and the other by herwig .
the cross sections are summarized in table [ masscross ] .
the susy cross section for the point c is the smallest among them , where the generated @xmath146 events correspond to an integrated luminosity of 600 fb@xmath7 .
the fit results of the @xmath1 distributions are listed in table [ fittable ] , where we follow the fitting procedure described in the previous section .
the weighted end point @xmath147 is defined in eq .
( [ mtbw ] ) and the relevant branching ratios are listed in table [ branch ] . in table
[ fittable ] , @xmath148 is the number of @xmath4 events after the sideband subtraction , while @xmath149 is the number of @xmath4 events after the subtraction , with one and only one gluino decaying through the mode ( iii)@xmath107 , ( iii)@xmath50 or ( iv)@xmath50 . generator information is used to obtain @xmath149 . the relation between @xmath113 and @xmath150 is shown in fig .
[ endpoint ] .
the fitted value @xmath125 increases linearly with the weighted end point @xmath147 in the mass range @xmath121 gev@xmath151 gev , and tends to be lower than @xmath147 . in the previous literature @xcite
, the lower mass value was understood as the effect of particles missed outside the jet cones . in the literature
invariant mass distributions of the events with same flavor and opposite sign leptons are studied , which comes from the cascade decay of the squarks @xmath152 .
the end point of the @xmath153 distribution is lower than the parton level @xmath154 end point by about 10% . in the study of the squark cascade decay @xmath155
, the end point agrees better with the parton level one by changing the jet cone size to @xmath156 . in our study at the reference point a1 , the reconstructed number of events is significantly reduced for @xmath156 .
we find that using the jet cone size @xmath157 leads to a better reconstruction and a larger @xmath113 , which is closer to @xmath51
. such comparison might be useful to estimate the true end point .
we discuss the dependence on the jet finding algorithm and the cone parameters in appendix .
we note that our definition of @xmath147 might be too simple if the reconstruction efficiencies of the decay modes ( iii ) and ( iv ) are very different .
in addition , the weighted average should not be applied when @xmath45 and @xmath46 differ by more than 80 gev , as we use the events with @xmath158 gev for the fit . & @xmath125[gev ] & @xmath114/(10gev ) & @xmath149 & @xmath148 a1 & py & 459 & 455.2 @xmath159 8.3 & [email protected] & 5846 & 10487 & hw & & 434.5 @xmath159 5.8 & [email protected] & 6685 & 11470a2 & py & 409 & 442.0 @xmath15917.5 & [email protected] & 3064 & 7525 & hw & & 394.4 @xmath1599.5 & [email protected] & 3095 & 7805 t1 & py & 468 & 460.3 @xmath159 5.4 & 327.0 @xmath159 21.6 & 6620 & 11659 & hw & & 452.0 @xmath159 3.9&447.5 @xmath159 23.5 & 8170 & 14050 t2 & py & 429 & 434.5 @xmath159 8.1&223.2 @xmath159 21.6 & 4461 & 8466 & hw & & 416.6 @xmath159 5.2 & 321.0 @xmath159 23.2 & 5378 & 9592 b & py & 371 & 385.6 @xmath159 6.3&226.9 @xmath159 19.6 & 2801 & 5396 & hw & & 361.7 @xmath159 7.3&223.5 @xmath159 21.3 & 3105 & 5935 c & py & 557 & 548.2 @xmath15914.1&142.4 @xmath159 17.5 & 4026 & 11228 & hw & & 556.3 @xmath1599.4 & 178.2 @xmath159 17.9 & 4395 & 12704 g & py & 533 & 498.5 @xmath159 8.6&244.1 @xmath159 23.8 & 5784 & 13630 & hw & & 506.9 @xmath159 6.4&325.5 @xmath159 22.9 & 6248 & 15039 i & py & 507 & 497.8 @xmath159 7.3&289.7 @xmath159 24.0 & 6016 & 13752 & hw & & 492.9 @xmath159 5.2&383.9 @xmath159 24.5 & 6661 & 14968 e1 & py & 360 & 345.5 @xmath159 5.3&270.6 @xmath159 23.7 & 2778 & 4595 & hw & & 348.0 @xmath159 6.3&251.5 @xmath159 23.9 & 3169 & 5167e2 & py & 453 & 430.8 @xmath159 7.5&352.0 @xmath159 33.9 & 5577 & 16490 & hw & & 444.7 @xmath159 8.0&324.6 @xmath15930.9 & 4394 & 15724 ( iv)@xmath160 & ( iii)@xmath30 & ( iii)@xmath160 & sum & @xmath161 a1 & 11.0 & 6.7 & 1.4 & 3.4 & 2.7 & 25.3 & 43.4 a2 & 3.1 & 6.5 & 1.6 & 1.4 & 0.4 & 13.1 & 32.0 t1 & 24.5 & 3.2 & 0.8 & 5.0 & 3.0 & 36.5 & 56.3t2 & 4.3 & 9.9 & 2.2 & 0.5 & 2.1 & 19.0 & 36.2b & 4.1 & 8.2 & 2.3 & 0.9 & 1.7 & 17.3 & 33.5c & 7.2 & 5.3 & 1.3 & 0.9 & 0.8 & 15.4 & 38.5 g & 6.6 & 7.5 & 1.2 & 0.5 & 0.8 & 16.6 & 40.4 i & 6.2 & 11.1 & 0.7 & 0.0 & 0.7 & 18.7 & 47.3e1 & 78.5 & 0 & 0 & 0 & 0 & 78.5 & 99e2 & 42.6 & 0 & 0 & 0 & 0 & 42.6 & 98
we now discuss the relation between @xmath127 and @xmath162 using the @xmath4 samples .
we find that sum of the masses @xmath163 has a linear dependence on @xmath127 as shown in fig .
[ ameff ] .
the deviation of the sample points from a linear fit is less than 5% .
the plot is for herwig samples and we find that pythia and herwig give consistent results for the relation between @xmath127 and the masses .
the @xmath4 sample contains two and only two @xmath83-jets originating from a gluino decay , therefore the @xmath42 dependence of @xmath164 is expected .
one may wonder that the @xmath165 production might affect the @xmath127 value .
this is , however , not the case for the points we study .
indeed , the points e1 and e2 , where @xmath166 , also satisfy the linear relation between @xmath127 and @xmath167 .
note that the quark parton distribution is harder than the gluon parton distribution .
therefore @xmath168 in a wide mssm parameter region .
this is why @xmath127 becomes a very good function of @xmath163 . in ref .
@xcite the relation between @xmath127 and the effective susy scale is studied , where the effective susy scale is defined as the cross section weighted mean of the masses of two sparticles initially produced in @xmath5 collisions . on the other hand
, we actively select the @xmath169 productions by requiring two tagged @xmath83-jets .
this leads to a clear dependence of @xmath164 and @xmath167 .
we now discuss the relation between the edge height @xmath114 , the number of reconstructed @xmath4 events , and the reconstruction efficiencies .
the total number of the `` edge '' events @xmath149 arising from the decay chains ( iii ) and ( iv ) may be estimated from @xmath113 , @xmath114 per the bin size @xmath170 as follows , @xmath171 this formula is obtained by assuming the parton level distribution , and equating the minimum of the @xmath1 distribution from the decay chain ( iii ) or ( iv ) to @xmath14 .
this is a good approximation for reasonable susy parameters . in fig .
[ reconstruct](a ) and ( b ) we compare @xmath149 in table [ fittable ] with @xmath172 .
we find a very good agreement between them both for the pythia and herwig samples . on the other hand ,
the correlation between @xmath148 and @xmath172 is much worse as shown in fig .
[ reconstruct](c ) .
the number @xmath148 receives contribution from other gluino cascade decay chains such as the modes ( i ) and ( ii ) , as well as contributions from the stop and the sbottom pair productions . by the end point fit , we extract number of events coming from only the modes ( iii)@xmath29 , ( iii)@xmath50 and ( iv)@xmath50 .
@xmath149 , and then @xmath172 , must be related to the number of gluino decays through the cascade decay chains ( iii ) and ( iv ) via the reconstruction efficiencies . for the points we study , the number of produced @xmath165 events , @xmath173 , is typically 10% to 14% of the total susy production events , while the number of @xmath174 and @xmath175 production events @xmath176 ranges from 42% to 51% .
the gluino decay branchings ratios are listed in table [ branch ] .
the number of events @xmath177 , where one gluino decays through the modes ( iii ) or ( iv ) and the other squark or gluino decay does not produce any bottom quark , is given as follows ; @xmath178 where @xmath179 is the branching ratio of the gluino decaying into stop or sbottom , thus having two bottom quarks in the final state .
the reconstruction efficiency of the @xmath4 edge mode , @xmath180 , is given as @xmath181 if the efficiency does not strongly depend on uncertainty in hadronization and the model parameter dependence can be corrected from other measurements , we can extract the number @xmath177 from the experimental data .
the major uncertainty in hadronization may be estimated by the generator dependence of the reconstruction efficiency @xmath182 .
the edge height @xmath114 for the herwig sample is significantly larger than that of the pythia sample in table [ fittable ] , except the points e1 and e2 , and the difference is more than 20% at many points .
we note the difference is small before we apply the sideband subtraction .
for example , the numbers of @xmath4 events before and after the sideband subtraction are 9695 ( 10180 ) and 2462 ( 2949 ) for the pythia ( herwig ) sample , respectively , at the point a1 .
the number of events before the sideband subtraction only differs by 5% between herwig and pythia , while the difference increases to be more than 20% after the subtraction .
this indicates that the resolution of the @xmath94 boson mass is somewhat better for the herwig sample .
indeed , the major difference of the two generators are in the fragmentation scheme .
pythia is based on the string model , while herwig is based on the qcd model .
currently we are only able to say that understanding the nature of fragmentation is essential for the interpretation of @xmath172 .
we compare the herwig and pythia samples in appendix .
let us discuss other effects that might change the reconstruction efficiency . in fig .
[ acceptance ] we plot the reconstruction efficiency @xmath182 as a function of @xmath42 by normalizing the efficiency to that for the point a1 .
here @xmath177 is calculated using eq .
( [ nprod ] ) , the generator information on @xmath173 , @xmath183 , and the monte carlo inputs of the gluino branching ratios .
the length of the bar in the plot shows twice of the statistical error of @xmath149 , @xmath184 .
we do nt discuss the points e1 and e2 here because events with two bottom quarks are not the dominant signature of squark and gluino production at these points .
the efficiencies are very close to one another among the points a1 , a2 , t1 , and t1 .
although the stop mass is very different among the points , the mass has little influence on the efficiency .
note that these points have the same gluino masses , and almost the same parameters for the chargino and neutralino sector .
for the other points , the efficiency can also be expressed as a linear function of @xmath42 except the sample point c. the gluino mass dependence arises from the pre - selection cut @xmath185 gev . the reconstruction efficiency at the point c ( a circle in fig .
[ acceptance ] ) is lower by 20% from the linear fit ( the dashed line in fig .
[ acceptance ] ) .
this is due to the loose lepton cut which reduces the @xmath66 background . as the chargino decay branching ratio into leptons is high ( about 47% ) at the point c
, the lepton cut kills a significant fraction of @xmath4 events .
the effect is estimated by comparing @xmath149 without any lepton cut and @xmath149 with the loose lepton cut , at the points c and a1 .
the efficiency at the point c without any lepton cut is plotted in fig .
[ acceptance ] , and is close to the linear fit of the other points .
the gluino mass dependence of the reconstruction efficiency would be easily corrected by estimating the gluino mass from the effective mass distribution .
the uncertainty from the chargino decay branching ratio may be corrected by studying @xmath4 final states with leptons as well .
therefore the parameter dependence of the efficiency may be corrected by analyzing the monte carlo and real data .
efforts to determine the mssm parameters are of course very important to this purpose . in the previous subsection , we find that the reconstruction efficiencies significantly differ between the pythia and herwig samples . however , as we can see in table [ fittable ] , the generator dependence mostly cancels in the ratio @xmath186 .
therefore the ratio may play a role to determine the fundamental parameters .
for example , when the contributions from the stop or sbottom pair productions are negligible , @xmath186 is determined by the branching ratios to the third generation squarks . however , there are several aspects one must consider .
@xmath148 consists of contributions from various decay modes , even if the contributions from the stop or sbottom pair productions are negligible : * the @xmath66 final state from the decay chains ( ii ) contributes to @xmath148 .
the branching ratio is typically 1/3 of that of the @xmath4 final state for @xmath187 .
the @xmath66 final state is reconstructed as @xmath4 with a high probability , although the @xmath1 distribution does not have an edge structure .
* we expect that events having no @xmath188 decay would be eliminated by the sideband subtraction , and should not contribute to @xmath148
. however , some events actually remain , because of @xmath94 and @xmath189 bosons from the decay of charginos and neutralinos , as well as accidental @xmath94 and @xmath189 bosons from other cascade decay chains .
an example where @xmath27 decays dominantly into the @xmath189 boson will be discussed in the next subsection .
+ in addition , the sideband subtraction is not necessarily perfect , as will be discussed in appendix .
for example , the efficiency of the mode ( i)@xmath25 is roughly a half of that of the mode ( iii)@xmath107 after the sideband subtraction at the point a1 ( see appendix ) .
* the cuts to reduce background events may induce decay mode dependence of the efficiency .
for example the lepton cut could efficiently reduce the edge mode ( iii ) if @xmath190 is large as we have seen in fig .
[ acceptance ] for point c. the interpretation of the ratio @xmath186 becomes more complicated for the following two cases : * if @xmath191 , squarks mostly decay into the gluino . as discussed previously ,
the squark / gluino production and decay typically produce events with four @xmath83-jets .
they are identified as a two @xmath83-jet event when two @xmath83-jets are tagged and the other two are mis - tagged .
these events decrease the fraction from the decay chain ( iii ) and ( iv ) in the two @xmath83-jet events . assuming a tagging efficiency of 60% for a single @xmath83-jet ,
35% of the four bottom quark events are identified as two @xmath83-jet events , and only 1/3 of them contain a @xmath83-jet pair originated from a single gluino decay .
the wrong @xmath83-jet pair combinations significantly dilute the edge mode .
study of events with three @xmath83-jets and four @xmath83-jets is necessary for this case .
* the @xmath4 signal from the stop pair production could be significant if the stop is much lighter than the gluino .
for example , only 4.7% of the susy events comes from the stop pair production at the point a1 , while it amounts to be about 17% at the point t1 . in the msugra model , both the right - handed stop and
the left - handed sbottom is lighter than gluino in a broad parameter space . as a result , the decay modes which involve @xmath94 bosons ( modes ( ii ) , ( iii ) and ( iv ) )
dominate over the gluino decays to @xmath161 in the region . because the events with @xmath94 bosons remain after the @xmath94 sideband subtraction , the reconstruction efficiency @xmath182 is expected to be similar for those decay modes .
thus , if the contributions from the stop or sbottom pair productions are negligible , the numbers of events with two bottom quarks are given approximately as @xmath192 , \cr { n_{\rm all}}&\sim & { \epsilon_{tb}}{{\rm br}}(\tilde{g}\rightarrow bbx)\times \cr & & \hskip -20pt \left [ 2n(\tilde{g}\tilde{g } ) \left(1-{{\rm br}}(\tilde{g}\rightarrow bbx)\right ) + n(\tilde{g}\tilde{q})+ n(\tilde{g}\tilde{q}^ * ) \right].\cr&&\end{aligned}\ ] ] if this simple formula holds , the ratio @xmath186 should provide the ratio of the branching ratios @xmath193 .
this is illustrated in fig .
[ ratio ] , where we plot the ratio @xmath194 as a function of @xmath195 .
here we plot @xmath194 instead of @xmath186 , because we have already seen @xmath196 , and the statistical fluctuation of @xmath149 is small ( @xmath197 2% ) with a help of the generator information .
some points in the plots are away from the line @xmath198 : the point c is off because the chargino has large branching ratios into leptons as discussed earlier . at the point t1
, the @xmath199 productions contributes to @xmath148 .
the points e1 and e2 are significantly off because the first and the second generation squarks dominantly decay into the gluino .
the systematics due to the chargino and neutralino decay patterns are difficult to estimate from the lhc data only . by combining the data from a future @xmath200 linear collider ( lc ) of @xmath201 gev such systematics would be significantly reduced .
note that susy points with @xmath202 gev are studied in this paper , which is within the reach of a tev scale lc .
if events with four @xmath83-jets dominate the susy events at the lhc , study of the events with four or three @xmath83-jets would be important , which is out of the scope of this paper .
if @xmath203 , it is important to constrain the stop mass in order to estimate the contribution from @xmath204 , as the cross section increases rapidly as @xmath205 decreases . the weighted end point @xmath51 would provide the information on @xmath205 . in this subsection
we check if we would reconstruct fake end points by any chance when the mode of our interest is insignificant . to this purpose
, we generate a set of monte carlo data for the snowmass points @xcite .
the snowmass points and slopes ( sps ) are a set of benchmark points and parameter lines in the mssm parameter space corresponding to different scenarios in the search for susy at the present and future experiments .
some of those points correspond to the case where decay modes ( iii ) or ( iv ) are not dominant , and we should not expect to see the edge structure in the @xmath1 distribution .
we list the fit results at sps 1 - 2 , and 4 - 6 in table [ snowmass ] . here
sps 3 is not listed as it is the same as the point c. the @xmath120 is lower by 20 gev than the @xmath51 at sps 1 and sps 6 , similar to the points discussed in subsection [ kinematical ] ( see table [ fittable ] ) .
sps 2 is so - called the `` focus point '' , where the scalar mass at the gut scale @xmath54 is larger than the gut scale gaugino mass @xmath206 .
the gluino two body decays into a squark and a quark are not kinematically open , therefore the dominant decay modes of the gluino are the three body decays ; @xmath207 , @xmath208 , and @xmath209 . the @xmath1 distribution should not have any edge structure because it dominantly consists of the gluino three body decays .
the reconstructed edge has only @xmath210 in height , and is statistically insignificant .
the @xmath211 value of the fit is also rather large , @xmath21244.7 for 30 degrees of freedom .
is sometimes large even when the decay mode ( iii ) dominates the distribution .
for example @xmath213 for the point a1 . ]
the fitted end point @xmath214 gev may be related to the mass difference between the gluino and the heavier chargino , @xmath215 gev and @xmath216 gev . the @xmath1 distribution at sps 2 is compared with that at sps 1 in fig . [ sps12 ] .
at sps 5 , the stop is so light that the decay @xmath217 is kinematically closed .
the gluino decay branching ratio into @xmath24 is large ( 9% ) , but the @xmath24 subsequently decays mostly into @xmath218(80% ) .
we thus do not find any significant edge .
sps 4 is an example that a straightforward application of our reconstruction technique fails to find @xmath51 .
the distribution is shown in fig . [ sps4 ] .
the sbottom is light because of large @xmath58 , and the dominant gluino decay mode is @xmath219 ( 78% ) .
the sbottom further decays into chargino ( 36% ) or the second lightest neutralino ( 30% ) .
decays of the chargino and the second lightest neutralino are dominated by @xmath220 ( 100% ) and @xmath221 ( 98% ) , respectively .
the dominance of the decay into @xmath189 or @xmath94 is common when the decay into the higgs boson or sleptons are kinematically forbidden .
some of the decays @xmath222 are reconstructed as @xmath4 events together with additional gauge bosons , resulting in the @xmath1 distribution having a quasi - edge structure .
the edge is reconstructed by our fit ( see table [ snowmass ] ) .
note that the end point of the @xmath3 distribution of the decay is 403 gev , while the reconstructed @xmath223 distribution of the events with the decay ( i)@xmath224 has a peak around 500 gev and the distribution ends around 580 gev , close to the kinematical limit @xmath225 gev .
the edge structure is identified as that of the mode ( i)@xmath224 with a tagged @xmath226 . in the @xmath1 distribution , the second edge structure from the decay mode ( iii ) or ( iv )
is weakly seen in the distribution . assuming that two edges exist
, we find the second end point @xmath227 gev and @xmath228/(10 gev ) .
@xmath125[gev ] & @xmath114/(10 gev ) sps 1 & 380.8 & 363.9 @xmath159 4.8 & 267.3 @xmath159 20.8 sps 2 & n / a & 484.9 @xmath159 24.9 & 92.2 @xmath159 23.5 sps 4 & 431.0 & 495.0 @xmath159 6.2 & 515.7 @xmath159 33.0 sps 5 & 416.2 & 442.6 @xmath159 42.8 & 56.6 @xmath159 69.6 sps 6 & 442.3 & 416.2 @xmath159 6.2 & 326.4 @xmath159 24.7
at the points given in table [ masscross ] , the lighter chargino and neutralinos ( inos ) @xmath229 , @xmath230 and @xmath231 are gaugino - like , while the heavier inos @xmath232 @xmath52 , @xmath233 and @xmath234 are higgsino - like .
unlike the first and second generation squarks , the third generation squarks can decay into the heavier inos .
this is because the third generation squarks couple to the higgsinos by the top ( or bottom ) yukawa coupling .
the identification of the cascade decay chain @xmath235 @xmath236 is a probe for the supersymmetric version of yukawa interactions .
in addition , the decay distribution may be sensitive to the mass difference @xmath237 .
in this section we show an example to select such events by requiring additional leptons in the final state .
we study the point c , where the stop and sbottom decay into @xmath236 with large branching fractions . at this point
the gluino decay branching ratios are br@xmath238% and br@xmath239% .
the squarks further decay into the higgsino - like inos with branching ratios of @xmath240% and @xmath241% , respectively .
the stop and sbottom decay branching ratios are comparable to those into @xmath242 . some of the decay branching
ratios of the charginos and neutralinos are listed in table [ pointcbr ] .
the higgsino - like inos have large branching ratios into the @xmath189 boson .
the heavier inos also have non - negligible branching ratios into triple leptons , because they may also decay into @xmath27 which further decays into leptons .
the signature of the heavier inos from stop or sbottom is an excess of the @xmath189 boson or three leptons in events with two @xmath83-jets . in fig .
[ ll ] we plot the invariant mass of the same flavor opposite sign leptons ( @xmath243 ) for the events with @xmath244 and 2 , where @xmath245 is the number of @xmath83-jets in an event . in the plots , accidental lepton pair distribution estimated with events with different - flavor opposite sign leptons are subtracted .
all plots show a common structure that the invariant mass distribution increases toward the edges around 70 gev and 105 gev , which correspond to the decay chains of @xmath246 and @xmath247 , respectively .
the @xmath189 peak is also seen in the plots .
as @xmath245 increases , the @xmath189 peak height becomes more significant relative to the edge structure .
the excess of @xmath248 in the @xmath249 sample indicates the @xmath189 boson originating primarily from stop or sbottom qualitatively , the edges in the @xmath243 distribution for the @xmath249 sample are completely subtracted by the distribution for the @xmath250 sample scaled by a factor of 0.125 , and the @xmath189 peak remains ( the bottom - right plot of fig .
[ ll ] ) .
% @xmath251 & 29 & @xmath252 & 59@xmath253 & 26 & @xmath254 & 1.7@xmath255 & 3.0 & @xmath256 & 25@xmath257 & 6.8 & & @xmath258 & 59 & @xmath259 & 9.8 @xmath260 & 1.7 & @xmath261 & 3.0@xmath262 & 1.3 & @xmath263 & 11@xmath264 & 3.0 & @xmath265 & 37@xmath266 & 0.9 & @xmath267 & 8.9 the same analysis can be performed by using events with three leptons . in fig .
[ lll ] we plot sum of invariant mass distributions ; @xmath268 for all possible combinations of opposite sign lepton pairs . in this mode
the @xmath189 peak is more significant .
the ratio of the edge heights in the @xmath243 distribution for the @xmath249 sample to that of the @xmath250 sample is 0.25 .
this means that the events originating from the decay @xmath269 dominate the three lepton events , and the events with @xmath250 arise due to misidentified @xmath83-jets .
note that these plots at the point c are made for a total of @xmath145 susy events , which corresponds to an integrated luminosity of 600 fb@xmath7 .
the statistics for 100 fb@xmath7 is not significant
. however , this procedure should be useful for lighter mass spectrum where the decay of the third generation squarks to @xmath236 is open .
after establishing the contribution of the decay @xmath270 in the tagged @xmath271 and @xmath272 events , we may reconstruct the @xmath4 system in the sample .
the @xmath4 reconstruction is similar to that described in section ii , except that the distributions are made without the cuts for the effective mass and the lepton , because the standard model background is expected to be small now . in fig .
[ tb3l ] the @xmath1 distribution with tagged three leptons is shown . the number of events after the sideband subtraction is 242.5 .
the @xmath1 distribution is concentrated just below the expected end point which is close to @xmath273 gev , although the statistics is low .
the distribution of tagged @xmath248 is similar and the number of events after the sideband subtraction is 142 .
in this section we interpret our study of the third generation squarks at the lhc in a framework of the msugra model . the lhc may determine the msugra parameter @xmath54 , @xmath206 , @xmath56 , and @xmath58 through the measurement of various distributions .
the decay chains @xmath274 followed by @xmath275 @xmath276,@xmath277 , @xmath278 , or @xmath279 are especially useful if the kinematical end points of the distributions are measured .
the effective mass measurement of the inclusive susy signal is also useful to determine the absolute susy scale . among these parameters ,
the @xmath56 parameter is only weakly constrained .
this is because the left - right mixings of sbottom and stau are almost fixed by @xmath280 , and @xmath13 is insensitive to @xmath56 as will be described below .
one of the possible ways to determine @xmath56 is to measure the mass difference between squarks or sleptons with different flavors , which is caused by the renormalization group equation ( rge ) running between the gut scale and the weak scale .
the squark - mass matrix is given as follows , @xmath281 where @xmath282 for stops , and @xmath283 for sbottoms .
in the msugra , the susy - breaking parameters at the weak scale in these mass matrices are evaluated by the rges with the universal gut scale boundary conditions . when the bottom - yukawa coupling is negligible , it is convenient to present the susy - breaking parameters by the infrared - fixed point value for the top - yukawa coupling constant @xmath284@xcite , @xmath285 here , @xmath286 where @xmath287 and @xmath59 is the renormalization scale .
@xmath288 is the gauge coupling constant at the gut scale .
the susy breaking parameters for the third generation at the low energy scale are approximately given using the value as @xmath289 up to @xmath290 , where @xmath291 here we use the gut relation for the gaugino masses , @xmath292 .
for the pole top mass @xmath293 gev , @xmath294 ranges @xmath295 . from eq .
( [ rg ] ) , the qualitative behavior of the susy - breaking parameters is clear .
the reduction of the @xmath296 and @xmath297 is more efficient for @xmath298 , although the dependence is suppressed by an overall factor of @xmath299 .
the right handed stop mass @xmath132 is also insensitive to @xmath54 . in a moderate @xmath58 region ,
@xmath18 and @xmath24 is almost right - handed and left - handed , respectively . in fig .
[ msugrafig1 ] we show the masses of @xmath17 , @xmath18 , and @xmath24 as functions of @xmath56 , where the other msugra parameters are fixed to be @xmath300 gev , @xmath301 gev , @xmath63 , and @xmath64 .
as expected from the discussion above , the masses @xmath35 and @xmath41 have sensitivity to @xmath56 .
the reduction of @xmath296 and @xmath297 by the top - yukawa coupling constant is larger when @xmath302 .
the stop @xmath303 , which is almost @xmath12 , is lighter than the gluino unless @xmath54 is very large .
similarly , since the sbottom @xmath304 is almost @xmath305 , the branching ratio to the edge events in eq .
( [ gluinodecay ] ) , which includes @xmath242 , is large since it has su(2 ) gauge and top - yukawa interactions .
the study of the edge events has therefore potentially high feasibility in the msugra .
[ msugrafig2 ] shows the regions , where the gluino decays into @xmath306 and @xmath307 are open , in a plane of @xmath54 and @xmath56 . in this plot
the other parameters are fixed to be @xmath308 gev , @xmath3095 or 30 , and @xmath64 . since the top quark mass deviates from the fixed - point value , the @xmath56 dependence on @xmath296 and @xmath297 survives , as shown in fig .
[ msugrafig1 ] .
this may lead to a measurable dependence on @xmath56 of the end point of the @xmath1 distribution . in fig .
[ msugrafig3](a ) , we show @xmath150 , @xmath310 , and @xmath311 as functions of @xmath56 . since @xmath304 is heavy , @xmath311 is lower than @xmath312 .
however , br@xmath313 is larger than br@xmath314 due to the phase space , which makes @xmath150 closer to @xmath310 . if @xmath150 is determined with a precision of a few @xmath315 as in our simulation study , the @xmath56 parameter can be evaluated with a precision of @xmath316 gev .
in addition to the weighted end point @xmath150 , the edge height @xmath114 of the @xmath1 distribution is measurable , and the height is roughly proportional to the branching ratio of the gluino to the edge events arising from the decay modes ( iii ) and ( iv ) in eq .
( [ gluinodecay ] ) . in subsection [ secbr ]
we proposed to investigate @xmath172/@xmath148 where @xmath172 is the number of the edge event estimated by the fit and @xmath148 is the total @xmath4 events selected .
it is related to the normalized branching ratio @xmath193 .
the normalized branching ratio is sensitive to the branching ratios for @xmath303 and @xmath304 , and uncertainty on fragmentation and @xmath83-tagging efficiency would be canceled out in the ratio . in fig .
[ msugrafig3](b ) we show the branching ratio br@xmath317 and the branching ratio of the gluino to the edge events normalized by br@xmath318 .
they are decreasing functions of @xmath56 , since the @xmath303 and @xmath304 is heavier for larger @xmath56 . while the behavior of br@xmath318 is moderate , the normalized branching ratio of the edge events is sensitive to @xmath56 , since it depends on the branching ratios of @xmath303 and @xmath304 . in fig .
[ msugrafig3](b ) the normalized branching ratio has two kinks , where some decay modes for @xmath303 or @xmath304 are open .
the first kink around @xmath319 gev comes from a change of the branching ratio of @xmath303 .
the dominant decay modes of @xmath303 are @xmath320 and @xmath321 , however , the mode @xmath322 is open around @xmath323 gev .
the decay modes @xmath324 and @xmath325 are open at the second kink around @xmath326 gev .
the decay modes are not included in the edge events , because the end point @xmath327 is significantly lower than @xmath328 . the additional tags for @xmath329 discussed in the previous section to identify the decay works for the region .
since @xmath329 is higgsino - like and the yukawa coupling to @xmath304 is enhanced by the top - yukawa coupling , the decay mode @xmath325 becomes dominant for @xmath330 gev , and then the normalized branching ratio become quickly diluted . in the msugra ,
the masses of @xmath331 and @xmath332 are related with each other in a broad parameter space . as the determination of @xmath333 has been studied in ref .
@xcite , we now discuss @xmath51 and the normalized branching ratio with a fixed @xmath333 value .
we fix @xmath334 gev and @xmath335 gev for @xmath336 and 30 , and show @xmath150 @xmath337 as a function of @xmath205 in fig .
[ msugrafig4](a ) .
our parameter scan is restricted to the range @xmath338 tev .
we have two disconnected solutions corresponding to @xmath339 gev@xmath340 gev and @xmath341 gev @xmath342 gev for @xmath343 , respectively .
the @xmath56 region between the two solutions is not allowed because of the experimental constraint of @xmath344 mass or charged lsp . for @xmath58=5 ( 30 ) , @xmath54 is smaller than 460 ( 750 ) gev for @xmath345 , and 500 ( 300 ) gev for @xmath346 . since @xmath347 in the msugra
, a larger @xmath348 corresponds to a larger @xmath54 for a fixed @xmath333 , as expected from eq .
. nevertheless @xmath303 could be much lighter than @xmath304 for a large and negative value of @xmath56 as can be seen in fig .
[ msugrafig4](a ) . for @xmath345 ,
satisfying a condition for the consistent minimum for the potential . ]
note that @xmath51 is determined for a fixed @xmath18 mass when @xmath205 is lighter than 370 gev .
if @xmath333 is constrained elsewhere , @xmath205 is strongly restricted by the @xmath51 measurement under the msugra assumption . in fig .
[ msugrafig4](b ) we show the solution in a @xmath51 and @xmath193 plane .
the normalized branching ratios is almost 1 for @xmath349 gev ( or @xmath350 gev ) .
when the mass difference of @xmath205 and @xmath333 is large , the decay of @xmath304 is dominated by @xmath351 .
furthermore , for @xmath352 gev , br@xmath353 is 100% .
however since @xmath354 in this region , squark and gluino production goes to the final states having four bottom quarks , where we have seen the disagreement between the measurement @xmath194 and @xmath355/br(@xmath356 ) in section iii d. in this case our study must be extended to events with more than three tagged @xmath83-jets .
note that the decay @xmath357 is mostly closed in fig .
[ msugrafig4 ] .
the decay is open only at the most right region of fig .
[ msugrafig4](b ) , near the end of the lines of the solutions .
if this decay is open , the masses of @xmath358 and @xmath359 are model - independently determined by the @xmath243 , @xmath360 and @xmath361 distributions .
when this decay mode is kinematically forbidden , one needs to combine various distributions for consistency check of the msugra assumptions . as is already shown in fig .
[ msugrafig2 ] , the decay @xmath362 is open up to @xmath363 , providing information of the third generation squarks in the wide region of the parameter space .
similar to the tau - lepton decay , we may measure the polarization of the top quark since it decays to @xmath364 via the @xmath365 interaction .
the top quarks from the @xmath17 , @xmath366 , and @xmath367 decays are polarized , and the polarization depends on the mixing angles for stops , charginos , and neutralinos .
the bottom quark angular distribution in the polarized top quark decay is the following , @xmath368 where @xmath369 is the angle between the direction of the bottom quark and the direction of the top quark spin in the rest frame of the top quark .
the terms proportional to @xmath370 come from the decay to the longitudinal @xmath94 boson .
the bottom quark thus tends to go to the opposite direction of the top quark spin . in the decay mode ( iii)@xmath107 , the top quark from a gluino decay
is polarized to be left - handed ( right - handed ) if @xmath303 is left - handed ( right - handed ) .
the polarization is reflected on the invariant mass distribution of the @xmath371 system ( @xmath3 ) , if the top quark is relativistic enough in the gluino rest frame . here
, one of the bottom quarks comes from the top quark decay and the other comes from @xmath372 .
especially , when the invariant mass @xmath1 is close to the end point @xmath39 of the decay mode ( iii)@xmath107 , the top and bottom quarks go to the opposite direction to each other in the gluino rest frame .
thus , the distribution of the invariant mass @xmath3 for events with @xmath1 close to @xmath373 is harder ( softer ) for left - handed ( right - handed ) top quarks . in fig .
[ polfig1 ] we show the @xmath3 distribution from the decay chain ( iii)@xmath107 . in this simulation
, we use the herwig generator , since it respects helicities for each particles in the processes .
we generated a large number of events which go through the decay ( iii)@xmath107 .
we use the mass spectrum at the reference point a1 in table [ masscross ] .
we use events with 400 gev@xmath374 gev to make the distribution , and the solid ( dotted ) line is for the left - handed ( right - handed ) stop .
the statistical significance in the difference between the left - handed and right - handed stops is about @xmath375 for @xmath376 events . in the above simulation
, we neglect contribution of other decay chains such as ( iii)@xmath30 and ( iv)@xmath30
. the other decay chains may contribute to the @xmath3 distribution even if we impose that @xmath1 is near the end point .
the @xmath3 distributions in the modes ( iii)@xmath30 and ( iv)@xmath30 do not depend on the polarization for the top quark , since the top quark distribution in the scalar boson decay is spherical . in fig .
[ polfig2 ] we show the @xmath1 distribution around the edge region at the reference point a1 .
the solid line is for the total distribution , the dashed line is for the distribution excluding the mode ( iii)@xmath107 , and the dotted line is for the distribution excluding the modes ( iii)@xmath107 , ( iii)@xmath30 , and ( iii)@xmath160 . about a half of the events near the end point come from the signal mode ( iii)@xmath29 , and the ratio between the signal events and the rest depends on the mssm parameters .
in this paper we study cascade decays @xmath377 at the lhc by reconstructing @xmath4 final state where the top quark decays hadronically .
the @xmath1 distribution of the cascade decay has an edge structure .
the measurement of the end point and the edge height of the @xmath1 distribution constrains a combination of the masses of @xmath378 and @xmath379 and the decay branching ratios of the particles involved in the decays . through a detailed simulation study
, we show in this paper the measurement of the end point and edge height on a continuum background is indeed possible .
namely , the end point of the cascade decay calculated in parton level agrees with the reconstructed edge position , and the ratio of @xmath149 ( the number of reconstructed edge events ) and @xmath148 ( the number of total reconstructed @xmath4 events ) is understood well by the ratio @xmath193 .
the end point and branching ratios depend on the mass and left - right mixing of the @xmath366 and @xmath367 , as well as the chargino and neutralino masses and mixings in the mssm . in the msugra
, these sparticle spectrum is expressed by a few parameters at the gut scale .
the decay mode @xmath380 is open for a wide region of the parameter space where @xmath381 .
the @xmath1 distribution is sensitive to the @xmath56 parameter , the trilinear coupling at the gut scale .
the distribution is most sensitive to @xmath56 when @xmath382 .
the stop and sbottom could decay both into the heavier and lighter charginos and neutralinos unlike the first and second generation squarks .
this is because the third generation squarks have the large top ( bottom)-yukawa coupling to the higgsinos . a strategy to search for such decays specific for the stop and sbottom is described in section iv .
the yukawa coupling is also related to the @xmath11-@xmath12 mixing .
the polarization of the top quark arising from the gluino decay @xmath383 depends on the stop left - right mixing .
the dependence of the @xmath3 distribution on the top polarization in the tagged @xmath4 sample is discussed in section vi . to understand the event distribution better
, one needs to know the nature of the quark and gluon fragmentations into jets .
the reconstruction efficiencies are significantly different between the two standard susy generators herwig and pythia , which adopt different models for the fragmentation .
we point out that the smearing of the jet - pair invariant mass originating from a @xmath94 decay affects the reconstruction efficiency .
the interplay between the lhc and the future lc would be useful to reduce the systematics coming from the uncertainties of sparticle masses and decay branching ratios .
they would be reduced dramatically if some of the charginos and neutralinos are accessible at the lc .
model independent and precise determination of the stop and sbottom masses may be possible in such cases and will be discussed elsewhere .
we thank the atlas collaboration members for useful discussion . we have made use of the physics analysis framework and tools which are the result of collaboration - wide efforts .
we especially thank dr .
kanzaki and mr .
we acknowledge icepp , univ . of tokyo , for providing us computing resources . this work is supported in part by the grant - in - aid for science research , ministry of education , science and culture , japan ( no.13135297 and no.14046225 for jh , no.11207101 for kk , and no.14540260 and 14046210 for mmn ) .
in this paper , we generate events both by the pythia and herwig generators .
the herwig generator uses a parton - shower approach for initial- and final- state qcd radiations including the color coherence effects and the azimuthal correlation both within and between jets @xcite .
the full available phase space for the parton shower is restricted to an angular order region , namely , the angle between the two emitted partons is smaller than that of previous branches . on the other hand ,
the pythia generator adopts the string model @xcite .
the two generators predict different @xmath4 reconstruction efficiencies , which may be considered as the uncertainty in fragmentation .
definition of jets also affects the number of reconstructed @xmath4 events and the reconstruction efficiency .
we try two algorithms , a cone - based algorithm and a @xmath80 algorithm ( montreal version ) , available in the jet finder library @xcite , which is interfaced to the atlfast packages .
the cone - based algorithm merges the calorimeter cells around a high @xmath82 cell in a fixed cone size @xmath79 . on the other hand , in the @xmath80 algorithm ,
a cell called a protojet @xmath384 grows until there are no more protojets @xmath385 which satisfy @xmath386 , therefore the cone size depends on @xmath82 s , and the shape of the jet depends on the distribution of the particles in the jet .
the @xmath80 algorithm is based on qcd , and it is considered to be advantageous to merge soft jets from an initial parton , although experimentally challenging .
the fit results with different generators and jet finding algorithms at the point a1 is summarized in table [ fitjet ] .
the end point @xmath120 becomes closer to @xmath51 as @xmath79 increases .
this is expected because more soft jets are merged to leading jets as @xmath79 increases .
the edge height @xmath114 is larger for herwig with the @xmath80 algorithm .
the difference of the height for @xmath387 is more than 25% in table ix .
the height significantly decreases for @xmath388 .
this is because we have to reconstruct a jet pair with relatively small invariant mass @xmath197 80 gev .
@xmath114/(10 gev)py & cone 0.4 & 455.2 @xmath159 8.2 & 271.4 @xmath159 22.7 & 0.5 & 436.1 @xmath159 7.7 & 272.9 @xmath159 33.4 & 0.6 & 461.1 @xmath159 10.4 & 217.8 @xmath159 22.8 & @xmath80 0.4 & 442.0 @xmath159 4.7&321.3 @xmath159 22.3 & 0.5 & 452.2 @xmath159 6.0 & 305.3 @xmath159 21.3 & 0.6 & 459.1 @xmath159 6.1 & 241.6 @xmath159 31.6hw & cone 0.4 & 434.5 @xmath159 5.8 & 354.8 @xmath159 23.3 & 0.5 & 460.2 @xmath159 4.9 & 349.2 @xmath159 22.8 & 0.6 & 440.9 @xmath159 7.1 & 305.3 @xmath159 33.7 & @xmath80 0.4 & 434.9 @xmath159 4.3 & 406.5 @xmath159 22.1 & 0.5 & 460.0 @xmath159 5.5 & 379.6 @xmath159 33.5 & 0.6 & 468.4 @xmath159 5.8 & 314.3 @xmath159 20.7 after sub & & & & & cone & 0.4 & 10180 & 2949 & 9695 & 2462 cone & 0.5 & 9659 & 2974 & 9213 & 2447
@xmath80 & 0.4 & 9405 & 3105 & 9226 & 2813@xmath80 & 0.5 & 9118 & 3239 & 8607 & 2808&&&&&cone & 0.4 & 2779 & 574 & 2630 & 500@xmath80 & 0.5 & 2086 & 448 & 2058 & 433 to see the origin of the difference more closely , we compare the @xmath95 distributions consistent to the top interpretation in fig .
[ modemjj ] , where , the solid histograms are the invariant mass distributions for jet pairs which satisfy @xmath389 gev and @xmath390 gev . dashed histograms show the distribution of accidental jet pairs in the @xmath94 mass region , that is , the `` fake @xmath94 '' background , estimated by the sideband method .
the left plots are the distributions of the events arising from the decay chain ( iii)@xmath107 , while the right plots are the @xmath1 distribution of all selected @xmath4 events .
the distribution at the bottom - left ( herwig and the @xmath80 algorithm for @xmath157 ) are more concentrated around @xmath391 gev compared to the others ( pythia and the cone - based or the @xmath80 algorithm ) , corresponding to better reconstruction efficiencies .
incompleteness of the sideband subtraction is seen in the same figure .
the sideband subtraction uses the jet pairs with invariant mass 35 gev@xmath392 gev and 95 gev@xmath393 125 gev to estimate the `` fake w '' background . on the other hand , as the jets are defined to have @xmath39410 gev , a jet pair near @xmath395 gev is suppressed .
we see an underestimate of background events in the range 65 gev@xmath393 70 gev , where the dotted line and solid line differ significantly .
we find that events with @xmath396 gev do not contribute to the edge structure of the @xmath1 distribution .
the events in this region might have to be removed from the candidate events .
the incompleteness of the sideband subtraction can also be seen by comparing the reconstruction efficiency of the events from specific decay chains . here
we take two decay chains for comparison ; @xmath397 the decay chain ( iii)@xmath107 is relevant to the @xmath4 events . on the other hand , as the decay chain ( i)@xmath25 has no @xmath94 ,
the sideband subtraction should remove the events arising from the decay chain completely , unless there are accidental weak bosons , or the understimation of the misreconstructed events that has just been discussed above .
the numbers of reconstructed @xmath4 events before and after the sideband subtraction are listed in table [ mode ] . for the @xmath80 algorithm with @xmath157 , the ratio of the numbers n(i)@xmath25/n(iii)@xmath107 are 0.229 and 0.15 before and after the sideband subtraction , respectively .
comparing these ratios to that of the branching ratios br(i)@xmath25/br(iii)@xmath107=0.32 , we see that the requirement @xmath400 gev reduces the contribution of ( i)@xmath25 to about 70% relative to ( iii)@xmath107 and it is further reduced to less than 50% after the sideband subtraction .
this ratio may be improved further by cutting the events with @xmath396 gev or by using more sophisticated background estimation .
m. dine , a. kagan and s. samuel , phys .
b * 243 * , 250 ( 1990 ) ; s. dimopoulos and g. f. giudice , phys .
b * 357 * , 573 ( 1995 ) ; a. pomarol and d. tommasini , nucl .
b * 466 * , 3 ( 1996 ) ; a. g. cohen , d. b. kaplan and a. e. nelson , phys .
b * 388 * , 588 ( 1996 ) ; j. hisano , k. kurosawa and y. nomura , phys . lett .
b * 445 * , 316 ( 1999 ) : nucl .
b * 584 * , 3 ( 2000 ) .
nelson and m.j .
strassler , jhep * 0009 * , 030 ( 2000 ) ; t. kobayashi and h. terao , phys . rev .
d * 64 * , 075003 ( 2001 ) ; t. kobayashi , h. nakano , and h. terao , phys .
d * 65 * , 015006 ( 2002 ) ; t. kobayashi , h. nakano , t. noguchi and h. terao , phys . rev .
d * 66 * , 095011 ( 2002 ) ; jhep * 0302 * , 022 ( 2003 ) .
h. e. haber and r. hempfling , phys .
lett . *
66 * , 1815 ( 1991 ) ; y. okada , m. yamaguchi and t. yanagida , prog .
. phys . * 85 * , 1 ( 1991 ) ; j. r. ellis , g. ridolfi and f. zwirner , phys .
b * 257 * , 83 ( 1991 ) ; r. barbieri and m. frigeni , phys .
b * 258 * , 395 ( 1991 ) .
j. hisano , k. kawagoe , r. kitano and m. m. nojiri , phys .
d * 66 * , 115004 ( 2002 ) .
h. baer , f. e. paige , s. d. protopopescu and x. tata `` isajet 7.48 : a monte carlo event generator for @xmath5 , @xmath401 , and @xmath200 reactions '' , hep - ph/0001086 . | in supersymmetric models a gluino can decay into @xmath0 through a stop or a sbottom .
the decay chain produces an edge structure in the @xmath1 distribution .
monte carlo simulation studies show that the end point and the edge height would be measured at the cern lhc by using a sideband subtraction technique .
the stop and sbottom masses as well as their decay branching ratios are constrained by the measurement .
we study interpretations of the measurement in the minimal supergravity model .
we also study the gluino decay into @xmath2 as well as the influence of the stop left - right mixing on the @xmath3 distribution of the tagged @xmath4 events . |
for some time now it has been standard practice in @xmath1 reactions to compare event - shape distributions with resummed perturbative predictions ( see for instance @xcite ) .
the resummation is necessary because in the two - jet limit ( small values of the shape variable ) the presence of large soft and collinear logarithms spoils the convergence of the fixed - order calculations .
such resummed analyses have led to valuable information about the strong coupling constant and also about non - perturbative effects . at hera ,
studies of event - shape distributions are being carried out by both collaborations @xcite but as yet no perturbative resummed calculations exist for comparison . here
we present preliminary results on such a calculation @xcite . in dis ,
event shapes are defined in the current hemisphere of the breit frame @xcite to reduce contamination from remnant fragmentation which is beyond perturbative control .
the distribution of emissions in the current hemisphere ( @xmath2 ) is analogous to that in a single hemisphere of @xmath1 .
differences arise from @xmath3 however because the momentum of the current quark ( the quark struck by the photon ) depends on remnant - hemisphere ( @xmath4 ) emissions through recoil effects .
this necessitates a resummed treatment of the space - like branching of the incoming parton .
currently such a simultaneous resummation of the space - like and time - like ( double logarithmic ) contributions exists only for jet multiplicities @xcite , or for cross sections at large @xmath5 @xcite . for the thrust @xmath6 ( as defined in the following section ) the recoil from remnant hemisphere emissions can be divided into two parts : a piece from soft and collinear emissions , which gives double logarithms @xmath7 , identical to those from half a hemisphere of @xmath3 , and a piece from purely collinear emissions which gives single logarithms @xmath8 that can be identified with a change of scale in the parton distribution .
this is outlined in the next section .
a valuable cross - check of any resummation is to expand it to nlo and compare the result to that from fixed order monte carlo programs such as disaster++ @xcite and disent @xcite .
this is illustrated in section [ sec : fixed ] . finally in section [ sec : future ] , we comment on possible future developments .
there are several possible definitions of the thrust in dis , which differ according to a choice of axis and the normalisation .
we here consider the thrust with respect to the photon axis , @xmath9 , and normalised to @xmath10 , @xmath11 where the sum extends over all particles in the current hemisphere , @xmath12 . at lowest order @xmath13 , so the region requiring resummation will be that of @xmath14 close to zero . expressing all momenta in terms of their sudakov components @xmath15 where @xmath16 and @xmath17 are along the remnant and current directions respectively ,
we can write the thrust as @xcite @xmath18 where the sums now run over emitted particles only and @xmath19 is the @xmath20 component of the current quark . ] .
dependence on the remnant hemisphere emissions arises both through the @xmath21 term and through @xmath22 ( we have made use of the fact that @xmath23 ) .
we now consider the cross section @xmath24 for @xmath14 to be smaller than @xmath25 .
roughly it will contain virtual corrections from the exclusion of all emissions whose contribution to is larger than @xmath25 . in the remnant hemisphere
we exclude soft and collinear emissions , @xmath26 , where the @xmath27 condition ensures that particle be in @xmath4 .
the integration over @xmath28 and @xmath29 gives a double logarithm .
in @xmath2 we exclude the soft and collinear region @xmath30 .
again this gives us a double logarithm .
the exclusion @xmath31 is just a limit on the maximum emitted transverse momentum : this implies ` stopping ' the dglap @xcite parton evolution at a scale @xmath32 , which gives single collinear logarithms .
these aspects are illustrated in the first order result for @xmath24 : @xmath33 \\ - \frac{\as}{2\pi } \frac{\ln 1/\tau}{q(x ) } \int_x^1 \frac{d z}{z } \left[q\left(\frac{x}{z}\right ) p_{qq } ( z ) + g\left(\frac{x}{z}\right ) p_{qg } ( z ) \right ] .\end{gathered}\ ] ] the first line contains the soft and collinear double logarithms ( which turn out to be identical to the @xmath3 result ) from the conditions on the @xmath34 and @xmath35 .
the second line contains the collinear logarithm associated with the restriction on the dglap evolution .
photon - gluon fusion plays a role only through this collinear logarithm , in the convolution with the gluon distribution .
the actual details of the resummation will be presented in @xcite .
schematically , the result is @xmath36 \frac{q(x,\tau q^2)}{q(x , q^2 ) } \ , \sigma(\tau),\qquad \sigma(\tau ) = e^ { - \frac{\as \cf}{2 \pi } \left[2 \ln^2 \frac1\tau - 3 \ln \frac1\tau \right ] + \co{\as^2\ln^3\frac1\tau}},\ ] ] where @xmath37 is just the corresponding @xmath3 quantity of @xcite and @xmath38 is a @xmath25-independent coefficient function which is given in @xcite .
the ratio @xmath39 comes from the suppression of radiation with @xmath40 , as mentioned above .
expanding to nlo it is possible to perform a comparison to the fixed order monte carlo programs disaster++ @xcite and disent @xcite .
we actually look at the coefficient of @xmath41 in @xmath42 where we have normalised to the the born cross section @xmath43 for simplicity . at second order eq . contains terms @xmath44 , @xmath45 and for the resummation to be correct to next - to - leading order we should correctly control terms @xmath46 .
so the difference between our expanded result and the exact result should at most be a constant for small @xmath25 .
the upper two plots of figure [ fig : disdis ] show our results for the quark and gluon - initiated components of the answer compared to the predictions from disaster++ .
the shape of the distributions is well reproduced for small @xmath25 .
the lower two plots show the difference between the disaster++ results and ours : one sees that for small @xmath25 this difference is indeed compatible with a constant , as required .
comparisons have also been made with disent , which seems to disagree with our result in the gluon sector at the level of a term proportional to @xmath47 , and perhaps also in the quark sector .
although we have results for the resummation of the thrust distribution in dis , some work remains to be done for the practical implementation of our results for comparison with experimental data .
in particular , prescriptions need to be defined for the matching of our resummed result with the fixed order results , in order extend the range of applicability of the calculation ( without the matching the results are applicable only to very small @xmath25 ) .
work is also in progress on the resummation of other dis event - shape variables . among these ,
the thrust normalised to the energy in the current hemisphere is close to completion .
other variables to be studied include the jet mass , the @xmath48-parameter and the jet - broadening ( the resummation of the latter is relevant also for predicting the form of the power correction @xcite ) .
once this programme is complete , we hope that it will lead to phenomenological studies analogous to those already being carried out in @xmath3 .
we benefited much from continuous discussions of this and related subjects with yuri dokshitzer , pino marchesini and mike seymour .
one of us ( va ) is grateful to yuri dokshitzer and pino marchesini for introducing him to perturbative qcd , and to the organisers and the participants of the 1999 u.k .
phenomenology workshop on collider physics for the invitation and for the stimulating atmosphere .
he would also like to thank n. brook and v. khoze for useful discussions .
this research was supported in part by murst , italy and by the eu fourth framework programme ` training and mobility of researchers ' , network ` quantum chromodynamics and the deep structure of elementary particles ' , contract fmrx - ct98 - 0194 ( dg 12-miht ) .
99 s. catani , l. trentadue , g. turnock and b.r .
webber , _ nucl .
* b 407 * ( 1993 ) 3 .
feynman r.p .
1972 _ photon hadron interactions _ ( w. a. benjamin , new york ) ; dokshitzer yu .
l. , khoze v.a . and troyan s.i .
1989,in _ perturbative quantum chromodynamics _
a. h. mueller , world scientific , singapore ) .
m. dasgupta and b.r .
webber , _ jhep _ * 10 * ( 1998 ) 001 [ hep - ph/9809247 ] .
gribov and l.n .
lipatov , _ sov .
_ * 15 * ( 1972 ) 438 ; + g. altarelli and g. parisi , _ nucl .
* b 126 * ( 1977 ) 298 ; + yu.l .
dokshitzer , _ sov .
* 46 * ( 1977 ) 641 . | we present preliminary results on the resummation of leading and next - to - leading logarithms for the thrust distribution in deep inelastic scattering .
our predictions , expanded to @xmath0 , are compared to corresponding results from the monte carlo programs disaster++ and disent .
bicocca
ft9931 + hep - ph/9910343 + october 1999 |
in @xcite has been obtained a representation in an explicit form of the solution of the linear partial differential equation of the higher order in two variables with initial condition whose coefficients were real - valued coefficients .
the aim of the present manuscript is resolve an analogous problem for a linear partial differential equation of the higher order in two variables with initial condition whose coefficients are real - valued simple step functions .
the paper is organized as follows . in section 2 , we consider some auxiliary results obtained in the paper @xcite . in section 3 , it is obtained a representation in an explicit form of the weak solution of the partial differential equation of the higher order in two variables with initial condition whose coefficients are real - valued simple step functions .
fourier differential operator @xmath0 in @xmath1 is defined as follows : @xmath2 for @xmath3 , let @xmath4 be a vector space of all @xmath5-times differentiable functions on @xmath6 such that for arbitrary @xmath7 , a series obtained by a differentiation term by term of the fourier series of @xmath8 pointwise converges to @xmath9 for all @xmath10 .
let @xmath11 .
let @xmath12 be an embedding of the @xmath13 in to @xmath1 which sends a function to a sequence of real numbers consisting from its fourier coefficients .
i.e. , if @xmath14 then @xmath15 . then , for @xmath11 , the following equality @xmath16 holds . assume that for @xmath11 , we have the following representation @xmath17 by the definition of the class @xmath13 , we have @xmath18 @xmath19 @xmath20 @xmath21 by the definition of the composition of mappings , we have @xmath22 @xmath23 @xmath24 by the scheme used in the proof of lemma 2.1 , we can get the validity of the following assertion .
let @xmath12 be an embedding of the @xmath4 in to @xmath1 which sends a function to a sequence of real numbers consisting from its fourier coefficients .
then , for @xmath25 and @xmath26 , the following equality @xmath27 holds .
@xcite if @xmath28 is the real matrix @xmath29 then @xmath30 for @xmath31 , let us consider a linear autonomous nonhomogeneous ordinary differential equations of the first order @xmath32 with initial condition @xmath33 where @xmath34 @xmath35 ; @xmath36 @xmath37 is the sequence of continuous functions of a parameter @xmath38 on @xmath39 . for each @xmath40
, we put @xmath41 @xmath42 then the solution of ( 2.6)-(2.7 ) is given by @xmath43 where @xmath44 denotes an exponent of the matrix @xmath45 and it exactly coincides with an infinite - dimensional @xmath46 -cellular matrix @xmath47 with cells @xmath48 for which @xmath49 and @xmath50 where for @xmath40 , @xmath51 and @xmath52 are defined by ( 2.8)-(2.9 ) , respectively . we know that if we have a linear autonomous inhomogeneous ordinary differential equations of the first order @xmath53 with initial condition @xmath54 where @xmath34 @xmath35 ; @xmath36 @xmath55 is the sequence of continuous functions of parameter @xmath38 on @xmath39 ; @xmath56 @xmath57 is an infinite dimensional @xmath46-cellular matrix with cells @xmath58 .
then the solution of ( 2.6)-(2.7 ) is given by ( cf .
@xcite , 6 , section 1 ) @xmath59 where @xmath60 and @xmath61 denote exponents of matrices @xmath62 and @xmath63 , respectively .
note that @xmath64 is an infinite - dimensional @xmath46-cellular matrix with cells @xmath65 such that @xmath66 and @xmath67 for @xmath40 . under notations ( 2.8)-(2.9 ) , by using example 2.4 we get that for @xmath68 , @xmath60 exactly coincides with an infinite - dimensional @xmath46 -cellular matrix @xmath47 with cells @xmath48 for which @xmath49 and @xmath69 note that , for @xmath70 , the matrix @xmath61 exactly coincides with an infinite - dimensional @xmath46 -cellular matrix @xmath71 .
the following proposition is a simple consequence of lemma 2.5 . for @xmath31 ,
let us consider a linear partial differential equation @xmath72 with initial condition @xmath73 if @xmath74 is such a sequence of real numbers that a series @xmath75 defined by @xmath76 @xmath77 belongs to the class @xmath78 as a series of a variable @xmath79 for all @xmath80 , and is differentiable term by term as a series of a variable @xmath38 for all @xmath10 , then @xmath81 is a solution of @xmath82-@xmath83 .
let @xmath84 and @xmath85 .
suppose that @xmath86 for @xmath31 , let us consider a partial differential equation @xmath87 with initial condition @xmath88 ( iii ) for each fixed @xmath10 , the function @xmath75 is continuous with respect to @xmath92 , and for each @xmath93 the function @xmath75 is continuous with respect to @xmath79 on @xmath6 except points @xmath94 .
@xmath118 setting @xmath119 and @xmath120 for @xmath116 we obtain @xmath121 and @xmath122 it is obvious that the system of equations ( 3.13)-(3.14 ) has the unique solution which can be done as follows : @xmath123 and @xmath124 for @xmath116 .
it is obvious that under nice restrictions on coefficients participated in ( 3.1 ) and ( 3.2 ) , we can continue our procedure step by step .
correspondingly we can construct a sequence @xmath128 such that @xmath129 satisfies a linear partial differential equation @xmath130 with initial condition @xmath131 notice that for each natural number @xmath207 , one can easily modify the mathlab program described in example 3.3 for obtaining the graphical solution of the linear partial differential equation ( 3.1)-(3.2 ) whose coefficients @xmath208 are real - valued simple step functions on @xmath209 and @xmath210 is a trigonometric polynomial on @xmath6 .
since each constant @xmath211 same times is a step function we can use mathlab program described in example 3.3 for a solution of the linear partial differential equation ( 2.17)-(2.18 ) with constant coefficients .
the approach used for a solution of ( 3.1)-(3.2 ) can be used in such a case when coefficients @xmath217 are rather smooth continuous functions on @xmath218 .
if we will approximate @xmath219 by real - valued simple step functions , then it is natural to wait that under some `` nice restrictions '' on @xmath219 the solutions obtained by theorem 3.2 , will give us a `` good approximation '' of the solution of the required linear partial differential equation of the higher order in two variables with corresponding initial conditions . | by using the method developed in the paper [ _ georg . inter .
j. sci . tech . ,
_ volume 3 , issue 1 ( 2011 ) , 107 - 129 ] , it is obtained a representation in an explicit form of the weak solution of a linear partial differential equation of the higher order in two variables with initial condition whose coefficients are real - valued simple step functions * 2000 mathematics subject classification : * primary 34axx ; secondary 34a35 , 34k06 . *
key words and phrases : * linear partial differential equation of the higher order in two variables , fourier differential operator . |
scattering resonances play an important role in physics . in the field of ultracold atoms ,
feshbach resonances induced with magnetic field are used to control the interaction between atoms.@xcite this unparalleled tunability of the interaction has been essential to the recent remarkable advance of the field and led to experimental realization of a rich variety of physics , such as the bcs - bec crossover in fermi gases@xcite and the efimov effect in bose gases.@xcite these strongly correlated phenomena in the vicinity of the scattering resonance are universal , i.e. , independent of microscopic details .
also , an atom loss peak caused by the three - atom or atom - dimer resonance has been used as a signature of the emergence of an efimov trimer.@xcite turning to the field of condensed matter , the scattering resonance can be induced between collective excitations in ferromagnetic spin systems ( magnons ) , for example , by tuning the exchange coupling.@xcite because the ferromagnetic coupling is sensitive to the structure of ions,@xcite it is experimentally possible to vary the exchange coupling significantly with pressure@xcite and bring the system close to the two - magnon resonance .
future achievement of the two - magnon resonance may have an impact comparable to the feshbach resonance in ultracold atoms and will open up a rich variety of strongly correlated physics such as the recently proposed efimov effect in quantum magnets.@xcite but , what is an experimental signature of the two - magnon resonance ? in this article , we study the electron spin resonance ( esr ) in a dilute magnon gas as a probe of magnon scattering resonances .
esr is a powerful experimental technique to investigate the dynamics of interacting spin systems,@xcite while theoretical calculations of its spectrum are in general challenging .
here we develop a quantum cluster ( or virial ) expansion which uses a small parameter called fugacity @xmath0 related to the diluteness of the system to perform a systematic expansion , where @xmath1 is a single - particle excitation energy and @xmath2 is the inverse temperature.@xcite we first demonstrate the quantum cluster expansion in sec .
[ sec : boson ] by computing a single - particle spectral function in a dilute bose gas , which serves as the universal formula for esr at low temperature and can be used to extract the scattering length between magnons . then in sec .
[ sec : magnon ] , we apply the same method to esr in a dilute magnon gas and show that the frequency shift of the single - magnon peak changes its sign and the linewidth reaches its maximum across a scattering resonance between magnons .
such characteristic behaviors are universal and can be used to experimentally locate the two - magnon resonance when an external parameter such as pressure is varied . finally , sec .
[ sec : summary ] is devoted to the conclusion of this article and discussion where we also suggest that the emergence of an efimov state of three magnons may be observed with esr through an additional peak structure in the linewidth caused by the three - magnon resonance .
below we set @xmath3 and the unspecified range of integration is assumed to be from @xmath4 to @xmath5 .
before we work on esr in a dilute magnon gas , it is instructive to demonstrate the quantum cluster expansion in a dilute bose gas , which is described by a hamiltonian @xmath6 here @xmath7 is the single - particle dispersion relation , @xmath8 is the chemical potential , and the bare coupling constant @xmath9 is related to the @xmath10-wave scattering length @xmath11 by @xmath12 with @xmath13 being a momentum cutoff .
our purpose in this section is to compute the single - particle spectral function @xmath14 where @xmath15 with @xmath16 being the matsubara frequency is the fourier transform of the imaginary - time propagator @xmath17 as we will show in the next section , @xmath18 serves as the universal formula for esr at low temperature and can be used to extract the scattering length between magnons .
the imaginary - time propagator is defined by @xmath19,\end{aligned}\ ] ] where @xmath20 is the grand canonical partition function : @xmath21 $ ] .
the systematic expansion over the fugacity @xmath22 can be developed by writing the grand canonical trace as a sum over canonical traces with fixed particle number @xmath23 : @xmath24=\sum_{n=0}^\infty{\mathrm{tr}}_n[\,\cdot\,]$ ] .
because @xmath25\propto z^n$ ] , we obtain @xmath26 where @xmath27 is the system volume and @xmath28 is the thermal de broglie wavelength .
accordingly , the particle number density is found to be @xmath29 therefore , the quantum cluster expansion is valid when the system is so dilute that the mean interparticle distance is much larger than the thermal de broglie wavelength : @xmath30 .
the numerator in eq .
( [ eq : propagator - time ] ) can be expanded over @xmath31 in the same way . by denoting the fourier transform of each term as @xmath32 \sim o(z^n),\end{aligned}\ ] ] the leading term is easily evaluated as @xmath33 on the other hand , after a straightforward calculation , the next - to - leading term is evaluated as @xmath34 where @xmath35 is the forward scattering amplitude between a particle with energy - momentum @xmath36 and an on - shell particle with momentum @xmath37 .
accordingly , the fourier transform of the imaginary - time propagator ( [ eq : propagator - freq ] ) can be brought into the standard form @xmath38 where the quantum cluster expansion of the self - energy is found to be @xmath39 therefore , the self - energy at @xmath40 is determined only by the two - particle physics , i.e , binary collisions with thermally excited particles . in eq .
( [ eq : boson_on - shell ] ) as functions of @xmath41 , represented by solid and dashed curves , respectively .
[ fig : boson ] ] the resulting single - particle spectral function ( [ eq : boson_spectrum ] ) at @xmath42 is simply a delta function located at @xmath43 corresponding to the single - particle energy . by including the self - energy correction @xmath44 ,
it becomes a sharp peak whose line shape within the accuracy up to @xmath40 is described by the lorentzian @xmath45 ^ 2+[{\mathrm{im}}\,\sigma({{\bm{p}}})]^2},\end{aligned}\ ] ] where we introduced the on - shell self - energy @xmath46 therefore , the energy shift and the decay width of a particle with momentum @xmath47 in a dilute bose gas are given by the real and imaginary parts of @xmath48 , respectively .
the corresponding normalized quantity @xmath49 at @xmath50 is plotted in fig . [ fig : boson ] as a function of @xmath41 .
we find that the energy shift changes its sign and the decay width reaches its maximum across the scattering resonance at @xmath51 .
such characteristic behaviors are universal and shared by esr at low temperature . for @xmath52 and @xmath53 as a function of @xmath54 .
the single - particle peak goes up to @xmath55 and there is a threshold singularity at the two - particle binding energy @xmath56 .
[ fig : bound ] ] besides the single - particle peak at @xmath57 , there exists an additional structure in @xmath58 at @xmath59 when two particles form a bound state with binding energy @xmath60 .
this structure is due to the pole of the two - particle scattering amplitude ( [ eq : boson_amplitude ] ) and shows a square - root threshold singularity at @xmath61 in the single - particle spectral function as @xmath62 .
this feature at @xmath50 is demonstrated in fig .
[ fig : bound ] by choosing @xmath52 and @xmath53 as an example .
we now apply the same method to esr in a dilute magnon gas . for definiteness
, we consider a spin-@xmath63 heisenberg model with exchange ( @xmath64 ) and single - ion ( @xmath65 ) anisotropies , which is described by @xmath66 here @xmath67 is a sum over six unit vectors of a simple cubic lattice and spin operators @xmath68 and @xmath69 obey the usual commutation relations : @xmath70=2s_{{\bm{r}}}^z\delta_{{{\bm{r}}},{{\bm{r}}}'}$ ] and @xmath71=\pm s_{{\bm{r}}}^\pm\delta_{{{\bm{r}}},{{\bm{r}}}'}$ ] .
esr experiments measure an absorption intensity of electromagnetic radiation polarized perpendicular to the magnetic field axis . within the linear response theory for a circular polarization ,
the absorption intensity normalized by the system volume and the intensity of the incident radiation is given by@xcite @xmath72 where @xmath73 is the fourier transform of the imaginary - time susceptibility @xmath74 while the esr experiments can measure the spectrum only at zero momentum , we develop the formulation for general @xmath47 and set @xmath50 at the end . the ground state for a sufficiently large magnetic field @xmath75 is a fully polarized state with all spins pointing downwards : @xmath76 and @xmath77 .
accordingly , we redefine the hamiltonian to absorb the ground state energy @xmath78 so that @xmath79
. then at low temperature , magnons are thermally excited and thus the system becomes a dilute magnon gas .
a single magnon has the dispersion relation @xmath80 $ ] with the excitation energy @xmath81 .
as long as the fugacity is small , @xmath22 , the quantum cluster expansion can be developed for the dilute magnon gas similarly to the previous dilute bose gas .
the imaginary - time susceptibility is defined by @xmath82,\end{aligned}\ ] ] where @xmath20 is the grand canonical partition function : @xmath21 $ ] . by writing the grand canonical trace as a sum over canonical traces with fixed magnon number @xmath23
, we obtain @xmath83 = 1 + v\frac{z}{\rho^3 } + o(z^2),\end{aligned}\ ] ] where we used @xmath25\propto z^n$ ] and introduced the analogue of the thermal de broglie wavelength by @xmath84 with @xmath85 being the lattice constant .
accordingly , the magnetization relative to the ground state ( @xmath86 ) is found to be @xmath87 the numerator in eq .
( [ eq : susceptibility - time ] ) can be expanded over @xmath31 in the same way . by denoting the fourier transform of each term as @xmath88 \sim o(z^n),\end{aligned}\ ] ] the leading term
is easily evaluated as @xmath89 on the other hand , after a straightforward calculation , the next - to - leading term is evaluated as @xmath90 where @xmath91\,\gamma(\hat{{\bm{e } } } ) \notag\\ & \quad - 2d\gamma({{\bm{0}}})\end{aligned}\ ] ] is the forward scattering amplitude between a magnon with energy - momentum @xmath36 and an on - shell magnon with momentum @xmath37 . here
the unknown function @xmath92 implicitly depends on @xmath93 and satisfies the lippmann - schwinger equation @xmath94\,\gamma(\hat{{\bm{e } } } ) - 2d\gamma({{\bm{0 } } } ) } { i\omega_n-\delta+{\varepsilon_{{\bm{q}}}}-\varepsilon_{({{\bm{p}}}+{{\bm{q}}})/2+{{\bm{k}}}}-\varepsilon_{({{\bm{p}}}+{{\bm{q}}})/2-{{\bm{k}}}}}.\end{aligned}\ ] ] by setting @xmath95 and @xmath96 , we obtain four coupled equations to determine @xmath97 and @xmath98 appearing in eq .
( [ eq : magnon_amplitude ] ) . accordingly ,
the fourier transform of the imaginary - time susceptibility ( [ eq : susceptibility - freq ] ) can be brought into the standard form @xmath99 where the quantum cluster expansion of the self - energy is found to be @xmath100 therefore , the self - energy at @xmath40 is determined only by the two - magnon physics , i.e , binary collisions with thermally excited magnons . the resulting esr spectrum ( [ eq : magnon_spectrum ] ) at @xmath42 is simply a delta function located at @xmath101 corresponding to the single - magnon energy at @xmath50 . by including the self - energy correction @xmath102 ,
it becomes a sharp peak whose line shape within the accuracy up to @xmath40 is described by the lorentzian @xmath103 ^ 2+[{\mathrm{im}}\,\xi({{\bm{0}}})]^2},\end{aligned}\ ] ] where we introduced the on - shell self - energy : @xmath104 . therefore , the frequency shift and the linewidth of the single - magnon peak are given by the real and imaginary parts of @xmath105 , respectively . also , when two magnons form a bound state with binding energy @xmath106 , the esr spectrum shows an additional structure at @xmath107 similarly to fig .
[ fig : bound ] , while we will not investigate it further .
+ + + + our remaining task is to solve the lippmann - schwinger equation ( [ eq : lippmann - schwinger ] ) with @xmath108 and @xmath50 : @xmath109\,\gamma(\hat{{\bm{e } } } ) - 2d\gamma({{\bm{0 } } } ) } { { \varepsilon_{{\bm{q}}}}-\varepsilon_{{{\bm{q}}}/2+{{\bm{k}}}}-\varepsilon_{{{\bm{q}}}/2-{{\bm{k}}}}+i0^+}\end{aligned}\ ] ] to determine @xmath97 and @xmath98 in the two - magnon scattering amplitude ( [ eq : magnon_amplitude ] ) . in the low - temperature limit @xmath110 , the integration over @xmath37 in the self - energy ( [ eq : magnon_self - energy ] )
is dominated by the region @xmath111 because of the boltzmann factor . in this case , by expanding the right hand side of eq .
( [ eq : lippmann - schwinger0 ] ) up to @xmath112 , we find @xmath113 and the integration over @xmath114 can be performed analytically .
then , by substituting the obtained analytical solutions for @xmath97 and @xmath98 into eq .
( [ eq : magnon_amplitude ] ) , we find that the two - magnon scattering amplitude takes the same form as that of bosons in eq .
( [ eq : boson_amplitude ] ) : @xmath115 where @xmath116 is the inverse effective mass of magnons and @xmath117 } { 2s-1+\frac{j_z}{j}(1-\frac{d}{6sj})+3w[1-\frac{d}{3j}-\frac{j_z}{j}(1-\frac{d}{6sj})]}\end{aligned}\ ] ] is the scattering length between magnons with @xmath118 } \notag\\ & = \frac{\sqrt6}{96\pi^3}\,\gamma\!\left(\frac1{24}\right)\gamma\!\left(\frac5{24}\right ) \gamma\!\left(\frac7{24}\right)\gamma\!\left(\frac{11}{24}\right)\end{aligned}\ ] ] being the watson s triple integral for a simple cubic lattice.@xcite we note that the same scattering length was obtained in ref . with a different approach .
accordingly , the low - temperature limit of the on - shell self - energy at @xmath50 appearing in eq .
( [ eq : magnon_result ] ) reduces to that in eq .
( [ eq : boson_on - shell ] ) , [ eq : universal ] @xmath119 and thus the line shape of the single - magnon peak is described by the universal formula @xmath120 where @xmath121 is the single - particle spectral function of bosons obtained in the previous section as eq .
( [ eq : boson_result ] ) .
this result is actually expected because the magnon gas at low temperature is so dilute that the system becomes independent of microscopic details and thus described by only a few low - energy parameters such as @xmath122 and @xmath11.@xcite therefore , it is possible to extract the scattering length between magnons by fitting the universal formula ( [ eq : universal ] ) to the experimentally measured temperature dependence of the line shape of the single - magnon peak .
away from the low - temperature limit , the line shape ( [ eq : magnon_result ] ) is model - dependent and has to be evaluated numerically by solving eq .
( [ eq : lippmann - schwinger0 ] ) for general @xmath37 . the frequency shift and the linewidth of the single - magnon peak
are given by the real and imaginary parts of @xmath105 , respectively , and the corresponding normalized quantity @xmath123 is plotted in fig . [ fig : magnon ] .
for demonstration , we choose three distinct cases where @xmath124 , @xmath125 with @xmath126 as functions of @xmath127 , and @xmath125 with @xmath128 as a function of @xmath129 at four different temperatures , @xmath130 .
equation ( [ eq : scattering - length ] ) indicates that the two - magnon resonances @xmath51 are located at @xmath131 , @xmath132 , and @xmath133 , respectively , where magnons interact strongly .
we find that the line shape of the single - magnon peak is well described by the universal formula ( [ eq : universal ] ) at low temperature @xmath134 and thus the frequency shift changes its sign and the linewidth reaches its maximum across the two - magnon resonance .
such characteristic behaviors become sharper with decreasing temperature and can be seen moderately even at intermediate temperature @xmath135 , while they disappear at higher temperature @xmath136 .
in this article , we studied esr in a dilute magnon gas that is realized in a ferromagnetic spin system at low temperature .
we developed the quantum cluster expansion up to @xmath40 which is determined by the two - magnon physics and showed that the frequency shift of the single - magnon peak changes its sign and the linewidth reaches its maximum across a scattering resonance between magnons .
such characteristic behaviors are universal and can be used to experimentally locate the two - magnon resonance when an external parameter such as pressure is varied . future achievement of the two - magnon resonance may have an impact comparable to the feshbach resonance in ultracold atoms and will open up a rich variety of strongly correlated physics such as the recently proposed efimov effect in quantum magnets.@xcite it is straightforward in principle to continue this systematic expansion to include higher - order corrections .
in particular , the @xmath137 term involves the three - magnon physics and thus it is possible to probe the efimov effect with esr .
when the system comes across the critical coupling where an efimov state of three magnons emerges from the scattering threshold , the linewidth of the single - magnon peak as a function of the external parameter is expected to show an additional peak structure caused by the three - magnon resonance on either side of the two - magnon resonance .
this feature is in analogy with ultracold atom experiments where an atom loss peak caused by the three - atom or atom - dimer resonance has been used as a signature of the emergence of an efimov trimer.@xcite similarly , a pair of universal four - magnon states associated with every efimov state@xcite may be observed with esr through an additional peak structure in the linewidth at @xmath138 caused by the four - magnon resonance.@xcite besides such characteristic behaviors in the single - magnon peak , we expect an additional structure in the esr spectrum at @xmath139 when @xmath23 magnons form a bound state with binding energy @xmath140 .
this structure shows a threshold singularity at @xmath141 as @xmath142 , which may be used to measure binding energies of magnon efimov states .
therefore , esr is a powerful experimental technique to investigate the interaction among magnons and their spectrum .
so far we considered ferromagnetic spin systems where the single - magnon dispersion relation has a minimum at zero momentum . on the other hand , it is also possible to induce the two - magnon resonance and thus the magnon efimov effect in spin systems with antiferromagnetic or frustrated exchange couplings where the single - magnon dispersion relation has a minimum at nonzero momentum.@xcite in these cases , however , the sharp signature of magnon scattering resonances discussed in this article will not appear in esr because it measures the spectrum only at zero momentum .
therefore , different experimental techniques such as inelastic neutron scattering that can scan momentum space should be used to probe magnon scattering resonances .
the quantum cluster expansion developed here will be useful to compute any relevant physical observables in a dilute magnon gas . c. chin , r. grimm , p. julienne , and e. tiesinga , rev .
phys . * 82 * , 1225 ( 2010 ) . c. a. regal , m. greiner , and d. s. jin , phys .
* 92 * , 040403 ( 2004 ) .
m. w. zwierlein , c. a. stan , c. h. schunck , s. m. f. raupach , a. j. kerman , and w. ketterle , phys .
* 92 * , 120403 ( 2004 ) .
t. kraemer , m. mark , p. waldburger , j. g. danzl , c. chin , b. engeser , a. d. lange , k. pilch , a. jaakkola , h .- c .
ngerl , and r. grimm , nature * 440 * , 315 ( 2006 ) .
s. knoop , f. ferlaino , m. mark , m. berninger , h. schbel , h .- c .
ngerl , and r. grimm , nature phys . * 5 * , 227 ( 2009 ) .
y. nishida , y. kato , and c. d. batista , nature phys .
* 9 * , 93 ( 2013 ) .
j. b. goodenough , _ magnetism and the chemical bond , _
( john wiley & sons , new york , 1963 ) .
m. oshikawa and i. affleck , phys .
b * 65 * , 134410 ( 2002 ) .
m. brockmann , f. ghmann , m. karbach , a. klmper , and a. weie , phys . rev .
b * 85 * , 134438 ( 2012 ) .
g. n. watson , q. j. math . * 10 * , 266 ( 1939 ) .
h .- w . hammer and l. platter ,
j. a * 32 * , 113 ( 2007 ) . j. von stecher , j. p. dincao , and c. h. greene , nature phys .
* 5 * , 417 ( 2009 ) .
f. ferlaino , s. knoop , m. berninger , w. harm , j. p. dincao , h .- c .
ngerl , and r. grimm , phys .
* 102 * , 140401 ( 2009 ) .
f. ferlaino and r. grimm , physics * 3 * , 9 ( 2010 ) . | we study the electron spin resonance in a dilute magnon gas that is realized in a ferromagnetic spin system at low temperature .
a quantum cluster expansion is developed to show that the frequency shift of the single - magnon peak changes its sign and the linewidth reaches its maximum across a scattering resonance between magnons .
such characteristic behaviors are universal and can be used to experimentally locate the two - magnon resonance when an external parameter such as pressure is varied .
future achievement of the two - magnon resonance may have an impact comparable to the feshbach resonance in ultracold atoms and will open up a rich variety of strongly correlated physics such as the recently proposed efimov effect in quantum magnets .
we also suggest that the emergence of an efimov state of three magnons may be observed with the electron spin resonance through an additional peak structure in the linewidth caused by the three - magnon resonance . |
blazars are a class of radio - loud active galactic nuclei ( agn ) characterized by a broadband nonthermal continuum from radio to @xmath2-rays , extreme variability and a high degree of polarization . in the framework of the unified agn model
( see e.g. * ? ? ?
* ) they are considered agn whose relativistic jet is aligned with the line of sight .
the blazar spectral energy distribution ( sed ) is thus dominated by the emission from the jet , enhanced by relativistic effects . in the @xmath2-ray sky , blazars are the dominant agn class , representing @xmath4 of the _ fermi_-lat agn ( between 100 mev and 100 gev , * ? ? ?
* ) , and @xmath5 of extragalactic sources detected at very high energies ( vhe ; @xmath6 gev ) by ground - based imaging atmospheric - cherenkov telescopes ( iacts ) .
+ the blazar class is composed of two subclasses , flat - spectrum radio quasars ( fsrq ) and bl lacertae objects , depending on the presence ( in the former ) or absence ( in the latter ) of emission lines in their optical / uv spectrum ( the threshold between the two subclasses is an emission line equivalent width of 0.5 nm , see e.g. * ? ? ?
the two subclasses are also characterized by different luminosity and redshift distributions .
the fsrqs are on average brighter and more distant than bl lac objects ( see e.g. * ? ? ?
this dichotomy in the blazar class reflects the dichotomy observed in the radio - galaxy population : fsrqs are considered the blazar version of the fanaroff - riley ii ( fr ii , * ? ? ?
* ) radio - galaxies , while bl lac objects are believed to represent the blazar version of fr i. the two subclasses share the same sed shape : a broadband continuum from radio to gamma - rays , composed of two separate bumps , peaking in ir - to - x - rays and mev - to - tev , respectively .
while fsrqs are generally characterized by a lower frequency of the first peak , bl lac objects show a variety of peak frequencies , and are further classified ( see e.g. * ? ? ?
* ) into low - frequency - peaked bl lacs ( lbl , with @xmath7 in infrared ) and high - frequency - peaked bl lacs ( hbl , with @xmath7 in uv and beyond ) .
when @xmath7 is located in the optical / uv the object is often classified as an intermediate - frequency - peaked bl lac ( ibl ) . + the position of the second peak is related to the position of the first one , as shown by observations in the mev - gev energy band with _ fermi_-lat . while fsrqs and lbls present a peak at mev / gev , hbls show a @xmath2-ray component peaking at higher energies , often above the _
fermi_-lat energy band @xcite .
@xcite proposed the existence of an anti - correlation between the blazar luminosity and @xmath7 ( the so - called blazar sequence ) , although there is not a general consensus on this point .
more recently @xcite extended this sequence into a more general `` blazar envelope '' .
the subclass of hbls , even though the least luminous among the other blazar subclasses , is the brightest one at vhe , and the majority of vhe blazars are indeed hbls ( see e.g. * ? ? ? * for a recent review ) .
+ the study of vhe blazars is complicated by two observational characteristics of blazars themselves : their broadband emission and their rapid variability
. a vhe detection on its own , sampling only a small part of the nonthermal continuum , does not provide sufficient information about the underlying blazar physics . rather , strictly simultaneous multiwavelength ( mwl ) campaigns are required to constrain the blazar sed .
the launch of the _ fermi _ satellite and its monitoring capabilities have had a large impact on blazar physics , providing for the first time long - term light curves of hundreds of blazars in @xmath2-rays , and assuring simultaneous observations of vhe blazars in the mev / gev energy band .
+ observations by iacts are limited by the high sensitivity of the camera photomultiplier tubes ( pmts ) , which in their standard configuration experience higher noise , higher current level , and accelerated aging when operated in bright moonlight conditions ( moon illumination @xmath8 of full moon ) .
this constraint particularly affects blazar studies , for example , limiting the organization of mwl campaigns , or prohibiting follow - up observations of a flare .
several experiments have successfully performed observations of cherenkov light under bright moonlight conditions @xcite . of the current generation of iacts , only _ magic _ and _ veritas _ ( see * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) observe under moderate moonlight ( @xmath9 moon illumination ) which significantly improves the duty cycle of the observatories .
_ fact _ , whose camera is composed of solid - state geiger - mode avalanche photodiodes ( also called silicon - photomultipliers ) , instead of pmts @xcite , is also capable of observing under moonlight . + in 2012 ,
the _ veritas _ collaboration began a new program of observing under bright moonlight , by applying reduced high voltage ( rhv ) to lower the pmt gain , or by utilizing uv - transparent filters .
the schott glass ug-11 filters bandpass is from @xmath10 nm to @xmath11 nm , reducing moonlight by a factor of ten and cherenkov radiation by a factor of three .
the details of the observing strategy have been presented at several conferences @xcite and will be discussed in an upcoming _ veritas _ publication ( 2015 , in preparation ) . in this paper , we concentrate on the capabilities of the _ veritas _ telescope array in the rhv configuration
. we also present the detection , during may 2013 , of vhe emission from the blazar _
1es 1727 + 502 _ at a flux of roughly five times the archival vhe flux measured by _ magic _
@xcite . the high - flux state was initially detected during bright moonlight observations , which represents an innovation for _ veritas_. + the blazar _
1es 1727 + 502 _ ( from the _ einstein _ slew survey catalog , * ? ? ?
* ) , also known as _
i zw 187 _ or _ ot 546 _ , is a nearby ( @xmath12 , * ? ? ?
* ) hbl , discovered as a @xmath2-ray source by _ fermi _ @xcite , and as a vhe source by _ magic _ @xcite .
radio observations show a compact core - jet morphology @xcite , typical of blazars , with an apparent jet opening angle of 11@xmath13 @xcite . past optical monitoring
shows only moderate variability @xcite and the presence of a weak 300 nm bump @xcite , quite unusual in bl lac objects .
prior to being detected by _ magic _ as a vhe source , _
1es 1727 + 502 _ was observed by both
_ hegra _ @xcite and the _ whipple _ 10-m telescope @xcite , with no detection .
( @xmath14 ) crab nebula units above @xmath15 ( @xmath16 ) gev was measured by _ whipple _ ( _ hegra _ ) . ]
the _ magic _
collaboration reported an integral flux of @xmath17 @xmath18 s@xmath19 above @xmath20 gev ( corresponding to @xmath21 of the crab nebula flux ) and a spectral index of @xmath22 ( see * ? ? ? * ) .
+ this paper is organized as follows : in section [ section2 ] we describe the capabilities of the _ veritas _ instrument in the rhv configuration ; in section [ section3 ] we describe the _ veritas
_ observations of _ 1es 1727 + 502 _ , in both rhv and standard configurations ; the details of the mwl campaign triggered by the detection of the high - flux state are reported in section [ section4 ] ; in section [ section5 ] we model the sed using a standard one - zone synchrotron - self - compton model , and in section [ section6 ] we discuss the results obtained from the mwl campaign .
_ veritas _ is an array of four imaging atmospheric cherenkov telescopes located at the whipple observatory in southern arizona ( 3140n , 11057w ) at an altitude of 1.3 km above sea level .
each telescope is of davies - cotton design @xcite with a 12-m diameter reflector , and the array is arranged in a diamond configuration with @xmath23 100 m to a side .
each reflector comprises 345 identical hexagonal mirror facets and has a collection area of 110 m@xmath24 .
+ each _ veritas _ telescope is instrumented with a camera made up of 499 pmts each with a 0.15 field of view ( fov ) , for a total fov diameter of 3.5 .
the pmt signals are digitized by 500 megasamples - per - second flash analog - to - digital converters ( fadcs ) .
_ veritas _ employs a 3-level trigger system @xcite and has an array trigger rate of @xmath23 450 hz .
+ _ veritas _ has an energy resolution of 15% , and a single - event angular resolution of 0.1 at 1 tev . during standard observations ( i.e. when
the moon is @xmath9 illuminated ) a source with an integrated flux of 1% of the crab nebula flux can be detected at the @xmath25 level in @xmath23 25 hours , and a 5% crab source in less than 2 hours .
more information on the _ veritas _ array can be found in @xcite and @xcite .
+ in the rhv observation mode , the pmt voltages are reduced to 81% of their standard values during dark - sky observations , allowing _
veritas _ to operate when the moon is 35 - 65% illuminated .
this reduces the absolute gain of each pmt by a factor of @xmath26 .
observing in rhv mode increases the yearly exposure of _ veritas _ by @xmath27 above 250 gev . for rhv data , systematic errors induced by uncertainty in the atmosphere , telescope optical point spread function , and mirror reflectivity are the same as those for standard data . in terms of sensitivity , a standard analysis of _ veritas _
observations of the crab nebula taken under dark skies at small zenith angles yields a sensitivity of @xmath28 and an analysis energy threshold of @xmath29 170 gev . in comparison ,
crab data taken in rhv mode under moonlight also have a sensitivity @xmath28 albeit with a higher energy threshold ( @xmath23 200 gev ) .
these values assume a crab - like spectral index .
note that _ veritas _
event - selection cuts are optimized for sensitivity and not energy threshold .
the cuts used in this rhv analysis would allow for a 5% crab source to be detected in less than 2 hours .
the _ veritas _ observations that allowed for the detection of _
1es 1727 + 502 _ were taken between may 1 , 2013 ( mjd 56413 ) and may 7 , 2013 ( mjd 56419 ) .
there were additional data taken on may 18 , 2013 ( mjd 56430 ) that did not result in a detection .
after quality selection , approximately 6 hours of data remain . of these , 3 hours were taken in rhv mode on the first two nights of the exposure .
all observations were made in `` wobble '' mode , wherein the telescopes were pointed 0.5 away from the target to allow simultaneous measurements of the target and background regions @xcite , and all data in the 2013 dataset were taken at elevations between 65@xmath13 and 70@xmath13 . +
two sets of gamma - hadron separation cuts , each optimized _ a priori _ on data taken on the crab nebula where the gamma - ray excess has been scaled down to 5% , were used in the analysis of the 2013 dataset . for the standard - voltage subset ,
only events with images in three or more telescopes were used , whereas in the rhv data subset two - telescope events were used and a smaller cut was made on the brightness of the shower images to account for the fact that the camera gain was reduced .
the motivation for using these two sets of cuts is to approximately match the energy threshold of the analyses , resulting in a threshold of @xmath23 220 gev for both the rhv and standard voltage datasets .
+ the analysis of the complete 2013 dataset used the `` reflected - region '' background model resulting in 159 on events and 850 off events with a background normalization factor @xmath30 of 0.077 , yielding a detection significance of @xmath31 ( * ? ? ?
17 ) and an average gamma - ray - like event rate of @xmath32 counts / minute with a background event rate of @xmath33 counts / minute . +
prior to the detection of _
1es 1727 + 502 _ , _ veritas
_ observed the object for 8.6 hours between 2007 and 2009 .
there was no detection in these data ; the upper limit ( 99% confidence level ) is @xmath34 ( @xmath35 crab ) above 350 gev assuming the spectral index from @xcite .
this is consistent with the blazar s low - flux state as detected by _ magic _ in @xcite . + [
cols="^,^,^,^,^,^,^ " , ] [ section5 ] the sed of _ 1es 1727 + 502 _ is shown in fig .
[ figsed ] .
simultaneous observations ( with _ swift_-xrt , _ swift_-uvot , and _ veritas _ ) were carried out only during mjd 56419 . given that none of the instruments involved in the campaign detected any significant variability , we make the assumption that the flux was constant during the observations , and we thus consider that the average spectral measurements from the different instruments are representative of the blazar emission .
it is possible that the emission was variable , but at a level below instrumental sensitivities .
flwo 48 " _ measurements also cover the infrared - to - optical part of the sed . in this case as well , given the absence of variability in the b filter between _
swift_-uvot and _ flwo 48 " _ observations
, we make the assumption that the _ flwo 48 " _
measurements can be considered as representative of the emission from _ 1es 1727 + 502 _ during the _ veritas _ observations . _
fermi_-lat upper limits calculated from observations performed during the _ veritas _ campaign ( but not strictly simultaneously ) are also included , as well as the average spectrum from the long - term analysis . the bow - tie from the _ magic _
detection and the archival mwl data from the ned ( nasa / ipac extragalactic database ) are also included .
+ in the context of the synchrotron - self - compton ( ssc ) model , the two components of the blazar sed are associated respectively with synchrotron emission from leptons ( e@xmath36 ) and inverse - compton scattering of the particles off synchrotron radiation from the same lepton population .
the emitting region is a spherical blob of plasma ( characterized by its radius @xmath37 ) in the relativistic jet , moving towards the observer with doppler factor @xmath38 , and filled with a tangled , homogeneous magnetic field @xmath39 .
the particle population is parametrized by a broken power - law function , and it carries six free parameters : the minimum , maximum , and break lorentz factors of the particles ( @xmath40 , @xmath41 , @xmath42 ) , the two indices ( @xmath43 and @xmath44 ) , and the normalization factor @xmath45 .
the minimum lorentz factor of the leptons can be fixed at a reasonably low value without affecting the modeling .
the remaining eight free parameters can be constrained by observations , as discussed , for example , in @xcite and @xcite .
+ given the lack of a simultaneous _
fermi_-lat detection , it is impossible to constrain the position and the luminosity of the inverse - compton peak during the high state , and a unique solution for the ssc model can not be provided . however , the synchrotron component is very well sampled , and can provide some constraints on the energy distribution of particles in the emitting region . at low energies , the subtraction of the host - galaxy contamination reveals the agn nonthermal continuum , which can be described by a power law from infrared to uv , with no sign of break .
a fit of the _ flwo 48 " _ and _ uvot _ data results in an index @xmath46 . in the case of synchrotron radiation ,
this index reflects directly the index of the underlying e@xmath36 population @xmath47 , which is thus equal to @xmath48 .
similarly , the index of the x - ray power law below the x - ray break corresponds to @xmath49 . in the following ,
@xmath43 has been fixed to @xmath50 and @xmath44 to @xmath51 , i.e. a spectral break of @xmath52 has been assumed .
+ in blazar physics , an additional observational constraint comes from the variability timescale @xmath53 and the causality argument : the emitting region size @xmath37 has to be smaller than @xmath54 .
however , for the case of _ 1es 1727 + 502 _ , there is no evidence of variability on short timescales in any of the light curves that can provide an estimation of @xmath53 and thus constrain the emitting region size . + as discussed above , the value of @xmath40 does not affect the modeling and it is therefore held fixed .
however , values lower than @xmath55 would overestimate archival radio measurements ( see figure [ figsed ] ) , even though , given their nonsimultaneity , these data should not be considered as a strong constraint .
an additional constraint on @xmath40 is provided by the _ fermi_-lat nondetection , and will be discussed in the next section . in the following
we study two different cases , for @xmath56 and @xmath57 .
the value of @xmath41 is constrained by the break observed in the x - ray spectrum of _
1es 1727 + 502_. in the following , we express it as a function of @xmath42 : @xmath58 .
the numerical factor corresponds to the ratio between the x - ray break and the synchrotron peak , which is estimated by extrapolating the uv and x - ray spectra . + once
the indices and maximum energy of the particle population have been fixed , the number of free parameters in the ssc model is thus five ( @xmath38 , @xmath39 , @xmath37 , @xmath42 , and @xmath45 ) , and , with only four observables ( the frequency and flux of the synchrotron peak , the vhe spectral index and the vhe flux at 620 gev : @xmath59 ; @xmath60 ; @xmath61 and @xmath62 ) , a unique solution can not be provided . however , assuming a reasonable value of the doppler factor , it is possible to study the parameter space of the remaining free parameters .
the best - fit solution is computed using the numerical algorithm described in @xcite .
the parameter space is systematically sampled , producing a set of ssc models and computing for each of them the expected observable values .
a fit is then performed in order to express each observable as a function of the free parameters , defining a set of equations which is solved for the particular set of observables of _
1es 1727 + 502_. the uncertainty in the observables is taken into account by iteratively solving the system of equations , producing a family of solutions and determining the allowed range for each free parameter .
+ for the particular case of _
1es 1727 + 502 _ , the doppler factor has been fixed to @xmath63 ( a value in line with standard ssc modeling of vhe blazars , see e.g. * ? ? ?
* ; * ? ? ?
* ) , while the other free parameters have been studied in the range : @xmath64 $ ] mg ; @xmath65 $ ] cm ; @xmath66 $ ] @xmath67 ; @xmath68 $ ] .
@xmath69 is defined as the particle density at @xmath42 : @xmath70 .
the solutions of the ssc model are computed iterating on the four observables : @xmath71 $ ] ; @xmath72 $ ] ; @xmath73 $ ] and @xmath74 $ ] . for the two
_ veritas _
observables , the systematic uncertainty has been taken into account and summed in quadrature with the statistical error .
the system of equations is completed by an inequality relating the variability timescale to @xmath37 and @xmath38 .
we considered @xmath75 days , which corresponds roughly to the interval between the last _ veritas _
detection on mjd 56419 and the observation during mjd 56430 , which indicated that the vhe flare may have ended .
all the ssc models that correctly describe the _ 1es 1727 + 502 _ sed are then recomputed , and plotted in fig .
[ figsed ] .
the parameter values are listed in table [ table1 ] , together with derived quantities such as the energy densities , the luminosity of the emitting region and the minimum variability timescale . for each solution ,
the @xmath76 with respect to the observational data is calculated in order to find the solution that minimizes the @xmath76 .
the _ veritas _ detection of vhe @xmath2-ray emission from _ 1es 1727 + 502 _ triggered mwl observations that allowed , for the first time , the study of the sed during a high @xmath2-ray flux state , a factor of five higher than the _ magic _ detection . +
even though the lack of simultaneous detection by _
fermi_-lat does not allow the parameter space to be fully constrained , the available data permit some important conclusions to be drawn regarding the properties of the particle population and the acceleration / cooling mechanisms . + in the presence of synchrotron radiation , a break in the stationary particle population is expected , and it is characterized by @xmath77 ( see e.g. * ? ? ?
deviations from @xmath77 can , however , occur if the emitting region is inhomogeneous , or if the emission from escaped particles is taken into account ( see e.g. * ? ? ?
* ; * ? ? ?
interestingly , the measurements of the synchrotron component of _ 1es 1727 + 502 _ are consistent with a simple break of 1.0 , indicating that a single particle population , injected with a power - law distribution with index equal to @xmath78 , could explain the blazar emission
. + the break energy corresponds to the equality of the escape ( or adiabatic ) timescale and the synchrotron and inverse - compton cooling timescales .
the former is energy independent ( if the escape is advective , and not diffusive ) , and equal to @xmath79 , while the cooling timescales are both proportional to @xmath80 . following ( * ? ? ?
* equation 30 ) , it is possible to relate @xmath42 to the model parameters and find the only unknown variable , @xmath81 $ ] .
this means that the stationary particle population is consistent with an injected power - law particle population cooled by synchrotron and inverse - compton emission only if the advective escape timescale is @xmath82 .
this value is low compared to the ones used by @xcite ( @xmath83 $ ] ) .
for example , assuming @xmath84 , the break energy would have been at @xmath85 , two orders of magnitude higher than the constraint provided by quasi - simultaneous observations .
alternatively , the hypothesis of a power - law population injected and cooled by synchrotron and inverse - compton processes could be too simplistic , and additional injection / cooling / escape terms could contribute to the final stationary particle population .
+ it is important to recall that these considerations are valid only for a given value of @xmath86 .
lower values of @xmath38 would imply higher values of @xmath39 and then higher values of @xmath87 .
for example , assuming @xmath88 , the overall normalization is reduced by a factor of @xmath89 and can be compensated by assuming a magnetic field four times stronger , @xmath90 mg . in order to maintain the same @xmath59 ( which is @xmath91 ) , @xmath42 has to be reduced by a factor of @xmath92 .
these transformations would affect the inverse - compton peak as well , and it is not guaranteed that these parameters could reproduce the vhe spectral index .
regardless , using these new values , @xmath93 , one order of magnitude higher than that computed for @xmath86 .
+ it is possible to extract information on the acceleration mechanism from the value of the index of the particle population .
acceleration by diffusive shocks naturally produces power - law particle populations , with an index close to 2.0 . for ultrarelativistic shocks , several authors ( see * ? ? ? * and references therein ) have shown that @xmath43 is expected to be around 2.2 .
the quasi - simultaneous multiwavelength ( from infrared to uv ) observations of _ 1es 1727 + 502 _ show that @xmath43 has to be between @xmath94 and @xmath95 , fully consistent with ultrarelativistic shock acceleration .
+ the maximum particle energy is another important model parameter related to the acceleration .
the break observed in x - rays at @xmath96 kev can indeed be related to the high - energy cutoff of the particle population ( @xmath41 ) , and can thus be used to constrain the maximum energy of the leptons . in the framework of the standard diffusive shock acceleration ,
the maximum particle energy can be computed assuming that the acceleration timescale equals that for the radiative losses : particles with energy higher than @xmath41 lose energy faster than they can be accelerated .
assuming @xmath97 , where @xmath98 represents a parameter characteristic of the shock acceleration , it is easy to estimate @xmath99 , which is consistent with the constrained values of @xmath41 only for @xmath98 of the order of @xmath100 .
such a high value , which implies rather inefficient acceleration , is typical of high - frequency - peaked bl lac objects ( see e.g. * ? ? ?
* ) . in this case ,
solutions computed for lower values of @xmath38 can not significantly lower the value of @xmath98 .
it is easy to show that , using the same values estimated in the discussion on @xmath42 , @xmath98 would be reduced only by a factor of two .
+ the minimum particle energy @xmath40 also carries important information on the physics of the emitting region . both archival radio measurements and the nondetection by _
fermi_-lat suggest that @xmath40 has to be at least of the order of @xmath55 .
the study of the models for two different values of @xmath40 ( equal to @xmath101 and @xmath57 ) shows that solutions with lower values of @xmath40 have higher emission in the 0.1 - 1 gev energy band .
a @xmath40 lower than @xmath55 would start conflicting with the _ fermi_-lat nondetection ( under the assumption that the _ veritas _ flux is representative of the average flux throughout the whole period used to compute the _
fermi_-lat upper limits ) .
such a high value of @xmath40 is not a surprise for hbls ( see * ? ? ?
. it can be explained , for example , by assuming that the particles are already injected with a truncated - power - law distribution , and they do not have time to cool down completely , or that the particles are somehow reaccelerated inside the blazar region @xcite .
+ it is also interesting to note that the one - zone ssc modeling suggests that the blazar gev spectrum during the high - flux state was harder compared to the average spectrum measured by _
fermi_-lat . a harder - when - brighter correlation in the _
fermi_-lat band is not common for hbls ( see for example * ? ? ?
* ; * ? ? ?
* ) , and is more similar , for example , to what is observed during the @xmath2-ray flaring activity of bl lacertae @xcite .
+ one of the open questions in blazar physics is related to the energy budget of the emitting region .
is it close to equipartition between particle , photon , and magnetic energy densities ( @xmath102 , @xmath103 and @xmath104 ) ?
the solutions estimated here are all characterized by a very high value of @xmath105 , which is of the order of @xmath106 . on the other hand ,
the photon energy density ( synchrotron emission by primary leptons ) is of the same order as the magnetic one .
equipartition is sometimes useful to constrain the parameter space of blazar models , especially when dealing with external photon fields , which increase the number of free parameters .
recently @xcite presented a general approach to compute blazar models close to equipartition , showing that , within this framework , it is possible to describe the sed of blazars , and in particular fsrqs , correctly fitting the spectral break observed by _
fermi_-lat in several low - frequency - peaked blazars @xcite . on the other hand ,
it is not clear if a general equipartition approach still holds for hbls , and indeed several modeling attempts ( within a standard one - zone ssc scenario ) on the most well studied hbls result in equipartition factors far from unity ( see * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
+ it is important to emphasize that the sed of _
1es 1727 + 502 _ represents a snapshot of a high - flux state , and that the ratio between the flaring and the archival fluxes is higher in the inverse - compton component than in the synchrotron one .
it is thus possible that the emitting region is only temporarily far from equipartition , and that the low state is indeed characterized by @xmath105 closer to one . to test this hypothesis ,
the constraining algorithm has been rerun assuming @xmath62 lower by a factor of five ( i.e. consistent with the _ magic _
detection ) . as expected , the equipartition factor is lowered by one order of magnitude , lying between @xmath107 and @xmath108 .
again , solutions computed for lower values of @xmath38 , implying a higher value of @xmath39 , would result in @xmath105 much closer to equipartition .
in addition , since the particle index @xmath43 is softer than @xmath109 , as constrained by optical / uv observations , the value of @xmath102 is strongly dependent on @xmath40 .
given that we can not really constrain the value of @xmath40 , it is possible that the equipartition factor is indeed closer to unity .
in this paper we presented the results of a multiwavelength campaign carried out on the bl lac object _
1es 1727 + 502 _ during may 2013 , triggered by a vhe high - flux state detected by _
veritas_. this represents the first detection of a blazar flare with _ veritas _ during bright moonlight observations .
+ within this campaign , no significant variability is detected at vhe , nor at lower energies ( x - rays and optical ) .
the _ veritas _ light curve is consistent with a constant flux ( @xmath110 crab above 250 gev , roughly five times the archival _ magic _ detection ) between may 01 , 2013 and may 07 , 2013 ; additional observations during may 18 , 2013 indicate ( at a @xmath111 level ) that the high - flux state may have ended at some point after the last _ veritas _
detection on may 07 .
+ the quasi - simultaneous sed is fitted by a standard one - zone ssc model . even though the nondetection by _
fermi_-lat did not enable a full study of the parameter space , the measurements are fully consistent with particle acceleration by relativistic diffusive shocks and simple synchrotron and inverse compton cooling , resulting in a power law with injection index around 2.2 and a spectral break of 1.0 .
+ in recent years the number of vhe blazars has significantly increased thanks to the current generation of cherenkov telescope arrays such as _ veritas _ , _ magic _ and _ h.e.s.s._. the broadband emission and the rapid variability of blazars require prompt simultaneous multiwavelength campaigns to fully understand the physics of their emitting region .
the new _ veritas _ observing strategy under bright moonlight will be particularly useful for blazar science , significantly increasing blazar monitoring capabilities at vhe .
+ combining data sets taken under various observing conditions is not a new aspect of vhe gamma - ray data analysis . in the case of
_ veritas _ , lookup tables for the analysis are generated from simulations which already assume a wide range of different zenith angles , azimuth angles ( and hence geomagnetic field strengths ) , sky brightness and local atmospheric conditions .
the introduction of a new observing mode for _ veritas _ is therefore not expected to adversely affect the scientific interpretation of the results .
exploring new modes is certainly worthwhile , as the results of this paper demonstrate .
bright moonlight observing time is particularly useful for variable vhe sources , and can also be used to obtain deeper exposures on steady sources , especially for targets with hard vhe spectra which can be detected despite an increased energy threshold .
the merging of data taken under multiple observing modes and detector configurations will be an important analysis challenge for future large arrays , such as the cherenkov telescope array ( _ cta _ ) , particularly with regard to deep vhe exposures and extended surveys . | during moonlit nights , observations with ground - based cherenkov telescopes at very high energies ( vhe , @xmath0 gev ) are constrained since the photomultiplier tubes ( pmts ) in the telescope camera are extremely sensitive to the background moonlight .
observations with the _ veritas _ telescopes in the standard configuration are performed only with a moon illumination less than 35@xmath1 of full moon . since 2012 ,
the _ veritas _ collaboration has implemented a new observing mode under bright moonlight , by either reducing the voltage applied to the pmts ( reduced - high - voltage configuration , rhv ) , or by utilizing uv - transparent filters . while these operating modes result in lower sensitivity and increased energy thresholds ,
the extension of the available observing time is useful for monitoring variable sources such as blazars and sources requiring spectral measurements at the highest energies . in this paper
we report the detection of @xmath2-ray flaring activity from the bl lac object _
1es 1727 + 502 _ during rhv observations .
this detection represents the first evidence of vhe variability from this blazar .
the integral flux is @xmath3 above 250 gev , which is about five times higher than the low - flux state .
the detection triggered additional _
veritas _
observations during standard dark - time .
multiwavelength observations with the _ flwo 48 "
_ telescope , and the _ swift _ and _ fermi _ satellites are presented and used to produce the first spectral energy distribution ( sed ) of this object during @xmath2-ray flaring activity .
the sed is then fitted with a standard synchrotron - self - compton model , placing constraints on the properties of the emitting region and of the acceleration mechanism at the origin of the relativistic particle population in the jet . |
* relationship between qrp and renormalization transformation . * here we offer a new perspective on qrp . as shown in fig .
1(a)-(c ) , the operations of qrp exhibit a hierarchical characteristic and self - similarly nested structure . using standard box - counting method , it s easy to compute fractal dimension @xmath1 . historically kadanoff s seminal picture that spin blocks hierarchically nest in a self - similar manner basically stimulated wilson s renormalization group theory , which is renowned as a powerful tool to the problem of phase transition @xcite .
here we point out that the implementation of qrp with nested structure actually corresponds to renormalization process in network setting . as shown in fig .
1(a)-(c ) , the quantum channel between nodes @xmath2 and @xmath3 is divided into @xmath4 segments , and every @xmath5 consecutive segments are grouped together into a unit which is extended to larger length - scale .
this procedure is repeated with @xmath6 nesting levels , where @xmath7 @xcite .
if we interpret @xmath5 as a parameter which is none other than the transforming length - scale , then above procedure can be viewed as real space renormalization transformation .
guided by this remarkable idea , we present an universal framework which allows one to apply qrp to arbitrary qrn with fractal structure .
we find clear correspondences between one - dimensional and high dimensional quantum repeaters .
undoubtedly , the exponential decay of fidelity imposes strong constraint on the way nodes are interconnected . basically nodes are interconnected in accordance with local attachment :
a node prefers to link to neighboring nodes via intermediate repeater nodes , rather than distant nodes .
this connection fashion gives rise to fractal structure @xcite .
then fractal qrn can be substituted for the 1d linear network . and the length of 1d chain is replaced by the diameter of underneath fractal qrn @xmath8 namely , largest distance between nodes ( in graphic sense , the distance between two nodes is defined as the number of links along the shortest path @xcite ) . compared with segmentation fashion of 1d chain , the entire network is divided into boxes , whose size is equal to @xmath5 , see fig .
. it will be desirable if the nesting level takes a similar form : @xmath9 we will prove that it is indeed the case .
it answers an intuitive problem : how many nesting levels @xmath10 are required for quantum networks with @xmath4 nodes ? also we will show that it implies a structural transition where small - world is obtained . auxiliary segments , avoiding exponential attenuation of fidelity .
entanglements are repeatedly created for each segments .
( b ) and ( c ) nested purification that combines entanglement swapping and purification are successively performed in a hierarchical way .
adjacent segments are connected and extended to longer distance , eventually two remote nodes are connected via perfect entanglements .
bottom panel : schematic illustration of coupled renormalization .
( d ) quantum node ( circle ) is a composite system composed of ensembles of qubits ( dots ) .
a pair of quantum nodes is connected by multiple copies of bipartite entangled states ( solid lines ) . for simplicity ,
they are represented by single dashed lines in fig .
1(e ) and ( f ) .
( e ) nodes are assigned to different boxes according to the memb algorithm .
( f ) single level cr is performed.,title="fig:",width=328 ] + renormalization was successfully introduced into complex networks by song _
_ , uncovering the self - similarity of complex networks @xcite .
a network is renormalized according to the box - covering technique @xcite ( see fig .
1(e)-(f ) ) .
the basic idea is as follows : tile the entire network with minimum number of boxes @xmath11 , where the distance between nodes within any box is smaller than the box size , namely , the transforming length - scale @xmath12 .
each box is then replaced by a supernode .
these supernodes are connected if there is at least one link between nodes in their respective boxes .
it defines the fractal dimension @xmath13 in terms of a power law : @xmath14 apply this transformation @xmath15 to a fractal network @xmath16 , which is scale invariant , then we have @xmath17 . several algorithms @xcite have been proposed to coarse - grain complex networks , nevertheless , not all of them are useful here .
we choose the memb algorithm to divide the entire network into boxes @xcite .
it s a geometric algorithm which has the advantages of guaranteeing connectivity within boxes , isolating hubs of different boxes and avoiding overlap between boxes .
we then introduce some modifications . for clarity , this is rephrased in network language .
we select the hub of a box ( most connected node ) as representative node , playing the role of supernode .
if there is one link between two boxes , then add one link connecting their hubs .
otherwise , no links are attached . above presentation
instructs us to create shortcuts ( long - distance entangled state with high fidelity ) between which pair of representative nodes . as shown in fig .
1(f ) , the corresponding shortcuts denoted by pink links form a coarse - grained network ( cgn ) . as a result ,
the cgn is reconstructed and coupled to the initial network .
because of the self - similarity of underling network , the resulting cgn is topologically equivalent to the original one , and above renormalization processing can be iteratively applied to previous cgn with fixed @xmath5 , and so on until the critical nesting level ( see eq . ( [ nl ] ) ) .
notice that the requirement of local attachment is fulfilled throughout entire process . eventually , the superposition of each level cgn forms a new network , namely , the coupled network ( cn ) . for the sake of distinction between the standard and this modified renormalization
, we name the later coupled renormalization ( cr ) .
we have generalize quantum repeaters to arbitrary high but finite fractal dimension , in the sense that when @xmath1 , cr is automatically reduced to 1d quantum repeaters .
one possible application of cr for 1d linear network is shown in fig .
1(c ) . in 1d case ,
each representative node is merely connected to two representative nodes ( nodes below arrows ) . in regard to fractal qrn , however , a representative node ( red circles ) has probability @xmath18 to link @xmath19 representative nodes ( it depends on the degree distribution ) via the short path composed of entangled links between two boxes ( green links ) .
for instance , in fig .
1(f ) , hubs @xmath2 and @xmath3 are connected through a path @xmath3-@xmath20-@xmath21-@xmath2 , whose length is @xmath22 .
each of such a path corresponds to one unit of 1d case ( segments between two arrows in fig .
then shortcuts are established between representative nodes , with quantum operations identical to 1d case . in other words ,
its definite physical realization relies on which qrp is utilized .
entanglement swapping @xcite and purification @xcite are two most important quantum operations . by entanglement swapping , adjacent entangled links are connected and extended to two representative nodes . to obtain high fidelity entangled states ,
entanglement purification is required , which extracts nearly perfect entangled states from states with lower degree of entanglement .
alternatively , some qrp use quantum error correction @xcite .
this class of qrp could circumvent the probabilistic fashion of purification - based protocols , and relax the strong requirement of long - lived quantum memory @xcite .
thus they have potential to extend entanglement to longer distance and speed up communication rate .
* why small - world and scale - free properties are relevant for quantum networks . * despite the great diversity of real - world networks , they share some common features .
most of real - world networks are scale - free networks with small - world property @xcite .
a network is scale - free if the probability to find a node with @xmath23 links @xmath24 , follows a power law @xmath25 ( for real - world networks , @xmath26 ) .
this is quite different from poisson distribution for erds - rnyi random graphs @xcite .
small - world is an influential concept describing such a phenomenon : despite the large size of networks , on average , any two nodes are separated by relatively short distance , which typically scales as : @xmath27 @xcite .
these features play a dominant role on dynamic functions of complex networks .
one naturally wonders whether the two fundamental characteristics make a difference to quantum networks the realistic significance can be illustrated from the perspective of limited - path - length entanglement percolation @xcite . without quantum processing such as entanglement purification ,
the fidelity of communication along a long noisy path of imperfect entangled states decreases exponentially @xcite .
thus it yields a very short distance @xmath28 of faithful communication , which severely limits effective size of quantum networks @xmath29 .
things seem to be bad . however , if @xmath30 , it turns out that most of nodes are reachable within the distance of reliable communication . for @xmath31-dimensional regular lattice , @xmath32 , and we have @xmath33 .
in contrast , @xmath34 for small - world networks . we therefore need a quantum small - world network , so that without further quantum processing , this desirable topological effect alone suffices to effectively mitigate the limitation of noise , and enhance the scalability of quantum networks . , @xmath35 , @xmath22 .
( a ) and ( b ) distance distribution of cn with different nesting levels . ( a ) is log - log plot of @xmath36 versus @xmath28 , the slope of the upper line is @xmath37 ( analytical estimate ) , and the lower line 0.24 ( fitting ) .
( c ) prediction of diameter of single level cn as a function of @xmath5 .
( d ) log - log plot of the minimal and average diameter of different size of single level cn.,title="fig:",width=328 ] + the striking effect of small - world originates from the existence of shortcuts between remote nodes .
a quantum network without shortcuts is not small - world network .
consequently , it s rather difficult to extend the influential concept of small - world to large - scale quantum networks , where shortcuts are not directly available .
cr can generate a set of shortcuts , which eventually leads to small - world transition at a relatively small nesting level .
we will clarify it in subsequent section . while cr can be applied to fractal networks such as erds - rnyi random graphs at criticality @xcite , apparently ,
such topology is not a realistic option .
we propose that qrn are scale - free and fractal observed at arbitrary spacial scale .
we justify this proposal with twofold reasons .
entanglement percolation is a combination of entanglement swapping and classical entanglement percolation@xcite . in their framework
, the threshold actually depends on the final topology .
notably , the percolation threshold for scale - free networks can be vanishing @xcite .
it means that , as far as scale - free networks , classical entanglement percolation is such a good strategy that the critical amount of entanglement required for the presence of giant cluster is zero in asymptotic limit .
therefore , scale - free qrn has exceptional ability to preserve connectivity in the presence of noise .
in addition , it s strongly supported by three facts .
other than local operations , classical communication is indispensable between quantum nodes . and entangled photon pairs can be transmitted through commercial telecom fiber .
thus , to some extent , qrn is embedded in classical communication networks . whereas both phone call networks and internet are scale - free networks @xcite .
if qrn is scale - free , we can make full use of the existing network infrastructures , without significantly altering them .
above facts suggest that scale - free network is a plausible and eligible candidate for qrn . without loss of generality , we apply cr to a scale - free fractal network generated by the minimal model ( see methods ) .
let @xmath38 be the @xmath6th level cgn . according to renormalization group theory , @xmath39 .
so @xmath38 can be seen as larger - scale network enlarged from @xmath40 , or equivalently it arises from single level cr with transforming length - scale @xmath41 .
larger - scale cgn here collectively act as shortcuts of the underlying smaller - scale ones , which drastically change the topology in such a way that nodes are globally separated by short path of entangled links .
this can be further unveiled by the squeezed distance distribution which follows gaussian distribution ( see fig .
2(b ) ) . hence , in the end , a hierarchically nested quantum small - world network is produced .
* proof of fractal to small - world transition .
* we proceed to make it clear whether single or iterative cr will lead to fractal to small - world transition .
two analytical proofs with numerical simulations are provided .
a rigorous and reliable method is to observe the behavior of average degree under renormalization flow ( see methods ) @xcite . in regard to single level cr , the expected transition does not arise .
however , it s safe to say that iterative cr can give rise to fractal to small - world transition .
evidences for the transition displayed in fig .
3 conform above conclusion .
nonetheless , the average diameter of cn , namely , the signature of small - world is unclear .
we begin with analyzing the impact of single level cr , and then generalize it to iterative case . plugging eq .
( [ fd ] ) into eq .
( [ d0 ] ) , we immediately obtain the diameter of cgn @xmath42 . in order to compute @xmath43 , the diameter of cn
, we suggest a hierarchical routing method which exploits the hierarchical structure and convert it into a routing problem .
a path connecting two remote nodes is divided into two parts : one part links one of the nodes and its corresponding hub within the box , and the other part , consisting of only representative nodes , connects the two hubs .
the total length of the first part is approximately @xmath44 , and that of the second is @xmath45 .
we thus have@xmath46 see fig
take note that there is an optimal transforming length - scale , @xmath47 , which yields minimal diameter @xmath48 .
meanwhile , numerical simulation gives the corresponding minimal average diameter @xmath49 ( see fig .
an analytical approximation shows that @xmath50 scales as a power - law too ( see supplemental information ) .
the behavior of @xmath51 suggests that single level cr is unable to trigger the transition , which is consistent with above conclusion . ) with parameters @xmath22 , and @xmath52 .
the analytical prediction ( straight line ) matches well with numerical simulation ( square ) .
( b ) log - log plot of @xmath53 versus @xmath54.,title="fig:",width=328 ] + now let s consider multi - level cr with fixed box size . in analogy with above results , it s easy to obtain the diameter of iterative cn @xmath55 for small @xmath5 , using eq .
( [ dc ] ) with recursive derivation , we find @xmath56 where @xmath6 is nesting level of cr .
taking into account finite size effect , eq .
( [ i d ] ) holds on condition that @xmath57 .
remarkably , the first term decays exponentially , whereas the second increases linearly .
hence , @xmath55 is governed by the linear term and grows slowly .
we readily obtain the criterion for the transition : @xmath58 implying that @xmath59 for large @xmath4 .
it s desirable that both @xmath60 and @xmath61 increases logarithmically with size of network , since @xmath62 and @xmath63 how to appreciate the implication of @xmath60 now is evident .
we identify @xmath60 as critical nesting level , at which small - world is achieved .
direct evidence is shown by eq .
( [ cd ] ) .
what s more , it s exactly in agreement with simulation result , see fig .
here @xmath60 is inversely proportional to fractal dimension @xmath13 , indicating its topological interdependence .
when @xmath1 , eq .
( [ nl ] ) reproduces the result of 1d case . while @xmath64 , @xmath65 , means that the initial network is already small - world , perfectly consistent with conclusion that when @xmath64 , these networks are small - world without fractality @xcite .
the highlights of our scenario are as follows .
our scenario is a fairly general framework .
cr is compatible with various qrp based on the aforementioned principle . in principle
, cr is applicable to arbitrary quantum networks with fractal structure , not restricted to the example of scale - free networks . and
the transition will arise as long as eq .
( [ cond ] ) or eq .
( [ nl ] ) is satisfied .
it s not difficult to check that the unique requirement of the whole derivations is the fractality of qrn .
furthermore , the collective distribution of shortcuts across entire network is mapped to renormalization transformation , where the self - similar fashion of operations is preserved at network level , while the scale - free fractal structure is kept at all length - scales .
each level transformation enlarges underling network into larger - scale qrn .
the simultaneous logarithmical scaling of critical nesting level and corresponding diameter suggests that to achieve small - world , cr is operable even for qrn of large size .
moreover , thanks to the scale - free nature , cn is particularly resilient to random failure of quantum nodes . to summarize , we have generalized one - dimensional quantum repeaters to high fractal dimension by introducing an approach called cr , which relates entanglement distribution over arbitrary fractal qrn to recursive renormalization transformations .
we assume that qrn is fractal and scale - free , which is the case for a large number of real - world networks . in spite of the large size of qrn ,
there exists a relatively small critical nesting level for cr , at which small - world is obtained .
small - world seems to be a necessary element for a scalable quantum network .
our study suggests that concepts and tools from statistical physics will play an important role in the joint study .
it has conceived another significant direction which may open new avenue to address the outstanding issue of complexity .
that is , quantum simulation of dynamic process on complex networks or design of quantum algorithms for sophisticated questions in network science @xcite .
all of these attempts may dramatically alter the landscape of both fields .
* minimal model for scale - free networks . *
the growth of the network is actually the inverse procedure of renormalization @xcite .
we begin from a triangle at time step @xmath66 .
\(i ) at time step @xmath67 , @xmath68 new nodes are connected to the endpoints of each link @xmath69 generated at time step @xmath70 .
\(ii ) then , with probability @xmath71 , we remove link @xmath69 of time step @xmath70 , and add one new link connecting a new pair of nodes attached to the endpoints of link @xmath69 .
\(iii ) repeat ( i ) and ( ii ) recursively until the wanted time step .
this model produces a scale - free network with degree distribution exponent @xmath72 .
we are particularly interested in two distinct types of networks with @xmath73 or @xmath74 , where a pure fractal network with fractal dimension @xmath75 , and a nonfractal , small - world network ( sw ) are achieved respectively . for simplicity and without loss of university , here let @xmath76 , then @xmath77 .
* renormalization group method for proof of fractal to small - world transition .
* literature @xcite studied networks constructed by randomly adding links to a fractal network with probability @xmath78 .
let @xmath79 , @xmath80 and @xmath81 be the average degree of the initial , the new and renormalized network respectively .
then @xmath82 where @xmath83 .
let @xmath84 . if @xmath85 , @xmath86 , otherwise @xmath87 .
notice that when @xmath88 , @xmath89 , a stable phase corresponding to fractal network is separated from the unstable phase moving toward complete graph , where small - world is achieved . with purpose of finding the location of cn in the phase diagram , we have to calculate the exponent @xmath90 . in our model , on one hand , @xmath91 so @xmath92 , and @xmath93 , @xmath94 .
on the other hand , numerical simulation shows that @xmath95 ( see fig .
the apparent deviation is mainly caused by the rough approximation that @xmath96 follows a power law ( see fig .
2(a ) ) . in spite of the deviation , it s definite that @xmath97
. thus , cn belongs to the unstable phase and multi - level cr can give rise to fractal to small - world transition . for single level cr
, we have @xmath98 when @xmath99 , the links we added will not emerge in the renormalized network with length - scale @xmath12 @xcite .
we only need to consider one case @xmath100 , which diverges in the large size limit .
so @xmath101 , and @xmath102 .
thus , coinciding with prediction of equation @xmath103 , single level cn stays in the stable phase , the expected transition does not occur .
we thank chaoming song and liang jiang for helpful suggestions .
this work is funded by the national natural science foundation of china ( grant nos .
11275186 , 10975126 , 91024026 , 11205040 ) , the major important project fund for anhui university nature science research ( grant no .
kj2011zd07 ) and the specialized research fund for the doctoral program of higher education of china ( grant no .
20093402110032 ) .
z .- w.w . conceived the research , performed analytical and numerical calculations , and wrote the manuscript .
p.h . discussed the manuscript and prepared the figures .
b .- h.w . supervised the research .
* competing financial interests : * the authors declare no competing financial interests .
99 horodecki , r. , horodecki , p. , horodecki , m. , horodecki , k. quantum entanglement .
phys . _ * 81 * , 865 ( 2009 ) .
pan , j .- w .
_ et al . _
multiphoton entanglement and interferometry .
phys . _ * 84 * , 777 ( 2012 ) .
yuan , z .- s .
_ et al . _
entangled photons and quantum communication .
rep . _ * 497 * , 1 ( 2010 ) .
kimble , h. j. the quantum internet .
_ nature _ * 453 * , 1023 ( 2008 ) .
ritter , s. _ et al .
_ an elementary quantum network of single atoms in optical cavities
. _ nature _ * 484 * , 195 ( 2012 ) .
albert , r. barabsi , a .-
statistical mechanics of complex networks . _ rev .
phys . _ * 74 * , 47 ( 2002 ) .
boccaletti , s. _ et al .
_ complex networks : structure and dynamics .
rep . _ * 424 * , 175 ( 2006 ) barabsi , a .-
l . the network takeover . _ nature phys .
_ * 8 * , 14 ( 2012 ) .
acn , a. , cirac , j. i. , lewenstein , m. entanglement percolation in quantum networks .
_ nature phys . _ * 3 * , 256 ( 2007 ) .
cuquet , m. , calsamiglia , j. entanglement percolation in quantum complex networks .
lett . _ * 103 * , 240503 ( 2009 ) .
cuquet , m. , calsamiglia , j. limited - path - length entanglement percolation in quantum complex networks .
rev . a _ * 83 * , 032319 ( 2011 ) . perseguers , s. , lewenstein , m. , acn , a. , cirac , j. i. quantum random networks .
_ nature phys . _
* 6 * , 539 ( 2010 ) .
garnerone , s. , giorda , p. , zanardi , p. bipartite quantum states and random complex networks .
_ new j. phys . _
* 14 * , 013011 ( 2012 ) .
ionicioiu , r. , spiller , t. p. encoding graphs into quantum states : an axiomatic approach .
rev . a _ * 85 * , 062313 ( 2012 ) . sangouard , n. , simon , c. , riedmatten , h. de , gisin , n. quantum repeaters based on atomic ensembles and linear optics . _
mod . phys.__**83 * * , 33(2011 ) .
briegel , h .- j . ,
dr , w. , cirac , j. i. , zoller , p. quantum repeaters : the role of imperfect local operations in quantum communication .
lett . _ * 81 * , 5932 ( 1998 ) .
duan , l .- m . ,
lukin , m. d. , cirac , j. i. , zoller , p. long - distance quantum communication with atomic ensembles and linear optics .
_ nature _ , * 414 * , 413 ( 2001 ) .
munro , w. j. _ et al .
_ from quantum multiplexing to high - performance quantum networking .
_ nature photon . _ * 4 * , 792 ( 2010 ) .
l. jiang , j. m. taylor , k. nemoto , w. j. munro , r. van meter , and m. d. lukin , quantum repeater with encoding .
a _ * 79 * , 032325 ( 2009 ) .
nadja k. bernardes , and peter van loock . hybrid quantum repeater with encoding .
a _ * 86 * , 052301 ( 2012 ) .
barabsi , a .-
albert , r. emergence of scaling in random networks .
_ science _ * 286 * , 509 ( 1999 ) . watts , d. j. , strogatz , s. h. collective dynamics of small - world networks .
_ nature _ * 393 * , 440 ( 1998 ) .
fisher , m. e. renormalization group theory : its basis and formulation in statistical physics . _
_ * 70 * , 653 ( 1998 ) .
wilson , k. g. the renormalization group : critical phenomena and the kondo problem . _ rev .
* 47 * , 773 ( 1975 ) .
song , c. , havlin , s. , makse , h. a. origins of fractality in the growth of complex networks .
_ nature phys . _
* 2 * , 275 ( 2006 ) .
song , c. , havlin , s. , makse , h. a. self - similarity of complex networks .
_ nature _ , * 433 * , 392 ( 2005 ) .
rozenfeld , h. d. , havlin , s. , ben - avraham , d. fractal and transfractal recursive scale - free nets._new j. phys . _ * 9 * , 175 ( 2007 ) .
kim , j. s. , goh , k .-
i . , kahng , b. , kim , d. a box - covering algorithm for fractal scaling in scale - free networks .
_ chaos _ , * 17 * , 026116 ( 2007 ) .
song , c. , gallos , l. k. , havlin , s. , makse , h. a. how to calculate the fractal dimension of a complex network : the box covering algorithm .
_ j. stat .
( 2007 ) , p03006 .
ukowski , m. , zeilinger , a. , horne , m. a. , ekert , a. k. event - ready detectors bell experiment via entanglement swapping .
* 71 * , 4287 ( 1993 ) .
dr , w. , briegel , h. j. entanglement purification and quantum error correction .
phys . _ * 70 * , 1381 ( 2007 ) .
pan , j .- w . ,
simon , s. , brukner , c. , zeilinger , a. entanglement purification for quantum communication .
_ nature _ * 410 * , 1067 ( 2001 ) .
bennett , c .
h. , brassard , g. , popescu , s. , schumacher , b. , smolin , j. a. , wootters , w. k. purification of noisy entanglement and faithful teleportation via noisy channels .
lett . _ * 76 * , 722 ( 1996 ) .
zhao , b. _ et al . _ a millisecond quantum memory for scalable quantum networks . _
nature phys .
_ * 5 * , 95 ( 2009 ) .
clausen , c. _ et al .
_ quantum storage of photonic entanglement in a crystal .
_ nature _ * 469 * , 508 ( 2011 ) .
n._et al . _ holographic storage of biphoton entanglement . _
lett . _ * 108 * , 210501 ( 2012 ) .
albert , r. , jeong , h. , barabsi , a .-
l . error and attack tolerance of complex networks .
_ nature _ * 406 * , 378 ( 2000 ) .
callaway , d. s. , newman , m. e. j. , strogatz , s. h. , watts , d. j. network robustness and fragility : percolation on random graphs .
* 85 * , 5468 ( 2000 ) .
vzquez , a. , moreno , y. resilience to damage of graphs with degree correlations .
e _ * 67 * , 015101 ( 2003 ) .
rozenfeld , h. d. , song , c. , makse , h. a. small - world to fractal transition in complex networks : a renormalization group approach .
lett . _ * 104 * , 025701 ( 2010 ) .
garnerone , s. , zanardi , p. , lidar , d. a. adiabatic quantum algorithm for search engine ranking .
lett . _ * 108 * , 230506 ( 2012 ) .
paparo , g. d. , martin - delgado , m. a. google in a quantum network .
* 2 * , 444 ( 2012 ) .
snchez - burillo , e. , duch , j. , gmez - gardees , j. , zueco , d. quantum navigation and ranking in complex networks .
* 2 * , 605 ( 2012 ) .
* calculation of average diameter for single cr . * for scale - free fractal networks which are not small - world , the average diameter @xmath104 @xcite and diameter @xmath105 are as follows : @xmath106 @xmath107 after renormalization , the size of the new network @xmath108 . and
the degree of a node @xmath109 is related with its original degree @xmath23 by a power law : @xmath110 , where @xmath111 @xcite .
combining eqs .
( [ ad ] ) and ( [ d0 ] ) gives @xmath112 , where @xmath113 .
so the average diameter of the renormalized network is @xmath114 , and within a box of size @xmath5 , it s @xmath115 .
accordingly , the average diameter for single cn approximately behaves as follows : @xmath116 once again , the optimal transforming length - scale @xmath47 , which yields minimal average diameter see fig .
and it is in good agrement with numerical results .
in fact , numerical simulation gives @xmath49 .
these can be further demonstrated distance distribution between nodes . as shown in fig .
4(b ) , in particular , with respect to the optimal transforming length - scales @xmath118 , it shrinks into gaussian distribution .
naturally , it s not surprising that @xmath119 .
although eq .
( [ lm ] ) is just an approximation of @xmath50 , it successfully predicts that single cr is unable to trigger fractal to small - world transition .
ref1 burda , z. , correia , j. d. , and krzywicki , a. statistical ensemble of scale - free random graphs .
, * 64 * , 046118 ( 2001 )
. song , c. , havlin , s. , makse , h. a. self - similarity of complex networks .
_ nature _ , * 433 * , 392 ( 2005 ) . | quantum networks provide access to exchange of quantum information .
the primary task of quantum networks is to distribute entanglement between remote nodes .
although quantum repeater protocol enables long distance entanglement distribution , it has been restricted to one - dimensional linear network . here
we develop a general framework that allows application of quantum repeater protocol to arbitrary quantum repeater networks with fractal structure .
entanglement distribution across such networks is mapped to renormalization .
furthermore , we demonstrate that logarithmical times of recursive such renormalization transformations can trigger fractal to small - world transition , where a scalable quantum small - world network is achieved .
our result provides new insight into quantum repeater theory towards realistic construction of large - scale quantum networks . by exploiting the probabilistic prediction nature of quantum mechanics and the nonlocal correlation of entanglement @xcite ,
new technology of quantum communication has been developed @xcite .
for example , quantum teleportation allows faithful teleportation of unknown quantum states , and quantum cryptography ( ekert91 protocol ) enables truly secure communication @xcite .
quantum nodes can store and manipulate photons locally , and their interconnection via quantum channel , e.g. , optical fiber , gives birth to quantum networks .
quantum networks are backbone of quantum communication and distributed quantum computing @xcite . a prototype of quantum network has been reported recently@xcite .
quantum networks can be seen as large and complex system of quantum states . how to characterize and understand such quantum system remains a challenge .
it is no longer effectively described by a global density operator @xmath0 @xcite .
however , complex networks which describe a wide range of natural and social system @xcite , provide conceptual basis for in - depth investigation of topological properties of quantum networks .
it involves exciting phenomena .
for instance , entanglement percolation @xcite and the peculiar behavior of quantum random networks @xcite have received much attention . what s more , it has opened new perspective for study of entanglements : map complex networks into entangled states , and vice versa @xcite . as essential ingredients for quantum information , however , entanglement is such fragile resource that suffer fatal photon loss , decoherence caused by noise , and imperfection of quantum local operations @xcite . in consequence ,
the state fidelity or degree of entanglement decreases exponentially with channel length .
quantum repeater protocols ( qrp ) are one of most promising solutions designed to tackle such challenging problem @xcite . the general principle of qrp is illustrated in fig .
1(a)-(c ) . in principle , qrp allow one to establish long - distance entanglement with fidelity close to unity , while the required time increases , e.g. , polynomially with channel length ( it depends on specific qrp ) .
thereby quantum repeaters hold promise for building large - scale quantum networks .
then a fundamental problem comes to us : what s the possible topology of quantum repeater network ( qrn ) , and how to perform qrp on such network ? in this work , we address this problem , with focus on the interplay between entanglement distribution and topology of quantum networks .
a practical scenario of entanglement distribution has to consider the complex topology of quantum networks .
recent studies suggest that the topology of quantum networks strongly affects their performance @xcite .
how well can we say about the topology of qrn ?
it s a problem that has not been seriously considered . motivated by the rich and intriguing topology of complex networks ,
we envisage the topology of qrn as follows .
the exponential decay of fidelity requires that qrn be fractal . and scale - free network is a plausible option for qrn @xcite .
moreover , as a generic characteristic of real - world networks , small - world effect @xcite is able to reinforce the scalability of quantum networks @xcite .
we combine these elements with qrn .
next , so far , qrp have been elucidated for one - dimensional linear network . in order to apply qrp to fractal qrn , we draw on concepts and methodology from statistical physics .
we find a clue to relate implementation of qrp to renormalization transformation . as a result
, entanglement distribution over arbitrary fractal qrn corresponds to a process of successive renormalization transformations . |
the lambert w function is defined as the multivalued inverse of the function @xmath0 .
its mathematical properties have been explored only quite recently after its implementation in the mathematical library of the computer algebra program _
maple_. the history , applications and properties of lambert @xmath1 are reviewed in corless _
et al_. @xcite mathematical developments and applications to physics ( mainly in quantum mechanics and electrostatics ) can be found in refs . .
since w is a very simple function , all its applications to physics or other fields are not exhausted , and , after a short enquiry , i have discovered that many of my colleagues , aware or unaware of its name , have met the function in their own works.@xcite here we reformulate some old and new problems of the statistical mechanics of classical liquids in terms of lambert w. our paper is organized as follows . in sec .
ii we commence by refreshing the reader ideas with some known mathematical properties of @xmath2 .
we focus on the principal branch @xmath3 of the lambert function and the related function @xmath4 , both being of prime importance for their applications to the problems of statistical mechanics considered in this article . some new mathematical properties of @xmath3 and @xmath5 used in the paper are presented in the appendix ( the legendre transforms of the real functions @xmath6 and @xmath7 are computed and the dispersion relations for @xmath3 and @xmath5 established ) . in sec .
iii we reformulate the properties of the tonks gas,@xcite i.e. a one - dimensional ( 1d ) classical fluid of hard rods , in terms of the principal branch @xmath3 of the lambert function .
our presentation is compared with the seminal work of hauge and hemmer on this system.@xcite the introduction of lambert @xmath3 allows to recover all the properties of the model in a unified treatment . in sec .
iv we focus on the second system considered by hauge and hemmer in ref . ,
i.e. a classical @xmath8 fluid made of particles with intermolecular pair repulsions of the form @xmath9 where @xmath10 is a positive integrable function .
the extension to arbitrary dimension is trivial . in the limit @xmath11 which will be considered , we deal with infinitely weak and infinitely long range repulsive interactions . once again
, the use of the functions @xmath3 and @xmath5 is of a great help for describing in a unified manner all the known properties of this model . sec .
v contains new material .
we consider the field theoretical representation of a fluid of particles interacting via repulsive potentials in the framework of the formalism developed in ref . .
it is shown that at the mean field ( mf ) level of the theory the expressions for the pressure and the density are identical to those derived in sec .
iv for the fluid with weak long range repulsive interactions . in other words ,
the mf approximation is exact in the case of infinitely long range interactions , a satisfactory , although expected result .
we conclude in sec .
in this section we summarize the main mathematical properties of the lambert @xmath2 function .
@xmath2 is defined as the root of @xmath12 where @xmath13 is a complex number .
@xmath2 is a multivalued function which has been studied recently in refs . . the different branches of lambert w ( i.e. the different possible solutions of eq
( [ w ] ) ) are labeled by an integer @xmath14 etc .
when @xmath13 is a real number eq .
( [ w ] ) can have either two real solutions for @xmath15 , in which case they are @xmath3 and @xmath16 , or it can have only one real solution for @xmath17 , this being @xmath3 while @xmath16 is now complex , or no real solutions for @xmath18 . @xmath3 and @xmath16 are the only branches of @xmath1 that take on real values . a brief survey of the properties of the principal branch @xmath3 will be sufficient for our purpose .
firstly , @xmath3 is analytic at @xmath19 .
this follows from the lagrange inversion theorem@xcite .
its power series about the point @xmath19 reads as @xmath20 the radius of convergence of the series ( [ wseries ] ) is equal to @xmath21 as it is easily shown using the cauchy test and , within the circle of convergence , @xmath22 . as a consequence of the relation ( [ w ] ) one has [ wprime ] @xmath23 where the radius of convergence of the series ( [ wprimeb ] ) is again equal to @xmath21 .
lagrange theorem gives more than eq .
( [ wseries ] ) ; it allows to obtain power series for @xmath24 , @xmath25 , @xmath26 , etc for @xmath13 within the circle of convergence of @xmath3 .
@xcite in particular , the power series of the function @xmath5 defined to be @xmath27 is given by @xmath28 note that if follows from eq .
( [ wprimea ] ) that @xmath29 @xmath3 has a second - order branch point at @xmath30 which it shares with both @xmath31 and @xmath16 and its branch cut is conveniently chosen to be @xmath32 with the convention that @xmath3 is defined on the upper lip of the branch cut .
the behavior of @xmath3 about the branch point is given by the series@xcite @xmath33 where @xmath34 ( the series converges for @xmath35 ) . in physics , it is quite common ( and sometimes enlighting ) to write dispersion relations for functions having a branch cut such that @xmath3 and @xmath5 .
@xcite we show in the appendix that , in the present case , these relations can be recast under the form [ dispersion ] @xmath36\ ; , \end{aligned}\ ] ] where @xmath13 is an arbitrary complex number which however is _ not _ on the branch cut .
these equations convey the interpretation of @xmath3 ( resp .
@xmath5 ) as the two - dimensional complex electric potential created by a distribution of charges @xmath37 ( resp .
@xmath38 ) located on the branch cut .
moreover , we show in the appendix that the distribution @xmath37 is normalized to unity ( cf eq .
( [ normaw ] ) ) while the total charge of the charge distribution @xmath38 is infinite .
other physical interpretations of the relations ( [ dispersion ] ) will be discussed in forthcoming developments .
in this section we consider the tonks gas , i.e. a 1d fluid of hard rods of length @xmath39.@xcite the equation of state ( eos ) is known exactly : @xmath40
( @xmath41 pressure , @xmath42 , @xmath43 temperature , @xmath44 number density ) .
of course @xmath45 . for @xmath46
one recovers the eos of the ideal gas , while for @xmath47 , which corresponds to the close packing of the hard rods , the pressure diverges .
less trivial is the expression of the pressure as a function of the activity @xmath13 . recall that @xmath48 where @xmath49 is the chemical potential and we have assumed that the debroglie thermal wavelength @xmath50 . starting from the thermodynamic relation @xmath51 it is not difficult to obtain the relation@xcite @xmath52 _ a priori _ eq .
( [ relatonks ] ) is valid only for real positive values of @xmath13 but it allows to define the pressure @xmath53 in the complex plane of the activities by analytic continuation .
the authors of ref . did not know the lambert function as we do ; clearly one has @xmath54 since the other branches @xmath55 of lambert w will not give a real pressure for real chemical potentials @xmath49 .
the mayer expansions for the pressure and the density of the tonks gas are therefore obtained by copying out eqs .
( [ wseries ] ) and([wprimeb ] ) : [ mayert ] @xmath56 the results of sec .
ii on lambert w enable us to conclude that 1 .
the radius of convergence of the mayer series ( [ mayert1 ] ) and([mayert2 ] ) is @xmath57 . within the circle of convergence @xmath58 .
the pressure @xmath53 is singular at the point @xmath59 of the circle of convergence , i.e. the branch point of @xmath60 , in agreement with one of the conclusions of the second theorem of groeneveld.@xcite 3 .
@xmath53 has a branch cut on the negative real axis : @xmath61 .
it can be identified with the distribution of zeros of the grand partition function in the thermodynamic limit.@xcite the distribution of the yang - lee zeros is given by @xmath62 ( cf eqs .
( [ dispersionw ] ) and ( [ dispersionw ] ) ) .
@xmath63 is a positive and increasing function over the interval @xmath64 which behaves as @xmath65 for @xmath66 and as @xmath67 for @xmath68 , @xmath69 ; therefore the integral of @xmath63 over the interval @xmath64 is convergent and equal to @xmath39 ( cf eq .
( [ normaw ] ) ) .
4 . for an arbitrary @xmath13
( however _ not _ on the branch cut ) the pressure @xmath53 may be seen as the @xmath70 complex electrostatic potential created by the charge distribution @xmath63 .
it follows from eqs .
( [ dispersionw ] ) and ( [ dispersionw ] ) that @xmath53 takes the form proposed by yang and lee@xcite @xmath71 a close examination of their paper reveals that the derivation of eq .
( [ disperstonks ] ) given by hauge and hemmer is restricted to a point @xmath13 inside the circle of convergence .
it is amusing to check the various conclusions of second theorem of groeneveld.@xcite recall that for a d - dimensional classical fluid of particles interacting via positive pair potentials @xmath72 such that ( twice ) the second virial coefficient @xmath73 converges , the radius of convergence @xmath74 of the virial series ( [ mayert1 ] ) and the mayer coefficients @xmath75 satisfy to the following inequalities [ groe ] @xmath76 in the present case @xmath77 , and one easily checks that @xmath57 and @xmath75 as given by eq .
( [ mayert3 ] ) , do satisfy to the groeneveld inequalities ( [ groe ] ) .
we end this section by determining the helmoltz free energy per unit volume @xmath78 as the legendre transform of @xmath79 viewed as a function of the reduced chemical potential @xmath80 ( see eg ref . ) .
it follows from the eq .
( [ legendref ] ) of the appendix that @xmath81 note that we have restored @xmath82 to make the argument of the @xmath83 dimensionless .
@xmath78 is a strictly convex function of the density defined for @xmath84 .
the first term in the rhs of the equation is the helmoltz free energy per unit volume of the ideal gas while the second one is the excess free energy .
the analytic continuation of @xmath78 to complex densities is obtained by defining @xmath78 according to eq .
( [ po ] ) where @xmath83 is the principal branch of the natural logarithm for @xmath78 must take on real values for @xmath85 , @xmath45 .
@xmath78 is then a multivalued function with two branch cuts @xmath86 and @xmath87 .
the eos of a gas of particles interacting via the pair potential ( [ weak ] ) is known exactly and given by @xcite @xmath88 where @xmath89 is a positive , @xmath90-independent constant . here
@xmath44 can take on all non negative real values .
it is not difficult to obtain the relation between the density and the activity which reads as @xcite @xmath91 therefore , since for a real @xmath13 the density must be real @xmath92 and , by integration of the thermodynamic relation ( [ thermo ] ) @xmath93 the pressure and the density for complex activities are obtained by an analytical continuations of eqs .
( [ lamb1 ] ) and ( [ lamb2 ] ) . the virial series for @xmath53 and @xmath94 follow from eqs .
( [ wseries ] ) and ( [ useries ] ) @xmath95 the radius of convergence of these two series is @xmath96 and it satisfies to the first groeneveld inequality ( [ groe1 ] ) .
one also checks that the coefficients of the power series ( [ toto ] ) do satisfy to the second groeneveld inequality ( [ groe2 ] ) .
@xmath53 has a branch cut on the negative real axis : @xmath61 .
for any complex activity @xmath13 _ not _ on the cut , it can be written once again under the form proposed by yang and lee @xcite ( cf eq .
( [ dispersionu ] ) ) @xmath97 where @xmath98 it follows from the analysis given in the appendix that the function @xmath63 is not integrable on the cut ( it behaves as @xmath99 for @xmath66 ) .
recall that , strictly speaking , the theory of yang and lee is valid only for interactions with a hard core contribution . in this case
the grand partition function @xmath100 for a finite volume @xmath101 is a polynomial in @xmath13 .
the distribution @xmath63 of its zeros after the passage to the thermodynamic limit allows to rewrite @xmath102 under the form ( [ dispersweak ] ) , where @xmath63 is normalizable .
@xcite for soft interactions as those considered in this section , the reasoning breaks down , leaving us with a non normalizable distribution @xmath63 ( one can pack an infinite number of particles in a finite volume and the density @xmath103 is not bounded ) .
finally the helmoltz free energy per unit volume @xmath78 is computed as the legendre transform of @xmath104 .
it follows from the eq .
( [ legendreh ] ) of the appendix that @xmath105 @xmath78 is a strictly convex function of @xmath44 on the interval @xmath106 .
we consider the statistical mechanics of a system made of @xmath107 classical point particles interacting via pair potentials of the form @xmath108 where @xmath109 is some reference potential ( eg a hard sphere repulsion for instance ) and @xmath110 admits a positive , well defined fourier transform @xmath111 .
@xmath109 and @xmath110 are supposed to meet all the requirements which are necessary for the existence of a thermodynamic limit ( tl ) .
@xcite we denote by @xmath112 the domain occupied by the molecules of the fluid
. it will be convenient to assume that @xmath112 is a cube of side @xmath113 with periodic boundary conditions ( pbc ) . in a given configuration @xmath114
the microscopic density of particles reads as @xmath115 and its fourier transform @xmath116 the configurational potential energy of the system can be decomposed as @xmath117 where @xmath118 is the configurational energy of the reference system , @xmath119 a self - energy contribution and @xmath120 we shall work in the grand canonical ( gc ) ensemble .
we denote by @xmath49 the chemical potential and by @xmath121 an external potential .
the local chemical potential will be noted @xmath122 . performing a kac - siegert - stratonovich - hubbard - edwards ( ksshe )
transform @xcite one can show that the gc partition function can be recast under the form of a functional integral@xcite @xmath123= { \cal n}^{-1 } \int { \cal d } \xi \exp(-{\cal h}[\xi ] ) \ : , \ ] ] where @xmath124 is the functional measure and the action @xmath125 $ ] of the ksshe field theory reads as @xmath126 = \frac{1}{2\beta } \left
< \xi \vert \varphi^{-1 } \vert \xi \right > - \log \xi_{0}\left[\overline{\nu } + i \xi \right ] \ ; , \ ] ] where @xmath127 and @xmath128 is the gc partition function of the reference system .
the field @xmath129 which enter eqs .
( [ xi ] ) and([h ] ) is a real scalar field .
note that , in the general case where the sign of @xmath130 is abitrary , the action @xmath131 involves _ two _ real scalar fields @xmath132 and @xmath133 ( or a complex field @xmath129 ) .
@xcite finally , the normalization constant @xmath134 is given by @xmath135 in ref .
the potential @xmath109 of the reference system was chosen to be a hard core potential of diameter @xmath39 . here
we specialize to the case @xmath136 ( or equivalently @xmath137 ) , i.e. we take for the reference system the ideal gas .
all the conclusions of ref . remain valid provided that the tl of our system is well behaved .
note that the configuration energy ( [ v ] ) can be rewritten as @xmath138 therefore if @xmath139 and @xmath140 ( which will be assumed henceforth ) then @xmath141 with @xmath142 , i.e. the system is h - stable in the sense of ruelle and the tl exists.@xcite note that , conversely , if @xmath143 , the system does not have a thermodynamic behavior and the introduction of a repulsive hard core is mandatory to ensure the existence of a tl . such potentials ( i.e. hard core plus an attractive tail ) can yield a liquid - vapor transition as explained in ref .
; in the case considered here ( soft repulsive tail ) , the possibility of such a transition has to be ruled out .
with the choice @xmath144 the ksshe action reads now @xmath145 = \frac{1}{2\beta } \left < \xi \vert \varphi^{-1 } \vert \xi \right > - \int_{\omega}d^3\vec{r } \ ; e^{\overline{\nu}(\vec{r } ) + i \xi(\vec{r})}\ ; .\ ] ] we turn now our attention to the mean field ( mf ) level of the theory .
the mf or saddle point approximation is defined by the equation @xmath146 where at @xmath147 the action @xmath131 is minimum .
if there are several local minima then one retains the absolute minimum .
the stationary condition @xmath148 can be recast under the form of the implicit integral equation @xmath149 moreover the mf density of the fluid is given by the density of the reference fluid -here the ideal gas- at the local chemical potential @xmath150@xcite , i.e. @xmath151 in the case of a homogeneous system to which we stick from now , ( i.e. @xmath152 ) the solution of eq .
( [ statio1 ] ) is clearly @xmath153 where @xmath154 is some branch of lambert w , @xmath13 the activity and @xmath155 from which it follows that @xmath156 it remains to determine the branch of w. for real activities @xmath13 , the mf density should be a positive real number ; it follows then from eq .
( [ rhomf2 ] ) that , necessarily , @xmath157 .
the mf pressure is easily derived from eq .
( [ ximf ] ) with the result @xmath158 note that the mf equation of state takes the familiar form @xmath159 .
finally , the mf free energy is obtained by a legendre transform of @xmath53 , which gives @xmath160 in agreement with the general expression of ref . .
some comments are in order . 1 .
the taylor series of @xmath161 and @xmath162 in terms of the activity @xmath13 are once again given by eqs.([wseries ] ) and ( [ useries ] ) with a radius of convergence equal to @xmath163 .
for an arbitrary complex activity @xmath13 , the pressure @xmath161 is given by the dispersion relation ( [ dispersweak ] ) .
we have shown elsewhere@xcite that the one - loop order approximation of the ksshe field theory coincides with the random phase approximation ( rpa ) of the theory of liquids.@xcite as well known , for long range potentials of the form ( [ weak ] ) the rpa corrections to the pressure vanishes as @xmath11.@xcite note that , in this limit , @xmath164 and @xmath165 ; thence , in the limit @xmath11 , one recovers for the pressure , the free energy etc the expressions derived in sec .
iv . stated otherwise , the mf - ksshe theory is exact for infinitely weak and long range repulsive potentials
we have shown in ref . that the mf pressure and free energy constitute _
exact _ bounds .
more precisely we have + @xmath166
in this paper we have discussed some applications of the lambert w function to the theory of liquids . in the case of @xmath167 hard rods or infinitely weak long range repulsive potentials ,
one obtains a close expression for the complex pressure @xmath53 as a function of complex activities in terms of either @xmath3 or the related function @xmath5 .
the dispersion relations derived for @xmath3 and @xmath5 in the appendix give a rigorous justification to the heuristic formula proposed by lee and yang for @xmath53.@xcite we have also shown that , in the framework of the ksshe field theory of liquids , the mf pressure @xmath53 of a gas of particles interacting via soft repulsive potentials can be expressed in terms of the function @xmath5 .
the author likes to thank p. viot for drawing his attention to ref . .
[ appendix ]
it follows from eq .
( [ derif2 ] ) that @xmath170 for all @xmath171 .
thence the function @xmath172 is strictly convex and one can define its legendre transform @xmath173 ( see eg ref . ) . by definition @xmath174 or ,
more precisely @xmath175 where @xmath176 is ( the unique ) solution of @xmath177 therefore @xmath173 is defined for @xmath178 .
it is also a strictly convex function , the legendre transform of which is @xmath179 ( i.e. the legendre transform is involutive ) . in order to get
the expression of @xmath173 one notes that eq .
( [ l1 ] ) implies that @xmath180 and that @xmath181 which yields @xmath182 similarly it can easily be shown that the function @xmath183 is strictly convex and that its legendre transform @xmath184 is given by @xmath185 note that @xmath184 is defined on @xmath186 . over this interval , @xmath184 is strictly convex . where @xmath188 is the contour shown in fig .
( [ contour ] ) and @xmath13 is _ not _ on the cut @xmath189 .
indeed @xmath190 as follows from eq .
( [ wseries ] ) .
we consider now eq . ( [ a1 ] ) in the limit @xmath191 and @xmath192 . the asymptotic formula for large ( complex ) @xmath13 , i.e. @xcite @xmath193 where @xmath194 is the principal branch of the natural logarithm , ensures that the contribution to eq .
( [ a1 ] ) from the large circle @xmath195 tends to zero as its radius @xmath74 tends to infinity ( jordan lemma ) .
note that we have substracted @xmath190 from @xmath3 in eq .
( [ a1 ] ) precisely in order to obtain this property .
therefore we have with the convention that @xmath199 is defined on the upper lip of the cut and by noting that @xmath200 ( for @xmath201 not on the cut ) we infer from eq.([cauchy2 ] ) that @xmath202 \ ; ds \ ; .\ ] ] the final step is to perform an integration by parts which yields @xmath203_{-\infty}^{-e^{-1 } } \nonumber \\ & + & \int_{-\infty}^{-e^{-1 } } \frac{-1}{\pi } \frac{d}{ds } \im \left ( w_0\left(s\right)\right ) \ ; \ln(1-\frac{z}{s } ) \ ; ds \ ; .\end{aligned}\ ] ] since , in the one hand @xmath204 and , on the other hand , @xmath205 as can be obtained for instance from the asymptotic behavior ( [ asymp ] ) of @xmath3 , then the first contribution to the rhs of eq .
( [ dispers0 ] ) vanishes and we are left with \(a ) the function @xmath37 is a positive increasing function on the cut @xmath189 . for @xmath207 ( @xmath208 , @xmath209 as can be seen from the series expansion of @xmath3 about the branch point ( cf eq .
( [ wseriesb ] ) ) . in the other hand @xmath210 for @xmath66
as can be infered from eq .
( [ asymp ] ) .
thence the integral of @xmath37 along the branch cut is convergent ; more precisely @xmath211/\pi \nonumber \\ & = & 1 \ ; .\end{aligned}\ ] ] \(b ) the dispersion relations for the function @xmath5 are similar to eqs.([dispersionw ] ) for the reasoning presented above can be reproduced without major changes .
the distribution @xmath38 is again a positive function on the interval @xmath212 .
it is given by 99 corless r m , gonnet g h , hare d e , jeffrey d j and knuth d e 1996 _ adv .
maths _ * 5 * 329 jeffrey d j , hare d e and corless r m 1996 _ mathematical scientist _ * 21 * 1 corless r m , jeffrey d j and knuth de 1997 in _ proceedings issac 97 _
( maui , wolfgang w. kuechlin , editor ) valluri s r , jeffrey d j and corless r m 2000 _ canadian j. physics _ * 78 * 823 see eg .
talbot j , tarjus g and viot p 1999 _ j. phys . a : math . gen _ * 32 * 2997 _ mathematics for physicists _ ( new york : harper and row ; tockyo : john weatherhill ) tonks l 1936 _ phys .
_ * 50 * 955 hauge e h and hemmer p c 1963 _ physica _ * 29 * 1338 caillol j .-
2003 _ arxiv : cond - mat/0211113 _ , to be published in _
_ carathodory c 1958 _ theory of functions , vol i _
( new york : chelsea publishing company ) dennery ph . and
krzywicki a. 1967 _ mathematics for physicists _
( new york : harper and row ; tockyo : john weatherhill ) groeneveld j. 1962 _ phys .
lett . _ * 3 * 50 yang c n and lee t d 1952 _ phys .
rev . _ * 87 * 404 lee t d and yang c n 1952 _ phys .
rev . _ * 87 * 410 j .- m .
caillol j .-
2002 _ j. phys . a : math .
gen . _ * 35 * 4189 ruelle d 1969 _ statistical mechanics , rigorous results _
( new york , w. a. benjamin ) kac m 1959 * 2 * 8 siegert a j f 1960 _ physica _ * 26 * 530 .
stratonovich r l 1958 _ sov .
solid state _ * 2 * 1824 hubbard j 1954 _ phys .
lett . _ * 3 * 77 ; hubbard j and shofield p 1972 _ phys .
lett . _ * 40a * 245 edwards s f 1959 _ phil . mag .
_ * 4 * 1171 hansen j p and mc donald i r 1986 _ theory of simple liquids _ ( london : academic press ) | we apply the recently defined lambert w function to some problems of classical statistical mechanics , i.e. the tonks gas and a fluid of classical particles interacting via repulsive pair potentials .
the latter case is considered both from the point of view of the standard theory of liquids and in the framework of a field theoretical description .
some new mathematical properties of the lambert w function are obtained by passing . |
whether the milky way dwarf spheroidal ( dsph ) satellite galaxies are undergoing tidal disruption remains a controversial question .
such tidal disruption would naturally lead to extended populations of stars that have been stripped from the satellite core .
that _ most _ of the milky way ( mw ) dwarf spheroidals exhibit radial density profiles with extended components was suggested by the large area photographic survey of most of the galactic dsph satellites by ( * ? ? ?
* hereafter ih95 ) .
a number of studies have addressed the question of the reality of extended structural components around individual dsph examples among them the carina dsph , for which the issue has prompted lively debate ( @xcite , ( * ? ? ?
* ; * ? ? ?
* hereafter paper ii ) , @xcite , @xcite , @xcite ) . in a recent review of past and new work on the carina system , (
* ; * ? ? ?
* hereafter paper vi ) attempted to resolve the previous , apparently discordant results regarding the photometric detection of an extended carina structural component .
paper vi showed that of all previous photometric surveys of carina , that of @xcite which makes use of the @xmath12washington @xmath13 filter technique to identify giant stars ( ( * ? ? ?
* ; * ? ? ?
* hereafter paper i ) ) at the distance of the carina system achieves the highest , and therefore most reliable , signal - to - background contrast in the diffuse outer parts of the carina system .
moreover an extended , power - law component detected around carina in @xcite is supported by spectroscopic confirmation of carina giant candidates to 1.4 times the nominal limiting radius ( @xmath2 ) of the central - fitted king profile in paper vi .
our previous work has therefore established the likely reality of the king + power law " density profile for the carina dsph . in this paper ( 3 )
we take advantage of a similar , but deeper and much wider area , photometric database of carina than that presented in @xcite and , in addition , contribute higher - quality radial velocities ( rvs ) of carina stars from echelle spectroscopy of giant star candidates to more than three times the angular separation from the carina center than we explored in @xcite .
carina - associated stars are now established to @xmath14 from the carina core , leaving no doubt as to the reality of an extended component to the carina system and imposing extreme limits on the mass of carina if these stars are bound to the dsph ( 4.6 ) . to further improve the kinematical mapping of the carina system at smaller radii
, we also take advantage of archived , publicly available vlt+giraffe spectra of more than 1000 stars near the center of carina , which contribute more than 300 additional rvs of carina - related stars within the king limiting radius .
the resulting velocity dispersion profile of the carina system is the most extensive yet determined for any dsph galaxy , yet shows a continuation to large radii of the same more or less flat trend found ( to smaller radii ) in other galactic dsphs ( @xcite ; @xcite ; @xcite ; @xcite ) .
the present photometric and spectroscopic database of stars in the direction of carina has yielded the additional discovery of a second apparently coherent stellar population in the foreground of the dsph ( 6 ) .
this other milky way substructure is as dynamically cold as the carina system itself and , ironically , represents the primary source of contamination within our previous ( @xcite ) and present photometric samples of carina stars outside @xmath2 .
the fifteen stars in our mike sample that are part of this other substructure share a number of properties ( color - magnitude diagram position , metallicity , and velocity - angular separation trend ) with stars of the large magellanic cloud ( lmc ) , but stretch some 22@xmath11 from the lmc center . as with carina
, these widely separated stars place new , very large lower limits on the lmc mass and tidal radius if the stars are bound to their parent satellite .
through spectroscopic follow - up of stars in the @xcite carina database , @xcite demonstrated the efficacy of washington @xmath0 photometry to produce high quality candidate lists of giant stars from the carina system to large separations from the core , and dispelled concerns ( @xcite ; @xcite ; @xcite ) that there may have been problems with the original methods or findings of @xcite .
nevertheless , a deeper , more uniform , and larger area @xmath0 survey of the carina system was desired : @xcite ( see 2.4 of that paper ) showed how the results of a survey with better photometry would improve the carina giant candidate selection , whereas surveying to a larger angular radius would give greater insight into the extent and character of this outer carina population .
thus , new carina photometry over a 10.74 deg@xmath1 area ( 9.3 times more area than covered in @xcite the area outlined below in fig .
1 ) centered on the carina dsph was obtained with the mosaic wide - field imaging camera on the blanco telescope on ut 2000 feb 24 - 27 under photometric conditions . daophot
ii / allstar ( @xcite ) psf - fitting photometry was derived for stars in each of the individual mosaic pointings , producing magnitudes with median errors of @xmath15 ( 0.018 , 0.020 , 0.015 ) at @xmath16 , which is approximately 3.4 mag below the carina red giant branch ( rgb ) tip ( fig . 2 ) .
the photometry of this new survey is about 2 times more precise at that magnitude than the carina data presented in @xcite .
instrumental magnitudes were calibrated into the standard system via multiple observations of washington@xmath17 standards in @xcite . each star in our catalog
has been corrected for reddening based on its galactic coordinates by using the reddening map constructed by @xcite .
we found an @xmath18 range of 0.0330.102 in our carina fields .
as in @xcite , the dereddened @xmath19 two - color diagram ( 2cd ) and the @xmath20 color - magnitude diagram ( cmd ) are used together to select stars most likely to be carina giant stars .
figure 3 illustrates the selection criteria we used to identify carina rgb stars : carina giant star candidates " are expected to fall primarily within the regions bounded by the solid lines in each of the diagrams in figure 3 . because we want to create the most reliable maps of carina density possible and because we choose to reserve the valuable magellan echelle spectroscopic follow - up ( 3 ) observing for the very best photometrically - selected candidates , our initial selection criteria were deliberately conservative .
for example , we did _ not _ employ the proposed wider cmd selection criteria discussed in 3.2 of @xcite , but maintained the more restrictive limits used in @xcite . in addition , our 2cd giant selection boundary is set far from the dwarf star locus to minimize photometric contamination ( fig .
3b ) of the giant sample .
however , the extensive , archived vlt+giraffe spectroscopic data set for the carina field , obtained for dsph candidates selected independently of our photometry and methodology , allows us in the velocity analysis described later ( 4 ) to search for additional carina stars with measured rvs that , while being excluded from our conservatively made best `` candidate lists , still occupy ' ' rgb - like " regions of the cmd and 2cd ( 4.1.3 ) . across our survey area
, the photometric sample is expected to be complete to @xmath16 , so we analyze the spatial distribution of giant candidates to this magnitude limit .
in addition , because our spectroscopic survey is almost complete outside the carina @xmath2 to @xmath21 , we also analyze the spatial distribution of giants using this magnitude limit .
figure 4 presents the radial density profile derived for the carina dsph for these two adopted magnitude limits . to create this profile
, stars have been binned into elliptically - shaped annuli matching the carina center , ellipticity and position angle found by @xcite . as discussed at length in @xcite and @xcite , proper assessment of the background level
( i.e. density of false positive detections ) is critical to deriving dsph radial density profiles .
here we adopt two strategies for assessing this backgound level . in the case of the @xmath22-limited density profile , we can
very accurately estimate the background directly from the results of our spectroscopic survey ( 4 ) , which is 90% complete for stars beyond the @xcite carina @xmath2 .
this spectroscopically - verified , true background of 2.3 deg@xmath23 within our carina giant candidate " star sample is subtracted from the observed density distribution of @xmath24 carina dsph giant candidates across the entire survey to reveal the density profile shown in figure 4a .
because of our near spectroscopic completeness for these stars , the density profile shown in figure 4a beyond the @xcite @xmath2 virtually reflects the exact distribution of all rv - verified members there . for our @xmath25 sample , the background is estimated using the cmd - shifting " method used in @xcite to estimate a background rate , with the one important difference that with our new survey here we are able to make exclusive use of the vast area outside of the carina king limiting radius to significantly reduce potential contribution of any carina stars not lying on the carina rgb ( e.g. , asymptotic and post - asymptotic giant branch stars ) that may have artificially inflated the estimated backgrounds in the @xcite execution of this method .
on the other hand , we acknowledge that this method may also underestimate the contribution of the newly discovered halo substructure discussed in 6 , since it has a similar cmd position as the carina dsph . to correct for this ,
we add back into our background estimate the fractional contribution of stars from this substructure among the spectroscopic sample of @xmath26 stars chosen as carina giant candidates .
this yields a conservative background level of 10.3 deg@xmath23 , which is then subtracted from the observed density distribution of @xmath25 carina rgb candidates .
the two samples with the two methods of background calculation produce remarkably consistent radial profiles ( fig .
4 ) . in both cases ,
the central part of the density profile is well described by the normalized @xcite king profile ( shown by the solid line ) for carina , which is characterized by @xmath27 arcmin and a core radius of 8.8 arcmin .
both radial distributions also show a prominent , second break population " roughly following a power - law decline to the limits of our present survey .
the dashed lines correspond to power - law indices of -1.5 , -2 and -2.5 respectively .
a -2 index power law appears to yield a reasonable match to the density fall - off of the break population , although power laws with indices of -1.5 or -2.5 can not be discounted ; in general , the power law here is steeper than found in @xcite , owing to a slightly higher background derived in this study ( see also a discussion of this steeper slope in @xcite ) . nevertheless , in this completely new photometric survey with substantial spectroscopic
follow - up we have independently borne out the general conclusion of @xcite and paper vi that carina exhibits a prominent , extended , power - law break population .
as discussed in @xcite , the density profile exhibited in figure 4 mimics that of model disrupting dsph galaxies ( see , e.g. , fig .
15 of @xcite ) . with our updated version of the carina density profile
, we can revisit the implied fractional mass loss rate according to the method of @xcite under the assumption that the power law population represents unbound tidal debris .
we derive a fractional mass loss rate for carina of @xmath28=0.075 gyr@xmath9 , but we must note that this method is technically derived for break populations following a -1 power law , and even in that case it only yields estimates good to within a factor of two .
perhaps a better estimate of the fractional mass loss rate comes from the @xmath29 method of @xcite using the corrections given by @xcite ; this method yields un upper limit for the carina mass loss rate of @xmath30 gyr@xmath9 . in a subsequent paper ( muoz et al .
, in preparation , hereafter m06 ) we use this newly derived density profile as well as the velocity dispersion profile derived in 4.3 to model the mass loss history using n - body simulations specific to the carina dsph , and derive likely mass loss rates generally between these two estimates .
the carina power - law population has been of particular , though not exclusive , interest during our follow - up spectroscopic observations .
@xcite presented radial velocity observations obtained with the blanco telescope + hydra multifiber system . only some of these observations were of sufficient resolution to contribute reliable information on the internal dynamics of ( rather than simply stellar membership in ) the carina dsph .
thus , on ut 2004 jan 27 - 28 and dec 29 - 30 good spectra of a total of 77 carina giant star candidates selected from the new photometric survey were obtained using the magellan inamori kyocera echelle ( mike ) spectrograph on the clay 6.5-m telescope at las campanas ; this instrument delivers @xmath31 resolution spectra over the red echelle orders we used for this work . of the 77 stars targeted ,
65 were selected to be giant candidates by the giant selection shown by the solid lines in figure 3 and 12 were selected using an expanded 2cd selection ( shown by the dotted lines in the same figure ) .
this wider 2cd giant selection corresponds to that derived in @xcite .
more than half ( 40 ) of the 77 stars observed with mike lie outside the nominal carina @xmath2 as determined by @xcite ; the rest are scattered throughout the region inside the king limiting radius , but primarily at larger radii where few previous carina spectra have been obtained . radial velocities ( rvs ) have been derived via cross - correlation of the mike spectra against a universal template " containing sets of stellar atmospheric absorption lines that typically give the strongest correlations to the spectra of late type stars ; apart from these lines , the bulk of the spectra and templates are masked out because these wavelengths contribute more noise than signal to the cross - correlation spectrum .
prior to cross - correlation , the spectra are also fourier - filtered to remove irrelevant low frequency features as well as features with higher frequency than the intrinsic resolution of the spectrograph . a fuller discussion of this cross - correlation technique is given in @xcite
; but we have found that the procedure works just as well , or even better , for @xmath32 spectra than for the moderate resolution spectra cross - correlated in that paper .
we observed to high @xmath33 a number of k giant velocity standard stars that we used to measure small systematic offsets imposed on the derived rvs that are particular to the nature of the adopted artificial template .
our cross - correlations here were conducted over the echelle order ( spanning 8468 - 8693 ) that contains the calcium infrared triplet and over a dozen other useful lines in stars as metal poor as carina ( [ fe / h ] @xmath34 ) .
tests with other , nearby orders yield similar rv results but of lower reliability , so the values given here are based solely on the calcium triplet order , where the typical @xmath33 of the stellar continua were 7 - 12 per pixel .
this particular echelle order also contains ample numbers of telluric absorption features with strengths great enough to yield useful velocities . since the stars were observed with a 0.9 arcsec slit whereas the seeing often was as good as 0.7 arcsec , significant fractional errors in derived rvs may arise from slit centering errors . to measure the velocity shifts that result from this effect
, we independently cross - correlate the telluric absorption features in each order against those in a set of observed rv standards as well as in dusk spectra ( see discussion in @xcite ) .
these rv standards were typically exposed by smoothly passing them across the slit during the integration to create a symmetric net slit function for the resulting spectra . by comparison of multiple spectra obtained of several carina giants as well as by comparison of results from cross - correlation of different echelle orders
, we find the random errors in the derived rvs to be better than 1.0 km s@xmath9 for the january run and 2.5 km s@xmath9 for the december run .
the degradation in the second run was due to significantly worse overall observing conditions that resulted in poorer @xmath33 spectra on average .
table 1 gives for the stars observed with mike the j2000.0 positions , date of spectroscopic observation , photometric data , rvs in both the heliocentric and galactic standard of rest ( @xmath35 ) conventions , as well as a parameter that characterizes the quality of the rv : an overall quality index , @xmath36 , which ranges from 1 ( lowest quality ) to 7 ( highest quality ) .
the precise meaning of the various @xmath36 grades is explained in @xcite and @xcite .
as an additional check on the rvs , we independently derived rvs for all mike spectra using the fxcor package from iraf following the method described in @xcite .
the mean rv difference between both methods is @xmath37 km s@xmath9 with a dispersion of @xmath38 km s@xmath9 showing a very close correspondence between the methods .
however , for some spectra with very low @xmath33 our standard methodology failed to yield an acceptable ( i.e. @xmath39 ) cross - correlation , whereas fxcor yielded a cross - correlation with higher apparent reliability . in these cases
we have adopted the fxcor rv in table 1 and given the derived velocity error in place of the @xmath36 value . in the end ,
all 65 carina - giant candidates observed with mike have a reliable velocity and these form the basis of most of the outer carina rv analyses below . among the 12 stars with rvs selected from the expanded giant selection criterion in figure 3b ,
none have been found to have a carina - like velocity ; however , two of these stars have velocites near @xmath40 km s@xmath9 and constitute members of the newly discovered halo substructure discussed in 6 .
thus we include these two stars in our discussions relevant to this halo substructure .
because our mike observing focused primarily on the most widely separated carina giant candidates , our resulting spectroscopic coverage leaves a significant statistical gap from the only other previously published echelle resolution carina rvs , which are in the carina core ( @xcite ) .
fortunately , there exists a substantial collection of archived vlt / flames observations of the carina system that bridges the gap .
these data were retrieved and reduced to rvs by s.z . and d.c .. flames is installed at the nasmyth a focus of the vlt kueyen telescope and is composed of a fiber positioner , ozpoz , that feeds the dedicated medium - high resolution giraffe ( resolving power @xmath41 ) and uves ( @xmath42 ) spectrographs with 132 and 8 science fibers , respectively , over a large field of view ( @xmath43 arcmin in diameter ) in the medusa " mode .
the vlt carina data set used in this paper was collected over a 9 night run at the end of 2003 ( 22 - 31 december ) and consists of 16 different pointings , each observed four times . the exposure time was @xmath44 seconds per pointing .
the four exposures for each pointing were taken in sequence and with the same medusa plate configuration .
all observations were done using giraffe in the low resolution , lr08 set - up having @xmath45 , and centered on the calcium infrared triplet to cover the region from 8206 to 9400 . at the end of each observing night , during daylight , a sample of calibration frames were taken by the vlt staff within the nominal vlt calibration plan .
spectroscopic calibration and extraction have been performed with the giraffe bldrs data reduction pipeline ( version 1.12 ) and the gircalib calibration reference file database ( version 2.3 ) .
the gircalib image database contains generic reference solutions for the calibration frames ( bias and dark frames , flat - fields , fiber slit geometry and fiber response correction frames as well as wavelength calibration for all the different flames observing modes ) that are used as initial guesses for specific night - to - night solutions for all the different calibration steps . for each night
, the calibration frames ( bias , flats , wavelength calibrations ) are grouped together and reduced with the appropriate recipe , starting with the reference solution in the gircalib database and then iterating corrections to it .
once all of the solutions are found they are applied with a single command to the science images .
no particular problems were encountered in the reduction of the calibration frames , but occasionally the wavelength calibration gave unstable and distorted solutions due to the uneven spacing and scarce number of thar calibration lines in the available spectral range . to overcome and check this problem , for each wavelength calibration frame used ( one per night taken during the daytime ) the emission line detection threshold and fitted polynomial order were readjusted until a satisfactory solution was obtained .
these solutions were then verified directly by the thar - calibrated science spectra ( before night sky subtraction ) , which were cross - correlated with a separate , emission line night sky spectrum calibrated externally with the detailed night sky line lists of @xcite .
it was found that the average rms velocity scatter from fiber to fiber based on the sky - lines was an acceptable @xmath46 km s@xmath9 .
we adopt this value as our wavelength zero point error for the giraffe spectra .
the same test revealed that the offset from plate to plate was less than @xmath47 km / s ; nevertheless , we corrected all plates to the same radial velocity zero point system based on the night sky lines .
the archived giraffe images contain spectra from all of the fibers for a given medusa plate . between 109 and 112 medusa fibers
were placed on target stars depending on the pointing , with the remaining fibers positioned on empty sky positions .
the identification of the target objects associated with each spectra is possible using associated archived tables containing the observers input values of target positions and magnitudes as well as details of the positioning of the fiber on the sky .
radial velocity derivations were performed using an implementation of the tonry and davis ( 1979 ) method in the midas environment .
we extracted radial velocities both for each single exposure of each medusa plate and then for the sum of the four exposures per plate . for each exposure we first extracted the sky fibers to create a sky spectrum for that exposure .
this sky spectrum was subtracted from each target fiber spectrum and the result was continuum - normalized and finally cross - correlated with a synthetic spectrum , log@xmath48 and [ fe / h]@xmath49 .
] of a low metallicity giant star to obtain the radial velocity .
in the second reduction method we summed the four extracted and sky - subtracted spectra for each star and cross - correlated _ that _ with the template spectrum .
the comparison of the single spectra and the summed spectra rvs for each object revealed that the rvs from the former were very poor , especially for the faintest stars : several times we failed completely to measure a reliable rv . in the cases where we were able to get four independent rvs we compared their average with the rv of the summed spectrum and found that the rms was much larger than the measurement error in 80% of the sample .
thus , we decided to use only the rvs derived from the summed spectra .
a total of 1771 independent radial velocity measurements were obtained across the sixteen medusa pointings . after removing 66 stars for which we could not get an rv and accounting for repeated targeting of some stars ,
rvs were obtained for 994 distinct stars . in the final definition of the rvs we found that among objects having more than three measurements ( @xmath50 stars ) the scatter was always compatible with the measurements errors except for the very faintest objects where we found a larger scatter .
the rv errors take into account this larger scatter . of the 994 individual giraffe target stars ,
975 were found in our washington+@xmath51 photometric catalog .
we only consider those 975 in our analysis because the remaining 19 stars not present in our photometric catalogues can not be checked for their giant status in the 2cd .
, visible in figure 1 .
we note that only 10 of these 19 stars have rvs consistent with the carina dsph . ]
table 2 presents the rv information for these stars . to the mike and giraffe data we also add blanco+hydra rvs for photometrically selected giant stars from @xcite that were observed at @xmath52 resolution in october 2001 .
we include these hydra rvs only for those stars not already having higher resolution echelle observations . in the end ,
our sample includes a total of 1123 rvs from table 1 , table 2 , @xcite and the @xcite contribution .
figure 5a shows the distribution of all derived rvs for stars in the carina field as a function of elliptical distance from the center , including stars having rvs from @xcite ( green points ) , hydra observed stars from @xcite ( cyan points ) , stars with mike rvs ( red points ) , and giraffe data ( blue points ) . the elliptical radius of a star is defined to be the semi - major axis radius of the ellipse centered on carina ( with the ellipticity , center and position angle for the dsph as found by @xcite ) that passes through the star .
figure 6a shows the integral of the rv distribution over all radius .
a most obvious characteristic of these full sample " rv distributions is the presence of the prominent rv peak associated with the carina core near @xmath53 km s@xmath9 .
however , a significant contribution of stars at other rvs may be seen , particularly from stars with @xmath54 km s@xmath9 from the milky way .
these contaminants come predominantly from the giraffe sample , which was apparently primarily selected on the basis of positions of stars in the cmd . while the carina rv peak still stands out , the substantial background of non - carina stars makes it difficult to define an accurate rv criterion for cleanly isolating carina members . figures 5b and 6b show the same rv distribution , but only for stars satisfying the conservative figure 3 criteria for identifying carina giant candidates by their washington @xmath0 photometry .
this distribution of the
best " photometric candidates makes it easier to define an appropriate additional criterion , based on rvs , for identifying carina members . anticipating that the velocity dispersion of carina members actually rises slightly outside the carina core
, we define as an rv membership criterion the 3@xmath55 range defined by rvs for carina stars beyond @xmath56 ( twice the core radius ) , where we find a mean @xmath57 and a @xmath58 km s@xmath9 .
this range is indicated by the dashed lines in figure 5b and the shaded region in figure 6b .
this new rv selection criterion for mike and giraffe stars is narrower than that applied in @xcite , but this is because the rvs in the present sample have smaller random errors .
this final set of stars , selected by our conservative cmd and 2cd criteria ( fig .
3 ) is shown in fig . 5b .
an additional feature apparent within the rv distribution of the best photometric sample " in figure 5b and 6b is the distinct group of stars with a clumped rv at an even _ more _ extreme velocity than the carina dsph .
this feature is even more clear in figure 6d , where we show a histogram for a subsample of stars from figure 6b , in particular , stars with @xmath59 .
their @xmath40 km s@xmath9 implies a significant retrograde motion for stars in this direction of the sky .
the magnitudes and colors of these stars ( table 1 ) are also rather clumped , indicating similar spectral characteristics and an apparently similar ( and substantial ) distance . in
6 we explore further this moving group of giant stars from what appears to be a newly found halo substructure .
a comparison of figures 5a and 6a with figures 5b and 6b suggests that the restrictiveness of our conservative " photometric selection of carina giants , while providing extremely pure samples of carina stars , also leads to a non - negligible level of incompleteness ( a well - known issue we have addressed before in 3.2 of paper vi ) .
given that we now have the advantage of three criteria for discriminating carina giants and a large number of rvs from giraffe in the carina main body , it is worth reinvestigating the tradeoffs between sample size / completeness and sample purity .
more specifically , can we expand any of the selection limits to admit substantially more carina stars from the giraffe sample without sacrificing the reliability of the membership census .
* 2cd outliers : * figure 7 demonstrates some possibilities for expanding our membership acceptance criteria by showing the 2cd and cmd of stars satisfying our newly established carina rv - membership criterion , but falling outside one or the other ( or both ) our conservative photometric criteria ( plotted as the solid lines in figs . 3 and 7 ) .
as may be seen in figure 7b , a large fraction of these stars lie _ just below _ our figure 3b giant selection in the 2cd .
however , inspection of the distribution of stars in figure 3b clearly shows a strong , almost vertical giant star 2cd concentration at @xmath60 that extends below the adopted diagonal limit there .
moreover , the 2cd analysis of giant and dwarf stars presented in @xcite makes clear that giant stars are commonly found at these positions of the cmd a point demonstrated by the superposition of the @xcite giant star boundary " in figure 7b ( dotted lines ) .
stars in figure 7 lying within the @xcite 2cd boundary but within the figure 3a cmd boundary are marked with red open triangles in figure 7 .
given that these stars satisfy the rv , cmd and the @xcite 2cd criteria , we regard these stars as carina giants from here on
. * cmd outliers : * we may also investigate those stars that satisfy the rv and 2cd criteria but not our initial cmd boundary . in figure 7 these stars are marked with open blue circles for the stars that satisfy the stricter of the 2cd boundaries and green open circles for the stars satisfying the @xcite 2cd limit .
almost all of these lie very close to the rgb limit .
a number of them lie at a cmd position just above the strong red clump .
given that carina has stellar populations as young as 0.6 - 1.0 gyr ( @xcite ) , it might not be too surprising to find some core he - burning stars lying above the canonical red clump from the dominant , older , more metal - poor carina population ( e.g. @xcite ) .
other modest cmd outliers are in cmd positions consistent with those expected for asymptotic giant branch stars .
a similar outlier trend was found in paper vi , where it was noted that slightly expanding the magnitude width of the cmd selection criterion by a few tenths of a magnitude would increase completeness with virtually no decrease in reliability .
given that previous conclusion , and that these stars satisfy the 2cd and rv criteria , we consider all of these outliers as carina members .
* rv outliers : * finally , what about stars that fall within the 2cd and cmd criteria but just outside the rv criterion ? several of these stars are conspicuous in figure 5b .
first , we note that there are @xmath61 stars satisfying our 3@xmath55 rv criterion in figure 5b . for a sample of this size and with a gaussian distribution , one expects @xmath62% , or @xmath10 outlier .
as may be seen in figure 5b , two stars with @xmath63 lie just below the rv cutoff and are probably very likely this kind of gaussian - wing outlier member .
these stars ( c2661 and c161179 ) are indicated by the blue solid squares symbols in figures 5 and 7 , where they can be seen to be very solidly photometric members . nevertheless , because they are in the well - populated central part of carina , whether or not they are included in our analyses has very little effect . on the other hand , as we discuss in 4.5
, the velocity dispersion of carina appears to grow beyond the king limiting radius , and , even though our rv selection criterion was derived from stars with @xmath64 specifically for this reason , the rv dispersion that sets the selection criterion is dominated by stars with @xmath65 . beyond this range ,
the dispersion not only grows , but , as we show in 5 , the rv distribution becomes flatter than gaussian .
both larger velocity dispersions as well as more platykurtic velocity distributions are fully consistent with models of disrupting dsphs systems ( @xcite ; m06 ) .
thus , even wider separated rv - outliers are not only conceivable at large radii , they are expected .
we mark three of these from our mike sample c1960448 , c2450090 and c2050415 with red square symbols in figures 5b and 7 .
these stars , which lie within @xmath66 km s@xmath9 ( 3@xmath55 ) , @xmath67 km s@xmath9 ( 2@xmath55 ) and @xmath68 km s@xmath9 ( 1@xmath55 ) , respectively , of our carina rv membership limit , are again solidly within the photometric carina giant candidate selection criteria ( figure 7 ) .
they are particularly interesting potential members , since all three lie approximately along the carina major axis , and at large radii @xmath69 to the east , @xmath70 southwest and @xmath71 northeast of carina center , respectively ( see fig .
indeed , the latter star is potentially the most widely separated carina giant in our sample , at @xmath72 .
nevertheless , unlike in the cases of the sample - admitted 2cd and cmd outliers above , even though we can make a compelling case for the membership of all five of these rv outliers , we _ exclude them _ from our dynamical analyses to follow , so that we do not unduly bias our velocity results .
figures 5c and 6c summarize the rv distributions of our final , expanded carina - member sample based on our two ( slightly widened ) photometric criteria and one velocity criterion . in tables 1 and 2
we designate by the column member " those 260 stars considered to be members by the most conservative criteria and those additional 116 stars that have been admitted as carina members by the exceptions described in this subsection .
we stress that ( 1 ) all 116 of these stars are from the giraffe sample , ( 2 ) all but 2 are within @xmath73 and so have no impact on the dynamical results at larger radii , and ( 3 ) the inclusion or exclusion of these 116 stars in our analysis has little effect on the general dispersion trends described later ( fig .
thus we have opted to include these 116 stars to improve our sampling and statistical uncertainties .
the five rv outliers discussed above but not included in our analyses are highlighted in this column by rv ? " . the azimuthal distribution of the carina rv - members on the sky ( fig .
8a ) shows them to lie predominantly along the carina major axis , even though , as shown in figure 1 , the azimuthal coverage of our photometric and spectroscopic efforts actually favors the _ minor _ axes ( see , e.g. , the distribution of carina giant candidates _ not _ found to be rv members in fig .
figure 9 , which shows the ratio of the circular to elliptical radius ( @xmath74 ) for each star in the survey versus its circular radius , demonstrates the tendency for carina rv members outside the king limiting radius to lie along an extension of the position angle of carina s ellipticity and , indeed , to have an apparently even more elliptical distribution in this direction at larger radii .
stars on the major axis will have @xmath75 and stars on the minor axis will have @xmath76 , according to the ellipticity of carina ( @xcite ) .
that the mean @xmath74 increases at larger @xmath77 shows the tendency for the extended population to become even more stretched along the major axis , evokes the character expected of tidal tails , and is a key characteristic of dsph tidal disruption models ( @xcite ; @xcite ; @xcite ; @xcite ; m06 ) . further surveying for carina members over larger radii to see whether and how this trend may continue
would provide a valuable check and important constraint on the nature of any tidal disruption .
@xcite has already focused on the reliability of our methodology to assess dsph structure into extremely low surface brightness regimes , with specific focus on carina . however , with the now much better spectroscopic coverage as well as better photometry of the carina field we may reassess the effectiveness of our washington @xmath78 survey strategy .
in addition , the mike spectroscopic sample , which was pre - selected based on the washington+ddo51 photometry , provides an interesting contrast with the giraffe sample , which was not . a straight calculation of our success rate from the 48 carina rv - members among all 65 carina giant candidates with mike spectroscopy yields a success rate of 74% in identifying true dsph members . restricting the analysis to only stars outside the nominal ( ih95 ) king limiting radius yields a success rate of 55% ( 22 dsph members among 40 @xmath79 carina giant candidates with rvs ) , and this includes candidates at extremely low densities ( 0.058% the density of the carina core ) .
however , 13 of the 40 carina giant candidates outside @xmath2 with determined rvs appear to be giant stars from _ another _ tidal stream with rather similar cmd characteristics as carina ( 6 ) . though these stars are not attributable to carina , this newly discovered milky way feature might be argued as a success of the overall methodology we have been using in this series of papers to identify just this kind of halo substructure .
were we to combine these stars with the true carina dsph members , our success rate in identifying
halo substructure " stars rises to 94% .
in contrast , the original giraffe sample was apparently selected only on the basis of the position of these stars in the cmd ( though not our cmd ) . among the 975 stars in the giraffe sample also in our catalogue , 390 , or 40.0% ,
are found to have carina rvs and this is for a sample highly concentrated to the main body of carina , with most stars having @xmath80 .
however , had we applied our photometric selection criteria to the giraffe catalog 97.3% of the stars identified as carina giant candidates would have been found to be rv - members ( almost tripling the telescope efficiency ) . combining all available rv data at all radii , the washington+ddo51 pre - selection results in a 90.5% rv - member efficiency .
thus , the combination of washington@xmath17 photometry with quality spectroscopy is found once again ( see palma et al .
2003 , westfall et al .
2006 , sohn et al .
2006 ) to be a very effective observational strategy for identifying very diffuse halo substructures .
the point is relevant to potential further work on the extended structure of the carina system
. continued searches for carina giants at large separations from the dsph center will require an efficient means to identify the best candidates to optimally take advantage of spectroscopic time on the largest telescopes .
we note that using _ only _ a selection for carina stars by their position along the carina rgb in the cmd becomes a very inefficient way to find carina giants at 3@xmath2 : at these radii , only one in 85 stars in the rgb selection region in the cmd we have used ( fig .
3b ) turns out to be an actual carina giant , and to @xmath81 , the density of such stars is only 7.4 deg@xmath23 , making even multifiber spectroscopic searches for members within a cmd - only target list a rather inefficient enterprise .
estimates for the central and global carina @xmath82 determined using standard prescriptions ( e.g. , core - fitting combined with the central velocity dispersion ) are given by mateo et al .
( 1993 ) as @xmath83 and @xmath84 ( all @xmath82 values in solar units ) , respectively , when isotropic , single - component @xcite models are adopted ; anisotropic models were argued to give similar global @xmath82 for the lowest possible central mass density .
these values were based on an observed central velocity dispersion of @xmath85 km s@xmath9 .
monte carlo analyses conducted @xcite show that it is unlikely that this dispersion has been inflated by either atmospheric jitter in the target k giants or the influence of binaries
. however , there seems to be no real consensus on derived @xmath82 s for carina .
for example , mateo ( 1998 ) quotes the carina @xmath86 as 31 , whereas @xcite , adopting the original mateo et al .
( 1993 ) central velocity dispersion , derive @xmath87 and @xmath88 ( where the large error bars reflect uncertainties in the velocity dispersion , core radius and at least a factor of two uncertainty for the central surface brightness ) .
@xcite estimate the carina mass and @xmath82 by assuming that its perigalactic tidal radius can be approximated by @xmath2 ( obtained from their photometric survey of the dsph ) and using the @xcite relationship between the tidal radius of a satellite and its mass and orbit .
circular orbits yield @xmath82 as low as 0.6 while more eccentric orbits can easily accommodate values as high as the ones derived by @xcite , but @xcite derive a carina @xmath89 based on an orbit with eccentricity 0.6 and apogalacticon twice that of carina s current distance .
the new rv dataset presented here invites yet another @xmath82 evaluation . unlike previous determinations making use of a
central `` velocity dispersion from a relatively small number of stars in the very core of the dsph , our extensive and radially continuous velocity coverage means that the definition of ' ' central " is not pre - defined by our available sample .
if we assume that at least the inner parts of the dsph are well represented by a king profile , figure 4.11 from @xcite shows that the velocity dispersion of stars begins to deviate from its central value at about half the core radius .
figure 10 shows the central velocity dispersion of carina as we grow the radius ( shown in units of core radius as measured by @xcite ) within which we include rvs in the dispersion computation . as we add successive stars out from the carina center the derived central " velocity dispersion ( calculated using the maximum likelihood method , @xcite ; @xcite ; @xcite ) reaches a value of @xmath90 km s@xmath9 at half the core radius ( computed from 87 total carina stars ) .
this value , which is slightly larger than ( but consistent with ) the 6.8 km s@xmath9 value used by @xcite , is adopted to rederive the carina @xmath82 s .
the central mass - to - light ratio can be determined as ( @xcite ) : @xmath91 where @xmath92 is a correction parameter dependent on the concentration value ( 0.955 for carina ) , @xmath93 is the geometrical mean of the half - light radii measured along the major and minor axis ( @xmath94 pc ) and @xmath95 is the central surface brightness ( @xmath96 l@xmath4/pc@xmath97 ) .
we adopt all these structural values from @xcite and obtain @xmath98 for carina where the main source of uncertainty comes from the uncertainty in the central surface brightness .
to illustrate this , we calculate the error in the @xmath99 not considering the uncertainty in the central surface brightness , and obtain @xmath100 . from @xcite :
@xmath101 where @xmath102 is now the geometric - mean king core radius in pc ( @xmath103 ) , @xmath104 is the @xcite dimensionless mass parameter , and @xmath105 is a model - dependent velocity parameter related to the observed velocity dispersion .
table 10 in @xcite gives values for both @xmath104 and @xmath106 of @xmath107 and 0.52 , respectively , for a carina concentration of @xmath108 .
this yields @xmath109 for a @xmath110 ( @xcite ) .
this translates into a total mass of @xmath111 m@xmath4 .
these results are in very good agreement with the ones found by @xcite despite the fact that the structural parameters they use are different from the ih95 ones adopted here . here
we adopt the updated distance of carina from mateo ( 1998 ) , which is larger than the value used by @xcite , and this results in a larger half - light radius that compensates for the slightly larger luminosity adopted here . with this large rv dataset in hand we can now assess the velocity dispersion behavior for carina to well past @xmath2 .
to ascertain this trend , we have studied the velocity dispersion as a function of both elliptical and circular angular distance from the carina center . because the true shape of the gravitational potential and tidal boundary of a dsph are likely to be somewhere in between these limiting shapes , it is helpful to explore these two limiting cases . in each calculation of an rv dispersion 3-@xmath55 outliers
have been removed iteratively , with the mean velocity for each bin reevaluated at each iteration and the dispersions estimated using the maximum likelihood method .
we note that this method assumes that the velocity distribution follows a gaussian distribution everywhere which is not strictly true for carina .
however , such non - gaussian behavior is apparent only in the outskirts of carina ( @xmath112 ; 5 ) , and the effect of the non - gaussian character found there is that the dispersion will tend to be slightly underestimated by the maximum likelihood method .
figure 11 shows the derived carina velocity dispersion profiles for both choices of angular separation : the left panels show profiles plotted against elliptical radius , the right shows the same for circular radius . to test binning effects ,
we have used both 23 and 46 stars per bin ( lower and upper panels respectively ) for stars inside @xmath2 , but because the number of stars with measured rv beyond this point is sparse , the last four dispersion points in each plot are binned at 10 stars each .
the figure 11 carina profiles remain fairly flat throughout the radial extent of the main body of the dsph , to @xmath113 .
such flat profiles over a comparable structural radial range have now been reported ( although not to the radial extent of this study ) for several dsphs : sculptor ( @xcite ; @xcite ) , draco ( @xcite ) , ursa minor ( @xcite ) , fornax ( @xcite ) , leo i ( @xcite ) and sagittarius ( majewski et al .
, in preparation ) .
note that while @xcite found a sudden drop in velocity dispersion at about @xmath2 for both ursa minor and draco , this feature could not be reproduced by @xcite when reanalysing these profiles when washington+@xmath51 photometric and additional spectroscopic data were used to check them .
kleyna et al .
( 2004 ) have also found sextans to have a predominantly flat profile but with a cold velocity dispersion at about @xmath2 ( and a kinematically cold center as well ) ; given that similar claims for cold points near @xmath2 in the ursa minor and draco dsphs have not held up under further scrutiny , the sextans result warrants further investigation .
flat velocity dispersion profiles are incompatabile with mass - follows - light dsph models ( with or without dark matter ) in complete dynamical equilibrium , where decreasing dispersions are expected at large radius , approaching zero as the cutoff radius of the distribution is approached . to explain the observed velocity behavior , @xcite suggest that the _ easiest _ assumption to discard is that mass follows light ; following this line of reasoning , a number of groups ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) have invoked two - component dsph models " , where the dark mass extends far beyond its luminous counterpart and is responsible for the flat dispersion profile at large radius . yet , our mike observations of carina have now yielded the most extensive coverage of velocities in any dsph , including , for the first time , the measurement of the velocity dispersion of a dsph ( apart from sgr ) with a reasonable sample of stars beyond @xmath114 . as may be seen in figure 11 , the velocity dispersion for carina approximately doubles at these large separations a result that is _ not _ explained with previous two - component models
. is abandoning mass - follows - light really the
easiest " assumption to discard in the dsph models ?
flat dispersion profiles arise _ naturally _ in tidal disruption models ( @xcite ; @xcite ; @xcite ) _ even if large amounts of dark matter are present _ and the central parts of dsphs are bound and in equilibrium ( @xcite ; @xcite ) . as we show in m06 , a single - component , mass - follows - light , tidally disrupting dsph model gives a good representation for both the density and velocity dispersion profile for the carina dsph we have derived here . further evidence for a disruption scenario is provided by the trend of velocity across the satellite . in figure 12
we show the mean rv ( in galactic standard of rest ) as a function of @xmath115-distance from the center of carina ( approximately the major axis of the satellite ) .
no significant rv trend in the central part of carina that resembles a rotation curve is observed .
however , beyond @xmath2 , a gentle velocity gradient is observed across the major axis of carina to the extent of our observations . over @xmath116 degree
( 2.1 kpc ) , a peak - to - peak difference of @xmath68 km s@xmath9 is seen in this trend a difference significantly larger than the error in the means for the binned points .
this velocity trend is interesting because it has been predicted as a hallmark of tidal disruption by several studies ( e.g. , @xcite ; @xcite , @xcite ) .
according to @xcite , a velocity gradient across the galaxy that is larger than the velocity dispersion is the clearest signature [ of tidal destruction ] " .
figures 5 and 8 show that we have found rv - verified carina member stars to 4.5 @xmath2 .
this limit may extend to 4.9 @xmath2 if we adopt a 3@xmath55 limit for rv - members specific to the outermost bins in figure 11 , in which case star c2050415 ( represented by the outermost square in figures 5b and 8) is the outermost detected carina giant .
if the rv member at 4.5 @xmath2 is bound to carina , it sets a new lower limit for the physical extent and tidal radius of the dsph at 96.5 arcmin , or 2.84 kpc for an assumed distance of 101 kpc to carina ( @xcite ) . using this radius in the tidal limit equation ( @xcite ) : @xmath117{\rm ln}[(1+e)/(1-e)]+1 } } \right \}^{1/3}\ ] ] where @xmath118 is the orbital semimajor axis , @xmath119 and @xmath120 are the mass of the dsph and the mw inside @xmath118 respectively and @xmath121 is the orbital eccentricity ( values for @xmath118 and @xmath121 taken from @xcite to be 61 kpc and 0.67 respectively ) , the lower limit to the carina mass becomes @xmath122 m@xmath4 assuming a mass of the milky way interior to @xmath118 of @xmath123 m@xmath4 ( @xcite ) .
this estimated mass limit is further underestimated because we are taking the _ projected _ radius of the star as the actual , three - dimensional distance from the center .
given the carina luminosity @xmath124 l@xmath4 ( @xcite ) , the above mass translates to a global mass - to - light of @xmath125 , which is more than 100 times higher than the central and total @xmath82 derived for carina in 4.4 . is ( 0.26 ; 0.94 ) which results in a @xmath82 range of ( 370 ; 470,000 ) .
even a value for @xmath121 of 0.24 corresponding to an orbit with peri : apogalacticon of 63:102 kpc , ( their 95% confidence bounds for these parameters ) yields a @xmath82 that is an order of magnitude higher than the central value . ] on the other hand , if the star at 4.9 @xmath2 is a carina member and it is bound , it sets the tidal radius at 133.7 arcmin , or 3.93 kpc , enclosing an astounding mass of @xmath126 m@xmath4 , which yields @xmath127 .
while some stars on trapped orbits can be found well outside the true tidal radius up to 2@xmath2 or even more ( see , e.g. , discussion in 7.3 of @xcite ) , the number should be extremely rare beyond 4@xmath2 . also , were one to expect the @xmath82 of a galaxy to grow with radius , the asymptotic values implied for carina are unreasonable high even when compared to values for galaxy clusters : 200 - 300 ( @xcite ) , which are thought to be approaching fair samples of the universe . from this line of reasoning
, we must therefore conclude that either carina has an enormous , extended dark matter halo to create a @xmath82 an order of magnitude higher than the universe , or , more simply , that these widely separated carina stars are simply not bound .
we ( muoz et al . 2005 ) have used similar arguments in our discussion of the ursa minor dsph , where a global @xmath82 of 1,400 to 14,400 was implied by the widest separated rv member , depending on the use of circular or elliptical radii , respectively . while the possibility that the widely separated muoz et al .
ursa minor stars could be interlopers that just happen to have the same rv and color - magnitude positions ( i.e. approximate distances ) as ursa minor was explored and shown to be very unlikely , this miniscule possibility can not presently be completely discounted . however , the case for the widely separated carina stars being interlopers is far more difficult to make because of the sheer number of them : six ( possibly eight ) farther than 2@xmath2 .
figure 13 , which shows the global mass and @xmath82 implied for carina as progressively more widely separated rv members are attributed as bound satellite members , demonstrates that the implication of an enormous implied carina @xmath82 is robust to the invalidation of any particular star , or even several , attributed as a sample interloper .
the @xmath82 s in figure 13 are derived in two ways that make use of equation 3 : ( 1 ) the implied mass of carina is found by assuming a spherical potential for the dsph and the star s linear projected distance from the center of carina used as @xmath128 ( _ open circles _ ; again , this is a conservative lower limit , because we are working with _ projected _ radii ) .
( 2 ) assuming that the distribution of stars around carina maintains a constant ellipticity with radius , we can assume there exists for every star not on the major axis a counterpart at the same _ elliptical radius _ on the major axis which is then used for @xmath128 .
this assumption raises the lower limits on the implied @xmath82 s ( _ solid circles _ ) .
the two methods for deriving the minimum implied @xmath82 probably span the actual limits , since galaxy potentials tend to be rounder than their density profiles .
figure 13 demonstrates that all of the stars with @xmath129 or @xmath130 exceeding @xmath131 would need to be discounted as carina - associated to bring the global minimum @xmath82 to more standard values for the carina dsph ( such as the @xmath132 found from core fitting with the central velocity dispersion in 4.4 ) . in other words , if one assumes that the global @xmath82 of carina is that obtained using the central velocity dispersion , then the tidal boundary _ coincides _ with the radius at which the break in the density distribution is indeed observed .
figure 5b attests to the relative purity of the carina dsph giant candidate sample created by our dual photometric selection criteria ( figs .
3a and 3b ) : very few rv outliers are found among our carina giant candidates overall , and , in addition the small number of giant candidates we find that do _ not _ share the carina dsph rv lie predominantly in the 332 km s@xmath9 group .
furthermore , figure 5b suggests that the outer halo is highly substructured ( at least when traced by giant stars ) , a result that is also evident from figure 2 in muoz et al .
in such circumstances , to obtain substantial contamination in our survey would require a considerably unfortunate conspiracy of phenomena to produce a _
second _ halo substructure with the same rv , approximate distance , and cmd distribution as carina ; we consider this possibility as unlikely .
taken alone , figure 13 can be argued as a validation of the notion that dsphs like carina are surrounded by large dark matter halos ( @xcite ; @xcite ) . according to @xcite , nfw - like halos that fit the carina central velocity dispersion ( adopted as 6.8 km s@xmath9 ) and central luminous king profile , even in the face of substantial tidal stripping of the dark halo , still maintain halos with ( 1 ) maxima in their circular velocity profile exceeding 50 km s@xmath9 that peak well outside @xmath2 , as well as ( 2 ) true tidal radii of 11 kpc or more .
making similar arguments for all of the milky way satellites alleviates _ at the high mass end _
the mismatch between the cdm - predicted subhalo mass function and that presented by the galactic satellite system ( i.e. , the missing satellites problem " ; @xcite ; @xcite ; @xcite )
. nevertheless , we believe that an alternative explanation of figure 13 i.e. that carina ( and other dsphs ) are surrounded by populations of _ unbound _ stars released through tidal disruption is not only simpler but also provides a better match to _ all _ of the available observations of carina : * density profile * : we have remeasured the carina density profile with new data , and confirm the existence of a two - component ,
king+power law break " shape suggested earlier by the photometric studies of @xcite , @xcite , @xcite , and @xcite .
this photometric work is now solidly backed by spectra of stars in the break population ( see also @xcite ) , proving the existence of rv - members in the extended power - law break population and leaving no doubt as to the reality of the feature ( cf .
@xcite ; @xcite ) .
this density profile matches ( 1 ) the classic shape of a disrupting dsph galaxy , as seen by n - body simulations of disrupting satellites ( e.g. , @xcite , @xcite ) as well as ( 2 ) profiles observed in archetype examples of tidal disriuption like the sagittarius system ( @xcite ) . in contrast , no published dark halo models predict a dynamical structure that would give rise to the observed _ luminous _ , two - component profile of carina .
it is difficult to imagine how the required structural transition between two bound , pressure - supported stellar populations . ]
could be produced so deeply inside an extended dark matter halo , and , coincidentally , exhibit _ no significant change _ in the observed dynamics ( velocity dispersion ) at this point ( see below ) .
moreover , the position of the break in the profile precisely matches that expected for a carina having a constant @xmath82 given by the core - fitting technique ( 4.4 ) . * azimuthal configuration * : the distribution of stars found in the outer carina structural component shows a preference to lie along the major axis , and to have an even greater ellipticity than the carina core , just as would be expected for emerging tidal tails ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ) . in contrast
, cdm halos tend to have rounder potentials ( @xcite ; @xcite ; @xcite ) so that either the carina halo is very unusual , or an explanation is required for why its embedded luminous component has a rather different spatial distribution than its dark halo . *
velocity shear * : as pointed out in 4.5 , the observed velocity trend observed in the carina system is that expected for tidally induced shear .
however , we regard this observed trend with caution appropiate to the still meager statistics for this measurement in the outermost parts of carina . * velocity dispersion profile * : we find a carina velocity dispersion profile that is flat and then rising well past the king limiting radius .
a characteristic of bound populations is that eventually the velocity dispersion of stars should decline with radius , eventually approaching 0 km s@xmath9 at radii where bound stars reach the apocenters of their internal orbits . that a dynamical cold point " radius is _
not reached _ even among our most widely separated rv - members suggests that , if bound , these stars are not near the tops of their orbits , and that the tidal radius of carina must be beyond even _ well beyond _ , given the still large velocity dispersion at @xmath133@xmath2 the observed typical radius of our rv - members .
thus , to explain the observed velocity dispersion trend requires an extremely extended dark halo of even larger dimensions and mass than implied by figure 13 .
in contrast , flat ( and rising ) dispersion profiles are a natural product of tidal disruption models ( @xcite , m06 ) .
* flattening of the velocity distribution * : as shown in recent studies ( @xcite ; @xcite ; m06 ) if the milky way tidal field strips stars from dsphs ( even if surrounded by a dm halo ) the velocity distribution at large radii deviates from a pure gaussian , in general becoming more platykurtic near and beyond @xmath2 .
we have shown for the case of ursa minor , draco ( @xcite ) , sculptor ( @xcite ) and leo i ( @xcite ) that the velocity distribution evolves from gaussian in the center to a flatter distribution with increasing radius .
the same is observed in carina , where the distribution seems to flatten out at large radii , with a kurtosis excess of @xmath134 for stars beyond 0.8@xmath2 contrasted with the near - gaussian @xmath135 for stars inside 0.8@xmath2 .
however , we note that such flattened outer rv distributions could also be observed in systems where the orbits are mostly circular ( @xcite ) . * an emerging too many satellites problem " ? * : 4.6 makes the case that to keep all of carina rv members bound requires a potential minimum mass for the dsph of @xmath136 m@xmath4 . @xcite
have performed a similar analysis on the ursa minor dsph system and find that to keep it s most widely separated rv - member bound requires a minimum mass of almost @xmath137 m@xmath4 , or @xmath138
m@xmath4 for a counterpart of that star moved along its elliptical isopleth to the major axis .
@xcite argue that , in fact , dsphs have masses of @xmath137
@xmath138 m@xmath4 , which would prevent them from undergoing tidal stripping , even in very extreme , radial orbits .
such @xmath139lmc - mass dark matter halos ( dmh ) are at the limits of the largest subhalo sizes predicted by @xmath140cdm ( @xcite ) ; the existence of _ several _ @xmath141-mass subhalos in a milky way - sized system is not expected ( see figure 14 of @xcite ) . if more examples of subhalos much more massive than previously inferred are found e.g. ,
if we continue to extend the radius over which rv - members are identified in carina and the other satellites of the milky way ( see , e.g. , 6 ) and attribute these stars as bound to the dsph a new problem for cdm will emerge , namely an _
of inferred massive satellites about the milky way . while the situation is not yet extreme enough to rule out the extended dark halo hypothesis on this basis , nevertheless , it is worth pointing out again that tidal disruption is a simple way to put stars at any arbitrary angular separation from a dsph , should even more extreme outliers be found .
moreover , as @xcite point out , inferring the existence of these extremely extended halos and large masses for satellite galaxies brings an inconsistency with the actual measured central velocity dispersions ( which are lower than predicted ) , even if significant tidal stripping and shocking are considered . * the sagittarius paradigm * : all of the observed spatial and dynamical features in carina are also found in the one undisputed case of dsph tidal disruption in the milky way the sagittarius dsph ( see sgr spatial and velocity properties given in @xcite ) .
moreover , we ( m06 ) have explored n - body simulations of modest mass , one component dsph systems ( originating as plummer models ) orbiting for significant fractions of a hubble time and can reproduce the observed properties of carina fairly well . that _ both _ ( 1 ) an actual , uncontested , _ tidally disrupting _ analogue of the carina system , as well as ( 2 ) successful tidal disruption models ( with fewer unexplained details than alternative , extended dark matter halo models ) exist makes it difficult to avoid the question : is carina simply another example of the established sgr paradigm ?
* commonality of disruption * : a number of discoveries of apparent halo moving groups or streams have recently been made ( including the one presented here in the foreground of carina , see 6 ) : the monoceros / gass stream ( @xcite ; @xcite ; @xcite ; @xcite ) , the triand structure ( @xcite , majewski et al .
2004 ) , the m31 giant southern stream ( @xcite ) and a recently discovered , second m31 halo substrucutre ( @xcite ) , the identification of an outer galactic halo stream using blue horizontal branch stars by @xcite , a potential system in virgo ( @xcite ) ; and a new halo moving group found with m giant stars ( majewski et al .
, in preparation ) .
this growing list of examples provides increasingly solid evidence of a highly substructured milky way halo , and to the _ commonality _ of tidal disruption of stellar systems in the milky way halo ( e.g. , @xcite ; @xcite ) .
such tidal streams must come from _ somewhere _ and dsph satellites are the most obvious available source .
the new mike rvs have revealed an additional coherent rv peak in the field centered on the carina dsph ( fig .
5 ) at @xmath142 km s@xmath9 , represented by 15 stars with the rather small velocity dispersion of @xmath143 km s@xmath9 ( figs . 5 and 6 ) .
the extreme rv of this system ( + 122 km s@xmath9 when converted to the galactic standard of rest ) implies a strong retrograde motion for these stars if they are nominal milky way stars at this galactic position ( @xmath144=[260,-22]^{\circ}$ ] ) .
the strong rv coherence of this group makes it even more unlikely that it is from a dynamically hot , well - mixed , random galactic halo population , but the dispersion is , however , of order what one sees in dwarf satellite galaxies : for example , the dispersion is comparable to those measured in the extended parts of the carina system ( fig . 11 ) which we have argued to be likely tidal debris as well as those measured all along the trailing tidal arm of the sgr dwarf debris stream ( @xcite ) .
however , the lack of any spatial concentration of these stars across the relatively large span of our survey fields ( see fig .
8b ) and their very low apparent density ( a factor of @xmath145 more diffuse than the mean @xmath146 giant star density for carina stars of the same apparent magnitude ) suggest that these stars represent either tidal debris from a satellite galaxy or an extremely low density part of a very extended satellite .
figure 14a shows the distribution of stars in this moving group within the cmd of all stars selected to be giants in our photometric survey ( according to the tenets of figure 3b ) , along with the carina dsph rgb " boundary we have used in figure 3a .
the cmd positions of the fifteen 332 km s@xmath9 group stars is both highly concentrated and slightly brighter in mean rgb position than the mean cmd locus of the carina rgb .
a similar concentration is also seen for the moving group members in the 2cd ( fig .
14b ) ; moreover , their relative position in the 2cd compared to carina stars suggests that the 332 km s@xmath9 stars are more metal rich than the mean carina star ( see @xcite ) , assuming similar [ mg / fe ] ratios .
an independent test of the relative metallicities of these stars comes directly from the spectra : despite the relatively low @xmath33 of the spectra ( which were taken for rvs ) , in many cases the strong calcium infrared triplet lines are clear . when possible the equivalent width for each triplet line within each mike spectrum was measured .
we found that for all three calcium lines the equivalent widths for the 332 km s@xmath9 group stars were about double those of carina stars with a similar @xmath147 color .
we also used a photometric bandpass method for measuring the calcium infrared triplet line strengths because ( 1 ) it is perhaps more reliable for relatively low @xmath33 spectra , ( 2 ) it averages results over three lines , and ( 3 ) a formalism exists to convert these photometric line measures into a formal [ fe / h ] value .
we limit this work to mike spectra with @xmath148 per pixel and follow the bandpass definitions summarized in @xcite .
we point out that since our original survey was not intended to measure metallicities , we did not observe an appropriate set of stellar calibrators of the metallicity scale . however , since a primary intention is to compare the relative metallicity between the carina and 332 km s@xmath9 group samples , precise calibration is not necessary . therefore , we followed the prescription outlined in @xcite for converting calcium equivalent width and stellar gravity to [ fe / h ] , adopting the calibration for this procedure from @xcite .
a. j. cenarro graciously made available the code used to measure the line strength indices ( @xcite ) . for studies of resolved galaxies and star clusters
an rgb star s cmd position relative to the system horizontal branch , @xmath149 , is often used as a proxy for surface gravity . to adopt this method , transformation equations from majewski et al .
( 2000a ) are used to translate the washington photometry into cousins @xmath150 and @xmath151 magnitudes .
we start by assuming all stars are at the same distance as the carina dsph and adopt @xmath152=20.8 as the mean magnitude of the carina red horizontal branch .
@xcite use a similar technique to study open clusters with spectra having only slightly better @xmath33 and derive a mean metallicity error of 0.3 dex .
therefore , we believe that 0.5 dex is a conservative estimate of our mean uncertainty , where the main contribution comes from uncertainties in the equivalent width measurements .
figure 15 shows the [ fe / h ] distribution derived for both carina and 332 km s@xmath9 group stars under the assumption of a similar distance .
the mean [ fe / h ] derived for carina stars is -1.86 with a dispersion of @xmath153 in good agreement with other studies ( @xcite ; @xcite ) whereas the mean [ fe / h ] derived for the 332 km s@xmath9 group is -0.93 with a dispersion of @xmath154 . barring possible variations in [ ca / fe ] between the two groups of stars , figure 15
suggests that the metallicity of the moving group may be @xmath155 dex higher in [ fe / h ] than the carina dsph were this group at the same distance . on the other hand , if these moving group stars are more metal rich ( as their calcium line strengths suggest ) , they are also _ intrinsically fainter _ in the @xmath150 band , whereas they are also _ brighter _ in apparent magnitude relative to carina stars of the same color .
all of this suggests that the moving group must be _ closer _ than carina , and by as much as a magnitude in distance modulus or more ( see , e.g. , fig .
12a of @xcite ) .
interestingly , this places the distance of these stars to be of order the distance of the large magellanic cloud ( lmc ) , _ the center of which not only lies only @xmath156 away from the center of our carina field in the sky but has a similarly high systemic heliocentric velocity ( 262 km s@xmath9 ; ( * ? ? ? * hereafter vdm02)_. even more intriguing , in other @xmath157 photometric survey work in fields encircling the lmc we have found additional giant stars with lmc - like velocities ranging from 4 to 18.5@xmath11 away from the lmc center in the general region between the lmc and the carina dsph .
preliminary results for this work have been shown in e.g. , fig .
6 of majewski ( 2004 ) , and a more complete discussion will be given elsewhere ( nidever et al .
, in preparation ) .
here we focus on the relative positions ( fig .
16 ) and velocities ( fig .
17 ) of our best - quality velocities for stars in fields that bridge the region between the lmc core and our carina survey field .
because the expanse of sky involved is sufficiently large that there is significant variation in the reflex motion of the sun in the rv , figure 17 shows velocities after conversion to the galactic standard of rest ( gsr ) frame .
range ; groups of stars with clumped , negative ( i.e. generally retrograde ) @xmath35 are also found ( e.g. , see @xcite ) , but are not relevant to the present discussion . ] in the galactic rotation , anticenter and @xmath158 directions . ] after conversion to @xmath35 , an even greater agreement is found ( fig .
17 ) between the actual velocities of the lmc ( big solid circle ) , the 332 km s@xmath9 group in the carina field ( smaller filled circles ) , and rgb stars we have found with similar velocities between these systems ( open triangles and circles ) . while the relative numbers of stars in each position on figure 17 are a function of widely varying survey areas , spectroscopic magnitude limits , and spectroscopic target selection ( i.e. whether or not stars were selected to be an lmc - like giant , a carina - like giant , or any kind of giant ) ; what is relevant is the smooth variation of the mean velocities in each survey field from the lmc to the 332 km s@xmath9 group , a trend that strongly suggests a dynamical association of all of these stars .
lower rv precision of the measurements for the stars found outside the carina survey region . ] even more intriguing is that this velocity trend matches that found for other lmc tracers ( e.g. , @xcite ; @xcite ) at similar position angles from the lmc core over a 13@xmath11 angular separation from the lmc center , and where the trend is attributed to the rotation curve of the lmc ( @xcite ; @xcite ; @xcite ) .
figure 17 shows the rv trend for the lmc disk ( solid line ) and halo ( dashed line ) from the best - fitted model to previously published outer lmc data by @xcite ( see their fig .
we show the trends at lmc position angles corresponding to our survey fields and to the 13@xmath11 limit of the model and previous data , as well as an extrapolation of the @xcite lmc rv trends to the carina field .
this figure suggests that the inner data follow the disk velocity trend , whereas the 332 km s@xmath9 moving group lies right on the extrapolation of the halo velocity trend to @xmath159 ( @xmath67 kpc ) radius from the lmc center . to further test an association of the magellanic clouds to the 332 km s@xmath9 group stars , we compare in figure 18 their distribution in the cmd and 2cd to that of stars found in the closest survey field to the lmc , shown by a green open circle in figure 16 .
the position of spectroscopically - confirmed lmc stars from this same inner rv survey field are shown by green open triangles in figure 18 .
figure 18a shows that the cmd position of the 332 km s@xmath9 group stars ( red solid circles ) is _ precisely _ where the locus of the lmc s prominent red clump slightly overlaps our carina rgb selection boundary .
moreover , inspection of our carina field sample of giant stars that fall outside our carina cmd selection region in figure 14a reveals : ( 1 ) a possible additional concentration of stars at @xmath160 just outside the carina selection boundary at the position of the lmc red clump seen in figure 18a ( although not stretching as blueward in fig .
14a because such stars are eliminated by the 2cd selection ) ; and ( 2 ) a slight excess of stars tracking the nominal position of the lmc rgb visible in figure 18a , above the carina rgb selection boundary . to test whether both of these groups of carina outliers
" may be magellanic in origin , on ut 2005 august 15 we observed two bright giant candidates in this lmc rgb position " of the cmd ( marked as solid squares in figure 18 ) using the mike spectrograph on the magellan telescope .
these turned out to have rvs ( 317 and 342 km s@xmath9 ) consistent with membership in the 332 km s@xmath9 group , which further vindicates a magellanic cloud provenance of this moving group .
km s@xmath9 , therefore we do not include them in the velocity dispersion calculation but only use them as membership information .
] comparison of figures 14a and 18a certainly evokes the notion of a diaphanous presence of lmc stars in the foreground of the carina dsph , which has given rise to the 332 km s@xmath9 group .
finally , within the giraffe rv dataset , we found four more stars with velocities matching the 332 km s@xmath9 group and positions in the cmd ( red open circles in figure 18 ) reasonably compatible with being lmc red clump stars . adding these four stars changes only marginally the mean velocity and the velocity dispersion of the moving group . with the possible connection to the magellanic clouds in mind , we can bring the abundance argument full circle to look for self - consistency of this hypothesis .
for example , if the originally identified 332 km s@xmath9 group members are parts of the red clump of the lmc , then for each star we can recalculate its [ fe / h ] from the infrared triplet strength assuming the @xmath161 of the lmc red clump .
the result yields a mean [ fe / h]=-0.67 ( with dispersion @xmath154 dex ) relatively more metal poor than , but still consistent with the mean metallicity ( [ fe / h]=@xmath162 ) of the dominant population of stars in the lmc found recently by @xcite using the same infrared triplet methodology . considering also that it would seem unlikely to find two such extreme velocity stellar systems at a similar distance and position in the sky
, the collective evidence compellingly suggests that we have found widely dispersed stars from one of the magellanic clouds the lmc being more likely in the foreground of the carina dsph .
as with the examples of the carina dsph explored earlier , and the ursa minor system explored in @xcite , the presence of extremely widely displaced , but satellite - associated stars would seem to have profound implications for the structure of the lmc .
one can use equation ( 45 ) from @xcite to estimate the mass of the lmc given a certain tidal radius . for our most widely separated star in the 332 km s@xmath9 group ( @xmath159 away from the center of the lmc ) to remain bound to the lmc implies a minimum lmc mass of @xmath163 m@xmath4 assuming a milky way mass interior to the lmc of @xmath164 m@xmath4 ( @xcite ; the @xcite model gives almost the identical milky way mass ) .
this inferred lmc mass is @xmath165 times more than that reported by @xcite ( @xmath166 m@xmath4 ) to a 13@xmath11 radius and consistent with the @xmath167 m@xmath4 lmc mass derived if we assume a flat lmc rotation curve to this distance .
the implied 20.2 kpc minimum tidal radius is now more than 33% ( 1@xmath55 ) greater than the @xmath168 kpc tidal radius estimated by @xcite .
these results immediately suggest two possible scenarios ( ignoring possible solutions offered by modified newtonian dynamics ; @xcite ; @xcite ) : ( 1 ) the lmc is substantially larger than previously appreciated .
the inferred total @xmath82 would exceed 10 in solar units .
an even larger mass is implied by the fact that the velocity dispersion of the 332 km s@xmath9 9.8 km s@xmath9 while @xmath169 smaller than the dispersions of tracers @xmath170 kpc from the lmc , as might be expected in the outer limits of a galaxy halo , are still quite larger than the expected , small asymptotic value at the edge " of a galaxy .
match well those found for the carbon stars summarized in figure 6 of @xcite .
] we note that an lmc extending out to @xmath67 kpc ( in the line of sight ) has been already proposed by @xcite based on the identification of a vertically extended red clump in the cmd of a field in the direction of the lmc .
( 2 ) the magellanic cloud stars we observe in the foreground are not bound to the lmc .
the colder dynamics of the 332 km s@xmath9 stars might be explained through a tidal debris origin .
but if unbound , stars in the direction of the carina dsph are _ not _ aligned with the expected direction of an lmc tidal tail , based on both the typical proper motions measured for the lmc ( summarized in table 1 of @xcite ) as well as the direction of the hi magellanic stream ( both the leading and trailing arms ) both lie in a roughly orthogonal direction .
this is not necessarily a problem , since stars will be tidally stripped anywhere along the satellite - milky way equipotential , whereas we have only explored one position angle from the lmc here . on the other hand , the carina survey field _
does _ happen to lie more or less along the axis defined by the lmc and smc .
a tidal disruption scenario involving an interaction of the lmc and smc might conceiveably throw magellanic stars out along this axis .
for example , the velocities of our magellanic giant stars are consistent with those of the carbon stars found by @xcite in the same general direction ( see figs . 16 and 17 ) , and which these authors attribute to a polar ring " of smc debris around the lmc .
alternatively , the widely separated lmc " stars may constitute residue from the disruption of a former greater magellanic galaxy " which has often been invoked as a possible explanation for the curious alignment of a number of milky way satellites and globular clusters along a magellanic plane " that also includes the hi magellanic stream ( @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite ) .
a dynamical association of ursa minor , draco , the lmc and the smc is suggested by their common motions along one great circle ( see , e.g. , fig . 3 of @xcite ) . were this group of milky way satellites truly daughters of the break up of a greater magellanic system or produced together as tidal dwarfs during a major merger with the milky way , their close , but not precise , alignment in a single plane might indicate the possibility of a potentially broad stellar swath of loosely coherent magellanic plane debris .
but if the 332 km s@xmath9 stars represent dynamically _ old _ tidal debris like this , one might not expect it to so well match the current distance of the lmc , nor its velocity ( or , even more coincidentally , the velocity extrapolated from the lmc velocity trend to this position in the sky ) . only with further surveying for additional 332 km s@xmath9 group " stars in other directions around the magellanic clouds can one hope to test such hypotheses .
we intend to explore these possibilities further elsewhere ( nidever et al .
, in preparation ) with a larger database of outer lmc stars collected over a larger area .
our survey for diffuse halo substructure in a large field around the carina dsph has yielded the following primary results on the structure of both the carina dsph and the lmc ( or magellanic clouds ) : using a combination of new washington+@xmath51 photometry and new echelle spectroscopy we have confirmed the existence of an extended , power law component in the density distribution of carina , which can be modeled as a king + power law " .
such density distributions are characteristic of those found in models of disrupted satellites and has also been observed in the tidally disrupting sgr dsph . with magellan+mike echelle spectroscopy of giant star candidates in the carina field we have establish the existence of carina stars to the limits of our photometric survey field , with confirmed carina members to at least 4.5@xmath2 , and likely 4.9@xmath2 . these detections represent the most widely separated stars ( in terms of @xmath2 ) found associated with any dsph ( apart from the sgr dsph ) to date . beyond verifying the existence of the extended carina population ,
these widely separated member stars have profound implications for the structure of carina : if the stars are bound , carina must have a minimum total @xmath82 of 6,300 in solar units , or 16,000 in the case of the 4.9@xmath2 example . with the addition of vlt+giraffe spectroscopic data and other published data ( @xcite ; @xcite ) within @xmath2 to our mike velocities at large @xmath2 , we have good and continuous sampling of the carina rv distribution to well past @xmath2 by 408 confirmed carina members . with these data
, we have rederived the central and global @xmath82 for carina , assuming a single - component model and using the core - fitting technique ; the results yield 43@xmath171 and 41@xmath172 ( m / l)@xmath4 respectively , where the main source of uncertainty comes from the luminosity .
these results are significantly at odds with the lower limits to the global @xmath82 found using the outlying carina members above .
with the extensive rv coverage we have also derived the line of sight radial velocity dispersion profile for carina to @xmath173 , the most extensive such profile so far ( by more than a factor of two ) for any dsph .
the profile is flat to past @xmath2 and then exhibits a rise in the dispersion to almost twice the inner value at @xmath174 .
such results are incompatible with completely bound , mass - follows - light dsph models , but also challenge two - component models that account for the flat dispersion via an extended dark halo surrounding the dsph . in the latter case an enormous halo
is needed , one significantly more massive than that implied above for simply keeping the @xmath175 carina stars bound , since the significant dispersion at large radius implies that the tidal radius is much farther out .
while with our new data we can not definitively rule out a _ very _ large , and extended dark halo for carina one producing a global @xmath82 approaching as much as 6,300 or more we conclude that a simpler , less contrived scenario that provides a good match to _ all _ available observations of carina is that it is tidally disrupting and we have identified some of its unbound stars .
this scenario simultaneously accounts for the following observed features of the carina system : ( 1 ) its
king+power law " density profile , which is a natural product of tidal disruption ; ( 2 ) the fact that the extended component of carina lies predominantly along its major axis and shows increasing ellipticity with radius , as would be expected in nascent tidal tails ; ( 3 ) carina stars extending from the core to the edge of the survey area ; ( 4 ) the flat , then rising velocity dispersion profile with radius ; and ( 5 ) a flattening of the rv distribution with radius , from gaussian in the core to platykurtic at large radius . explaining this combination of observed carina properties with extended dark halo scenarios will require substantial efforts to create successful ad hoc models . on the other hand ,
all of the above carina features not only resemble those seen in the established , tidally disrupting sgr dsph system , but have been well - matched by mass - follows - light models ( presented in a companion paper , muoz et al . 2006 ) of disrupting dsphs having the nominal central @xmath132 derived here .
finally , we have detected a second , strongly velocity - coherent structure in the carina field with even higher rv than carina .
the more metal rich stars constituting this moving group have cmd positions consistent with lmc red clump stars and their velocities follow the extrapolated velocity trend expected for lmc halo stars . with additional washington@xmath17 photometry and follow - up spectroscopy we have traced this population from 4@xmath11 separation from the center of the lmc out to the 22@xmath11 separation of the carina field center .
these stars either represent the detection of magellanic stellar tidal debris , or , if bound to the lmc , imply a significantly larger mass and tidal radius for the lmc than previously determined .
traditionally , debate over the kinematical and structural properties of the diffuse , low surface brightness dsphs has tended to polarize around two primary interpretations : ( 1 ) that dsphs are dark matter dominated ( e.g. , @xcite ) galaxies , with @xmath82 reaching to as much as 100 ( m / l)@xmath4 , making them structurally different compared to globular clusters and de systems . the prime observational evidence to support this claim is the relatively high measured central velocity dispersions that
coupled with an assumption of dynamical equilibrium imply masses far in excess of that inferred by the luminous component .
alternatively , ( 2 ) dsphs have also been discussed as systems partly or completely out of virial equilibrium ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) .
such an assertion seeks to explain the large central velocity dispersions of dsphs through inflation by tidal heating or other dynamical effects , allowing for much more modest dsph masses , consistent with no dark matter . to date , despite much observational and theoretical effort , the physical evidence has generally remained unpersuasive enough to dislodge the most ardent adherents to these models .
reinforcing viewpoints have been several all or nothing " notions introduced into the debate , including : ( 1 ) the assumption that dark matter dominated systems are in dynamical equilibrium throughout their entire physical extent ( e.g. , @xcite ; @xcite ) ; or , ( 2 ) if evidence of tidal stripping is found around a dsph , the system must be devoid of dark matter ( e.g. , @xcite ) .
remarkably , more recent work intended to _ clarify _ the physical nature of dsphs has , instead , increased the apparent gulf between diametrical viewpoints .
@xcite had previously suggested that only @xmath176 additional observations [ of dsph star rvs ] at 0.75 times the tidal radius
would be required to distinguish clearly between an mfl distribution and an extended halo or disrupted remnant model with a flat or radially rising velocity dispersion . " yet , despite the fact that the latest dsphs spectroscopic surveys have provided rvs of hundreds of dwarf members to beyond 0.75@xmath2 in several systems ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) , we are apparently no closer to a consensus view of dsph dynamics . while this is partly due to technical differences in interpretation of even the same databases ( e.g. , wilkinson et al .
2004 , okas et al .
2005 , muoz et al .
2005 ) , in general most studies are finding flat dsph velocity dispersion profiles to be the norm .
as discussed several times here , such profiles are produced naturally in dsph models undergoing tidal disruption ( see also @xcite ; @xcite ; @xcite , @xcite ; @xcite ; m06 ) .
however , rather than settling the issue , these rather flat velocity dispersion profiles have prompted the development of even more extreme , two - component , extended dark halo dsph models with _ substantially higher _ bound masses and total @xmath82 exceeding 400 or even 1000 @xmath177 ( e.g. , @xcite ; @xcite ; @xcite ) .
these models are partly motivated by the proposition that the so - called missing satellite problem of @xmath140cdm cosmologies could be alleviated if galactic dsph satellites inhabit the most massive sub - halos i.e. , m@xmath178 m@xmath4 , or equivalently v@xmath179 30 40 km s@xmath9 ( @xcite ; @xcite ) .
25 km s@xmath9 can succesfully reproduce the velocity dispersion profiles of these dsphs . ] from the numerous arguments laid out thus far , we are persuaded that the weight of evidence militates against the extreme halo hypothesis for carina in favor of a tidal disruption scenario . yet
one more argument favors the latter hypothesis : an extended dm halo of the magnitude our data would require in this scenario has ancillary implications for the _ chemical evolution _ of carina that are problematical . despite having a complex and episodic star formation history ,
carina has a relatively low mean metallicity of [ fe / h]@xmath180 ( @xcite ; @xcite ) .
@xcite note that galaxy masses of order a few times @xmath181 m@xmath4 consistent with the mass of carina derived from central velocity dispesion ( @xcite ; 4.4 ) are low enough to suffer metal - enriched winds , which promote preferential depletion of metals but retention of sufficient gas to allow further star formation at a continued , relatively low mean metallicity ( like carina s ) .
a larger galactic potential diminishes the possibility of blow - out / blow - away of either gas or metals ( e.g. , @xcite ; @xcite ; @xcite ) , leading to closed - box enrichment ( @xcite ) . but @xcite find that the metallicity distribution function of carina is not well matched by a closed - box model .
if carina has an enormous extended dm halo , it would have resulted in an enrichment that it is not observed ( @xcite ) . however , a more modest carina dark matter content is not discounted by this argument .
one of our goals in this paper ( see also @xcite ) has been to push the measurement of physical parameters in one dsph to hitherto unexplored regions to see if , in at least one system , new data in the extrema can definitively rein in the range of possible models .
we conclude that the new breed of extremely large @xmath82 , extended dark matter halos is less likely to apply to the present carina dsph than a tidal disruption scenario , which more readily explains all present observational data on the satellite.cdm if the models produce subhalos that are sufficiently stripped to reach the luminous matter . ]
that said , our results do not rule out _ any _ dark matter in the dsph , and , indeed , as we shall show in muoz et al .
( 2006 ) , an easily workable ( and therefore likely ) model for carina is one with elements of _ both _ of the originally debated dsph scenarios : a tidally disrupting , mass - follows - light dsph , but one with relatively high ( @xmath182 ) dark matter content , as suggested by the central velocity dispersion .
it has recently been claimed ( @xcite ) that `` sgr was a rare event , not a paradigm for the average '' .
this conclusion has been motivated by the notion that were systems like the carina dsph tidally disrupting , then the halo should have large numbers of youngish ( e.g. , blue main sequence ) stars in larger numbers than seen ( @xcite ) .
such an analysis presumes that the _ present _ carina system is representative of the typical stellar population that would have been contributed to the halo by tidally disrupting dsph systems _ including the former carina_. in contrast , as was previously demonstrated in @xcite , if for a hubble time carina were disrupting at the fractional mass loss rate implied by its density profile , i.e. @xmath183 gyr@xmath9 ( see 2.2 ) then far more stars from carina s oldest population would have been lost by now than from either the intermediate - aged or young populations in carina .
this is also a reasonable explanation for why the carina system today is dominated by it s intermediate - aged population , and even for why there seems to be a radial metallicity and age gradient in carina : the present balance of populations bound to carina likely reflects the competing interplay of star formation history and mass loss history ( @xcite ) in this disrupting analogue of the sgr dsph galaxy .
we appreciate useful discussions with david law and andrew mcwilliam .
we thank an anonymous referee for helpful suggestions that helped improve the paper and to carlton pryor for providing us with his code to calculate velocity dispersions using the maximim likelihood method .
we acknowledge funding by nsf grant ast-0307851 , nasa / jpl contract 1228235 , the david and lucile packard foundation , and the generous support of frank levinson and through the celerity foundation . additionally , p. m. f. is supported by the nasa graduate student researchers program , a university of virginia faculty senate dissertation - year fellowship , and by the virginia space grant consortium .
is supported by the arcs foundation , a university of virginia president s fellowship , and by the virginia space grant consortium .
frinchaboy , p. m. , muoz , r. r. , majewski , s. r. , friel , e. d. , phelps , r. l. , & kunkel , w. e. 2005 , in `` chemical abundances and mixing in stars in the milky way and its satellites '' , eds .
l. pasquini & s. randich , eso astrophysics symposia , _ in press _
( astro - ph/0411127 ) smecker - hane , t. a. , stetson , p. b. , hesser , j. e. , & vandenberg , d. a. 1996 , in `` from stars to galaxies : the impact of stellar physics on galaxy evolution '' , asp conference proceedins vol 98 , eds . c. leitherer , u. fritze - von alvensleben , & j. huchra ( san francisco : asp ) ,
328 l c c c c c c c c c c c c830356 & 6:35:11.95 & @xmath18451:25:05.1 & 27jan2004 & 18.84 & 1.30 & @xmath1840.00 & 328.0 & 119.7 & 6 & y & n + c2520066 & 6:36:40.86 & @xmath18451:58:07.0 & 27jan2004 & 19.04 & 1.39 & 0.02 & 213.9 & 5.3 & 7 & y & y + c2640634 & 6:38:22.77 & @xmath18451:11:00.4 & 27jan2004 & 18.08 & 1.62 & 0.01 & 221.2 & 12.2 & 7 & y & y + c2680057 & 6:38:36.82 & @xmath18451:16:23.9 & 28jan2004 & 18.33 & 1.57 & 0.03 & 222.1 & 13.1 & 2.6 & y & y + c2411078 & 6:38:47.04 & @xmath18450:50:31.2 & 28jan2004 & 18.65 & 1.42 & 0.04 & 229.4 & 20.4 & 4 & y & y + l c c c c c c c c c c c c2501778 & 6:38:31.92 & @xmath18451:07:04.8 & 19.76 & 1.22 & 0.01 & 230.3 & 21.3 & 2.9 & 2cd & y + c2680118 & 6:38:45.15 & @xmath18451:11:26.6 & 19.56 & 1.20 & 0.02 & 216.3 & 7.2 & 2.6 & y & y + c1400762 & 6:38:47.04 & @xmath18451:00:46.8 & 18.60 & 1.27 & @xmath1840.07 & 20.8 & @xmath184188.2 & 1.1 & cmd2 & n + c1401432 & 6:38:54.60 & @xmath18451:04:01.2 & 20.17 & 1.19 & 0.05 & 216.8 & 7.7 & 7.6 & y & y + c1402042 & 6:39:03.60 & @xmath18450:57:43.2 & 20.63 & 1.91 & @xmath1840.02 & 13.8 & @xmath184195.3 & 2.1 & cmd1 & n + c2413901 & 6:39:11.88 & @xmath18450:58:40.8 & 19.28 & 1.31 & @xmath1840.11 & 69.5 & @xmath184139.7 & 1.0 & n & n + c2413890 & 6:39:12.60 & @xmath18450:54:10.8 & 19.68 & 1.19 & 0.02 & 225.0 & 15.8 & 3.4 & y & y + | a new large - area washington @xmath0 filter survey of more than 10 deg@xmath1 around the carina dsph galaxy reveals a _ spectroscopically confirmed _ power law radial density break " population of carina giant stars extending several degrees beyond the central king profile .
magellan telescope mike spectroscopy establishes the existence of carina stars to at least 4.5 times its central king limiting radius , @xmath2 and primarily along carina s major axis . to keep these stars bound to
the dsph would require a global carina mass - to - light ratio of @xmath3 ( m / l)@xmath4 .
the mike velocities , supplemented with @xmath5 additional carina field velocities from archived vlt+giraffe spectra with @xmath6 , demonstrate a nearly constant carina velocity dispersion ( @xmath7 ) to just beyond @xmath8 , and both a rising @xmath7 and a velocity shear at still larger radii .
together , the observational evidence suggests that the discovered extended carina population represents tidal debris from the dsph .
of 65 giant candidates at large angular radii from the carina center for which mike spectra have been obtained 94% are associated either with carina or a second , newly discovered diffuse , but strongly radial velocity - coherent ( @xmath7=9.8 km s@xmath9 ) , foreground halo system .
the fifteen stars in this second , retrograde velocity population have ( 1 ) a mean metallicity @xmath10 dex higher than that of carina , and ( 2 ) colors and magnitudes consistent with the red clump of the large magellanic cloud ( lmc ) .
additional spectroscopy of giant star candidates in fields linking carina and the lmc show a smooth velocity gradient between the lmc and the retrograde carina moving group .
we conclude that we have found magellanic stars almost twice as far ( 22@xmath11 ) from the lmc center than previously known . |
we have entered into a new era , where synoptic photometry for millions of stars is currently being collected and analysed ( e.g. , ogle , gaia , wasp ) . in the relatively near future , lsst will obtain synoptic photometry for billions of stars . in order to make the most of those data , we need to develop automated routines to sift through enormous datasets in order to identify light curves of interest .
a precursor to that activity is to develop a better understanding of all the types of photometric variability present in real stars , and where possible to assign physical mechanisms to those variability classes .
the _ kepler _ mission provided one very well characterized training set of light curves for @xmath4100,000 stars . however , because it was tied to a single position in the sky and the target stars were primarily chosen for one science goal , the _ kepler _ data only provide templates for a limited range of stellar variability types .
k2 _ mission ( howell et al .
2014 ) is providing a greatly augmented set of data by observing a new group of @xmath220000 stars every three months , with those stars having a much wider range of characteristics .
one specific variability type for which _
k2 _ is providing a wealth of new data is young stellar objects ( ysos ) . during campaign 2 of the _ k2 _
mission , nearly 1500 candidate members of the star - forming regions comprised of the @xmath0 oph dark - cloud complex and the upper scorpius association were targeted . the data from that 78 day campaign form the largest set of precision yso optical light curves that have ever been obtained .
those light curves provide a hunting ground for the search for newly formed planets via their transits ( david et al .
2016a ; mann et al . 2016 ) , for the identification of yso eclipsing binaries ( alonso et al .
2015 ; david et al . 2016b
; kraus et al . 2015 ; lodieu et al .
2015 ) , and for the determination of the rotational periods for a large sample of low mass stars of 5 - 10 myr age ( rebull et al .
, in preparation ) .
a small fraction of the upper sco ysos have ir excesses and are likely classical t tauri stars ( ctts ) ; a sub - group of burster " lightcurves associated with some of these accreting systems is discussed in cody et al .
( 2017a ) .
the discussion of our process for identifying a best set of probable upper sco members , and the categorization of light curve morphologies for the disked stars is provided in cody et al .
( 2017b , in preparation ) .
our sorting of those members into stars with primordial disks and those without disks ( weak - lined t tauri stars , wtts ) is described briefly below , and more fully in cody etal ( 2017b ) .
about 1100 of the upper sco and @xmath0 oph stars appear not to have infrared ( ir ) excesses and presumably have shed their primordial circumstellar disks .
in the course of our processing and analysis of those light curves , the vast majority were identified as showing periodic variability characteristic of stars with dark , non - axisymmetrically distributed spots .
a handful were found to be eclipsing binaries , and another handful ( all relatively high mass stars ) are pulsators . however , twenty - three of the stars showed periodic photometric variability which did not fit into any of those categories .
it is those stars that are the subject of this paper .
all of these stars are m dwarfs , with most having spectral types between m4 and m6 .
all have short periods more than half less than one day , and all but one less than two days . in
2 , we describe the sources of data for this paper . our process for identifying unusual light curve classes , and the working definitions we have developed for our three classes is provided in 3 .
the empirical characteristics of the variability for each of the three classes is laid out in 4 - 6 .
for the first three _ k2 _ campaigns , the _ k2 _ science center did not produce detrended light curves .
instead , what was provided was the raw ccd pixel data for each star at each epoch .
we ( cody et al.2017a ) therefore created our own detrended light curves .
we produced two primary versions
one where the aperture centers were held fixed over the course of the _ k2 _ campaign , and another where a new center was derived at each epoch . for each flavor of reduction ,
photometry was produced with 2 and 3 pixel radius apertures . for each of these four light curve versions
, we then applied a version of the detrending routine described in aigrain et al .
( 2016 ) to remove the short - term , non - astrophysical trends in the data .
after we had begun writing this paper , two other light curve reductions became publicly available for campaign 2 , the vanderburg reduction ( vanderburg & johnson 2014 ) and the everest reduction ( luger et al .
we downloaded light curves for each of the stars that we considered as possibly relevant to this paper . for each star addressed in this paper , we examined all of the light curve variants that we had available , and chose the one which appeared to have the least non - astrophysical contribution ( when several variants were about the same , we simply adopted our own detrended light curve ) . in all cases , the unusual features which caused us to include
the star in our paper were present in multiple light curve variants .
in a few cases , one of the light curve variants did not have the same signature .
we believe this indicates that the detrending algorithm employed in that version removed some real astrophysical signature , or that the aperture size used in that version or the centering algorithm in that version resulted in the light curve being dominated by light from a different star .
we made special effort in these cases to closely examine all the available data ( finding charts , literature , all light curve variants and different aperture sizes ) in order to make sure that the light curve we present here is the best available for that star .
the ecliptic plane input catalog ( epic ; huber et al .
2016 ) catalog available at mast provides accurate coordinates for all of the stars in our sample . using those coordinates , we downloaded all available near and mid - ir photometry for our stars from the 2mass ( skrutskie et al .
2006 ) , wise ( wright et al .
2010 ) , spitzer ( werner et al .
2004 ) seip and feps ( meyer et al . 2006 ) , akari ( murakami et al . 2007 ) , and sdss ( e.g. , ahn et al . 2014 ) archives as well as @xmath5 and @xmath6 band photometry from apass ( henden & munari 2014 ) .
we also downloaded @xmath7 photometry for most of our stars from the pan - starrs database ( flewelling et al .
spectral types were available in the literature ( preibisch et al .
2002 , erickson et al .
2011 , lodieu et al .
2011 , luhman & mamajek 2012 , rizzuto et al .
2015 ) for 80% of our stars ; for most of the stars , those same references also provided h@xmath3 and lithium ( 6708 ) equivalent widths .
we have used our own keck hires spectra to provide a small amount of new spectral information , as described in the appendix .
one goal for the ancillary data was to provide a color - magnitude diagram which would include all the stars in our sample , and for that we decided to attempt to determine estimated @xmath8 colors for every star . because only a small fraction of the upper sco members have published @xmath6 band photometry , this required transforming photometry or colors obtained at other optical bands .
as the first step in that process , we downloaded @xmath9 band photometry for all of our stars from the newly released gaia dr1 archive ( gaia collaboration 2016 ) .
we then derived a formula to convert @xmath10 colors to @xmath8 colors based on a large sample of low mass stars with published @xmath8 colors in praesepe ( rebull et al .
2017 , submitted ) . for the stars lacking gaia photometry
, we obtained @xmath11 magnitudes from the deep near - infrared survey of the southern sky ( denis ; epchtein et al .
1997 ) and derived a conversion formula in a similar way between @xmath12 and @xmath8 .
we used all of the available photometry and spectral types to produce spectral energy distributions ( seds ) for each star ( the individual seds are shown in the appendix ) . from those seds , we identified stars with possible ir excesses . for each of the sources
thought to have an ir excess , we looked at the images , where available , to assess the quality of the photometry .
the longest wavelength data we have is either wise 22 @xmath13 m or , for just four stars spitzer 24 @xmath13 m . in several cases , the 22 @xmath13 m point from wise ,
in particular , was not a secure detection , and the apparent ir excess therefore vanished .
only one star in our final sample has a relatively secure , though small ir excess ( only at 24 @xmath13 m ) , and that is epic 205024957 .
carpenter et al .
( 2009 ) also determined that epic 205024957 has a 24 @xmath13 m excess and identified it as a debris disk .
for the stars where our longest wavelength detection is at 12 @xmath13 m , the lack of an excess at that wavelength precludes them from having primordial disks . because of the paucity of good data longward of 12 @xmath13 m for most of our stars
, a larger fraction could harbor debris disks .
the first step in our sample selection was to eliminate stars with active accretion ( ctts ) from the upper sco and @xmath0 oph _ k2 _ sample ,
since the accretion process in ctts and interactions with the primordial circumstellar disk can produce unusual light curves of many types , and addressing those types of variability was beyond the scope we wanted here ( but see cody et al .
none of the stars we retained for this paper have seds consistent with class i or ii primordial disks ( see the sed plots in the appendix , and compare , for example , to the seds of class i and ii sources in upper sco shown in cody et al .
2017a or similar sources in taurus as shown in rebull et al .
two of the stars are considered as @xmath0 oph members in the literature ( epic 203962559 and epic 203927435 ) ; all of the others are considered as members of upper sco .
once we had identified the full set of upper sco weak - lined t tauri ( wtts ) members with _ k2 _
data , three of the authors ( tjd , lmr , amc ) independently examined the data for each star , including each of the light curve variants described above .
they also independently searched for periodicities in those light curves and examined the phased light curves .
those efforts were aimed primarily at identifying possible planet transits , or eclipsing binary candidates , or variability related to accretion or circumstellar disk structure , or in compiling stellar rotation periods . during the course of those efforts ,
a few dozen stars were identified as having light curves that were not easily explained by any of those mechanisms . in those cases ,
further analysis was done and other members of the team were brought in to perform more manual detrending of the light curves .
this more detailed analysis resulted in some of these light curves being ascribed to artifacts in the data or to allowing them to fit into one of the established light curve classes .
however , 23 of these stars survived .
these stars are listed in table 1 .
we have sorted our unusual light curves into three classes in order to facilitate our discussion of their properties .
we acknowledge that this sorting is probably imperfect , and that the boundaries between classes are indistinct .
an alternative view is that all of our stars belong to a single group , and their differing light curve morphologies arise due to a range in our view angle to their rotation axes or to a range in their rotation rates ( e.g. , because that affects the location of the keplerian co - rotation radius relative to the sublimation radius ) or to some other hidden parameter . after assembling the data we had available
, we nevertheless felt it was helpful to sort the stars into the three groups in order to emphasize the shared properties within each group .
table 1 provides a list of all 23 stars we have identified as belonging to these three classes , along with some of their overall defining properties .
spectral types are known for 20 of these stars ; all are m dwarfs with 17 of the 20 having spectral types between m4.0 and m5.5 .
most of the upper sco _
k2 _ sample are m dwarfs , but the stars in table 1 are predominantly weighted towards later m sub - types ( this is also reflected in their cmd locations
see figure [ fig : uscocmd.pvmk ] ) .
eighteen of the stars in table 1 have known h@xmath3 equivalent widths , all but one of which have equivalent widths appropriate for wtts ( @xmath143.0 @xmath1 eqw @xmath1 @xmath1416 ) .
thirteen of the stars in table 1 have measured li 6708 equivalent widths , all of which are @xmath2 0.5 , also as appropriate for wtts mid - to - late m dwarfs ( the spectral types for two stars and the lithium data for six stars are from our own hires spectra , as described in the appendix ) .
eleven of 23 ( 48% ) of the stars in table 1 have two lomb - scargle periods ( i.e. are apparent binaries ) , which seems high . from the full sample of _ k2
_ m dwarf wtts with 2 @xmath1 p @xmath1 8 days ( @xmath2280 stars ) , only 16% are apparent binaries by the same measure .
however , for the @xmath2140 _ k2 _
m dwarfs with p @xmath1 1 day , 38% are apparent binaries .
thus , the stars in table 1 may have a high binary fraction simply because the rapidly rotating low mass stars in upper sco are often binaries .
this is consistent with the conclusion from stauffer et al .
( 2016 ) that m dwarfs in the pleiades with two lomb - scargle peaks were both more rapidly rotating and the two periods were closer together than would have been true if their periods were drawn at random from the single star population .
figure [ fig : prototype_light_curves ] shows phased light curves for two examples of each of the three classes ; phased light curves for all members of the scallop - shell and persistent flux dip classes are provided in the appendix , along with a description of the processing we have done to the original light curves .
the class definitions are as follows : * * scallop - shells * phased light curves that look like the rim of a scallop shell .
they have many wave - shaped undulations in their phased light curves .
usually the shape remains nearly constant for the entire _
k2 _ campaign , but in some cases it changes over time .
peak to trough amplitudes from a few percent to more than 10% ; periods always less than 0.65 days .
eleven of our upper sco stars fall in this category . *
* persistent flux dip class * discrete , triangularly shaped flux dips , generally covering less than 0.15 in phase ( for each of up to 4 dips ) .
the flux dips are often more or less stable in depth over the _ k2 _ campaign , with dip depths generally a few percent . outside of the dips , the phased light curves are either nearly flat or consistent with sinusoidal variation due to cool starspots .
periods range from 0.5 to 1.8 days .
a half dozen rapidly rotating , late m dwarf members of the pleiades with this light curve signature were identified from _ k2 _ campaign 4 data ( rebull et al .
2016 , 3.6 ; collier cameron et al .
eight of our upper sco stars fall in this category . * * transient , narrow dip class * more or less triangularly shaped flux dips , narrow in phase , but always variable in depth , on both short and long timescales , over the 78 day _ k2 _ campaign . only one prominent dip present in the phased light curve .
dip depths generally a few percent , but up to 20% for epic 205483258 ( rik-210 ) .
all four stars also show a spotted - star waveform , whose period is identical to the dip period within our uncertainties .
periods ranging from 1 to 5.5 days . in the next sections ,
we discuss specific examples of these categories .
lccccccccccccc 204918279 & rik-21 & 239.1046 & -20.2711 & 9.86 & 6.90d & 0.38 & 12 & m5.0 & -10.2 & 0.61 & 0.4594 & 0.4665 * & 1 + 204066898 & uscoctio 80a & 239.6509 & -23.8005 & 10.19 & 5.59 & 0.27 & 12 & m3 & -9.9 & & 0.3956 * & 0.5386 & 1 + 203462615 & rik-42 & 239.9086 & -26.0565 & 10.25 & 6.13 & 0.47 & 12 & m5.5 & -16.2 & 0.81 & 0.5201 * & 0.4421 & 1 + 204897050 & uscoctio 56 & 240.4208 & -20.3689 & 10.86 & 7.16d & 0.42 & 12 & m5 & & & 0.2639 & & 1 + 202724025 & rik-90 & 242.2373 & -28.5993 & 9.63 & 5.78 & 0.33 & 12 & m4.5 & -10.2 & 0.21 & 0.2595 * & 0.2795 & 1 + 204117263 & lhj-65 & 242.7589 & -23.5973 & 10.95 & 6.42 & 0.35 & 12 & m5 & -11.6 & & 0.6423 & & 1 + 204367193 & lhj-77 & 242.9766 & -22.6137 & 13.30 & 8.33 & & 4.6 & m6.3 & -10.3 & & 0.4835 & & 1 + 203534383 & & 243.5125 & -25.8148 & 11.71 & 7.73 & 0.73 & 12 & m4 : & -9.5 & & 0.2784 * & 0.3234 & 1 + 205110559 & & 243.8322 & -19.3520 & 10.46 & 7.74 & 0.61 & 12 & & & & 0.4031 & & 1 + 203050730 & rik-246 & 247.7741 & -27.4295 & 10.53 & 5.87 & 0.53 & 12 & m4.5 & -9.1 & 0.38 & 0.4865 * & 0.7740 & 1 + 203185083 & rik-253 & 248.6464 & -26.9675 & 10.48 & 6.62 & 0.18 & 12 & m4.5 & -27.1 & 0.54 & 0.4400 & & 1 + + 204364515 & & 240.3398 & -22.6240 & 10.05 & 6.84 & 0.34 & 24 & m4 & -8.4 & 0.55 & 3.0863 & 1.456 * & 2 + 203849738 & rik-100 & 242.4703 & -24.6982 & 10.93 & 6.69d & 0.37 & 12 & m5.5 & -15.1 & 0.40 & 0.6190 & & 2 + 203692610 & & 242.6318 & -25.2671 & 11.60 & 5.92 & 0.28 & 12 & m4 & -3.0 & 0.54 & 1.821 & & 2 + 205374937 & & 242.8255 & -17.9580 & 9.33 & 6.35 & 0.34 & 24 & m4 & -4.8 & 0.55 & 0.6345 * & 0.5436 & 2 + 202873945 & & 245.4164 & -28.0518 & 11.24 & 8.21 & 0.75 & 12 & & & & 0.6258 & & 2 + 204296148 & & 246.2158 & -22.8952 & 10.99 & 8.06k & 0.10 & 12 & & & & 0.5314 * & 0.4717 & 2 + 203962559 & aoc 64 & 246.7103 & -24.2312 & 10.80 & 8.83d & & 8 & m4 & -6.0 & & 1.5402 & &
2 + 203927435 & wmr 2 - 23 & 247.1794 & -24.3812 & 10.14 & 9.11d & 0.13 & 12 & m4.5 & -6.8 & 0.76 & 0.4820 * & 0.4162 & 2 + + 205024957 & & 242.5459 & -19.7678 & 11.38 & 6.53 & 0.46 & 24@xmath15 & m5 & -4.4 & 0.56 & 1.6656 & & 3 + 205046529 & & 242.6099 & -19.6642 & 10.40 & 6.52k & 0.35 & 24 & m4 & -4.4 & 0.54 & 2.5619 & 1.8358 * & 3 + 204143627 & lhj-120 & 243.9969 & -23.4934 & 11.31 & 6.29 & 0.13 & 12 & m5 & & & 1.125 & & 3 + 205483258 & rik-210 & 245.8523 & -17.2909 & 9.65 & 5.04 & 0.37 & 12 & m2.5 & -5.3 & 0.59 & 5.667 & & 3 +
as illustrated in the top row of figure [ fig : prototype_light_curves ] , these stars show complex phased light curves , with many local maxima and minima . for half of these stars ,
their waveform is relatively stable for the entire 78 day _ k2 _ campaign .
however , for five of the eleven stars we put in this category , there is distinct evolution in the waveform within the campaign . because we believe the nature of this evolution provides important clues as to the physical mechanism responsible for creating this light curve morphology
, we discuss in detail these five stars and their light curve evolution during the campaign .
first , however , we illustrate the class by describing the light curve of one of the six members whose phased light curve shape does not change appreciably over the duration of the _ k2 _ campaign .
epic 204066898 ( a.k.a .
uscoctio 80a ; kraus & hillenbrand 2009=kh2009 ; kraus & hillenbrand 2012=kh2012 ) is an m3 member of upper sco with no apparent ir excess .
it has been identified as a wide binary , with separation about 13@xmath16 and @xmath17 = 1.9 mag ( kraus & hillenbrand 2009=kh2009 ) .
kh2012 also identify uscoctio 80a as a close ( 0.05@xmath16 ) , nearly equal mass binary ( see their table 3 ) .
we find two independent peaks in the lomb - scargle periodogram with periods of 0.3956 and 0.5386 days .
we have experimented with one pixel and 1.5 pixel apertures for the light curve extraction in order to assess the contribution of uscoctio 80b to the epic 204066898 light curve ; our conclusion is that the two lomb - scargle peaks are both correctly associated with uscoctio 80a ( uscoctio 80b is also epic 204066097 , for which we derive a period of 0.36d ) .
we thus assume the two stars associated with the two lomb - scargle peaks are also the close binary pair identified in kh2012 .
the longer period system for epic 204066898 ( which we designate as the secondary component ) has a more or less sinusoidal light curve with about 1% peak - to - trough amplitude . after removing the signal from this system and phasing the remaining signal to the 0.3956 day period of the primary component ,
the light curve shape for this star ( figure [ fig : epic204066898 ] ) is fairly typical of the scallop - shell class , with a full amplitude of about 5% and multiple peaks and troughs in its phased light curve .
there are no published spectra for epic 205110559 , but its colors suggest it is a mid to late m dwarf .
its light curve ( figure [ fig : epic205110559 ] ) is also fairly typical of the scallop - shell class , with a full amplitude of about 5% and multiple peaks and troughs in its phased light curve . for phase 0.45 to 1.0
, the waveform is relatively stable over the duration of the campaign .
however , as emphasized in the second panel of figure [ fig : epic205110559 ] , the waveform shows considerable evolution over the campaign for the phase range 0 to 0.45 . the primary shift in waveform shape appears to occur almost exactly at the boundary between the first window ( day 2060 - 2099 ) and the second window ( day 2099.2 - 2107.5 ) .
the third panel of figure [ fig : epic205110559 ] shows an expanded view of the timeframe when the waveform changed state , which reveals that a strong flare occurred at day 2099 .
this is by far the strongest flare for this star during the _ k2 _ campaign , and it occurs at exactly the point that the waveform made its state transition .
this latter point is illustrated in the last panel in figure [ fig : epic205110559 ] .
the excess " flux at phase @xmath20.15 gradually declines , returning to approximately the pre - flare level by day @xmath22128 .
we conclude that there is very likely a physical link between the occurrence of the flare and the change in morphology of the phased light curve .
the _ k2 _ light curve for epic 204918279 has a beating appearance and the lomb - scargle periodogram shows two strong peaks at very similar periods , 0.4594 and 0.4665 days , which we interpret as indicating it is a binary with the two components having nearly the same period . after separating the variability at those two periods , we find that the star with @xmath18=0.4665 days has a light curve shape which places it in the scallop - shell category at least for the first @xmath260 days of the _ k2 _ campaign , as shown in the first panel of figure [ fig : epic204918279 ] .
the phased light curve changed abruptly in shape at day @xmath22123.5 , as illustrated in the second and third panels of figure [ fig : epic204918279 ] .
the state transition of the light curve shape again appears to have been triggered by a flare ; see the fourth panel of figure [ fig : epic204918279 ] . in this case , this is a much weaker flare than for epic 205110559=0.4665 days , we believe it is highly probable the flare originated on that star . ] .
nevertheless , this was the strongest flare seen in the _ k2 _ light curve for epic 204918279 , at least during the last 60 days of the campaign . prior to and after the flare
, the light curve shape appears quite stable during the _ k2 _ campaign .
epic 203185083 is an m4.5 yso upper sco member with unusually strong h@xmath3 emission ( equivalent width = @xmath1427 ) for a wtts of that spectral type ( rizzuto et al .
however , it appears to have no ir excess , so we consider it to be a wtts .
the lomb - scargle periodogram indicates a period of 0.4400 days . after phasing the light curve to the 0.44 day period
, we find a waveform shape which is not plausibly explained by cool spots and which seems to evolve over the duration of the _ k2 _ campaign .
after some investigation , we find the waveform evolution basically occurs at two rapid transitions on day 2093 and day 2102 , as shown in the left panel of figure [ fig : epic203185083 ] .
the first state transition of the light curve shape again appears to have been triggered by a flare ( right panel of figure [ fig : epic203185083 ] ) , though again a much weaker flare than for epic 205110559 .
nevertheless , this was the strongest flare seen in the _ k2 _ light curve for epic 203185083 .
the second state transition appears to happen without any obvious trigger as seen from our vantage point .
finally , for both epic 203185083 and epic 204918279 , the flares at the state - transition seem unusual in that they have nearly symmetric shape ( rather than a rapid rise and slow decay as true for most flares and for the flare on epic 205110559 ) .
the lomb - scargle periodogram for epic 204117263 shows a single strong peak at @xmath18=0.6423 days ; when phased to that period , the light curve has a scallop - shell appearance for the entire campaign .
there was , however , a small , abrupt change in the phased light curve shape at day 2080.0 , as illustrated in figure [ fig : epic204117263 ] . in this case
, the state transition in the light curve shape does not appear to be associated with a flare ( at least on the side of the star that is facing us ) .
the lomb - scargle periodogram for epic 204897050 shows a dominant strong peak at a very short period of @xmath18=0.2639 days ; when phased to that period , the light curve has a scallop - shell appearance for the entire campaign .
as illustrated in figure [ fig : epic204897050 ] , there is a small portion of the phased light curve around phase=0.6 where the waveform shape gradually changes throughout the length of the campaign .
this could be consistent with the other three stars discussed above , if there was a trigger event just prior to the beginning of campaign 2 ; no significant flares are seen during the time period of our light curve .
the stars having scallop - shell light curves have many shared characteristics , which presumably provide clues to the physical process driving their unusual light curve morphology .
below we collect the most salient of those shared characteristics : * all have periods less than 0.65 days . we do not know that this is , in fact , the stellar rotation period , but we will assume that is the case . with that assumption ,
their periods place them among the most rapidly rotating m dwarfs in upper sco ( see also 7 ) . * all
have mid to late m spectral types ( m4 to m6.5 ) .
* they have full amplitudes for their _ k2 _ light curves of 2% to 10% . *
six of eleven of them show relatively stable phased light curve shapes over the entire 78 day _
k2 _ campaign . *
five of them show evolving waveforms . in all but one case
, there is an abrupt transition from one waveform to another with the switch happening in generally less than a day . in three cases ,
the phase change takes place when a strong flare occurs . in four of the five cases
, the change in shape only significantly affects 20 to 30% of the waveform , with the remaining waveform staying essentially constant in shape . for epic 205110559 ,
the fractional flux change between states at phase=0.1 is of order 4% ( figure [ fig : epic205110559 ] , second panel ) . any physical mechanism to explain this light curve morphology must be able to accomodate events of this size . for all three events with triggering flares , the affected portion of the light curve brightened after the flare .
* the number of distinct peaks ( or troughs ) in their phased light curves ranges from 3 to 6 .
* there are 83 wtts in our sample with p @xmath1 0.65 days and @xmath19 5 .
the fraction that have scallop - shell light curve morphologies for that period range is therefore 13% .
* we have also examined this wtt control group for obvious flares .
based on a simple visual check of the detrended light curves of our table 1 stars and our wtt control group , we find no obvious difference in either the frequency or strength of flares .
there are eight stars in table [ tab : basicdata ] in this category .
the phased light curves of these stars are dominated by shallow , triangularly shaped flux dips , superposed either on a flat continuum " or on a waveform consistent with that due to rotational modulation of cool spots at the same period as the flux dips .
four of these stars show little or no evolution in their phased light curve morphology over the time frame of the _ k2 _ campaign ; we describe one of these four stars ( epic 205374937 ) in detail .
the other four members of the class show demonstrable evolution in their light curve morphology over the _ k2 _ campaign , and we discuss each case individually in order to illustrate the nature of this evolution .
preibisch et al .
( 2001 ) first identified epic 205374937 as a low mass member of upper sco .
carpenter et al .
( 2009 ) and luhman & mamajek ( 2012 ) both reported no ir excess . using gemini north adaptive optics ( ao ) imaging , lafreniere et al .
( 2014 ) identified epic 205374937 as a nearly equal mass binary with a 0.1@xmath16 separation .
the _ k2 _ light curve shows two approximately equal strength close peaks , corresponding to periods of 0.6345 and 0.5436 days .
we assume the two nearly equal mass components seen with the ao imaging give rise to these two periodogram peaks ; we somewhat arbitrarily assign the the larger periodogram peak as the a component .
the phased light curve for the a component is shown in figure [ fig : epic205374937 ] .
we interpret the phased light curve for epic 205374937a as the superposition of a more or less sinusoidal variation due to star spots and two ( or possibly three ) short - duration flux dips .
the two most prominent flux dips are centered at phases about 0.2 and 0.6 ; the full width at zero intensity ( fwzi ) of the dips are 0.16 and 0.12 in phase ( or about 2 hours in time ) and their depths are about 3% of the continuum .
there is no suggestion of the dips varying in depth or width over the course of the campaign ( that is , the noise level in the phased light curve appears the same inside and outside the dips ) .
the period that phases the dips best also produces the least scatter in the rest of the light curve , indicating that the dip period is essentially the same as the spot period .
two of the other stars in this group ( epic 204296148 and 203849738 ) also have both spotted star waveforms and persistent flux dips where the periods are the same ( see figure [ fig : batwing_medians ] ) .
epic 203692610 is an essentially anonymous , red star with proper motion consistent with upper sco membership .
our own keck hires spectra for this star shows it to have a spectral type of about m4 , with weak h@xmath3 emission ( equivalent width = @xmath143.0 ) , and strong lithium absorption ( equivalent width = 0.54 ) , typical of upper sco m dwarfs .
the lomb - scargle periodogram for its _ k2 _ light curve shows a single strong peak at @xmath18 = 1.821 days away , we have analysed a light curve derived using a 1.5 pixel aperture for epic 203692610 , rather than the 2 or 3 pixel radius used throughout the rest of the paper . ] .
when the light curve is phased to that period , it shows a relatively flat continuum with three sharp , triangular shaped flux dips ( figure [ fig : epic203692610 ] ) .
the dip depths are about 3% to 7% , and the fwzi for each dip is about 0.1 in phase .
close examination of the light curve shows that the dip depths vary in time , with the time dependence being somewhat different for each dip , as illustrated in the top - right panel of figure [ fig : epic203692610 ] .
the left - most dip is more or less stable in its depth until day 2119 , when it essentially disappears .
the central dip is more or less stable in depth until day 2131 , when it too disappears .
the strongest ( right - most ) dip is quite deep until day 2089 ; from 2089 to 2131 it is about half as deep ( and somewhat shifted in phase ) , and at day 2131 it becomes significantly weaker again .
there are no strong flares in the _ k2 _ light curve for this star .
however , the strongest flare - like event ( or events ) occurs at the time of the first state - transition of the flux dips at day 2089 , as illustrated in the bottom - left panel of figure [ fig : epic203692610 ] .
there are no obvious flares of any sort at the other two state - transitions .
finally , we note that the dip properties and the general appearance of the phased light curve for epic 203692610 are quite similar to the _ k2 _ light curve for hhj 135 , one of the six late m dwarfs with persistent , short - duration flux dips identified in the pleiades by rebull et al .
( 2016 ) . while epic 203962559 has been identified as a @xmath0 oph member ( wilking et al . 2005 ) ; it appears to have no ir excess and has relatively weak h@xmath3 emission , from which we conclude it is a wtts .
the lomb - scargle periodogram shows one strong , real peak at p = 1.5402d .
when phased to that period and examined closely , the light curve appears to undergo a major change in shape at about day 2106 .
this is illustrated in figure [ fig : epic203962559 ] .
prior to that day , there were four well - defined flux dips in the light curve ; after that day , only two flux dips remain , and one of those is significantly narrower and less deep .
we are unable to determine if there was a flare at day@xmath22106 because that date also is when mars transited the _
k2 _ fov ( near the position of epic 203962559 ) , corrupting its light curve for @xmath210 hours .
we can think of no possible way for the mars transit to induce systematic flux errors in a specific phase range in the extracted light curve for this star , so we assume this was a bizarre coincidence .
epic 203927435 has also been identified as a @xmath0 oph member ( wilking et al.2005 ) ; it too appears to have no ir excess and relatively weak h@xmath3 emission ( equivalent width @xmath146.8 ) .
the lomb - scargle periodogram shows two resolved peaks , with periods of 0.4817 and 0.4162 days . after removing the signal from the star with @xmath18 = 0.4162 days ( which we refer to as the b component ) , we find that the phased light curve for the a component shows a waveform which is not explainable by cold spots , but that also shows some evolution during the _ k2 _ campaign .
the first panel of figure [ fig : epic203927435 ] shows the phased light curve for the first 63 days of the campaign ; the light curve for the last 13 days of the campaign is shown in the second panel of figure [ fig : epic203927435 ] .
a very strong flare occurred at the transition point
third panel of figure [ fig : epic203927435 ] on day 2123.3 ; this was by far the strongest flare for this star during the _ k2 _ campaign .
prior to that day , there were four well - defined flux dips in the light curve ; after the flare , only one ( weaker ) dip remains . in this case , the transition between states took place over about 3 days . during the middle of that transition period , a second weaker flare - like event occurred ;
see the last panel of figure [ fig : epic203927435 ] .
that event , however , has a symmetrical shape ( as was the case for the triggering events for epic 204918279 and epic 203185083 ) .
bouy et al .
( 2006 ) showed that epic 204364515 is a possible resolved triple system , consisting of a close , equal brightness pair with a 0.15@xmath16 separation and a fainter possible companion at a separation of about 1@xmath16 ( with its @xmath24@xmath16 pixels , the _ k2 _ light curve includes light from all three stars ) .
the lomb - scargle periodogram for epic 204364515 shows two dominant periods of 3.0863 and 1.456 days .
after removal of the light from the star with @xmath18=3.086 days , the phased light curve for the star with @xmath18=1.456 days shows four well - defined flux dips .
figure [ fig : epic204364515b ] shows the phased light curve , where we have split the data into two windows of time ( black for day 2060 - 2105 ; red for day 2105 - 2139 ) .
the two strongest flux dips appear to be approximately stable in position , shape and depth .
the flux dip at phase @xmath20.2 , however , appears to shift in phase by about 0.03 in phase ( 10 degrees in azimuth ) between the two windows .
the fourth flux dip at phase @xmath20.75 appears to become considerably deeper in the second time window .
we have measured widths and depths for all the prominent flux dips in these eight stars ; see table [ tab : persistentdata ] .
the widths we provide are full - width , zero - intensity values , fwzi , in units of fraction of the period meaning that they are intended to indicate the entire duration of the event ( rather than , for example , a fwhm ) .
most of the dips have fwzi of 0.1 to 0.15 in phase ; for a typical star with @xmath18 = 0.6 days , a fwzi of 0.12 corresponds to about 1.7 hours .
the depths are the difference in counts between a nearby continuum " level and the bottom of the dip relative to the continuum level , expressed as a per cent .
most of the dips have depths of 1% to 3.5% , with the strongest being 7% ( the deepest dip for epic 203692610 ) .
lcccc 204296148a & 1 & 0.32 & 0.16 & 3.0 + 204296148a & 2 & 0.62 & 0.11 & 1.4 + 204364515b & 1 & 0.19 & 0.12 & 1.5 + 204364515b & 2 & 0.38 & 0.11 & 3.1 + 204364515b & 3 & 0.62 & 0.11 & 1.9 + 204364515b & 4 & 0.74 & 0.07 & 0.9 + 202873945 & 1 & 0.10 & 0.09 & 0.8 + 202873945 & 2 & 0.42 & 0.10 & 2.3 + 202873945 & 3 & 0.63 & 0.09 & 1.2 + 203962559w1 & 1 & 0.37 & 0.08 & 3.2 + 203962559w1 & 2 & 0.52 & 0.06 & 0.9 + 203962559w1 & 3 & 0.67 & 0.11 & 3.1 + 203962559w1 & 4 & 0.87 & 0.15 & 3.5 + 203962559w2 & 1 & 0.65 & 0.12 & 3.2 + 203962559w2 & 2 & 0.85 & 0.15 & 2.2 + 203692610 & 1 & 0.19 & 0.06 & 2.5 + 203692610 & 2 & 0.29 & 0.06 & 2.5 + 203692610 & 3 & 0.41 & 0.13 & 7.5 + 203849738 & 1 & 0.13 & 0.19 & 3.9 + 203849738 & 2 & 0.79 & 0.07 & 1.2 + 205374937a & 1 & 0.17 & 0.16 & 3.2 + 205374937a & 2 & 0.60 & 0.12 & 3.0 + 203927435a & 1 & 0.26 & 0.2 & 4.5 + 203927435a & 2 & 0.39 & 0.2 & 3.6 + 203927435a & 3 & 0.63 & 0.14 & 1.4 + 203927435a & 4 & 0.78 & 0.14 & 1.3 + additional shared characteristics of these stars include : * all eight of these stars have @xmath20 2 days .
five of them have @xmath21 0.65 days . in three cases ,
the period derived for the flux dips agrees with the period inferred from a spotted - star waveform , confirming that the dip period is also the stellar rotation period . for the other five , we believe it is likely that the dip period is the rotation period , but that is not proven .
* all have mid to late m dwarf spectral types ( m4 to m5.5 ) .
* the deepest flux dips for each star have depths from 2% to 7% ; their widths ( fwzi ) are generally 0.1 to 0.15 in phase , corresponding to about 1 to 5 hours duration in time .
* they have two to four discrete flux dips in their phased light curves ( at the sensitivity level of the _ k2 _ data ) . * for four of these stars ,
the dip depths appear fairly stable over the duration of the _ k2 _ campaign . in three of the other stars , one or more of
the flux dips disappears abruptly ( or becomes markedly weaker ) during the course of the campaign .
at least two of the abrupt changes in flux dip depth occurs immediately following a flare ( or flare - like event ) .
there are only four members of this class ; we discuss each of them here .
the lomb - scargle periodogram for epic 205046529 shows two strong peaks indicating it is a binary , with @xmath22 = 2.5619d and @xmath23 = 1.837d .
this star appeared as a close pair in the keck / hires guider camera , and we assume the two spatially resolved stars seen with keck correspond to the two periodogram peaks . after removing the signal from the longer period star and then phasing to 1.837 d , we noted the presence of an apparent flux dip in the light curve for the second star .
after further processing , we determined a best - fit period to highlight the flux dip of @xmath18 = 1.8358 days .
figure [ fig : epic205046529b ] shows the phased light curve ; the dominant feature is a spotted star waveform , superposed on which is a well - defined flux dip centered at phase @xmath20.6 .
the flux dip , however , appears somewhat noisy either due to jitter in the individual dip depths or their timing or both .
we believe this fuzziness is primarily due to variations in the depth of the dip both on short and long timescales .
figure [ fig : epic205046529b.4win ] shows four 10 day windows in the light curve for epic 205046529b .
close examination of those traces shows that the dip depth seems to decrease slowly with time over the _ k2 _
campaign , but also varies in depth from cycle to cycle . in order to determine better the nature of the variations in the dip times and depths , we measured the depths and first - moment times of occurrence of all the dips where we could identify a dip was present . those times and depths
are provided in table [ tab:205046529bdepths ] .
histograms of the dip depths in two time windows ( day 2060 - 2086 and day 2087 - 2106 ) are shown in figure [ fig : epic205046529b.histo ] , demonstrating the decreasing dip depth with time .
after day 2107 , the flux dip was no longer detected with certainty .
figure [ fig : epic205046529b.phasedgrps ] shows the phased light curves for those same two time windows , also illustrating the decrease in dip strength as the _ k2 _ campaign progressed . for both time windows , the dip profile appears to show a shallower ingress slope compared to the egress slope . for the stronger ,
better defined dips , we measure a mean width ( fwzi ) of 0.195 days , or as a fraction of the period , ( dip duration)/period = 0.106 .
lccc 2061.45 & 1.5 & 2087.18 & 1.2 + 2063.31 & 1.6 & 2089.05 & 0.4 + 2065.14 & 1.3 & 2090.89 & 1.0 + 2066.99 & 1.4 & 2092.69 & 0.8 + 2068.81 & 1.3 & 2094.52 & 1.0 + 2070.66 & 1.4 & 2096.34 & 0.8 + 2072.48 & 1.8 & 2098.19 & 0.9 + 2074.31 & 1.3 & 2100.04 & 0.9 + 2076.15 & 1.5 & 2101.87 & 1.0 + 2083.50 & 1.4 & 2103.70 &
1.2 + 2085.34 & 1.6 & 2105.52 & 1.1 + epic 205024957 is the one star in our sample that appears to have a significant 24 @xmath13 m ir excess ; however , as originally concluded by carpenter et al .
( 2009 ) , the lack of an excess at shorter ir wavelengths indicates that this is likely a debris disk and not a primordial disk .
its weak h@xmath3 emission ( table 1 ) also supports it being a wtts and not a ctts .
the lomb - scargle periodogram for epic 205024957 shows a single strong peak at @xmath18 = 1.6672d , largely driven by a nearly sinusoidal spotted - star waveform ; however , the phased light curve also shows a definite but fuzzy flux dip ( figure [ fig : epic205024957 ] ) .
we have analysed the light curve for this star in the same way as for epic 205046529 , and reach similar conclusions .
that is , we believe the fuzzy appearance of the flux dip is due to both short and long term variations in the dip depth , with little contribution from variations in the occurrence times for the dips .
figure [ fig : epic205024957.4win ] illustrates the short and long term variability of the flux dip depths with four 10 day portions of its light curve . based on gaussian fits to the well - detected flux dips , we derive a best period for the dips of @xmath18=1.6656d . for the stronger , better defined dips , we measure a mean dip width ( fwzi ) of 0.203 days , or as a fraction of the period , ( dip duration)/period = 0.122 . times and depths for the individual flux dips we could measure
are provided in table [ tab:205024957depths ] . as for epic 205046529 , in figure
[ fig : epic205024957.histo ] and figure [ fig : epic205024957.phasedgrps ] we also provide histograms of the dip depths and phased light curves for epic 205024957 in two time windows to illustrate the evolution of the dip character during the _ k2 _ campaign .
lccc 2072.48 & 4.3 & 2115.75 & 1.7 + 2074.13 & 2.5 & 2117.43 & 1.6 + 2080.78 & 1.0 & 2119.06 & 1.9 + 2084.11 & 0.9 & 2120.78 & 1.8 + 2090.83 & 1.6 & 2122.45 & 1.9 + 2095.9 & 1.0 & 2124.08 & 2.3 + 2099.08 & 0.8 & 2125.77 & 2.1 + 2100.75 & 0.7 & 2127.44 & 2.8 + 2102.43 & 1.0 & 2129.08 & 3.2 + 2104.1 & 0.7 & 2130.78 & 3.2 + 2105.75 & 1.6 & 2132.43 & 2.5 + 2109.1 & 1.5 & 2134.12 & 2.6 + 2110.8 & 0.9 & 2135.73 & 2.6 + 2112.5 & 1.0 & 2137.42 & 2.9 + 2114.1 & 1.3 & & + the lomb - scargle periodogram for epic 204143627 shows a single strong peak at @xmath18 = 1.125 days , largely driven by a nearly sinusoidal spotted - star waveform ; however , the phased light curve also shows a definite but weak flux dip ( figure [ fig : epic204143627 ] ) .
we have analysed the light curve for this star in the same way as for epic 205046529 , and reach similar though more limited conclusions due to the lower s / n for this star .
that is , we believe the fuzzy appearance of the flux dip is due to variations in the dip depth , with little contribution from variations in the occurrence times for the dips .
the four panels of figure [ fig : epic204143627 ] show that the flux dip was present and well - defined in the first 20 days of the _ k2 _ campaign , and slowly diminished in depth over the duration of the campaign . for the first time window in figure [ fig : epic204143627 ]
, we measure a dip width ( fwzi ) of 0.06 days , or relative to the period , ( dip duration)/period = 0.055 .
epic 205483258 is a relatively anonymous ( only two references in simbad as of this date ) upper sco member .
its literature name is rik-210 ( rizzuto et al .
it is both the earliest type star in table 1 and has the longest period .
its phased light curve is dominated by a well - defined , stable sinusoidal variation presumably due to a highly non - axisymmetric spot distribution .
superposed on that waveform is a single strong , narrow flux dip as shown in figure [ fig : epic205483258 ] .
the dip width ( fwzi ) varies from epoch to epoch , but is about 0.4 to 0.5 days , or relative to the period , ( dip duration)/period @xmath2 0.08 .
the dips are generally asymmetric , with a shallower and more variable ingress profile .
we have much more followup data for rik-210 than for any of our other stars . in david
et al . ( 2016c ) , we provide a detailed examination of the _ k2 _ light curve for rik-210 and an analysis of synoptic spectra for it which we obtained with keck hires .
radial velocites at 15 epochs exclude any eb scenario to explain the narrow flux dips .
the variability in the shape , depth and duration of its flux dip can only be explained by some type of cloud of material orbiting the star at the keplerian co - rotation radius and with physical dimension similar in size to the stellar radius .
we return to the implications of these data in 8 .
all of the stars in this class share these characteristics : * all four stars have @xmath24 1 day . in all four cases ,
there is a well defined spotted star waveform in addition to the flux dip , and both share the same period to within the uncertainties .
therefore , for these stars , we are certain that the flux dip period is the same as the rotation period . *
all have m spectral types ( m2.5 to m5 ) .
* in all cases , there is only one major flux dip per star .
the maximum depths are 1.5% , 2.5% , 3% , and 20% ( the latter for rik-210 ) .
the widths ( full width zero intensity ) are of order 0.1 in phase or slightly less , corresponding to about 10 hours for rik-210 and to a few hours for the other three stars .
* the dip depths vary significantly on both short ( cycle to cycle ) and long ( campaign duration ) timescales . from one cycle to the next , the dip can go from present and near maximum depth to absent . * in one case for sure , and probably in two of the other cases , the dip shape is asymmetric , with the slope of the ingress being steeper or shallower than the slope of the egress .
* of all 23 stars discussed in this paper , only one has a relatively secure ir excess and that is epic 205024957 .
how do the stars we have discussed in the previous three sections compare to other low mass upper sco wtts in terms of their basic physical properties ?
figure [ fig : uscocmd.pvmk ] provides three such comparisons a color - magnitude diagram , a spectral type histogram , and a period - color diagram .
the color - magnitude diagram shows that despite their unusual light curves , our stars do not appear to have unusual locations in the @xmath6 vs. @xmath8 cmd , other than their being concentrated to colors corresponding to mid to late m spectral types .
there is also no obvious displacement of one of our groups relative to the other two groups in this diagram ( as might have been true if they sampled different age or mass ranges ) .
spectral types are available for about a third of the 1100 wtts with _
k2 _ light curves , and for @xmath480% of the stars in table 1 .
the spectral type histogram reinforces the conclusion from the cmd that our stars are concentrated to mid to late m dwarfs . a very large fraction of the _ k2 _ upper sco stars are m dwarfs , so the fact that our stars are m dwarfs is not unexpected .
however , 40% of the full sample have spectral types earlier than m4 , whereas that is true for @xmath1 10% of the stars from table 1 (
i.e. , if drawn at random , we would have expected eight of our 20 stars with spectral type to have to be early m , vs. only two that we observe ) .
conversely , the period - color diagram does show striking correlations .
most importantly we believe , this diagram emphasizes that the stars with scallop - shell light curves segregate to the most rapidly rotating stars in upper sco .
several of them have periods quite close to break - up at ages appropriate for upper sco .
the persistent , short - duration flux dip class are also relatively rapidly rotating on average , but less so than for the scallop - shells .
the stars with transient short - duration flux dips have periods that are essentially typical of upper sco members for their @xmath8 color .
figure [ fig : uscoirexc.halpha ] provides two additional plots , both of which relate to whether the m dwarfs with unusual phased light curves in upper sco have significant signposts of circumstellar disks or active accretion from a disk .
the left panel of figure [ fig : uscoirexc.halpha ] compares the @xmath25-[w3]$ ] ( 3.5 @xmath13 m @xmath14 12.0 @xmath13 m ) color of our stars versus the other upper sco stars with and without primordial disks .
we label our stars as pev " in this diagram , for periodic exospheric variables ; for the moment , this is simply a conveniently short alias .
the pev stars appear to be quite normal for their @xmath8 color , and very well separated from the stars of the same mass with primordial disks .
the right panel of figure [ fig : uscoirexc.halpha ] compares the h@xmath3 equivalent widths for the pev stars to the other stars without obvious ir excesses . with one exception ,
the pev stars are again unremarkable .
the one exception is epic 203185083 ( @xmath8 = 6.6 , h@xmath3 = @xmath1427 ) .
our conclusion from this is that the types of variability associated with active accretion from a close - in disk ( accretion hot spots , flux dips due to an inner warped disk or dust entrained in an accretion stream ) are unlikely to be relevant for our stars .
it is also useful to compare the size of our stars to their keplerian co - rotation radii and to distances from the star where grains would be expected to sublimate . that comparison is provided in figure
[ fig : mdwarf_radii ] . for the calculations
, we have adopted the baraffe et al .
( 2015 ; hereafter bhac15 ) evolutionary models to provide stellar luminosities and radii at 8 myr .
the radii in those models may be underestimated because they do not include radius - inflation " due to magnetic fields or spots ( macdonald & mullan 2013 ; jeffries et al . 2016 ; somers & stassun 2016 ) , but we believe the radii are adequate for our current purposes .
we use the formula in monnier & millan - gabet ( 2002 ) to determine sublimation radii , with the two curves corresponding to minimum and maximum @xmath26 values of 1.0 and 4.0 , corresponding to a plausible range of sizes for silicate or carbon grains .
based on the spectral types in table 1 , most of our stars should have masses about 0.3 m@xmath27 .
the bhac15 radius for such a star is 0.8 r@xmath27 at 8 myr ; the corresponding keplerian co - rotation radii for 0.5 , 1.5 and 5.5 days are 1.8 , 3.7 and 8.7 r@xmath27 .
the curves in figure [ fig : mdwarf_radii ] show that for the stars with period near 0.5 days , dust is unlikely to be able to form or survive ; dust may be able to survive at the keplerian co - rotation radius for the stars with @xmath281.5 days ; and dust would certainly be able to form and survive for extended periods for rik-210 , the star with @xmath285.5 days . for the persistent and transient short - duration flux dip stars , the duration , shape and depth of the flux dip encodes information on the size and nature of the object that is transiting ( or being eclipsed by ) the star , if that is the correct physical explanation for their variability . for a point source at the keplerian co - rotation radius ( @xmath29 ) whose orbit passes through the earth - star line , the duration of a transit ( or eclipse ) is just : @xmath30 for radii and masses in solar units and rotation periods in days , the durations are just : @xmath31 for an adopted age of 8 myr , our stars have masses from the bhac15 models ranging from about 0.2 to 0.6 m@xmath27 , and corresponding radii of 0.64 to 1.02 r@xmath27 .
figure [ fig : dip_durations ] plots the dip durations we have measured in 5 and 6 for the short - duration flux dip stars as a function of their periods , and compares those data to the duration predicted for a point source for stars with mass of 0.2 and 0.6 m@xmath27 .
for the shortest periods , most of the stars have their maximum observed dip durations consistent with the prediction for a point source ; for periods longer than about 1.5 days , however , the observed durations are significantly longer than the point source prediction , indicating that the occulting ( occulted ) object has size comparable to the star .
the relationship between dip duration and period appears tight based on the limited amount of data that exists , but is clearly much steeper than our simple keplerian model . to illustrate this point in another way
, we have estimated stellar radii for each of the persistent and transient flux dip stars .
for that , we have used our spectral types and estimated @xmath8 color and table 6 of pecaut & mamajek ( 2013 ) to provide intrinsic @xmath32 , @xmath33 , and @xmath8 color . those data and the stefan - boltzmann law yield direct stellar radius estimates ; for stars identified as binaries in table [ tab : basicdata ]
, we have divided the system luminosity by two prior to calculating the radius .
we then derive estimates of the keplerian co - rotation radius , adopting masses from our estimated luminosities and the bhac15 8 myr isochrone .
the maximum duration for a point source at the keplerian co - rotation radius then follows from the formula above .
table [ tab : dip_durations ] provides our estimated stellar radii and compares the predicted maximum dip durations to the longest duration flux dips for each star .
the uncertainties in the stellar radius estimates , and hence in the durations , from propagating uncertainties in the distance and age and in the stellar properties are likely 15 - 20% ( see , for example , similar calculations for individual planet - hosting stars by mann et al .
2016 and david et al .
because radii two to three times as large as we estimate would be required for the dip durations to be consistent with observations for several of our stars ( including 3 of 4 of the transient dip class ) , we conclude that radius inflation due to magnetic effects ( e.g. , macdonald & mullan 2017 ) can not explain these discrepancies and instead that the transiting / occulting body must have size comparable to the star .
lccccccc 204364515 * & m4.0 & 1.07 & 1.46 & 4.13 & 2.90 & 4.2 & pers + 203849738 & m5.5 & 0.92 & 0.62 & 1.78 & 2.45 & 2.8 & pers + 203692610 & m4.0 & 0.62 & 1.82 & 4.79 & 1.80 & 6.1 & pers + 205374937 * & m4.0 & 1.34 & 0.63 & 2.36 & 2.72 & 2.6 & pers + 203962559 & m4.0 & 1.28 & 1.54 & 4.28 & 3.52 & 5.5 & pers + 203927435 * & m4.5 & 1.20 & 0.48 & 1.81 & 2.43 & 2.3 & pers + 205024957 & m5.0 & 0.76 & 1.67 & 3.72 & 2.61 & 8.0 & trans + 205046529 * & m4.0 & 0.88 & 1.84 & 4.82 & 2.57 & 8.6 & trans + 204143627 & m5.0 & 0.75 & 1.13 & 2.87 & 2.26 & 1.6 & trans + 205483258 & m2.5 & 1.30 & 5.67 & 11.24 & 5.01 & 12.2 & trans + the shape of the flux dips can also provide qualitative information on the nature of the occulting body .
the roughly triangular flux dips for our transient flux dip group are roughly similar in shape to the flux dip shapes found in many eclipsing binaries ( ebs ) . in most ebs ,
the flux dips are very stable in their depth and shape and the dip profile is symmetric .
however , if the primary of the eb were heavily spotted , the flux dips could be asymmetric if the transit path crosses regions of higher or lower spot coverage .
if , in addition , the eb orbit is precessing and the spots are changing on timescales comparable to the orbital period , flux dips with variable depth and shape could result .
our four transient flux dip stars are indeed heavily spotted .
the spotted - star waveforms are very stable over the entire _ k2 _ campaign .
therefore , even a precessing eb orbit would at most produce changes in the dip shape and depth that vary smoothly with time , which is not the case for our stars .
the combination of dip depths of a few percent , and the above arguments from durations and shapes therefore suggests that the occulting ( occulted ) bodies are likely to have sizes comparable to that of the star and optical depths ( or surface brightnesses or filling factors ) that are small .
one can also draw inferences as to the physical mechanism(s ) driving the variability in our group of stars from the presence or absence of similar light curve morphologies in surveys of the light curves of other samples of m dwarfs .
first , to our knowledge none of the @xmath23000 m dwarfs in the original _ kepler _ field show light curve morphologies like those in our three groups ( mcquillen , aigrain , & mazeh 2013 ) .
this likely arises both from the relative paucity of mid to late m dwarfs in the _ kepler _ sample , and to a paucity of young and/or rapidly rotating m dwarfs .
k2 _ light curves have been obtained for several hundred mid to late m dwarfs in both the pleiades ( age @xmath2 125 myr ) and praesepe ( age @xmath2 700 myr ) open clusters .
we ( rebull et al . 2016 ) found six
late m dwarfs in the pleiades with persistent , short - duration flux dips ; we find no such stars in the older praesepe cluster .
we have found no examples of scallop - shell waveforms nor transient , short - duration flux dips in either the pleiades or praesepe .
0.65 d , and about 400 pleiades stars with period @xmath1 2 d. for praesepe , in that same mass range , there are about 50 stars with period @xmath1 0.65 d and about 200 stars with period @xmath1 2 d. ] we discuss the implications of these results in the next section .
light curves from _ corot _ , _ kepler _ , and _
k2 _ have identified a small but growing population of stars with photometric variability that often seems mysterious and often takes the form of short - duration flux dips .
this small menagerie of objects includes pre - main sequence stars ( e.g. , cody et al .
2014 ) , but also includes main sequence stars ( e.g. , rappaport et al . 2012 , boyajian et al .
2016 ) and even white dwarfs ( vanderburg et al . 2015 ) .
possible physical mechanisms for producing the short - duration flux dips range from occultations by warped or structured circumstellar disks to dust streaming away from a disintegrating close - in planet ( as well as other even more speculative thoughts ) .. in many cases , a consensus has not yet been reached as to which mechanism is responsible for the observed variability .
our goal for this paper has been to add a new set of objects to this menagerie , and to provide detailed empirical data for these stars in order to begin the process of categorizing their variability and placing them in the context of the existing objects with similar properties . in this section ,
we briefly consider several broad categories of physical models as possible mechanism to explain our light curves .
we expect that this is just the beginning of a lengthy process , to which we hope many other groups will contribute .
non - axisymmetrically distributed spots on the surfaces of low mass stars typically produce relatively simple and smooth phased light curve shapes with one or two maxima and one or two minima per period .
however , it is not inconceivable that some combination of spot sizes and locations might produce light curve signatures similar to our unusual upper sco stars . in order to address this hypothesis ,
we have created a spotted star modeling routine and have produced simulated light curves for a variety of spot configurations .
we describe the spot simulation program and some of the tests we have run in the appendix .
based on those simulations , it _ is _ possible to get light curve shapes that are very similar to the scallop - shell and persistent dip classes if one puts spots very close to the upper ( or lower ) limb of the star as viewed from earth .
however , the spot distributions that can mimic the light curve shapes _ can not _ also yield light curve amplitudes that match our observations . for several of the scallop - shells , the full amplitude is of order 10% , and for a number of the persistent flux dip class , the dip depths are between 3 and 8% with fwzi @xmath10.2 in phase . in order to yield
the very structured phased light curves of the scallop - shell class , or the very narrow ( in phase ) flux dips of the persistent dip class , the spots have to be _ very _ near the edge of the visible hemisphere of the star . because of the geometric fore - shortening and limb darkening when at the edge of the visible hemisphere , such spots yield amplitudes or dip depths less than 1% , even if the spots are assumed to be totally black .
larger spots or spots further from the terminator " yield light curves that are smoother and do not match the morphologies we see .
therefore , we believe that cool spots on the stellar surface do not provide a suitable model to explain the light curve morphologies for our upper sco stars .
there is some similarity between the light curves of some of the stars in table 1 and some of the stars identified in recent years as dippers " in orion , ngc 2264 and upper sco ( morales - calderon et al .
2011 ; cody et al . 2014
; stauffer et al . 2015 ; ansdell et al . 2016 ) ; these stars are also sometimes referred to as aa tau analogs ( bouvier et al .
1999 ; alencar et al . 2010 ; mcginnis et al .
the light curves of aa tau stars show a wide diversity , with most of them showing very broad and structured flux dips which are quite dissimilar from our upper sco stars and thus do not serve as plausible comparison objects . a small subset of them , however , do have periodic , short - duration flux dips sometimes superposed on sinusoidal variations with the same period ( stauffer et al . 2015 ) , thus superficially linking their light curve morphology with that of the stars we have discussed in this paper . for these stars ,
the periodic short - duration flux dips have generally been attributed to extinction by localized structures in or above the star s inner circumstellar disk , where the inner disk rim and stellar photosphere are assumed to be locked into at least approximate co - rotation .
could some / many of the stars in table 1 simply be another subset of this class , with the same basic physical processes producing the observing photometric variability ?
we believe the answer to that question is primarily no , based on one major difference between our stars and the stars of the dipper class .
all but two of the 40 dippers in orion identified by morales - calderon et al . and
all of the dippers identified in ngc 2264 by cody et al . and by stauffer et al .
have strong ir excesses and seds appropriate for class i or ii ysos .
all of the dippers in upper sco identified by ansdell et al .
( 2016 ) have strong 24 @xmath13 m excesses , and all but one of them have strong 12 @xmath13 m excesses .
those ir excesses arise from warm dust at or near the inner edge of their primordial circumstellar disks .
as discussed in 2.2 and in the appendix , with one exception our stars do not have ir excesses and have seds consistent with bare photospheres .
also , for most of our stars ( with p @xmath1 1 day ) , the keplerian co - rotation radius lies interior to the likely dust sublimation radius , and therefore any cloud " responsible for the periodic light curve variability should be dust free .
the corot light curves for ngc 2264 are similar in quality to the _ k2 _ light curves for upper sco , and corot light curves were obtained for more than 300 wtt cluster members ( about twice as many wtts as ctts ) ; however , none of those wtts showed light curve morphologies like our upper sco stars in table 1 .
could that absence somehow affect the conclusion we have just reached ? probably not .
ngc 2264 is five times further away than upper sco , and the corot limiting magnitude is brighter than for _ k2_. therefore , the sample of stars with good light curves in ngc 2264 contains very few stars in the mass range of the stars in our table 1 ( venuti et al .
2016 include only two stars with @xmath20 1 day and mass @xmath1 0.4 m@xmath27 in their table of ngc 2264 rotation periods from corot ) . to the extent that the light curve signatures we see require young ages , rapid rotation , and low masses
, we would not expect to have found them in ngc 2264 because we were not able to sample the needed mass range . as discussed in 7 , 8.2 , and in david et al .
( 2016c ) , the transient short - duration flux dips found for four of our stars can not be produced by cool spots on the stellar photosphere or by any other plausible phenomenon associated with the surface of the star .
the equality of the period for the narrow flux dip and the sinusoidal waveform ( which we assume _ is _ due to spots ) therefore places the material responsible for the narrow flux dips at the keplerian co - rotation radius . that material must be distributed over a region with size comparable to the star ( figure [ fig : dip_durations ] and table [ tab : dip_durations ] ) , but have low average optical depth given flux dip depths generally just a few percent .
several recently discovered stars have light curve properties that seem reasonably analogous to our four transient flux dip stars .
the closest analog is ptf 08 - 8695 ( van eyken et al .
2012 ; johns - krull et al .
2016 ) , an m3 star in the @xmath23 myr old 25 ori subgroup of orion .
light curves for ptf 08 - 8695 show narrow flux dips with depths up to 3% , fwzi @xmath20.18 in phase , and a period of 0.48 days . the flux dip depths are variable and at some epochs are not present at all .
the dips are superposed on a more or less sinusoidal spotted star waveform with identical period to that for the flux dips .
these properties are remarkably similar to our four upper sco transient flux dip sources .
van eyken et al .
( 2012 ) concluded that the best interpretation of the data was to ascribe the flux dips to transits of a recently formed or forming planet undergoing roche - lobe overflow mass loss . two main sequence field stars with _ kepler _ or _
k2 _ data have light curve properties that make them somewhat analogous to our transient flux dip stars .
kic 12557548 ( rappaport et al . 2012 ) is a 16th magnitude , field k dwarf with a rotation period @xmath34 23 days and hence a likely age of a few gyr .
the _ kepler _ data show it to have asymmetric flux dips with a period of 15.7 hours ; the flux dips vary irregularly in depth , with a maximum depth of about 1.3% .
the fwzi of the dips is about 0.15 in phase .
_ k2_-22b ( sanchis - ojeda et al .
2015 ) is a field m0 dwarf with a rotation period @xmath34 15 days and hence a likely age of about a gyr .
k2 _ light curve shows asymmetric flux dips with a period of about 9.1 hours ; the flux dips are highly variable in depth , with dip depths from zero to 1.3% .
the fwzi of the dips is about 0.09 in phase .
the flux dips for both stars have been modeled by the above authors as arising from a planet that is in the process of disintegrating .
the surface of each planet would have temperatures near 2000 k. in the rappaport et al .
( 2012 ) model , a dusty wind from the surface generates a dust stream / tail that is responsible for the transits , with the curved geometry of the tail explaining the asymmetric dip profiles and their variability . if the model proposed for kic 1255b and _ k2_-22b is correct , those planets are at a kind of fiery goldilocks distance from their star
close enough for their surfaces to be molten , but not so close that dust grains could not survive in their planetary wind stream long enough to create a tail of sufficient extent to explain the transit duration and asymmetric shape . for our four transient flux dip stars , using the parameters in table [ tab : dip_durations ] , the surface temperature of a hypothetical planet would be in the range 1200 - 1500 k , and thus the specific mechanism proposed for kic 1255b and _ k2_-22b is not likely applicable .
however , we believe the similarities in light curve properties between these two stars and our four stars make a compelling case to ascribe the flux dips for our stars to dust associated with a planet . in our case , the distance of the planet from the star is not a function of the star s luminosity but is instead a function of its rotation period and mass . planets at about that location could arise from orbital migration within a circumstellar disk , and then having that migration stop when the planet reaches the inner rim of the disk which is expected to be close to the keplerian co - rotation radius for the star ( papaloizou 2007 ) . what is not obvious is the mechanism for creation of the dusty cloud that would cause the transits for our stars .
several possibilities are explored in david et al.(2016c ) .
the salient features of the persistent flux dip class are multiple short - duration flux dips with fwzi 0.1 to 0.15 in phase and depths of a few percent .
periods are always less than two days , usually less than 1 day .
long term behavior is akin to the puntuated equilibrium models in evolutionary theory long periods of stability , but often with abrupt shifts where the depth of one or more dips changes significantly . at some level ,
the scallop - shell class is just a more extreme version of the persistent flux dip class more structure in the phased light curve , shorter periods ( all @xmath1 0.65 days ) , the same or more frequent abrupt shifts in light curve morphology .
we therefore hope to find a single physical mechanism that could explain the light curve variability of both classes .
the closest analog to these stars we can find in the literature is hhj 135 , in the pleiades ( rebull et al .
hhj 135 is a 125 myr , dm4.5 star ; its _
k2 _ data yields a period of 0.61 days and a phased light curve showing three short - duration flux dips with a very similar appearance to our upper sco persistent flux dip stars epic 203692610 and epic 204364515 .
this primarily tells us that extreme youth ( low gravity ) is not required for this morpological signature .
the next best analogy in the literature we can find for these stars is , surprisingly , an 8 m@xmath27 star .
@xmath35 ori e is a rapidly rotating be star .
ground - based and most light curves for @xmath35 ori e show it to have two short - duration flux dips ( widths about 0.2 in phase ) superposed on a wave - shaped undulation with a period of 1.19 days ( townsend et al . 2013 ) .
the light curve shape is apparently stable over many years .
the most light curve closely ressembles that for our persistent flux dip stars epic 204296148 and epic 205374937 .
townsend & owocki ( 2005 ) have proposed that the @xmath35 ori e light curve is a natural outcome for a rapidly rotating star with a wind and a strong magnetic dipole field .
gas from the wind accumulates in a lumpy warped disk with inner - rim at the keplerian co - rotation radius , oriented so that the normal to the disk lies between the rotation and magnetic axes .
the short - duration flux dips in the light curve arise when the disk transits our line of sight . their model ( see also ud - doula et al .
2006 ) also predicts break - out events , resulting when too much mass accumulates onto the field lines .
these magnetic reconnections could ressemble flares ( which are otherwise not predicted for b stars ) .
while the townsend & owacki ( 2005 ) model is for a high mass star with a radiative wind , it turns out that a very similar model had been proposed to explain some properties of rapidly rotating low mass stars .
optical spectra of ab doradus itself ( @xmath18 = 0.51 days ; collier cameron & robinson 1989a , b ) and speedy mic ( hd 197890 ; @xmath18 = 0.38 days ; jeffries 1993 , dunstone et al .
2006 ) and other similar rapidly rotating k dwarfs show narrow h@xmath3 absorption dips which sweep through their broad , quiescent h@xmath3 profiles , best explained as due to orbiting warm gas clouds at or above the keplerian co - rotation radius that transit the observers line of sight . some be stars show similar absorption dip transients ( smith & balona 2006 ) . in ab dor ,
often as many as six individual features ( i.e. , individual clouds ) are seen at a given epoch .
many of the h@xmath3 transients repeat from night to night for periods up to the end of a week - long campaign ( dunstone et al .
similar features have also recently been detected in a rapidly - rotating ( p=0.44 days ) m dwarf , v374 peg ( vida et al .
jardine & van ballegooijen ( 2005 ) have proposed a model for these systems in which gas from the stellar ( parker ) wind accumulates in a lumpy torus at or above the keplerian co - rotation radius where net gravitational and magnetic forces vanish , with density maxima at the places where the radial component of the magnetic field changes polarity .
these density maxima could be the clouds " that give rise to the light curve structure we see in the persistent flux dip and scallop - shell light curves .
note that in this model , the flux dips in the _ k2 _ data arise when one of these glowing clouds is eclipsed by the star , rather than when the cloud transits the face of the star and produces the h@xmath3 absorption transients .
the scallop - shells may differ from the persistent flux dip class because they have an additional source of supra - photospheric gas
gas lost through their equatorial plane because they are rotating at breakup velocity .
the above speculations are at most proof of concept .
the next step is to determine if the gas in these structures have densities , ionization fractions and temperatures consistent with that needed to generate the light curve features we see .
we intend to address those issues in a future paper ( collier cameron et al .
2017 , in prep ) .
by targeting the upper sco association , the second campaign of the _ k2 _ mission provided the largest collection of high quality light curves for low mass pre - main - sequence stars ever obtained .
those data made it possible to search for rare light curve types which only occur to a significant degree at very young ages and very low masses .
we have , in fact , identified 23 stars with light curve morphologies which have few or no previous counterparts in the literature .
we have attempted to characterize these light curves as well as possible and to identify correlations or shared characteristics among these stars and their light curves that might point to the physical mechanism responsible for creating their unusual light curve morphologies .
the most striking shared characteristics of the 19 stars in our two main groups are : * all have mid - to - late m dwarf spectral types ; * more than half of the stars have rotation periods within a factor of two of the predicted breakup period ; * eight of these stars show rapid state - transitions " in their light curve morphology , in many cases apparently triggered by a strong flare ; * the short - duration flux dips and highly structured phased light curves seem incompatible with star - spots as the driver of the photometric variability .
warm gas clouds organized into a structured , toroidal shape located at the keplerian co - rotation radius seems to be the most plausible explanation for these characteristics .
a discussion of a detailed physical model for creating and maintaining these clouds and for their producing our observed light curve morphologies is provided in collier cameron et al .
( 2017 , in prep ) .
the four other upper sco stars we have identified have single short - duration flux dips superposed on normal spotted - star light curves , with the dip period and spot period being identical .
the flux dips have highly variable depths and shapes , with shapes that are often asymmetric .
these characteristics seem better matched to some type of dusty debris , orbiting a young planet located at or near the keplerian co - rotation radius .
a detailed examination of the best studied of this group ( rik-210 ) and a comparison of those data to a variety of physical models is provided in david et al .
( 2016c ) . while the unusual light curve morphologies we find in upper sco may only be present for a small fraction of stars during a brief time interval in their early evolution , understanding these stars may have more far reaching implications .
if an ab dor slingshot prominence type model is found to be appropriate for the scallop - shell and persistent flux dip morphologies , then analysis of those stars may provide important new clues to the magnetic field topologies , winds and mass loss rates of low mass stars .
these data could also help explain the comparatively uniform and rapid rotation found for most field brown dwarfs . if a young planet is present near the keplerian co - rotation radius for the four transient flux - dip stars , that also could have significant implications for models of planet formation , particularly around m dwarfs .
we thank leslie hebb for timely advice on debugging and testing our spot simulation routine .
we also thank kevin covey , lee hartmann , nuria calvet , steve saar and jeremy drake for helpful discussions of these objects .
some of the data presented in this paper were obtained from the mikulski archive for space telescopes ( mast ) .
support for mast for non - hst data is provided by the nasa office of space science via grant nnx09af08 g and by other grants and contracts .
this paper includes data collected by the _
kepler _ mission .
funding for the _ kepler _ mission is provided by the nasa science mission directorate .
this research has made use of the nasa / ipac infrared science archive ( irsa ) , which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration .
this research has made use of nasa s astrophysics data system ( ads ) abstract service , and of the simbad database , operated at cds , strasbourg , france .
this research has made use of data products from the two micron all - sky survey ( 2mass ) , which is a joint project of the university of massachusetts and the infrared processing and analysis center , funded by the national aeronautics and space administration and the national science foundation .
the 2mass data are served by the nasa / ipac infrared science archive , which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration .
this publication makes use of data products from the wide - field infrared survey explorer , which is a joint project of the university of california , los angeles , and the jet propulsion laboratory / california institute of technology , funded by the national aeronautics and space administration .
ahn , c. , alexandroff , r. , allende prieto , c. , et al .
2014 , , 211 , 17 aigrain , s. , llama , j. , ceillier , t. et al .
2015 , , 450 , 3211 aigrain , s. , parvianinen , h. , pope , b. , 2016 , , 459 , 2408 alencar , s. , teixeira , p. , guimaraes , m. et al .
2010 , , 519 , 88 .
alonso , r. , deeg , h. , hoyer , s. et al .
2015 , , 584 , l8 .
alves de oliveira , c. & casali , m. 2008 , , 485 , 155 .
ansdell , m. , gaidos , e. , mann , a. , et al .
2016 , , 816 , 69 .
ardila , d. , martin , e. , & basri , g. 2000 , , 120 , 479 .
bouvier , j. , chelli , a. , carrasco , l. , et al . 1999 , , 349 , 619 .
bouy , h. & martin , e. 2009 , , 504 , 981 .
bouy , h. , moraux , e. , bouvier , j. et al .
2006 , , 637 , 1056 .
boyajian , t. , lacourse , d. , rappaport , s .
et al .
2016 , , 457 , 3988 .
byrne , p. , eibe , m. , & rolleston , w. 1996 , , 311 , 651 .
carpenter , j. , bouwman , j. , mamajek , e. et al .
2009 , , 181 , 197 .
chiang , e. & laughlin , g. 2013 , , 431 , 3444 .
claret , a. , hauschildt , p. , witte , s. 2012 , , 546 , 14 .
cody , a. , stauffer , j. , baglin , a. et al .
2014 , , 147 , 82 .
cody , a. , hillenbrand , l. , david , t. , et al .
2017 , arxiv 1612.05599 .
collier cameron , a. , & robinson , r. 1989a , , 236 , 57 .
collier cameron , a. , & robinson , r. 1989b , , 238 , 657 .
collier cameron , a. , duncan , d. , ehrenfreund , p. , et al . 1990 , , 247 , 415 .
cutri , r. m. , wright , e. l. , conrow , t. , et al .
2012 , http://wise2.ipac.caltech.edu/docs/release/allsky/expsup/ dahm , s. , slesnick , c. , & white , r. 2012 , , 745 , 56 .
david , t. , hillenbrand , l. , petigura , e. , et al .
2016a , nature , 534 , 658 .
david , t. , hillenbrand , l. , cody , a. , et al .
2016b , , 816 , 21 .
david , t. , hillenbrand , l. , cody , a. , et al .
2016c , arxiv:1612.03907 .
dunstone , n. , collier cameron , a. , barnes , j. , & jardine , m. 2006 , , 373 , 1308 .
epchtein , n. , de batz , b. , capoani , l. et al .
1997 , eso messenger 87 , 27 .
erickson , k. , wilking , b. , meyer , m. , et al .
2011 , , 142 , 140 .
van eyken , j. , ciardi , d. , van braun , k. , et al .
2012 , , 755 , 42 .
flewelling , h. , magnier , e. , chambers , k. , et al .
2016 , arxiv:1612.05243 .
gaia collaboration , prusti , t. , de bruijne , j.h.j .
, et al . , 2016 , arxiv:1609.04172 henden , a. , & munari , u. , 2014 , coska , 43 , 518 howell , s. , sobeck , c. , haas , m. , et al . 2014 , , 126 , 398 huber , d. , bryson , s. , haas , m. , et al . , 2016 , , 224 , 2 irwin , p. , fletcher , l. , tide , d. et al .
2016 , , 271 , 418 jardine , m. & van ballegooijen , a. 2005 , , 361 , 1173 .
jeffries , r.d .. 1993 , , 262 , 369 .
jeffries , r. , jackson , r. , franciosini , e. , et al . , 2016 , arxiv:1609.07150 johns - krull , c. , prato , l. , mclane , j. , et al .
2016 , , 830 , 15 kraus , a. , cody , a. , covey , k. et al .
2015 , , 807 , 3 kraus , a. & hillenbrand , l. 2007 , , 662 , 413 .
kraus , a. & hillenbrand .
2009 , , 703 , 1511 .
kraus , a. & hillenbrand .
2012 , , 757 , 141 .
lodieu , n. , hambly , n. , jameson , r. , et al .
2007 , , 374 , 372 .
lodieu , n. , dobbie , p. , & hambly , n. 2011 , , 527 , 24 .
lodieu , n. , alonso , r. , gonzalez hernandez , j. et al .
2015 , , 584 , 128 luger , r. , agol , e. , kruse , e. , et al .
2016 , aj , 152 , 100 .
luhman , k. & mamajek , e. 2012 , , 758 , 31 .
macdonald , j. & mullan , d. 2013 , , 765 , 126 .
macdonald , j. & mullan , d. 2017 , , 834 , 67 .
mann , a. , newton , e. , rizzuto , a. , et al.2016 , , 152 , 61 .
mcginnis , p. , alencar , s. , guimaraes , m. 2015 , , 577 , 11 .
mcquillan , a. , aigrain , s. , & mazeh , t. 2013 , 432 , 1203 .
meyer , m. , hildebrand , l. , backman , d. , et al .
, 2006 , , 118 , 1690 morales - calderon , m . ,
stauffer , j. , hillenbrand , l. et al .
2011 , , 733 , 50 .
murakami , h. , baba , h .
, barthel , p. et al .
2007 , , 59 , 369 papaloizou , j. 2007 , , 463 , 775 .
preibisch , t. , guenther , e. & zinnecker , h. 2001 , , 121 , 1040 .
preibisch , t. , brown , a. , bridges , t. et al .
2002 , , 124 , 404 .
rappaport , s. , levine , a. , chiang , e. et al.2012 , , 752 , 1 .
rebull , l. , padgett , d. , mccabe , c. , et al.2010 , , 186 , 259 .
rebull , l. , stauffer , j. , bouvier , j. , et al.2016 , , 152 , 114 .
rigliaco , e. , wilking , b. , meyer , m. , et al .
2016 , , 588 , 123 .
rizzuto , a. , ireland , m. , & kraus , a. 2015 , , 448 , 2737 .
sanchis - ojeda , r. , rappaport , s. , palle , e. , et al . , 2015 , , 812 , 112 .
skrutskie , m. , cutri , r. m. , stiening , r. , et al . , 2006 , , 131 , 1163 smith , m. & balona , l. 2006 , , 640 , 491 .
somers , g. & stassun , k. 2016 , arxiv:1609.04841 .
stauffer , j. , spinrad , h. , & thorstensen , j. 1979 , , 91 , 59 .
stauffer , j. , cody , a. , mcginnis , p. , et al .
2015 , , 149 , 130 .
stauffer , j. , rebull , l. , bouvier , j. , et al .
2016 , , 152 , 115 .
townsend , r. & owocki , s. 2005 , , 357 , 251 .
townsend , r. , rivinius , th . , rowe , j. , et al . , 2013 , , 769 , 33 .
ud - doula , a. , townsend , r. , & owocki , s. 2006 , , 640 , l191 .
vanderburg , a. , johnson , j. , rappaport , s. et al .
2015 , nature 526 , 546 .
vanderburg , a. , & johnson , j. , 2014 , , 126 , 948 venuti , l. , bouvier , j. , cody , a. , et al . , 2016 ,
arxiv 1610.08811 .
vida , k. , kriskovics , l. , olah , k. et al .
, 2016 , , 590 , 11 .
vogt , s. , allen , s. , bigelow , b. et al .
, 1994 , spie , 2198 , 362 .
werner , m. , roellig , t. , low , f. , et al .
, 2004 , apjs , 154 , 1 wilking , b. , meyer , m. , robinson , j. , & green , t. 2005 , , 130 , 1733 .
wright , e. , eisenhardt , p. r. m. , mainzer , a. k. , et al .
, 2010 , , 140 , 1868
spectral energy distributions for all the stars from table 1 are provided here as figure [ fig : seds1 ] - figure [ fig : seds4 ] .
plots are log @xmath36 in cgs units ( ergs s@xmath37 @xmath38 ) against log @xmath39 in microns .
symbols : @xmath40 : optical literature ( sdss , apass ) ; box at short bands : denis @xmath41 ; diamond : 2mass @xmath42 ; circle : irac ; box at long bands : mips ; stars : wise ; arrows : limits ; vertical bars ( often smaller than the symbol ) denote uncertainties . a kurucz - lejeune model for the corresponding spectral type ( taken to be m3 for those without known types )
is also shown as the grey line , normalized to the observations at @xmath43 .
note that this is not a robust fit , but just to `` guide the eye . ''
some models have been lightly reddened to better fit the optical points ( @xmath44 in all cases ) .
almost all the stars have seds consistent with pure photospheres ; none have large ir excesses .
epics 204117263 and 202873945 have apparent excesses , but the uncertainties are large and we do not believe these apparent excesses are real .
epic 203534383 has a possible w3 detection but the psf in the image is not circular , so we do not believe this is a secure excess .
epic 203050730 has a w4 detection in the wise catalog , but nothing is present at the source location in the image , so we have changed it to an upper limit .
epic 204364515 and 205024957 both have possible w4 detections that are contradicted by the mips 24 @xmath13 m data ; since mips is higher spatial resolution and more sensitive , we take the mips measurement as reality , rather than the w4 .
epic 205024957 has a small apparent excess at 24 @xmath13 m , and it is the only reasonably secure ir excess for the stars in table 1 .
carpenter et al .
( 2009 ) had previously identified this star as having a 24 @xmath13 m excess , which they interpreted as due to a debris disk ( because the shorter wavelength points are consistent with photospheric emission ) .
we have obtained high resolution spectra for seven of the stars in table 1 using the keck hires spectrograph ( vogt et al .
additional hires spectra of one of these star from table 1 epic 205483258 ( rik-210 ) are reported in david et al.(2016c ) .
the spectra cover the wavelength range roughly 4800 to 9200 , at a spectral resolution of about r = 45,000 .
we measured equivalent widths for h@xmath3 and lithium using the splot routine in iraf . for two of the stars
epic 203534383 and epic 203692610 there was little spectral information in the published literature , and we have used the hires spectra to determine spectral types as well as to measure h@xmath3 and lithium equivalent widths .
figure [ fig : hires_spectra ] show the h@xmath3 profiles for these two stars , and the li 6708 feature for epic 203692610 ; the s / n at the lithium line for epic 203534383 is too poor to detect lithium . for epic 203692610
, we determine equivalent widths of @xmath143.0 for h@xmath3 and 0.55 for li 6708 . for epic 203534383 ,
our derived h@xmath3 equivalent width is @xmath149.5 .
we determined spectral types for the two stars based on the strength of the tio bandheads near 7100 (stauffer et al .
1979 ; preibisch et al . 2001 ) . for both stars
, we estimate a spectral type of m4 , though for epic 203534383 this is uncertain due to the low s / n of the spectrum . for the other four stars , epic 203849738 , 204364515 , 205024957 and 205046529 we have only used the spectra to determine equivalent widths for the lithium 6708 doublet .
those lithium equivalent widths are reported in table 1 .
phased light curves for all of the stars with scallop - shell light curves are provided in figure [ fig : scallops1_light_curves ] and figure [ fig : scallops2_light_curves ] .
for those stars whose phased light morphology underwent significant evolution during the _ k2 _ campaign , we split up the 78 day _ k2 _ campaign into several time windows and plot each window with a different color point .
note that phase = 0 is arbitrary for these light curves .
figure [ fig : scallops_medians ] shows median smoothed versions of these light curves . for the stars whose phased morphology evolved during the _ k2 _ campaign , we derived the median shape from only the first time window .
figure [ fig : batwing_light_curves ] and figure [ fig : batwing_medians ] show similar figures for all stars in the persistent flux dip class .
our standard processing ( see 3 ) produced light curves from which short - term non - astrophysical trends had been removed and identified the best lomb - scargle period ( or periods ) for each star .
plots were created showing the original light curve and the light curve phased to each period identified for the star .
those latter plots were the basis for selecting stars that might be included in this paper . in order to produce higher quality versions of the phased light curves and to validate stars for inclusion in the paper , we conducted additional processing . for stars with long - term trends in their light curve , we fit low order polynomials to the data and then removed the long - term trend ( assuming that it was likely to be non - astrophysical but that in any case the presence of the long term trend would degrade the ability to phase - fold the light curve and examine its detailed structure ) . in some cases , even after this step , the light curve retained bumps and wiggles on timescales much longer than the rotation period but shorter than the _ k2 _ campaign length . in those cases , we determined the median count - rate at each period and then normalized the data in each period to the same rate ( for stars with very short periods , we instead adopted an interval of 4 periods for the calculation ) .
this again could remove real flux variations from the light curve , but removing these trends allowed us to produce phased light curves with significantly less noise .
more than half of the scallop - shell and persistent narrow flux dip stars had two significant , narrow - peaked , independent ( non - harmonic ) lomb - scargle periods , indicating that they are binaries ( or at least that light from two stars is contributing to the _ k2 _ light curve ) . in all cases ,
the plots from our standard processing showed one of the two stars to be relevant for this paper with the other star appearing to have a normal , spotted star light curve .
for these systems , we adopted an iterative approach to the final processing .
first , we phased the initial light curve to the period of the star that appears to have the larger light curve amplitude .
we then derived a median smoothed version of that light curve , and then subtracted that median - smoothed light curve from the initial light curve resulting in a residual light curve that should largely be the signal from the second star .
we then phased that residual light curve to the second lomg - scargle period , derived a median - smoothed version of that , and then subtracted this median - smoothed light curve from the initial light curve - resulting in a new residual light curve that should largely be the signal from the 1st star ( now with the signal from the second star largely removed ) .
this process was repeated if necessary with the final product being a phased light curve for the star of interest , from which most or all of the signal from the binary companion had been removed .
cool spots generally produce phased light curve morphologies ressembling sinusoids or broad dips or humps whose fwzi is approximately 50% in phase .
that is because spots seen from most view angles will be visible for nearly half the rotation period of the star ( or more ) , and therefore the signatures they impose on the light curve will span half or more in extent of the phased light curve .
however , for a view angle which places a spot very near the top or bottom of the visible hemisphere of the star , the spot could be visible for significantly less than half the rotation period .
therefore , it is conceivable that in such geometries , single spots could produce narrow flux dips ( as observed for our persistent or transient flux dip class ) or that several spots at widely spaced longitudes could produce the highly structured , wavy phased light curves we find for the scallop - shell class . to test this hypothesis ,
we have created a program to produce simulated light curves for stars with up to four cool spots placed at arbitrary latitudes and longitudes , and viewed from an arbitary view angle .
the spots are required to be circular and to have a uniform surface brightness that is a factor of between 0.0 and 1.0 of the surface brightness of the pristine photosphere of the star .
the star is assumed to be spherical .
we adopt a linear limb darkening law , and a limb darkening coefficent ( normally designated as `` u '' ) of 0.8 , following the tables of claret et al .
( 2012 ) for a mid - m dwarf and for the _ kepler _ camera .
we then allow the star to rotate , and measure its brightness after every 1 degree in rotation , following the star for a complete rotation period .
for the present purposes , we set all the spots to be `` black '' that is , non - emitting because that will produce the largest possible amplitude for a given spot size .
we have first used the program to attempt to produce narrow flux dips that might be similar to those we observe .
the persistent flux dip stars have fwzi for their dips ranging from 0.06 to 0.2 ( in fractional phase ) and depths ranging from 1% to 7.5% , with an average of about 2.5% .
the transient flux dip stars have fwzi for their dips of about 0.06 , 0.08 , 0.20 and 0.20 and maximum depths of 1.5% , 2.5% , 3% and 20% . to determine if those properties can be matched by cool spots , we first ran models with single spots of 8 degree radius , viewed from a range of locations up to a view angle which places the spot essentially at the upper limb of the visible hemisphere when the spot is facing the observer . figure [ fig : onespot_8d ] shows the flux dips resulting from those spots for the selected viewing angles .
figure [ fig : spotwidth_depth_8d ] shows how the width and depth of these flux dips vary with view angle .
based on the two preceding figures , spots with radii of 8@xmath45 or less can not be responsible for our short - duration flux dips because they are incapable of producing dip depths of 1% or more .
we therefore decided to try a larger spot ; specifically a spot of radius 18@xmath45 .
figure [ fig : onespot_18d ] and figure [ fig : spotwidth_depth_18d ] show the flux dip shapes and their depth and width as a function of the observer s view to the star . a spot this size can produce flux dip depths of several percent
; however , the dips are relatively broad .
the plots demonstrate that no view angle will produce a dip with depth @xmath4 1% and fwzi @xmath46 0.2 .
we have tried other combinations of parameters , but the answers are always the same . for any cold spot that can produce a fwzi of 0.2 or less , the spot depth will always be much less than 1%
. therefore , cold spots can not be the explanation for the short - duration flux dips for either our persistent flux dip or transient flux dip light curves . as a test to determine if appropriately arranged spots could produce phased light curves ressembling the scallop - shell class , we placed four spots of 8 to 10@xmath45 radius at latitudes between 60 and 62@xmath45 and longitudes of 0 , 80 , 180 and 250@xmath45 .
we first put the observer at a latitude of @xmath1424@xmath45 in order for those spots to be near the upper limb of the star as seen by the observer .
as shown in the first panel of figure [ fig : model_scallop ] , this does produce a phased light curve whose shape is reasonably similar to several of the stars in our scallop shell class .
however , the light curve amplitude is tiny .
the other panels in figure [ fig : model_scallop ] show that while the amplitude does increase as the observer to spot angle decreases , the degree of structure in the light curve also decreases . by the point where the light curve amplitude is @xmath21.0% ,
the morphology of the light curve is much smoother and no longer ressembles the scallop - shell stars .
choosing larger spots does not help there is again a trade off between degree of structure in the light curve and amplitude , and in no case have we been able to produce shapes which approximate our stars and had amplitudes @xmath4 0.4% . because many of the scallop - shell stars have light curve amplitudes of 5 - 10%
, we conclude that cold spots also are unable to explain the scallop - shell light curve morphologies . | using _ k2 _ data ,
we have identified 23 very low mass members of the @xmath0 oph and upper scorpius star - forming region as having periodic photometric variability not easily explained by well - established physical mechanisms such as star spots , eclipsing binaries , or pulsation .
all of these unusual stars are mid - to - late m dwarfs without evidence of active accretion , and with photometric periods generally @xmath11 day .
often the unusual light curve signature takes the form of narrow flux dips ; when we also have rotation periods from star spots , the two periods agree , suggesting that the flux dips are due to material orbiting the star at the keplerian co - rotation radius .
we sometimes see state - changes " in the phased light curve morphologies where @xmath225% of the waveform changes shape on timescales less than a day ; often , the state - change " takes place immediately after a strong flare . for the group of stars with these sudden light curve morphology shifts ,
we attribute their flux dips as most probably arising from eclipses of warm coronal gas clouds , analagous to the sling - shot prominences postulated to explain transient h@xmath3absorption features in ab doradus itself and other rapidly rotating late type stars . for another group of stars with somewhat longer periods , we find the short duration flux dips to be highly variable on both short and long timescales , with generally asymmetric flux dip profiles .
we believe that these flux dips are due to particulate clouds possibly associated with a close - in planet or resulting from a recent collisional event .
et al.et al . |
quantum states formally represent the addressable information content about the system they describe . during their evolution quantum systems may suffer the presence of noise , for instance due to the interaction with another system , generally referred as an external environment .
this may cause a loss of information on the system , and leads to a modification from its initial to its final state . in quantum communication theory , stochastic channels , that is completely positive trace preserving ( cpt )
mappings , provide a formal description of the noise affecting the system during its evolution .
the most detrimental form of noise from the point of view of quantum information , is described by the so - called entanglement breaking ( eb ) maps @xcite .
these maps when acting on a given system destroy any entanglement that was initially present between the system itself and an arbitrary external ancilla .
accordingly they can be simulated as a two stage process where a first party makes a measurement on the input state and sends the outcome , via a classical channel , to a second party who then re - prepares the system of interest in a previously agreed state @xcite . for continuous variable quantum systems
@xcite , like optical or mechanical modes , there is a particular class of cpt maps which is extremely important : the class of gaussian channels @xcite .
almost every realistic transmission line ( _ e.g. _ optical fibers , free space communication , _ etc .
_ ) can be described as a gaussian channel . in this context
the notion of eb channels has been introduced and characterized in ref .
gaussian channels , even if they are not entanglement breaking , usually degrade quantum coherence and tend to decrease the initial entanglement of the state @xcite .
one may try to apply error correction procedures based on gaussian encoding and decoding operations acting respectively on the input and output states of the map plus possibly some ancillary systems .
this however has been shown to be useless @xcite , in the sense that gaussian procedures can not augment the entanglement transmitted through the channel ( no - go theorem for gaussian quantum error correction ) . here
we point out that such lack of effectiveness does nt apply when we allow gaussian recovering operations to act _ between _ two successive applications of the same map on the system .
specifically our approach is based on the notion of _ amendable channels _ introduced in @xcite , whose definition derives from the generalization of the class of eb maps ( gaussian and not ) to the class of eb channels of order @xmath0 .
the latter are maps @xmath1 which , even if not necessarily eb , become eb after @xmath0 iterative applications on the system ( in other words , indicating with @xmath2 " the composition of super - operator , @xmath1 is said to be eb of order @xmath0 if @xmath3 is eb while @xmath4 is not ) .
we therefore say that a map is amendable if it is eb of order 2 , and there exists a second channel ( called _ filtering _ map ) such that when interposed between the two actions of the initial map , prevents the global one to be eb . in this context
we show that there exist gaussian eb channels of order @xmath5 which are amendable through the action of a proper gaussian unitary filter ( i.e. whose detrimental action can be stopped by performing an intermediate , recovering gaussian transformation ) .
the paper is structured as follows . in section
i we focus on the formalism of gaussian channels , the characterization of eb gaussian channels and their main properties . in section
ii we explicitly define two types of channels which are amendable via a squeezing operation and a phase shifter respectively . for each channel
we also propose a simple experiment based on finite quantum resources and feasible within current technology .
let us briefly set some standard notation .
a state @xmath6 of a bosonic system with @xmath7 degrees of freedom is gaussian if its characteristic function @xmath8 $ ] has a gaussian form @xcite , @xmath9 @xmath10 is the unitary weyl operator defined on the real vector space @xmath11 , @xmath12 $ ] , where @xmath13 is the symplectic form , @xmath14 and @xmath15 , @xmath16 are the canonical observables for the bosonic system .
@xmath17 is vector of the expectation values of @xmath18 , and @xmath19 is the covariance matrix @xmath20_{ij}=\frac{\langle r_i r_j+r_j r_i\rangle_\rho}{2}-\langle r_i\rangle_\rho\langle r_j \rangle_\rho\,.\ ] ] a cpt map @xmath1 is called gaussian if it preserves the gaussian character of the states , and can be conveniently described by the triplet @xmath21 , @xmath22 and @xmath23 , @xmath24 being @xmath25 matrices , which fulfill the condition @xmath26/2,\quad\ ] ] and act on @xmath27 and @xmath28 as @xmath29}}= k^\top \mathbf{v}_{\rho } k + \beta\ , \\ { \langle\vec{r}\rangle}_{\rho } & \to & { \langle\vec{r}\rangle}_{\phi[\rho]}= k^\top { \langle\vec{r}\rangle}_\rho + l.\end{aligned}\ ] ] a special subset of gaussian channels is constituted by the unitary gaussian transformations , characterized by having @xmath30 : they include multi - mode squeezing , phase shifts , displacement transformations and products among them .
the composition of two gaussian maps , @xmath31 , described by @xmath32 and @xmath33 respectively , is still a gaussian map whose parameters are given by @xmath34 finally , a gaussian map @xmath1 is entanglement - breaking @xcite if and only if its matrix @xmath24 can be expressed as @xmath35 with @xmath36 one - mode attenuation channels @xmath37 are special examples of gaussian mappings such that : @xmath38 where @xmath39 , @xmath40 and @xmath41
. this transformation can be described in terms of a coupling between the system and a thermal bosonic bath with mean photon number @xmath42 , mediated by a beam splitter of transmissivity @xmath43 . in ref .
@xcite the eb properties of the maps @xmath37 under channel iteration were studied as a function of the parameters @xmath44 and @xmath45 . for completeness
we report these findings in fig .
[ fig : att ] . in the plot
the solid lines represent the lower boundaries between the regions which identify the set of transformations @xmath46 which are eb of order @xmath0 .
they are analytically identified by the inequalities @xmath47 or , in terms of the parameter @xmath48 which gauges the bath average photon number , by @xmath49 notice that for @xmath50 , @xmath51 for all finite @xmath0 , that is if the system is coupled with the vacuum ( zero photons ) the reiterative application of the map , represented by the action of a beam - splitter on the input signal , does not destroy the entanglement between the system and any other ancilla with which it is maximally entangled before the action of the map .
it is a well known fact that a map @xmath1 is eb if and only if when applied to one side of a maximally entangled state it produces a separable state @xcite .
this fact gives an operationally well defined experimental procedure for characterizing the eb property of a channel @xmath1 based on the ability of preparing a maximally entangled state to be used as probing state for the map .
unfortunately however , while feasible for finite dimensional systems , in a continuous variable setting this approach is clearly problematic due to the physical impossibility of preparing such an ideal probe state since it would require an infinite amount of energy . quite surprisingly , the following property will avoid this experimental issue . * property ( equivalent test states ) . * _ given @xmath52 an orthonormal set , let @xmath53 be an un - normalized maximally entangled state and @xmath54 a full - rank @xmath55 density matrix .
then the ( normalized ) state @xmath56 is a valid resource equivalent to @xmath57 in the sense that a channel @xmath1 is eb if and only if @xmath58 is separable .
_ _ proof .
_ we already know that @xmath1 is eb if and only if @xmath59 is separable @xcite . we need to show that @xmath7 is separable if and only if @xmath60 is separable
. this must be true because the two states differ only by a local cp map which can not produce entanglement namely : @xmath61 and @xmath62 .
the same property can be extended to continuous variable systems where @xmath57 is not normalizable but it can still be consistently interpreted as a distribution @xcite .
now , let us consider a two - mode squeezed vacuum ( tmsv ) state with finite squeezing parameter @xmath63 , i.e. @xmath64 where @xmath65 is now the fock basis .
it can be expressed in the form of eq .
( [ property1 ] ) by choosing @xmath66 is a valid resource for the eb test .
the previous property implies that _ it is sufficient to test the action of a channel on a two - mode squeezed state with arbitrary finite entanglement in order to verify if the channel is eb or not .
_ surprisingly , even a tiny amount of entanglement is in principle enough for the test .
however , because of experimental detection noise and imperfections , a larger value of @xmath63 may be preferable as it allows for a clean - cut discrimination .
the previous results are obviously extremely important from an experimental point of view since , for single mode gaussian channels , one can apply the following operational procedure : * prepare a realistic two - mode squeezed vacuum state @xmath67 with a finite value of @xmath68 , * apply the channel @xmath1 to one mode of the entangled state resulting in @xmath69 , * check if the state @xmath70 is entangled or not .
probably the experimentally most direct way of witnessing the entanglement of @xmath71 is to apply the so - called product criterion @xcite . in this case
, entanglement is detected whenever @xmath72with @xmath73we indicate with @xmath74 and @xmath75 , @xmath76 , the position and momentum quadratures associated to each mode of the twin beam . if inequality ( crit ) is satisfied ,
@xmath71 is entangled and so @xmath77 is not eb .
this test , is a witness but it does not provide a conclusive separability proof . for this reason it is useful to compare it with a necessary and sufficient criterion .
we will use the logarithmic negativity @xmath78 , which is an entanglement measure quantifying the violation of the @xmath79 separability criterion ppt .
let @xmath80 be the covariance matrix of @xmath81 written in the block form @xmath82 the entanglement negativity @xmath78 is a function of the four invariants under local symplectic transformations @xmath83,\det [ \mathbf{b}],\det [ \mathbf{c}],\det [ \mathbf{v}_{\rho } ] $ ] and can be analytically computed @xcite : @xmath84}}{2 } } \label{eq : nu}\end{aligned}\]]where @xmath85+\det [ \mathbf{b}]-2\det [ \mathbf{c}]$ ] .
notice that @xmath86 is the minimum symplectic eigenvalue of the partially transposed state and can be interpreted as an _ optimal product creterion _ since we have that @xmath71 is entangled if and only if @xmath87while eq .
( [ crit ] ) is only a sufficient condition .
both tests eq . ( [ crit ] ) and ( [ eq : nuquadromis ] ) will be used for assessing , in the next section , the entanglement breaking property of two possible realization of amendable gaussian channels .
we note that , direct simultaneous measurements , in a dual - homodyne set - up , on the entangled sub - systems allow a direct evaluation of the product criterion @xcite . while , the experimental evaluation of @xmath78 requires the reconstruction of the bipartite system covariance matrix that in many cases can be gained by a single homodyne @xcite .
in this section we aim to prove the existence of amendable gaussian maps constructing explicit examples and propose experimental setups that would allow one to implement and test them .
to do so we will look for gaussian single mode maps @xmath88 and @xmath1 , where @xmath88 is unitary , such that @xmath89 ( notice that the second condition requires that @xmath1 can not be eb ) . under these assumptions
, it follows that the channel @xmath90 is an eb map of order 2 which can be amended by the unitary filter @xmath91 . indeed exploiting the fact that local unitary transformation can not alter the entanglement ,
the above expressions imply : @xmath92 even though ( [ eq : phiuphieb ] ) , ( [ eq : phiuphieb1 ] ) and ( [ eq : phiuphiu ] ) , ( [ argum ] ) are formally equivalent it turns out that the former relations are easier to be implemented experimentally . for this reason in the following we will focus on such scenario . [ example1 ] here we provide our first example of a channel @xmath1 and of a unitary transformation @xmath88 fulfilling eqs . ( [ eq : phiuphieb ] ) and ( [ eq : phiuphieb1 ] ) .
we will consider two mode gaussian maps . by exploiting the property explained in sec .
[ sec : choi ] regarding the equivalence of test states , without loss of generality we will apply our channels to twin - beam states with finite squeezing parameter , that is with finite energy , rather then to maximally entangled states which would require an infinite amount of energy to be realized .
( eq : phiuphieb ) and ( [ eq : phiuphieb1 ] ) will be implemented by the two setups of fig . [
fig : simplifiedsetups ] : * the first one ( setup @xmath93 ) is used to realize the transformation @xmath94 .
it consists in an optical squeezer , implementing the unitary @xmath88 , coupled on both sides with a beam - splitter ( one for each side ) of transmissivity @xmath43 . * the second setup ( setup @xmath5 of fig .
[ fig : simplifiedsetups ] ) instead is used to realize the transformation @xmath95 : it is obtained from the first by removing the squeezer between the beam splitters .
as anticipated we will use @xmath96 states as entangled probes .
the aim of the section is to show that by properly choosing the system parameters , the squeezing and the beam - splitter transmissivities , it is possible to realize an amendable gausssian channel .
the transformation induced by the beam splitter can be described by an attenuation map with @xmath97 , @xmath98 .
on the other hand , we indicate as @xmath99 the unitary map depending on the real parameter @xmath100 , referring to the action of an optical squeezer @xmath101 we set the initial state of the two modes to be a twin - beam @xmath102 , with covariance matrix given by @xmath103 the states at the output of our two setups are described by the following 2-mode density matrices , @xmath104 $ ] and @xmath105 $ ] with @xmath106 we stress that @xmath107 and @xmath108 act only on one of the two modes of the incoming twin - beam . .
both setups are divided in three stages : a @xmath96 state is prepared , the desired sequence of channels is applied to one mode of the entangled probe , and finally the output state is measured .
the beam - splitters implement the attenuation channels @xmath109 of eqs .
( [ conc2 ] ) , ( [ conc1 ] ) which represent the transformations @xmath1 of eqs .
( [ eq : phiuphieb ] ) , ( [ eq : phiuphieb1 ] ) , while the squeezing transformation @xmath99 implements the unitary @xmath88 . ] the entanglement properties of the two setups can be established by applying the criterion ( [ eq : ebtgauss])-([eq : ebtgauss2 ] ) to @xmath110 .
as already recalled , in @xcite it was shown that @xmath111 never becomes @xmath112 for any value of the transmissivity @xmath43 .
on the contrary , it can be shown that @xmath113 , given by @xmath114 is eb if and only if @xmath115 or equivalently @xmath116 in fig .
[ fig : valtilde ] we report plots of @xmath117 vs. @xmath100 and @xmath118 vs. @xmath43 to better visualize the eb regions for the two parameters .
it follows then , that for all values of @xmath43 and @xmath100 fulfilling the condition ( [ eq : tildeeta ] ) [ or its equivalent version ( [ eq : rtilda ] ) ] the channel concatenations ( [ conc2 ] ) and ( [ conc1 ] ) provide an instance of the identities ( [ eq : phiuphieb ] ) and ( [ eq : phiuphieb1 ] ) .
consequently , following the argument ( [ argum ] ) we can conclude that the map @xmath119 is an example of gaussian channel that is eb of order 2 , and can be amended by the filtering map @xmath120 : @xmath121 for all @xmath43 s .
we conclude this section , by introducing an experimental proposal for testing the entanglement - breaking properties of the maps discussed above .
a possible procedure is to use in both setups the product criterion given in eq .
( [ crit ] ) in order to test the entanglement of the twin - beam after applying @xmath113 and @xmath122 [ i.e. the entanglement of the states @xmath123 and @xmath124 . otherwise ,
if we are able to measure the full covariance matrix of the state , we can apply the optimal criterion of eq .
( [ eq : nuquadromis ] ) .
we will take into account both criterions since the first one could be experimentally simpler while the second one provides a conclusive answer . in our case ,
the covariance matrix for @xmath123 is given by @xmath125 where @xmath126 if follows that @xmath127 and @xmath128 in ( [ eq : wgauss ] ) are given by @xmath129 and for what concerns the computation of @xmath130 we get @xmath131&=&-\frac{1}{4 } \left(2 \gamma(\eta , r , r^{\prime 2}-\alpha(\eta , r , r^{\prime } ) \cosh ( r)\right ) \nonumber \\ & & \times \left(\alpha(\eta ,- r , r^{\prime } ) \cosh ( r)-2 \gamma(\eta ,- r , r^{\prime 2}\right)\ , .
\nonumber\end{aligned}\ ] ] as already observed , the state @xmath132 which describes the system at the output of the second configuration can be obtained from @xmath123 by simply setting @xmath133 : therefore , in this same limit the above equations can also be used to determine the corresponding values for the state @xmath132 .
the results for both channels are presented in fig .
[ fig : r0andr1 ] which shows the values of @xmath134 and @xmath130 as functions of the beam splitter transmittivity @xmath43 .
the comparison with the entanglement measure @xmath130 is useful to determine the values of @xmath43 and @xmath100 for which the product criterion provides a reliable entanglement test . in the second setup [ @xmath133 ]
we expect the state of the twin - beam to be entangled , since @xmath136 for all @xmath43 s . on the one hand ,
as expected we have that @xmath130 is always lower that @xmath137 , the bound being saturated when @xmath138 or @xmath139 ( see fig .
[ subfig : r0 ] ) .
on the other hand , for @xmath140 @xmath141 we get @xmath142 , and thus we can not distinguish @xmath143 from a separable state if the product criterion is used . we conclude that the product criterion , directly accessible by a dual homodyne set - up , is reliable for @xmath144 . on the contrary
the ppt criterion , requiring the full experimental reconstruction of the state covariance matrix , can be used all the way down to @xmath139 , as shown in fig .
[ subfig : r0 ] .
if we switch on the optical squeezer [ @xmath145 for @xmath146 ( see eq .
( [ eq : rtilda ] ) ) , we will get @xmath147 and the same we expect for @xmath148 , as @xmath149 .
equivalently , for any fixed @xmath100 , from eq.([eq : tildeeta ] ) we know that @xmath150 for @xmath151 , as also proved by the behavior of @xmath130 in fig .
[ subfig : r1 ] where we have set @xmath152 . on the contrary
, @xmath148 is always greater than @xmath137 , and thus our test based on @xmath148 is not conclusive for @xmath153 .
this comes from the fact that the product criterion , while being directly accessible by measurements , gives a sufficient but not necessary condition for entanglment . summarizing
if we fix the squeezing parameter @xmath100 , in order to get a reliable test by measuring @xmath148 for both setups , the transmissivity @xmath43 of the beamsplitter should be fixed such that @xmath154 under these conditions the witness measurement we have selected allows us to verify that @xmath132 is entangled [ meaning that @xmath122 is not eb ] . at the same time
the state @xmath123 will not pass the entanglement witness criterion in agreement with the fact that @xmath113 is eb .
of course this last result can not be used as an experimental _ proof _ that @xmath155 is eb since , to do so , we should first check that no other entanglement witness bound is violated by @xmath123 .
notice that this drawback can be avoided if we are able to compute the optimal witness @xmath130 by measuring the full covariance matrix of the output state . finally , let us stress that @xmath156 in the final relation ( eq : reliability ) does not depend on the two - mode squeezing of the incoming twin - beam ( see inset of fig .
[ fig : r0andr1 ] ( a ) ) and thus we do not need to test the eb properties of our maps on states characterized by an infinite amount of energy , that is on maximally - entangled states .
this represents an important observation , especially from the point of view of the experimental implementation of our scheme .
a more detailed analysis of possible experimental losses and detection errors will be addressed in a future work @xcite . in the previous section
we have seen a class of eb gaussian channels which are amendable through a squeezing filtering transformation @xmath157 .
here we focus on channels which are amendable with a different unitary filter : a phase shift @xmath158 .
according to the previous notation , the phase shift @xmath158 can be represented with the triplet : @xmath159 where @xmath160 is a phase space rotation of an angle @xmath161 . following the analogy with the previous case we look for a channel @xmath1 , such that the concatenation @xmath162 is eb or not eb , depending on the value of @xmath161 .
it is easy to check that @xmath1 can not be an attenuation channel because in this case it would simply commute with the filtering operation @xmath158 .
a good candidate is instead the channel @xmath163 , given by @xmath164 where @xmath165 , @xmath166 and @xmath167 .
notice that this corresponds to an attenuation channel where the noise affects only the @xmath168 quadrature of the mode .
this channel does not commute with a phase shift @xmath158 and , as we are going to show , the composition @xmath169 is eb only for some values of the angle @xmath161 . from the composition law in eq .
( [ eq : composition ] ) we have that the total map @xmath170 is given by @xmath171 the entanglement breaking condition given in eq .
( [ eq : ebtgauss ] ) , is equivalent to @xmath172 as explained in sec .
[ sec : choi ] .
this implies that @xmath173 where @xmath174 and @xmath175 are solutions of the equation @xmath176 . they can be explicitly determined : @xmath177 and @xmath178 , where @xmath179 the two solutions make sense only in the cases in which @xmath180 .
we may identify this as an _
amendability condition_. otherwise , in the cases in which there are no admissible solutions , it means that the channel is constantly eb or not eb independently of the filtering operation .
if we want to experimentally test the eb property of the channel @xmath170 as a function of the filtering parameter @xmath161 , we should be able to realize the operations @xmath163 and @xmath158 .
a phase shift operation @xmath158 applied to an optical mode can be realized by changing the effective optical path .
this is a classical passive operation and it is experimentally very simple .
the main difficulty is now the realization of the channel @xmath163 . a possible way to realize @xmath163 is to combine a beam splitter with an additive phase noise channel @xmath181 .
this is defined by the triplet @xmath182 and is it essentially a random displacement @xmath183 of the @xmath168 quadrature , where the shift @xmath184 is drawn from a gaussian distribution of variance @xmath185 and mean equal to zero .
this could be realized via an electro - optical phase modulator driven with classical electronic noise or by other techniques .
it is immediate to check that @xmath186 _ i.e. _ a beam splitter followed by classical phase noise is a possible experimental realization of the channel @xmath163 .
the proposed experimental setup is sketched in fig .
[ exp2 ] . .
as in fig .
[ fig : simplifiedsetups ] the setup is divided in three stages ( preparation of the probing state @xmath96 , application of the channels , and finally measurement of the output state ) .
the global map is obtained by applying twice the gaussian channel @xmath187 with the intermediate insertion of a unitary phase shifter @xmath158 .
depending on the value of the phase shift @xmath161 the global channel is eb or not . ] a two - mode squeezed state is prepared and the desired sequence of channels is applied on one mode of the entangled pair .
the presence of entanglement after the application of all the channels is verified by measuring the variances of @xmath188 and @xmath168 defined in ( [ eq : wgauss ] ) after a unitary correction @xmath189 .
this correction does not change the entanglement of the state but it is important for optimizing the entanglement criterion ( [ crit ] ) .
a possible experiment could be to measure the witness for various choices of the filtering operation , or in other words for various values of @xmath161 .
one should check that the condition for entanglement @xmath190 is verified only for some angles @xmath161 while for @xmath191 we must have @xmath192 because the channel is eb . as a figure of merit for the quality of the experiment ,
the witness @xmath148 can be compared with the corresponding optimal witness @xmath130 . and optimal theoretical witness @xmath193 as functions of the angle @xmath194 for the setup of fig . [ exp2 ] with parameters : @xmath195 , @xmath196 and @xmath197 . in this case
we find that the global channel is entanglement breaking only in the region @xmath198 where , @xmath199 and @xmath200 . ]
the results are plotted in fig .
[ plot2 ] . for some values of @xmath161
, one can experimentally show that the channel is not eb . on the other hand , inside the entanglement breaking region ,
the witness is consistently larger than @xmath137 .
again , we underline that , if we are able to measure the covariance matrix of the output state , the product criterion can be replaced by the optimal one @xmath201 ( see eq .
( [ eq : nuquadromis ] ) ) . as a final remark
we stress that , even though it is realistic to consider @xmath202 to account for experimental losses , the same qualitative results are possible in the limit of @xmath203 , _ i.e. _ without the two beam splitters . in this case the amendability condition @xmath204 ( see eq .
( [ c - eta ] ) ) implies @xmath205 and the global map is eb for @xmath206
in this paper we proved the existence of amendable gaussian maps by constructing two explicit examples . for each of them we put forward an experimental proposal allowing the implementation of the map .
we took as benchmark model the set of entanglement breaking maps , and presented a sort of `` error correction '' technique for gaussian channels . differently from the standard encoding and decoding procedures applied before and after the action of the map @xcite , it consists in considering a composite map @xmath207 with @xmath208 and applying a unitary filter between the two actions of the channel so as to prevent the global map from being entanglement breaking .
we focused on two - mode gaussian systems .
we recall that in order to test the entanglement breaking properties of a map we have to apply it , tensored with the identity , to a maximally - entangled state , which in a continuous variable setting would require an infinite amount of energy . however in sec .
sec : choi we have proved that without loss of generality it is sufficient to consider a two - mode squeezed state with finite entanglement . this property is crucial for the experimental feasibility of our schemes .
finally , in order to verify if the entanglement of the input state survives after the action our gaussian maps , we applied the product criterion to the out coming modes @xcite , and compared it with the entanglement - negativity .
the latter analysis enabled us to properly set the intervals to which the experimental parameters have to belong in oder to consider the product criterion reliable .
this analysis paves the way to a broad range of future perspectives .
one possibility would be to extend it to the case of multimode gaussian or non gaussian maps .
another compelling isuee would be determining a complete characterization of amendable gaussian maps of second or higher order .
we recall that , according to the definition introduced in @xcite , a map @xmath1 is amendable of order @xmath209 , if @xmath208 and it is possible to delay its detrimental effect by @xmath210 steps by applying the same intermediate unitary filter after successive applications of the channel .
one possible outlook in this direction would be to allow the choice of different filters at each error correction step and determine an optimization procedure over the filtering maps .
of course this analysis would be extremely difficult to be performed for arbitrary noisy maps .
a first step would be to focus on set of the gaussian maps using the conservation of the gaussian character under combinations among them and their very simple composition rules to perform this analysis . s. l. braunstein and p. van loock , rev .
* 77 * , 513 ( 2005 ) ; c. weedbrook , s. pirandola , r. garcia - patron , n. j. cerf , t. c. ralph , j. h. shapiro and s. lloyd , rev .
phys . * 84 * , 621 ( 2012 ) ; a. s. holevo and r. f. werner , phys .
a * 63 * 032312 ( 2001 ) ; a. ferraro , s. olivares and m. g. a. paris , _ gaussian states in continuous variable quantum information _ , ( bibliopolis , napoli ) ( 2005 ) ; j. eisert and m. m. wolf , _ gaussian quantum channels in quantum information with continous variables of atoms and light _ , 23 ( imperial college press , london ) ( 2007 ) ; f. caruso , j. eisert , v. giovannetti and a. s. holevo , phys . rev .
a * 84 * , 022306 ( 2011 ) ; w. p. bowen , r. schnabel , p. k. lam , and t. c. ralph phys .
a * 69 * , 012304 ( 2004 ) ; j. laurat , g. keller , j. a. oliveira - huguenin , c. fabre , t. coudreau , a. serafini , g. adesso , and f. illuminati , j. opt .
b : quantum semiclass . opt .
* 7 * , s577 ( 2005 ) ; | we show that there exist gaussian channels which are _ amendable_. a channel is amendable if when applied twice is entanglement breaking while there exists a _ unitary filter _ such that , when interposed between the first and second action of the map , prevents the global transformation from being entanglement breaking [ phys .
rev .
a * 86 * , 052302 ( 2012 ) ] .
we find that , depending on the structure of the channel , the unitary filter can be a squeezing transformation or a phase shift operation .
we also propose two realistic quantum optics experiments where the amendability of gaussian channels can be verified by exploiting the fact that it is sufficient to test the entanglement breaking properties of two mode gaussian channels on input states with finite energy ( which are not maximally entangled ) . |
arikan s _ polar codes _
@xcite are a class of ` symmetric capacity'-achieving codes for binary - input channels .
their block error probability behaves roughly like @xmath5 @xcite , where @xmath6 is the blocklength , and they achieve this performance at an encoding / decoding complexity of order @xmath7 .
polar codes for non - binary input channels were considered in @xcite . as in the binary case ,
their construction is based on recursively creating new channels from several copies of the original : let @xmath1 be a discrete memoryless channel with input alphabet @xmath8 . throughout this note
, @xmath0 will be assumed to be a prime number .
the output alphabet @xmath9 may be arbitrary .
we will let @xmath10 $ ] denote the mutual information developed across @xmath1 with uniformly distributed inputs .
] , i.e. , @xmath11 let @xmath12 , @xmath13 be independent , uniformly distributed inputs to two independent copies of @xmath1 , and let @xmath14 , @xmath15 be the corresponding outputs . consider the one - to - one mapping @xmath16
@xmath17 where ` @xmath18 ' denotes modulo-@xmath0 addition .
observe that @xmath19 and @xmath20 are independent and uniformly distributed over @xmath21 .
define the channels @xmath22 described through the conditional output probability distributions @xmath23 it follows from the chain rule of mutual information that @xmath24 .
it is also easy to see that @xmath3 is better than @xmath1 , whereas @xmath2 is worse , in the sense that @xmath25 since @xmath2 and @xmath3 are also @xmath0-ary input channels , the above procedure can be applied to each of them , creating the channels @xmath26 , @xmath27 , @xmath28 , and @xmath29 . repeating this procedure @xmath30 times , one obtains @xmath31 channels , @xmath32 , @xmath33 , with @xmath34 .
the main observation that leads the author of @xcite to construct polar codes is that these channels are _ polarized _ in the following sense : [ thm : q - polarization ] @xmath35\big\}=i(w),\ ] ] @xmath36 for all @xmath37 .
the proofs given in @xcite and @xcite for theorem [ thm : q - polarization ] are based on the following arguments : the symmetric mutual informations of the channels @xmath38 created by the above procedure have a martingale property , from which it follows that they must converge for almost all paths in the construction .
this shows that both limits in theorem [ thm : q - polarization ] exist .
to prove the claim on these limits values , it would be sufficient to show that holds with strict inequalities for all @xmath38 , unless @xmath39 .
observe , however , that since the output alphabets of channels @xmath38 grow as the construction size increases , this approach would require the aforementioned inequality to hold uniformly for all @xmath0-ary input channels .
this difficulty is circumvented in @xcite and @xcite by appropriately defining an auxiliary channel parameter @xmath40 and proving the convergence of @xmath41 to @xmath42 by the above arguments , which then implies the convergence of @xmath43 to @xmath42 .
the purpose of this note is to provide a proof of theorem [ thm : q - polarization ] that avoids this indirect approach . in order to do so
, we will need the following theorem .
[ thm : main ] if @xmath44 for some @xmath37 , then there exists an @xmath45 such that @xmath46 the dependence of @xmath47 on the channel @xmath1 is only through @xmath48 , and not through particular channel specifications ( e.g. , output alphabet size )
. theorem [ thm : main ] will be proved as a corollary to the following lemma , which is the main result reported here .
[ lem : entropy ] let @xmath49 , @xmath50 be random variables with joint probability density @xmath51 if @xmath52 for some @xmath37 , then there exists an @xmath45 such that @xmath53 we will prove lemma [ lem : entropy ] in section [ sec : proof ] .
it suffices to show that @xmath54 , as the equality @xmath24 will then imply the second half of the claim .
let @xmath49 denote two independent and uniformly distributed inputs to two copies of @xmath1 , and let @xmath50 be the corresponding outputs . since @xmath1 is memoryless , @xmath55 are jointly distributed as in .
further , @xmath44 implies @xmath56 it then follows from lemma [ lem : entropy ] that @xmath57 completing the proof .
let @xmath58 be @xmath59-valued i.i.d .
random variables with @xmath60=\pr[b_1=+]=\tfrac12 $ ] .
let @xmath61 be random variables defined as @xmath62 note that @xmath63 takes values in @xmath64 $ ] .
further , it follows from the relation @xmath24 that @xmath65=i_n$ ] .
hence , the process @xmath61 is a bounded martingale , and therefore converges almost surely to a @xmath64$]-valued random variable @xmath66 .
note , on the other hand , that @xmath67=\frac1{2^n}\#\big\{\mathbf{s}\in\{-,+\}^n\colon i(w^\mathbf{s})\in(\delta,1-\delta)\big\}.\ ] ] to conclude the proof , it thus suffices to show that @xmath68=i(w)$ ] and @xmath69=1-i(w)$ ] . to that end , note that the almost sure convergence of @xmath63 implies @xmath70=\mathbb{e}[i(w^{b_1\dotsc b_n+})-i(w^{b_1\dotsc b_n})]\to0 $ ] .
it follows from theorem [ thm : main ] that the latter convergence implies @xmath71 with probability @xmath72 . due to the martingale property of @xmath63
we have @xmath73=\mathbb{e}[i_0]=i(w)$ ] , from which it follows that @xmath68=1-\pr[i_\infty=0]=i(w)$ ] , completing the proof .
in what follows , @xmath74 and @xmath75 will both denote the entropy of a random variable @xmath76 with probability distribution @xmath77 .
we will let @xmath78 , @xmath79 denote the probability distribution with @xmath80 the cyclic convolution of vectors @xmath77 and @xmath81 will be denoted by @xmath82 .
that is , @xmath83 we will also let @xmath84 denote the uniform distribution over @xmath21 .
we will use the following lemmas in the proof : [ lem : variation - divergence ] let @xmath77 be a distribution over @xmath21 .
then , @xmath85 .\ ] ] lemma [ lem : variation - divergence ] partially complements pinsker s inequality by providing a lower bound to the @xmath86 distance between an arbitrary probability distribution and the uniform distribution by their kullback leibler divergence .
@xmath87\\ & \le q\log e \sum_i p(i)|p(i)-1/q|\\ & \le q\log e\|p-\uni(\mathcal{x})\|_1,\end{aligned}\ ] ] where we used the relation @xmath88 in the first inequality
. lemma [ lem : variation - divergence ] holds for distributions over arbitrary finite sets . that @xmath89 is a prime number has no bearing on the above proof .
[ lem : shift - distance ] let @xmath77 be a distribution over @xmath21 . then , @xmath90 for all @xmath91 , @xmath92 . that is , unless @xmath77 is the uniform distribution , its cyclic shifts will be separated from each other in the @xmath86 distance .
let @xmath93 for some @xmath94 .
we will show that there exists a @xmath95 satisfying @xmath96 which will yield the claim since @xmath97 .
suppose that @xmath98 , as the claim is trivial otherwise .
let @xmath99 denote the @xmath100th largest element of @xmath77 , and let @xmath101 .
note that @xmath102 is a proper subset of @xmath21 .
we have @xmath103&=p^{(1)}-p^{(|s|+1)}\\ & \ge p^{(1)}-1/q\\ & \ge \frac1{2(q-1)}\|p-\uni(\cx)\|_1\\ & \ge \frac{1-h(p)}{2q(q-1)\log e}.\end{aligned}\ ] ] in the above , the second inequality is obtained by observing that @xmath104 is smallest when @xmath105 , and the third inequality follows from lemma [ lem : variation - divergence ] .
therefore , there exists at least one @xmath106 such that @xmath107 given such an @xmath100 , let @xmath108 . since @xmath0 is prime , @xmath21 can be written as @xmath109 for any @xmath95 and @xmath110 .
therefore , since @xmath111 is a proper subset of @xmath21 , there exists a @xmath112 such that @xmath113 , implying @xmath114 which yields the claim .
[ lem : h - convolution ] let @xmath77 and @xmath81 be two probability distributions over @xmath21 , with @xmath115 and @xmath116 for some @xmath117 .
then , there exists an @xmath118 such that @xmath119 let @xmath120 denote the distribution with a unit mass on @xmath79 . since @xmath121 , it follows from the continuity of entropy that @xmath122 for some @xmath123 . on the other hand , since @xmath116 , we have by lemma [ lem : shift - distance ] that @xmath124 for all pairs @xmath92 .
relations , , and the strict concavity of entropy implies the existence of @xmath118 such that @xmath125 let @xmath126 and @xmath127 be two random probability distributions on @xmath21 , with @xmath128 it is then easy to see that @xmath129,\\ h(x_2\mid y_2)&=\mathbb{e}[h(p_2)],\\ h(x_1+x_2\mid y_1,y_2)&=\mathbb{e}[h(p_1\ast p_2)].\end{aligned}\ ] ] suppose , without loss of generality , that @xmath130 .
it suffices to show that if @xmath131,\mathbb{e}[h(p_2)]\in(\delta,1-\delta)$ ] for some @xmath37 , then there exists an @xmath45 such that @xmath132\ge\mathbb{e}[h(p_1)]+\epsilon(\delta)$ ] . to that end , define the event @xmath133 observe that @xmath134\\ & \le \big(1-\pr[h(p_1)>\delta/2]\big)\cdot\delta/2+\pr[h(p_1)>\delta/2],\end{aligned}\ ] ] implying @xmath135>\tfrac{\delta}{2-\delta}$ ] .
it similarly follows that @xmath136>\tfrac{\delta}{2-\delta}$ ] .
note further that @xmath137 and @xmath138 are independent since @xmath14 and @xmath15 are .
thus , @xmath111 has probability at least @xmath139 . on the other hand , lemma [ lem : h - convolution ] implies that conditioned on @xmath111 we have @xmath140 for some @xmath141 .
thus , @xmath142\\&=\pr[a]\cdot\mathbb{e}[h(p_1\ast p_2)\mid a]+\pr[a^c]\cdot\mathbb{e}[h(p_1\ast p_2)\mid a^c]\\ & \ge\pr[a]\cdot \mathbb{e}[\big(h(p_1)+\epsilon_1(\delta/2)\big)\mid a]\\ & \hspace{10em}+\pr[a^c]\cdot\mathbb{e}[h(p_1)\mid a^c]\\ & \ge \mathbb{e}[h(p_1)]+\epsilon_1(\delta/2)\epsilon_2(\delta),\end{aligned}\ ] ] where in the first inequality we used and the relation @xmath143 . setting @xmath144 yields the result .
the proof of theorem [ thm : main ] does not extend trivially to the case of composite input alphabet sizes .
in particular , that the cyclic group @xmath145 is generated by each of its non - zero elements is crucial to the proof of lemma [ lem : shift - distance ] . on the other hand ,
a weaker statement holds when the input alphabet size is composite : consider replacing the mapping with @xmath146 where @xmath147 is a permutation over @xmath21 , and define the channels @xmath148 and @xmath149 accordingly .
then , it can be shown that there exists a permutation @xmath147 for which theorem [ thm : main ] holds , irrespective of the input alphabet size .
the proof of this statement is similar to that of theorem [ thm : main ] , and therefore is omitted .
it then follows that channels with composite input alphabet sizes can be polarized in the sense of theorem [ thm : q - polarization ] if the mapping in is chosen appropriately at each step of construction . whether such channels can be polarized by recursive application of a _ fixed _
mapping is an open question .
i would like to thank emre telatar for helpful discussions
. 1 e. arkan , `` channel polarization : a method for constructing capacity - achieving codes for symmetric binary - input memoryless channels , '' _ ieee trans .
inform . theory _
it-55 , pp .
30513073 , july 2009 . | it is shown that given two copies of a @xmath0-ary input channel @xmath1 , where @xmath0 is prime , it is possible to create two channels @xmath2 and @xmath3 whose symmetric capacities satisfy @xmath4 , where the inequalities are strict except in trivial cases .
this leads to a simple proof of channel polarization in the @xmath0-ary case .
channel polarization , polar codes , entropy inequality . |
a 47@xmath4 significant pentaquark with a mass of @xmath51540mev@xmath0 decaying to @xmath1 or @xmath6 has been reported by ten [email protected] the mass measurements of these experiments we find @xmath8mev@xmath0 .
the @xmath9 for this averaging is @xmath10 giving a confidence level of @xmath11 , a 5.2@xmath4 problem .
pentaquarks with two strange quarks@xcite and with a charm quark@xcite have also been reported .
focus ran during the 19967 fixed - target run at fermilab .
a photon beam from brehmsstrahlung of a 300 gev electron and positron beam impacts beo targets .
16 silicon strip planes provide vertexing and tracking .
charged particles are tracked and momentum analyzed as they pass through up to two dipole magnets and up to five sets of multiwire proportional chambers .
three erenkov counters , two em calorimeters , and two muon detectors identify particles . a hadronic trigger requiring @xmath525gev of energy passed 7 billion events for reconstruction .
thus , these events are well above threshold for pentaquark production .
charge conjugates are assumed for these analyses and all pentaquarks are assumed to decay strongly .
we search for @xmath13 and measure the production relative to two similar decays , @xmath14 and @xmath15 .
the data is from events with a reconstructed @xmath16 or @xmath17@xcite . selecting vee candidates within @xmath18 of the nominal mass we obtain @xmath19 @xmath20 ( @xmath21 @xmath22 ) candidates with 92% ( 96% ) purity .
the remaining good quality tracks must form a good vertex ( cl @xmath23 1%).the proton candidate must pass stringent erenkov i d cuts , reducing the misidentification rate to @xmath50 .
the @xmath24 and @xmath25 are fit with a simple breit - wigner plus background of @xmath26 where @xmath27 is the energy release .
we find @xmath28 @xmath24 , @xmath29 @xmath30 , and @xmath31 @xmath32 signal events .
no pentaquark evidence is seen in the the @xmath1 mass plot ( fig.[fig : mpks ] ) .
we obtain 95% cl limits on the yield by determing how much the fitted yield must be increased to change the log - likelihood by 1.92 ( continually minimizing the background parameters ) .
this procedure is performed with a breit - wigner width of @xmath33 and 15mev@xmath0 , both with gaussian resolution varying from 2.363.07mev@xmath0 .
the maximum limit on the yield over the mass range 1.511.56gev@xmath0 is 754 ( 2252 ) for @xmath34 of 0 ( 15 ) mev@xmath0 . to set cross section limits we generate pentaquarks the same as @xmath30 . using pythia production and focus mc simulation of @xmath24 , @xmath25 , and @xmath35 events
we convert the yield limits into cross section ratio limits . for @xmath36@xmath37@[email protected]@xmath0 ,
the 95% cl limit on the production of @xmath35 relative to combined @xmath30 and @xmath32 is 0.7% ( 2.1% ) for @xmath34 of 0 ( 15 ) mev@xmath0 .
relative to @xmath24 , the limit is 0.06% ( 0.17% ) for @xmath34 of 0 ( 15 ) mev@xmath0 .
we account for all branching ratios and assume @xmath39 . in search for @xmath40.,width=480,height=154 ]
focus reconstruction of @xmath42 is described in ref.@xcite .
we select 800,000 @xmath43 candidates of which 75% are signal .
we search for @xmath44 and the @xmath45 pentaquark candidate @xmath46 .
we require the production and @xmath47 vertices have cl @xmath48 1% and separated by less than @xmath49 . the pion candidate must have a erenkov signature consistent with a pion .
we find @xmath50 @xmath51 events and no evidence for @xmath52 as shown in fig.[fig : xiplots ] .
the yield upper limit calculated at a mass of @xmath53gev@xmath0 is 114 ( 170 ) for @xmath34 of 0 ( 15 ) mev@xmath0 with resolution @xmath54mev@xmath0 .
assuming production like @xmath51 we find @xmath55 at 95% cl for @xmath34 of 0 ( 15 ) mev@xmath0 .
( left ) for @xmath51 and @xmath56 for @xmath52.,title="fig:",width=238,height=172 ] ( left ) for @xmath51 and @xmath56 for @xmath52.,title="fig:",width=238,height=172 ]
using standard focus charm reconstruction techniques we obtain a clean sample of @xmath58 @xmath59 events and @xmath60 events ( fig .
[ fig : thetacplots ] ) . combining @xmath61 @xmath62 and @xmath63 @xmath64 candidates with a positively identified proton we find no evidence for a charm pentaquark as shown in fig .
[ fig : thetacplots ] . 0 t. nakano _ et al .
_ [ leps collaboration ] , ( 91,012002,2003 )
. v. v. barmin _ et al . _ [ diana collaboration ] , ( 66,1715,2003 ) .
s. stepanyan _ et al . _ [ clas collaboration ] , ( 91,252001,2003 ) .
j. barth _ et al .
_ [ saphir collaboration ] , ( b572,127,2003 ) .
a. e. asratyan , _ et al .
_ , ( 67,682,2004 ) . v. kubarovsky _ et al .
_ [ clas collaboration ] , ( 92,032001,2004 ) .
a. airapetian _ et al . _
[ hermes collaboration ] , ( b585,213,2004 ) .
a. aleev _ et al .
_ [ svd collaboration ] , arxiv : hep - ex/0401024 m. abdel - bary _ et al . _ [ cosy - tof collaboration ] , ( b595,127,2004 ) .
s. chekanov _ et al . _
[ zeus collaboration ] , ( b591,7,2004 ) c. alt _ et al . _
[ na49 collaboration],(92,042003,2004 ) .
a. aktas _ et al .
[ h1 collaboration ] , ( b588,17,2004 ) . j. m. link _ et al . _
[ focus collaboration ] , ( a484,174,2002 ) . | we find no evidence for high - energy photoproduction of pentaquarks at 1540mev@xmath0 , 1862mev@xmath0 , or 3099mev@xmath0 using decay modes @xmath1 , @xmath2 , and @xmath3 , respectively . |
due to their interesting mathematical structure and possible physical applications deformed algebras have attracted much attention in the last twenty years . in 1982
kulish @xcite showed that the underlying algebra of the @xmath2-heisenberg spin model was a deformation of the @xmath3 algebra , called nowadays @xmath4 . following the technique developed by schwinger to compose two heisenberg oscillators to obtain @xmath3 algebra , in 1989
macfarlane and biedenharn @xcite constructed the @xmath4 algebra using two @xmath5-oscillators .
@xmath5-oscillator algebra is a deformation of the heisenberg algebra through a parameter @xmath5 , which reproduces heisenberg algebra when the parameter @xmath6 . since heisenberg algebra has an important role in several areas of physics it was tempting to search for applications of this new structure .
thus , there was an intense study of possible physical applications of @xmath5-oscillators and since then some succesfull ones have been described in the literature @xcite .
they have also been used as a phenomenological approach to study composite particles @xcite , spectra of atoms and molecules @xcite , generalized field theory @xcite and coherent states @xcite .
recently , a new algebraic structure was proposed @xcite that generalizes the heisenberg algebra .
this structure contains also @xmath5-oscillators as a particular case and has been successfully used in some different physical problems @xcite . in this new algebra ,
called generalized heisenberg algebra ( gha ) , the commutation relations among the operators @xmath7 , @xmath8 and @xmath9 depend on a general analytic functional @xmath0 and are given by : @xmath10 @xmath11 @xmath12=f(j_0)-j_0,\ ] ] with @xmath13 and @xmath14 . identifying the operator @xmath9 with the hamiltonian of a physical system , this algebra tell us that the @xmath9 eigenvalues ( @xmath15 in fock space ) , which are the energy spectrum of the physical system , are obtained by an one - step recurrence ( @xmath16 ) , _ i.e _ , each eigenvalue depends on the previous one .
thus , the eigenvalue behavior can be studied by dynamical system techniques , enormously simplifying the task of finding possible representations of the algebra @xcite . when the functional @xmath0 is linear we re - obtain the @xmath5-oscillator algebra @xcite ; other kinds of functionals give more general algebraic structures than @xmath5-oscillators @xcite .
the functional @xmath0 can be , for example , a polynomial and therefore it depends on some parameters ( the polynomial coefficients ) .
depending on the kind of functional and their parameters , finite or infinite representations are allowed @xcite .
the @xmath9 operator eigenvalues depend on the characteristic function @xmath0 and it can present a variety of different behaviors ( monotonously increasing or decreasing , etc , upper limited or not , etc ) @xcite . for instance , for the functional @xmath17 , where @xmath18 with @xmath19 and @xmath20 being the mass and the length of the well , one obtains the square - well potential algebra @xcite .
however , there are several systems which can not be described by such a class of algebras .
quasi - periodic structures have attracted much interest lately @xcite .
in particular , fibonacci sequences were considered in several areas of mathematics and physics @xcite .
moreover , it was shown that an electron gas under sudden heating has a quasiequilibrium regime with a quasiperiodic energy distribution @xcite . since most systems having a quasiperiodic energy spectrum can not be described by the gha , in this paper we are going to present a first step in this direction .
we construct here a heisenberg - type algebra for a system having the energy spectrum described by a generalized fibonacci sequence . in order to realize this
, we propose a simple algebraic structure depending on two functionals , @xmath0 and @xmath1 .
we introduce an operator @xmath21 , being the hamiltonian of a system , as it was done in eqs .
( [ eq0a ] - [ eq0c ] ) , we will call this operator , from now on , directly as @xmath21 . ]
having eigenvalues obeying the equation @xmath22 , i.e. , the eigenvalue of the @xmath23-th state depends on the two previous eigenvalues .
we call this structure a _ two - step algebra_. in order to extend the `` one - step algebra '' described by the gha , we introduce also an additional operator @xmath24 . when both functionals @xmath0 and @xmath1 appearing in the two - step algebra are linear functions , we show that we can re - write both the @xmath21 and @xmath24 operators in fock space as functions of a single number operator , @xmath25 . the algebraic structure generated by these functionals
contains the gha , @xmath5-oscillator algebra and is related to other interesting algebras as special cases . in section
ii we introduce the two - step generalized heisenberg algebra .
in section iii we study the linear case for the algebra , that contains a fibonacci - like spectra and many other interesting quasi - periodic sequences and we also discuss the types of representations we could find . in section iv we present our conclusions and in the appendix we classify possible representations for the algebra when @xmath0 and @xmath1 , appearing in the algebra , are linear functions .
we propose an structure generated by the operators @xmath21 , @xmath24 , @xmath26 and @xmath7 , with @xmath7 and @xmath8 , @xmath27 , being the step operators , @xmath28 the hamiltonian and @xmath29 , obeying the following relations : + where we have assumed that @xmath0 and @xmath1 are analytical functions . in the fock space representation of this structure , one has the normalized vacuum state @xmath30 defined by the relations : @xmath31 where @xmath32 and @xmath33 are real numbers . a casimir operator associated with this algebraic relations
is @xmath34 satisfying the following relations : @xmath35=[c^{(1)},a ] = [ c^{(1)},h]=[c^{(1)},j_3]=0.\ ] ] in order to build a representation theory for the structure given by ( [ eq1]-[eq6 ] ) , let us start with the vacuum state defined by relations ( [ eq7a]-[eq7c ] ) .
the @xmath36 operator acting on the vacuum state produces another vector , say @xmath37 : @xmath38 .
it is trivial to see that the new vector @xmath37 is orthogonal to vector @xmath39 .
let us call @xmath40 .
the @xmath41 constant can be determined by performing the inner product of @xmath42 and its dual and by using relation ( [ eq5 ] ) : @xmath43 in order to determine the @xmath21 eigenvalue on the state @xmath44 , we have , using equation ( [ eq1 ] ) : @xmath45 just as @xmath21 , we can calculate the @xmath24 eigenvalue using equation ( [ eq3 ] ) , @xmath46 therefore , we have @xmath47 . in general , it can be proved that : @xmath48 where @xmath49 if we combine equations ( [ eq29a ] ) and ( [ eq29b ] ) , the two - step dependence of the @xmath21 eigenvalue becomes clear : @xmath50 , where @xmath51 , for all @xmath52 .
choosing different functions and different initial values @xmath32 and @xmath33 , we can construct different fock space representations for this mathematical structure . in matrix representation ,
the operators of the algebra can be written as @xmath53 in the next section , we shall study the case where @xmath0 and @xmath1 are linear functions .
if we assume @xmath54 and @xmath55 ( @xmath56 ) , the equations ( [ eq1]-[eq6 ] ) reduce to @xmath57=&(r-1)h+j_3\end{aligned}\ ] ] using eqs .
( [ eq29a ] ) and ( [ eq29b ] ) for this linear case we get : @xmath58 @xmath59 the recurrence equation ( [ eq55 ] ) is a difference equation which generates a generalized fibonacci sequence .
hence , the sequence of eigenvalues of the @xmath21 operator follows a generalized fibonacci series .
we can write equations ( [ eq54 ] ) and ( [ eq55 ] ) as a two - dimensional linear map , which is more adequate in order to analyze the representation theory of the algebra .
this map can be written as : @xmath60 in @xcite , the authors have analyzed , for one quantum number , the representation theory of the gha using dynamical systems techniques .
the representations were obtained analyzing the stability of the fixed - points and the cycles of the equation @xmath61 . in this work
we have extended the previous effort and analyzed the stability of the fixed - points of the two dimensional system given by eq .
( [ eq56 ] ) .
the fixed - points of equation ( [ eq56 ] ) can be found by solving the system @xmath62 the solutions of the previous equation are @xmath63 and @xmath64 for @xmath65 , where @xmath66 .
the stability of these fixed points are given by the eigenvalues of the @xmath67-matrix in eq .
( [ eq56 - 0 ] ) , which can be calculated by means of the characteristic equation , @xmath68 .
these eigenvalues are independent of a specific point in the @xmath69-space , as the recurrence given by eq .
( [ eq56 ] ) is linear , and can be written as : @xmath70 the main possibilities are : +
( i ) the fixed point @xmath71 is asymptotically stable if @xmath72 ; an initial point @xmath73 , iterated using eq .
( [ eq56 ] ) , approaches @xmath71 as the number of iterations increases ; + ( ii ) if at least one of the eigenvalues is , in modulus , bigger than one , the fixed point @xmath71 is called unstable and the iterations of an initial point @xmath73 by eq . ( [ eq56 ] )
move this point far away the fixed point .
+ ( iii ) if both eigenvalues are , in modulus , equal to one - edge @xmath74 in figure ( 3 ) - we call the fixed point @xmath75 marginally stable .
the iterations of an initial point in the @xmath69-space will not move it far away the fixed point , but it will not approach the fixed point too as the iterations increase . in figure ( [ f1a ] )
we show the stability regions of the fixed points @xmath76 and @xmath77 in the parameter space @xmath78 . when @xmath79 the fixed point @xmath80 is the only fixed point in the @xmath81 space .
it is asymptotically stable for values of @xmath82 and @xmath83 inside the triangle @xmath84 and unstable for values of @xmath82 and @xmath83 outside this triangle .
it is marginally stable on the edge @xmath74 and , on the edge @xmath85 , where one eigenvalue ( @xmath86 ) is always equal to @xmath87 and the other ( @xmath88 ) is equal to @xmath89 , with @xmath90 $ ] , the iteration of any initial point @xmath73 will approach a cycle two , with the line joining them crossing the fixed point @xmath91 in the @xmath81-space .
if @xmath92 , the dotted line in figure ( [ f1a ] ) , there are a line of fixed points in the @xmath81-space , @xmath93 , with @xmath94 .
all these fixed points are unstable if , in the line @xmath65 , @xmath95 or @xmath96 , i.e. , for points in this line outside the edge @xmath97 ( see figure ( 3 ) ) . for points
@xmath78 in this line belonging to the edge @xmath97 , the associated eigenvalues are @xmath98 and @xmath99 , with @xmath100 $ ] . the fixed points in the @xmath81-space , @xmath101 , are stable in one direction , crossing the line of fixed points , associated with the eigenvalue @xmath99 , @xmath102 .
the direction along the line of fixed points is marginally stable , associated with the eigenvalue @xmath98 . in the linear case ,
given the values of @xmath82 and @xmath83 ( or , equivalently , the values of @xmath88 and @xmath86 , see fig .
5 ) , the possible values of @xmath32 and @xmath33 for the representations of the linear case of the algebra we are analyzing are presented in the appendix . if the two - dimensional map , eq .
( [ eq56 ] ) , has an unstable fixed point ( i.e. , if the values of @xmath82 and @xmath83 are outside the triangle of fig .
3 ) , the possible representations are infinite - dimensional and their corresponding spectra have no upper limit ; the difference between two consecutive numbers in the corresponding sequence increases as the number of iterations increases . if this map has an asymptotically stable fixed point ( i.e. , if the values of @xmath82 and @xmath83 are inside the triangle of fig .
3 ) , the possible representations are also infinite - dimensional but they have superiorly - limited spectra ; the difference between two consecutive numbers in the corresponding sequence decreases as the number of iterations increases .
the representations associated with marginally stable cases are more complex and , besides infinite - dimensional representations , it is also possible to have finite - dimensional ones . see the appendix for a discussion of the possible values of @xmath32 and @xmath33 for these representations .
if we assume @xmath103 the relation given by equation ( [ eq55 ] ) becomes @xmath104 which yields the usual fibonacci sequence if we choose @xmath105 .
the fibonacci series have as eigenvalues @xmath106 .
the @xmath107 eigenvalue is the so - called golden - number .
the single fixed - point @xmath76 is , therefore , unstable .
the representations are infinite dimensional and the sequences ( depending on the initial values ( @xmath108 ) ) increase without an upper limit , as we know .
specific values of @xmath32 and @xmath33 describing possible representations for this case will be discussed in the next section .
the values of @xmath109 and @xmath110 for @xmath103 are shown in table [ t1 ] .
the eigenvalues @xmath109 and @xmath110 provide the number of elements of kind `` a '' and `` b '' , respectively , in the @xmath23-th inflation of the fibonacci chain ( figure [ f1 ] ) .
thus , the operator @xmath36 acts as an `` inflation operator '' of the fibonacci chain , while the @xmath21 eigenvalues provide the number of elements of the quasiperiodic chain generated at the @xmath23-th inflation .
the generalized fibonacci series ( eq . ( [ eq55 ] ) ) can be written by means of the so - called binet formula : @xmath111_{p , q}+ \beta_0[n]_{p , q } , \end{aligned}\ ] ] where @xmath112 are the roots given by eq .
( [ eq56a ] ) , @xmath113 and @xmath114_{p , q}=\frac{p^n - q^n}{p -q}$ ] is the two - parameter @xmath115-number ( gauss number ) @xcite . as @xmath21 is the hamiltonian ,
the choice of @xmath33 is not completely free inasmuch as we should guarantee that @xmath116 for any @xmath117 ( the energy of the ground state must be the lowest one ) .
this condition implies that , once the value of @xmath32 had been chosen , the allowed values of @xmath33 should satisfy the relation @xmath118_{p , q } \geq ( 1-[n+1]_{p , q } ) \ , \alpha_0 , \ ] ] for any value of @xmath23 , @xmath52 .
in particular , if @xmath119_{p , q } > 0 $ ] for any value of @xmath52 , the condition to be satisfied by @xmath33 can be written as : @xmath120_{p , q } } { [ n]_{p , q } } \ , \alpha_0 , \ ] ] for any value of @xmath23 , @xmath52 . the solution for @xmath33 for the whole parameter space
will be described in the appendix . for the specific case of fibonacci chain ( @xmath103 )
it is simple to see that the term @xmath121_{p , q } ) / [ n]_{p , q}$ ] varies from @xmath122 up to @xmath123 .
then , @xmath124 implies @xmath125 and @xmath126 implies that @xmath127 .
we can show that there exist another casimir operator @xmath128 when @xmath0 and @xmath1 are linear functions : @xmath129_{p , q}-1)- \beta_0([n+1]_{p , q}-1),\ ] ] where @xmath130_{p , q}=\frac { p^n- q^{n } } { p - q}$ ] is the @xmath131deformed number operator and @xmath25 is the usual number operator , defined in fock space as @xmath132 using eq .
( [ eq59b ] ) , it is possible to reduce the algebraic structure ( [ eq53a]-[eq53e ] ) to a heisenberg - like structure composed by the usual operators @xmath133 , @xmath134 and @xmath25 . from eq .
( [ eq59b ] ) @xmath21 can be written in fock space as : @xmath135 and , using eq .
( [ eq54 ] ) , @xmath24 can be written as @xmath136 note that the effect of the application of @xmath21 and @xmath24 on @xmath137 is exactly the same as if we use eqs .
( [ eq60 ] ) and ( [ eq61 ] ) on @xmath137 . substituting @xmath21 and @xmath24 in eq .
( [ eq53e ] ) by eqs .
( [ eq60 ] ) and ( [ eq61 ] ) and using eqs .
( [ eq28a ] ) , ( [ eq28b ] ) and ( [ n ] ) we get : @xmath138&=&a^\dagger,\\\label{eq62b } [ n , a]&=&-a,\\\label{eq62c } [ a , a^\dagger]&=&(r-1)\left ( \alpha_0[n+1]_{p , q}+ \beta_0[n]_{p , q } \right ) + \nonumber\\ & + & s \left ( \alpha_0[n]_{p , q}+ \beta_0[n-1]_{p , q } \right).\end{aligned}\ ] ] the above algebraic structure is the fock space form of a two - step heisenberg algebra .
notice that the algebraic structure described in eqs .
( [ eq53a]-[eq53e ] ) correspond to eqs .
( [ eq62a]-[eq62c ] ) plus the hamiltonian @xmath139_{p , q}+ \beta_0[n]_{p , q}$ ] , whose eigenvalues follow the generalized fibonacci sequence . remembering that @xmath140 and @xmath141 ( @xmath142 and @xmath5 are roots of @xmath143 ) , the right hand side of eq .
( [ eq62c ] ) can be written as @xmath144_{p , q } + \beta_0[n]_{p , q})-pq(\alpha_0[n]_{p , q } + \beta_0[n-1]_{p , q}),\ ] ] which , after some algebra allows us to write eq .
( [ eq62c ] ) as : @xmath145&=\alpha_0\left ( [ n+2]_{p , q}-[n+1]_{p , q}\right ) + \beta_0 \left ( [ n+1]_{p , q } - [ n]_{p , q } \right).\end{aligned}\ ] ] considering @xmath146 and @xmath147 in eq .
( [ eq63a ] ) , eqs .
( [ eq62a]-[eq62b ] ) and ( [ eq63a ] ) can be related to the @xmath148-oscillator , reference @xcite , which was used in nuclear spectroscopy @xcite .
thus , the @xmath21 eigenvalues of the @xmath148-oscillator can be written as a particular case of the generalized fibonacci numbers with initial conditions " @xmath146 and @xmath147 . in our extended algebra " , the @xmath21 eigenvalues are written as generalized fibonacci numbers where the initial values @xmath32 and @xmath33 can assume any value since that @xmath149 for any @xmath52 . thereby , each set of initial values ( @xmath108 ) define a new representation .
we have constructed a heisenberg - type algebra having as generators the hamiltonian , the step - operators and an auxiliary operator . moreover ,
this structure depends on two analytic functions @xmath0 and @xmath1 .
since the energy spectra of the systems described by this algebra depend on the two previous energy levels , this algebra was called generalized two - step heisenberg algebra .
this algebra describe quantum systems having eigenvalues of the operator @xmath21 ( hamiltonian ) analytically depending on the two previous eigenvalues , i. e. , @xmath150 , where @xmath151 is a function and @xmath152 is the eigenvalue of @xmath21 acting on the eigenstate @xmath153 , i. e. , @xmath154 .
a quantum system having the hamiltonian eigenvalues obeying the fibonacci sequence ( quasi - periodic systems ) could , for example , be described by this algebra .
possible spectra generated by this algebra can be seen in figure [ level ] , showing that the typical possible spectra found in nature can be generated by this algebraic structure .
we have discussed the representations of this algebra and we have classified in detail the possible representations for linear @xmath0 and @xmath1 . to perform this classification we have used tools from dynamical system techniques as the analysis of the stability of the fixed points of the characteristic functions of the algebra , @xmath0 and @xmath1 . in @xcite , an algebraic phenomenological approach , based on the gha , was implemented for the @xmath155 molecule .
we think that the algebra constructed here could be used to generalize the mentioned approach to more general spectra .
we also consider mathematically interesting to analyse in detail the representations of this algebra at least for simple nonlinear examples of the characteristic functions .
in this appendix we discuss the representations associated with the linear case . in order to classify the possible values of @xmath32 and @xmath33 according to the value of the eigenvalues @xmath88 ( or @xmath142 ) and @xmath86 ( or @xmath5 ) we divide the plane ( @xmath156 ) in regions labeled by the numbers @xmath157 , @xmath158 , @xmath159 , @xmath160 , @xmath161 and @xmath162 , see figure [ lambda ] . as , by definition , @xmath163
, only the semi - plane equal and above the diagonal line at @xmath164 in this plane makes sense .
the representations of the extended algebra can be obtained providing @xmath32 and @xmath33 satisfy the conditions given by eqs .
( [ betaoa ] ) and ( [ betao ] ) . analysing these equations for the regions labeled from @xmath157 to @xmath165 in figure [ lambda ] , it is possible to obtain representations for the extended algebra since @xmath32 and @xmath33 satisfy the following conditions ( the results for regions iv , vi and vii are numerical ) : region iv : @xmath181 and @xmath182 + a ) if @xmath168 @xmath169 @xmath183 + b ) if @xmath171 @xmath169 @xmath184 and @xmath185 + in this region there is no simple general analytical solution for the values of @xmath32 and @xmath33 and the possible representations should be computed case by case . -space .
the asymptotically stable region of the @xmath205 fixed - point , i.e. , @xmath72 , lies inside the triangle @xmath84 ( edges are not included ) . outside the triangle @xmath84
the @xmath91 fixed - point in the @xmath81-space is unstable ( at least one eigenvalue has modulus greater than one ) .
the hached area corresponds to the asymptotically stable region where the eigenvalues are complex .
the marginally stable region of the @xmath91 fixed - point lies on the dashed edges . on the edge
@xmath206 both eigenvalues @xmath207 are complex , with unitary modulus .
the dotted line is the region in the parameter space allowing the existence of the fixed - points @xmath208 in the dynamical space @xmath81 .
these fixed points are marginally stable for values of the parameters @xmath78 on the edge @xmath97 and unstable outside .
the eigenvalues at the points @xmath133 , @xmath209 , and @xmath210 are , respectively , @xmath211 , @xmath212 and @xmath213 . ] | we have constructed a heisenberg - type algebra generated by the hamiltonian , the step operators and an auxiliar operator .
this algebra describes quantum systems having eigenvalues of the hamiltonian depending on the eigenvalues of the two previous levels .
this happens , for example , for systems having the energy spectrum given by fibonacci sequence .
moreover , the algebraic structure depends on two functions @xmath0 and @xmath1 . when these two functions are linear we classify , analysing the stability of the fixed points of the functions , the possible representations for this algebra . xxxxxxxxxxxxxxxxxx= * keywords : * heisenberg algebra ; fibonacci series ; quasi - periodic systems ; + fock representation .
+ * pacs numbers : * 03.65.fd , 02.10.de |
low - mass protostars spend the earliest stages of their evolution embedded in large amounts of gas and dust .
these envelopes are dispersed over the course of a few times @xmath13 yr , after which the pre main - sequence object , possibly surrounded by an accretion disk or its remnant , is revealed at near - infrared and optical wavelengths . even though the general characteristics of this evolution are well understood ( see shu et al .
1993 for a recent overview ) , many details remain uncertain . during the early embedded phase ,
various physical processes occur simultaneously within a few thousand au of the forming star .
these include the formation and evolution of a circumstellar disk , accretion of matter onto the protostar , the onset of an outflow , and the dissipation of the envelope .
the evolution of these phenomena is best studied by observing a well - defined sample of low - mass young stellar objects ( ysos ) , and by probing all components of their environment on the relevant spatial scales .
several surveys of the ( sub- ) millimeter continuum emission of various classes of ysos have been undertaken in recent years using single - dish telescopes ( e.g. , beckwith et al .
1990 , barsony & kenyon 1992 , moriarty - schieven et al.1994 , andr & montmerle 1994 , osterloh & beckwith 1995 ) .
these studies have established that @xmath14 of t tauri stars ( class ii objects ; lada 1987 ) have circumstellar disks , and that the mass of these disks decreases over a period of a few times @xmath15 yr . the interpretation of such data in the earlier embedded phase ( class i ) is much more ambiguous , since , in addition to the disks , the envelopes also contribute to the observed single - dish flux . during the embedded phase , matter is transferred from the envelope to the ( growing ) disk .
since the latter is likely to be optically thick even at millimeter wavelengths the spectral slope of its emission is determined by geometry and orientation , and only lower limits of its mass can be inferred .
the high spatial resolution offered by interferometer observations is essential to separate the relative contributions of disk and envelope , as demonstrated by keene & masson ( 1990 ) , terebey , chandler & andr ( 1993 ) and butner , natta & evans ( 1994 ) .
circumstellar disks have been tentatively resolved for only a few embedded objects , with typical semi - major axes of 6080 au ( l1551 irs 5 , hl tau : lay et al .
1994 ; t tau : van langevelde , van dishoeck & blake 1995 ; l1551 irs 5 : mundy et al . 1996 ) .
caution in the interpretation of these limited interferometry data is urged , since the recent results of looney et al .
( 1996 ) show that the allegedly resolved disk around l1551 irs 5 is in fact a binary , with each component surrounded by an unresolved disk .
an alternative approach for studying circumstellar envelopes is to use spectral line observations to test models of cloud collapse such as those developed by shu ( 1977 ) , terebey , shu & cassen ( 1984 ) , galli & shu ( 1993 ) , fiedler & mouschovias ( 1992 ) , ( 1993 ) , boss ( 1993 ) , and foster & chevalier ( 1993 ) .
whereas the continuum data are only sensitive to the total mass in the beam , line observations also test the density and velocity structure .
detailed calculations of the molecular excitation and radiative line transfer in model envelopes have been performed by zhou ( 1992 , 1995 ) , zhou et al .
( 1993 ) , walker , narayanan & boss ( 1994 ) , and choi et al .
( 1995 ) , but comparison with observations has been limited to a few specific , very young class 0 objects such as b335 and iras 16293@xmath162422 . although the results seem to be consistent with the simple inside - out collapse models of shu ( 1977 ) and terebey et al.(1984 ) , the models have not yet been tested over the entire time span of the embedded phase .
ohashi et al.(1991 , 1996 ) and moriarty - schieven et al .
( 1992 , 1995 ) have surveyed a large sample of embedded objects using various transitions of the cs molecule to probe a range of excitation conditions .
this molecular approach works best if the abundance of the adopted species does not change with time through , e.g. , depletion , and if it uniquely traces the envelope but none of the other components .
as will be shown in this work , the hco@xmath0 ion may be better suited for this purpose than cs .
there are several observational developments which make the detailed study of a larger sample timely .
first , millimeter interferometers have expanded in size over the last few years , increasing both the sensitivity and mapping speed by large factors .
second , single - dish submillimeter telescopes have been equipped with low - noise sis detectors at high frequencies , which makes observations of the less massive , presumably more evolved , envelopes feasible .
third , powerful bolometer arrays to trace the continuum emission from dust are just coming on line .
stimulated by the availability of these new techniques , we have carried out a detailed study of nine embedded low - mass ysos in the nearby ( 140 pc ) taurus auriga star - forming region , using both single dish and interferometric techniques .
these objects span a large range in luminosity and outflow activity , presumably reflecting variations in age and mass . both continuum emission and a variety of molecular transitions
have been observed , tracing densities from @xmath17 to @xmath18 @xmath19 , temperatures between 10 and 150 k , and angular scales of @xmath20 to 2@xmath21 ( @xmath22 au to 15,000 au at 140 pc ) , independent of orientation .
this large range in probed physical conditions and scales , achieved through the combination of single - dish and interferometer observations , distinguishes our work from most studies mentioned above .
the aim of this paper , the first in a series , is to investigate the evolution of the mass and density structure of the envelope using both continuum and line observations .
the results will be tested against one of the simplest models of protostellar collapse ( shu 1977 ) , and closely related power - law density distributions . a second objective is to investigate whether even the most embedded ysos are already surrounded by circumstellar disks .
ohashi et al .
( 1996 ) found that only a small fraction , @xmath23 , of the embedded objects shows compact 98 ghz emission in the nobeyama interferometer .
compared to the number of t tauri stars with disks , this detection rate is low , and has been interpreted as an indication of disk growth during the embedded phase . a final objective is to study the usefulness of certain molecules , especially hco@xmath0 , as reliable tracers of the circumstellar environment .
this is important not only for constraining the physical structure , but also for establishing a baseline for future studies of the chemical evolution .
the choice of the taurus
auriga region is motivated by its close proximity , the relative isolation in which its ysos appear to form , and the extensive literature on this region including the objects of our sample ( see references throughout this paper ) . by focusing on a single star - forming region , the effects of different environment
are minimized .
systematic infrared and millimeter surveys of taurus have identified most of the embedded class i objects in this cloud down to 1 mm fluxes of 10 mjy , corresponding to envelope masses of 0.0015 m@xmath24 ( tamura et al .
1991 , ohashi et al.1996 ) . as a subset of this complete sample ,
our nine objects are well defined and representative of the embedded phase .
the outline of the paper is as follows . in 2
the nine sources studied in this paper are introduced , together with their selection from the sample defined by tamura et al .
the observations are discussed in 3 . in
4 the millimeter - continuum emission of the disks and envelopes around the sources are analyzed .
the interferometer observations and literature values of the 1.1 mm single - dish fluxes are used to obtain ` pure ' envelope fluxes .
the results of the hco@xmath0 observations are presented in 5 , and further analyzed together with the envelope fluxes within the framework of the spherically symmetric inside - out collapse model of shu ( 1977 ) and related power - law density distributions in 6 .
the implications for the full sample of tamura et al.(1991 ) are discussed in 7 , where a relative evolutionary ordering of the objects is proposed .
our main conclusions are summarized in 8 .
our sources have been chosen from the flux- and color - limited sample of 24 iras sources located in taurus auriga identified as embedded ysos by tamura et al .
( 1991 , hereafter tgww ) .
their main selection criteria are infrared color @xmath25 < -0.25 $ ] , and infrared flux @xmath26 jy at either 60 or 100 @xmath27 m . sources with a known identification as an sao star or a galaxy were rejected . in table 2
the sources of tgww are listed with some of their properties .
one source ( 04154 + 1755 ) is listed as a galaxy in the simbad database , and is dropped from the sample .
the color criterion limits the sources to objects more embedded than the majority of t tauri stars ; only eight out of 23 are optically visible t tauri - like objects .
the bolometric luminosity of the sample ranges between 0.7 and 22 @xmath28 . near - infrared imaging of the sample by tgww and kenyon et al .
( 1993b ) shows that most sources have associated reflection nebulosity with sizes between 1000 and 3000 au .
seven out of the 23 sources have a clear monopolar or bipolar morphology , suggesting that the bipolar outflow plays an important role in the appearance of ysos at these short wavelengths .
co outflow emission is detected toward 20 out of the 23 objects ( moriarty - schieven et al .
1994 ) .
one source , haro 610 ( 04263 + 2426 ; identified with gv tau = elias 37 , located in the l1524 cloud ) , was added to this sample following kenyon , calvet & hartmann ( 1993a ) and leinert & haas ( 1989 ) .
the latter authors show that this object is a t tauri star ( gv tau ) with a more embedded companion .
its 2560 @xmath27 m color index of @xmath29 is only just above the selection criterion of tgww .
fourteen of the 24 objects of the sample were found to show hco@xmath0 32 emission of @xmath30 k in the @xmath4 beam of the james clerk maxwell telescope ( see 3.2 , fig . 1 , and table 2 ) . the nine strongest of these objects were subsequently selected for our subsample , and are listed in table 1 with some of their basic properties .
seven are embedded sources ( class i according to the classification of lada 1987 ) ; t tau and haro 610 ( gv tau ) are optically visible ( class ii ) , but have more embedded companions .
all nine show evidence of co outflow emission ( terebey , vogel & myers 1989 ; moriarty - schieven et al.1994 ; hogerheijde et al .
it should be kept in mind that , given the observed binary frequency of t tauri stars of 4060% ( ghez et al .
1993 ; leinert et al .
1993 ) , four to six of our nine sources may in fact be multiple systems .
in addition to haro 610 and t tau ( dyck , simon & zuckerman 1982 ; ghez et al.1993 ) , companions have been reported for l1551 irs 5 ( looney et al .
1996 ) and l1527 irs ( fuller , ladd & hodapp 1996 ) .
the pointing centers of our observations were based on optical or near - infrared observations , which have an accuracy of 2@xmath313@xmath31 . because of its deeply embedded nature , no reliable position of l1527 irs was available , and the iras position was used . in table 1
the positions derived from the millimeter - interferometer data are listed ( see 4.1 ) , which have an accuracy of @xmath32 . only for l1535 irs and l1527 irs
do these positions differ by 4@xmath315@xmath31 from the values listed by tgww .
upper limits to the mass of the central stars can be obtained by assuming that all bolometric luminosity is stellar and that the object is located on the birthline in the hertzsprung
russell diagram ( stahler 1988 ; palla & stahler 1993 ) . the inferred maximum masses range between 0.15 @xmath33 for tmc 1 and 2.62.7 @xmath33 for l1551 irs 5 and t tau , and are listed in table 1 . for t
tau n and s masses of 2 and 1 @xmath33 , respectively , are inferred by bertout ( 1983 ) and beckwith et al.(1990 ) .
leinert & haas ( 1989 ) quote a mass of 1.01.5 @xmath33 for haro 610 .
both of these objects are located close to the birthline , and the inferred values from table 1 are near their true masses . as part of the tgww sample , our sources have been studied previously by various authors .
moriarty - schieven et al.(1992 , 1994 , 1995 ) observed the sample in single - dish continuum emission at 800 @xmath27 m and 1.1 mm , and in transitions of co , cs and h@xmath10co , deriving envelope masses and beam - averaged densities .
we will use their 1.1 mm fluxes in our analysis .
kenyon et al .
( 1993a , 1993b ) modeled the spectral energy distributions ( seds ) of an overlapping sample of ysos , and observed and modeled the near - infrared emission .
they found that the seds and the near - infrared images can be reproduced by the collapse model of terebey et al.(1984 ) , but that a cleared - out bipolar cavity is required in many cases to simultaneously fit the near - infrared and far - infrared emission .
ohashi et al .
( 1991 , 1996 ) performed interferometric observations of cs 21 of seven objects of our sample , as well as of some more evolved t tauri objects .
an overview of the data obtained for each source is given in table 3 . in the following the details of the observations
are discussed .
observations of the continuum emission at 3.4 mm and 2.7 mm were obtained with the owens valley radio observatory ( ovro ) millimeter array between 1992 and 1997 , simultaneously with the hco@xmath0 10 , and the @xmath2co and c@xmath34o 10 transitions , respectively . during the 3.4 mm observations the array consisted of five antennas ; the 2.7 mm observations were made with a six - element array .
two sources were observed per track .
data taken in the low - resolution and equatorial configurations were combined , resulting in a @xmath35 coverage with spacings between 4 and 40 k@xmath36 at 3.4 mm and between 4 and 80 k@xmath36 at 2.7 mm .
this corresponds to naturally weighted , synthesized beams of @xmath37 and @xmath20 fwhm , respectively .
the observations of t tau were made in five different array configurations ( van langevelde , van dishoeck & blake 1994a ) .
the lower and upper - sideband continuum signals were recorded separately over the full instantaneous 1 ghz if - bandwidth .
the data were calibrated using the mma package , developed specifically for ovro ( scoville et al .
the quasars pks 0333 + 321 and 0528 + 134 served as phase calibrators ( 0420@xmath16014 for the observations of t tau ) ; the amplitudes were calibrated on 3c 454.3 and 3c 273 , whose fluxes at the time were determined from observations of the planets .
the interferometer data were edited in the usual manner by flagging data points with clearly deviating amplitudes and phases .
editing was especially necessary for day - time observations at 2.7 mm , when the phase stability of the atmosphere can be low .
the final @xmath38 rms noise value of the visibilities is approximately 4 mjy when vector averaged over 10 k@xmath36 wide @xmath35 intervals ( see 4.1 ) .
the naturally weighted , cleaned images have a typical @xmath38 noise level of 2 mjybm@xmath39 . reduction and analysis of the visibility data was carried out within the miriad software package .
maps of hco@xmath0 @xmath40 10 ( 89.18852 ghz ) emission were obtained with the iram 30 m telescope , covering regions between @xmath41 and @xmath42 ( approximately 16,00024,000 au in diameter ; see fig . 4 ) .
the maps were sampled at intervals between @xmath43 and @xmath44 depending on the source , with a beam size of @xmath44 .
the data were obtained in position - switched mode , with a typical switch of 15@xmath2130@xmath21 in right ascension .
special care was taken to ensure that the off positions were free of hco@xmath0 emission . pointing was checked regularly , and the maps were obtained in such a way as to minimize systematic effects of pointing drifts .
the remaining pointing error is smaller than @xmath45 .
the spectra were obtained at a frequency resolution of 40 khz ( 0.14 kms@xmath39 ) , and were converted to the main - beam temperature scale using @xmath46 , resulting in an rms noise level of typically 0.4 k per channel . for t tau a region of @xmath47 was mapped with a grid spacing of @xmath48 , and a rms noise level of 0.16 k at a velocity resolution of 0.33 kms@xmath39 . in the hco@xmath0 @xmath40 32 ( 267.55762 ghz ) and 43 ( 356.73429 ghz ) transitions
the sources were mapped over a @xmath49 ( @xmath50 au ) region at the james clerk maxwell telescope ( jcmt ) .
the maps were sampled at 5@xmath3110@xmath31 intervals , corresponding to @xmath51@xmath52 times the beam sizes of @xmath4 ( 267 ghz ) and @xmath53 ( 356 ghz ) , respectively .
the observations were obtained in position - switched mode , with a typical switch of 15@xmath2130@xmath21 in right ascension , again to ensure that the spectra are not contaminated by hco@xmath0 emission at the off position .
pointing was checked regularly , and the remaining uncertainty is less than @xmath45 .
the spectra were recorded with the digital autocorrelation spectrometer ( das ) backend with a typical resolution of 156 khz ( 0.150.18 kms@xmath39 ) .
since the local oscillator of the 267 ghz receiver at the jcmt has no phase - lock loop , the hco@xmath0 32 spectra have a minimum effective line width of @xmath54 0.51.0 kms@xmath39 .
the spectra have been converted to the main - beam antenna temperature scale using @xmath55 ( 267 ghz ) and @xmath56 ( 356 ghz ) , obtained from measurements of jupiter and mars by the jcmt staff .
typical resulting rms noise levels are 0.20.4 k per channel on the main - beam temperature scale .
observations of hco@xmath0 32 toward all 24 sources of tgww were obtained with the jcmt in 1993 august . during these observations
the effective calibration was ill determined because of technical difficulties . from comparison with spectra obtained in 1995 , main beam efficiencies between @xmath57 and 0.31 were obtained , resulting in a calibration uncertainty of at least 30% .
observations of h@xmath2co@xmath0 32 ( 260.25548 ghz ) and 43 ( 346.99854 ghz ) were also obtained at the jcmt toward the source positions . instead of a position switch , a beam switch of @xmath58 was used in 1996 may .
typical rms noise levels are 60100 mk on the main - beam temperature scale at 0.14 kms@xmath39 resolution .
nobeyama 45 m telescope maps in h@xmath2co@xmath0 10 of five of our sources have been presented by mizuno et al.(1994 ) .
the single - dish observations were further reduced and analyzed with the class software package .
in order to study the evolution of the circumstellar disks and envelopes around ysos , it is necessary to accurately separate their respective contributions to the millimeter - continuum emission . for this purpose ,
the interferometer observations are best represented in the @xmath35 plane .
since visibilities are complex quantities , only combinations of the real and imaginary parts can be plotted .
the vector average of the amplitudes with respect to a given phase center corresponds to the fourier transform of those components of the sky - brightness distribution which are symmetric around that position .
it is important to use the correct source position as the phase center , because an offset of the order of the synthesized beam ( 3@xmath316@xmath31 ) or more creates an artificial decrease of flux with @xmath35 distance .
since amplitudes are never negative , random noise in the complex visibilities translates to a non - zero expectation value for the vector - averages even in the absence of emission . in fig . 2 the vector - averaged and @xmath35-binned visibilities of our sources are shown as functions of @xmath35 distance , with their @xmath38 error bars and zero - signal expectation values .
the positions of the unresolved point sources were adopted as phase centers ( see below and table 1 ) . as a reference , the naturally weighted , cleaned images are also presented .
all nine sources are detected in 3.4 and 2.7 mm continuum emission , with total fluxes ranging between 4 and 200 mjy .
three sources , l1489 irs , l1535 irs and tmc 1 , are detected above the zero - signal expectation value in only one or two @xmath35 bins .
however , the continuum positions listed in table 1 agree well with the 2 @xmath27 m positions , and the objects can be clearly discerned in the cleaned images .
it is therefore concluded that these are true detections .
the other six sources are confidently detected above the noise level in all @xmath35 bins .
because the vector - averaged fluxes form the fourier transform of the sky brightness , the signal of a point source is constant with @xmath35 distance in the absence of noise , while that of an extended structure is a decreasing function of @xmath35 separation .
emission on scales larger than @xmath59 ( corresponding to the shortest spacing of @xmath60 k@xmath36 ) is resolved out altogether by the array ; on intermediate scales only part of the flux is recovered , depending on the @xmath35 coverage .
such resolved emission is detected on top of unresolved emission toward five sources : t tau , haro 610 , l1551 irs 5 , tmr 1 and l1527 irs .
the uncertain @xmath35 dependence of the visibilities of the weaker sources l1489 irs , l1535 irs , and tmc 1 precludes any statement about the spatial distribution of their continuum emission .
the observed visibilities , and not just their vector averages , were fit with a source model consisting of a point source plus an extended gaussian component whenever necessary .
this method has been applied previously to the millimeter - continuum emission of ysos by , e.g. , keene & masson ( 1990 ) , terebey et al.(1993 ) , and butner et al.(1994 ) .
these authors , however , use a complete envelope model to describe the extended component .
because of the limited signal - to - noise ratio and @xmath35 coverage of our data , such level of detail is unwarranted .
the obtained fit parameters ( position , point - source flux , flux and fwhm of the extended component ) are listed in table 4 .
the unresolved point sources have sizes @xmath3 ( @xmath61 au in diameter ) .
their positions are listed in table 1 and agree within 2@xmath31 with the near - infrared positions , except for l1527 irs and l1535 irs , which are shifted by @xmath62 to the northeast and north , respectively .
ladd et al.(1991 ) and fuller et al .
( 1996 ) found a similar shift for l1527 irs from single - dish millimeter continuum maps . in the following
, only the point - source fluxes will be used quantitatively in the analysis ; the inferred values for the extended sources are not sufficiently reliable for further interpretation . of the sources in our sample
only haro 610 had not been observed previously by millimeter interferometry ; all others have been observed before , though generally at a lower signal - to - noise ratio , resulting in fewer detections .
the total 2.7 mm flux levels range between 4 mjy and 200 mjy , and agree well with those listed by keene & masson ( 1990 ) , ohashi et al .
( 1991 , 1996 ) and terebey et al .
( 1993 ) . at the position of the reported companion of l1527 irs ( fuller et al.1996 )
no continuum emission is detected at 2.7 mm at an estimated @xmath38 noise level of 4 mjybm@xmath39 , taking into account primary beam attenuation and @xmath35 coverage .
the unresolved point - source emission seen by millimeter interferometers around ysos is usually attributed to thermal emission from an optically thick circumstellar disk ( e.g. , terebey et al.1993 ) .
lay et al .
( 1994 ) and mundy et al .
( 1996 ) have resolved the circumstellar disks around hl tau and l1551 irs 5 with the jcmt
cso single - baseline and bima interferometers .
they find sizes of 6080 au for the semi - major axes of these disks , consistent with the upper limits found here .
spectral indices over @xmath63 mm to 870 @xmath27 m of @xmath64 and 2.5 , respectively , are inferred .
the recent identification of l1551 irs 5 as a binary system instead of a resolved disk ( looney et al .
1997 ) does not influence the inferred value for @xmath65 .
van langevelde et al .
( 1995 ) find @xmath66 between 2.7 mm and 840 @xmath27 m for t tau , although possible variability and decomposition into thermal and non - thermal contributions for t tau n and s complicate the interpretation ( see also hogerheijde et al .
1998b ) .
non - thermal contribution to the continuum flux of ysos at millimeter wavelengths could come from free free emission of an ionized wind .
fluxes at 2 , 3.6 and 6 cm are given for some of our sources by rodrguez et al .
( 1989 ) , keene & masson ( 1990 ) and skinner & brown ( 1994 ) . from these values and the expected flat free free spectrum
, it is estimated that this mechanism does not provide more than 10% of the 2.7 mm flux of the ysos observed here ( cf . keene & masson 1990 ; lay et al.1994 ) . only for t tau
may up to 30% of the 2.7 mm flux be attributed to free free emission ( skinner & brown 1994 ; van langevelde et al .
1995 ; hogerheijde et al .
1998b ) .
the millimeter - continuum visibilities show that at least two - thirds of the embedded sources selected here are surrounded by compact disks .
this detection rate is significantly larger than that of 15% reported by ohashi et al .
( 1996 ) for a largely identical sample , owing entirely to the differences in sensitivity ( 2 mjybm@xmath39 vs. 37 mjybm@xmath39 ) .
the inferred 2.7 mm disk fluxes range between 7 and 100 mjy , which may correspond to ranges in disk size , density , temperature , orientation , or mass .
it is therefore not straightforward on the basis of these fluxes alone to investigate evolutionary effects , such as disk growth .
comparison with point source fluxes of more evolved class ii objects is further complicated by the fact that this phase lasts ten times longer than the embedded phase . during this evolutionary period significant dispersal of the disk
is expected , and it is not clear which objects represent the youngest class ii sources ( cf .
ohashi et al.1996 ) .
dutrey et al .
( 1996 ) find an average flux at 2.7 mm of @xmath67 mjy for a sample of 12 t tauri stars with the plateau de bure interferometer , comparable to the flux levels found toward our embedded sources .
moriarty - schieven et al .
( 1994 ) obtained the continuum flux at 1.1 mm of the tgww sample in a @xmath4 beam with the jcmt .
the flux of haro 610 is given by kenyon et al .
( 1993a ) .
these single - dish measurements contain contributions from the extended envelope as well as any compact disk .
the point - source fluxes derived in the previous section can be used to separate these components if a spectral index for the point - source emission between 3.4/2.7 mm and 1.1 mm is known or adopted .
since all our sources have good signal - to - noise single - dish fluxes , reliable envelope fluxes can be obtained even for those sources which have only marginal detections in the interferometer beam . for typical densities and values of the dust emissivity ,
the envelope is optically thin at 1.1 mm , and its flux traces all the mass within the beam .
the unresolved circumstellar disks , on the contrary , are likely to be optically thick . for a sharp - edged , isothermal , unresolved disk and optically thick emission , a spectral slope of @xmath69
is expected .
more realistic disks , with radial surface - density and temperature gradients , may have @xmath70 , because the optically thick area increases for shorter wavelengths .
a maximum value for @xmath65 is found from the ratio of the 3.4/2.7 mm ovro and 1.1 mm jcmt fluxes , and ranges between 2.4 and 4.0 .
these are strict upper limits since extended emission is actually observed toward half of the sources in the interferometer beam . in the following , @xmath71 will be assumed as a best estimate for the spectral slope of the disk based on the measured values for hl tau , l1551 irs 5 , and t tau . the effect of over- or underestimating @xmath65 is to attribute too little , or too much , respectively , of the 1.1 mm flux to the envelope .
sources with relatively weak fluxes at 1.1 mm are worst affected by the uncertainty in @xmath65 . however , none of the results from the subsequent analysis depend critically on the assumed value . in table 5 the 1.1 mm single dish fluxes from moriarty - schieven et al .
( 1994 ) and the estimated envelope contributions are listed .
the envelopes and disks are found to contribute equally to the single - dish fluxes : approximately 3075% of the single - dish flux at 1 mm originates in the envelope ; for @xmath69 this would be 5085% . toward haro 610
all single - dish flux can be attributed to the point source detected at 2.7 mm , although an extended component appears also present in the visibilities ( see fig . 2 ) .
no constraints on the spatial distribution of the 3.4 and 2.7 mm emission were obtained in 4.1 for l1489 irs , l1535 irs and tmc 1 , resulting in a range of possible envelope fluxes .
these caveats are included in the estimated error bars on the fluxes in the analysis below .
the envelope mass traced in the @xmath4 jcmt beam can be obtained , assuming optically thin emission from the relation @xmath72 where @xmath73 is the mass , @xmath74 is the 1.1 mm ( envelope ) flux , @xmath75 is the distance to taurus ( 140 pc ) , @xmath76 is the dust emissivity at 1.1 mm per unit total mass , and @xmath77 is the planck function at the dust temperature @xmath78 . for @xmath76 a value of 0.01 @xmath79g@xmath39 is adopted ( agladze et al.1994 ; pollack et al .
1994 ; ossenkopf & henning 1994 ) .
a dust temperature @xmath80 k is assumed , consistent with the average temperature inferred from sed fitting by moriarty - schieven et al.(1994 ) .
the obtained envelope masses range from @xmath81 @xmath33 for haro 610 to 0.26 @xmath33 for l1551 irs 5 , and are listed in table 5 . using the same expression , a lower limit to the mass of the disks
can be obtained , again assuming @xmath82 .
since the disks are likely to be optically thick at 3.4/2.7 mm , only lower limits are found .
using the same dust emissivity of 0.01 @xmath79g@xmath39 scaled by @xmath83 and a dust temperature of 30 k , disk masses between @xmath84 @xmath33 for l1489 irs and @xmath85 @xmath33 for l1551 irs 5 are inferred ( see table 4 ) , i.e. , a factor of @xmath86 less than the envelope masses . for a higher dust temperature the inferred disk masses decrease as @xmath87 , while for significant optical depth they increase as @xmath88 for @xmath89 .
given the fact that the disks likely have @xmath90 , it is concluded that the disks and envelopes typically are equally massive , with each carrying @xmath91 of the mass of the central object ( cf .
table 1 ) .
in fig . 3 the hco@xmath0 10 , 32 and 43 , and h@xmath2co@xmath0 32 and 43 spectra obtained toward the source positions are presented . in fig .
4 contour maps of the emission integrated over the full width of the hco@xmath0 lines are shown .
the 50% intensity contours are indicated by the thick solid lines in all maps .
the observed velocity - integrated intensities toward the source centers and the fwhm of the maps are listed in table 6 . toward the majority of sources
the hco@xmath0 spectra are dominated by a relatively narrow line ( @xmath92 23 kms@xmath39 ) .
double peaked line profiles are present toward most sources .
since care was taken to obtain emission - free reference positions ( 3.2 ) , these can be attributed to self - absorption .
red - shifted self - absorption features in a high - excitation line like hco@xmath0 43 are sometimes invoked as a tracer of protostellar infall ( walker et al . 1986 ; zhou et al.1993 ; zhou 1995 ; ward - thompson et al .
it should be noted that such features are only prominently present toward two of our sources , l1489 irs and l1527 irs .
the occurrence of blue - shifted absorption , or the absence of any features altogether , indicates that other factors like orientation are equally important in determining the shape of the line profiles .
for example , toward t tau the absorption seen in interferometer data is filled in by emission from the surrounding cloud in the single - dish observations ( van langevelde et al.1994a ) . in many spectra , emission from the bipolar outflow
is visible as line wings , but only toward t tau do the wings dominate the total intensity .
the integrated intensities listed in table 6 and the contour maps shown in fig . 4 therefore predominantly reflect the quiescent envelope material . for t tau , estimates of the relative contributions of the envelope and the outflow are given in the table .
the h@xmath2co@xmath0 lines are narrow , with @xmath933.0 kms@xmath39 , peaking at the same velocity as the hco@xmath0 absorption features , if present .
the average optical depth @xmath94 over the hco@xmath0 line profiles can be estimated from the ratio of the hco@xmath0 and h@xmath2co@xmath0 integrated intensities , assuming an abundance ratio of 65:1 for [ hco@xmath0]:[h@xmath2co@xmath0 ] .
typical opacities of 410 or less are found ( see table 6 ) .
larger optical depths are inferred for the 10 lines of l1527 irs and tmc 1a , with @xmath96 1821 .
the h@xmath2co@xmath0 lines are optically thin in all observed transitions and toward all sources
. a beam - averaged hco@xmath0 abundance can be obtained from the h@xmath2co@xmath0 10 intensities toward l1551 irs 5 , l1535 irs , tmr 1 , and l1527 irs ( mizuno et al .
1994 ) , and the c@xmath34o 10 data of hayashi et al .
( 1994 ) , in respective beams of @xmath4 and @xmath48 .
assuming lte at an excitation temperature @xmath97 , the beam - averaged column density in @xmath98 is given in cgs units by @xmath99 ( scoville et al .
1986 ) , where @xmath100 is the integrated intensity in kkms@xmath39 of the @xmath101@xmath102 transition with frequency @xmath103 and opacity @xmath104 .
the permanent dipole @xmath27 is 0.112 debye and 3.91 debye for c@xmath34o and h@xmath2co@xmath0 , respectively ( millar et al .
1991 ) . with critical densities of @xmath105 @xmath19 and @xmath106 @xmath19 , respectively ,
the excitation of h@xmath2co@xmath0 10 and c@xmath34o 10 is likely to be thermalized , especially in the inner dense regions of the envelope . assuming the same abundance ratio for [ hco@xmath0]:[h@xmath2co@xmath0 ] of 65:1 , a ratio of [ h@xmath10]:[c@xmath34o ] of @xmath108 , and an excitation temperature of 30 k , an abundance of hco@xmath0 of @xmath109 is found .
the uncertainty in this value is dominated by the spread in the data points of the five sources .
this value is comparable to that inferred for dark clouds of @xmath110 ( irvine et al .
1987 ) , and it is assumed that it can be applied to all sources in our sample . in the maps of integrated hco@xmath0 10 intensity ,
each source shows up as a distinct core of roughly 50@xmath31150@xmath31 in diameter ( @xmath44 beam size ) , superposed on extended emission from the surrounding cloud .
the enhancement in integrated intensity around the sources arises partially from the increased line widths . at positions @xmath111 ( 8400 au ) away from the center , the average hco@xmath0 10 line width has decreased from 2.03.0 kms@xmath39 to 0.51.0 kms@xmath39 .
the emission in hco@xmath0 32 and 43 is much more concentrated around the sources , with typical sizes of 20@xmath3130@xmath31 in diameter compared with the respective beam sizes of @xmath4 and @xmath53 .
the 50% intensity contours are nearly circular and all cores appear marginally resolved with fwhm @xmath112 times the beam size .
this is typical for density distributions following a radial power law , i.e. , distributions without an intrinsic size scale within the sampled range .
this point is illustrated by ladd et al.(1991 ) and terebey et al .
( 1993 ) for continuum emission ; line emission does possess an intrinsic density scale .
the hco@xmath0 32 and 43 emission around t tau shows a prominent extension toward the southwest , which is also seen in several transitions of co ( edwards & snell 1982 ; schuster et al .
1993 ) and is probably associated with the reflection nebula ngc 1555 . for the scope of this paper ,
i.e. , the comparison of the continuum flux with the hco@xmath0 line emission as tracers of the circumstellar envelopes , the spherically averaged quantities listed in table 6 suffice . in a future paper ( hogerheijde et al .
1998c ) , the line profiles will be studied in greater detail .
the observed integrated hco@xmath0 32 and 43 intensities are compared with the 1.1 mm envelope fluxes in fig . 5 , and are seen to correlate well . since the hco@xmath0 line intensities depend primarily on density and mass ,
whereas the continuum fluxes scale with dust temperature and mass , the apparent correlation suggests that both quantities primarily trace envelope mass .
the inferred envelope continuum fluxes at 1.1 mm and the hco@xmath0 observations presented above allow further investigation into the mass and density structure of the circumstellar material around the objects of our sample . from the beam - convolved hco@xmath0
source sizes it was concluded that the envelopes do not posses any intrinsic scale within the sampled range , i.e. , their density follows a radial power law .
many models of protostellar collapse predict such a density distribution .
the precise value of the power - law index and its evolution with time depend on the formulation of the problem and distinguish the various proposed models ( e.g. , shu 1977 , terebey , shu & cassen 1984 , galli & shu 1993 , fiedler & mouschovias 1992 , 1993 , boss 1993 , and foster & chevalier 1993 ) . in this section the envelope fluxes and integrated hco@xmath0 intensities
will be compared with one such model , namely the self - similar , inside - out collapse model of shu ( 1977 ) .
although this model is obviously oversimplified , it has the advantage that it depends on only two parameters : the sound speed @xmath113 , and the initial mass of the cloud core , or , equivalently , its initial outer radius @xmath114 . since the objects in our sample are all located within the similar environment of a single star - forming region , the dependence on these parameters is minimized .
because of the simple formulation of this model , it has been used widely for comparison to observations ( zhou 1992 ; zhou et al.1993 ; choi et al .
1995 ; ceccarelli et al .
formally , the assumption of self - similarity breaks down when the so - called collapse expansion wave reaches the outer radius .
however , we will continue to use this formulation even after this time as a qualitative description .
the inside - out collapse model has only two main parameters : the sound speed @xmath113 and the initial core radius @xmath114 .
in addition , the dust emissivity @xmath76 at 1.1 mm ( 4.3 ) , the hco@xmath0 abundance ( 5 ) , and the temperature of the gas and the dust are required to calculate the emergent continuum flux and hco@xmath0 line strengths .
the density configuration of the initial state is given by @xmath115 , and its mass by @xmath116 .
estimates for the sound speed in the taurus cloud cores vary between @xmath117 and 0.43 kms@xmath39 ( terebey et al . 1984 ) .
a value of @xmath118 kms@xmath39 is used here , corresponding to the isothermal sound speed for the typical temperatures of @xmath54 3040 k derived by moriarty - schieven et al .
in addition , it is the smallest value which can explain the observed 1.1 mm continuum flux of l1551 irs 5 ( @xmath54 3 jy ) .
it should be noted that the density , flux and mass are strongly dependent on @xmath113 , with @xmath119 and @xmath120 .
estimates for the core radius can be obtained from the hco@xmath0 10 and 32 maps , and from the c@xmath34o 21 emission observed at positions more than 30@xmath3160@xmath31 away from the source centers ( hogerheijde et al .
the density at these positions has dropped to ( 0.33.0)@xmath121 @xmath19 , representative of the surrounding cloud . with @xmath118 kms@xmath39 , this corresponds to a radius of @xmath12228,000 au where the core merges into the surrounding cloud . since most of the continuum and line emission originates from the central regions of the core , especially for the hco@xmath0 32 and 43 lines which possess critical densities @xmath123 @xmath19 , @xmath122 au ( @xmath124 ) will be used .
another estimate of the core size can be obtained from the observed line widths in the surrounding cloud , typically @xmath125
a turnover radius can be defined where the kinetic energy contained in these motions , @xmath126 , balances the gravitational binding to a @xmath127 @xmath7 object , @xmath128 .
from such an analysis , a value of @xmath129 au is found , in good agreement with the adopted @xmath122 au .
the total envelope mass depends strongly on @xmath114 , with @xmath130 , and is therefore not well constrained by the observations . for @xmath122
au , @xmath131 @xmath7 is obtained , a factor of 3 higher than the average stellar mass in taurus .
however , the bipolar outflow may disperse part of the envelope , or even reverse infall before all envelope mass has accreted onto the protostar .
the mass - accretion rate corresponding to @xmath118 kms@xmath39 is @xmath132 @xmath7yr@xmath39 .
if the accretion rate onto the star is the same as this large scale envelope accretion rate , a luminosity of @xmath133 @xmath28 is generated , assuming a stellar mass of 0.5 @xmath7 and radius of 3 @xmath134 .
this accretion luminosity is much larger than the typical observed values for our sources , but a non - constant accretion rate onto the star can result in a significantly reduced luminosity during large periods of time separated by short outbursts of high luminosity ( cf .
fu orionis events ; see hartmann , kenyon & hartigan 1993 , and 6 ) . the inside - out collapse model is based on isothermal initial conditions .
after the onset of collapse and the formation of a central object , this assumption no longer holds .
a simple power - law behavior for the dust temperature distribution is assumed , @xmath135 , as is valid for any centrally heated envelope which is optically thin to the photons carrying the bulk of the heating ( e.g. , rowan - robinson 1980 ) . at a characteristic radius of 1000 au , @xmath136 k is assumed , consistent with recent calculations of the dust temperature distribution in yso envelopes ( e.g. , ceccarelli et al .
although the gas and the dust are not closely coupled thermally at the lower densities , the simplifying assumption @xmath137 is made for all radii .
recent calculations by ceccarelli et al.(1996 ) indicate that the gas kinetic temperature does not deviate from the dust temperature by more than a factor of @xmath138 . for comparison ,
calculations are also made for power - law density distributions , @xmath139 , with @xmath140 , 2 and 3 . the same outer radius @xmath114 , dust emissivity , hco@xmath0 abundance and temperature structure as for the collapse model are used .
since the continuum emission at 1.1 mm is optically thin , the flux as observed in a @xmath4 beam can be easily calculated as a function of time by direct integration along all lines of sight .
the results are shown in fig .
it should be stressed that the time - axis has no quantitative meaning , since the mass - accretion rate , and hence the mass of the envelope as function of time , is strongly dependent on the ill - constrained sound speed .
furthermore , the evolution is followed beyond the point where the assumption of self - similarity breaks down , as indicated by the vertical dashed line , in that the envelope fluxes observed toward almost the entire source sample fall beyond this line .
the results are therefore more realistically presented as a function of density at the arbitrary radius of 1000 au .
this approach is used in fig .
6b , which presents curves for the collapse and three power - law models . in table 5
the h@xmath10 number densities at 1000 au corresponding to the inferred envelope fluxes are listed for the collapse model curve ; values range between @xmath141 and @xmath142 @xmath19 , and depend on the adopted dust temperature .
since the disks are estimated to contribute 3075% to the total 1.1 mm flux , the inferred densities do not depend critically on the assumed value for the spectral index @xmath65 .
for haro 610 , where all single - dish flux can be attributed to a disk , only an upper limit to the density is found . to calculate the hco@xmath0 emission of the model envelopes , the excitation must to be solved together with the radiative transfer , since the lines are generally optically thick .
a one - dimensional monte carlo code has been used , based on the description of bernes ( 1979 ) .
the radiation field at all transition frequencies is represented simultaneously by 200 model photons propagating through the model envelope .
the hco@xmath143h@xmath10 collision rates of monteiro ( 1985 ) and green ( 1975 ) have been used , together with the line frequencies of blake et al .
the model is divided into 15 concentric shells , each with a density following the inside - out collapse model or any of the three power - law descriptions , a kinetic temperature , and fractional abundances for hco@xmath0 and h@xmath2co@xmath0 .
no abundance variations with radius have been adopted . instead of a detailed velocity field
, a turbulent line width of 1.0 kms@xmath39 fwhm has been used throughout the model envelope .
this produces line widths comparable to the observed spectra . for the excitation of lines for which the envelope is transparent ( e.g. , h@xmath2co@xmath0 ) , the details of the velocity field are not important .
significant optical depths are obtained for hco@xmath0 , but it is found that the resulting integrated line intensities do not depend strongly on the details of the velocity field , and stay well within the estimated uncertainty of a factor of @xmath144 originating from the other parameters of the source model , i.e. , sound speed , temperature , and abundance . after the molecular excitation has converged , the radiative transfer is solved .
the resulting sky - brightness distribution is then convolved with the appropriate beam sizes .
line intensities and fwhm source sizes predicted by the monte carlo code are shown in fig .
6c and 6d as functions of density for the four models .
the behavior of the curves reflects the increasing molecular column density with larger @xmath145 , resulting in higher line intensities and larger source sizes . at high column density ,
the intensities level off because of optical depth .
since the curves are parameterized by the density at 1000 au , the central density is lower for the @xmath140 model than for the @xmath146 model at the same @xmath145 .
this explains why the model curves can cross .
the behavior of the collapse - model curve is best understood in terms of time , indicated along the top of the panels , rather than density .
for example the decrease of the hco@xmath0 line intensities with time of 2030 kkms@xmath39 to a few kkms@xmath39 between @xmath147 yr and a @xmath148 yr reflects the decreasing density and column density of the envelope .
the h@xmath2co@xmath0 10 , 32 and 43 line intensities decrease from 1015 kkms@xmath39 to @xmath149 kkms@xmath39 over the same time span .
lower bounds to the observational sensitivities are typically 0.20.7 kkms@xmath39 , so that the data can probe the full parameter range .
the fwhm sizes of the beam - convolved model hco@xmath0 emission vary from 70@xmath31110@xmath31 ( 10 ) , 30@xmath31110@xmath31 ( 32 ) , and 20@xmath3160@xmath31 ( 43 ) .
the maximum in the convolved source size around @xmath150 yr corresponds to the moment when the collapse expansion wave reaches the outer radius of the envelope , and the density follows @xmath151 over a large region . a flatter density distribution results in a more extended sky - brightness distribution , and hence a larger convolved source size .
the time variation of the column density and the steepening of the density distribution drive the evolution of the integrated intensities and fwhm source sizes .
the small observed source sizes in 32 and 43 exclude density distribution as flat as @xmath152 .
otherwise , no strong constraints are placed on the model parameters by the source sizes . to investigate
how well the different models can describe the envelopes of our sources , the calculated hco@xmath0 intensities are plotted against the corresponding envelope fluxes in fig . 5 .
the individual sources are identified in the hco@xmath0 32 panel . whereas the continuum flux only traces mass , the hco@xmath0 lines probe both density and mass . for the inside - out collapse model
, time runs from the upper - right to lower - left corners of the panels ; the density at 1000 au decreases in that same direction for all four curves .
the observations are plotted with error bars corresponding to the @xmath153% calibration uncertainty . for t
tau , the relatively large contribution made by the outflow to the hco@xmath0 43 line wings is incorporated in the error bar .
overall , the model curves are seen to describe the observations well .
the uncertainty in the model parameters and the scatter in the data points preclude any constraints on the density model , although the h@xmath2co@xmath0 43 data seem to favor a slope @xmath154 .
to a large extent , the agreement between models and observations reflects the unsurprising fact that less massive envelopes are weaker in both continuum and line emission .
however , the correlation between the envelope flux and the hco@xmath0 intensity indicates that the observed range in either of these quantities is not due to variations in the temperature of the bulk of the dust or of the hco@xmath0 abundance .
this means that within the @xmath155 beam , no significant depletion of molecules has occurred . only for l1551 irs
5 do the observations fall significantly below the model curves in many of the hco@xmath0 and h@xmath2co@xmath0 panels .
one possible explanation is offered by the identification of this source as undergoing a fu orionis outburst ( cf .
mundt et al.1985 ) , which may result in an enhanced dust temperature and a larger continuum flux .
for the sources in general , the correspondence between model curves and observations is better for the hco@xmath0 32 and 43 lines , consistent with the maps , which show that significant contribution to the 10 line comes from the surrounding cloud .
it is concluded that the hco@xmath0 32 and 43 emission is a good tracer of the envelope mass .
in the previous section it was shown that the envelopes around our nine sources , as traced by their 1.1 mm continuum emission and hco@xmath0 32 and 43 lines , are characterized by a radial power - law density distribution with slopes @xmath11
this is consistent with the inside - out collapse model of shu ( 1977 ) , but more detailed comparisons with line profiles should be made before further statements about the validity of any specific collapse model can be made ( see hogerheijde et al .
1998c for such an analysis of the hco@xmath0 data presented here ) .
especially the hco@xmath0 32 and 43 lines are found to be especially robust tracers of the envelopes .
similar conclusions were reached by blake et al.(1994 ) for ngc 1333 iras 4 and van dishoeck et al.(1995 ) for iras 16293@xmath162422 .
a beam - averaged hco@xmath0 abundance of @xmath9 @xmath19 is inferred from c@xmath34o 10 observations , and is found to be well - matched to the continuum fluxes , i.e. , essentially undepleted abundances over @xmath156 scales .
typical masses within the @xmath4 jcmt beam of 0.0010.26 @xmath33 are found .
moriarty - schieven et al .
( 1995 ) used observations of cs 32 , 54 and 76 to investigate the envelopes around all sources from the tgww sample .
these transitions probe densities between @xmath15 and @xmath157 @xmath19 and temperatures between 14 and 66 k , and as such should be well suited as envelope tracers . in fig .
7 a comparison is made between their cs data and our hco@xmath0 32 measurements for the full tgww sample .
the correlation of the hco@xmath0 32 ( @xmath4 beam ) with the cs 32 line ( @xmath158 beam ) and with the 54 line ( @xmath159 cso beam ) is found to be poor , reflecting the large contribution by the surrounding cloud to these transitions .
for the 54 line ( @xmath156 jcmt beam ) and the 76 line ( @xmath156 beam ) , essentially the same trend is observed as in hco@xmath0 32 , although the hco@xmath0 32 is , on average , stronger by a factor of 510 .
thus , hco@xmath0 may be a more sensitive and easier to observe envelope tracer than cs .
does the observed trend of hco@xmath0 and 1.1 mm continuum flux as tracers of the envelopes also hold for the full sample defined by tgww ?
in particular , are those sources which remained undetected in hco@xmath0 32 also different in 1.1 mm dust continuum from the selected sources ? in fig . 8 the hco@xmath0
32 observations of the full tgww sample are plotted against their 1.1 mm fluxes from moriarty - schieven et al .
this plot is an extension of the hco@xmath0 32 panel of fig . 5 , with the only difference that no correction is made for the contribution of possible disks to the continuum flux , which is not known for the other sources . as a reference
, the model curve for inside - out collapse is also shown .
note that this plot shows that the point - source subtraction is not critical for the conclusions of 6 ; the data points of our source sample ( indicated by filled symbols ) are still reasonably well fit by the model curve .
the sources which were not included in our sample are indicated by the open symbols , and can be split into two groups : those with detected , but weak hco@xmath0 32 emission , and those undetected in this line .
the first group forms the low - brightness tail of the [email protected] mm - continuum distribution of our sample .
these are sources with envelopes of low mass , either because they already have accreted much of their core material , or because they are intrinsically low - mass objects .
the ysos in the second group are found to span the same range in 1.1 mm flux as our subsample , and are also weak in c@xmath34o 32 ( hogerheijde et al .
1998a ) , indicating that the lack of hco@xmath0 32 emission is due to a low total column density , and not to a low hco@xmath0 abundance or low density .
it is therefore concluded that the 1.1 mm continuum flux of these sources originates primarily in a circumstellar disk , much smaller than the @xmath4 jcmt beam , instead of an extended envelope .
interferometric observations of millimeter continuum toward several of these sources ( dg tau : dutrey et al .
1996 ; gg tau : ohashi et al .
1996 ) are consistent with this interpretation , assuming a spectral index of @xmath71 .
no detectable hco@xmath0 32 emission is expected from disks with such small beam - filling factors at the noise level achieved by our observations .
dutrey et al .
( 1997 ) detected an integrated hco@xmath0 32 intensity of 1.5 kkms@xmath39 toward gg tau in the @xmath160 iram 30 m beam , consistent with our upper limit of @xmath161 kkms@xmath39 in the @xmath4 jcmt beam for an unresolved source . in the simple evolutionary picture sketched in 1 , sources without envelopes but with disks are more evolved than those still surrounded by envelopes
obviously , intrinsically less massive objects appear to be more evolved if judged by the absolute mass of their envelope alone
. bontemps et al .
( 1996 ) and saraceno et al.(1996 ) used the envelope mass , as traced by the 1.1 mm flux , together with the bolometric luminosity to obtain an evolutionary ordering of an extended sample of ysos .
if the mass accretion rate @xmath162 is assumed to be constant , then the mass of the envelope at time @xmath163 is given by @xmath164 , where @xmath165 is the initial core mass .
the mass of the central object is given by @xmath166 .
the bolometric luminosity @xmath167 is a measure of @xmath168 , whether it is mainly stellar in origin , or whether it is assumed to be generated by mass accretion onto the star , @xmath169 , with @xmath170 the radius of the protostar . in this simple scheme
, the ratio of @xmath171 over @xmath167 gives the relative evolutionary ` phase ' of the object ; a yso starts off with large @xmath171 and low luminosity ( low @xmath168 ) , and evolves toward high luminosity ( large @xmath168 ) and low @xmath171 . in fig . 2 of saraceno et al.(1996 )
this ratio corresponds to the relative distance traveled along the evolutionary tracks .
based on the results of the previous section , the hco@xmath0 32 line strength can be used to trace the envelope mass uncontaminated by any contribution from a disk , which may be a problem with the 1.1 mm flux . in fig .
9 , the cumulative distribution of the quantity @xmath172 is shown for our nine sources , the five sources from the tgww sample with weak but detected hco@xmath0 , and the sources undetected in hco@xmath0 .
for this last group , the @xmath173 upper limits are used as if they were detections .
the first two groups appear to follow a similar distribution , clearly separated from the third . a kolmogorov
smirnov test indicates that there is a chance of @xmath174 that the third group is similar to either our sample or the union of the weak - hco@xmath0 sources with our sample .
this difference is based on the strict upper limits acquired on their hco@xmath0 32 line strengths ; it also shows , however , that they do not have the low bolometric luminosity expected from young , but intrinsically low - mass objects .
it can therefore be concluded that these sources form a group of more evolved objects than the ysos selected in 2 .
the hco@xmath0 line strength seems to be a better signpost of young age than either the 1.1 mm continuum flux or the iras color .
the assumption of a constant mass - accretion rate is not critical for this conclusion .
two types of variations may exist in @xmath175 .
firstly , the accretion rate may gradually change over time , and probably decrease ( cf .
foster & chevalier 1993 ) .
this results in an enhanced luminosity for young sources and a smaller range in @xmath176 .
since a clear range in @xmath177 is observed , its use as an evolutionary marker is not precluded by this type monotonic of change in @xmath175 .
secondly , long periods of relatively low accretion rates may be separated by short bursts of high accretion ( fu orionis effect , cf .
hartmann et al .
protostars are expected to spend @xmath91 of their lifetime in a fu ori phase , which means that only two or three sources of the tgww sample may be in this phase , and only one of our nine sources .
indeed , l1551 irs 5 is sometimes referred to as undergoing a fu ori outburst ( mundt et al .
the effect of the sudden increase in @xmath167 is that an object is shifted temporarily to lower @xmath176 ; i.e. , it appears more evolved and more massive than it really is .
the short period spend in a fu ori phase , as inferred from the very small number of known fu ori objects , ensures that for the 24 objects studied here the statistical distribution of @xmath176 is not seriously affected by an individual object undergoing a fu ori outburst .
a difficulty in using @xmath178 to trace evolution is the opacity of the hco@xmath0 32 line . in fig .
9 the cumulative distribution of our sources is shown after taking into account the known 32 line opacity ( see table 6 ) .
this correction is small for most sources .
in addition , it makes the difference between sources detected and undetected in hco@xmath0 32 larger , strengthening the conclusion that they form two distinct evolutionary groups . along the @xmath104-corrected curve
the position of our objects is indicated , giving a relative evolutionary ordering of our sample .
note that no absolute timescale can be assigned , and that the ordering only reflects how evolved an object is along its route from cloud core to star .
objects of different mass may proceed at different speeds along this track .
no correlation is found between either the point - source flux or the ratio of the point - source flux over the total 1.1 mm flux , and the evolutionary phase @xmath178 of our nine sources .
this suggests that the mass of the disks does not change significantly during the embedded phase .
there is a trend , however , between the point - source flux , the envelope flux , and the luminosity , indicating that the disk flux primarily depends on the mass or activity of the central object .
myers & ladd ( 1993 ) proposed a diagram of @xmath179 versus the bolometric temperature @xmath180 as an extension to the hertzsprung
russell diagram ( the so - called blt diagram ) .
the quantity @xmath180 is defined as the temperature of a blackbody having a maximum at the same frequency as the sed of the source . in fig .
10 the blt diagram of the tgww sample is shown ( chen et al .
1995 ) , with our subsample again indicated by filled symbols .
it is seen to populate the upper right - hand region of the diagram , corresponding to young sources .
although no calculations of evolutionary tracks have been published for a blt diagram yet , a source is expected to proceed from low @xmath180 and @xmath167 to the main sequence via a maximum in @xmath167 .
compared to our evolutionary tracer , @xmath181 , which requires the presence of envelope material , the blt approach extends over a much larger range in object age .
our method has the advantage , however , that mass and evolution can be separated independent of any collapse model , and that only a sufficiently well - behaved @xmath162 is required .
a more extended survey of hco@xmath0 and h@xmath2co@xmath0 in a larger variety of objects will therefore be interesting to further investigate this scenario .
the envelopes around a sample of nine embedded ysos have been investigated with 3.4 and 2.7 mm continuum interferometry , and with single - dish observations of the hco@xmath0 and h@xmath2co@xmath0 10 , 32 and 43 transitions .
the sources were selected from the iras flux and color - limited sample defined by tgww .
the results of this paper can be summarized as follows . 1
. continuum emission at 3.4 and 2.7 mm is detected in the ovro beam toward all nine sources with integrated fluxes ranging between 4 mjy and 200 mjy .
the emission can be described by an unresolved ( @xmath3 ) point source and , in about half of the objects , emission from an extended , partially resolved envelope .
the point - source emission is attributed to thermal dust emission from a circumstellar disk , and its high detection rate suggests that such disks are established already early in the formation and evolution of protostars .
no significant change in the ratio of disk mass over envelope mass is inferred during the subsequent duration of the embedded phase . 2 . by extrapolating the point - source fluxes to 1.1 mm , assuming a spectral index @xmath71 , the relative contributions of the disk and the envelope to the single - dish fluxes observed by moriarty - schieven et al .
( 1994 ) are estimated .
the point sources typically contribute 3075% to the single - dish flux .
the inferred envelope masses within the @xmath4 jcmt beam are 0.0010.26 @xmath33 , assuming an average dust temperature of 30 k. 3 .
@xmath8=10 emission maps reveal extended cores ( @xmath182 ) superposed on emission from the surrounding cloud . in the 32 and 43 lines
the emission is concentrated toward the source positions and marginally resolved ( @xmath183 ) , as is expected for a power - law density distribution .
the spectra are generally best fit by a narrow emission line ( @xmath184 23 kms@xmath39 ) , broad line wings , and , in a few cases , a narrow self - absorption component .
the h@xmath2co@xmath0 32 and 43 lines are detected toward half of our sources , with spectra well fit by a single gaussian at the @xmath185 of the hco@xmath0 self - absorption , if present . from observations of h@xmath2co@xmath0 10 by mizuno et al .
( 1994 ) and c@xmath34o 10 by hayashi et al .
( 1994 ) , a beam - averaged hco@xmath0 abundance of @xmath9 with respect to h@xmath10 is derived .
4 . the 1.1 mm envelope flux and the integrated hco@xmath0 line strengths
are well correlated , especially for the higher @xmath8 transitions .
they can both be modeled with the inside - out collapse model of shu ( 1977 ) and with simple power - law density distributions ( slopes @xmath11 13 ) , assuming that the temperature of the dust and the gas has a power law with slope @xmath186 , and a value of 30 k at a radius of 1000 au .
typical densities at @xmath187 au of @xmath188 to @xmath142 @xmath19 are inferred .
the fact that the models agree well with the observations indicates that hco@xmath0 32 and 43 can be used as envelope tracers , and that no significant depletion of molecules takes place over @xmath183 scales .
however , more detailed comparison with line - profile calculations are required before further constraints can be placed on possible collapse models .
among the remaining objects of the original sample , i.e. , those not selected for our subsample , there is a group that have similar 1.1 mm single - dish fluxes compared to our ysos in spite of undetected hco@xmath0 32 emission .
it is proposed that all their 1.1 mm flux originates in a disk rather than in an extended envelope .
available c@xmath34o 32 and continuum - interferometer data support this interpretation .
the quantity @xmath178 is proposed as an evolutionary tracer independent of intrinsic mass , and measures the ratio of the mass of the central star to that of the envelope .
it is found that those tgww sources which are detected in hco@xmath0 32 have a significantly different distribution in this tracer compared with sources that remain undetected in hco@xmath0 .
this supports the interpretation of the latter group as more evolved objects , and shows that hco@xmath0 is very well suited as a sensitive probe of the early embedded phase of low - mass star formation .
the authors wish to thank remo tilanus , gran sandell , fred baas , and floris van der tak for carrying out part of the jcmt observations .
lee mundy is acknowledged for useful discussions concerning the analysis of the continuum observations .
the staffs of ovro , jcmt and iram 30 m are thanked for support during various observing runs .
mrh is indebted to the caltech divisions of geological and planetary sciences and of mathematics , physics and astronomy , as well as the owens valley radio observatory for their hospitality .
the _ stimuleringsfonds internationalisering _ of the netherlands organization for scientific research ( nwo ) and the leids kerkhoven
bosscha fonds provided travel support for mrh .
research in astrochemistry in leiden is supported by nwo / nfra through grant no .
gab gratefully acknowledges support provided by nasa grants nagw2297 and nagw1955 , and hjvl support by the european union under contract chgect920011 .
the referee is thanked for providing insightful comments , which led to a much improved revised manuscript .
agladze , n. i. , sievers , a. j. , jones , s. a. , burlitch , j. m. , & beckwith , s. v. w. 1994 , , 372 , 243 andr , ph . , & montmerle , t. 1994 , , 420 , 837 barsony , m. , & kenyon , s. j. 1992 , , 384 , l53 beckwith , s. v. w. , sargent , a. i. , chini , r. s. , & gsten , r. 1990 , , 99 , 924 bernes , c. 1979 , , 73 , 67 bertout , c. 1983 , , 126 , l1 blake , g. a. , laughlin , k. b. , cohen , r. c. , busarow , k. l. , & saykally , r. j. 1987 , , 316 , l45 blake , g. a. , van dishoeck , e. f. , jansen , d. j. , groesbeck , t. d. , & mundy , l. g. 1994 , , 428 , 680 bontemps , s. , andr , ph . , terebey , s. , & cabrit , s. 1996 , , 311 , 858 boss , a. p. 1993
, , 410 , 157 butner , h. m. , natta , a. , & evans ii , n. j. 1994 , , 420 , 326 ceccarelli , c. , hollenbach , d. j. , & tielens , a. g. g.m .
1996 , , 471 , 400 chen , h. , myers , p. c. , ladd , e. f. , & wood , d. o. s. 1995 , , 445 , 377 choi , m. h. , evans ii , n. j. , gregersen , e. m. , & wang , y. s. 1995 , , 448 , 742 dutrey , a. , guilloteau , s. , duvert , g. , et al .
1996 , , 309 , 493 dutrey , a. , guilloteau , s. , gulin , m. 1997 , , 317 , l55 dyck , h. m. , simon , t. , & zuckerman , b. 1982 , , 255 , l103 edwards , s. , & snell , r. l. 1982 , , 261 , 151 fiedler , r. a. , & mouschovias , t. c. 1992 , , 391 , 199 fiedler , r. a. , & mouschovias , t. c. 1993 , , 415 , 680 foster , p. n. , & chevalier , r. a. 1993 , , 416 , 303 fuller , g. a. , ladd , e. f. , & hodapp , k .- w .
1996 , , 463 , l97 galli , d. , & shu , f. h. 1993 , , 417 , 220 ghez , a. m. , neugebauer , g. , & matthews , k. 1993 , , 106 , 2005 green , s. 1975 , , 201 , 366 hartmann , l. , kenyon , s. , & hartigan , p. 1993 , in protostars and planets iii , eds . e. h. levy , j. i. lunine ( tucson : univ .
arizona ) , 497 hayashi , m. , hasegawa , t. , ohashi , n. , & sunada , k. 1994 , , 426 , 234 hogerheijde , m. r. , van dishoeck , e. f. , blake , g. a. , & van langevelde , h. j. 1998a , in preparation .
hogerheijde , m. r. , van langevelde , h. j. , blake , g. a. , mundy , l. g. , & van dishoeck , e. f. 1998b , in preparation .
hogerheijde , m. r. , et al . 1998c , in preparation .
irvine , w. m. , goldsmith , p. f. , & hjalmarson , .
1987 , in interstellar processes , proc .
iau symp .
120 , eds .
d. j. hollenbach , h. a. thronson ( dordrecht : reidel ) , 561 keene , j. , & masson , c. r. 1990 , , 355 , 635 kenyon , s. j. , calvet , n. , & hartmann , l. 1993a , , 414 , 676 kenyon , s. j. , whitney , b. a. , gomez , m. , & hartmann , l. 1993b , , 414 , 773 lada , c. j. 1987 , in star forming regions , eds .
m. peimbert , j. jugaku ( dordrecht : reidel ) , 1 ladd , e. f. , adams , f. c. , fuller , g. a. , myers , p. c. , casey , s. , davidson , j. a. , harper , d. a. , & padman , r. 1991 , , 382 , 555 lay , o. p. , carlstrom , j. e. , hills , r. e. , & phillips , t. g. 1994 , , 434 , l75 leinert , c. , & haas , m. 1989 , , 342 , l39 leinert , c. , zinnecker , h. , weitzel , n. , christou , j. , ridgway , s. t , jameson , r. , haas , m. , & lenzen , r. 1993 , , 278 , 129 looney , l. w. , mundy , l. g. , & welch , w. j. 1996 , , 189 , 53.02 millar , t. j. , bennett , a. , rawlings , j. m. c. , brown , p. d. , charnley , s. b. 1991 , 87 , 585 mizuno , a. , ohishi , t. , hayashi , m. , ohashi , n. , sunada , k. , hasegawa , t. , & fukui , y. 1994 , , 368 , 719 monteiro , t. s. 1985 , , 214 , 419 moriarty - schieven , g. h. , wannier , p. g. , tamura , m. , & keene , j. 1992 , , 400 , 260 moriarty - schieven , g. h. , wannier , p. g. , keene , j. , & tamura , m. 1994 , , 436 , 800 moriarty - schieven , g. h. , wannier , p. g. , mangum , j. g. , tamura , m. , & olmsted , v. k. 1995 , , 455 , 190 mundt , r. , stocke , j. , strom , s. e. , strom , k. m. , anderson , e. r. 1985 , 297 , l41 mundy , l. g. , looney , l. w. , erickson , w. , grossman , a. , welch , w. j. , forster , j. r. , wright , m. c. h. , plambeck , r. l. , lugten , j. , & thornton , d. d. 1996 , , 464 , l169 myers , p. c. , & ladd , e. f. 1993 , , 413 , l47 ohashi , n. , kawabe , r. , hayashi , m. , & ishiguro , m. 1991 , , 102 , 2054 ohashi , n. , hayashi , m. , kawabe , r. , & ishiguro , m. 1996 , , 466 , 317 ossenkopf , v. , & henning , th .
1994 , , 291 , 943 osterloh , m. , & beckwith , s. v. w. 1995 , , 439 , 288 palla , f. , & stahler , s. w. 1993 , , 418 , 414 pollack , j. b. , hollenbach , d. , beckwith , s. v. w. , simonelli , d. p. , roush , t. , & fong , w. 1994 , , 421 , 615 rodrguez , l. f. , myers , p. c. , cruz - gonzlez , i. , & terebey , s. 1989 , , 347 , 461 rowan - robinson , m. 1980 , , 44 , 403 saraceno , p. , andr , ph . , ceccarelli , c. , griffin , m. , & molinari , s. 1996 , , 309 , 827 schuster , k. f. , harris , a. i. , andersen , n. , & russel , a. p. g. 1993 , , 412 , l67 scoville , n. z. , sargent , a. i. , sanders , d. b. , claussen , m. j. , masson , c. r. , lo , k. y. , & phillips , t. g. 1986 , , 303 , 416 scoville , n. z. , carlstrom , j. e. , chandler , c. j. , phillips , j. a. , scott , s. l. , tilanus , r. p. j. , & wang , z. 1993 , , 105 , 1482 shu , f. h. 1977 , , 214 , 488 shu , f. h. , najita , j. , galli , d. , ostriker , e. , & lizano , s. 1993 , in protostars and planets iii , eds .
e. h. levy , j. i. lunine ( tucson : univ .
arizona ) , 3 skinner , s. l. , & brown , a. 1994 , , 107 , 1461 stahler , s. w. 1988 , , 332 , 804 tamura , m. , gatley , i. , waller , w. , & werner , m. w. 1991 , , 374 , l25 ( tgww ) terebey , s. , shu , f. h. , & cassen , p. 1984
, , 286 , 529 terebey , s. , vogel , s. n. , & myers , p. c. 1989 , , 340 , 472 terebey , s. , chandler , c. j. , & andr , ph .
1993 , , 414 , 759 van dishoeck , e. f. , blake , g. a. , jansen , d. j. , & groesbeck , t. d. 1995 , , 447 , 760 van langevelde , h. j. , van dishoeck , e. f. , & blake , g. a. 1994 , , 425 , l45 van langevelde , h. j. , van dishoeck , e. f. , & blake , g. a. 1997 , in co : twenty - five years of millimeter - wave spectroscopy , proc .
iau symp .
170 , eds .
w. b. latter et al .
( dordrecht : kluwer ) , 469 walker , c. k. , lada , c. j. , young , e. t. , maloney , p. r. , & wilking , b. a. 1986 , , 309 , l47 walker , c. k. , narayanan , g. , & boss , a. p. 1994
, , 431 , 767 ward - thompson , d. , buckley , h. d. , greaves , j. s. , et al . 1996 , , 281 , l53 zhou , s. 1992 , , 394 , 204 zhou , s. 1995 , , 442 , 685 zhou , s. , evans ii , n. j. , kmpe , c. , & walmsley , c. m. 1993 , , 404 , 232 lrrrcrrrrrr l1489 irs & 04016 + 2610 & 04 01 40.5 & + 26 10 48 & emb & 9.3 & 2.0 & 56.0 & @xmath189 & 3.70 & 0.4 t tau & 04190 + 1924 & 04 19 04.1 & + 19 25 06 & vis & ( 5.4 ) & 0.9 & 98.1 & @xmath190 & 25.50 & 2.7 haro 610 & 04263 + 2426 & 04 26 21.9 & + 24 26 29 & vis & 7.6 & 2.6 & 49.3 & @xmath29 & 6.98 & 0.9l1551 irs 5 & 04287 + 1801 & 04 28 40.2 & + 18 01 42 & emb & 9.3 & 2.8 & 457.9 & @xmath191 & 21.90 & 2.6l1535 irs & 04325 + 2402 & 04 32 33.4 & + 24 02 13 & emb & 11.1 & 1.8 & 23.0 & @xmath192 & 0.70 & 0.15 tmr 1 & 04361 + 2547 & 04 36 09.7 & + 25 47 29 & emb & 10.6 & 2.5 & 33.1 & @xmath193 & 2.90 & 0.3tmc 1a & 04365 + 2535 & 04 36 31.1 & + 25 35 54 & emb & 10.6 & 2.2 & 38.0 & @xmath194 & 2.20 & 0.3l1527 irs & 04368 + 2557 & 04 36 49.6 & + 25 57 21 & emb & ( 13.0 ) & 2.1 & 71.0 & @xmath195 & 1.30 & 0.2tmc 1 & 04381 + 2540 & 04 38 08.4 & + 25 40 52 & emb & 12.0 & 2.3 & 12.6 & @xmath196 & 0.66 & 0.15 llcrrrrrr 04016 + 2610 & l1489 irs & emb & & & 238 & 3.70 & @xmath197 & @xmath19804108 + 2803 & & emb & & & 205 & 0.72 & @xmath199 & @xmath200 04113 + 2758 & & emb & & & 606 & 2.0 & @xmath201 & @xmath202 04169 + 2702 & & emb & & & 170 & 0.80 & @xmath203 & @xmath204 04181 + 2655 & & emb & & & 278 & 0.43 & @xmath205 & @xmath206 04190 + 1924 & t tau & vis+emb & & & & 25.50 & @xmath207 & @xmath208 04191 + 1523 & & emb & & & & 0.48 & @xmath209 & @xmath210 04239 + 2436 & & emb & & & 236 & 1.27 & @xmath211 & @xmath212 04240 + 2559 & dg tau & vis & & & 1440 & 6.36 & @xmath213 & @xmath214 04248 + 2612 & hh 31 irs & emb & & & 334 & 0.36 & @xmath215 & @xmath216 04263 + 2426 & haro 610 & vis+emb & 4730 & & 253 & 6.98 & @xmath217 & @xmath218 04287 + 1801 & l1551 irs 5 & emb & & & 97 & 21.90 & @xmath219 & @xmath220 04288 + 2417 & hk tau & vis & 3785 & 1.30 & 2148 & 0.81 & @xmath221 & @xmath222 04292 + 2422 & haro 613 & vis & & & 910 & 1.30 & @xmath223 & @xmath224 04295 + 2251 & l1536 irs & emb & & & 447 & 0.44 & @xmath225 & @xmath224 04296 + 1725 & gg tau & vis & 4060 & 1.31 & 2621 & 2.00 & @xmath226 & @xmath227 04302 + 2247 & & emb & & & 202 & 0.34 & @xmath228 & @xmath224 04325 + 2402 & l1535 irs & emb & & & 157 & 0.70 & @xmath229 & @xmath230 04328 + 2248 & hp tau & vis & 4730 & 1.22 & 2748 & 2.40 & @xmath199 & @xmath222 04361 + 2547 & tmr 1 & emb & & & 144 & 2.90 & @xmath231 & @xmath232 04365 + 2535 & tmc 1a & emb & & & 172 & 2.20 & @xmath233 & @xmath234 04368 + 2557 & l1527 irs & emb & & & 59 & 1.30 & @xmath235 & @xmath236 04381 + 2540 & tmc 1 & emb & & & 139 & 0.66 & @xmath237 & @xmath238 04390 + 2517 & lkh@xmath65 332/g2 & vis & 4060 & 1.94 & 2563 & 1.60 & @xmath199 & @xmath214 llll 1992/4 , 1993/7 & ovro & @xmath239 & t tau 1993/10 , 1994/24 & ovro & @xmath239 & all , except t tau 1993/1,6 & ovro & @xmath240 & t tau 1995/25 & ovro & @xmath240 & haro 610 , l1551 irs 5 , l1535 irs , tmr 1 , l1527 irs , tmc 1 1996/10 , 1997/2 & ovro & @xmath240 & l1489 irs , tmc 1a 1991/5 & iram 30 m & hco@xmath0 10 & t tau 1995/5 & iram 30 m & hco@xmath0 10 & all , except t tau 1993/8 & jcmt & hco@xmath0 32 & full source sample ( see table 2 ) 1995/8 & jcmt & hco@xmath0 32 & l1489 irs , t tau , tmr 1 , tmc 1a , tmc 1 1994/12 & jcmt & hco@xmath0 43 & all1994/12 , 95/10 , 96/5 & jcmt & h@xmath2co@xmath0 32 , 43 & all lrrrrrrrr l1489 irs & @xmath241 & & & & @xmath242 & & & @xmath243 t tau & @xmath244 & & & & @xmath245 & @xmath246 & @xmath247 & @xmath248 haro 610 & @xmath249 & & & & @xmath250 & @xmath251 & @xmath252 & @xmath253 l1551 irs 5 & @xmath254 & @xmath255 & @xmath256 & & @xmath257 & @xmath258 & @xmath247 & @xmath259 l1535 irs & @xmath260 & & & & @xmath261 & & & @xmath262 tmr 1 & @xmath263 & & & & @xmath264 & @xmath265 & @xmath266 & @xmath267 tmc 1a & @xmath268 & @xmath269 & @xmath270 & & @xmath271 & & & @xmath272 l1527 irs & @xmath273 & @xmath274 & @xmath275 & & @xmath276 & @xmath277 & @xmath278 & @xmath279 tmc 1 & @xmath280 & & & & @xmath281 & & & @xmath282 lrrrr l1489 irs & @xmath283 & 112180 & 0.0160.025 & @xmath284@xmath285 t tau & @xmath286 & @xmath287 & @xmath288 & @xmath289 haro 610 & @xmath290 & @xmath291 & @xmath292 & @xmath293 l1551 irs 5 & @xmath294 & @xmath295 & @xmath296 & @xmath297 l1535 irs & @xmath298 & 074 & @xmath299 & @xmath300 tmr 1 & @xmath301 & @xmath302 & @xmath303 & @xmath304 tmc 1a & @xmath305 & @xmath306 & @xmath307 & @xmath308 l1527 irs & @xmath309 & @xmath310 & @xmath311 & @xmath312 tmc 1 & @xmath313 & 38116 & 0.0050.016 & @xmath314@xmath315 lrrrrrrrrr l1489 irs & @xmath316 & 60 & & @xmath317 & 24 & @xmath318 & @xmath319 & 17 & @xmath320 t tau & @xmath321 & & & @xmath322 & 25 & @xmath323 & @xmath324 & 20 & @xmath325 haro 610 & @xmath326 & 90 & & @xmath327 & & @xmath328 & @xmath329 & 23 & @xmath330 l1551 irs 5 & @xmath331 & 140 & @xmath332 & @xmath333 & & @xmath334 & @xmath335 & 22 & @xmath336 l1535 irs & @xmath337 & 110 & @xmath338 & @xmath339 & & @xmath340 & @xmath341 & 24 & @xmath342 tmr 1 & @xmath343 & 48 & @xmath264 & @xmath344 & 30 & @xmath345 & @xmath346 & 24 & @xmath347 tmc 1a & @xmath348 & 75 & @xmath349 & @xmath350 & 19 & @xmath340 & @xmath351 & 20 & @xmath352 l1527 irs & @xmath353 & 63 & @xmath354 & @xmath355 & & @xmath356 & @xmath357 & & @xmath358 tmc 1 & @xmath348 & 66 & & @xmath359 & 26 & @xmath340 & @xmath360 & 19 & @xmath361 l1489 irs & & & & @xmath362 & & & @xmath363 & & t tau & & & & @xmath364 & & & @xmath365 & & haro 610 & & & & @xmath366 & & & @xmath367 & & l1551 irs 5 & @xmath368 & & & @xmath369 & & & @xmath370 & & l1535 irs & @xmath371 & & & @xmath366 & & & @xmath227 & & tmr 1 & @xmath372 & & & @xmath373 & & & @xmath214 & & tmc
1a & @xmath374 & & & @xmath366 & & & @xmath375 & & l1527 irs & @xmath376 & & & @xmath377 & & & @xmath378 & & tmc 1 & & & & @xmath366 & & & @xmath379 & & | the envelopes and disks around embedded low - mass young stellar objects ( ysos ) are investigated through millimeter continuum and hco@xmath0 line emission .
nine sources , selected on the basis of their hco@xmath0 32 emission from an iras flux- and color - limited sample of 24 objects , are observed in @xmath1 and 2.7 mm continuum emission with the owens valley millimeter array , and in the hco@xmath0 and h@xmath2co@xmath0 43 , 32 , and 10 transitions at the james clerk maxwell and iram 30 m telescopes .
all nine sources are detected at 3.4 and 2.7 mm in the interferometer beam , with total fluxes between 4 and 200 mjy .
the visibilities can be fit with an unresolved ( @xmath3 ) point source , and , in about half of the sources , with an extended envelope .
the point sources , presumably thermal dust emission from circumstellar disks , typically contribute 3075% of the continuum flux observed at 1.1 mm in a @xmath4 beam , assuming a spectral slope of 2.5 .
the fact that at least two - thirds of our sources show point - source emission indicates that circumstellar disks are established early in the embedded phase .
the remainder of the 1.1 mm single dish flux is attributed to an extended envelope , with a mass of @xmath5@xmath6 @xmath7 within a @xmath4 beam . in hco@xmath0 ,
the @xmath8=10 line is seen to trace the surrounding cloud , while the emission from @xmath8=32 and 43 is concentrated toward the sources .
all sources look marginally resolved in these lines , indicative of a power - law brightness distribution .
a beam - averaged hco@xmath0 abundance of @xmath9 with respect to h@xmath10 is derived .
the 1.1 mm continuum fluxes and hco@xmath0 line intensities of the envelopes correlate well , and are modeled with the simple inside - out collapse model of shu ( 1977 ) and with power - law density distributions of slopes @xmath11 13 .
all models provide satisfactory fits to the observations , indicating that hco@xmath0 is an excellent tracer of the envelopes .
of the 15 sources of the original sample that were either undetected in hco@xmath0 32 or too weak to be selected , seven show 1.1 mm single - dish fluxes comparable to our objects .
it is proposed that all of the 1.1 mm flux of the former sources should be attributed to compact circumstellar disks .
the relative evolutionary phase of a yso , defined as the current ratio of stellar mass over envelope mass , is traced by the quantity @xmath12 .
sources which are undetected in hco@xmath0 are found to to have significantly lower values in this tracer than do the objects of our subsample , indicating that the former objects are more evolved .
the sources which are weak in hco@xmath0 32 are indistinguishable from our subsample in this tracer , and have intrinsically low masses .
it is concluded that hco@xmath0 , especially in its 32 and 43 transitions , is a sensitive tracer of the early embedded phase of star formation . |
with the experimental set [ setup ] , analyze the dependence between current and voltage of the `` black box '' , that must not be opened during the olympiad . here
follow two quality problems , described in section [ quatity_tasks ] , which aim to verify whether the device , hidden in the `` black box '' , works .
section [ black_box_investigation ] describes in detail the experiments that must be carried out with the set to analyse vac of the `` black box '' .
then in section [ the_world_is_not_perfect ] will make a more detailed analysis and you will see that the physical properties which are found up to now on this subject are valid within certain limits . then follows the static analysis section [ dynamics ] , which will explore the dynamic characteristics of the `` black box '' by mechanical experiments .
there is also the purely theoretical section [ theoretical problem ] connected to the theoretical description of the proposed experiments .
we suggest the students who are not very confident in the experiment , but better cope with mathematics , to concentrate on the theoretical problem . for the tireless
there is a homework assignment , described in section [ homework ] .
in the set of . [ setup ] , there is one `` black box '' and one led with long cables and two alligator clips .
connect the leds to the `` black box '' and if the led does not light , swap the `` alligator clips '' . at least in one of the polarities the led lights .
upon further quality problems will have to measure current and voltage in leds and explain how these variables are related to the current - voltage characteristics of the elements .
parallel to the `` black box '' and the led connect a resonant circuit .
move the led and you will see that the light pulsates now . moreover ,
if you include parallel and piezo element , as shown in . [ lcnr ] , you will hear a faint buzz .
confirm whether the led flashes and the piezo element buzzes .
the following tasks are related to the detailed quantitative analysis of these oscillations and their theoretical explanation . with alligator clips.,width=332 ] in which of the two quality problems the led is brighter constantly or pulsating emission ?
_ the subpoints of the section [ black_box_investigation ] , are addressed to younger students .
the problems are easier and give less points .
_ 1 . * measure the voltage and the current through the `` black box '' , related to the led and without it .
( 7 points ) * + connect the circuit in figure [ hypothesis_reject ] and measure the voltage @xmath0 on the `` black box '' and the current @xmath1 , which flows through it .
if the led of the first or the last circuit does not light up , change its polarity .
the results of the measurements write down in the table , as shown in the exemplary table [ template_4_setups ] . + what causes the small difference between the voltages @xmath2 and @xmath3 ?
+ add an amperemeter and a voltmeter .
their values are @xmath2 and @xmath4 .
( b ) replace the led with wire and the values are @xmath5 and @xmath6 .
( c ) we leave the the amperemeter and the value is @xmath7 .
( d ) replace the amperemeter and the voltmeter and the value is @xmath8 .
( e ) measure the voltage @xmath3 and the current @xmath9 according to the following scheme .
, width=612 ] + @xmath0 [ @xmath10 ] + a ) & @xmath11 & @xmath12 + b ) & @xmath13 & @xmath14 + c ) & @xmath15 & + d ) & & @xmath16 + e ) & @xmath17 & @xmath18 + + [ template_4_setups ] 2 .
* measure the voltage of the battery @xmath191.5 v. ( 1 point ) * + use the multimeter as a voltmeter and measure the voltage of the battery .
the device displays the sign of the voltage .
you can remember the rules for characters and that the black lead is connected to the input of the device marked with `` ground '' ( ) or com and red to one of the other inputs .
* measure the resistance of the big white resistor .
( 1 point ) * + turn the multimeter as an ohmmeter and measure and note the resistance @xmath20 of the big white resistor with thermal lining of white cement .
work with an accuracy of 1 @xmath21 @xmath20= ?
4 . * measure the resistance of the five small resistors .
( 3 points ) * + on the connecting wires of the five resistors , that are given to you , stick yellow label .
switch the multimeter to operate as an ohmmeter .
write the numerical values of the resistances on their labels .
arrange them in size and write numbers on the stickers .
represent the resistances in a table @xmath22 , work with an accuracy of 1 @xmath21 5 . * using the five small resistors and battery of @xmath191.5 v measure the relationship between current and voltage of the big white resistor with resistance @xmath20 .
( 7 points ) * + an electrical circuit to measure the relationship between current and voltage is shown on fig .
[ resistance - measurement ] and fig .
[ r_neg_r_with_crocodiles ] .
use one multimeter as an amperemeter and plug it consecutively to the white resistor .
watch for the signs the current has direction ! the other multimeter turn as a voltmeter , parallel to the white resistor and again watch out for signs and polarity . for the big white resistor
acts the ohm s law @xmath23 .
if the voltage is positive , the current is positive , otherwise the voltage is negative and the current is negative . if the signs of voltage and current are opposite see where you went wrong in linking of the devices .
+ and the ampermeter ( a ) is connected in series and measure the current @xmath1 . when the circuit is closed with different resistances @xmath24 current and voltage are different .
thus are obtained in several points of vac . for small voltages
ratio @xmath25 is a constant and this is one possible way to verify ohm s law and signs of the measured current and voltage .
resistor @xmath20 on the scheme on the left is replaced by the `` black box '' of the figure to the right and this is the only difference in the two circuits . ,
width=434 ] + with alligator clips.,width=566 ] + ampermeter connect in series with the battery of 1.5 v. close the circuit consistent with resistances 0 @xmath26 separately the 5 @xmath27 for each measurement note in a table 5 columns and 6 lines in the form given in table [ template ] : 1 ) number of resistors @xmath28 ,
2 ) resistance @xmath29 , 3 ) current @xmath30 and 4 ) the voltage of the voltmeter @xmath31 5 ) the calculated resistance of the measuring element @xmath32 .
on the electric circuit shown in figure [ circuit ] , of the circuit with the three resistors @xmath73 @xmath74 @xmath75 and the triangle shown is a voltage amplifier with an amplification @xmath76 that is powered from two batteries with a voltage @xmath77 .
the voltages @xmath78 , @xmath79 , as well as the current @xmath1 are unknown
. the notation @xmath78 originates from @xmath80.,width=340 ] the input currents in the nodes ( + ) and ( ) are zero , and the output currect in the node ( 0 ) is such that the corresponding voltages are related @xmath81 , where the gain coefficient @xmath82 has very big value .
the triangle sign notes an electric element ( called amplifier ) that is sourced with two batteries with electromotive force @xmath83 the node between the two batteries is connected with wire to the resistor @xmath84 and one of the inputs of the schematics .
it is very convinient to choose the potential of this node to be zero @xmath85 .
the engineers call this node `` ground '' ( ) .
the index @xmath86 originates from the english words common point . on the multimeter
this node is noted as com . in the opposite case if the @xmath87 , then the equation for the voltage amplification becomes @xmath88 .
the current that flows through the ground node is equal to zero . calculate the ratio between the input voltage @xmath0 and the current @xmath1 that flows through the entire circuit with accuracy 1% .
find the effective resistance @xmath89 as a function of the three resistors in the circuit .
for simplicity you can assume that the gain @xmath90 goes to infinity @xmath91 .
replace the following example values of the resistors @xmath92 in the found expression . in short : find the ( 1 ) final expression for the resistance of the entire circuit ( 2 ) calculate the value of the resistance with accuracy 1% .
what is the sign of the expression @xmath25 and what is the module of its value . how the resistance light on the led ?
red and robust flying .
_ after the olympiad , find a screwdriver and remove the screws from the cover of the `` black box '' .
take out the batteries or turn one of the switches from position `` on '' to `` off '' .
_ the `` black box '' vac ( volt - ampere curve ) for small voltages is a straight line with a constant relation @xmath94 with the dimension of resistance , just like the ohm s law .
try to measure the resistance of the `` black box '' with an ohmmeter .
compare the readings of the ommeter and the determined resistance via an examination of the vac .
explain why the determined resistance via the vac could not be measured with an ohmmeter ?
what modification has to be done in the `` black box '' circuit shown in fig . [ circuit ] and
why will it make possible the measurement with an ohmmeter , for instance with the multimeter dt-830b that you were given at the olympiad in the range 20 k@xmath26 ?
the first participant that finds the answer of at least one of both questions and sends it from his / her registered for the olympiad email address to ` [email protected] ` by 07:00 on november 1 2015 earns the prize of 137 $ .
you can work in teams , read books , use _ google _ and consult radio engineers and university professors in electronics worldwide through internet . at midnight in kumanovo , it is late afternoon in california , while the working day in japan starts there are always working colleagues around the globe .
the authors are thankful to albert varonov for the critical reading of the manuscript .
, . [ bg_setup ] , @xmath38 - ( ) `` '' , . , 5 , , 1 50 cm , - , , 4 ( ) , 9 v , ( @xmath64 ) , - `` '' .
.,width=332 ] [ bg_quatity_tasks ] , `` '' . [ bg_black_box_investigation ] , `` '' .
[ bg_the_world_is_not_perfect ] - , , , . [ bg_dynamics ] , `` '' . [ bg_theoreticalproblem ] . , , - . [ bg_homework ] 137 $ ; ( 01.11.2015 ) 7:00
. , . . , , .
, . [ bg_setup ] , `` '' , 2 `` '' .
`` '' `` '' .
. - - . `` '' @xmath64 @xmath70 @xmath69 @xmath95 .
, . , - , . [ bg_lcnr ] , . - . . `` '' .
, ,
- . , width=332 ] , , ( `` '').,width=332 ] - , ?
_ [ bg_black_box_investigation ] , - . ( 18 ) - - . , , . _ 1 .
* `` '' . ( 7 ) * + . [ bg_hypothesis_reject ] @xmath0 `` '' @xmath1 , . . , . [ bg_template_4_setups ] .
+ @xmath2 @xmath3 ?
+ . @xmath2 @xmath4 .
( b ) @xmath5 @xmath6 .
( c ) @xmath7 .
( d ) @xmath8 .
( e ) @xmath3 @xmath9 . . [ bg_template_4_setups ] .
, width=612 ] + @xmath0 [ @xmath10 ] + a ) & @xmath11 & @xmath12 + b ) & @xmath13 & @xmath14 + c ) & @xmath15 & + d ) & & @xmath16 + e ) & @xmath17 & @xmath18 + + [ bg_template_4_setups ] 2 .
* @xmath191.5 v. ( 1 ) * + ( , ) @xmath96 1.5 v. . , , `` '' ( ) com .
* . ( 1 ) * + @xmath20 . 1 @xmath21 @xmath20= ?
* . ( 3 ) * + . . . . @xmath22 , 1 @xmath21 5 .
* @xmath191.5 v @xmath20 .
( 7 ) * + . [ bg_resistance - measurement ] . [ bg_r_neg_r_with_crocodiles ] .
m , , . !
. @xmath97 .
, , , . . + , ( a ) @xmath1 .
@xmath24 . . @xmath25
. @xmath20 `` '' . , width=434 ] + , `` ''.,width=566 ] + 1.5 v. 0 @xmath26 5- @xmath27 5 6 . [ bg_template ] : 1 ) @xmath28 , 2 ) @xmath29 , 3 ) @xmath30 4 ) @xmath31 5 ) @xmath32 .
, . [ bg_circuit ] , @xmath73 @xmath74 @xmath75 @xmath104 @xmath77 .
@xmath78 , @xmath79 , @xmath1 .,width=340 ] @xmath78 @xmath105 [ bg_circuit ] ( + ) ( ) , ( 0 ) , @xmath81 , @xmath82 . ( ) , - @xmath83 @xmath84 . @xmath85 `` '' ( ) .
@xmath86 common point com .
, @xmath87 , @xmath88 .
, `` '' , . 1% ( ) u @xmath1 @xmath89 . , @xmath91 .
@xmath92 .
: ( 1 ) ( 2 ) , 1% . @xmath25 . ?
. ?
_ , `` '' .
`` on '' `` off '' . _ - ( ) `` ''
@xmath94 , . `` '' . , . ?
. [ bg_circuit ] `` '' , dt-830b 20 k@xmath26 , ?
, , ` [email protected] ` 07:00 1 2015 . 137 $ .
, , _ google _ . , ,
.
, * +
2- 3 : 5 ) @xmath25 , .. 4- , 6 ) ( ) @xmath106 , .. 3- , 7 ) @xmath107 .
7 [ bg_template_ext ] .
+ [ @xmath26 ] & @xmath1 [ @xmath55 a ] & @xmath0 [ v ] & @xmath109[k@xmath26 ] & @xmath110[@xmath26 ] & @xmath111[@xmath26 ] + 0 & 0 & & & & & + 1 & & & & & & + 2 & & & & & & + 3 & & & & & & + 4 & & & & & & + 5 & & & & & & + + [ bg_template_ext ] 2 .
* . ?
* + 2- @xmath112 @xmath113 , .. 5- 7- @xmath114 . 3 . *
`` '' @xmath115 ( ) @xmath29 ( ) .
* @xmath116 .
* + @xmath116 `` '' @xmath20 .
, ?
+ `` '' `` '' .
* @xmath117 `` '' . . * + : @xmath29 , . @xmath118 .
. . , @xmath29 @xmath1 . 6
. * @xmath29 . * + `` '' @xmath116 @xmath29 @xmath119 .
. @xmath120 .
, ?
?
( ) . ? 7 .
`` '' , . `` '' . . [ bg_lednr ] , . . . @xmath9 , `` '' . . @xmath3 `` '' . . @xmath1 `` ''
@xmath59 , . `` '' @xmath121 @xmath122 @xmath123 . . ma @xmath9 , , ?
@xmath3
? , @xmath124 . 8
. * . @xmath3 . * + . [ bg_lcnr ] : `` '' , @xmath64 , - ( ) , , , - . + , , . @xmath125 .
@xmath125 @xmath3 ?
. * `` '' * 10 .
* *
. : , . . . [ @xmath26 ] & @xmath1 [ @xmath55 a ] & @xmath0 [ v ] & @xmath109[k@xmath26 ] & @xmath110[@xmath26 ] & @xmath111[@xmath26 ] + 0 & 0 & -1123 & 1.76 & -1.57 & -1455 & -1455 + 1 & 100 & -1209 & 1.89 & -1.56 & -1352 & -1452 + 2 & 180 & -1288 & 2.02 & -1.57 & -1269 & -1449 + 3 & 300 & -1429 & 2.24 & -1.57 & -1144 & -1444 + 4 & 390 & -1554 & 2.43 & -1.56 & -1052 & -1442 + 5 & 510 & -1764 & 2.76 & -1.56 & -926.3 & -1436 + [ bg_experimental_data_black_box_6 ] [ @xmath26 ] & @xmath1 [ @xmath55 a ] & @xmath0 [ mv ] & @xmath109[@xmath26 ] & @xmath110[@xmath26 ] & @xmath111[@xmath26 ] + 1 & 0 & 981 & 1528 & 1558 & 1666 & 1666 + 2 & 100 & 923 & 1439 & 1559 & 1770 & 1670 + 3 & 180 & 882 & 1375 & 1559 & 1853 & 1673 + 4 & 300 & 827 & 1288 & 1557 & 1976 & 1676 + 5 & 390 & 790 & 1231 & 1558 & 2068 & 1678 + 6 & 510 & 744 & 1160 & 1559 & 2196 & 1686 + [ bg_experimental_data_white_resistor_6 ]
. [ bg_resistance - measurement ] @xmath27 `` '' .[bg_experimental_data_black_box_6 ] , .[bg_experimental_data_white_resistor_6 ] .
@xmath126 , @xmath127 .
. , @xmath128 . , width=566 ] . [ bg_resistance - measurement ] @xmath27 . [ bg_i - r_plot ] `` '' .[bg_experimental_data_black_box_6 ] , .[bg_experimental_data_white_resistor_6 ] .
@xmath129 , @xmath130 .
@xmath131 , @xmath132 , . , , . @xmath133 `` '' . , width=566 ] 6 `` '' . @xmath1 , @xmath134 . [ bg_resistance - measurement ] , a [ bg_experimental_data_black_box_6 ] [ bg_experimental_data_white_resistor_6 ] . ( ) @xmath135 , `` '' ( ) @xmath136 , @xmath137 . @xmath40 , . . `` '' . . ; `` '' . , width=566 ] . @xmath138 .
, @xmath139 , `` '' @xmath140 @xmath20 . ,
width=566 ] `` '' ( ) [ bg_experimental_results_black_box ] , ( ) [ bg_experimental_results_white_resistor ] ( ) [ bg_experimental_results_red_led ] .
. [ bg_iv6 ] @xmath1 ( ) @xmath0 ( ) .
. [ bg_resistance - measurement - i - v - curve ] .
- @xmath141 @xmath142 .
`` '' @xmath67- @xmath143 ; @xmath144 . - @xmath145 @xmath146 .
@xmath147,width=566 ] @xmath148 @xmath149 .
@xmath150 @xmath151 `` '' . . @xmath152 . [ bg_lednr ] @xmath153 [ bg_ivdetail ] 1 & -8.45 & 0.18 + 2 & -8.01 & 0.42 + 3 & -7.64 & 0.63 + 4 & -6.96 & 0.99 + 5 & -6.53 & 1.23 + 6 & -5.74 & 1.65 + 7 & -5.30 & 1.90 + 8 & -4.68 & 2.22 + 9 & -4.26 & 2.43 + 10 & -4.01 & 2.53 + 11 & -3.83 & 2.46 + 12 & -3.43 & 2.20 + 13 & -2.07 & 1.33 + 14 & -1.40 & 0.90 + 15 & -0.68 & 0.44 + 16 & -0.30 & 0.20 + 17 & 0.14 & -0.08 + 18 & 0.43 & -0.31 + 19 & 1.31 & -0.84 + 20 & 1.86 & -1.18 + 21 & 2.71 & -1.73 + 22 & 3.58 & -2.29 + 23 & 4.29 & -2.73 + 24 & 4.36 & -2.55 + 25 & 5.10 & -2.22 + 26 & 5.60 & -1.98 + 27 & 6.13 & -1.70 + 28 & 6.71 & -1.39 + 29 & 7.13 & -1.17 + 30 & 7.56 & -0.94 + 31 & 7.98 & -0.72 + 32 & 8.58 & -0.40 + [ bg_experimental_results_black_box ] 1 & -6.77 & -4.35 + 2 & -6.05 & -3.88 + 3 & -4.66 & -2.99 + 4 & -2.76 & -1.77 + 5 & -1.38 & -0.88 + 6 & -0.47 & -0.30 + 7 & 0.65 & 0.42 + 8 & 1.82 & 1.17 + 9 & 3.42 & 2.20 + 10 & 6.78 & 4.35 + [ bg_experimental_results_white_resistor ] @xmath154\qquad$ ] + 1 & -2.16 & -16.83 @xmath155 + 2 & -2.12 & -14.16 @xmath155 + 3 & -2.09 & -12.31 @xmath155 + 4 & -2.08 & -10.20 @xmath155 + 5 & -2.02 & -7.90 @xmath155 + 6 & -1.99 & -5.80 @xmath155 + 7 & -1.94 & -3.95 @xmath155 + 8 & -1.88 & -1.94 @xmath155 + 9 & -1.63 & -0.01 @xmath155 + 10 & -1.678 & -25.8 @xmath156 + 11 & -1.620 & -9.6 @xmath156 + 12 & -1.492 & -2.2 @xmath156 + 13 & -0.914 & -0.9 @xmath156 + 14 & 0.020 & 0.1 @xmath156 + 15 & 0.132 & 0.2 @xmath156 + 16 & 0.589 & 0.6 @xmath156 + 17 & 1.105 & 1.1 @xmath156 + 18 & 1.743 & 1.7 @xmath156 + 19 & 2.52 & 2.4 @xmath156 + 20 & 3.72 & 3.6 @xmath156 + 21 & 4.24 & 4.1 @xmath156 + 22 & 5.25 & 5.1 @xmath156 + 23 & 6.37 & 6.1 @xmath156 + 24 & 8.03 & 7.7 @xmath156 + 25 & 8.34 & 8.8 @xmath156
+ [ bg_experimental_results_red_led ] , @xmath1 `` ''.,width=377 ] @xmath64 .[bg_lcnr ] .
@xmath157 . .,width=377 ] o @xmath157 @xmath64 coach-5 .
,
; . . . : coach-5 nova ?,width=377 ] @xmath158 . , .,width=377 ] . , width=377 ] `` . '' , . i - v . , i - v . @xmath140 - , , - , . , . , width=377 ] @xmath159 ( . [ bg_circuit ] ) @xmath160 , ( + ) @xmath161 @xmath162 1% , ( ) ( + ) @xmath163 @xmath164
@xmath165 . ( ) `` '' @xmath166 . @xmath167 @xmath74 .
, ( 0 ) @xmath168 @xmath169 : @xmath79 @xmath167
@xmath170 @xmath171 @xmath1 @xmath75 @xmath0 @xmath78 @xmath172 .
@xmath173
@xmath78 @xmath174\frac{u}{r_3 } = -\frac{r_2}{r_1 r_3}u = \frac{u}{r},\ ] ] @xmath175 @xmath116 .
@xmath140 .
@xmath176 @xmath177 @xmath178 , , . , @xmath179 - @xmath180 @xmath0 - @xmath181 @xmath182 @xmath183 e @xmath75 .
@xmath184 @xmath185 @xmath95 ; . @xmath186 @xmath187 . @xmath188 @xmath189 @xmath190 .
@xmath191 @xmath192 q- @xmath193 @xmath194 , @xmath195 . @xmath196 , @xmath197 . @xmath198 @xmath199 . , . e e @xmath65 @xmath200 , @xmath201 ; @xmath202 @xmath203 @xmath204 , . @xmath66 @xmath205 . . . k , @xmath206 ( [ bg_criterion_oscillations ] ) .
- @xmath207 , w @xmath208 @xmath209 .
- @xmath190 @xmath210 e . @xmath211 @xmath212 , . . @xmath213 . , `` '' @xmath214 k@xmath26 .
@xcite , _
google _ _ negative resistance _
_ negative impedance converter(nic)_. . . , @xcite ( 1978 . ) @xcite ( 1953 . ) @xcite ( 1968 . ) .
, . , . 9 https://en.wikipedia.org/wiki/negative_impedance_converter ; + https://en.m.wikibooks.org/wiki/circuit_idea/negative_resistance ; + https://en.wikipedia.org/wiki/negative_resistance
. + j.c .
linvill , `` transistor negative - impedance converters '' , proceedings of the ire 41 ( 6 ) : 725729 ( 1953 ) , doi:10.1109/jrproc.1953.274251 g.j .
deboo , `` gyrator type circuit '' , us patent 3493901 , filed 5 march 1968 , issued 1970 - 02 - 03 , assigned to nasa .
pippard , _ the physics of vibration . the simple classical vibrator _ ( cambridge university press . , 1978 ) sec .
11 , fig . 11.14 . https://en.wikipedia.org/wiki/operational_calculus ; + o heaviside , proc .
( london ) * 52 * , 504 - 529 ( 1893 ) , * 54 * , 105 - 143 ( 1894 ) ; + j.r .
carson , _ electric circuit theory and the operational calculus _ ( mcgraw - hill , new york , 1926 ) .
data sheet ada4817 - 1/ada4817/2 , low noise , 1 ghz fastfet op amps , eq . ( 7 ) and eq .
( 4 ) , + http://www.analog.com/media/en/technical-documentation/data-sheets/ada4817-1_4817-2.pdf . k. g. mcconnel and w. r. riley in _
handbook on experimental mechanics _
, ed . by a.
s. kobayashi ( prentice - hall , new jersey , 1987 ) , eqn .
( 3.10 ) and eqn . ( 3.12 ) .
ecircuit center , op amp model - level 1 , basic op amp model , current source model , http://www.ecircuitcenter.com/circuits/opmodel1/opmodel1.htm p.e .
allen and d.r .
holberg _ cmos analog circuit design , _ third edition , the oxford series in electrical and computer engineering , ( oxford university press , 2011 ) , chap
. 6 , sec . 6.7 macromodels for op amps , figures 6.7 - 4 , 6.7 - 5 , 6.7 - 6 , example 6.7 - 2 , isbn : 9780199765072 , http://www.aicdesign.org/sec.6.7%281=4=02%29.pdf , + https://global.oup.com / academic / product / cmos - analog - circuit - design-9780199765072?cc = bg&lang = en&#. power line chokes , 100 mh , http://www.farnell.com/datasheets/1465589.pdf tl07xx low - noise jfet - input operational amplifiers , fig .
9 , http://www.ti.com/lit/ds/symlink/tl071.pdf , . , . , @xmath215 - @xmath216 ( [ bg_exp ] ) , @xmath217 e . @xmath218 @xmath219 @xmath220 @xmath221 @xmath222 @xmath223 .
@xmath1 , @xmath224 , .@xcite @xmath225 @xmath226 @xmath227 . @xmath4 a @xmath228 @xmath229 @xmath230 @xmath231,\\ & & \hat{g}^{-1}=\frac{1}{g_0}+\tau_0\mathrm{d}_t , \quad g({\mathrm{d}_t})\equiv\frac{g_0}{1+\tau_a\mathrm{d}_t}. \quad g_0\equiv \frac{r_a}{r_a } , \quad \tau_a\equiv r_a c_a , \quad \tau_0\equiv r_a c_a=\frac{\tau_a}{g_0}\end{aligned}\ ] ] @xmath232 - @xmath233 ( pad ) , , . spice . @xmath234 @xmath235 tlo71@xcite spice @xmath236 @xmath237 @xmath238 @xmath239 ( [ bg_generalopampeq ] ) - @xmath240 . - , @xmath241 @xmath242 @xmath243 , . ( ) ( opamp ) @xmath244 @xmath245 .
( [ bg_masteropampequation ] ) @xmath246 @xcite @xmath247 @xmath248 , @xmath249 . @xmath250 @xmath251 o @xcite d. ( [ bg_masteropampequation ] ) @xmath252 , , . , , . . `` [ bg_circuit ] , : '' @xmath253 e @xmath254 @xmath255 = \frac{u}{z},\ ] ] @xmath256 . ( [ bg_generalopampeq ] ) @xmath257 @xmath258 @xmath259 @xmath260 _
( negative - impedance converter , nic ) _ @xmath261 @xmath206 @xmath262 @xmath134 @xmath263 , @xmath264 @xmath250 , @xmath265 @xmath266 , . @xmath267 , @xmath268 @xmath269 .
`` '' . `` '' , @xmath270 ,
@xmath271 . _ stability , negative impedance converter _ , , . , ( + ) ( - ) . . @xmath64 @xmath95 . @xmath272 @xmath273-. . @xmath70 @xmath69 @xmath274 @xmath189 @xmath275 a @xmath276 @xmath277 @xmath278 , @xmath279 , @xmath280 @xmath281 : @xmath282 @xmath187 .
@xmath283 @xmath284 @xmath95 @xmath285 ( [ bg_instab ] ) .
@xmath286 @xmath287 . @xmath288 @xmath289 @xmath84 , @xmath74 , @xmath75 , ( _ printable circuit board , pcb _ ) . pcb .
( led ) . @xmath66 @xmath67 @xmath290 : @xmath291 @xmath292 , @xmath293 @xmath294 .
@xmath295 , @xmath296 o @xmath297 @xmath298 @xmath299 @xmath300 @xmath301 @xmath302 q- @xmath285 1 . q- @xmath303 , 1 . q- @xmath304 , `` '' .
, . : @xmath305
@xmath306 ( [ bg_stabilitycriterion ] ) .
. [ bg_circuit ] , ( + ) ( - ) , . @xmath307 .
, ( + ) ( - ) , , , . @xmath308
. @xmath309 @xmath310 @xmath311 @xmath189 @xmath190 @xmath312 @xmath313 @xmath314 @xmath315 @xmath316 @xmath317 @xmath318 , @xmath319 @xmath320 @xmath321 @xmath322 @xmath323 @xmath324 @xmath325 @xmath326 @xmath327 : , @xmath328 . @xmath329 @xmath330 k ( [ bg_reff ] ) -e @xmath312 @xmath331 @xmath332 @xmath333 @xmath334 . @xmath197 , @xmath335 , @xmath336 . @xmath337 . @xmath338 `` '' `` . '' @xmath9 @xmath3 @xmath339 @xmath340 @xmath341 ( [ bg_var_epsilon - delta ] ) - . @xmath190 @xmath190 . @xmath342 2 .
( ) . [ mk_setup ] , @xmath38 ( ) `` '' , . , 5 , , 1 50 cm , - , , 4 ( ) , 9 v , ( @xmath64 ) , . .,width=332 ] [ mk_quatity_tasks ] , `` '' . [ mk_black_box_investigation ] - ( ) `` '' .
[ mk_the_world_is_not_perfect ] . [ mk_dynamics ] `` '' . , [ mk_theoreticalproblem ] . 3
, . [ mk_homework ] c 137 $ ; 7:00 .
, . . , , .
. [ mk_setup ] `` '' 2 ( ) .
- `` '' `` '' .
. . led - . `` '' @xmath64 @xmath343 @xmath69 @xmath95 .
. . , . [ mk_lcnr ] , . led - . . `` '' .
led , - .,width=332 ] , , ( ).,width=332 ]
. ?
_ [ mk_black_box_investigation ] , ( ) .
( 18 ) . , . _ 1 . * `` '' . ( 7 ) * + . [ mk_hypothesis_reject ] @xmath0 `` '' @xmath1 , . .
. [ mk_template_4_setups ] , a . + @xmath2 @xmath3 ?
+ . @xmath2 @xmath4 .
( b ) @xmath5 @xmath6 .
( c ) ja @xmath7 .
( d ) @xmath8 .
( e ) @xmath3 @xmath9 . [ mk_template_4_setups ] .
, width=612 ] + @xmath0 [ @xmath10 ] + a ) & @xmath11 & @xmath12 + b ) & @xmath13 & @xmath14 + c ) & @xmath15 & + d ) & & @xmath16 + e ) & @xmath17 & @xmath18 + + [ mk_template_4_setups ] 2 . * @xmath191.5 v. ( 1 ) * + .
. , com ( ) , v , , , . . 3
. * . ( 1 ) * + - @xmath20 , ( ) .
1 @xmath21 @xmath20= ?
* . ( 3 ) * + -. . @xmath344 , @xmath345 , @xmath346 . @xmath22 .
1 @xmath21 5 .
* @xmath191.5 v rwr . ( 7 ) * + . [ mk_resistance - measurement ]
. [ mk_r_neg_r_with_crocodiles ] . . !
. @xmath23 .
. . + , a ( ) @xmath1 .
@xmath24 . . @xmath25 e . @xmath20 `` '' . , width=434 ] + , `` ''.,width=566 ] + 1.5 v. 0 @xmath26 5- @xmath27 5 6 [ mk_template ] : 1 ) @xmath28 , 2 ) @xmath29 , 3 ) @xmath30 , 4 ) @xmath31 5 ) @xmath32 .
. [ mk_circuit ] , ( + ) ( ) , ( 0 ) @xmath81 , @xmath90 , @xmath82 .
( ) @xmath83 @xmath84 . @xmath85 `` '' ( ) .
@xmath86 common point u com .
e , @xmath87 , n @xmath88 .
, `` '' , . 1% ( ) @xmath0 @xmath1 @xmath89 . , @xmath91 .
@xmath92 .
: ( 1 ) ( 2 ) , 1% .
@xmath25 u . led ?
@xmath73 @xmath74 @xmath75 @xmath104 @xmath77 .
@xmath78 , @xmath79 , @xmath1 c . @xmath78 @xmath105,width=340 ] o a . ?
_ , n `` '' . n on on ff . _ - ( ) `` '' @xmath94 . `` ''
. . ?
. [ mk_circuit ] `` '' , dt-830b od 20 k@xmath26
?
ro , ` [email protected] ` 07:00 1 2015 137 $ .
, , _ google _ . , ,
.
pomou datog kompleta eksperimentalnih ureaja ( set ureaja ) prikazanog na slici [ sr_setup ] .
ispitajte zavisnost izmeu struje i napona @xmath38 ili kako se jo naziva voltamperska karakteristika ( vak ) `` crne kutije '' , koja se ne sme otvoriti za vreme olimpijade . , druge otpornike , ute etikete , jedan kondenzator sa piezo - ploicom koji igra ulogu minijaturnog zvunika , led diodu povezanu na dugake ice , plastini lenjir , etiri prikjuna kabla za multimetre ( crveni i crni ) , potenciometar povezan na kontakte za baterije od 9 v , paralelno povezane kondenzator i kalem ( predstavljajui rerzonantni @xmath64 krug ) i najvaniji deo
jednu crnu kutiju.,width=332 ] slede dva kvalitativna zadadtka opisanih u delu [ sr_quatity_tasks ] , iji je cilj da proverite dali ureaj sakriven u `` crnoj kutiji '' radi .
u poglavlju [ sr_black_box_investigation ] su detaljno opisani eksperimenti koje trebate izvriti datim kompletom da bi istraili volt - ampersku karakteristiku ( vak ) `` crne kutije '' .
dalje , u poglavlju [ sr_the_world_is_not_perfect ] uradiete detaljnije istraivanje i uvideti da su fizika svojstva ovog objekta koja ste do sad otkrili validna u odreenim granicama .
nakon statine analize sledi poglavlje [ sr_dynamics ] u kom ete mehanikim opitima istaivati dinamika svojstva `` crne kutije '' .
poglavlje [ sr_theoreticalproblem ] i m a i isto teoretski deo , povezan teoretskim opisom predloenih eksperimenata .
uenicima koji nisu mnogo sigurni u eksperiment , ali se bolje nose sa matematikom predlaemo da se skoncentriu na teoretski deo zadatka .
za neumorne je domai zadatak opisan u poglavlju [ sr_homework ] , uz novanu premiju od 137 $ krajnji rok za predavanje ( 01.11.2015 ) sutra u 7:00 .
nakon zavretka olimpijade komplet dobijate na poklon kabinetu fizike .
molimo vas da ga predate vaem nastavniku da bi pokazali eksperiment acima iz vae kole .
organizatorima jedino trebate vratiti primljeni multimetar .
elimo vam dobro raspoloenje , interesantno istaivanje , zabavu i uspeh .
u kompletu prikazanom na sl .
[ sr_setup ] imate `` crnu kutiju '' i led diodu povezanu na krajevima duim kablovima sa tipaljkama .
poveite led diodu na `` crnu kutiju '' i ako ne zasvetli premenite mesta tipaljkama .
u jednoj kombinaciji za pvezivanje dioda zasveti .
pri daljnjih kvantitativnih zadataka potrebno je mjerenje struje i napona na led ( light - emitting diode ) i objasniti kako su ove varijable povezane s karakteristikama struje napona od elemenata .
paralelno na crnu kutiju i led diodu prikjuite @xmath64 konturu sastavljenu od paralelno povezanog kondenzatora kapaciteta @xmath70 i kalema induktivnosti @xmath69 i unutranjeg otpora @xmath95 . kalem je namotan na feritno jezgro .
pomerite led diodu i vedeete da svetlost pulsira .
ako paralelno ukljuite
i piezo - ploicu , kako je prikazano na slici [ sr_lcnr ] , uete i slabo zujanje .
proverite dali led dioda treperi i piezo - ploica zuji .
naredni zadaci se odnose na detaljno kvantitativno istrivanje i teoretsko objanjenje svetlosti i zujanja .
rezonantnom oscilatzornom krugu pomou `` crne kutije '' . ako pomerimo led diodu u ovom sluaju vidimo da svetlost pulsira , a ujemo zujanje piezo - pliice izazvano naizmeninim naponom.,width=332 ] , pomou prikljunih kablova koji zavravaju tipaljkama.,width=332 ] kod kog od ova dva kvalitativna zadatka dioda svetli jae
u sluaju kada je dioda povezana sama na `` crnu kutiju '' ili kada je prikljuen i rezonantni oscilatzorni krug , t.j .
kod postojane ili pulsirajue svetlosti ?
* merenje napona i struje koja tee kroz `` crnu kutiju '' povezanu sa led diodom i bez nje .
( 7 poena ) * + poveite eme kao na sl .
[ sr_hypothesis_reject ] i izmerite napon u na `` crnoj kutiji '' i struju @xmath1 koja tee kroz nju .
ako led dioda iz prve ili druge eme ne svetli , poveite je obrnuto , t.j .
promenite joj polaritet .
rezultate merenja unesite u tabelu identinu tabeli [ sr_template_4_setups ] , kako je priazana nie .
usled ega se pojavljuje mala razlika izmeu napona @xmath2 i @xmath3 ?
+ dodaju se jo ampermetar i voltmetar . na njima se mogu uitati vrednosti @xmath2 u @xmath4 .
( b ) zamenimo led diodu provodnikom i zabeleimo vrednosti na ureajima @xmath5 i @xmath6 .
( c ) uklonimo voltmetar i uitamo jainu koju pokazuje ampermetar @xmath7 .
( d)zamenite ampermetar voltmetrom i upiite vrednost napona @xmath8 .
( e ) izmerite napon @xmath3 jainu @xmath9 po poslednjoj emi.upiite merenja u tabelu kao u primeru tabela [ sr_template_4_setups].,width=612 ] + @xmath0 [ @xmath10 ] + a ) & @xmath11 & @xmath12 + b ) & @xmath13 & @xmath14 + c ) & @xmath15 & + d ) & & @xmath16 + e ) & @xmath17 & @xmath18 + + [ sr_template_4_setups ] 2 .
* izmerite napon baterije iz kompleta sa ems @xmath191.5 v. ( 1 poen ) * + ukljuite kombinovani ureaj multimer da radi kao voltmetar i izmerite napon baterije na njenim krajevima . instrument moe da registruje i znak napona .
upamtite , ako je crni kabel povezan na ulaz instrumenta oznaenim kao com ili znakom uzemljenja ( ) , a crveni kabel za drugi ulaz za napon @xmath10 , i ako se drugi kraj crnog kabla povee na minus baterije a crveni kabel na plus baterije instrument pokazuje pozitivnu vrednost .
ako se promene mesta kablova pokazae negativnu vrednost .
upiite vrednosti na voltmetru u oba sluaja .
* izmerite otpor velikog belog otpornika .
( 1 poen ) * + ukljuite multimetar kao om - metar
i izmerite i upiite vrednost otpora @xmath20 , ( otpor velikog belog otpornika ) .
radite sa tanou od 1 @xmath21 @xmath20= ?
* izmerite otpor pet malih otpornika .
( 3 poena ) * + nastavite da koristite multimetar kao om-metar.izmerite elektrini otpor svakog od pet malih otpornika u vaem kompletu .
zalepite utu etiketu iz kompleta na svaki od otpornika i upiite @xmath22 i upiite odgovarajuu vrednost .
radite sa tanou od 1 @xmath21 5 . * pomou pet malih otpornika i baterije od @xmath191.5 v izmerite zavisnost izmeu struje
i napona velikog belog otpornika sa otporom @xmath20 .
( 7 poena ) * + elektrina ema za merenje zavisnosti izmeu struje i napona data je na sl .
[ sr_resistance - measurement ] i sl .
[ sr_r_neg_r_with_crocodiles ] .
multimetar upotrebite kao ampermetar , poveite ga serijski sa belim otpornikom.pazite na znakove za povezivanje instrumenta sa izvorom struje
struja i m a svoj smer
! drugi multimetar poveite kao voltmetar paralelno na prouavani beli otpornik i ponovo pazite na znakove za povezivanje voltmetra i polaritet ukljuivanja .
za veliki beli otpornik vai omov zakon @xmath23 .
ako voltmetar pokazuje pozitivnu vrednost i ampermetar treba da pokazuje pozitivnu vrednost , i obrnuto .
ako su znaci napona i struje suprotni proverite gde ste pogreili kod povezivanja instrumenata . + , a ampermetar ( a ) je povezan serijski i meri struju @xmath1 . kada se elektrino kolo zatvara razliitim otporima @xmath24 struja i napon su razliiti .
tako se dobija nekoliko taaka voltamperske karakteristike .
za male napone odnos @xmath25 je postojan i to je jedan mogui nain provere omovog zakona i znakova izmerene struje i napona .
otpornik @xmath20 na emi levo zamenjen
je `` crnom kutijom '' na emi desno i to je jedina razlika izmeu ova dva elektrina kola.,width=434 ] + pomou kablova koji zavravaju tipaljkama.,width=566 ] + serijski poveite na ampermetar bateriju od 1.5 v. zatvorite strujno kolo sa 0 @xmath26 i pojedinano svakim od pet otpornika @xmath27 rezultate svakog pojedinanog merenja upiite u odgovarajua polja tabele sa 5 kolona i 6 redova , kao kod tabele [ sr_template ] : 1 ) broj otpornika @xmath28 , 2 ) vrednost otpornika @xmath29 , 3 ) jainu struje @xmath30 , 4 ) napon voltmetra @xmath31 , 5 ) izraunatu vrednost otpora @xmath32 . +
u elektrinom kolu prikazanom na sl .
[ sr_circuit ] ulazne struje u takama ( + ) i ( - ) su jednake nuli , a izlazna struja u taki ( 0 ) je takva da odgovarajui naponi su povezani relacijom @xmath81 , gde
je g koeficijent pojaavanja i i m a jako veliku vrednost , @xmath82 .
elektronski element oznaen trouglom ( pojaiva ) se napaja iz dve baterije i u uslovu zadatka se pretpostavlja da su pomenuti naponi manji od napona napajanja baterija @xmath83 kola u zavisnosti od tri otpornika @xmath73 @xmath74 i @xmath75 na trouglom oznaenim naponskim pojaivaem sa koeficijentom pojaavanja @xmath104 koji se napaja sa dve baterije sa naponom @xmath77 .
naponi @xmath78 , @xmath79 , kao i struja i su nepoznati .
sa @xmath78 je oznaen napon izveden iz @xmath105,width=340 ] taka izmeu dve baterije
je povezana provodnikom sa krajem otpornika @xmath84 i jednom od ulaznih elektroda na emi .
je da se napon u toj zajednikoj taki izabere za nulu , @xmath85 , ili kako elektrotehniari kau to je uzemljenje ( ) .
indeks @xmath86 dolazi iz engleskog common point i na multimetrima se oznaava kao com .
u suprotnom sluaju , ako je @xmath87 , za jednainu pojaavanja napona imamo @xmath88 .
struja koja istie u `` uzemljenje '' je nula.izraunajte uz tanost od 1% ( tri znaajna broja ) odos izmeu ulaznog napona u i struje i koja tee kroz kolo i izrazite taj efektivni otpor @xmath89 u zavisnosti od tri otpora na emi .
da bi uprostili , moete pretpostaviti da koeficijent pojaavanja stremi ka beskonanosti @xmath91 .
u dobijenom izrazu zamenite @xmath92 .
ukratko : 1 ) trai se krajnja formula za otpor eme i 2 ) izraunati po njoj numeriku vrednost otpora uz tanost od 1% .
kakav je znak otpora @xmath25 i koliko oma i m a njegov modul ?
kako taj otpor pali led diodu ?
crveno i vrsto , a leti .
_ po zavretku olimpijade , naite rafciger
i odvrnite rafove sa poklopca `` crne kutije '' .
izvadite baterije ili promenite poloaj jednog od prekidaa sa `` on '' na `` off ''
. _ za male napone voltamperska karakteristika ( vak ) `` crne kutije '' je prava linija sa postojanim odnosom @xmath94 u skladu sa omovim zakonom .
pokuajte da izmerite otpor `` crne kutije '' ommetrom .
pokuajte da uporedite vrednosti na ommetru sa otporom dobijenim kroz ispitivanje vak - a .
objasnite zato se otpor `` crne kutije '' definitivno ne moe izmeriti ommetrom .
kakve promene iz eme na sl .
[ sr_circuit ] treba realizovati u `` crnoj kutiji '' da bi bilo mogue merenje otpora ommetrom , na primer multimetrom dt-830b od 20 k@xmath26 kakvim raspolaete na olimpijadi ?
prvi koji odgovori na bar jedno od dva pitanja i poalje odgovor sa adrese
kojom se registrovao za olimpijadu na ` [email protected]`s do 07:00 1.novembra 2015 osvojie novanu nagradu od 137 $ .
moete raditi u timu , koristiti literaturu , _ google _ , i da se na internetu konsultujete sa radioinenjerima i univerzitetskim profesorima elektronike svuda po svetu . kada
je pono u kumanovu , u kaliforniji je kasno popodne , u japanu poinje dan svuda na svetu postoje kolege koje rade . | this problem was given at third experimental physics olympiad `` the day of the resistor '' , 31 october 2015 , kumanovo , organized by the regional society of physicists of strumica , macedonia and the sofia branch of the union of physicists in bulgaria . |
une immersion isomtrique dhypersurface @xmath15 de classe @xmath16 est localement rigide en dehors du lieu des points de courbure nulle et on sattend ce que les proprits mtriques de @xmath10 dterminent de nombreuses proprits de @xmath17 .
il y a pourtant peu de rsultats quantitatifs qui estiment des caractristiques extrinsques en termes dobjets intrinsques et qui soient valables en toute dimension ( voir @xcite pour plus de rfrences ) .
le rsultat qui est la source du prsent travail est une ingalit de burago @xcite 6.2 _ en dimension @xmath4_. si @xmath18 dsigne la courbure totale de limmersion , @xmath19 est le rayon dune boule contenant @xmath20 et @xmath21 dsigne laire de @xmath10 alors @xmath22 avec galit uniquement pour le plongement isomtrique standard de la sphre de rayon @xmath19 .
nous nous sommes poss la question sil y a un analogue de ( [ burago depart ] ) en dimension suprieure .
la rponse est _ ngative _ pour ce qui est dune ingalit globale du type @xmath23 nanmoins , nous obtenons des ingalits caractre isoprimtrique _ semi - local _ faisant intervenir la courbure totale dun ouvert sur lequel lapplication de gauss est non - dgnre et exhibant le lieu des points de courbure nulle comme une obstruction lexistence dingalits globales .
larticle est structur comme suit .
nous prsentons dans la section [ le contre - exemple ] un exemple explicite qui prouve la non - existence danalogues de ( [ burago depart ] ) en dimension suprieure .
la section [ ineg globales ] dveloppe des formules intgrales concernant les polynmes symtriques des courbures principales et la fonction support de limmersion .
elles gnralisent celles dj obtenues par minkowski et kubota dans le cas convexe @xcite , p.64 et par hsiung @xcite dans le cas non - convexe .
elles sont utilises dune part pour donner des conditions intrinsques suffisantes pour lingalit globale en dimension @xmath24 et dautre part pour raffiner par la suite lingalit semi - locale .
nous obtenons notamment la [ ricci ] soit @xmath25 une immersion isomtrique de classe @xmath16 avec @xmath10 une varit riemannienne lisse compacte sans bord . on suppose que @xmath26 .
si @xmath10 vrifie @xmath27 avec @xmath28 alors @xmath29 il serait bien - sur intressant de trouver des contre - exemples lingalit ( [ desir ] ) satisfaisant @xmath30 .
dans la section [ ineg locales ] nous obtenons des ingalits caractre semi - local . on utilise une mthode destimation a - priori pour les quations de monge - ampre relles dveloppe par rauch et taylor @xcite et reprise par aubin @xcite .
loprateur de type monge - ampre qui apparait naturellement dans notre contexte est @xmath31 agissant sur les fonctions dfinies sur @xmath14 .
ceci suggre une relation - qui nous reste pour linstant cache - avec la thorie spectrale du laplacien sur @xmath14 .
notre rsultat principal est le on suppose que @xmath32 .
soit @xmath33 un ouvert sur lequel lapplication de gauss est non - dgnre .
il existe une constante @xmath34 qui dpend uniquement de la dimension telle que , si @xmath35 est un compact , on ait @xmath36 et @xmath37 o @xmath38 dsigne la courbure totale de @xmath11 et @xmath39 est la distance sphrique locale entre @xmath9 et @xmath40 ( voir la dfinition [ la distance locale ] ) . le terme @xmath39 tend vers zro lorsque @xmath11 approche @xmath9 .
cela fait que lestimation ( [ the best ] ) est dautant meilleure que le compact @xmath9 est situ une grande distance du lieu des points de courbure nulle .
cest dans ce sens quon interprte le lieu des points de courbure nulle comme une obstruction lingalit globale .
lobtention de ( [ the best ] ) partir de ( [ the semi best ] ) laide des formules intgrales globales de type minkowski occupe la section [ local to global ] .
loptimalit de lingalit ( [ the best ] ) est discute dans la section [ loptimalite ] et larticle clt sur une ingalit isoprimtrique inverse traite dans lappendice . _ remerciements .
_ ce travail constitue la premire partie de ma thse .
il a t prpar sous la direction bienveillante et inspire de claude viterbo , qui saura trouver ici lexpression de ma gratitude .
jai profit de remarques , suggestions et patience dcoute de la part de jrme bertrand , charles boubel , emmanuel ferrand , paul gauduchon , taoufik hmidi , jean lannes , christophe margerin , nicolae mihalache , liviu ornea , ferit oztrk , pierre pansu , jean - marc schlenker et costin vlcu .
je leur suis reconnaissant . | le thme de larticle est fourni par la conjecture empirique suivante : une hypersurface ferme @xmath0 de diamtre extrieur born et de `` grand '' volume devrait ncessairement tre `` trs '' courbe .
on prend comme mesure pour la courbure le volume recouvert sur la sphre unit @xmath1 par lapplication de gauss , savoir la courbure totale @xmath2 .
notre tude est motive par la preuve de burago @xcite de lingalit @xmath3 pour @xmath4 et @xmath5 .
nous donnons un exemple explicite qui prouve quune ingalit de la forme @xmath6 ne peut tre vraie en toute gnralit en dimension @xmath7 .
nous mettons toutefois en vidence une condition suffisante sur la courbure de ricci sour laquelle lingalit est vrifie en dimension @xmath8 .
le rsultat principal de larticle est une ingalit caractre semi - local qui majore le volume dun compact @xmath9 de @xmath10 par la courbure totale dun ouvert @xmath11 qui le contient , sous lhypothse que la courbure de gauss - kronecker ne sannule pas sur @xmath9 et @xmath11 .
pour @xmath5 on prouve que @xmath12 avec @xmath13 une distance mesure _ sur la sphre @xmath14 _ via lapplication de gauss .
le lieu des points de courbure nulle apparait de cette faon comme une obstruction avoir une ingalit globale comme celle de burago en dimension suprieure .
nous obtenons diffrentes variantes de lingalit ayant un caractre isoprimtrique et nous montrons quelles sont optimales .
au passage , nous obtenons une ingalit isoprimtrique `` inverse '' valable dans les espaces courbure constante . |
[ [ generalities ] ] generalities + + + + + + + + + + + + the present contribution is devoted to the mathematical analysis of the equation @xmath0 equation arises in the modeling of some non - newtonian fluid flows .
some details on the modeling will be given below .
the variable @xmath1 is one - dimensional , varies on the real line @xmath2 , and models a quantity homogeneous to a stress ( actually to a certain entry of the stress tensor ) .
the variable @xmath3 of course denotes the time , and equation is supplied with some initial condition @xmath4 .
the unknown real - valued function @xmath5 , solution to , satisfies the two properties : it is nonnegative @xmath6 and normalized to one @xmath7 for all times @xmath8 . the function @xmath9 models the density of probability to have a certain ( elementary microscopic ) stress @xmath1 , at time @xmath3 , at a macroscopic space position @xmath10 . the actual , deterministic stress within the fluid is thus given by @xmath11 , of equation below .
equation is thus implicitly parameterized by this position @xmath10 ( thus the multiscale nature of the problem , as will be seen below ) .
the function @xmath12 , also a function of time , is assumed given .
it models the shear rate at the position @xmath10 , under which we wish to compute , using , the mesoscopic response of the fluid .
the notation @xmath12 is traditional in fluid mechanics , hence its use here . in ,
we denote by @xmath13 the characteristic function @xmath14}\ ] ] where @xmath15 a scalar positive parameter , fixed once and for all .
it models some local threshold value of the stress , which plays a crucial role in the modeling .
as is usual , we denote by @xmath16 the dirac mass at zero . from the definitions of @xmath13 and @xmath16 , it is immediately seen
that , at least formally ( and this will indeed be made rigorous , see lemmata [ lem : max ] and [ lem : mass ] in section [ sec : properties ] below ) , equation preserves in time the two properties and .
two quantities are typically computed using the solution @xmath9 to : first the so - called _ fluidity _ @xmath17 and next the ( real - valued ) stress @xmath18 our purpose in this article is to mathematically study equation ( in terms of existence and uniqueness of the solution @xmath9 , properties and long - time behavior of that solution ) and to derive a macroscopic equation equivalent to this equation . by _
macroscopic _ equation , we mean an equation ( actually a differential equation , or a system of differential equations ) that directly relates the shear rate @xmath12 , the fluidity @xmath19 and the stress @xmath20 without the explicit need to compute @xmath9 . in these macroscopic equations
, the scalar @xmath21 will be the inverse of the mechanical relaxation time , thus its name `` fluidity '' .
we will be able , in particular , to obtain a macroscopic equation which is close to models that have been proposed for aging fluids @xcite , see the discussion at the end of section [ sec : passage ] .
[ [ some - elements - on - the - modeling ] ] some elements on the modeling + + + + + + + + + + + + + + + + + + + + + + + + + + + + + equation is the simplest possible form of an equation describing the mesoscopic behavior of a complex fluid , such as a concentrated suspension , or more generically a soft amorphous material , with properties intermediate between those of a fluid and those of a solid .
these materials exhibit a highly non - newtonian behavior and may give rise to a macroscopic yield stress . at low stress , such a material behaves in an elastic way .
but above a certain stress threshold , here denoted by the critical value @xmath22 , one observes a relaxation toward a completely relaxed state .
this behavior is modeled by equation .
the probability of finding the fluid in the state of stress @xmath1 at time @xmath3 evolves in time for two different reasons : the term @xmath23 models the modification of the stress induced by the existence of the shear rate , while the term @xmath24 encodes the relaxation toward zero of the part of the stress above the threshold @xmath22 . from a probabilistic viewpoint ,
the stochastic process @xmath25 associated to the fokker - planck equation evolves deterministically when @xmath26 and jumps to zero with an exponential rate @xmath27 when @xmath28 .
the process @xmath25 belongs to the class of _ piecewise - deterministic markov processes _ , which have been introduced in the probabilistic literature in the 1980 s for biological modeling for example . in particular
, coupling arguments have been proposed to study the longtime behavior of such processes ( see @xcite ) .
we argue on the fokker - planck equation and proceed differently .
the argument we are using here to study the longtime behavior is purely deterministic in nature , and is based on a delay equation related to the fokker - planck equation .
we would like to mention that for more realistic models , a third phenomenon is typically at play , in addition to the stress induced by the ambient fluid , and to the relaxation to zero .
all states of stress are not independent of one another , and they may also depend on the state of stress at neighboring points within the fluid .
a certain redistribution of the stress therefore always occurs .
this redistribution can be encoded in various ways , depending on some more detailed elements of modeling . in the so - called hbraud - lequeux model introduced in the seminal article @xcite ( and
then extensively studied mathematically in the works @xcite ) , the redistribution is performed by a diffusion term in the stress space , at the given location @xmath10 in the ambient physical space , and the complete equation thus writes @xmath29 where @xmath30 is some parameter . in an alternative model introduced by bocquet and coll . in @xcite
, the redistribution is achieved by some type of local `` convolution '' in the physical space .
the equation ( we recall , set at the physical location @xmath10 ) writes @xmath31 with a function @xmath32 related to the green function of some local stokes - type problem .
the equation which we study in the present article ignores the redistribution phenomenon , which amounts to taking @xmath33 in or @xmath34 in . in the absence of such a simplification , we are unable to proceed with the main result of this article , which is the derivation of the macroscopic equation from our multiscale model .
the well posedness result contained in our theorem [ th : exu ] , on the other hand , also holds for whith @xmath35 and has indeed been established some years ago in @xcite . some more detailed comments on the modeling , as well as some formal foundations of the model based on a system of interacting particles
are presented in @xcite .
[ [ plan - of - our - contribution ] ] plan of our contribution + + + + + + + + + + + + + + + + + + + + + + + + our article is organized as follows . to start with
, we study in section [ sec : ss ] the stationary solutions to .
we next show in section [ sec : exist - uniq ] existence and uniqueness of the solutions to the time - dependent equation . our result is stated in theorem [ th : exu ] .
section [ sec : properties ] follows , establishing some useful properties of the solution . in order to understand the macroscopic equivalent of equation for a given shear rate @xmath36 , which we assume varies slowly as compared to the characteristic time of equation , we need to understand the long - time behavior of the solution to .
we therefore study this behavior in sections [ sec : tlc ] and [ sec : long - time ] , respectively in the case of a constant shear rate @xmath37 , and in the case of a slowly varying shear rate @xmath38 .
the results are stated in theorems [ th : tlc ] and [ th : tle ] .
we are then in position to derive , in section [ sec : passage ] , the macroscopic differential equations equivalent to in this limit , namely system .
our final section , section [ sec : numerics ] , presents some numerical experiments which confirm and illustrate our theoretical results .
we study in this section the stationary states of .
we therefore assume that @xmath39 is a fixed scalar and consider the solutions @xmath40 to the following equation @xmath41 here and in the following , for a subset @xmath42 , @xmath43 denotes the set of distributions on @xmath44 . by convention ,
since the time - dependent version of the equation is linear and formally preserves positiveness and the integral over the real line , we are only interested in the stationary solutions @xmath45 that additionally satisfy and , that is , @xmath46 we have the following result : [ lem : ss ] when @xmath47 , the solutions to - are exactly all nonnegative normalized densities with compact support in @xmath48 $ ] .
when @xmath49 , there exists a unique solution @xmath45 to - . in the case
@xmath47 , the equation implies @xmath50 in @xmath51)$ ] , hence the result .
up to a change of @xmath45 into @xmath52 , we may , without loss of generality , consider only the case @xmath53 for our proof .
we first note that @xmath45 defined by @xmath54 is a solution to , hence the existence result .
we now show uniqueness . by linearity
, we assume that @xmath55 is a solution of with @xmath56 and show that @xmath50 .
equation implies @xmath57 in @xmath51)$ ] . because @xmath55 , this leads to @xmath58 a.e . on @xmath59 with @xmath60 a scalar and @xmath50 a.e . on @xmath61 . in the case
@xmath33 , this implies @xmath62 .
consequently , rewrites @xmath63 on the whole real line .
this readily implies @xmath50 since @xmath64 . in the case @xmath65
, we obtain that writes @xmath66 so that @xmath67 a.e . on @xmath68 . using that @xmath58 a.e . on @xmath59 and @xmath69
, we find @xmath33 and thus @xmath50 .
this concludes the proof .
this section is devoted to the proof of the following result : [ th : exu ] consider @xmath12 a function of time that satisfies @xmath70 denote by @xmath71 .
consider @xmath72 .
then , for any @xmath73 , there exists a unique @xmath9 in @xmath74 such that @xmath75 for almost all @xmath76 and such that holds for @xmath9 in the sense of distributions on @xmath77 .
in addition , introduce @xmath78 and @xmath79 defined by induction on @xmath80 as follows @xmath81 both @xmath82 and @xmath79 belong to @xmath83 .
then , the solution @xmath9 to is explicitly given by @xmath84 and @xmath19 defined by is equal to @xmath79 : @xmath85 the above results also hold for @xmath86 negative .
however , it is unclear how to extend these results if @xmath12 is allowed to vanish .
we will see in the next section that if the initial condition @xmath87 is nonnegative and normalized , then this property is preserved in time for the solution @xmath9 .
we first note that @xmath88 is a strictly increasing continuous function of the time because of . throughout this proof , we assume @xmath89 . when @xmath90 , the arguments are similar and actually simpler .
we first show uniqueness , then @xmath9 given by belongs to @xmath74 and is a solution of in @xmath91 .
[ [ step - uniqueness ] ] step : uniqueness equation is linear , we therefore consider a solution @xmath92 associated to the zero initial condition @xmath93 and we intend to show that @xmath94 .
denote by @xmath95 we now show @xmath96 with @xmath19 defined from @xmath9 by .
we have , for all @xmath97 , ( where @xmath98 denotes the set of @xmath99 functions with compact support in @xmath77 ) @xmath100 for @xmath101 , denote now by @xmath102 a mollifier , @xmath103 and @xmath104 the fact that @xmath9 is solution to in @xmath91 yields @xmath105 this rewrites @xmath106 as @xmath107 goes to infinity , @xmath108 converges to @xmath13 in @xmath109 and for almost all @xmath110 , @xmath111 because @xmath9 and @xmath112 respectively belong to @xmath113 and @xmath114 , all terms of are bounded from below and from above by an integrable function independent of @xmath107 .
we apply the dominated convergence theorem to pass to the limit in @xmath115 hence , using .
define @xmath116 .
we now show that @xmath117 in @xmath118 , this will prove that @xmath94 in @xmath118 . from
, we deduce that @xmath119 in @xmath120 with @xmath121 using that @xmath93 , we find @xmath122 in @xmath123 and therefore in @xmath124 .
this implies @xmath94 in @xmath125 . in particular ,
since @xmath126 , we have , for all @xmath127 and thus , the equation reads @xmath128 in @xmath129 .
we deduce that @xmath117 and therefore that @xmath94 in @xmath118 . taking @xmath130 as initial time , we find @xmath94 in @xmath131 with the previous arguments . iterating
, we obtain @xmath94 in @xmath132 .
this concludes the proof of uniqueness .
our next two steps are respectively devoted to proving that @xmath9 defined by belongs to @xmath133 and that it satisfies equation .
[ [ s : espace ] ] step : regularity of expression first , the function @xmath82 defined by belongs to @xmath83 ( with @xmath134 ) and therefore @xmath79 defined by the recurrence relation also belongs to @xmath83 .
this implies , for almost all @xmath135 , @xmath136 that is @xmath9 defined by belongs to @xmath132 .
denote @xmath137 so that @xmath138 .
we now check that @xmath139 and @xmath140 belong to @xmath74 and @xmath141 and @xmath142 belong to @xmath143 . using
that @xmath144 is 1-lipschitz on @xmath145 , this will prove that @xmath146 .
consider @xmath147 and @xmath148 such that @xmath149 . by density of @xmath150 in @xmath151
, there exists @xmath152 such that @xmath153 we obtain @xmath154 with @xmath155 .
moreover , by continuity of @xmath156 and @xmath157 , we have , for all @xmath76 , @xmath158 and @xmath159 on a bounded interval ( depending on the support of @xmath156 and on @xmath160 ) . using lebesgue
dominated convergence theorem , we deduce that there exists @xmath161 such that for all @xmath162 , @xmath163 so that @xmath164 this yields @xmath165 .
we now turn to @xmath141 . for almost all @xmath76
, we have @xmath166 this leads to @xmath167 which yields @xmath168 . for any @xmath169 and @xmath148 such that @xmath149 ,
we have @xmath170 since @xmath79 belongs to @xmath83 , @xmath171 belongs to @xmath172 and we introduce a sequence @xmath173 in @xmath174 such that @xmath175 and obtain @xmath176 using the dominated convergence theorem for the second term , this implies that @xmath177 vanishes with @xmath148 .
we have obtained that @xmath140 belongs to @xmath178 .
the continuity holds also at @xmath179 because for all @xmath180 @xmath181 finally , the function @xmath182 is in @xmath183 ( and so is @xmath142 ) . indeed , for any @xmath135 and @xmath148 such that @xmath149 , for almost all @xmath76 , @xmath184 so that , with the change of variable @xmath185 , @xmath186 this concludes the proof of the continuity in time of @xmath9 , with values in @xmath187 . [ [ step - expression - satisfies ] ] step : expression satisfies we first show that @xmath9 defined by satisfies @xmath188 for all @xmath97 , @xmath189 where we have made the changes of variables @xmath190 and @xmath191 . introducing @xmath192 this rewrites @xmath193}d\xi -\int_0^t \phi(v ) { \left[\int_v^t { \dfrac{\partial\mu}{\partialt}}(t , v)dt\right ] } dv\\ & \quad=\int_{\mathbb{r } } p_0(\xi ) { \left[\nu(0,\xi ) - \nu(t,\xi)\right]}d\xi\\ & \quad + \int_0^t
\phi(v ) { \left[\mu(v , v)-\mu(t , v)\right ] } dv = \int_0^t
\phi(v ) \psi(v,0 ) dv\end{aligned}\ ] ] thus .
we finally show that @xmath194 a.e . on @xmath195 , where @xmath9 is defined by .
this will prove that holds in @xmath196 .
first , for almost all @xmath197 , definition of @xmath9 implies @xmath198 .
the definition of @xmath79 implies @xmath199 , thus @xmath200 for such a time @xmath3 .
we next take @xmath201 .
we have @xmath202 where we have respectively simplified the exponential term , made the change of variables @xmath203 and used the definition of @xmath79 .
this concludes the proof of this step , and thus that of theorem [ th : exu ] .
in this section we prove various properties of the solution of the existence and uniqueness of which has been established in theorem [ th : exu ] .
we therefore assume throughout this section that , as for theorem [ th : exu ] , @xmath87 in @xmath151 and @xmath12 satisfies .
[ lem : max ] if @xmath204 for almost all @xmath76 then @xmath205 for almost all @xmath206 . using the definition of @xmath82
, we first have @xmath207 for almost all @xmath135 . because of recurrence relation on @xmath79
, we then find that @xmath208 for almost all @xmath135 .
consequently , the expression on @xmath9 gives the result .
[ lem : mass ] if @xmath209 then @xmath210 for all @xmath211 .
denote @xmath212 which is continuous since @xmath9 belongs to @xmath74 .
using , we obtain that @xmath148 reads @xmath213 for all @xmath214}$ ] , @xmath148 rewrites @xmath215 differentiating @xmath148 in the sense of distributions in time , we obtain @xmath216 using the definitions of @xmath82 and of @xmath79 . using that @xmath148 is continuous on @xmath217 with @xmath218
, we find @xmath219 on @xmath220 $ ] . for all @xmath221
, @xmath148 rewrites @xmath222 differentiating @xmath148 in @xmath223 , we obtain @xmath224 because of definitions of @xmath82 and of @xmath79 .
this implies @xmath219 on @xmath217 , hence the result .
[ lem : li ] assume @xmath87 satisfies @xmath225 then , @xmath79 defined by ( or , equivalently , @xmath19 , given ) satisfies @xmath226 if in addition @xmath87 satisfies @xmath227 then @xmath9 belongs to @xmath228 and there exists a constant @xmath229 which depends only on @xmath230 and the bound @xmath231 in such that @xmath232 _ proof . _
first , we notice that @xmath233 successively using the definition of @xmath19 , lemma [ lem : max ] and lemma [ lem : mass ] .
this immediately implies @xmath234 using the lower bound on @xmath12 , the @xmath235-bound on @xmath87 and the expression of @xmath9 in terms of @xmath79 , we find that @xmath9 belongs to @xmath228 and @xmath236 [ lem : dde ] assume @xmath237 .
then , @xmath82 and @xmath238 respectively defined by , and , belong to @xmath239 thus are continuous in time .
moreover , @xmath82 and @xmath79 satisfy , for almost all @xmath240 , @xmath241 \nonumber\end{aligned}\ ] ] and for almost all @xmath242 , @xmath243 as stated in theorem [ th : exu ] , the functions @xmath82 and @xmath79 belong to @xmath244 .
we show that both @xmath245 and @xmath246 belong to @xmath247 .
the expression on @xmath82 rewrites @xmath248 } p_0(\xi){{\rm e}^}{-\int_0^t \chi(\xi+\gamma(u))du } d\xi .
\end{aligned}\ ] ] differentiating the above expression in the sense of distributions in time , we obtain equation . since @xmath87 belongs to @xmath83 and @xmath12 and @xmath82
belong to @xmath249 belongs to @xmath247 and the equation holds for almost all @xmath169 . since @xmath250 on @xmath251
, @xmath246 belongs to @xmath252 .
differentiating the recurrence relation in the sense of distributions on @xmath253 , we have @xmath254 using that @xmath255 and @xmath79 respectively belong to @xmath256 and @xmath83 , the right - hand side of the above equation and thus @xmath246 belong to @xmath257 .
moreover , the equation holds for almost all @xmath242 .
this ends the proof .
[ lemma : tau ] assume that @xmath87 satisfies @xmath258 then @xmath259 belongs to @xmath132 so that the average stress @xmath20 defined by belongs to @xmath83 .
moreover , if there exists a scalar @xmath260 independent from @xmath261 such that @xmath262 , then there exists a constant @xmath263 independent from @xmath261 such that @xmath264 we multiply the expression on @xmath9 by @xmath265 and integrate in @xmath1 .
this implies @xmath266 using that @xmath267 and then the upper bound on @xmath268 .
we deduce @xmath269 so that @xmath259 belongs to @xmath132 .
moreover , from , we obtain , for all @xmath270 , @xmath271 additionally , with the change of variable @xmath272 , the last integral satisfies @xmath273 inserting the above inequality in yields . this concludes the proof . throughout our article
, we now assume that @xmath274 so that the five lemmata [ lem : max ] , [ lem : mass ] , [ lem : li ] , [ lem : dde ] and [ lemma : tau ] hold .
in this section , we assume that @xmath275 where @xmath53 is a given _ fixed _ constant and we study the longtime convergence for equation .
this allows us to lay some ground work which will prove useful for the general case of a slowly varying shear rate addressed in the next sections .
we prove the following .
[ th : tlc ] assume that @xmath275 with @xmath276 a given constant .
supply equation with an initial condition @xmath87 that satisfies .
consider @xmath9 the solution to and @xmath45 the associated stationary solution , the existence and uniqueness of those has been respectively established in theorem [ th : exu ] and lemma [ lem : ss ] .
then @xmath9 converges exponentially fast in time to @xmath45 for almost all @xmath277 .
in addition , @xmath19 defined by converges exponentially fast in time to @xmath278 .
more precisely , there exist @xmath279 such that , for all @xmath3 and for almost all @xmath277 , @xmath280 in addition , there exists a positive continuous function @xmath281 of @xmath282 such that for all @xmath283 , @xmath284 in the two estimates and , the rate @xmath285 can be chosen as @xmath286 for any @xmath287 , where @xmath288 is defined by : @xmath289 notice that the estimates and rely on the two natural timescales of the original problem : the exponential rate @xmath27 of the jump process to zero ( the hidden coefficient @xmath27 multiplying the right - hand side of the fokker - planck equation ) , and the typical time @xmath288 required for the process to leave the domain @xmath290 , when the shear rate is @xmath291 .
let us notice from that we may assume in the following , without loss of generality , that @xmath292 this will be an assumption on @xmath293 in the forthcoming sections .
this assumption is exclusively technical .
we require it to simplify some proofs where convolutions with exponential kernels are involved , in particular to use estimates such as @xmath294 ( which is only true if @xmath295 ) .
before we get to the proof of theorem [ th : tlc ] , to which the rest of this section is devoted , we need to introduce some notation and make some preliminaries .
denote by @xmath296 and @xmath297 . by linearity , and in place of @xmath82 and
@xmath79 introduced in theorem [ th : exu ] , we similarly introduce @xmath298 and @xmath299 we then have from theorem [ th : exu ] @xmath300 using lemma [ lem : dde ] , @xmath301 satisfies , for almost all @xmath302 , @xmath303 the equation on @xmath301 is a delay differential equation with constant coefficients .
the proof of theorem [ th : tlc ] is based upon three lemmata for such a delay differential equation , denoted in generality by @xmath304 where @xmath305 is a constant and @xmath306 is a locally integrable function .
equation is understood in the sense of distribution in time .
to such a delay differential equation is classically associated the unique function @xmath307 satisfying the following properties : 1 .
@xmath308 ; 2 .
@xmath309 ; 3 .
@xmath307 is continuous on @xmath145 ; 4 .
@xmath307 satisfies for all @xmath310 , @xmath311 the three lemmata useful for the proof of theorem [ th : tlc ] are lemma [ lem : cons ] , itself proved using lemma [ lem : residue ] , and lemma [ lem : varconst ] .
the latter two lemmata , lemma [ lem : residue ] and lemma [ lem : varconst ] , are borrowed respectively from @xcite and @xcite .
they are valid for more general delay differential equations , but for simplicity , we state them here for our specific delay differential equation .
* equation ( 5.10 ) , p. 22 ) [ lem : residue ] consider @xmath312 defined by and the associated properties above .
denote by @xmath313 then , for all @xmath314 such that no root of @xmath148 has real part equal to @xmath315 , the function @xmath312 writes , for all @xmath310 , @xmath316 where @xmath317 are the roots of @xmath148 such that @xmath318 and @xmath319 denotes the residue at @xmath320 . intuitively , is obtained as follows .
the function @xmath312 solves @xmath321 on @xmath322 , which by laplace transform and using the notation , yields @xmath323 ( where @xmath324 denotes the laplace transform of @xmath312 ) .
it remains then to divide by @xmath148 and apply a reverse laplace transform to finally obtain @xmath312 .
the difficulty is of course related to the zeros of the function @xmath148 .
[ lem : varconst ] ( * ? ? ?
* theorem 3.7 , p. 75 ) consider two functions @xmath325}$ ] and @xmath326 .
then , there exists a unique solution @xmath327 verifying in the sense of distribution , and @xmath328 satisfies : for @xmath329 , @xmath330 the result of lemma [ lem : varconst ] is stated in @xcite for @xmath306 continuous but holds for @xmath331 .
indeed , the existence of a unique solution is still valid in this more general setting ( see @xcite ) and expression satisfies almost everywhere .
as announced above , we first use lemma [ lem : residue ] to prove the following [ lem : cons ] assume @xmath332 then there exist @xmath285 and @xmath333 which depend only on @xmath334 and @xmath335 and such that for all @xmath310 @xmath336 with @xmath337 moreover , @xmath293 can be chosen as , for any @xmath287 .
we immediately emphasize that the point in lemma [ lem : cons ] is to show that the prefactor @xmath338 and the exponent @xmath293 appearing in do not depend on @xmath288 itself , but can be chosen locally uniformly , that is , depend only on the bounds @xmath334 and @xmath335 of the interval where @xmath288 lies .
proving for a fixed @xmath288 is a simple consequence of the classical results contained e.g. in @xcite .
[ rem : alternate ] using numerical experiments , we will show in section [ sec : numerics ] that the rate @xmath293 given by for the estimation is indeed sharp .
it is interesting to note that our result [ lem : cons ] in the present section does not explicitly require such a sharpness .
a simpler alternate proof ( which we owe to one of the anonymous referees ) shows a similar , however non sharp estimation .
that proof is based on the observation @xmath339 which shows that the function @xmath340 necessarily cancels on any interval @xmath341 .
this leads to the following induction relation on the maximum value @xmath342 of @xmath343 on @xmath341 @xmath344 indeed , denoting @xmath345 a real such that @xmath346 , @xmath340 solution of satisfies , for all @xmath347 , @xmath348 hence .
denote @xmath349 the integer part .
the induction relation then implies @xmath350 and therefore , using that for all @xmath351 , @xmath352}(t)-\frac1{1+\omega},$ ] @xmath353 where @xmath354 and @xmath355 are respectively decreasing and increasing functions of @xmath288 .
this proves lemma [ lem : cons ] , with the values @xmath356 and @xmath357 .
_ proof of lemma [ lem : cons ] . _
the proof falls in three steps .
we first derive an upper bound on the real part of the nonzero roots of the function @xmath148 defined by .
this upper bound actually yields the exponent @xmath293 in the exponential estimates of lemma [ lem : cons ] and thus of theorem [ th : tlc ] . in the second step ,
we apply lemma [ lem : residue ] . in the third and final step ,
we conclude .
[ [ step - upper - bound - on - the - real - part - of - the - nonzero - roots - of - h ] ] step : upper bound on the real part of the nonzero roots of @xmath148 the roots of the function @xmath148 defined by are @xmath358 and the complex numbers @xmath359 ( with @xmath360 ) that satisfy @xmath361 it is easy to check that @xmath33 implies @xmath362 and conversely , so that in the following , we assume @xmath363 and @xmath364 .
the equation rewrites @xmath365 .
since the function @xmath366 is non - positive on @xmath367}$ ] , @xmath368 satisfies @xmath369 moreover , we combine and and obtain @xmath370 this implies that @xmath60 is negative and therefore , using the bounds and respectively on @xmath288 and @xmath368 , @xmath371 the function @xmath372 is continuous , satisfies @xmath373 and @xmath374 so that by the intermediate value theorem , there exists @xmath375}. \end{aligned}\ ] ] the scalar @xmath285 depends only on @xmath335 .
additionally , the real part @xmath60 of the nonzero roots of @xmath148 satisfies @xmath376 and , combining and , @xmath377 therefore @xmath293 can be chosen as , for any @xmath287 .
[ [ step - applying - lemmalemresidue ] ] step : applying lemma [ lem : residue ] from the previous step , we know that the only root of @xmath148 with real part strictly above @xmath378 is @xmath358 .
we now apply lemma [ lem : residue ] with @xmath379 .
since the root @xmath358 is a simple root of @xmath148 , the residue of @xmath380 at @xmath358 is @xmath381 .
equation therefore writes @xmath382 proving that there exists @xmath333 which depends only on @xmath334 and @xmath335 , such that for all @xmath310 , @xmath383 therefore amounts to concluding the proof of lemma [ lem : cons ] .
actually , we will show that this holds up to changing @xmath293 to @xmath384 in the right hand side , for any positive @xmath385 .
this will conclude the proof .
[ [ step - exponential - bound ] ] step : exponential bound we first show , for all @xmath310 , @xmath386 by integration by parts , we have @xmath387}_{-t}^t.\end{aligned}\ ] ] introduce @xmath388 such that for all @xmath389 , @xmath390 then , for all @xmath391 , @xmath392 so that , for all @xmath310 , @xmath393}_{-t}^t \right\vert } \le \frac{4{{\rm e}^}{-bt}}{{\left\vert t \right\vert}t}.\end{aligned}\ ] ] by passing to the limit @xmath394 in , we thus obtain . now , for all @xmath395 , @xmath396 introduce @xmath397 which we may take depending only on @xmath335 , such that for all @xmath398 , @xmath399 so that @xmath400 for @xmath401 and @xmath310 , this implies @xmath402 the function @xmath403 is continuous for @xmath404}$ ] and is therefore bounded by a constant that only depends on @xmath334 and @xmath335 . from and
the bound , we deduce that there exists a constant @xmath333 that also only depends on @xmath334 and @xmath335 such that @xmath405 for any positive @xmath385 .
this concludes the proof .
we are now in position to turn to the _ proof of theorem [ th : tlc ] . _
the proof proceeds in five steps . in step [ s : varconst ] , we apply the above lemmata to the delay differential equation . in steps
[ s : terme0 ] and [ s : terme1 ] , we derive some estimates that will be useful , in the last two steps , to show convergence of @xmath301 , and eventually @xmath406 . [ [ s : varconst ] ] step : applying lemmata [ lem : cons ] and [ lem : varconst ] the function @xmath301 defined by satisfies in the particular case @xmath407 and @xmath408 .
notice that , by the lemma [ lem : dde ] , the function @xmath409 defined by belongs to @xmath239 and thus , in particular , is continuous in time .
we apply lemma [ lem : varconst ] and find ( by integration by parts ) : for all @xmath302 , @xmath410 we now recall that , in this section , the value of @xmath288 is fixed by at @xmath411 .
we can apply lemma [ lem : cons ] and insert the decomposition of @xmath312 into the previous equation on @xmath301 .
we obtain @xmath412 where @xmath340 satisfies with @xmath413 only depending on @xmath411 .
moreover , @xmath293 can be chosen as for any positive @xmath385 as stated in lemma [ lem : cons ] .
our next two steps consist in deriving a couple of estimates ( see and below ) on the terms of .
[ [ s : terme0 ] ] step : longtime convergence of @xmath414 using , the function @xmath409 defined by satisfies , for almost all @xmath310 , @xmath415 computing @xmath416 and integrating from @xmath358 to @xmath3 yield @xmath417 so that , respectively with the changes of variables @xmath418 and @xmath419 in the last two integrals , we obtain @xmath420 for @xmath421 , this implies that @xmath422 using that @xmath423 .
we deduce @xmath424 [ [ s : terme1 ] ] step : longtime convergence of @xmath425 using on @xmath426 and the estimate on @xmath340 , we have , for @xmath421 , @xmath427 here , we have used the assumption on @xmath293 . [
[ step - longtime - convergence - of - gt ] ] step : longtime convergence of @xmath428 using the decomposition and the estimates and derived in steps [ s : terme0 ] and [ s : terme1 ] , we have , for @xmath421 , @xmath429 recall that @xmath301 defined by satisfies by linearity @xmath430 and that @xmath431 a.e .
( see ) .
we have obtained estimate on @xmath432 . [ [ step - longtime - convergence - of - q ] ]
step : longtime convergence of @xmath406 we now turn to @xmath296 . using , we have @xmath433 in view of the estimate on @xmath301 , we deduce , for almost all @xmath76 and @xmath310 , @xmath434 this concludes the proof of theorem [ th : tlc ] .
our main result in this section is the following .
[ th : tle ] consider @xmath12 a lipschitz function with lipschitz constant @xmath435 , which satisfies , for all @xmath329 , @xmath436 consider an initial condition @xmath87 which satisfies .
for @xmath437 such that @xmath438 , consider the functions @xmath439 and @xmath440 respectively solutions to @xmath441 the existence and uniqueness of which have been respectively established in theorem [ th : exu ] and lemma [ lem : ss ] . then , there exist constants @xmath442 independent from @xmath443 and @xmath444 ( satisfying @xmath438 ) such that , @xmath445 and , for almost all @xmath277 such that @xmath446 , @xmath447 .
\label{conv : pmp}\end{aligned}\ ] ] in order to prove theorem [ th : tle ] , we need the following technical lemma . [
lem : intt ] consider @xmath12 a function of time that satisfies .
denote by @xmath448 .
then , for all @xmath449 and almost all @xmath76 , we have @xmath450 denote by @xmath451 . for all @xmath449 and almost all @xmath76 ,
we have @xmath452 using that @xmath453 .
we now estimate the above expression depending on @xmath1 . for almost all @xmath454 ,
the function @xmath455 is decreasing in @xmath1 so that @xmath456 .
moreover , because of , the function @xmath12 satisfies @xmath457 so that @xmath458 , hence for almost all @xmath454 . additionally , because of , the function @xmath157 satisfies , for all @xmath459 , @xmath460 this yields that @xmath461 is lipschitz with a lipschitz constant @xmath462 on @xmath145 .
therefore , for almost @xmath463 , @xmath464 hence . for almost all @xmath465 ,
the function @xmath455 is increasing in @xmath1 so that @xmath466 .
moreover , @xmath467 so that @xmath468 , hence for almost all @xmath465 .
the result also holds in the case @xmath469 $ ] where @xmath470 .
this ends the proof .
now that we have proved the technical lemma [ lem : intt ] , we turn to the _ proof of theorem [ th : tle ] . _
the proof is divided into six steps .
the first step establishes a delay differential equation on a function @xmath471 , for which an explicit decomposition is known thanks to the lemma [ lem : varconst ] .
we then rewrite @xmath471 in a different form whose terms are estimated in steps [ s : k0 ] and [ s : k1 ] . in the last two steps , we use these estimates to obtain and then . before we get to the proof we introduce some notation .
the scalars @xmath472 are fixed and satisfy @xmath473 . in section
[ sec : tlc ] , we have introduced @xmath474 which from bounds on @xmath12 satisfies @xmath475 we can therefore apply lemma [ lem : cons ] to the function @xmath476 satisfying with @xmath477 , so that there exist @xmath413 which depend only on @xmath22 , @xmath231 and @xmath260 such that and hold for all @xmath310 .
notably , @xmath293 and @xmath338 are independent from @xmath443 ( and @xmath444 ) .
[ [ step - applying - lemmalemvarconst ] ] step : applying lemma [ lem : varconst ] for a fixed @xmath443 , denote by @xmath478 and by @xmath479 the solution to @xmath480 with the initial condition @xmath481 .
consistently with , let us also introduce @xmath482 where @xmath156 satisfies .
then @xmath483 belongs to @xmath239 ( because @xmath484 and @xmath479 do , see lemma [ lem : dde ] ) and satisfies , for almost all @xmath485 @xmath486 introduce @xmath487 .
we apply lemma [ lem : varconst ] and obtain ( using the same computations as in above and the fact that @xmath488 ) , @xmath489 [ [ step - rewriting - g_epsilon ] ] step : rewriting @xmath471 in order to rewrite @xmath471 we show that , for almost all @xmath490 , @xmath491 first , the function @xmath156 solution to with @xmath87 as initial condition satisfies , for all @xmath492 , @xmath493 denote @xmath494 a function of @xmath495 , @xmath102 a mollifier on @xmath2 and @xmath496 . inserting @xmath497 in yields , for @xmath107 sufficiently large such that @xmath498 , @xmath499 which rewrites @xmath500 the function @xmath501 belongs to @xmath502,l^1)$ ] , see theorem [ th : exu ] , so that , by the dominated convergence theorem , for all @xmath503 $ ] , @xmath504 and @xmath505 converge to @xmath506 defined by as @xmath107 goes to infinity
. moreover @xmath507 belongs to @xmath508 ( the proof is similar to the one in step [ s : espace ] of theorem [ th : exu ] ) so that @xmath509 is continuous in @xmath1 .
passing to the limit @xmath510 in the above equation yields @xmath511 hence since @xmath484 and @xmath512 belong to @xmath513 .
we then denote @xmath514 so that the expression on @xmath471 rewrites @xmath515 using the decomposition @xmath516 ( see ) that was established in lemma [ lem : cons ] .
we now derive estimates on the two terms of the above expression , when @xmath517 .
[ [ s : k0 ] ] step : estimate of @xmath518 introduce @xmath519}}(\sigma ) { { \rm e}^}{\frac{\sigma + \sigma_c}{{\dot\gamma}(\theta)}}\end{aligned}\ ] ] which satisfies , in @xmath520 , @xmath521}}(\sigma ) { { \rm e}^}{\frac{\sigma + \sigma_c}{{\dot\gamma}(\theta)}}\end{aligned}\ ] ] and @xmath522}}(\sigma ) { { \rm e}^}{\frac{\sigma + \sigma_c}{{\dot\gamma}(\theta)}}.\end{aligned}\ ] ] for @xmath523 , take as a test function @xmath524}}(\sigma ) { { \rm e}^}{\frac{\sigma + \sigma_c}{{\dot\gamma}(\theta ) } } \zeta^m_{[0,s)}(t)\end{aligned}\ ] ] in and pass to the limit in @xmath107 and then @xmath525 . here and in the following , @xmath526 denotes a @xmath527 function with compact support in @xmath528 , such that @xmath526 converges pointwise to @xmath529 .
we omit the details , the arguments being similar to those in .
we obtain @xmath530 with changes of variable @xmath531 and @xmath532 in the last two integrals , this rewrites @xmath533 let us assume that @xmath534 and introduce @xmath535}}(\sigma ) { { \rm e}^}{\frac{\sigma+\sigma_c}{{\dot\gamma}(\theta ) } } { \rm 1\mskip-4mu l}_{(0,s-2\omega_\theta)}(t ) \nonumber\\&\quad + { \rm 1\mskip-4mu l}_{{\left[-\sigma_c,\sigma_c-{\dot\gamma}(\theta){\left(s - t\right)}\right]}}(\sigma ) { \rm 1\mskip-4mu l}_{(s-2\omega_\theta , s)}(t ) -{\rm 1\mskip-4mu l}_{[-\sigma_c,\sigma_c]}(\sigma ) \end{aligned}\ ] ] which satisfies , in @xmath536 , @xmath537}}(\sigma ) { \rm 1\mskip-4mu l}_{(0,s-2\omega_\theta)}(t ) - \delta_{\sigma_c-{\dot\gamma}(\theta){\left(s - t\right)}}(\sigma ) { \rm 1\mskip-4mu l}_{(s-2\omega_\theta , s)}(t)\end{aligned}\ ] ] and @xmath538}}(\sigma ) { \rm 1\mskip-4mu l}_{(0,s-2\omega_\theta)}(t ) \\&\qquad -({\dot\gamma}(\theta)-{\dot\gamma}(\epsilon t ) ) \delta_{\sigma_c-{\dot\gamma}(\theta){\left(s - t\right)}}(\sigma ) { \rm 1\mskip-4mu l}_{(s-2\omega_\theta , s)}(t).\end{aligned}\ ] ] we again use a regularization @xmath539}}(\sigma ) { { \rm e}^}{\frac{\sigma+\sigma_c}{{\dot\gamma}(\theta ) } } \
\zeta^m_{[0,s-2\omega_\theta)}(t ) \\&\quad + \rho^n*{\rm 1\mskip-4mu l}_{{\left[-\sigma_c,\sigma_c-{\dot\gamma}(\theta){\left(s - t\right)}\right]}}(\sigma ) \ \zeta^m_{(s-2\omega_\theta , s)}(t ) \\&\quad -\rho^n*{\rm 1\mskip-4mu l}_{[-\sigma_c,\sigma_c]}(\sigma ) \ \zeta^m_{[0,s)}(t)\end{aligned}\ ] ] and pass to the limit in @xmath540 in addition , from its definition , we know that @xmath541 satisfies ( see ) @xmath542 summing up expressions , and , we obtain @xmath543 taking @xmath544 and summing up the second and the third term ( with the change of variable @xmath545 in the third term ) , this rewrites @xmath546 using the lipschitz property of @xmath12 and @xmath235-bound on @xmath156 , this implies @xmath547 for a constant @xmath548 , we have @xmath549 using estimate or variants , we find @xmath550 since @xmath551 , we obtain @xmath552 with @xmath553 a constant that is independent from @xmath443 and @xmath444 . throughout the rest of the proof below
, we will likewise denote by @xmath553 such a constant , whose precise value may change from one occurrence to another .
[ [ s : k1 ] ] step : estimate of @xmath554 inserting expression of @xmath541 ( see above for a similar computation ) , @xmath555 defined by satisfies @xmath556 for further use , notice that the integral in the second line above can be rewritten as : @xmath557 in order to rewrite the first term of the right - hand side , we use again the functions @xmath558 and @xmath559 respectively defined by and . using a regularization of @xmath560 as test function in ( with @xmath561 ) and then passing to the limit
, we obtain @xmath562 this rewrites ( using similar computations as in above ) @xmath563 similarly , using a regularization of @xmath564 as test function in , we obtain @xmath565 so that ( using similar computations as in above ) @xmath566 we now perform the linear combinations : - ( + ) .
the last two term of the right - hand side of cancel out with the right - hand sides of and ( using in particular ) so that @xmath567 using lemma [ lem : cons ] , @xmath568 satisfies @xmath569 ( this is a consequence of and ) so that ( using ): @xmath570 using the lipschitz property of @xmath12 , the @xmath235-bound of @xmath156 and estimates and of @xmath568 and @xmath571 , we obtain @xmath572 using the estimate or variants and the bounds on @xmath573 , one concludes @xmath574 [ [ step - estimate - of - g_epsilonleftdfracthetaepsilonright ] ] step : estimate of @xmath575 using the decomposition and estimates and , we find @xmath576 moreover , the estimate established in theorem [ th : tlc ] yields , using the bounds on @xmath573 @xmath577 where @xmath578 is independent from @xmath444 and @xmath443 . by the definition of @xmath471 , we have @xmath579 so that @xmath580 where @xmath581 is independent from @xmath444 and @xmath443 .
this concludes the proof of .
[ [ step - estimate - of - p_epsilonleftdfracthetaepsilonsigmaright --- p_inftythetasigma ] ] step : estimate of @xmath582 recall that the scalars @xmath472 are fixed and satisfy @xmath473 . for all @xmath583 , denote @xmath584 where @xmath585 .
the expression of @xmath9 , that was established in theorem [ th : exu ] , reads , for almost all @xmath277 such that @xmath446 , @xmath586 note that the condition @xmath587 is not restrictive because we are interested in the limit @xmath588 for a fixed @xmath1 .
the rest of the proof depends on the value of @xmath1 .
let us start with the case @xmath589 .
we have @xmath590 we now apply lemma [ lem : intt ] with @xmath591 and @xmath592 and obtain @xmath593 notice that actually holds for all @xmath594 ( this will be used below ) . from the expression of @xmath45 , @xmath595 when @xmath596 .
this gives for almost all @xmath596 .
let us now consider the case @xmath597 $ ] .
let us introduce the notation @xmath598 .
we have , for all positive @xmath1 , @xmath599 using the estimate on @xmath600 , the lipschitz property and the boundedness of @xmath12 and the expression which gives @xmath601 . here , we also used the fact that @xmath602 , which is a consequence of the lipschitz property of @xmath461 : @xmath603 .
for almost all @xmath597 $ ] , we deduce ( using again ) @xmath604 moreover , we have , from , @xmath605 for almost all @xmath606 $ ] .
this proves in this region of @xmath1 .
we now eventually consider the case @xmath607 . applying the taylor - lagrange theorem on @xmath461 at @xmath608
, we have ( using the fact that @xmath12 is bounded from above ) @xmath609 using that the function @xmath144 is 1-lipschitz on @xmath145 , this implies @xmath610 for almost all @xmath607 , @xmath45 reads ( see ) @xmath611 and thus , we have ( using and ) @xmath612 this proves and ends the proof .
the purpose of this section is to obtain a macroscopic limit for the equation we have been studying . for @xmath613 presumably small
, we first introduce the notation @xmath614 , for any real - valued function @xmath615 of the real variable @xmath616 . the small parameter @xmath444 encodes the discrepancy between the typical time of variation of the macroscopic shear rate @xmath36 and the typical mesoscopic time @xmath3 with which the solution to varies .
the parameter @xmath444 will therefore be the small parameter on which our macroscopic limit is performed . to obtain a macroscopic limit of our equation
, we will look at a specific macroscopic time , denoted by @xmath443 , related to the mesoscopic time @xmath3 by @xmath617 . letting @xmath444 vanish ,
we will obtain the corresponding macroscopic behavior of our mesoscopic quantities . in this latter process
, we will of course use the results of the previous section , since formally , given a macroscopic time @xmath443 , the corresponding mesoscopic time @xmath618 is a long time limit .
we begin by formally multiplying respectively by @xmath1 and @xmath619 and integrating in @xmath1 .
we next evaluate at the time @xmath620 the two equations obtained .
this gives the following system of equations on @xmath621 and @xmath622 , the quantities associated with @xmath156 solution to , [ eq : sapprox ] align ( ) & = - ( ) ( , ) d+ ( ) , [ eq : sapproxa ] + ( ) & = -()+ ( ) ( ( , _ c)-(,-_c)).[eq : sapproxb ] the difficulty is that this system is not closed in the couple of unknown functions @xmath623 since @xmath624 still appears .
we next intend to `` eliminate '' @xmath624 from this system , and thereby obtain a system of ordinary differential equations , the solution of which is an approximation , for @xmath444 sufficiently small , of @xmath621 and @xmath622 .
there are indeed many options to do so .
we present two sets of equations which we can derive and that , in a sense made precise below , are equivalent to system .
the precise results are contained in two theorems we now successively state and prove , namely theorem [ th : mac1 ] and theorem [ th : mac2 ] .
[ th : mac1 ] for @xmath625 , consider @xmath626 the solution to the following differential equation in `` macroscopic '' time @xmath443 : @xmath627 supplied with any initial condition @xmath628 independent of @xmath444 . under the assumptions of theorem
[ th : tle ] , consider the function @xmath156 solution to .
then , there exists a constant @xmath629 , independent from @xmath444 and @xmath443 provided @xmath438 ( with @xmath231 the lower bound of @xmath12 in ) , such that @xmath630 where ( in accordance with ) @xmath631 denotes the stress associated to @xmath156 solution to .
the above result holds whatever the initial condition @xmath628 for the equation .
indeed , we are only considering macroscopic times such that @xmath438 and , in the limit @xmath588 , the boundary layer in time around @xmath632 does not affect the result . _
proof_. we first give the arguments to derive system . in the proof of theorem [ th : tlc ] , we established equation , which reads @xmath633 denoting @xmath494 a function of @xmath495 , @xmath102 a mollifier on @xmath2 and @xmath496 , we used the test function @xmath497 in the weak form of equation and then passed to the limit @xmath634 .
we now establish an equation on @xmath635 with the same method .
denote @xmath636})$ ] and use @xmath637 as a test function in : @xmath638 we pass to the limit @xmath634 , using that the function @xmath639 belongs to @xmath502,l^1)$ ] ( see theorem [ th : exu ] ) , @xmath640 , the dominated convergence theorem for the terms in the left - hand side , and that @xmath641 for the right - hand side .
we obtain , for all @xmath642 , @xmath643 so that @xmath644 changing the variable @xmath3 in @xmath645 in equations and , we obtain system . now that we have established system , we rewrite in the form @xmath646 using the definition of @xmath13 .
moreover , we use the expression of @xmath45 to rewrite as follows @xmath647 subtracting the above two equations yields @xmath648 denote @xmath649 so that @xmath650 .
we next apply the duhamel formula and find @xmath651 using the estimate on @xmath635 , the bounds on @xmath12 , we find @xmath652 where we recall that @xmath553 denotes a constant that is independent from @xmath443 and @xmath444 and whose precise value may change from one occurrence to another . inserting the above estimate and the estimate on @xmath653 in yields @xmath654}\label{eq : xex}\\ & \le k { \left({{\rm e}^}{-\frac\theta\epsilon } + \frac1{1-b}{\left({{\rm e}^}{-b\frac\theta\epsilon}- { { \rm e}^}{-\frac\theta\epsilon}\right ) } + \frac\theta\epsilon { { \rm e}^}{-\frac\theta\epsilon } + \epsilon\right)}\nonumber\\ & \le k{\left(\frac{b}{{\left\vert
1-b \right\vert}}\frac{{{\rm e}^}{-1}}{\theta } + \frac1{{\left\vert 1-b \right\vert}}\frac{{{\rm e}^}{-1}}{b\theta } + \frac{4{{\rm e}^}{-2}}{\theta}+1\right)}\epsilon,\nonumber \ ] ] using that the functions @xmath655 and @xmath656 are respectively bounded by @xmath657 and @xmath658 on @xmath659 in order to derive the last line .
this concludes the proof .
[ th : mac2 ] for @xmath613 , consider @xmath661 satisfying the following system of equations in macroscopic time @xmath443 : [ eq : mac2 ] align ( ) & = - ( ) f_^ * * ( ) _ ^ * * ( ) + ( ) , [ eq : mac2a ] + ( ) & = -f_^ * * ( ) + , [ eq : mac2b ] where we have introduced the notation @xmath662 and where the equations are supplied with any couple of scalars ( independent of @xmath444 ) @xmath663 as initial conditions .
consider @xmath664 such that @xmath665 . under the assumptions of theorem
[ th : tle ] , consider @xmath156 the solution of .
then , there exists a constant @xmath629 independent from @xmath444 and @xmath443 such that , @xmath666 as in theorem [ th : mac1 ] , the above result holds whatever the set of initial conditions @xmath663 for the system of differential equations .
_ proof_. the proof falls in three steps .
we first study @xmath484 , then an auxiliary function @xmath667 and finally @xmath635 . [
[ step - approximation - of - f_epsilon ] ] step : approximation of @xmath668 applying the duhamel formula to yields @xmath669 using the lipschitz property of @xmath12 and denoting by @xmath670 ( following and ) , we easily obtain @xmath671 collecting the above equation and the estimate established in theorem [ th : tle ] , we obtain @xmath672 using that the function @xmath655 is bounded by @xmath657 on @xmath659 . [ [ step - introduction - of - the - auxiliary - function - beta_epsilon ] ] step : introduction of the auxiliary function @xmath667 we now introduce @xmath673 defined for a density probability @xmath9 such that @xmath674 . denote @xmath494 a function of @xmath495 , @xmath102 a mollifier on @xmath2 and @xmath675 . using @xmath676})*\rho^n(\sigma )
\rho(t)$ ] as test function in and passing to the limit with the same arguments as in , we obtain @xmath677 consider @xmath678 the solution of the ordinary differential equation @xmath679 supplied with a scalar @xmath680 as initial condition . subtracting the above two equations and using that ( see ) @xmath681 and @xmath682 , we obtain @xmath683 with @xmath684 recall that @xmath649 and @xmath685 . applying the duhamel formula yields , @xmath686 using the upper bound
, @xmath687 satisfies @xmath688 moreover , @xmath678 solution of satisfies @xmath689 using , the boundedness of @xmath12 and the estimate on @xmath690 , the right - hand side @xmath691 of satisfies , @xmath692 inserting the three above inequalities in implies @xmath693 additionally , applying the duhamel formula to and using the explicit formula @xmath694 yield @xmath695 using the lipschitz property of @xmath12 and the estimate , we obtain @xmath696 combining and leads to @xmath697 and eventually , @xmath698 respectively using and to estimate the two terms of the right - hand side . here
we have used the notation @xmath699 , and the fact that @xmath700 . [ [ step - approximation - of - tau_epsilon ] ] step : approximation of @xmath635 we now turn to @xmath701 . combining and yields @xmath702 from the formula on @xmath45 , we compute @xmath703 , @xmath704 , @xmath705 so that the term @xmath706 cancels out because of the definition of @xmath707 .
we therefore obtain @xmath708 with @xmath709 we have @xmath710 ( see ) so that @xmath711 the duhamel formula then implies @xmath712 using the upper bound , @xmath621 satisfies @xmath713 moreover , the solution @xmath714 of satisfies ( using the non negativity of @xmath707 and @xmath668 ) : @xmath715 so that @xmath716 collecting established in theorem [ th : tle ] , and , the right - hand side @xmath717 of satisfies @xmath718 inserting the three above inequalities in implies @xmath719 we end this section with a discussion on the two macroscopic limits and we have obtained .
first , as mentioned above , there are many ways to close the system in the limit @xmath588 .
we have proposed here two possible macroscopic limits , which are indeed close up to terms of order @xmath720 to the original problem .
second , we would like to argue that the system derived in theorem [ th : mac2 ] is physically more relevant .
indeed , up to changing the coefficient @xmath721 defined by by a constant , this system belongs to a class of equations introduced in @xcite to model the evolution of aging fluids .
these equations read ( see ( * ? ? ?
* eq . ( 1 ) ) ) [ eq : derec ] align & = -f+ , [ eq : derectau ] + & = - u(f ) + v(f , , ) , [ eq : derecf ] where @xmath722 and @xmath723 are positive functions .
the formal similarity between and is clear . for this class of systems ,
the fluidity @xmath19 appears as the inverse of the relaxation time for the stress @xmath20 in equation . in equation
the evolution results from the competition between the two terms with opposite signs .
aging , meaning solidification of the fluid , is modeled by the negative term .
it makes the fluidity decrease so that the relaxation phenomenon is slower with time .
the opposite effect , flow - induced rejuvenation , is modeled by the positive term , which makes the fluidity ( the inverse relaxation time ) increase .
note that the assumption @xmath721 constant is a reasonable approximation when @xmath12 is small . in this case , system is close to system , which is indeed of the form . in section [ sec : numerics ] ,
we present numerical results that confirm that the solutions to and are indeed close when @xmath12 is small .
this section is devoted to some numerical experiments .
we consider three different situations , depending on the value of the function @xmath36 for @xmath724 $ ] . * in our first series of tests , we consider the _ constant _ shear rate @xmath37 , for different values @xmath725 of @xmath291 . in that case , the final time is @xmath726 * in our second series of tests , we consider @xmath727 , and the final time @xmath728 with @xmath729 and @xmath444 varying between the values @xmath730 and @xmath731 . * in our third and final series of tests , we take @xmath732 , and the same values of @xmath261 , @xmath443 , and @xmath444 as in case ( ii ) . in all our tests , the reference equation , namely ( or more precisely ) , is simulated over the time interval @xmath733 $ ] .
since in theory it is posed on the whole real line , we need to truncate the domain and thus actually solve the equation on the bounded interval @xmath734 $ ] ( with periodic boundary conditions ) , for @xmath735 , with a constant space step @xmath736 .
the initial condition @xmath87 is the normal centered gaussian density , appropriately renormalized on the interval @xmath737 $ ] .
the threshold value for the stress is @xmath738 .
the time discretization is performed using a splitting method : over the time interval @xmath739 $ ] , equation is solved explicitly , pointwise for each @xmath740 , while the equation for @xmath741 is indeed solved using the conservation of the total mass of the density @xmath9 . in short , the value of the density at zero is adjusted so that the integral of @xmath9 is one .
see @xcite for more details . the advection equation is solved using an upwind finite difference scheme .
all computations are performed using c++ .
\(i ) for our first series of tests , performed for the shear rate @xmath742 , we first check the optimality of the long time convergence result stated in theorem [ th : tlc ] . for the various values of @xmath291
indicated above , we simulate . using a least - square fit
, we then estimate , in function of @xmath291 , the exponent of the exponential rate of convergence of @xmath743 to zero as time goes to infinity , more precisely what corresponds to the parameter @xmath293 of the right - hand side of .
the function @xmath45 is of course the stationary solution , itself function of @xmath291 , explicitly determined in our theoretical study .
note that for this practical experiment we intentionally make a confusion between the rate of convergence of the pointwise difference @xmath744 in and its @xmath745 norm .
figure [ fig : agamma ] shows , as a function of @xmath291 , a comparison between the rate of convergence fitted on the numerical results and the theoretical value of the inferior bound on this rate of convergence provided by our theoretical estimate .
the two sets of data agree , thereby showing the quality of our estimate .
\(ii ) we next consider @xmath727 , and implement the change of time scale by taking @xmath746 as input for equation , that is , we solve equation .
we do this for various values of the small parameter @xmath444 . in this case
( ii ) , our purpose is twofold .
first , we consider the quantities @xmath747 and @xmath748 where @xmath156 and @xmath45 are the solutions to and , respectively .
we wish to check that , as the estimate proved in theorem [ th : tle ] suggests , these two quantities behave linearly in function of @xmath444 for small @xmath444 . the figures [ fig : directfk_f ] and [ fig : directfk_tau ]
show this is indeed the case .
our second purpose is to illustrate that the macroscopic equations obtained in theorems [ th : mac1 ] and [ th : mac2 ] indeed behave similarly to the original mesoscopic equation , that is , provide a fluidity @xmath19 and a stress @xmath20 comparable to that computed from the solution to the latter equation . for the different values of @xmath444
mentioned above , we therefore simulate equations and on the time interval @xmath749}$ ] , with the time step @xmath750 .
the values of @xmath751 , @xmath752 et @xmath753 are displayed on figures [ fig : directfk_f ] and [ fig : directfk_tau ] .
we observe that the convergence is linear in @xmath444 , as predicted by our theoretical results theorems [ th : mac1 ] and [ th : mac2 ] .
the macroscopic behavior is thus suitably reproduced , up to an error of size @xmath720 .
\(iii ) our final test case addresses the case @xmath732 .
we again rescale the time and consider @xmath38 .
a similar experiment as that performed in the previous case ( ii ) again shows that equation reproduces well the stress tensor computed from the solution to equation , for the different values of @xmath444 . simulating and , we compute @xmath754 .
the results are displayed on figure [ fig : directfk001_tau ] .
but our purpose here is also to illustrate another fact .
when @xmath12 is small , and it is indeed the case for our specific choice of @xmath12 in this case ( iii ) , the value of the parameter @xmath721 defined by and appearing in the macroscopic system is approximately 2 . system is thus close to the system as explained above , this system of differential equations belongs to the class of systems explicitly suggested in @xcite as a macroscopic system of evolution of @xmath19 and @xmath20 for a non newtonian aging fluid . our theoretical results of the previous sections can therefore be interpreted as a _ derivation _
, from a model at a finer scale , of the macroscopic system , present in the applicative literature . on figure [ fig : directfk001_tau ]
, we indeed check that the stresses solution to the systems and are close , up to an error of size @xmath720 , to the stress provided by , | we study a one - dimensional equation arising in the multiscale modeling of some non - newtonian fluids . at a given shear rate ,
the equation provides the instantaneous mesoscopic response of the fluid , allowing to compute the corresponding stress . in a simple setting , we study the well - posedness of the equation and next the long - time behavior of its solution . in the limit of a response of the fluid much faster than the time variations of the ambient shear rate
, we derive some equivalent macroscopic differential equations that relate the shear rate and the stress .
our analytical conclusions are confronted to some numerical experiments .
the latter quantitatively confirm our derivations .
non - newtonian fluids ; micro - macro model ; longtime behavior . |
the initial model for the afterglows of gamma - ray bursts ( grbs ) involved synchrotron emission from nonthermal electrons accelerated to a power - law spectrum in a relativistic spherical blast wave expanding into a constant - density , presumably interstellar , medium ( ism ; mszros & rees 1997 ) .
this simplest model has difficulty , however , explaining quantitatively the dozen or so sources whose afterglows are observed well enough to allow for detailed modeling ( e.g. , panaitescu & kumar 2002 ) .
the most commonly discussed complication is a collimated energy injection ( rhoads 1997 ; sari , piran & halpern 1999 ) .
this jet model provides a reasonable fit to the majority of the well observed afterglows , assuming a constant - density ambient medium ( panaitescu & kumar 2002 ) .
for some sources , a blast wave expanding into an ambient medium of @xmath5 density distribution , as expected of a stellar wind , can fit the data equally well or even better ( chevalier & li 2000 , cl00 hereafter ; li & chevalier 2001 ; panaitescu & kumar 2002 ) .
notable examples include grb 970508 ( see frail , waxman & kulkarni 2000 for a different view ) and grb 011121 ( price et al . 2002 ) .
there is evidence for two types of grb ambient environments , with implications for their progenitors .
they are not immediately distinguishable because , at an age of a few days , the preshock wind density is comparable to an interstellar density . at earlier times , the density contrast is higher and the jet effects are less important .
the early afterglow observations are expected to provide a better handle on the nature of the ambient medium . in this letter
, we summarize the characteristics of the early afterglows expected in the wind model ( [ windmodel ] ) , and argue that the afterglow observations of grb 021004 are consistent with the source interacting with a wolf - rayet type wind ( [ grb021004 ] ) .
the strongest evidence for wind interaction comes from the initial slow decay of the r - band light curve and its prominent steepening around day 0.1 .
analytic light curves for the standard ism model are given in sari , piran & narayan ( 1998 ) , assuming a power - law electron energy spectrum with index @xmath3 and constant fractions of blast wave energy , @xmath6 and @xmath7 , going into nonthermal electrons and magnetic field respectively .
at any given time , the synchrotron spectrum is determined by a set of characteristic frequencies : the typical frequency @xmath0 , cooling frequency @xmath8 , and self - absorption frequency @xmath9 .
the light curve at any given frequency @xmath10 is determined by the characteristic times @xmath11 , @xmath12 and @xmath13 , when @xmath0 , @xmath8 and @xmath9 cross @xmath10 , and the critical time @xmath14 when @xmath0 and @xmath8 become equal .
the light curve of the ism model was extended to the wind case by cl00 ( see also panaitescu & kumar 2000 and granot & sari 2002 ) . in the wind model ,
the cooling and self - absorption frequencies are expected to be lower at early times than those in the ism model , because of a higher ambient density at small radii .
the cooling frequency has a strong effect on the early emission in the optical and the self - absorption frequency in radio . at the earliest times
, the optical flux may be dominated by the emission from the reverse shock front , but the forward shock front is not much fainter for wind interaction ( cl00 , eq . [ 58 ] ) , assuming that the two shocks have similar efficiencies for the production of synchrotron radiation .
once the reverse shock front has passed through the initial shell , which is expected to occur on a timescale similar to that of the gamma - ray burst , the reverse shock emission drops sharply because it is in the fast cooling regime . unless the reverse shock is `` refreshed '' by a continued flow , the decline is determined by off - axis emission that arrives at a later time ( kumar & panaitescu 2000 ) .
optical frequencies are expected to be between @xmath8 and @xmath0 , so the flux is @xmath15 and decays as @xmath16 ( kumar & panaitescu 2000 ) . to describe the forward shock emission , we use the characteristic times mentioned above , which were estimated by cl00 assuming @xmath17 .
we rescale the estimates to the r - band with @xmath18 hz ) , and obtain @xmath19 which is about an hour for standard parameters .
the parameter @xmath20 is the cosmological redshift , @xmath21 is the blast wave energy in units of @xmath22 ergs , and @xmath23 .
note that @xmath11 does not depend on the wind density @xmath24 directly , where @xmath25 , although the wind must be sufficiently dense that the transition from fast cooling to slow cooling occurs after @xmath11 .
it depends most sensitively on @xmath6 .
the condition on @xmath6 for the r - band break to occur between one minute and one hour is @xmath26 which covers a reasonable range , and has only a weak dependence on other parameters .
the transition to slow cooling occurs around the time @xmath27 which is typically later than @xmath11 in the r - band . here , @xmath28 .
the cooling frequency crosses the frequency @xmath29 at a time @xmath30 which is typically later than both @xmath11 and @xmath14 .
the self - absorption frequency @xmath9 is typically well below the optical , and can be ignored . before @xmath14 , the synchrotron electrons are in the fast cooling regime , and the flux peaks at the cooling frequency @xmath8 , so @xmath31 with @xmath32 where @xmath33 is the luminosity distance in units of 10 gpc .
the flux at the typical frequency @xmath0 is lower , and is given by @xmath34 which is independent of the wind density .
after the transition to slow cooling at @xmath14 , the flux peaks at @xmath0 rather than @xmath8 , and now @xmath35 ( eq . [ [ fnumax ] ] ) .
there is a general scaling for the late - time r - band flux after the break ( @xmath36 ) @xmath37 where @xmath38 is the break time at the r - band .
sources with later breaks tend to be brighter .
the gamma - ray burst grb 021004 was detected by the hete ii satellite ( shirasaki et al .
2002 ) and had an optical afterglow detected within minutes of the @xmath39-ray burst ( fox et al .
the early light curves of the only other afterglows detected at such early times , grbs 990123 ( akerlof et al .
1999 ) and 021211 ( fox et al .
2003a ) , have similar shapes , both showing a rapid initial decline followed by a slower decay , although grb 021211 is fainter by about 3 magnitudes at similar epochs . in both sources ,
the initial rapidly decaying emission is interpreted as coming from the reverse shock of grb ejecta running into a constant density medium ( sari & piran 1999 ; li et al .
2003 ; fox et al . 2003a
; wei 2003 ) .
the early optical afterglow of grb 021004 shows a different behavior : it has a very slow initial decay of @xmath40 , followed by a steepening around day 0.1 into approximately @xmath41 ( fox et al .
kobayashi & zhang ( 2003 ) interpreted the early afterglow data in terms of emission from a combination of reverse and forward shocks expanding into a constant - density medium .
fox et al .
( 2003b ) questioned this interpretation , and suggested instead a continued energy injection into the blast wave after the @xmath39-ray burst to explain the initial slow decay .
we propose that the slow decay is a natural consequence of the fast cooling ( @xmath42 ) expected in a wind model at early times and that the steepening is caused by the typical frequency @xmath0 passing through the optical band from above while still in the fast cooling regime .
we show that this interpretation , besides fitting the r - band light curve , is in a reasonable agreement with the radio data available at several frequencies .
the free parameters that appear in the wind model can be estimated analytically using the formulae given in the previous section .
our identification of the break time @xmath43 day in the r - band light curve yields , using eq .
( [ tm ] ) , the relation @xmath44 for a redshift of @xmath45 ( chornock & filippenko 2002 ) . for a cosmological model with @xmath46 , @xmath47 , and @xmath48
, this redshift corresponds to @xmath49 . at the break ,
the r - band flux is approximately @xmath50 mjy , which yields a second relation @xmath51 from eq .
( [ fnum_1 ] )
. the wind density @xmath52 does not enter into either of the two relations , which enables us to express @xmath6 and @xmath21 in terms of @xmath7 : @xmath53 to constrain the wind density @xmath52 , we note that the typical frequency @xmath0 decreases with time as @xmath54 .
it should cross the 8.46 ghz wavelength around day 140 .
this is much later than the time @xmath14 for transition to slow cooling , which occurs around @xmath55 for typical parameters .
therefore , we can use eq .
( [ fnumax ] ) to find the expected peak flux at 8.46 ghz .
the result is @xmath56 which for standard parameters is comparable to the r - band flux at the break @xmath38 .
the 8.46 ghz flux is observed at 598@xmath5733 @xmath58jy on day 5.7 .
if this flux is comparable to the peak flux at the time @xmath11 ( which is true if the observed frequency is close to or beyond the self - absorption frequency ; see cl00 ) , then one can use eq . ( [ c5 ] ) to provide a rough estimate for @xmath52 in terms of @xmath7 : @xmath59 the above analytic estimates assumed @xmath17 , and are rather crude . they do indicate that the optical and radio data may be fitted with a wind model with parameters not far from the standard values .
we now demonstrate that this is indeed the case using a numerical model .
the model treats synchrotron emission from a spherical ( trans-)relativistic blast wave propagating in an @xmath5 density medium , with the light travel time effects , synchrotron self absorption and cooling included .
it was previously applied to grb 980508 ( cl00 ) , among others . like grb 980508
, the decay of the r - band flux on the time scale of days and longer is relatively slow , with @xmath60 approximately ; the wiggles on the light curve of this source makes a precise determination of the decay slope difficult .
the slope implies that @xmath3 is close to 2 , although the exact value is somewhat uncertain .
we pick @xmath61 , which corresponds to a decay slope of @xmath62 in the optical .
after some experimentation , we find a solution that fits the r - band and radio data reasonably well with the following set of parameters : @xmath63 , @xmath64 , @xmath65 , and @xmath66 , which corresponds to a wind mass loss rate of @xmath67 m@xmath68 yr@xmath69 ( assuming a nominal wind speed of @xmath70 km s@xmath69 ) . now all the parameters are approximately determined ( within a factor @xmath4 ) because of the inclusion of self - absorption .
the fits are shown in panels ( a ) and ( b ) of fig . 1 .
we did not attempt to fit the bumps on the r - band light curve ; they have been interpreted as arising from either late energy injections or inhomogeneities in the ambient density ( lazzati et al .
2002 ; nakar , piran & granot 2003 ; heyl & perna 2003 ) .
the bumps introduce some uncertainty to the model parameters we obtained .
radio emission was detected at 4.86 , 8.46 , 15 , 22.5 and 86 ghz at various times ( frail & berger 2002 ; berger , frail & kulkarni 2002 ; pooley 2002b , c ; bermer & castro - tirado 2002 ) . one upper limit each exists at 15 , 232 and 347 ghz ( pooley 2002a ; bermer & castro - tirado 2002 ; wouterloot et al .
the flux measurement of [email protected] mjy at 86 ghz flux at an average time of 1.5 days is particularly interesting .
this flux is three times higher than the r - band flux at the break around day 0.1 .
it presents a problem to the identification of the break as @xmath11 in a _ constant density _
medium when the cooling frequency has already passed the r - band from below ( kobayashi & zhang 2002 ) . in such a case , the maximum fluxes at lower frequencies should be the same as that of the r - band at the break , namely about 0.83 mjy , which is well below the 86 ghz measurement .
this discrepancy was also noted by pandey et al .
( 2002 ) .
the relatively high 86 ghz flux is not a problem for our model , where the transition to slow cooling occurs around day 2 , much later than day 0.1 . in a wind model
, the flux can be much higher in the radio ( broadly defined to include millimeter and sub - millimeter wavelengths ) , particularly at early times when the cooling frequency is expected to be in the spectral region .
this behavior shows up clearly in panel ( b ) of fig . 1 , where the peak fluxes at the three highest frequencies are all above 3 mjy .
such high fluxes are naturally expected in a wind model but not in an ism model , as emphasized by panaitescu & kumar ( 2000 ) .
however , it is difficult to make a strong case for wind interaction based on a single data point at 86 ghz .
a stronger case can be made if the 8.46 ghz flux starts to decline around 100 days , when the typical frequency @xmath0 is expected to pass through the frequency from above .
this expectation needs to be modified in the case of an early jet break .
x - ray afterglows are observed with chandra at two epochs .
the first epoch started about 0.87 days after the burst and lasted for 88.1 ksec ( sako & harrison 2002a ; fox et al .
2003b ) . within this epoch ,
the x - ray afterglow has a power - law spectrum , with index @xmath71 , and decays roughly as a power - law , with index @xmath72 .
both are consistent with our model , where the cooling frequency @xmath8 around day 1 is well below the x - ray band , and @xmath73 and @xmath74 are predicted to be 1.05 and 1.075 , respectively , for @xmath61 . the fact that the temporal decay slope in the r - band , @xmath75 , is close to 1 indicates that the cooling frequency is below the r - band as well around this time , which is in agreement with the optical spectral index of @xmath76 determined by pandey et al .
( 2002 ) and @xmath77 by matheson et al .
( 2003 ) in the absence of a substantial host galaxy extinction ( see also bersier et al . 2003 ; holland et al .
our best fit model yields an x - ray flux of @xmath78 erg @xmath79 s@xmath69 between 2 and 10 kev at day 1.4 , close to the middle of the first observing epoch .
it is lower than , but within a factor of two of , the mean flux of the entire epoch @xmath80 erg @xmath79 s@xmath69 ( sako & harrison 2002a ) .
we therefore conclude that the wind model is consistent with the first epoch of x - ray observations .
the second epoch of x - ray observations started 52.23 days after the burst and yielded a 2 - 10 kev flux of [email protected]@xmath81 erg @xmath79 s@xmath82 ( sako & harrison 2002b ) .
the flux implies a decay slope between the two epochs of approximately @xmath83 , which is steeper than that predicted in our spherical model .
the steepening may be due to a jet break between the two epochs of observation , which was suggested to have occurred around day 7 by pandey et al .
( 2002 ) and holland et al .
( 2002 ) based on their interpretation of the ( wiggly ) r - band light curve .
the jet break , if exists , should show up in a well - sampled radio light curve as well .
alternatively , the steepening could be due to a steepening of the energy distribution of nonthermal electrons well above the minimum energy of the electrons accelerated at the shock front ( e.g. , li & chevalier 2001 ; panaitescu & kumar 2002 ) .
the question of a jet break is related to the energy in the source .
the energy we find in a spherical model for the afterglow , @xmath65 , is comparable to the isotropic burst energy in @xmath39-rays , @xmath84 ergs ( bloom , frail , & kulkarni 2003 ) .
pandey et al .
( 2002 ) find , in a fit to the optical data through day 21 , that there is a break in the light curve at @xmath85 days .
bloom et al .
( 2003 ) interpret this as a jet break ; the correction for collimated flow reduces the @xmath39-ray energy by a factor of @xmath86
. however , fig .
1 shows that the variability in the light curve and the late light curve points make a clear designation of the jet break difficult . in a wind model
, the jet break evolves slowly ( kumar & panaitescu 2000 ) , which makes any determination of a jet break from afterglow data more uncertain .
the mass loss rate we deduced is typical for a wolf - rayet type wind .
there are other indications that grb 021004 may be interacting with a wolf - rayet type wind .
wolf - rayet winds are thought to be clumpy ( e.g. , hamann & koesterke 1998 ) , and the clumpiness may provide an explanation for the prominent bumps on the r - band light curve of grb 021004 ( lazzati et al .
2002 ; nakar et al . 2002 ; heyl & perna 2003 ) .
in addition , there are multiple absorption components in the spectrum of the afterglow , separated by speeds up to 3000 km s@xmath69 .
these components could come from substructures in a wolf - rayet wind ( mirabal et al .
2002 ; schaefer et al .
interestingly , grb 990510 , which is best modeled by interaction with a constant density medium ( cl00 ; panaitescu & kumar 2002 ) , shows a smoothly evolving optical afterglow ( stanek et al .
1999 ) .
grb 021004 differs from the other two grbs with detected early afterglows ( grbs 990123 and 021211 ) in several ways : it has a slow decay in the r - band light curve followed by a steepening rather than a steep decline followed by a flattening ( which occurs at a much earlier time than the break in grb 021004 ) , a higher optical flux at late times after the break , and a bright , long - lived radio afterglow .
the early emission from grb 990123 was convincingly interpreted as coming from the reverse shock of a blast wave expanding into an ism ( sari & piran 1999 ; mszros & rees 1999 ) , and the late - time afterglow data are consistent with an ism model ( e.g. , panaitescu & kumar 2002 ) .
the afterglow of grb 021211 resembles that of grb 990123 , and was interpreted similarly ( fox et al .
2003a ; li et al .
2003 ; wei 2003 ) .
the rate of initial decline in the r - band flux of approximately @xmath87 ( grb 990123 ) or shallower ( grb 021211 ) is difficult to reproduce in the reverse shock of a wind model .
the available data on the three early afterglows therefore appear to point to two types of grb ambient environments : ism and a stellar wind .
these two types of environments are also inferred in the detailed modeling of a larger number of afterglows at later times ( cl00 ; panaitescu & kumar 2002 ; price et al .
2002 ) .
the evidence for wind and constant density environments in the early afterglow evolution supports the hypothesis that these represent different progenitors ( chevalier & li 1999 ; cl00 ) because a small radial distance from the star is explored ( @xmath88 cm ) where a freely expanding wind is expected .
a prediction of the cl99 scenario is that wind interaction should be correlated with supernova light ; this hypothesis was supported by grb 011121 ( price et al .
2002 ) . however , grb 021211 , which has an early afterglow indicating ism interaction , shows some evidence for a supernova - like bump in the light curve ( fruchter et al .
2002 ; testa et al .
grb 020405 is another case of apparent ism interaction and a supernova - like bump ( berger et al .
spectra confirming the supernova nature of light curve bumps , as well as detailed afterglow observations , are needed to clarify the situation .
long term monitoring of radio afterglows will be crucial in testing the wind model of early afterglows ( by examining the evolution of @xmath0 ) , and in determining the wind density .
another prediction of the wind model is that the early optical emission before the break ( @xmath89 ) should have the spectrum @xmath90 , which is a flatter spectrum than is typically observed in optical afterglows .
in addition , the break is chromatic , occurring at a later time for a longer wavelength , which can be tested with densely sampled ir observations . | wind - interaction models for gamma - ray burst afterglows predict that the optical emission from the reverse shock drops below that from the forward shock within 100s of seconds of the burst .
the typical frequency @xmath0 of the synchrotron emission from the forward shock passes through the optical band typically on a timescale of minutes to hours .
before the passage of @xmath0 , the optical flux evolves as @xmath1 and after the passage , the decay steepens to @xmath2 , where @xmath3 is the exponent for the assumed power - law energy distribution of nonthermal electrons and is typically @xmath4 .
the steepening in the slope of temporal decay should be readily identifiable in the early afterglow light curves .
we propose that such a steepening was observed in the r - band light curve of grb 021004 around day 0.1 .
available data at several radio frequencies are consistent with this interpretation , as are the x - ray observations around day 1 .
the early evolution of grb 021004 contrasts with that of grb 990123 , which can be described by emission from interaction with a constant density medium . |
the standard model ( sm ) of electroweak interactions is in excellent agreement with the experimental data and has explained many features of particle physics throughout the years . despite its success there are some points in the model as , for instance , the enormous range of masses between the lightest and heaviest fermions and other peculiarities that could be better explained at a deeper level assuming the introduction of new fields or symmetries .
recently the atlas and cms collaborations reported the observation of a new boson with a mass @xmath2gev which is suspected to be the sm higgs boson . the current data on this boson diphoton event rate exhibit the signal strength about ( 1.5 - 2 ) larger than the one expected for the standard model higgs boson .
there are already many sm extensions trying to explain this possible enhancement of the @xmath5 decay .
in particular , the increase of this decay rate is natural in the context of a 3 - 3 - 1 model @xcite and its alternative version with exotic leptons@xcite , due to the presence of an extra charged vector boson and a doubly charged one as discussed in ref.@xcite .
this class of models predicts interesting new physics at tev scale @xcite and addresses some fundamental questions that can not be explained in the framework of the standard model@xcite .
these models also contain a set of fundamental scalar bosons , with many parameters and clearly suffering from the problems of naturalness and hierarchy @xcite .
however in refs.@xcite it was suggested that the gauge symmetry breaking in some versions of the 3 - 3 - 1 model @xcite could be promoted dynamically , because at the scale of a few tevs the @xmath6 coupling constant becomes strong and the exotic quark @xmath1 that appears in the model forms a condensate breaking @xmath7 to the sm electroweak symmetry .
this is a very interesting feature and peculiar to this class of models .
unfortunately the sm gauge symmetry still remains intact , and the nice characteristics of the model could be missed with the introduction of a elementary scalar field in order to break the electroweak gauge symmetry , leading to an unpleasant system of composite and elementary fields responsible for the gauge symmetry breaking . in ref.@xcite the full realization of the dynamical symmetry breaking of a @xmath8 extension of the sm @xcite was explored .
this was accomplished assuming the gauge symmetry @xmath9 , where the electroweak symmetry is broken dynamically by a technifermion condensate generated by the @xmath10 technicolor ( tc ) gauge group , i.e. besides the exotic @xmath1 quark condensate and respective composite scalar , we do now have another composite scalar boson formed by @xmath10 technifermions . this symmetry breaking
also occurs when we exchange the @xmath10 group by the @xmath11 group , as well as when we deal with different technifermion representations @xcite . in 3 - 3 - 1 models where the gauge symmetry breaking is promoted by elementary scalar fields ,
the many parameters in the scalar potential can be variated in a large range leaving space to scape , up to now , to the lhc experimental constraints . however , in the case where the gauge symmetry breaking it totally dynamical , once we describe the possible dynamics of the theory , we may already have some limitation on the possible models .
the study of possible constraints in this class of models is the main motivation of this work .
we compute the effective potential for composite operators of a class of 3 - 3 - 1 models where the gauge symmetry is dynamically broken , with the main purpose of determining the composite scalar masses .
if , for instance , we consider @xmath12 , we verify that it is quite difficult to generate a scalar boson mass of @xmath13 gev , assuming that it is a composite scalar that has been observed at the lhc , obeying , at the same time , the lower limit on the @xmath14 .
the composite scalar system of these 3 - 3 - 1 models have a mixing related to the @xmath15 and @xmath14 mixing , which is present in the exotic @xmath1 quark self - energy .
this mixing will appear in the calculation of the effective potential for composite operators @xcite , which , when minimized , supply the physical scalar masses , and it is important to know its amount because the scalar masses may be modified by this effect .
there are other possible contributions to this mixing , that would appear in an extended theory necessary to explain the fermion masses , whose spectrum has no explanation in any dynamical symmetry breaking model up to now .
however these corrections are expected to be small compared to the ones that we discuss here , and it will become clear that already at this level some versions of these models may be excluded by the recent lhc data .
the distribution of our paper is the following : in section ii we present the main aspects of the @xmath16 models . in section iii
we discuss the mixing of the neutral gauge boson system , its relation to the @xmath1 quark dynamical mass and present the self - energies that will be used in the effective action calculation .
the mixing in the @xmath1 quark self - energy will be responsible by the coupling between the composite scalars ( @xmath17 , @xmath18 ) , associated respectively with the breaking of symmetries @xmath8 and @xmath19 in the effective potential .
section iv contains the calculation of the effective action and in section v we compute numerically the scalar boson masses for different tc groups .
finally , in section vi we draw our conclusions .
on the lines below we describe the main features of the models , which are similar to that proposed in ref.@xcite , the fermionic content has the following form & & q_3l = ( c t + b + t ) _ l~(*1 * , * 3 * , 2/3 ) + + & & t_r~(*1 * , * 1 * , 2/3),b_r~(*1*,*1*,-1/3 ) + + & & t_r~ ( * 1*,*1 * , 5/3 ) & & q_l = ( c d + u + d ) _ l~(*1 * , * 3^ * * , -1/3 ) + + & & u_r ~ ( * 1*,*1 * , 2/3 ) , d_r~(*1 * , * 1 * , -1/3 ) + + & & d_r~ ( * 1 * , * 1 * , -4/3 ) where @xmath20 is the family index and we represent the third quark family by @xmath21 . in these expressions
@xmath22 , @xmath23 or @xmath24 denote the transformation properties under @xmath16 and @xmath25 is the corresponding @xmath26 charge .
the leptonic sector includes besides the conventional charged leptons and their respective neutrinos the charged heavy leptons @xmath27@xcite . & & l_al = ( c _ a + l_a + e^c_a ) _ l~(*1*,*3 * , 0 ) where @xmath28 is the family index and @xmath29 transforms as triplets under @xmath30 .
moreover , we have to add the corresponding right - handed components , @xmath31 and @xmath32 .
the fermionic content associated with the tc sector has the form & & _
1l = ( c u_1 + d_1 + u ) _ l~(*n_tc * , * 3 * , 1/2 ) + + & & u_1 r ~ ( * n_tc * , * 1 * , 1/2),d_1 r ~(*n_tc * , * 1*,-1/2 ) + & & u_r~ ( * n_tc*,*1 * , 3/2 ) , & & _ 2l = ( c d + u_2 + d_2 ) _ l~(*n_tc * , * 3^ * * , -1/2 ) + + & & u_2 r ~ ( * n_tc * , * 1 * , 1/2),d_2 r~(*n_tc * , * 1*,-1/2 ) + & & d_r~ ( * n_tc*,*1 * , -3/2 ) . [ 3 ] where @xmath33 and @xmath34 label the first and second techniquark families , @xmath35 and @xmath36 can be considered as exotic techniquarks making an analogy with quarks @xmath1 and @xmath37 that appear in the ordinary fermionic content of the model .
the model is anomaly free if we have equal numbers of triplets and antitriplets , counting the color of @xmath38 .
therefore , in order to make the model anomaly free two of the three quark generations transform as @xmath39 , the third quark family and the three leptons generations transform as @xmath40 .
it is easy to check that all gauge anomalies cancel out in this model , in the tc sector the triangular anomaly cancels between the two generations of technifermions . in the present version of the model the technifermions are singlets of @xmath38 .
as pointed out in refs.@xcite , one interesting feature of the versions@xcite of 3 - 3 - 1 models is the following relationship among the coupling constants @xmath41 and @xmath42 associated to the gauge group @xmath8 @xmath43 where @xmath44 , @xmath45 and @xmath46 is the electroweak mixing angle . according to the discussion presented in@xcite ,
it is precisely this feature of the model that allows the gauge symmetry breaking of this version of the @xmath8 model to the sm symmetry , because at the scale of a few tevs the @xmath6 coupling constant becomes strong as we approach the peak existent in eq.([equ1 ] ) .
therefore , in the model described in this section the exotic quark @xmath1 will form a condensate breaking @xmath8 to the electroweak gauge symmetry , while the sm gauge symmetry will be broken dynamically by a technifermion condensate . in order to compute the effective action generated for the composite scalar bosons resulting from the two symmetry breaking stages , in the next section we discuss the @xmath1 quark self - energy , that is related to the mixing between the standard model neutral gauge boson z with the z boson .
in the models that we consider here there is a mixing between the standard model neutral gauge boson z with the z boson , the mass eigenstates are@xcite & & z_1 = z- z + & & z_2 = z+ z where the mixing angle(@xmath47 ) is given by@xcite = , in this case ( @xmath48 ) represents the sm neutral boson and ( @xmath49 ) corresponds to the additional 3 - 3 - 1 heavy neutral boson .
therefore , assuming this mixing we can write in the euclidean space the following linearized gap equation ( @xmath50 ) for the @xmath1 quark : @xmath51}\frac{1}{[(p- k)^2 + m^2_{z ' } ] } + \nonumber \\ & & \,\ , a\sin\theta\!\!\int \ !
+ \mu^2_{x}]}\frac{1}{[(p - k)^2 + m^2_{z } ] } \nonumber \\
\label{10}\end{aligned}\ ] ] where @xmath52 @xmath53 is the energy scale where the @xmath26 interaction becomes sufficiently strong to break dynamically the @xmath8 to @xmath19 , @xmath54 is the @xmath55 coupling constant , @xmath56 and @xmath57 are respectively @xmath55 charges attributed to the chiral components of the exotic quark @xmath1 . besides the condensate and composite states ( scalar and pseudoscalar ) associated to eq.([10 ] )
, we have similar entities due to the @xmath58 group condensation at the scale @xmath59 , generated by a non - trivial technifermion self - energy ( @xmath60 ) . as discussed in ref.@xcite the technifermions multiplets , @xmath61 and @xmath62 described in eq.([3 ] ) , lead to the formation of composite scalar bosons ( @xmath63 and @xmath64 ) that are equivalents to the set of fundamental scalar fields , @xmath65 and @xmath66@xcite , so , in order to obtain a structure of the scalar potential similar the described in@xcite , we will assume that = , [ 35 ] this normalization results from the fact that @xmath67 ( as we shall describe in eq.([28 ] ) ) , and from this it is possible to verify that ^2 = ^2_1 + ^2_2 , [ 38 ] once @xmath68 , and @xmath69 , typical of the two technifermion generations of eq.([3 ] ) .
the @xmath70 are the technipions decay constants that can be computed through the linearized pagels and stokar relation@xcite f^2 _ [ eq34 ] whereas the pseudoscalar decay constant associated to t quark self - energy , @xmath71 will be written as f^2 _ _ .
[ eq341 ] to compute the effective potential for composite operators @xcite we need to know the self - energies of the strongly interacting fermions : the @xmath1 quark and the fermions with @xmath11 charges .
eq.([10 ] ) has two possible solutions , and in the program developed in refs.@xcite it was verified that the solution falling slowly with the momentum is the dominating one if suitable new interactions are assumed to be relevant at the scale of the ( uv ) cutoff . in their case
the gauge boson mass integrals receive significant contributions from a very large range of loop momenta and the sm gauge boson masses @xmath72 and @xmath73 turn out to be of similar magnitude when compared to the top quark mass .
this is exactly the situation that we have in the approach proposed to promote the gauge symmetry breaking of @xmath74 to the electroweak symmetry , where at the scale of a few tevs the @xmath55 coupling constant becomes strong as we approach the peak existent in eq.([equ1 ] ) . in the ref.@xcite after the numerical calculation of @xmath75 , it was found that the magnitudes of @xmath76 and @xmath75 are the same order .
therefore , considering the above comments we will assume that the solution of eq.([10 ] ) is giving by _
t(p^2 ) _t ( p^2)(1 - h ( ) ( ) ) , + [ 18 ] and is the one that will be used for the @xmath1 quark self - energy to determine the effective potential ( @xmath77 ) . to write eq.([18 ] )
we defined the following quantities : @xmath78 where @xmath79 , @xmath80 and h ( ) = ( 1 - + ) .
[ hi ] in the case of fermions with @xmath11 charges the self - energy will be given by ( p^2 ) ~_tc ^- , [ eq12 ] where @xmath81 is the @xmath11 characteristic scale of mass generation , = 3c/16 ^ 2 b [ eq12a ] and @xmath82 $ ] where @xmath83 are the casimir operators for fermions in the representations @xmath84 and @xmath85 that condense in the representation @xmath86 , @xmath87 for the @xmath11 group with @xmath88 flavors , @xmath89 is the coupling constant for which we assume the expression ^2(k^2)= , [ eq4 ] where , @xmath90 is an infrared dynamical gauge boson mass , whose phenomenologically preferred value is @xmath91 @xcite , and we will set @xmath92 . note that , using the above coupling constant , we are assuming that non - abelian gauge theories generate dynamical masses for their gauge bosons @xcite . as a consequence
it is expected that confinement should be necessary to generate non - trivial fermionic self - energies @xcite , and the expression for the self - energy is the one of eq.([eq12 ] ) , as discussed at length in ref.@xcite .
the main features of eq.([eq12 ] ) , is that it causes the decoupling of heavier degrees of freedom in models where there is an interaction connecting different fermionic families @xcite , it leads to the deepest minimum of energy , with a vacuum expectation value proportional to @xmath93 @xcite , it is the only self - energy able to naturally explain fermion masses as heavy as the top quark @xcite , as well as the unique possible form of solution that may generate a light composite scalar boson @xcite . in the self - energies that we discussed above the characteristic scales @xmath94 and @xmath81 have not been determined up to now .
however they should be constrained by the value of the @xmath8 gauge boson masses . in order to do
so we notice that for @xmath95 , it is possible to show that@xcite , [ 19 ] assuming the result described in ref.@xcite we obtain the masses m^2_z & = & ( f^2__1 + f^2__2 ) + & = & f^2 _ , [ 191 ] m^2_z = f^2 _ . [ 20 ] . with these masses and eq.([19 ] ) , we can write eq.([18 ] ) in the form _ t ( p^2 ) _t ( p^2)(1 + a ( ) ( ) ) , [ 21 ] where for @xmath11 @xmath96 the @xmath97 decay constant is related to the sm vacuum expectation value(vev ) through f^2 _
= ( f^2__1 + f^2__2 ) = v^2 = = ( 246 gev)^2 . and in the case of @xmath1 quark self - energy , since there are no evidences for the @xmath14 boson , we just assume @xmath98
. equations ( [ eq12 ] ) and ( [ 21 ] ) are the main ingredients to compute the effective action for the model described in section ii .
the effective potential for composite operators @xcite is a function of the green s functions of the theory , in particular it can be written as a function of the complete fermion ( @xmath99 ) and gauge boson ( @xmath37 ) propagators as v(s , d ) & = & - i tr ( s_0 ^ -1 s - s_0 ^ -1 s + 1 ) + & & + v_2 ( s , d ) , [ eq5 ] where @xmath100 ( and @xmath101 ) stands for the bare fermion ( gauge boson ) propagator and @xmath102 is the sum of all two - particle irreducible vacuum diagrams .
the physically meaningful quantity that we must compute is the vacuum energy density given by _
v = v(s , d ) - v(s_0 , d_0 ) , [ eq8 ] where we are subtracting the symmetric part of the potential from the potential that admits condensation in the scalar channel , that is denoted by @xmath103 and is a function of the perturbative propagators ( @xmath100 and @xmath101 ) . the vacuum energy density ,
if we remove all indices and integrations , can be written as @xcite _ v & = & - i tr ( s_0 ^ -1s - s_0 ^ -1s + 1 ) + i tr ( s - s_0 ) + & + & i tr ( s s - s_0 s_0)d .
[ eq10 ] the self - energies described by eqs.([eq12 ] ) and ( [ 21 ] ) enter into the definition of @xmath99 .
how the effective action for composite scalar bosons emerge from eq.([eq10 ] ) in the case of a dynamically broken gauge theory including the kinetic term has been detailed by cornwall and shellard @xcite , and is also discussed in refs.@xcite . here
we will skip lengthy details and follow closely the work of ref.@xcite , where it was shown that the effective action generated for a tc model could be written in the following way _ tc = d^4x .
[ 23 ] the effective scalar field @xmath104 acts like a dynamical effective scalar field with anomalous dimension @xmath105 , is related to the bilinear self - energy @xmath106 , it is seem as a variational parameter , and the kinetic term for our effective theory is obtained inserting @xmath107 in the effective action and expanding around @xmath108 @xcite . in the above equation z__tc , [ eq21 ] and & & _ 4_tc= ( + ) + & & _ 6_tc = - .
[ eqa7 ] in this expressions , @xmath109 , @xmath88 denote the number of technifermions , @xmath110 has been defined in eq.([eq12a ] ) and is calculated for the respective tc representations . in eq.([23 ] ) a term proportional to @xmath111 does not appear because we assume that the self - energies are exact solutions of the linearized gap equations @xcite , also odd terms in @xmath18 do not appear because we do not have current fermion masses . the constant @xmath112 arises when the contribution of the kinetic term is included in the calculation of the effective action@xcite , and this acts as a normalization constant .
this contribution is important in our calculation because it will give the correct normalization of the effective fields , @xmath113 and @xmath114 , as discussed in the refs.@xcite . in terms of these fields
we can also write the decay constants for the tc and @xmath1 fermions as f^2 _
= [ 28 ] . while for the self - energy , @xmath71 , we obtain the following relation f^2 _ = [ 29 ] , and in this case @xmath115
. we can now present the contribution to the effective action due to the composite scalar boson formed by the strong interaction of the exotic @xmath1 quark .
below we show the @xmath116 and @xmath117 terms of the effective potential @xmath118 and the corrections @xmath119 assuming the mixture in eq.([18 ] ) and the comments leading to eq.([21 ] ) : & & _ t = d^4x[eq32 ] where we identify & & z__t 8 ^ 2 a , _ 4 t ( + ) + & & _ 6 t - , _ 4 t = ( + ) + & & _ 6 t = - + & & _ 4 = ( ) + & & , in these expressions we assumed the existence of just one exotic quark that condenses in the most attractive channel(mac ) @xcite . in order to reproduce a standard scalar effective field theory we introduce in our effective lagrangian the normalized fields ( x ) = z__tc^-1/2 ( x ) , _ _
t(x ) = z__t^-1/2 _ _ t(x ) . now ,
considering the eqs.([35 ] ) , where we see that the field @xmath18 actually represent two fields ( @xmath63 and @xmath64 ) , and adding eq.([23 ] ) to eq.([eq32 ] ) we can write down the full effective action in terms of the normalized fields @xmath120 and @xmath121 ( composed by @xmath122 and @xmath123 ) : ( _ t , ) = & & d^4x [ efp ] where the normalized couplings for the composite fields , @xmath114 , @xmath122 and @xmath123 , are given respectively by & & ^r_4_t = ( + ) , ^r(a)_4_tc = ( + ) + & & ^r(b)_4_tc = ( + ) , ^r_6_t = - + & & ^r(a)_6_tc = - , ^r(b)_6_tc = - . [ eq - coupl ]
to compute the scalar masses from the effective lagrangian described in eq.([efp ] ) we use m^2___i = |____i=_i_min .
where @xmath124 , and @xmath125 label the first and second techniquark families . after neglecting terms of higher order
when substituting the minimum value in the potential we obtain m^2___i 2^r(a)_4_tc ( ) + 2^r(b)_4_tc ( ) .
[ eq41 ] assuming the above equation , with couplings @xmath126 , @xmath127 , @xmath128 and @xmath129 defined by eq.([eq - coupl ] ) , we finally can make a scan in the parameter space @xmath130 $ ] limiting the higgs mass to the range @xmath131 , in order to exclude possible candidates for tc models . note that for this purpose we assume that the atlas and cms are indeed observing a new scalar boson with a mass @xmath2gev , and consider also the strong limit on the @xmath132 mass announced by these collaborations@xcite .
the extra @xmath132 boson is predicted in many extensions of the standard model at the tev mass scale , as in the sequential standard model @xmath133@xcite , with sm like couplings . with lhc data at @xmath134 ,
recently the atlas and cms collaborations placed strong constraints on the mass of these particles@xcite .
these constraints depend on the knowledge of the coupling of this boson with sm fermions . in the case of the @xmath135 model with sm like couplings
the @xmath132 mass can be excluded below @xmath136 tev .
this limit can also be taken as a lower limit on the @xmath132 mass of the models discussed here . in this particular case , if @xmath137 the energy scale @xmath138 should be limited to @xmath139 . in fig.[fig - a ]
we present the allowed region of parameters for the @xmath140 , @xmath141 and @xmath142 cases , with a scalar composite
higgs " mass range @xmath131 .
the solid black line corresponds to the lower limit on @xmath138 , and from this figure we verify that if @xmath4 the model barely survives the confrontation with the lhc data . and @xmath138 , for a composite scalar mass range @xmath131 in the case of @xmath140 , @xmath141 and @xmath142 tc groups .
the region in red corresponds to @xmath140 , the blue region to @xmath141 and the yellow region to @xmath142 . in this figure we assumed @xmath143 for all groups .
the solid line corresponds to the lower limit on @xmath144 ( i.e. @xmath145 ) . ] in order to complement the analysis , in the fig.[fig - b ] we show the interval of parameters in the @xmath141 case assuming @xmath146 and @xmath147 . as can be noticed in the fig.[fig - b ] , the range of values that the parameters @xmath59 and @xmath138 can assume decreases when the number of technifermions is increased .
the @xmath148 dependence on the kinetic term is important and this one is transferred to the @xmath149s couplings and results in a decrease of the scalar mass .
we can write the eq.([eq41 ] ) in the following approximate form m___i_1 _ _ tc + _ 2 _ _ x where @xmath150 and @xmath151 are constants .
the increase of the number of fermions implies in an increase of @xmath152 , then to keep the interval to the higgs mass @xmath131 we observe a decrease in the area @xmath130 $ ] in the parameter space .
this effect is the same that appears in the normalization of the bethe - salpeter wave function discussed in ref.@xcite . and
@xmath138 , for a composite scalar mass range @xmath131 in the @xmath141 case : in this figure the blue region corresponds to @xmath143 , the light blue to @xmath146 and magenta to @xmath147 . ] in the fig.[fig - c ] we show the interval of parameters in the @xmath140 case assuming @xmath153 ( in green ) , and again we verify that @xmath140 case with this number of fermions can be ruled out in confrontation with the lhc data . and
@xmath138 , for a composite scalar mass range @xmath131 in the @xmath140 case : the small green region corresponds to @xmath153 , and the solid line corresponds to the lower limit on @xmath138 . ]
in this work we computed an effective action for the composite scalar boson system , @xmath114 , @xmath63 and @xmath64 , formed by the fermions and technifermions @xmath154 , @xmath155 and @xmath156 described in the eqs.(1 ) and ( 3 ) .
we include in this calculation the kinetic term of the effective theory .
this term is important because it provides a normalization factor for the effective scalar boson lagrangian .
the effective lagrangian is then normalized in order to reproduce a standard scalar effective field theory , leading to a non - trivial set of scalar self - couplings . from this lagrangian
we can determine the scalar boson masses of the theory . to compute the effective action for the model described in the section
ii , we firstly determined the correction to the @xmath1 quark self - energy that result from the mixing between the standard model neutral gauge boson @xmath15 with the @xmath14 boson .
we show that this correction is the responsible by introduce the coupling between the composite scalars , @xmath17 and @xmath157 associated respectively to the @xmath1 quark and to the technifermions . in section iii
we discussed the self - energies used to compute the effective potential ( @xmath77 ) , where we assumed that the interaction @xmath55 plays a role analogous to the ultraviolet dynamical symmetry breaking program proposed in ref.@xcite .
we also assume a tc self - energy that decays slowly with the momentum whose origin is due to the introduction of confinement in the gap equation , as discussed in ref.@xcite , that is typical of the gauged nambu - jona - lasinio ( njl ) type of models where the anomalous dimension is @xmath158 .
note that it is hardly possible to generate light scalar composite bosons without this particular choice @xcite .
we finally determined an effective scalar boson lagrangian and from it we obtained the scalar boson masses associated to these models .
as already discussed in the introduction , the models that we consider here are interesting due to the particular form of their anomaly cancellation , due to the fact that it is one example of model where we have a naturally strong abelian theory at the tev scale , capable of producing a dynamical symmetry breaking of the model to the sm symmetry , whereas the sm gauge symmetry is broken by a tc condensate . within this class of models
we can also explain the larger decay rate of the @xmath13 gev boson into photons that is observed by the lhc experimental groups , caused by the presence of extra charged vector boson(@xmath159 , @xmath160 ) @xcite .
the only ingredients of our calculation are the self - energies of the strongly interacting theories . from these ones
we can compute the scalar masses and the neutral gauge boson masses . assuming that the @xmath13 gev observed at the lhc is a scalar boson and using their lower limit in the @xmath14 mass
, we compare the experimental results with the mass values that we obtained for these quantities .
we verified that the models are strongly constrained , showing that it is rather difficult to have light scalar composites and at the same time to generate masses for neutral gauge bosons where one of them ( the @xmath14 ) is quite heavy .
it is possible that in models with the presence of fundamental scalar bosons such difficulty is not present due to the many parameters that can be adjusted in the scalar potential , however , in this case , we may also foresee that this adjustment may lead to unnatural values of the coupling constants , unless some discrete symmetries are introduced by hand in order to avoid undesirable terms in the scalar potential .
this research was partially supported by the conselho nacional de desenvolvimento cientfico e tecnolgico ( cnpq ) .
atlas collaboration , g. aad _
et al . _ ,
b * 716 * , 1 ( 2012 ) ; cms collaboration , s. chatrchyan _
et al . _ ,
b * 716 * , 30 ( 2012 ) .
f. pisano and v. pleitez , phys .
rev . d*46 * , 410 ( 1992 ) .
p. h. frampton , phys .
lett . * 69 * , 2889 ( 1992 ) .
v. pleitez and m.d .
tonasse , phys .
rev . d*48 * , 2353 ( 1993 ) .
a. alves , e. ramirez barreto , a.g .
dias , c.a .
de s.pires , f.s .
queiroz and p.s .
rodrigues da silva , phys .
d * 84 * , 115004 ( 2011 ) ; idem , hep - ph/1207.3699v2 .
alex g. dias , c.a . de s.pires , v. pleitez and p.s .
rodrigues da silva , phys . lett . * b621 * , 151 ( 2005 ) ; alex g. dias , a. doff , c. a. de s. pires and p.s .
rodrigues da silva , phys . rev .
* d72 * , 035006 ( 2005 ) ; alex gomes dias , phys . rev . * d71 * , 015009 ( 2005 ) ; alex g. dias , j.c . montero and v. pleitez , phys .
b637 * , 85 ( 2006 ) ; alex g. dias and v. pleitez , phys . rev . *
d73 * , 017701 ( 2006 ) , a. doff , c. a. de s. pires and p. s. rodrigues da silva , phys . rev . * d74 * , 015014 ( 2006 ) .
a. doff and f. pisano , mod .
a15 * , 1471 ( 2000 ) .
c.a de s. pires and o. p. ravinez , phys .
rev . * d58 * , 035008 ( 1998 ) ; a. doff and f. pisano , mod .
a14 * , 1133 ( 1999 ) ; a. doff and f. pisano , phys . rev . *
d63 * , 097903 ( 2001 ) .
l. susskind , phys .
d * 20 * , 2619 ( 1979 ) g. t hooft , naturalness , chiral symmetry and spontaneous chiral symmetry breaking " , _ nato adv .
study inst .
* 59 * , 135 ( 1980 ) .
prasanta das and pankaj jain , phys . rev . * d 62 * , 075001 ( 2000 ) .
a. doff , phys . rev . *
d 76 * , 037701 ( 2007 ) .
a. doff , phys . rev . *
d 81 * , 117702 ( 2010 ) .
a. doff and a. a. natale , int
a * 27 * , 1250156 ( 2012 ) .
h. pagels and s. stokar , phys .
rev . d*20 * , 2947 ( 1979 ) .
j. m. cornwall , r. jackiw and e. tomboulis , phys .
rev . d*10 * , 2428 ( 1974 ) .
daniel ng , _ phys .
d _ * 49 * , 4805 ( 1994 ) ; hep - ph/9212284 .
j. carpenter , r. norton , s. siegemund - broka and a. soni , phys . rev
. lett . * 65 * , 153 ( 1990 ) .
j. d. carpenter , r. e. norton and a. soni , phys .
b*212 * , 63 ( 1988 ) .
j. m. cornwall , phys .
d * 26 * , 1453 ( 1982 ) .
a. a. natale , pos qcd - tnt * 09 * , 031 ( 2009 ) .
a. c. aguilar , d. binosi and j. papavassiliou , phys .
d * 78 * , 025010 ( 2008 ) .
j. m. cornwall , phys .
d * 83 * ( 2011 ) 076001 .
a. doff , f. a. machado and a. a. natale , annals of physics * 327 * 1030 ( 2012 ) .
a. doff , f. a. machado and a. a. natale , new . j. phys .
* 14 * , 103043 ( 2012 ) .
a. a. natale , nucl .
* b226 * , 365 ( 1983 ) . j. c. montero , a. a. natale , v. pleitez and s. f. novaes , phys .
b * 161 * , 151 ( 1985 ) .
a. doff and a. a. natale , phys .
b * 537 * , 275 ( 2002 ) .
a. doff and a. a. natale , phys .
d * 68 * , 077702 ( 2003 ) , a. doff and a. a. natale , eur .
j. * c32 * , 417 ( 2003 ) .
a. doff , a. a. natale and p. s. rodrigues da silva , phys .
d * 80 * , 055005 ( 2009 ) .
a. doff , e. g. s. luna and a. a. natale , hep - ph/1303.3248 .
j. m. cornwall and r. c. shellard , phys .
d * 18 * , 1216 ( 1978 ) .
a. doff , a. a. natale and p. s. rodrigues da silva , phys .
d * 77 * , 075012 ( 2008 ) .
s. raby , s. dimopoulos and l. susskind , nucl .
b169 , 373 ( 1980 ) .
atlas collaboration ,
search for high - mass dilepton resonances in @xmath161 of pp collisions at @xmath162tev with the atlas experiment , * atlas - conf-2012 - 129 * , 2012 ; cms collaboration ,
search for resonances in dilepton mass spectra in pp collisions at @xmath163 tev , * cms pas exo-12 - 015 * , 2012 .
s. chatrchyan et al .
[ cms collab . ] , jhep 1105 , 093 ( 2011 ) . | we verify that @xmath0 models , where the gauge symmetry breaking is totally dynamical and promoted by the non - abelian technicolor ( tc ) group and the strong abelian interactions , are quite constrained by the lhc data .
the theory contains a @xmath1 quark self - energy involving the mixing between the neutral gauge bosons , which introduces the coupling between the light and heavy composite scalar bosons of the model .
we determine the lightest scalar boson mass for these models from an effective action for composite operators , assuming details about the dynamics of the strong interaction theories . comparing the value of this mass with the atlas and cms observation of a new boson with a mass @xmath2gev and considering the lower bound determined by the lhc collaborations on the heavy neutral gauge boson @xmath3 present in these models
, we can establish constraints on the possible models .
for example , if @xmath4 , with technifermions in the fundamental representation , the model barely survives the confrontation with the lhc data . |
dynamics of vortices has been the subject of intensive study since the pioneering works of helmholtz , lord kelvin and poincar in the 19th century ( see e.g. @xcite ) .
relatively simple dynamics of a thin vortex filament in an incompressible inviscid fluid and in an infinite three - dimensional domain has attracted a particular interest .
an approach based on the localized induction approximation ( lia ) has been proposed in 1906 by da rios @xcite . within
such approximation one assumes that the vorticity @xmath0 is confined to a thin vortex core region and the core radius @xmath1 of the vortex tube is much smaller than the radius of curvature @xmath2 at the point and , hence , the induced velocity at this point is dominated by the contribution from the neighboring segment . using the biot - savart formula @xmath3 under the assumption that the circulation @xmath4 of the vortex is a constant , one gets @xmath5 where @xmath6 denotes a point on the filament , @xmath7 is the arclenght , @xmath8 is time , @xmath9 is curvature , @xmath10 denotes the binormal vector and @xmath11 is the length of the segment of the filament whose contribution into the induced velocity is taken into account in the integral ( [ bs - law ] ) . assuming that one can neglect the variation of @xmath11 along the filament , rescaling time by the constant @xmath12 and using the expression of @xmath10 in terms of @xmath6 , one obtains @xmath13 thus , within lia approach , one has a simple binormal motion of the filament with the induced velocity proportional to the curvature @xmath9 . using the serret - frenet formulae for the space curves , one can transform ( [ bin - mot ] ) into the system of intrinsic equations for the curvature @xmath9 and torsion @xmath14 which is of the form @xcite @xmath15 these lia and da rios system have been rediscovered sixty years later in @xcite .
the detailed historical account of this story is presented in @xcite .
an important step in the development of the lia approach has been done by hasimoto in 1972 @xcite .
he observed that the system ( [ bin - intrinsic ] ) is equivalent to the focusing nonlinear schroedinger equation ( nls ) @xmath16 via the transformation @xmath17 the nls equation was already widely known in plasma physics . moreover in 1971 zakharov and shabat @xcite discovered that nls equation ( [ nls ] ) is integrable by the inverse scattering transform ( ist ) method .
hasimoto s result has demonstrated that the whole powerful machinery of the ist method ( solitons , infinite sets of integrals of motion , symmetries etc . )
is applicable to the vortex filament dynamics .
this fact has lead to the explosion of interest to this field and since that time various aspects of the lia in the vortex filament dynamics have been studied by different methods ( see e.g. @xcite-@xcite ) .
instabilities of vortex filament and formation of singularities were two important problems partially addressed during this period @xcite .
at the same time it was clear that lia is too restrictive since it neglects several important effects like axial flow , stretching , shortening , filament core structure etc .
attempts to go beyond lia have been done in @xcite-@xcite . in particular
, the inclusion of the axial flow into consideration has lead to the following expression for the induced velocity @xmath18 k \vec{b } + \alpha_3 \vec{t } \times \vec{a } + \alpha_4 \vec{t } \times ( k\vec{b})_s+\vec{q}\ ] ] where @xmath19 is the tangent vector , @xmath20 are some parameters defined by the called matching condition , @xmath21 is a certain vector and @xmath22 denotes the contribution from the nonlocal effects . in situation
when one can neglect @xmath22 and variation of @xmath23 along the filament , the vector @xmath21 can be chosen as @xmath24 .
after a rescaling of time equation ( [ eqn - axial ] ) takes the form @xcite @xmath25 where @xmath26 is the normal vector , @xmath27 is a certain parameter related to the intensity of the axial flow .
after the hasimoto transformation ( [ has - trans ] ) equation ( [ eqn - axial - approx ] ) becomes @xmath28 that is the hirota equation or the linear combination of the nls and first higher nls equation @xcite .
it is again an integrable equation . on the other hand
it describes new effects such as axial flow , local stretching proportional to @xmath29 , binormal velocity depending on torsion @xmath14 , effects which were not caught by lia equation ( [ bin - mot ] ) .
it was shown in @xcite that axial flow affects essentially stability properties of the filament.the observation that the simple generalizations of equation ( [ bin - mot ] ) or da rios system ( [ bin - intrinsic ] ) may describe important nonlinear effects essential for vortex filament dynamics was one of the main results of @xcite .
another observation made in @xcite based on the paper @xcite is that the study of the quasiclassical ( or dispersionless ) limit of the equation ( [ hirota ] ) clarifies essentially an analysis of destabilizing effects in the vortex filament motion .
the latter result was strongly supported by the recent study of the modulation instability of the focusing nls and higher nls equations ( see e.g. @xcite-@xcite ) . in these papers
it was shown that the dispersionless limit of the nls equation ( dnls ) is quite relevant to the study of the modulation instability . in this limit nls
is the elliptic system of quasilinear pdes which exhibits the gradient catastrophe behavior at finite time .
behavior of solutions near to the points of catastrophe and beyond has been studied in details in @xcite-@xcite .
it was argued in these papers that the gradient catastrophe for dnls is a source of the modulation instability for nls .
the above two facts clearly indicate the importance of the integrable systems modeling certain effects of vortex filament dynamics and of their quasiclassical ( or dispersionless ) limit . moreover , the results of the papers @xcite , showing that the cutoff @xmath30 is not a constant but strongly depends on curvature and torsion of filament , manifestly show the necessity to go beyond the approximation described by equations ( [ eqn - axial - approx ] ) and ( [ hirota ] ) .
indeed , in general case , the parameters @xmath20 in ( [ eqn - axial ] ) are some functions of @xmath9 , the vectors @xmath31 and @xmath22 are certain superpositions of @xmath19 , @xmath10 , and @xmath26 and then the induced velocity is of the form @xmath32 where a , b , c are some functions .
equation ( [ x - generic - motion ] ) implies certain intrinsic equations for @xmath9 and @xmath14 which , for particular choices of a , b , c , are integrable systems @xcite different from ( [ hirota ] ) .
relevance of these equations for the vortex filament dynamics is definitely of interest .
study of the quasiclassical approximation for the intrinsic equations associated with the motion ( [ x - generic - motion ] ) and their implications for the vortex filament dynamics is the main goal of the present paper .
quasiclassical intrinsic systems describe motions of vortex filament with curvature and torsion slowly varying along the filament and in the motion , i.e. they are systems of equations for @xmath33 , @xmath34 where @xmath35 and @xmath36 are slow variables defined as @xmath37 , @xmath38 , @xmath39 ( see e.g. @xcite ) . for the da rios system ( [ bin - intrinsic ] ) one has @xmath40 or , in terms of riemann invariants @xmath41 @xmath42 for the intrinsic system associated with equation ( [ hirota ] ) ( see @xcite ) and others the corresponding quasiclassical equations are of the form @xmath43 where the characteristic velocity is an appropriate function of @xmath44 and @xmath45 . in this paper
we will concentrate on the study of the equation @xmath46\beta_x\ ] ] which is the the quasiclassical limit of the intrinsic equations corresponding to the evolution of the filament of the type ( [ x - generic - motion ] ) with @xmath47 where @xmath48 and @xmath49 being real constants .
first two terms in r.h.s . of ( [ nls - mkdv - toda - riem - inv ] ) represents themselves the quasiclassical limit of equation ( [ hirota ] ) ( see equations ( 3.76),(3.77 ) ) from ( @xcite with @xmath50 ) , while the last term is of different type . thought the corresponding contribution to the velocity @xmath51 ( [ x - generic - motion ] ) is quite complicated ( [ nls - mkdv - toda - x - motion ] ) the equation describing
this flow separately is a very simple one .
it is equivalent to the equation @xmath52 which is the dispersionless toda ( dtoda ) or the boyer - finley equation @xcite-@xcite .
dtoda equation is integrable and plays an important role in various physical phenomena ( see e.g. @xcite-@xcite ) .
it is known also that dtoda and dnls equations belong to the same dtoda / dnls hierarchy ( see e.g. @xcite ) .
while the first two terms in r.h.s . of ( [ nls - mkdv - toda - riem - inv ] ) represent `` positive '' members the last ( dtoda ) term is the first `` log - type '' term .
the presence of such dtoda contribution into the induced velocity ( [ x - generic - motion ] ) , even if not completely justified in a model independent way , could be of phenomenological importance since it might describe nonlinear effects different from those captured by equation ( [ hirota ] ) .
ellipticity of equation ( [ nls - mkdv - toda - riem - inv ] ) or , in general , of ( [ riem - inv - gen ] ) suggests to write it , following to @xcite , as the nonlinear beltrami equation @xmath53 so , a solution of equation ( [ nls - mkdv - toda - riem - inv ] ) defines ( see e.g. @xcite ) quasiconformal mapping if @xmath54 in particular , any solution of the ddr system ( [ bin - intrinsic - disp ] ) defines a quasiconformal mapping on the plane @xmath55 .
quasiconformal mappings definitely play an important role in the vortex filament dynamics . according to the result of papers @xcite-@xcite , equation ( [ nls - mkdv - toda - riem - inv ] )
well approximates the corresponding full equation of motion ( equation at @xmath56 ) until the derivatives @xmath57 and @xmath58 are not large .
analysis of behavior of solutions of equation ( [ nls - mkdv - toda - riem - inv ] ) near the point of gradient catastrophe , where @xmath57 and @xmath58 become unbounded , is the main subject of our study .
some results concerning ddr equation ( equation [ nls - mkdv - toda - riem - inv ] with @xmath59 ) have been reported briefly in @xcite . here , following the same approach
, we will analyze the full equation ( [ nls - mkdv - toda - riem - inv ] ) .
we will see that the inclusion of these two terms essentially modify the behavior of vortex filament .
our approach starts with the observation that the hodograph solutions of equation ( [ nls - mkdv - toda - riem - inv ] ) describe critical points @xmath60 of the function @xmath61 which obeys the euler - poisson - darboux equation @xmath62 , i.e. @xmath63 this fact essentially simplifies the analysis of the singular sector for equation ( [ nls - mkdv - toda - riem - inv ] ) associated with the gradient catastrophe . since @xmath64 the points of gradient catastrophe belong to a singular sector @xmath65 in the space of solutions of the system ( [ nls - mkdv - toda - riem - inv ] ) for which @xmath66 .
generic points of @xmath65 are characterized by the condition @xmath67 while for deeper singular subsectors @xmath68 , @xmath69 all the derivatives of @xmath61 vanish up to @xmath70 . for these subsectors
one has @xmath71 constraints on all variables parameterizing @xmath61 . in particular
, generic gradient catastrophe happens in a single point @xmath72 .
vortex filament exhibits a very particular behavior at the point of gradient catastrophe .
while at a regular point ( with bounded @xmath9 , @xmath14,@xmath73 , @xmath74 ) a curve modeling the filament always lies of one side of the rectifying plane ( spanned by the tangent vector @xmath19 and the binormal vector @xmath10 ) , at the point of gradient catastrophe it oscillates fast around the rectifying plane .
in addition , center of the osculating sphere ( and its radius ) blows up to infinity .
moreover , the induced velocity @xmath75 becomes unbounded at the point @xmath76 , @xmath77 .
so , the gradient catastrophe for the system ( [ nls - mkdv - toda - riem - inv ] ) give rises to the fast oscillations which resemble the classical airfoil flutter ( see e.g. @xcite ) .
numerical results @xcite show that such oscillations begin to expand along the filament @xcite .
such a behavior of filament clearly indicates on its instability . analytically at the point of gradient catastrophe one has a catastrophe of certain type . using the double scaling limit technique
we show that at the generic point of gradient catastrophe ( i.e. in the sector @xmath78 ) one has the elliptic umbilic catastrophe in terminology of r. thom @xcite with the principal part given by @xmath79 .
deeper gradient catastrophes are associated with higher singularities .
for example ,
in the sector @xmath80 the corresponding singularity is given by @xmath81 singularity from the v. arnold list @xcite with the principal part of the form @xmath82 . at the point of gradient catastrophe
the quasiclassical approximation becomes invalid and the system ( [ nls - mkdv - toda - riem - inv ] ) should be substituted by its full dispersive version ( by the intrinsic version of equation ( [ eqn - axial - approx ] ) at @xmath56 ) .
first order corrections can be obtained by the double scaling technique together with an appropriate modification of the function @xmath83 to a functional @xmath84 such that the equation @xmath85 for critical points are substituted by the euler - lagrange equation @xmath86 for @xmath87 .
it is shown that for generic point of gradient catastrophe ( i.e. the sector @xmath78 ) the resulting equation for small corrections is equivalent to the painlev i equation @xmath88 the result of the paper @xcite allows us to conjecture that any generic solution of the da rios system ( [ bin - intrinsic ] ) at the point of gradient catastrophe behaves as @xmath89 and @xmath90 where @xmath91 is described by the tritronque solution of the painlev i equation .
it is shown also that a class of double scaling regularizations of the higher gradient catastrophes associated with sectors @xmath68 is described by the equations @xmath92 the paper is organized as follows . in section 2
the relation of hodograph solutions with euler - poisson - darboux equation is discussed . in section 3
it is shown that the hodograph solutions of dtoda / dnls equation describe critical points of a function .
hydrodynamic analogy and quasiconformal mappings for the governing equations are considered in section 4 .
next section is devoted to an analysis of gradient catastrophe .
geometrical implications of gradient catastrophe and flutter of vortex filament are described in section 6 . in section 7
it is shown that in the point of gradient catastrophe one has elliptic umbilic , @xmath81 and other singularities .
double scaling regularization of these catastrophes is analyzed in section 8 .
the ddr system ( [ bin - riemann - inv ] ) , equation ( [ nls - mkdv - toda - riem - inv ] ) and general two - component elliptic quasilinear system written in terms of riemann invariants , i.e. ( [ riem - inv - gen ] ) are linearizable by the classical hodograph transformation @xmath93 ( see e.g. @xcite ) .
one has @xmath94 for equation ( [ riem - inv - gen ] ) in the form of nonlinear beltrami equation ( [ nonl - beltrami ] ) the corresponding linearized equation is @xmath95 the compatibility condition for the system ( [ hodo - gen ] ) is given by the second order equation @xmath96 solutions of which provide us with solutions of hodograph equations ( [ hodo - gen ] ) . in the hyperbolic case , i.e. for hydrodynamic type equations ,
equation of the type ( [ hodo - epd ] ) has been widely used for the construction of the solutions and an analysis of their properties ( see e.g. @xcite ) . in our elliptic case ,
equation of type ( [ hodo - epd ] ) written in term of variables @xmath97 , @xmath98 and @xmath44 defined as @xmath99 , i.e. the equation @xmath100 has been applied to the study of the system @xmath101 which appears in the theory of wave processes in unstable media within nonlinear geometric optics approximation @xcite . at @xmath102
it is the quasiclassical limit of the focusing nls equation ( [ nls ] ) or the ddr system ( [ bin - intrinsic - disp ] ) for @xmath103 , @xmath104 . in particular , in the paper @xcite ,
the behavior of the solutions for the system ( [ suleiman - eqn ] ) near the singular point @xmath105 of equation ( [ lin - suleiman ] ) has been studied . for the hyperbolic version
see also @xcite for the ddr system ( [ bin - riemann - inv ] ) equation ( [ hodo - epd ] ) is particularly simple one @xmath106 it is the euler - poisson - darboux ( epd ) equation of type @xmath107 ( see e.g. @xcite ) . for the dtoda equation @xmath108 , equation ( [ hodo - epd ] ) is of the form @xmath109 that is again the simple epd equation of the type @xmath110 .
in contrast , for equation ( [ hodo - gen ] ) with @xmath111 which corresponds to the second term in ( [ nls - mkdv - toda - riem - inv ] ) and to the quasiclassical limit of the first higher nls equation ( see formula ( [ hirota ] ) ) , equation ( [ hodo - epd ] ) is @xmath112 for higher dnls equations this equation is more and more complicated and hence an analysis similar to that performed in @xcite becomes much more involved . a way to simplify drastically equations of type ( [ hodo - epd ] ) associated with equation ( [ hodo - epd ] ) can be extracted from the tsarev s @xcite discussion of the generalized hodograph equations .
he noted that it is convenient to rewrite equation ( [ hodo - gen ] ) as @xmath113 denoting @xmath114 , one observes that @xmath115 and equation ( [ hodo - gen - tsarev ] ) assumes the form @xmath116 so , equation ( [ hodo - gen - tsarev ] ) and , hence , the hodograph equation ( [ hodo - gen ] ) takes the form @xmath117 where @xmath118 is any function obeying the constraint ( [ hodo - constr - tsarev ] ) . in the form ( [ hodo - new - tsarev ] )
the hodograph equations can be formulated also for multi - component hydrodynamic type systems @xcite .
an observation here is that for the ddr system ( [ bin - intrinsic - disp ] ) and equation ( [ bin - riemann - inv ] ) one has @xmath119 and , hence , @xmath120 furthermore , these equations imply that @xmath121 where @xmath122 and @xmath123 are certain real valued functions . for instance , for equation ( [ nls - mkdv - toda - riem - inv ] )
@xmath124.\ ] ] in virtue of ( [ lambda - eqn ] ) and ( [ omega - eqn ] ) the functions @xmath125 and @xmath126 obey the epd equation @xmath62 @xmath127 so , the hodograph equation ( [ hodo - gen - tsarev ] ) is of the form @xmath128 where @xmath129 and @xmath126 are solutions of the same epd equation @xmath62 .
it remains to note that @xmath130 is a solution of the equation ( [ hodo - epd ] ) with @xmath131 .
thus , in contrast to the usual approach in which hodograph solutions of the commuting quasilinear systems contributing into the r.h.s .
of equation ( [ nls - mkdv - toda - riem - inv ] ) can be found via different linear second order equations ( see equations ( [ epd - nls]),([epd - toda ] ) , ( [ nonepd - mkdv ] ) ) , in the approach described above all of them can be build via solutions of the same and simple epd equation @xmath62 .
epd equations of different types are well studied and have number of remarkable properties ( see e.g. @xcite ) .
their relevance for the theory of hydrodynamic type equations and whitham equations has been understood in the papers @xcite-@xcite .
in addition to the observations made in the previous section , one readily concludes that the hodograph equation ( [ hodo - fin ] ) can be represented as @xmath132 where @xmath133 and the function @xmath61 obeys the epd @xmath62 equation @xmath134 thus , one has hodograph solutions of equations ( [ nls - mkdv - toda - riem - inv ] ) describes critical points ( [ crit - eqn ] ) of the function @xmath61 defined in ( [ w ] ) which obeys the epd @xmath62 equation . in the hyperbolic case , i.e. for the quasiclassical shallow water equation ( or one - layer benney system ) this fact has been observed in @xcite ( see also @xcite ) .
so , in order to construct a solution of equation ( [ nls - mkdv - toda - riem - inv ] ) one takes a function @xmath61 of the form ( [ w ] ) with function @xmath129 given by ( [ gen - w ] ) , arbitrary functions @xmath126 solving epd @xmath62 equation and finds its critical points solving equation ( [ crit - eqn ] ) .
the class of solutions obtained in this way is controlled by the class of solutions of the epd @xmath62 equation .
various classes of solutions of this equation has been described a century ago ( see e.g. @xcite ) as well as in recent papers concerning the application of the epd equation in whitham theory .
for example , a class of solutions of the epd @xmath135 equation is given by @xmath136 where @xmath137 is the positively oriented circle of large radius and @xmath138 is an arbitrary polynomial of the complex variable @xmath139 parameterized by the constants @xmath140 ( see e.g. @xcite ) . in particular , the function @xmath125 ( [ w ] ) with @xmath56 is given by ( [ e - w ] ) with @xmath141 and @xmath142 .
parameters @xmath140 in @xmath143 and @xmath61 can be viewed as the variables parameterizing a class of initial data for equation ( [ nls - mkdv - toda - riem - inv ] ) . in terms of curvature and torsion the epd equation ( [ epd - w ] )
is of the form @xmath144 that is the axisymmetric three dimensional laplace equation studied by beltrami @xcite ( see also @xcite ) . here
@xmath9 and @xmath14 play the role of the radial and axial coordinates , respectively , in the cylindrical system of coordinates .
function @xmath61 ( [ w ] ) is @xmath145-\widetilde{\widetilde{w}}\ ] ] and hodograph equations are given by @xmath146-\widetilde{\widetilde{w}}_k=0 , \\ & w_\tau =- x+y \left [ 2\tau + \gamma_1 \left ( -3\tau^2 + 3 k \tau \right ) \right]-\widetilde{\widetilde{w}}_\tau=0 . \end{split}\ ] ] the corresponding system of quasilinear pdes in terms of @xmath9 and @xmath14 is of the form @xmath147 k_x+\left [ k-3\gamma_1 \tau k -\frac{\gamma_2}{k}\right ] \tau_x = 0 , \\
\tau_y - & \left [ k-3\gamma_1 \tau k -\frac{\gamma_2}{k}\right ] k_x + \left[2\tau-\gamma_1\left ( 3\tau^2-\frac{3}{2}k^2\right ) \right ] \tau_x = 0 .
\end{split}\ ] ] at @xmath56 it coincides with the dispersionless limit of the equations given in @xcite .
we note that the contribution from dtoda terms ( with @xmath148 ) are quite different from those of dispersionless hirota equation ( ( [ kt - sys ] ) with @xmath56 ) considered in @xcite .
formally at small @xmath9 , the last terms in ( [ kt - sys ] ) becomes dominant .
essential difference between the ddr and dhirota type evolution and dtoda contribution is manifested by the expression ( [ nls - mkdv - toda - x - motion ] ) of the velocity of filament .
the former are small for filaments close to a straight line ( small @xmath9 ) and becomes larger for more curved filaments .
in contrast , the dtoda type contribution to the induced velocity ( proportional to @xmath149 ) is smaller for larger @xmath9 and increase with decreasing @xmath9 . since in the approximation under consideration
always @xmath150 the dtoda type contribution might be relevant for the description of some nonlinear and nonlocal effects appearing for slightly curved vortex filaments . in virtue of ( [ epd - w ] ) at the critical points of the function @xmath61 given by ( [ crit - eqn ] ) or ( [ crit - eqn - kt ] )
one has @xmath151 or @xmath152 the point @xmath153 is a singular point for these equations .
analysis of the behaviors of solutions of these equations near this point can be performed in a way similar to that discussed in @xcite .
geometrically the case @xmath153 corresponds to a straight line . in this paper
we will be interested in properties of curved vortex filament .
in the rest of the paper we will assume that @xmath154 .
so , equations ( [ epd - crit ] ) and ( [ epd - crit - kt ] ) take the form @xmath155 and @xmath156 we will see that these conditions simplify essentially an analysis of singular sector for equation ( [ nls - mkdv - toda - riem - inv ] ) .
ellipticity of the ddr system ( [ bin - intrinsic - disp ] ) and more general system ( [ kt - sys ] ) means that they have properties drastically different from those of the hyperbolic version of the ddr system ( or the system ( [ suleiman - eqn ] ) with @xmath157 ) , that is the classical shallow water equation ( see e.g. @xcite ) .
the latter has standard riemann wave solutions while in the elliptic case the cauchy problem is ill posed . in spite of such a difference a representation of the elliptic systems in the form of hydrodynamic type systems , but with negative pressure , has been widely used in the analysis of their local properties ( see @xcite ) .
following this line one can rewrite the system ( [ kt - sys ] ) in terms of @xmath158 and @xmath159 as @xmath160 where @xmath161 here @xmath162 , @xmath51 , and @xmath163 play the roles of density , velocity and pressure .
note that the system considered in @xcite is not a system of quasi - linear equations .
it contains the terms nonlinear in first order derivatives in addition to those quasi - linear ones in the system ( [ kt - sys ] ) with @xmath56 .
moreover our definition of the velocity and pressure is slightly different from those from @xcite .
consequently , the system ( [ euler ] ) is composed exactly by the continuity and euler equations in contrast to the system ( 3.79)-(3.80 ) from @xcite .
properties of the system ( [ euler ] ) are fully determined by the properties of the pressure ( [ pressure ] ) . within lia
the pressure @xmath164 is always negative and that causes the well - known instability of vortex filament .
inclusion of the hirota and dtoda contributions changes the situation drastically . for large curvature axial flow contributions
dominate and the sign of the pressure is different for different values of @xmath9 and @xmath14 .
for instance , if @xmath165 then the pressure p for large @xmath9 and @xmath14 is positive , while in the opposite case it is negative .
then , for the points with small @xmath14 ( near flattening points ) one has stabilizing effect ( @xmath166 ) while , with large torsion @xmath167 dhirota contribution ( i.e. axial flow contribution ) always destabilize the filament ( recall that @xmath168 ) . for small @xmath9 ( small density )
pure dtoda contribution equal to @xmath169 dominates the pressure .
so , for small @xmath9 the sign of the pressure coincides with the sign of @xmath170 .
hence , once again one has a threshold dividing the regimes with positive and negative pressure and consequently stabilizing and destabilizing effects of contributions different from the original lia model of da rios .
if one considers the contribution only from axial flow ( @xmath171 ) the pressure is @xmath172 @xmath163 vanishes for @xmath9 and @xmath14 connected by the relation @xmath173 , while it is positive for vortex filaments with curvature and torsion obeying the condition @xmath174 and it is negative in the opposite case .
this discussion shows that stability or instability of vortex filaments strongly depends on their geometry ( relations between curvature and torsion ) .
there is another way to formulate and emphasize an ellipticity of governing equations .
the ddr system ( [ bin - riemann - inv ] ) and more generally system ( [ riem - inv - gen ] ) are elliptic systems of quasilinear equations and , hence , under certain conditions their solutions define quasiconformal mappings on the plane ( see e.g. @xcite ) .
introducing the independent variable @xmath175 , one can rewrite general equation ( [ hodo - gen ] ) in the form @xmath176 it is the nonlinear beltrami equation @xcite solution of which defines a quasiconformal mapping @xmath177 iff the dilation coefficient @xmath178 obeys the constraint @xmath179 .
this condition , i.e. @xmath180 imposes constraints on @xmath44 and , hence , on @xmath9 and @xmath14 . for the ddr system ( [ bin - intrinsic - disp ] ) , @xmath181 , and one has @xmath182 since @xmath183 one always has @xmath179 for any nonstraight filament .
so , the slow motions of nonstraight vortex filaments within original lia are described by quasiconformal mappings on the plane defined by the equation @xmath184 for the generalization of the lia given by equation ( [ hirota ] ) or ( [ nls - mkdv - toda - riem - inv ] ) situation is more complicated . for equation ( [ nls - mkdv - toda - riem - inv ] ) with
@xmath56 one has @xmath185}{8-i[4(3\beta+{{\overline{\beta}}})+3\gamma_1(5\beta^2 + 2\beta{{\overline{\beta}}}+{{\overline{\beta}}}^2)]}\ ] ] and the condition @xmath186 is verified if @xmath9 and @xmath14 and parameter @xmath27 obey the constraint @xmath187 so , only part of the vortex filament configurations with axial flow is associated with the quasiconformal mappings , a sufficient condition being @xmath188 . for the complete flow ( [ nls - mkdv - toda - riem - inv ] )
the constraint on @xmath189 and @xmath149 is more complicated .
it is of interest to note that for pure dtoda equation @xmath190 corresponding to the last term in ( [ nls - mkdv - toda - riem - inv ] ) , one has @xmath191 so , solutions of the dtoda equation define a quasiconformal mapping iff @xmath192 . at @xmath193 equation ( [ riem - inv - dtoda ] ) is equivalent to the equation @xmath194 and @xmath195 and ,
hence , any solution of the dtoda equation ( [ bf - eqn ] ) defines a quasiconformal mapping on the plane .
the results of the previous sections indicate also a relation between certain class of quasiconformal mappings on the plane and epd equation .
indeed , solutions of equation ( [ crit - eqn ] ) with @xmath196 , i.e. solutions of the beltrami equation @xmath197 are critical points @xmath60 of the function @xmath198 which obeys the epd @xmath62 equation .
this means that the class of quasiconformal mappings on the plane defined by equation ( [ beltrami - w ] ) with arbitrary real - valued function @xmath122 obeying epd @xmath62 equation describe critical points of the function @xmath199 where @xmath126 is an arbitrary solution of the epd @xmath62 equation .
inverse statement is valid too : critical points of the function ( [ wz ] ) obeying the epd @xmath62 equation and constraint @xmath200 are described by quasiconformal mappings defined by equation ( [ wz ] ) .
similar results are valid also for hydrodynamic type system of @xmath201-type and other types . the interrelation between quasiconformal mappings and vortex filament dynamics will be studied elsewhere .
hodograph equations @xmath60 and @xmath202 provide us with a single - valued solution of the ddr system ( [ bin - riemann - inv ] ) if the standard condition @xmath203 is satisfied . in virtue of the epd equation ( [ epd - w ] ) on the solution of the systems ( [ bin - intrinsic - disp ] ) and ( [ bin - riemann - inv ] ) with @xmath154 the function @xmath61 obeys the equation @xmath204 .
hence , at @xmath154 @xmath205 consequently , the regular sector of equations ( [ bin - intrinsic - disp ] ) and ( [ bin - riemann - inv ] ) for which hodograph equations are uniquely solvable is defined by the conditions @xmath206 in this sector both @xmath44 and @xmath207 are bounded . indeed , differentiating the hodograph equation @xmath60 with respect to @xmath35 and @xmath36 and using the equation ( [ wbbb ] ) , one gets @xmath208 gradient catastrophe sector or singular sector for equations ( [ nls - mkdv - toda - riem - inv ] ) is composed by those of their solutions for which @xmath44 ( i.e. curvature @xmath9 and torsion @xmath14 ) remain bounded while derivatives @xmath57 , @xmath58 explode ( for standard definition see e.g. @xcite ) .
thus , for the gradient catastrophe sector @xmath209 condition of solvability of these equations is @xmath210 and , hence , generic singular sector is characterized by the conditions @xmath211 in terms of curvature and torsion these conditions at @xmath154 have the form @xmath212 equations @xmath213 , @xmath214 , @xmath215 are the consequences of ( [ grad - cat - sec - cond - kt ] ) .
critical points of functions which obey the condition ( [ grad - cat - sec ] ) are called degenerate critical points ( see e.g. @xcite ) . degenerate critical points are typical for families of functions parameterized by several variables @xcite . in our case the family of functions @xmath61 ( [ w ] ) is parameterized by @xmath216 and other variables defining initial values of @xmath44 encoded in @xmath126 . in the generic situation
the last condition in ( [ grad - cat - sec - cond ] ) allow us to solve equations @xmath217 ,
@xmath218 with respect to @xmath44 and @xmath219 , i.e. k and @xmath14 . substituting these expressions for @xmath9 and @xmath14 into the first equation ( [ grad - cat - sec - cond ] ) ,
one gets two relations @xmath220 for equation ( [ nls - mkdv - toda - riem - inv ] ) .
thus , the generic gradient catastrophe for equation ( [ nls - mkdv - toda - riem - inv ] ) is characterized by two constraints on the independent variables @xmath221 and parameters @xmath49 . for fixed @xmath27 and @xmath149
the gradient catastrophe for equation ( [ nls - mkdv - toda - riem - inv ] ) occurs in a single point @xmath222 for given @xmath123 , i.e. for given initial data for @xmath9 and @xmath14 . for the dnls equation
@xmath59 this fact has been observed within another approach for the first time in @xcite . in a different way it was derived in @xcite . for the quasiclassical counterpart of the hirota equation ( [ hirota ] ) , i.e. equation ( [ nls - mkdv - toda - riem - inv ] ) with @xmath56
, one has a one - parametric family of points of gradient catastrophes for given initial data .
recall that the parameter @xmath27 in equation ( [ hirota ] ) and ( [ nls - mkdv - toda - riem - inv ] ) encodes the matching data conditions for inner and outer expansions for the vortex filament with axial flow @xcite-@xcite .
thus , the meaning of a true physical `` time '' of the gradient catastrophe should be assigned to the minimal value @xmath223 of the solutions @xmath224 of the equation ( [ crit - point - f ] ) at @xmath56 .
it happens in the corresponding point @xmath225 of the filament . for equation ( [ nls - mkdv - toda - riem - inv ] ) and for the class of solutions of the ddr equation corresponding to a family of initial `` data '' @xmath126 parameterized by additional variables
, one can go deeper into the singular sector .
indeed it may happens that in addition to the conditions ( [ grad - cat - sec ] ) in the singular sector the condition @xmath226 is verified . for solutions for which @xmath227 one has a subsector defined by @xmath228 continuing this procedure one can show that the singular sector @xmath65 of the ddr equation or equations of type ( [ nls - mkdv - toda - riem - inv ] ) has the following structure @xcite @xmath229 where the subsector @xmath68 is defined as @xmath230 solutions belonging to @xmath68 are defined on a subspace of codimension @xmath71 in the space of the variables @xmath231 given by equations @xmath232 in order to have nonempty singular subsector @xmath65 one should have class of solutions characterized at least by @xmath71 parameters .
so , for equation ( [ nls - mkdv - toda - riem - inv ] ) with fixed `` initial data '' @xmath122 one may have only two nonempty sectors @xmath78 and @xmath80 .
for appropriate family of the initial data one can have gradient catastrophe of arbitrary deepness .
a gradient catastrophe for the system ( [ kt - sys ] ) means a special behavior of a filament at the point @xmath76 at time @xmath77 . using the formulae ( [ x - generic - motion ] ) and ( [ nls - mkdv - toda - x - motion ] )
, one observes that the induced velocity @xmath233 has a particular behavior at the point of gradient catastrophe . within pure lia @xmath59 , @xmath234 and , hence ,
velocity remain bounded at the point @xmath76 .
however , the acceleration given by @xmath235\ ] ] explodes at @xmath76 .
axial flow ( @xmath236 ) and dtoda contribution ( @xmath148 ) qualitatively change this behavior .
indeed , the formula ( [ nls - mkdv - toda - x - motion ] ) shows that at the point of gradient catastrophe the coefficient @xmath237 , i.e. the normal component of the velocity becomes very large both for dhirota and dtoda contributions .
so , in the point of gradient catastrophe the local rate of self - stretching , proportional to @xmath238 ( see e.g. @xcite ) becomes very large .
note that such an effect is absent in pure ddr dynamics .
sign of normal component depends on the sign of @xmath73 and @xmath74 and , hence , at the point @xmath76 the filament begins to oscillate changing fast the sign of normal components of velocity .
local intrinsic consideration reveals a geometric origin and nature of such oscillations of the filament at the point of gradient catastrophe .
it is well known , that the coordinates of a curve in a neighborhood of a point in the reference system formed by the tangent vector @xmath19 , normal @xmath26 and binormal @xmath10 and with origin in the point ( see e.g. formula ( 6.7 ) from @xcite and figure [ fig - vorline ] ) are @xmath239 at a regular point.,width=302,height=188 ] + at ordinary point where @xmath9 , @xmath14 and @xmath73 , @xmath74 are bounded the first terms in r.h.s . of ( [ xxx ] )
dominate and the curve , locally , is a twisted cubic .
its projections on the osculating plane ( spanned by @xmath19 and @xmath26 ) , normal plane ( spanned by @xmath10 and @xmath26 ) and rectifying plane ( spanned by @xmath19 and @xmath10 ) are parabola ( @xmath240 ) , cusp ( @xmath241 ) and cubic ( @xmath242 ) respectively ( fig .
[ fig - pcc ] , see e.g. @xcite ) . .
at @xmath243 one should reflect the curve with respect to the axis @xmath19 in the last figure.,width=566,height=188 ] characteristic feature of a curve ( and a filament ) in an ordinary point ( i.e. , outside the points of gradient catastrophe ) is that it lies always on one side of positive direction of the normal and on one side of rectifying plane . at the point @xmath76 of gradient catastrophe
the behavior of a curve changes drastically . indeed , when @xmath29 and @xmath244 ( i.e. @xmath245 and @xmath246 ) become large , the r.h.s .
of the formula ( [ xxx ] ) contains two parameters : small @xmath7 and large @xmath29 and @xmath244 .
so , the second terms in @xmath247 and @xmath248 become relevant and different balance between @xmath7 and @xmath29 ( @xmath244 ) lead to different regimes in behavior of a curve around the origin .
for instance , parabola in the osculating plane may change sign or even convert into cubic curve ( fig .
[ fig - bn ] ) while in the normal plane it could be a plane @xmath249 curve and so on ( fig .
[ fig - nt ] ) .
so , around the point @xmath76 of gradient catastrophe a filament oscillates from one side of the rectifying plane to another and back .
such oscillation is quite similar to that of airfoil ( see e.g. @xcite ) and one can refer to these oscillations of a filament in the point of gradient catastrophe as a flutter .
another peculiarity of the point of gradient catastrophe can be seen in the behavior of the focal curve associated with filaments .
the focal curve ( see e.g. @xcite ) is formed by the totality of centers of osculating spheres .
osculating sphere has a contact of third order with the curve in a points .
its center has coordinates @xmath250 where @xmath6 is the position vector of a point on the curve .
the radius @xmath2 of the osculating sphere is given by @xmath251 for a set of regular points the corresponding piece of the focal curve is regular too and it lies on a finite distance from the original curve . at the point of gradient catastrophe @xmath252 and ,
hence , the center of the osculating sphere as well as its radius blow up to infinity in the direction of the binormal vector ( catastrophe direction , see figure [ fig - cat - dir ] ) .
so , the point of gradient catastrophe for the original curve at which both curvature and torsion are finite corresponds to a point of the focal curve at which the second focal curvature ( @xmath253 ) becomes singular .
we emphasize that even if our consideration for the original curve is purely local at the point of gradient catastrophe one observes a nonlocal effect of unbounded increase of the distance between a point on the original curve and the corresponding point on the associated focal curve .
we note that the blow - up of @xmath254 and @xmath2 for the focal curve happens also at the flattening points at which @xmath255 ( for a discussion on the flattening points see e.g. @xcite ) .
it would be of interest to understand fully a difference between the flattening points on a curve and points with unbounded @xmath29 in pure geometrical terms . as an obvious difference one may observe that at the flattening point the projection of the curve on the osculating plane remains to be a parabola and
the velocity @xmath233 and acceleration @xmath256 remain bounded too .
in order to understand better the phenomenon described above one should analyze in more detail the structure of the singularity and the regularizing mechanism for the gradient catastrophe . for the focusing nls equation
such an analysis based on the @xmath201-expansion of the integrals of motions of the nls / toda equation has been performed in @xcite . here
we will follow a different approach proposed recently in @xcite .
thus , we consider a neighborhood of the gradient catastrophe point @xmath72 and denote the values of @xmath44 at this point by @xmath257 . following the double scaling limit method ( see e.g. @xcite-@xcite ) we will look for solutions of the system ( [ kt - sys ] ) near the point of gradient catastrophe in the form @xmath258 where @xmath259 and numbers @xmath260 should be fixed by further consideration . in our approach
the function @xmath61 is the principal object for such an analysis .
for the system ( [ kt - sys ] ) @xmath261 where @xmath262 for the solutions of the system ( [ kt - sys ] ) one has @xmath263 and also , for @xmath154 , equation ( [ wbbb ] ) , i.e. @xmath264 .
furthermore differentiating epd equation ( [ epd - w ] ) with respect to @xmath44 , one has @xmath265 hence , at the point of gradient catastrophe @xmath257 , @xmath266 due to equation ( [ wbbb ] ) and @xmath267 , one gets @xmath268 where @xmath269 .
so , at the expansion ( [ gc - point ] ) for the function @xmath61 , the third order derivatives @xmath270 and @xmath271 are the only nonvanishing third order derivatives in the singular sector @xmath78 . as a result
, an expansion of the function @xmath61 near the point of gradient catastrophe is given by @xmath272 y^ * + \frac{\varepsilon^{3\eta}}{6 } \left({w}_{\beta \beta \beta}^0 { \beta^*}^3 + c.c.\right ) + \dots \ . \end{split}\ ] ] consequently @xmath273 y^ * \\ & + \frac{\varepsilon^{3\eta}}{2 } { w}_{\beta \beta \beta}^0 { \beta^*}^2 + \dots = 0 \ .
\end{split}\ ] ] for generic point in @xmath78 the third term is smaller than the second one and the nontrivial balance of three terms in ( [ wbstar ] ) is achieved if @xmath274 , i.e. @xmath275 . thus , near to the point of gradient catastrophe @xmath276 where @xmath277 and @xmath278 denoting @xmath279 , one has @xmath280 where @xmath281 and @xmath282 are certain linear combinations of @xmath283 and @xmath284 . in thom
s catastrophe theory language a function of this form is known as describing elliptic umbilic catastrophe @xcite .
the type of behavior of @xmath57 and @xmath58 near to the point of gradient catastrophe follows from ( [ bxby ] ) and ( [ wbstar ] ) .
indeed , near to the point @xmath72 , one has @xmath285 hence @xmath286 and @xmath287 the elliptic umbilic singularity ( [ ellumbcat ] ) and formula ( [ bz - cat ] ) describe generic point of gradient catastrophe which corresponds to the sector @xmath78 . for variable parameters @xmath288 , @xmath289 and other parameters describing the family of initial data , higher singularities corresponding to the singular sectors @xmath68 discussed in section 5 may appear .
for instance , if for some @xmath288 , @xmath289 , @xmath290 and @xmath291 then instead of the expansions ( [ ww0 ] ) one has @xmath292 y^ * + \frac{\varepsilon^{4\eta}}{24 } \left({w}_{\beta \beta \beta \beta}^0 { \beta^*}^4 + c.c.\right ) + \dots \ .
\end{split}\ ] ]
so , @xmath293 y^ * \\ & + \frac{\varepsilon^{4\eta}}{6 } { w}_{\beta \beta \beta \beta}^0 { \beta^*}^3 + \dots = 0 \ .
\end{split}\ ] ] and , hence , in this case @xmath294 thus , at the sector @xmath80 one has @xmath295 where @xmath296 and @xmath297 as in the previous case ( when @xmath298 ) , while now @xmath299 . denoting @xmath300
, one gets @xmath301 the last three terms reproduce the canonical form of a ( nonsimple ) unimodular parabolic singularity @xmath81 in terms of arnold s classification .
near to the point of gradient catastrophe @xmath72 the behavior of @xmath302 is given by @xmath303 therefore @xmath304 and @xmath305 in the sector @xmath68 generically one has @xmath306 and therefore @xmath307 where @xmath308 with @xmath309 and @xmath310 appearance of these singularities and particularly the time order of their appearance strongly depends on concrete structure and properties of vortex filament .
for nongeneric point of singular sector one might have a different behavior of function @xmath61 .
for instance , for the singular sector @xmath78 in situation when @xmath311 the third term in ( [ ww0 ] ) of the order @xmath312 becomes relevant and the balance of order requires that @xmath313 .
hence , in such situation @xmath314 , @xmath315 and instead of the expression ( [ wbstar ] ) one has @xmath316 where @xmath317 y^ * \\ & + \frac{\varepsilon^{3\eta}}{6 } \left({w}_{\beta \beta \beta}^0 { \beta^*}^3 + c.c.\right ) + \dots \ . \end{split}\
] ] so , in the case ( [ xy - diff ] ) the variables @xmath283 and @xmath284 play a quite different role in contrast to the generic case ( [ wstar ] ) , ( [ zstar ] ) .
such type of behavior will be studied elsewhere .
at the point of gradient catastrophe the quasiclassical approximation ceases to be valid and contribution of the higher order derivatives terms in the governing equation becomes relevant .
a standard way to deal with this problem is to begin with the full ( dispersionfull ) version of the governing equation ( i.e. da rios system ( [ bin - intrinsic ] ) or intrinsic version of equation ( [ hirota ] ) ) and , then , performing an appropriate double scaling limit , calculate the required regularization terms . for the nls equation ( [ nls ] )
such an analysis has been performed in @xcite .
the first step is based on the observation ( section 3 ) that governing equations describe critical points of the function @xmath61 .
thus , in order to incorporate derivatives of @xmath44 , it is quite natural to modify @xmath61 to @xmath318 in such a way that equation ( [ crit - eqn ] ) is substituted by the euler - lagrange equations @xmath319 for the functional @xmath320 .
second point is to use the specific properties of the function @xmath61 .
namely , the real - valuedness of @xmath61 @xmath321 and the fact that in the perturbations @xmath322 ( see formulas ( [ wstar ] ) , ( [ wstargen ] ) ) the contributions of @xmath302 and @xmath323 are separated .
it is natural to look for the modifications of @xmath61 which preserve these properties .
thus , near the point of gradient catastrophe one has the following @xmath87 ( omitting inessential terms in ( [ w3 ] ) and ( [ wstar ] ) and choosing @xmath324 ) @xmath325 + \frac{1}{2}\epsilon^\delta \left [ b { \beta_{z^*}^*}^2 + \overline{b}\ , { \overline{\beta}_{\overline{z}^*}^*}^2 \right]\ ] ] where @xmath326 and @xmath327 are appropriate constants .
euler - lagrange equations @xmath328 implies that @xmath329 .
so , one has the equation @xmath330 - b { \beta_{z^ * z^*}^ * } = 0.\ ] ] this equation is converted into the classical painlev i equation @xmath331 by the change of variables @xmath332 with @xmath333 .
so , the regularized behavior at the point of the gradient catastrophe for the vortex filament dynamics is governed by the painlev i equation .
the relevance of this equation for the nls / toda system near to the critical point has been first observed in a different way in @xcite .
the correspondence between the solution of equation ( [ p - iour ] ) and those given by the formula ( 5.21 ) in @xcite is ( at @xmath334 ) @xmath103 , @xmath104 and @xmath335 where @xmath336 , @xmath45 , @xmath337 and @xmath338 are defined in ( 2.19 ) and ( 2.22 ) of @xcite .
one also concludes that @xmath339 .
in the paper ( @xcite ) it was conjectured that any generic solution of the nls equation near to the critical point is described by the tritronque solution of the painlev i equation .
extension of this observation to our case allows us to formulate the generic behavior of the vortex filament near the point @xmath72 of gradient catastrophe within the localized induction approximation is given by @xmath340 where @xmath341 is the tritronque solution of the painlev i equation ( [ p - ibis ] ) .
such an approach can be extended also to the singular sectors @xmath68 . in these sectors
the modified @xmath87 , for example , can be chosen in the following form @xmath342 where @xmath343 are appropriate constants .
the euler - lagrange equation readily implies that @xmath344 .
so , the modified @xmath61 is given by @xmath345 and the corresponding euler - lagrange equation for @xmath302 is @xmath346 by the appropriate change of variables @xmath347 one gets the equation @xmath348 eventual relation of these equations with appropriate limits of nls hierarchy , higher painlev equations and other types of regularizations containing higher order derivatives will be considered in a separate paper . *
acknowledgements * the authors are grateful to m. bertola , b. dubrovin and e. ferapontov for useful discussions .
this work has been partially supported by prin grant no 28002k9kxz and by far 2009 ( _ sistemi dinamici integrabili e interazioni fra campi e particelle _ ) of the university of milano bicocca .
99 batchelor g.k . _ an introduction to fluid mechanics _ , cambridge univ .
press , cambridge ( 1967 ) saffman p.g .
_ vortex dynamics _ , cambridge univ
. press , new york ( 1992 ) da rios l.s .
sul moto dun liquido indefinito con un filetto vorticoso ( in italian ) rend .
circ . mat .
palermo * 22 * 117 - 135 ( 1906 ) hama f.r .
localized induction concept on a curved vortex and motion of an elliptic vortex ring , phys .
fluids * 5 * 1156 - 1162 ( 1962 ) arms r.j . and hama f.r .
localized induction concept on a curved vortex and motion of an elliptic vortex ring , phys .
fluids * 8 * 553 - 559 ( 1965 ) betchov r. on the curvature and torsion of an isolated vortex filament .
j. fluid . mech . * 22 * 471 - 479 ( 1965 ) ricca r.l .
the contributions of da rios and levi - civita to asymptotic potential theory and vortex filament dynamics fluid dyn .
res . * 18 * 245 - 268 ( 1996 ) hasimoto h. soliton on a vortex filament j. fluid .
* 51 * 477 - 485 ( 1972 ) zakharov v.e . and
shabat a.b .
exact theory of two - dimensional self - focusing and one - dimensional self - modulation of waves in nonlinear media zetf * 61 * 118 ( 1971 ) ; sov .
jtep * 34 * 62 ( 1972 ) lamb jr .
solitons on moving space curves j. math .
* 18 * 1654 - 1661 ( 1977 ) kida s. a vortex filament moving without change of form j. fluid . mech .
* 112 * 379 - 409 ( 1981 ) levi d. sym a. and wojcieckowski s. n - solitons on a vortex filament phys .
a * 94 * 408 ( 1983 ) cieslinski j. gragert p.k.h . and sym a. exact solution to localized - induction - approximation equation modeling smoke ring motion phys .
* 57 * 1507 - 1510 ( 1986 ) fukumoto y. , j. phys .
, * 56 * , 4207 - 4209 ( 1987 ) langer j. and perline r. poisson geometry of the filament equation j. nonlinear sc .
* 1 * 71 - 93 ( 1991 ) klein k. and majda a. self - stretching of a perturbed vortex filament i physica d * 49 * 323 ( 1991 ) ; ii , physica d * 53 * 267 ( 1991 ) nakayma k. , segur h. and wadati m. integrablity and the motion of curves phys . rev . lett .
* 69 * , 2603 - 2606 ( 1992 ) ricca r. physical interpretation of certain invariants for vortex filament motion under lia phys .
fliuds a * 4 * 938 - 944 ( 1992 ) klein r. , majda a. j. , mclaughlin r. m. asymptotic equations for the stretching of vortex filaments in a background flow field .
fluids a 4 * 10 * , 22712281 ( 1992 ) langer j. and perline r. , local geometric invariants of integrable evolution equations j. math .
phys . * 35 * , 17321737 ( 1994 ) nishiyama t. and tani a. solvability of the localized induction equation for vortex motion comm . math .
phys . * 162 * no .
3 , 433445 ( 1994 ) fukumoto y. , miyajima m. , the localized induction hierarchy and the lund - regge equation . j. phys .
a * 29 * no . 24 , 80258034 ( 1996 ) kuznetsov e. a. , ruban v. p. hamiltonian dynamics of vortex and magnetic lines in hydrodynamic type systems phys .
e ( 3 ) * 61 * no . 1 831841 ( 2000 ) yasui y. and sasaki n. differential geometry of the vortex filament equation j. geom .
* 28 * 195 - 207 ( 1998 ) grinevich p.
g. and schmidt m. u. closed curves in @xmath349r3 and the nonlinear schrdinger equation proceedings of the workshop on nonlinearity integrability and all that : twenty years after needs 79 ( gallipoli 1999 ) 139145 world sci .
river edge nj 2000 schief w. k. and rogers c. binormal motion of curves of constant curvature and torsion .
generation of soliton surfaces r. soc .
a math . phys . * 455 * 3163 - 3188 ( 1999 ) ruban v. p.
applicability of the hodograph method for the problem of long - scale nonlinear dynamics of a thin vortex filament near a flat boundary phys .
e , * 67 * 066301 ( 2003 ) gutierrez s. rivas j. and vega l. formation of singularities and self - similar vortex motion under the localized induction approximation commun . partial diff .
equations * 28 * 927968 ( 2003 ) holm d. d. and stechmann s.n .
hasimoto transformation and vortex soliton motion driven by fluid helicity arxiv : nlin/0409040 konno k. and kakukhata kh a new type of stretched solutions , excited by initially stretched vortex filaments , of the local induction equation theoret . and math .
phys . * 144 * 11811189 ( 2005 ) schief w. k. and rogers c. the da rios system under a geometric constraint : the gilbarg problem .
. phys . * 54 * no .
3 , 286300 ( 2005 ) holm d. d. , nitsche m. and putkaradze v. euler - alpha and vortex blob regularization of vortex filament and vortex sheet motion . j. fluid mech . * 555 * 149176 ( 2006 ) calini a. and ivey j. finite - gap solutions of the vortex filament equation : isoperiodic deformations nonlinear sci . *
17 * 527 - 567 ( 2007 ) onodera e. generalized hasimoto transform of one - dimensional dispersive flows into compact riemann surfaces sigma * 4 * , paper 044 , 10 pp.(2008 ) banica v. and vega l. on the stability of a singular vortex dynamics commun .
phys . * 286 * 593 - 627 ( 2009 ) strong s.a . and
generalized local induction equation , elliptic asymptotics , and simulating superfluid turbulence arxiv : 1102.2258 moore d. w. and saffman p. g. vortex motion with axial flow phil .
royal soc .
london ser .
a * 272 * 403 - 429 ( 1972 ) callegari a. j. and ting l. motion of a curved vortex filament with decaying vortical core and axial velocity siam j. appl . math .
35 no . 1 148175 ( 1978 ) leibovich s. and ma h. y. soliton propagation on vortex cores and the hasimoto soliton .
fluids 26 no . 11 31733179 ( 1983 ) maxworty t. , hopfinger e. j. and redekopp l. g. wave motions on vortex core j. fluid mech . * 151 * 141 - 165 ( 1985 ) fukumoto y. and miyazaki t. three - dimensional distortions of a vortex filament with axial velocity j. fluid .
mech . * 222 * 369 - 416 ( 1991 ) tani a. and nishiyama t. solvability of equations for motion of a vortex filament with or without axial flow .
inst . math .
* 33 * no . 4 509526 ( 1997 ) hirota r , exact envelope - soliton solutions of a nonlinear wave equation j. math
* 14 * 805 ( 1973 ) forest m.g . and lee j.e .
_ geometry and modulation theory for the periodic nonlinear schroedinger equation _ in : i m a volumes in mathematics and its applications * 2 * 36 - 69 springer new york ( 1986 ) el g.a .
gurevich a.v .
khodorovskii v.v . and krylov a.l .
modulational instability and formation of a nonlinear oscillatory structure in a `` focusing '' medium phys lett .
a * 177 * 357 - 361 ( 1993 ) miller p.d . and kamvissis s. on the semiclassical limit of the focusing nonlinear schrdinger equation phys lett .
a * 247 * 75 - 86 ( 1998 ) bronski j.c .
spectral instability of the semiclassical zakharov - shabat eigenvalue problem .
advances in nonlinear mathematics and science physica d * 152 - 153 * 163 - 170 ( 2001 ) cai d. mclaughlin d.w . and
mclaughlin k.t.r . _ the nonlinear schrdinger equation as both a pde and a dynamical system _ handbook of dynamical systems * 2 * 599675 north - holland amsterdam ( 2002 ) kamvissis s. mclaughlin k.d.t .- r . and miller , p.d .
_ semiclassical soliton ensembles for the focusing nonlinear schrdinger equation _ in : annals of mathematics studies * 154 * princeton university press princeton nj ( 2003 ) ercolani n. m. , jin s. , levermore c. d , and macevoy w. d. jr . , the zero - dispersion limit for the odd flows in the focusing zakharov - shabat hierarchy .
res . not .
* 47 * 25292564 ( 2003 ) tovbis a. , venakides s. and zhou x. on semiclassical ( zero dispersion limit ) solutions of the focusing nonlinear schrdinger equation commun .
pure appl . math . * 57 * 877 - 985 ( 2004 ) dubrovin b. grava t. and klein c. on universality of critical behavior in the focusing nonlinear schrdinger equation , elliptic umbilic catastrophe and the tritronque solution to the painlev - i equation j. nonlinear sci .
* 19 * 5794 ( 2009 ) difranco j.
c. and miller p. d. the semiclassical modified nonlinear schrdinger equation .
i. modulation theory and spectral analysis .
d * 237 * no .
7 , 947997 ( 2008 ) tovbis a and venakides s. semiclassical limit of the scattering transform for the focusing nonlinear schrdinger equation arxiv:0903.2648 tovbis a and venakides s. , nonlinear steepest descent asymptotics for semiclassical limit of integrable systems : continuation in the parameter space .
phys . * 295 * no . 1 139160 ( 2010 ) bertola m. and tovbis a. universality for the focusing nonlinear schroedinger equation at the gradient catastrophe point : rational breathers and poles of the tritronque solution to painlev i arxiv:1004.1828 bertola m. and tovbis a. universality in the profile of the semiclassical limit solutions to the focusing nonlinear schrdinger equation at the first breaking curve int .
res . not .
imrn * 11 * 21192167 ( 2010 ) jenkins r. and mclaughlin k. d. t .- d .
, the semiclassical limit of focusing nls for a family of non - analytic initial data arxiv:1106.1699 ( 2011 ) zakharov v.e .
benney equations and quasiclassical approximation in the inverse problem method func .
* 14 * 89 - 98 ( 1980 ) boyer c. p. and finley j. d. , iii killing vectors in self - dual , euclidean einstein spaces . j. math . phys . * 23 * no . 6 , 11261130 ( 1982 ) kodama y. solutions of the dispersionless toda equation .
a * 147 * no .
8 - 9 477482 ( 1990 ) takasaki k. and takebe t. sdiff(2 ) toda equation hierarchy , tau function , and symmetries .
* 23 * no .
3 , 205214 ( 1991 ) takasaki k. and takebe t.integrable hierarchies and dispersionless limit .
* 7 * no . 5 , 743808 ( 1995 ) gerasimov a. , marshakov a. , mironov a. , morozov a. and orlov a. matrix models of two - dimensional gravity and toda theory .
nuclear phys .
b * 357 * no . 2 - 3 , 565618 ( 1991 ) wiegmann p. b. and zabrodin a. conformal maps and integrable hierarchies .
phys . * 213 * no . 3 523538 ( 2000 ) krichever i. , mineev - weinstein m. , wiegmann p. and zabrodin a. laplacian growth and whitham equations of soliton theory .
d * 198 * no
. 1 - 2 , 128 ( 2004 ) konopelchenko b. , martnez alonso l. and medina e. on the singular sector of the hermitian random matrix model in the large n limit . phys
a * 375 * no . 5 867872 ( 2011 ) lavrentev m. a general problem of the theory of quasi - conformal representation of plane regions .
( russian ) mat .
sbornik n.s .
* 21 * ( 63 ) 285320 ( 1947 ) ahlfors l. v. , _ lectures on quasiconformal mappings .
_ van nostrand mathematical studies , no .
10 d. van nostrand co. , inc . ,
toronto , ont .- new york - london ( 1966 ) bojarski b. , quasiconformal mappings and general structural properties of systems of non linear equations elliptic in the sense of lavrentev .
symposia mathematica * 18 * 485499 ( 1976 ) konopelchenko b. g. and ortenzi g. gradient catastrophe and flutter in vortex filament dynamics .
j. phys . a * 44 * no . 43 , 432001 , 12 pp ( 2011 ) milne - thompson l.m . ,
_ theoretical hydrodynamics _
mcmillan & co. ( 1960 ) thom r. , _ structural stability and morphogenesis .
an outline of a general theory of models _ addison - wesley publishing company redwood city ca ( 1989 ) arnold v.i .
gusen - zade s.m . and varchenko a.n .
_ singularities of differentiable maps .
i. the classification of critical points , caustics and wave fronts _ monographs in mathematics * 82 * birkhuser boston inc .
boston ma ( 1985 ) weinstein a. the singular solutions and the cauchy problem for generalized tricomi equations commun .
pure appl . math .
* 7 * 105 - 116 ( 1954 ) rodestvenski b. l. and janenko n. n. , _ systems of quasilinear equations and their applications to gas dynamics
mon . * 55 * ams providence ri ( 1983 ) landau l. d. and lifshitz e. m. fluid mechanics course of theoretical physics , vol .
6 pergamon press ltd ( 1959 ) gurevich a.v . and
shvartsburg a.b . exact solutions of nonlinear geometrical optics equations jetp , * 58 * 2012 - 2022 ( 1970 ) shvartsburg a.b .
nonlinear geometric optics in nonlinear wave theory .
nauka moscow ( 1976 ) kudashev v. r. and suleimanov b.i .
characteristic features of some typical spontaneous intensity collapse processes in unstable media .
jetp letters * 62 * 358 - 363 ( 1995 ) garifullin r. n. and suleimanov b. i. from weak discontinuities to nondissipative shock waves .
jetp * 110 * 133 - 146 ( 2010 ) darboux g. _ leons sur la thorie gnrale des surfaces ii _ gauthier villars ( 1915 ) tsarv s. p. the geometry of hamiltonian systems of hydrodynamic type . the generalized hodograph method .
( russian ) izv .
nauk sssr ser
. mat . * 54 * no . 5 10481068 ( 1990 ) ; translation in math .
ussr - izv .
* 37 * no . 2 397419 ( 1991 ) kudashev v. r. and sharapov s. e. hydrodynamic symmetries for the whitham equations for the nonlinear schrdinger equation ( nse ) .
a * 154 * no .
9 , 445448 ( 1991 ) ; inheritance of symmetries in the whitham averaging of the korteweg - de vries equation , and hydrodynamic symmetries of the whitham equations .
theoret . and
math . phys .
* 87 * , no .
1 , 358363 ( 1991 ) gurevich a.v . , krylov a.l . , el g.a . breaking of a riemann wave in dispersive hydrodynamics .
jetp letters * 54 * 104 - 109 ( 1991 ) tian f. r. , oscillations of the zero dispersion limit of the korteweg - de vries equation .
pure appl . math . *
46 * no . 8 10931129 ( 1993 ) pavlov m.v .
, integrable hydrodynamic chains . j. math .
phys . * 44 * no . 9 4134 - 4156 ( 2003 ) konopelchenko b. martnez alonso l. and medina e. hodograph solutions of the dispersionless coupled kdv hierarchies , critical points and the euler - poisson - darboux equation j. phys .
a : math . theor . * 43 * 434023 ( 2010 ) konopelchenko b. martnez alonso l. and medina e. singular sectors of the 1-layer benney and dtoda systems and their interrelations theor .
* 168 * ( 1 ) , 963 - 973 ( 2011 ) beltrami e. sulla teoria delle funzioni potenziali simmetriche mem r. accad .
bologna * 2 * 461 - 505 ( 1880 ) ; opere mat .
* 3 * 349 - 382 ( 1911 ) eisenhart l.p . , _ an introduction to differential geometry _
princeton univ .
press princeton ( 1947 ) uribe - vargas r. on the stability of bifurcation diagrams of vanishing flattening points .
( russian ) funktsional .
i prilozhen .
* 37 * no .
3 , 8894 ( 2003 ) ; translation in funct . anal .
appl . * 37 * no .
3 , 236240 ( 2003 ) konopelchenko b. and martnez alonso l. , _ catastrophes , double scaling regularization and integrable systems _ ( in preparation ) di francesco p. ginsparg p. and zinn - justin z. 2d gravity and random matrices phys . rep .
bf 254 2 ( 1995 ) martnez alonso l. and medina e. the double scaling limit method in the toda hierarchy j. phys .
a : math . theor . * 41 * 335202 ( 2008 ) | quasiclassical approximation in the intrinsic description of the vortex filament dynamics is discussed . within this approximation
the governing equations are given by elliptic system of quasi - linear pdes of the first order .
dispersionless da rios system and dispersionless hirota equation are among them . they describe motion of vortex filament with slow varying curvature and torsion without or with axial flow . gradient catastrophe for governing equations is studied .
it is shown that geometrically this catastrophe manifests as a fast oscillation of a filament curve around the rectifying plane which resembles the flutter of airfoils .
analytically it is the elliptic umbilic singularity in the terminology of the catastrophe theory .
it is demonstrated that its double scaling regularization is governed by the painlev i equation . |
under the hierarchical galaxy formation scenario , galaxy interactions play a key role in the evolution of galaxies ( e.g. @xcite ) .
owing to the nonlinear nature of the galaxy interactions , theoretical studies mostly rely on @xmath0-body / hydrodynamic simulations .
the initial conditions of the simulations often adopt isolated galaxy models .
galaxy models in many galaxy interaction simulations include dark matter halo , gas disk , stellar disk , and stellar bulge @xcite .
unfortunately , generating an isolated stable galaxy model is not trivial procedure .
it is required to solve complicated orbit integrations to make stable stellar systems @xcite .
the interactions among the different components significantly modify the very initial setups @xcite .
furthermore , it is needed to satisfy the initially hydrostatic condition of the gas components . in order to isolate the galaxy interaction effects
, it is necessary to carefully build the initial galaxy model and understand the evolution of the initial conditions , if there are any .
however , the basic steps to construct various types of initial galaxy models are often omitted or discussed shortly .
the detailed comparison studies of the evolution of different galaxy models in isolation are also not found sufficiently in the literature .
moreover , it has been reported that some spiral galaxies like our own milky way as well as elliptical galaxies possess hot diffuse gas in the halo .
importance of hot halo gas in the evolution of galaxies has been also evidenced by @xcite , @xcite , @xcite , and @xcite who studied the effects of galaxy interactions on galaxy properties such as morphology , color , and star formation activity among others at various environment and redshifts .
it has been consistently suggested by these studies that the hydrodynamic interaction between galaxies through exchange of cold gas and/or impact of hot halo gas makes a crucial role in the evolution of galaxy color and star formation activity .
however , only a few numerical studies taking account of halo gas in the models have been presented .
recently , @xcite performed hydrodynamic simulations of major mergers of disk galaxies , including for the first time a gradually cooling hot gas halo in their galaxy models .
they studied the impact of a gaseous halo and galactic winds on the star formation activity during mergers and found that the gas halo component strongly affects the kinematics and internal structure of the merger remnants as well as the star formation history on the course of the merger . in a follow - up study @xcite
, they also investigated the role of a hot gas component on disk thickening during minor merger and showed the dependency of the final scale height on the mass of the hot halo , the efficiency of the winds , and the merger mass ratio .
they found that the cooling of the hot gaseous halo regenerates the thin disk , so that the remnants of 1:10 mergers exhibit a newly formed thin disk .
these findings suggest that the existence of a hot gaseous halo significantly affects the results of galaxy interactions .
it may be necessary to reinterpret many previous galactic scale simulation results that did not consider a hot gaseous halo . in this paper , we present detailed descriptions of the various choices of the isolated milky way - like galaxy models .
we also study the effects of the gas halo component with different properties , such as the cumulative mass profile , total mass , and rotation of the gas halo on their evolution .
this paper is the first in a series of the galaxy interaction study that includes a hot gas halo in the initial galaxy model with a hope to explain the observational studies by @xcite , @xcite , @xcite , @xcite , and @xcite .
recent galaxy formation studies suggest that the galactic outflow is an essential mechanism for balance between gas accretion and galaxy growth ( e.g. @xcite ) .
it also regulates the star formation rate of galaxies . in our study
, we also include galactic outflow driven by supernovae with different wind parameters to study the effects of a gas halo and galactic winds on the evolution of isolated galaxies . in sec . 2
, we give a brief account of the zeno software package and explain our galaxy models and their initial parameters .
we describe in sec .
3 the simulation code gadget-3 and the results of evolution of the models .
finally , we summarize and discuss our main results in sec . 4 . the step - by - step procedure of constructing our galaxy models is presented in appendix a , and the long - term evolution of one of our models in appendix b.
to generate the initial galaxy models , we use the zeno software package ( version 008 ) .
zeno is a collection of programs for @xmath0-body / sph simulations and analysis , provided by j. e. barnesbarnes / software.html ] .
zeno allows one to build multiple spherical and disk components in mutual equilibrium with user - specified density profiles in collisionless or gaseous form , that fulfills our needs for constructing various galaxy models .
we use about a dozen programs in the zeno package to complete our multi - component galaxy models .
the procedure of constructing the models using those programs is described in appendix a in detail .
the zeno programs use dimensionless units with newton s constant @xmath1 = 1 .
we set the total disk mass ( including both stars and gas ) as the unit mass and the radial scale length of the stellar disk as the unit length in all models .
we set the components of our models to follow those of the milky way galaxy , and the units of length , mass , and time are 3.5 kpc , @xmath2 m@xmath3 , and 14 myr , respectively .
the unit velocity then corresponds to 247 km s@xmath4 . in our models both star and gas disks have an exponential surface density profile and a sech@xmath5 vertical profile ( c.f .
@xcite ) : @xmath6 where ` @xmath7 ' stands for either ` @xmath8 ' for the star disk component or ` @xmath9 ' for the gas disk component .
@xmath10 is the cylindrical radius @xmath11 and @xmath12 is the mass of the star or the gas disk .
@xmath13 is the radial scale length and @xmath14 is the vertical scale height of the disk .
a few distant particles farther than a cut - off radius ( @xmath15 ) are removed .
the gas disk is radially more extended than the star disk in our models .
the radial and vertical scales of the gas disk are chosen as @xmath16 and @xmath17 .
the mass density of the dm halo of all models and the gas halo of some models follows an nfw profile ( @xcite ) .
since the cumulative mass of the profile diverges at large radii , an exponential taper is applied at radii larger than @xmath18 : @xmath19 \displaystyle \rho_{\rm hc}^ { * } \ , \left(\frac{b_{\rm hc}}{r}\right)^2 \ , e^{-2 \beta ( r / b_{\rm hc } -1 ) } & { \rm for } \,\ , r > b_{\rm hc } \ , , \\ \end{array } \right . \label{eq2}\ ] ] where ` @xmath7 ' represents either ` @xmath20 ' for the dm halo component or ` @xmath9 ' for the gas halo component .
@xmath21 is the radial scale of the dm or gas halo model and @xmath22 is the mass of the halo within @xmath21 .
the radial scales of both dm and gas halos are chosen to be the same . @xmath23 and @xmath24 are calculated by requiring both @xmath25 and @xmath26 to be continuous at @xmath18 . the gas halo of the rest of our models follows a non - singular isothermal profiles with truncation : @xmath27 where @xmath28 and @xmath29 are the radius of core and taper , respectively .
@xmath30 is a function of @xmath28 and @xmath29 , and @xmath31 is the total mass of the halo gas .
the mass density of the bulge , which consists of stars only , in all models follows a @xcite profile , with a cut - off at large radii for efficient computation : @xmath32 \displaystyle \rho_{\rm b}^ { * } \ , \left(\frac{b_{\rm b}}{r}\right)^2 \ , e^{-2 r / b_{\rm b } } \ , & { \rm for } \,\ , r > b_{\rm b } \ , , \\
\end{array } \right .
\label{eq : hernquist}\ ] ] where @xmath33 is the radial scale of the bulge model and @xmath34 is the radius at which truncation starts .
@xmath35 is the mass of the bulge .
@xmath36 is determined by the continuity of @xmath37 at @xmath38 . for @xmath39 ,
@xmath40 is continuous at @xmath38 as well .
we construct seven different galaxy models in total .
their basic properties are summarized in tables 1 and 2 .
we first build model dhi as our fiducial model .
it contains a gas halo component following an isothermal density profile .
we then generate six more comparison models without or with a gas halo of varying properties . lcccl dhi & isothermal & @xmath41 & 0.01 & 0.12 + dhi - f5 & isothermal & @xmath41 & 0.05 & 0.12 + dhir & isothermal & gas disk [email protected] & 0.01 & 0.12 + + dhn & nfw & @xmath41 & 0.01 & 0.12 + dhn - f5 & nfw & @xmath41 & 0.05 & 0.12 + + d & @xmath41 & @xmath41 & @xmath41 & 0.12 + + hi & isothermal & @xmath41 & 0.01 & @xmath41 + llcccc * star disk : + disk model & & exponential & exponential & exponential & exponential + @xmath43 [ kpc ] & length scale of star disk & 3.5 & 3.5 & 3.5 & 3.5 + @xmath44 [ kpc ] & vertical disk scale height & 0.35 & 0.35 & 0.35 & 0.35 + @xmath45 [ kpc ] & outer disk cut - off radius & 42 & 42 & 42 & 42 + @xmath46 [ @xmath47 m@xmath3 ] & total mass of star disk & 4.4 & 4.4 & 4.4 & 5.0 + @xmath49 & number of particles & 16384 & 16384 & 16384 & 16384 + @xmath50 [ @xmath47 m@xmath3 ] & mass of individual particles & @xmath51 & @xmath51 & @xmath51 & @xmath52 + @xmath53 [ kpc ] & gravitational softening length & 0.609/0.621 & 0.609/0.621 & 0.606 & 0.649 + + * gas disk : + disk model & & exponential & exponential & exponential & @xmath41 + @xmath54 [ kpc ] & length scale of gas disk & 8.75 & 8.75 & 8.75 & @xmath41 + @xmath55 [ kpc ] & vertical disk scale height & 0.35 & 0.35 & 0.35 & @xmath41 + @xmath56 [ kpc ] & outer disk cut - off radius & 105 & 105 & 105 & @xmath41 + @xmath57 [ @xmath47 m@xmath3 ] & total mass of gas disk & 0.6 & 0.6 & 0.6 & @xmath41 + @xmath58 & number of particles & 16384 & 16384 & 16384 & @xmath41 + @xmath59 [ @xmath47 m@xmath3 ] & mass of individual particles & @xmath60 & @xmath60 & @xmath60 & @xmath41 + @xmath61 [ kpc ] & gravitational softening length & 0.225/0.229 & 0.225/0.229 & 0.224 & @xmath41 + + * dm halo : + halo model & & nfw & nfw & nfw & nfw + @xmath62 [ kpc ] & radial scale of dm halo & 21 & 21 & 21 & 21 + @xmath63 [ kpc ] & radius to begin tapering & 84 & 84 & 84 & 84 + @xmath64 [ @xmath47 m@xmath3 ] & mass within radius @xmath62 & 12.23/11.74 & 12.23/11.74 & 12.35 & 12.23 + @xmath65 & total mass of dm halo & 118.8/114 & 118.8/114 & 120 & 118.8 + @xmath66 & number of particles & 163840 & 163840 & 163840 & 163840 + @xmath67 [ @xmath47 m@xmath3 ] & mass of individual particles & @xmath68 / & @xmath68 / & @xmath69 & @xmath68 + & & @xmath70 & @xmath70 + @xmath71 [ kpc ] & gravitational softening length & 1 & 1 & 1 & 1 + + * gas halo : + halo model & & isothermal & nfw & @xmath41 & isothermal + @xmath28 [ kpc ] & radial scale , or & @xmath41 & 21 & @xmath41 & @xmath41 + & @xmath72 radius of core & 10.5 & @xmath41 & @xmath41 & 10.5 + @xmath29 [ kpc ] & radius to begin taper , or & @xmath41 & 84 & @xmath41 & @xmath41 + & @xmath72 radius of taper & 420 & @xmath41 & @xmath41 & 420 + @xmath73 [ @xmath47 m@xmath3 ] & mass within radius @xmath28 & @xmath41 & 0.12/0.62 & @xmath41 & @xmath41 + @xmath31 [ @xmath47 m@xmath3 ] & total mass of gas halo & 1.2/6.0 & 1.2/6.0 & @xmath41 & 1.2 + @xmath74 & number of particles & 32768/163840 & 32768/163840 & @xmath41 & 32768 + @xmath75 [ @xmath47 m@xmath3 ] & mass of individual particles & @xmath60 & @xmath60 & @xmath41 & @xmath60 + @xmath76 [ kpc ] & gravitational softening length & 0.225/0.229 & 0.225/0.229 & @xmath41 & 0.225 + + * bulge : + bulge model & & hernquist & hernquist & hernquist & hernquist + @xmath33 [ kpc ] & length scale of bulge & 0.7 & 0.7 & 0.7 & 0.7 + @xmath34 [ kpc ] & radius at which truncation starts & 140 & 140 & 140 & 140 + @xmath35 [ @xmath47 m@xmath3 ] & total mass of bulge & 1 & 1 & 1 & 1 + @xmath77 & number of particles & 8192 & 8192 & 8192 & 8192 + @xmath78 [ @xmath47 m@xmath3 ] & mass of individual particles & @xmath79 & @xmath79 & @xmath79 & @xmath79 + @xmath80 [ kpc ] & gravitational softening length & 0.410/0.419 & 0.410/0.419 & 0.408 & 0.410 + * * * * *
our models are characterized by the existence of a gas halo and denoted by initials dh , d , and h. ( d and h stand for disk and halo , respectively . ) : the models of type dh , such as the fiducial model dhi , consist of three collisionless components , dark matter ( dm ) halo , stellar disk , and stellar bulge , as well as two gaseous components , gas halo and gas disk .
the models of types d and h also possess the three collisionless components but have only one gaseous component at the initial time , either gas disk in type d or gas halo in type h. in all types of our models , the total mass is set to @xmath81 , i.e. , as similarly massive as the milky way galaxy .
the total masses of the halo ( dm + gas , if the model includes a gas halo ) , the disk ( stars + gas , if the model has a gas disk ) , and the bulge ( stars only ) components are @xmath82 , @xmath83 , and @xmath84 , respectively in all models .
we test two different cumulative mass profiles for the gas halo , the isothermal density profile ( eq . 3 ) and the nfw profile ( eq . 2 ) . according to the adopted mass profile , the models of type dh are divided into two subtypes - dhi and dhn .
the gas halo of the models of subtype dhi ( models dhi , dhi - f5 , and dhir ) and model hi follows the isothermal profile , and that of subtype dhn ( models dhn and dhn - f5 ) follows the nfw profile .
we also try two distinct values of halo gas fraction @xmath85 ( table 1 ) , in order to examine how galaxy models with different amounts of hot gas evolve differently . in one case ,
the halo gas fraction in models dhi , dhir , dhn , and hi ( without ` -f5 ' in the model names ) is chosen to be @xmath86 , which corresponds to @xmath87 .
this mass of a gaseous halo is motivated from anderson & bregman ( 2010 ) where the upper limit of the hot halo mass of the milky way galaxy is estimated to be @xmath88 from the observed dispersion measure of pulsars in the large magellanic cloud , assuming an nfw profile for the hot gas . in the other case , in models dhi - f5 and dhn - f5 , a more massive gas halo with @xmath89
is adopted , which corresponds to @xmath90 .
the more massive gas halo is motivated from @xcite . in their model with the maximum hot gas fraction , the mass of the gas halo within the virial radius is @xmath91 , such that the baryonic fraction within the virial radius has the universal value .
the gas fraction in the disk is chosen to be a single value of @xmath92 = 0.12 and thus @xmath93 in all models possessing a gas disk .
so the total mass of cold gas in a disk is either half ( in models dhi , dhir , and dhn ) or a tenth ( in models dhi - f5 and dhn - f5 ) of the total mass of the hot halo gas .
the effects of the rotation of the hot gas halo is also examined by comparing models dhi and dhir ; the latter is made from the former by adding an initial spin to the gas halo in such a way that the gas halo has the rotation curve similar to that of the gas disk , but with about half the amplitude ( rotational velocity ) .
the hot gas particles in model dhir thus have non - zero initial velocities in @xmath94 and @xmath95 directions but zero in @xmath96 direction ; in all other models with a gas halo , halo gas particles have zero initial velocities in all three directions .
the initial temperatures of the disk gas particles in all models possessing a gas disk are set to the single value of @xmath97 .
the temperatures of the halo gas particles are determined by the hydrostatic equilibrium in the zeno code ( appendix a ) .
the gas temperature in the central part of our models within @xmath98 kpc generally has a value between @xmath99 and @xmath100 .
we summarize the key properties of each initial galaxy model below .
model dhi is the fiducial model .
as noted earlier , it consists of five components : star disk , gas disk , dm halo , gas halo , and bulge .
the total mass of each of the components in the unit of @xmath101 are @xmath102 , @xmath103 , @xmath104 , @xmath105 , and @xmath106 as summarized in table 2 .
hence , the mass of the hot gas initialized in the halo is twice that of the cold gas in the disk . in terms of gas fractions in the disk and the halo , @xmath107 and @xmath86 as indicated in table 1 .
the number of star or gas particles set in each component is @xmath108 , @xmath109 , @xmath110 , @xmath111 , and @xmath112 .
the number of gas particles in the halo is twice of that in the disk , making the same mass per gas particle in the system .
the particle mass is different for different components .
to minimize the effects of the mass difference of simulation particles we set @xmath113 to be equal for particles in each component , where @xmath114 is the maximum acceleration experienced by a particle .
we choose the gravitational softening length for the dm halo particles to @xmath71 = 1 kpc ( in all models ; table 2 ) . then the smoothing lengths for the gas , disk star , and bulge particles are determined to be 0.225 , 0.609 , and 0.410 kpc , respectively .
the initial distribution of the star and gas particles and some basic profiles of the components are shown in figs 1 and 2 , respectively .
as for both star and gas disks , the particle distributions in the @xmath94-@xmath95 and @xmath94-@xmath96 projections are presented in the first two columns of fig . 1 , and their surface density profiles in the top - left panel of fig . 2 .
both disks have the exponential surface density profile ( eq . 1 ) .
the gas disk is seen more extended than the stellar one , with the larger radial scale length of @xmath115 kpc ( table 2 ) .
the particle distributions of the three spheroidal components are shown in the third to fifth columns in fig . 1 , and
the spherically averaged accumulated density profiles are plotted in the top - right panel of fig . 2 .
the dm halo follows the nfw profile with the exponential taper ( eq . 2 ) .
the radial scale is set to @xmath116 kpc , the same as in the galaxy model of @xcite , and the radius to begin taper to @xmath117 kpc .
the mass of the dm halo within @xmath118 is set to @xmath119 so that the total mass of the dm halo becomes @xmath120 .
the gas halo has the truncated isothermal profile ( eq . 3 ) , with the core radius @xmath121 kpc and the taper radius @xmath122 kpc
. the choice of the core radius is motivated by the recent model of moster et al .
given the core radius , the gas halo has nearly flat density within @xmath28 , as seen in the magenta line in the top - right panel of fig . 2 .
the bulge follows the hernquist profile with a truncation ( eq . 4 ) .
the length scale of the bulge is set to @xmath123 kpc , adopted from mcmillan and dehnen ( 2007 ) .
the cylindrically averaged tangential velocities @xmath124 of all components with respect to @xmath10 are shown in the bottom - left panel of fig . 2 .
the gas disk particles are initially given to have the local circular velocities ( cyan solid line ) in clockwise direction with zero vertical motion ( @xmath125 ) , the star disk particles to have additional dispersions ( black dashed line ) .
note that the gas halo particles of this model are set with zero initial velocities in all three directions .
the temperature @xmath126 as a function of the local gas density @xmath127 in the disk or halo is presented in the bottom - right panel of fig . 2 .
( the temperature is converted from the specific internal energy @xmath128 of the gas particles , assuming that @xmath126 below 10000 k , a mean molecular weight corresponding to neutral gas of primordial abundance , otherwise full ionization . )
the temperature of the disk gas particles are set uniformly to @xmath129 k ( cyan dots ) .
the temperature of the halo gas varies ( magenta dots ) to achieve hydrostatic equilibrium of the hot halo gas .
those halo gas particles in the @xmath126-@xmath127 diagram appear well - separated from the cold disk gas at the initial time . as the system evolves ( the model evolution is discussed in sec .
3 ) , @xmath126 and @xmath127 of the gas particles will be redistributed and some of the halo and the disk particles will be mixed .
for example , some cold gas particles which are originally set as the disk gas can be heated and have physical characteristics of the halo gas , and vice versa .
therefore , we establish a simple criterion , shown by the black line in the @xmath126-@xmath127 plane , to determine if a gas particle at a certain time should be considered as the disk or the halo particle .
the criterion is fixed in all our models at all times ( c.f .
fig . 3 ) .
the criterion is originally set in the @xmath128-@xmath127 plane and is converted on the assumption described above .
the line is chosen such that ( 1 ) the disk and the halo gas particles of all initial galaxy models can be well - separated by the common line , ( 2 ) all disk gas particles in the model without a gas halo and without considering winds can stay below the line at all times , ( 3 ) and all halo gas particles in the model without a gas disk and without winds can always stay above the line .
any gas particles lie below the black line at a time are considered as the disk gas and those above the line as the halo gas particles , with their origin identified by cyan or magenta in the @xmath126-@xmath127 plot .
this model is intended to have a more massive gas halo than that of the fiducial model dhi , with the other properties in common ( tables 1 and 2 ; to keep the total mass of the dm + gas halo the same , the dm halo of model dhi - f5 is set to be lighter than that of model dhi by the mass difference in the gas halos ) .
specifically , the gas halo of this model follows the same truncated isothermal density profile of eq . 3 as in the fiducial model with the common values of @xmath28 and @xmath29 , but is five times more in mass @xmath130 as well as in the number of particles @xmath74 .
we keep the single ( hot or cold ) gas particle mass the same at the initial time ( in all our models ) . as intended , the components except the gas halo in this model , and also in all other models of type dh , have ( statistically ) identical particle distributions at the initial time to those of the fiducial model dhi shown in fig . 1 .
in all models of type dh , the initial profiles of the four components , such as the surface density profiles of both disks , the spherically averaged density profiles of the dm halo and the bulge , and the tangential velocity profiles of the four components , are also the same ( or nearly the same for the properties of the dm halo ) as those of the fiducial model presented in fig . 2 . but
the gas halo of this model has higher density at each radius than that of the fiducial model dhi as shown in the second row - right column in fig .
4 ( see the black line for the initial density profile ) .
the initial velocities of the halo gas particles in this model are set to zero as in the fiducial model .
the temperature @xmath126 with respect to the local gas density @xmath127 for the gas particles in model dhi - f5 is presented in fig .
as seen in the second row - leftmost column in the figure , the disk gas particles are initialized to have a single value of @xmath131 ( cyan dots ) as in the fiducial model ( and as also in all models with a gas disk ) , but the temperature of the halo gas particles is assigned in accordance with the hydrostatic equilibrium of the halo gas ( magenta dots ) .
model dhir is designed to examine how the rotation of the gas halo makes a difference in the evolution .
thus the only difference between model dhir and the fiducial model dhi is whether the gas halo is spinning with non - zero initial velocities ( in model dhir ) or stationary at the beginning .
model dhir shares all parameter values of all components given in table 2 with the fiducial model .
the initial distribution of the particles and some basic profiles are presented in figs 1 and 2 , which are common to both models dhir and dhi except the tangential velocity profile of the gas halo as designed ( bottom - left panel of fig . 2 ) .
the tangential velocities of the hot gas in model dhir are assigned in such a way that the rotation curve of the gas halo has the same shape but has half the amplitude of that of the gas disk .
the vertical velocity @xmath132 is set to zero in this model .
the choice of the gas halo rotation is not yet strongly restricted by current observations .
for example , @xcite chose to set the spin of a gas halo in such a way that the specific angular momentum of the gas halo is four times that of the dm halo ( see sec . 2 of their paper for details ) .
the test of different gas halo rotations would be the subject of future research . as one of the dh type models ,
model dhn consists of the five components of both star and gas disks , both dm and gas halos , and a bulge .
differently from the fiducial model dhi , the gas halo of this model follows the nfw profile instead of the isothermal profile ( tables 1 and 2 ) .
the initial total mass of the gas halo is the same as that of the fiducial model .
this model is used to examine how the different initial density profiles of the hot gas with equal mass make differences in the evolution .
the particle distributions in the four components except the gas halo in model dhn are identical to those of the fiducial model in fig . 1 .
the gas halo follows the nfw profile like the dm halo but has five times fewer particles than in the dm halo .
the spherically averaged density profile of the gas halo is presented in fig .
4 ( fourth row - right column ) .
it is much higher within a few tens of kpc than that of the fiducial model and monotonically increasing towards the center .
the initial tangential velocity of the gas halo is zero in this model .
the temperature @xmath126 of the gas particles at the initial time is presented in fig . 3 with respect to the local density @xmath127 ( fourth row - leftmost panel ) .
model dhn - f5 , like the previous model dhn , possesses a gas halo following the nfw profile but has five times more hot halo gas particles ( tables 1 and 2 ) .
the spatial distribution of the hot gas following the nfw profile has the same shape as that of the dm halo shown in fig . 1 .
the number of the hot halo gas particles is equal to that of the halo dm particles .
a halo gas particle has 1/19 times lower mass than a dm particle .
the spherically averaged density profile of the gas halo is specifically shown in fig . 4 ( fifth row - right column ) .
model dhn - f5 has the highest density at the inner part among all our models .
the initial velocity of the gas halo is set to zero .
the temperature of the gas as a function of the local gas density is plotted in fig .
3 ( fifth row - leftmost panel ) .
this model is composed of star and gas disks , a dm halo , and a bulge , but has no gas halo . the disk gas fraction and the total mass and the number of particles of each component are summarized in tables 1 and 2 .
the initial particle distributions of the four components in this model are equal to those of model dhi ( fig . 1 ) .
the initial @xmath126 versus @xmath127 plot for the gas disk is presented in fig . 3 ( bottom row - leftmost column ; cyan dots only ) .
model hi is similar to the fiducial model dhi but has no gas disk .
its gas halo follows the isothermal density profile ( tables 1 and 2 ) .
the initial particle distributions of the four components are equal to those of model dhi ( fig . 1 ) .
the spherically averaged initial density profile of the hot halo is shown in fig . 4 ( bottom row - right column ; black line ) , which is the same as that of model dhi . the initial velocity of the gas halo is zero in this model .
the surface density of the star disk ( bottom row - left column in fig .
4 ; black line ) is highest among those of all models , as it has the more massive star disk without a gas disk .
the initial @xmath126 versus @xmath127 diagram for the gas halo is presented in fig .
3 ( bottom row - leftmost column ; magenta dots only ) ; it is different from that of the fiducial model due to the different galactic potentials of the models .
we use an early version of gadget-3 code ( originally described in springel 2005 ) to evolve our galaxy models listed in table 1 in isolation for 3 gyr .
the simulation code provides a model for galactic winds driven by star formation .
we first run the seven models without including galactic winds .
we then perform four more simulations using two of our models - models dhir and d - including winds with different sets of wind parameter values as summarized in table 3
. we will refer to the four runs with winds as ` wind test runs ' ( table 3 ) and name them dhir - wa , dhir - wa - e1 , dhir - wi , and d - wa . in this section ,
we first describe the simulation code to explain some key processes
. then we present our simulation results .
2.0pt .wind test runs [ tab03 ] [ cols="<,^,^,^",options="header " , ] the gadget-3 code is a parallel treesph code .
the gravitational force is computed with a hierarchical multipole expansion and the hydrodynamical force with the smoothed particle hydrodynamics ( sph ) technique in the entropy conservative formulation @xcite .
the code adopts radiative cooling and heating by photoionization @xcite .
it also includes star formation and supernova feedback using the sub - resolution multiphase model of the interstellar medium ( ism ) developed by @xcite .
the ism is pictured as a fluid comprised of condensed clouds in pressure equilibrium with an ambient hot medium .
a thermal instability is assumed to be operating in the two - phase region exceeding a threshold density @xmath133 .
star formation occurs in the dense regions , and converts cold clouds into stars on a time - scale constrained by observations @xcite .
the mass fraction of massive stars among the newly formed stars is given by the parameter @xmath24 .
the massive stars are subject to die instantly as supernovae ( with metal enrichment ) releasing energy to the ambient phase .
some cold clouds are supposed to evaporate inside the supernova bubbles .
the evaporation of cold clouds inside of supernova bubbles is handled by the simplified treatment of @xcite ( @xcite ; see sec . 3 of their paper for details ) .
only the density dependency on the supernova evaporation is taken into account for simplicity @xcite .
the radiative cooling of the hot gas causes a corresponding increase of the cold gas .
these processes - star formation , cloud evaporation and cloud growth - lead to a self - regulated cycle for star formation .
the multiphase model is numerically implemented with the simplified treatment described in sec . 5.2 in @xcite based on the fact that the conditions of the self - regulation are achieved quickly . in the treatment
, the mass of each new star particle is set identically to @xmath134 , where @xmath135 is the initial gas mass before the star formation and @xmath136 is number of star particles that may be generated in a gas particle at each star formation event .
the code also includes a model for galactic winds driven by star formation .
the rate of the disk mass loss through a wind is assumed to be proportional to the star formation rate @xmath137 , where @xmath138 is a coefficient measuring the wind efficiency and @xmath139 is the formation rate of long - lived stars .
it is also assumed that the rate of energy loss through the wind is @xmath140 , which is a fraction @xmath141 of the rate of energy generation by supernovae .
here @xmath142 is the velocity of the wind , and @xmath143 represents the average return of the supernova energy from the stars formed following the @xcite initial mass function .
the velocity of a particle is modified to @xmath144 by the wind .
the wind can be axial or isotropic @xcite ; the direction of the unit vector @xmath145 is chosen to be either random on the unit sphere in isotropic winds or as a random orientation along the direction of @xmath146 ( @xmath147 is the gravitational potential ) in axial winds . in all of our simulations , the standard parameters for the multiphase model are used .
they are mostly the same as those in @xcite , otherwise adjusted based on later observations or findings . the star formation time - scale @xmath148 is set to 1.47 gyr , and the mass fraction of massive star to @xmath24 = 0.1 .
the ` supernova temperature ' @xmath149 is chosen to be @xmath150 k which is equivalent to @xmath151 erg/@xmath152 , the temperature of cold clouds @xmath153 to 1000 k , and the parameter for supernova evaporation @xmath154 to 1000 .
we adopt @xmath136 = 2 , so a new star particle has a half the gas particle mass before the star formation event .
correspondingly , the total number of particles increases due to star formation even though the total mass of a system is kept constant . in the wind test runs ( table 3 ) , the effects of galactic winds driven by supernovae are included .
axial winds are tested in runs dhir - wa , dhir - wa - e1 , and d - wa , and isotropic winds in run dhir - wi . the parameter of the wind efficiency is set to the standard value @xmath138 = 2 in runs dhir - wa , dhir - wi , and d - wa , while the half the value in dhir - wa - e1 for comparison .
the wind energy fraction parameter is set to the default value @xmath141 = 1 in the runs with winds .
here we describe the evolution of our galaxy models without winds ( listed in table 1 ) , focusing on the effects of the hot gas halo .
the distributions of particles in each simulation at four epochs are presented in figs 5 through 11 .
we present the snapshots of the runs together with the evolution of the @xmath126-@xmath127 plots ( fig .
3 ) , the density profiles of the star disk , the gas disk , and the gas halo ( fig . 4 ) , and the specific mass and
the particle numbers of the components ( table 4 ) .
we begin with the runs of type dhi models .
l |rrrr|rrr|rrr + + & @xmath49 ( @xmath46 ) & @xmath155 & @xmath156 & @xmath157 & @xmath58 ( @xmath57 ) & @xmath158 & @xmath159 & @xmath74 ( @xmath31 ) & @xmath160 & @xmath161 + + * dhi & & & & & & & & & + @xmath162 = 0 gyr & 16384 ( 4.399 ) & 16384 & 0 & 0 & 16384 ( 0.6000 ) & 16384 & 0 & 32768 ( 1.200 ) & 32768 & 0 + @xmath162 = 0.1 gyr & 16780 ( 4.407 ) & 16384 & 396 & 0 & 16270 ( 0.5888 ) & 16260 & 10 & 32877 ( 1.204 ) &
32758 & 119 + @xmath162 = 0.5 gyr & 18165 ( 4.432 ) & 16384 & 1780 & 1 & 16127 ( 0.5629 ) & 16115 & 12 & 32890 ( 1.204 ) &
32756 & 134 + @xmath162 = 1 gyr & 19678 ( 4.460 ) & 16384 & 3286 & 8 & 15794 ( 0.5359 ) & 15771 & 23 & 32871 ( 1.203 ) &
32745 & 126 + @xmath162 = 2 gyr & 23182 ( 4.524 ) & 16384 & 6756 & 42 & 14089 ( 0.4726 ) & 14057 & 32 & 32849 ( 1.203 ) & 32724 & 125 + & & & & & & & & & + * dhi - f5 & & & & & & & & & + @xmath162 = 0 gyr & 16384 ( 4.399 ) & 16384 & 0 & 0 & 16384 ( 0.6000 ) & 16384 & 0 & 163840 ( 5.998 ) & 163840 & 0 + @xmath162 = 0.1 gyr & 16851 ( 4.408 ) & 16384 & 465 & 2 & 16190 ( 0.5848 ) & 16149 & 41 & 164022 ( 6.004 ) & 163799 & 223 + @xmath162 = 0.5 gyr & 19244 ( 4.452 ) & 16384 & 2813 & 47 & 16074 ( 0.5450 ) & 15904 &
170 & 163912 ( 6.000 ) & 163668 & 244 + @xmath162 = 1 gyr & 24232 ( 4.543 ) & 16384 & 7624 & 224 & 15200 ( 0.4598 ) & 14911 & 289 & 163745 ( 5.994 ) & 163505 & 240 + @xmath162 = 2 gyr & 36350 ( 4.765 ) & 16384 & 19290 & 676 & 9795 ( 0.2411 ) & 9648 & 147 & 163658 ( 5.991 ) & 163416 & 242 + & & & & & & & & & + * dhir & & & & & & & & & + @xmath162 = 0 gyr & 16384 ( 4.399 ) & 16384 & 0 & 0 & 16384 ( 0.6000 ) & 16384 & 0 & 32768 ( 1.200 ) & 32768 & 0 + @xmath162 = 0.1 gyr & 16724 ( 4.405 ) & 16384 & 340 & 0 & 16363 ( 0.5932 ) & 16344 & 19 & 32784 ( 1.200 ) & 32749 & 35 + @xmath162 = 0.5 gyr & 18079 ( 4.430 ) & 16384 & 1689 & 6 & 16212 ( 0.5670 ) & 16172 & 40 & 32823 ( 1.202 ) &
32728 & 95 + @xmath162 = 1 gyr & 19532 ( 4.457 ) & 16384 & 3134 & 14 & 15903 ( 0.5410 ) & 15850 & 53 & 32805 ( 1.201 ) &
32713 & 92 + @xmath162 = 2 gyr & 22621 ( 4.513 ) & 16384 & 6173 & 64 & 14452 ( 0.4856 ) & 14389 & 63 & 32775 ( 1.200 ) & 32687 & 88 + & & & & & & & & & + * dhn & & & & & & & & & + @xmath162 = 0 gyr &
16384 ( 4.399 ) & 16384 & 0 & 0 & 16384 ( 0.6000 ) & 16384 & 0 & 32768 ( 1.200 ) & 32768 & 0 + @xmath162 = 0.1 gyr & 16860 ( 4.408 ) & 16384 & 473 & 3 & 16378 ( 0.5915 ) & 16218 & 160 & 32763 ( 1.199 ) & 32608 & 155 + @xmath162 = 0.5 gyr & 18752 ( 4.443 ) & 16384 & 2173 & 195 & 16438 ( 0.5665 ) & 16015 & 423 & 32499 ( 1.190 ) &
32318 & 181 + @xmath162 = 1 gyr & 21935 ( 4.501 ) & 16384 & 4977 & 574 & 15862 ( 0.5164 ) & 15318 & 544 & 32276 ( 1.182 ) & 32087 & 189 + @xmath162 = 2 gyr & 30479 ( 4.657 ) & 16384 & 12743 & 1352 & 12060 ( 0.3646 ) & 11832 & 228 & 32148 ( 1.177 ) & 31955 & 193 + & & & & & & & & & + * dhn - f5 & & & & & & & & & + @xmath162 = 0 gyr & 16384 ( 4.399 ) & 16384 & 0 & 0 & 16384 ( 0.6000 ) & 16384 & 0 & 163840 ( 5.998 ) &
163840 & 0 + @xmath162 = 0.1 gyr & 17419 ( 4.418 ) & 16384 & 849 & 186 & 17443 ( 0.6214 ) & 16077 & 1366 & 162739 ( 5.957 ) & 162462 & 277
+ @xmath162 = 0.5 gyr & 27734 ( 4.607 ) & 16384 & 7821 & 3529 & 19764 ( 0.5906 ) & 14926 & 4838 & 158422 ( 5.799 ) &
158242 & 180 + @xmath162 = 1 gyr & 43271 ( 4.892 ) & 16384 & 16762 & 10125 & 17101 ( 0.4384 ) & 11364 & 5737 & 154810 ( 5.667 ) &
154646 & 164 + @xmath162 = 2 gyr & 75100 ( 5.474 ) & 16384 & 31139 & 27577 & 8256 ( 0.2252 ) & 1213 & 7043 & 144716 ( 5.298 ) & 144570 & 146 + & & & & & & & & & + * d & & & & & & & & & + @xmath162 = 0 gyr & 16384 ( 4.399 ) & 16384 & 0 & @xmath41 & 16384 ( 0.6000 ) & 16384 & @xmath41 & 0 ( 0.0 ) & @xmath41 & 0 + @xmath162 = 0.1 gyr & 16761 ( 4.406 ) & 16384 & 377 & @xmath41 & 16376 ( 0.5931 ) & 16376 & @xmath41 & 0 ( 0.0 ) & @xmath41 & 0 + @xmath162 = 0.5 gyr & 18839 ( 4.444 ) & 16384 & 2455 & @xmath41 & 16150 ( 0.5551 ) & 16150 & @xmath41 & 0 ( 0.0 ) & @xmath41 & 0 + @xmath162 = 1 gyr & 20767 ( 4.480 ) & 16384 & 4383 & @xmath41 & 15507 ( 0.5198 ) & 15507 & @xmath41 & 0 ( 0.0 ) & @xmath41 & 0 + @xmath162 = 2 gyr & 22860 ( 4.518 ) & 16384 & 6476 & @xmath41 & 14342 ( 0.4815 ) & 14342 & @xmath41 & 0 ( 0.0 ) & @xmath41 & 0 + & & & & & & & & & + * hi & & & & & & & & & + @xmath162 = 0 gyr & 16384 ( 5.000 ) & 16384 & @xmath41 & 0 & 0 ( 0.0 ) & @xmath41 & 0 & 32768 ( 1.200 ) & 32768 & @xmath41 + @xmath162 = 0.1 gyr & 16384 ( 5.000 ) & 16384 & @xmath41 & 0 & 0 ( 0.0 ) & @xmath41 & 0 & 32768 ( 1.200 ) & 32768 & @xmath41 + @xmath162 = 0.5 gyr & 16384 ( 5.000 ) & 16384 & @xmath41 & 0 & 0 ( 0.0 ) & @xmath41 & 0 & 32768 ( 1.200 ) & 32768 & @xmath41 + @xmath162 = 1 gyr & 16384 ( 5.000 ) & 16384 & @xmath41 & 0 & 0 ( 0.0 ) & @xmath41 & 0 & 32768 ( 1.200 ) & 32768 & @xmath41 + @xmath162 = 2 gyr & 16384 ( 5.000 ) & 16384 & @xmath41 & 0 & 0 ( 0.0 ) & @xmath41 & 0 & 32768 ( 1.200 ) & 32768 & @xmath41 + & & & & & & & & & + & & & & & & & & & + * dhir - wa & & & & & & & & & + @xmath162 = 0 gyr & 16384 ( 4.399 ) & 16384 & 0 & 0 & 16384 ( 0.6000 ) & 16384 & 0 & 32768 ( 1.200 ) &
32768 & 0 + @xmath162 = 0.1 gyr & 16720 ( 4.405 ) & 16384 & 336 & 0 & 16054 ( 0.5823 ) & 16040 & 14 & 33089 ( 1.211 ) &
32754 & 335 + @xmath162 = 0.5 gyr & 17521 ( 4.420 ) & 16384 & 1137 & 0 & 15022 ( 0.5338 ) & 15004 & 18 & 34071 ( 1.245 ) & 32750 &
1321 + @xmath162 = 1 gyr & 18137 ( 4.431 ) & 16384 & 1753 & 0 & 14184 ( 0.4985 ) & 14162 & 22 & 34817 ( 1.269 ) & 32746 & 2071 + @xmath162 = 2 gyr & 19124 ( 4.449 ) & 16384 & 2737 & 3 & 12594 ( 0.4441 ) & 12575 & 19 & 36102 ( 1.305 ) & 32749 & 3353 + & & & & & & & & & + * dhir - wa - e1 & & & & & & & & & + @xmath162 = 0 gyr & 16384 ( 4.399 ) & 16384 & 0 & 0 & 16384 ( 0.6000 ) & 16384 & 0 & 32768 ( 1.200 ) &
32768 & 0 + @xmath162 = 0.1 gyr & 16699 ( 4.405 ) & 16384 & 315 & 0 & 16195 ( 0.5877 ) & 16179 & 16 & 32951 ( 1.206 ) &
32752 & 199 + @xmath162 = 0.5 gyr & 17783 ( 4.425 ) & 16384 & 1398 & 1 & 15505 ( 0.5469 ) &
15484 & 21 & 33564 ( 1.227 ) & 32747 & 817 + @xmath162 = 1 gyr & 18642 ( 4.441 ) & 16384 & 2256 & 2 & 14830 ( 0.5150 ) & 14806 & 24 & 34080 ( 1.243 ) & 32744 & 1336 + @xmath162 = 2 gyr & 20218 ( 4.469 ) & 16384 & 3828 & 6 & 13168 ( 0.4580 ) & 13145 & 23 & 35134 ( 1.271 ) & 32743 & 2391 + & & & & & & & & & + * dhir - wi & & & & & & & & & + @xmath162 = 0 gyr & 16384 ( 4.399 ) & 16384 & 0 & 0 & 16384 ( 0.6000 ) & 16384 & 0 & 32768 ( 1.200 ) & 32768 & 0 + @xmath162 = 0.1 gyr & 16752 ( 4.406 ) & 16384 & 368 & 0 & 16117 ( 0.5840 ) & 16110 & 7 & 33026 ( 1.209 ) & 32761 & 265 + @xmath162 = 0.5 gyr & 17609 ( 4.422 ) & 16384 & 1224 & 1 & 14993 ( 0.5314 ) & 14985 & 8 & 34095 ( 1.246 ) & 32760 & 1335 + @xmath162 = 1 gyr & 18268 ( 4.434 ) & 16384 & 1883 & 1 & 14164 ( 0.4958 ) & 14156 & 8 & 34822 ( 1.269 ) & 32760 & 2062 + @xmath162 = 2 gyr & 19434 ( 4.455 ) & 16384 & 3047 & 3 & 12495 ( 0.4369 ) & 12489 & 6 & 36151 ( 1.307 ) & 32761 & 3390 + & & & & & & & & & + * d - wa & & & & & & & & & + @xmath162 = 0 gyr & 16384 ( 4.399 ) & 16384 & 0 & @xmath41 & 16384 ( 0.6000 ) & 16384 & @xmath41 & 0 ( 0.0 ) & @xmath41 & 0 + @xmath162 = 0.1 gyr & 16716 ( 4.405 ) & 16384 & 332 & @xmath41 & 16355 ( 0.5932 ) & 16355 & @xmath41 & 21 ( 0.0007507 ) & @xmath41 & 21 + @xmath162 = 0.5 gyr & 18193 ( 4.432 ) & 16384 & 1809 & @xmath41 & 15395 ( 0.5387 ) & 15395 & @xmath41 & 849 ( 0.02823 ) & @xmath41 & 849 + @xmath162 = 1 gyr & 18947 ( 4.446 ) & 16384 & 2563 & @xmath41 & 14435 ( 0.5019 ) & 14435 & @xmath41 & 1629 ( 0.05125 ) & @xmath41 & 1629 + @xmath162 = 2 gyr & 19417 ( 4.455 ) & 16384 & 3033 & @xmath41 & 13808 ( 0.4824 ) & 13808 & @xmath41 & 2071 ( 0.06211 ) & @xmath41 & 2071 * * * * * * * * * * * here we describe the results for the galaxy models dhi ( the fiducial model ) , dhi - f5 , and dhir .
these models contain a gas halo following the isothermal density profile ( eq .
the snapshots of run dhi are presented in fig .
the first two rows show the particle distribution of the star disk projected in the @xmath94-@xmath95 and @xmath94-@xmath96 planes , the third and fourth rows the gas disk , and the bottom two rows the gas halo , at @xmath162 = 0.1 , 0.5 , 1 , and 2 gyr from the first to the fourth column sequentially . in the first two rows in fig . 5 , the black dots represent the stars originally set as the star disk , and the blue and red dots indicate the stars added into the star disk from the gas originating from ( i.e. initially set as ) the gas disk and gas halo , respectively . in table
4 we give the total mass and number of star disk particles ( @xmath46 and @xmath49 , respectively ) at each time ; we also specify the number of star disk particles that are originally star disk particles ( @xmath155 ) , produced from the disk gas ( @xmath156 ) , and transformed from the halo gas ( @xmath157 ) separately . the total mass of the star disk as a whole is @xmath46 @xmath163 .
similarly , the total number of the star disk particles is @xmath164 .
as time passes , spiral patterns and a central bar develop in the disk . some deviation from the initial exponential surface density profiles
are seen in fig . 4 ( first row - left column ) due to those structures generated .
the star disk grows in mass continuously ( table 4 ) , as some gas turns into disk stars .
the mass increases from the initial value @xmath46 = 4.399 @xmath42 @xmath101 to 4.524 @xmath42 @xmath101 at @xmath162 = 2 gyr .
the total number of star disk particles also increases from @xmath49 = 16384 at the beginning to 23182 at @xmath162 = 2 gyr .
the disk gas particles shown in the middle two rows in fig .
5 are those determined as the disk gas by the criterion in the @xmath126-@xmath127 plane ( first row in fig . 3 ) .
the black dots in the snapshots are the original disk gas particles still remaining in the disk ( i.e. in the @xmath126-@xmath127 plot , the gas particles among cyan dots lie below the criterion line ) .
the magenta dots superimposed on the black dots are the originally hot halo particles that become the disk gas according to the @xmath126-@xmath127 criterion ( i.e. the gas particles among magenta lie below the criterion line ) . in the @xmath126-@xmath127 diagrams at later times ( second to fifth panels of the first row in fig .
3 ) , the disk gas particles , which are located below the black line , follow the finger - like pattern in the higher density region and show the sharp cut - off at @xmath165 k in the lower density region . the negative correlation between @xmath126 and @xmath127 as shown by the finger - like feature results from the effective equation of state for star - forming gas in the multiphase model @xcite .
the sharp cut - off appears because the minimum temperature a gas particle can reach through atomic radiative cooling processes is about 10000 k. as shown in the snapshots , some spiral patterns develop in the gas disk as in the star disk ; warps are also generated as seen in the side view .
the gas disk dissipates as time passes .
although some halo gas particles cool down to become the gas disk particles ( mostly to the central part of the gas disk , as indicated with magenta dots ) , more gas particles turn into the disk stars .
some heated disk gas join the halo .
the number of particles that are initially disk gas particles ( @xmath158 ) and that converted from the halo gas ( @xmath159 ) at each epoch are given in table 4 .
as seen in the surface density profile of the gas disk ( first row - middle column in fig .
4 ) , the size of the disk decreases by dissipation and the profile also evolves .
the density at the innermost part of the disk increases at later times .
the total mass of the gas disk @xmath166 decreases from the initial value @xmath57 = 0.6 @xmath42 @xmath101 to 0.4726 @xmath42 @xmath101 after 2 gyr ( table 4 ) . the black dots in the bottom two rows in fig .
5 are the gas particles that remain as the halo gas from the beginning ( i.e. in the @xmath126-@xmath127 diagram , the gas particles among magenta dots lie above the criterion line ) .
the number of those particles is denoted by @xmath160 in table 4 .
the cyan circles superimposed on the black dots are those particles among originally set as the cold disk gas particles but having the properties of the gas halo according to the @xmath126-@xmath127 criterion ( i.e. the gas particle among cyan lie above the criterion line ) .
the number of these transformed halo gas particles is @xmath161 in table 4 .
the cyan circles at early times are located mostly at the outer part of the gas disk . at later times
, they appear more dispersed in both @xmath94-@xmath95 and @xmath94-@xmath96 views .
the spherically averaged density profile of the gas halo is shown in fig .
4 ( first row - right column ) . at the initial time
, the gas halo follows the isothermal density profile with the halo gas fraction @xmath167 = 0.01 ( black line ) .
it changes a bit ( see olive and orange lines ) from the initial profile mainly in the central part , as some halo gas are removed down to the disk ( which subsequently turn into disk stars ) and the gas particles are redistributed .
the total mass of the gas halo @xmath168 changes only a little from the initial value 1.2 @xmath42 @xmath101 to 1.203 @xmath42 @xmath101 at @xmath162 = 2 gyr .
6 shows the evolution of model dhi - f5 .
the model starts with five times more hot gas ( @xmath89 ) than the fiducial model dhi .
more blue and red dots are seen in the first two rows , compared to the previous run . due to the massive gas halo ,
the gas disk particles ( originating from the gas disk or the gas halo ) are compressed , cooled , and turn into stars more quickly .
consequently , the surface density profile of the star disk ( second row - left column in fig .
4 ) deviates from the initial exponential profile more than that seen in run dhi .
the star disk grows in mass from the initial value @xmath46 = 4.399 @xmath42 @xmath101 to 4.765 @xmath42 @xmath101 at @xmath162 = 2 gyr ( table 4 ) . the gas disk and the gas halo particles of run dhi - f5 plotted in fig .
6 are those determined in accordance with the @xmath126-@xmath127 criterion .
the gas disk dissipates very quickly as clearly seen in the third and fourth rows in fig . 6 and also in the surface density profile ( second row - middle column in fig . 4 ) .
the gas disk diminishes in mass by star formation from the initial value @xmath46 = 0.6 @xmath42 @xmath101 to 0.2411 @xmath42 @xmath101 at @xmath162 = 2 gyr ( table 4 ) .
the gas halo changes mostly within a few tens of kpc as seen in both the snapshots ( fig .
6 ) and the density profile ( fig . 4 ) .
the change is contributed by the halo gas particles in the central part which cool down to the disk ( magenta dots in the snapshots ) as well as the gas particles originating from the gas disk which are heated by the hot gas and moved to the more diffuse gas halo ( cyan circles in the snapshots ) .
overall , the gas halo changes in mass slightly from the initial value of @xmath46 = 5.998 @xmath42 @xmath101 to 5.991 @xmath42 @xmath101 at @xmath162 = 2 gyr ( table 4 ) . four snapshots of run dhir are shown in fig .
the model has a rotating gas halo following the isothermal density profile with @xmath86 .
compared to the first run dhi , fewer number of gas particles originating from the gas disk turn into stars ( blue dots in the snapshots ; also refer to @xmath156 in table 4 ) . on the other hand ,
more gas particles originating from the hot halo cool down to the gas disk ( magenta dots ; @xmath159 ) .
the accreted gas particles mostly appear in the inner part of the disk but rarely in the outer part .
the rotation of the hot halo contributes the gas halo particles to settle down to the disk a bit more efficiently , but it also make the cooling time of the compressed gas particles in the disk longer .
this results in less massive star disk ( @xmath46 ) and more massive gas disk ( @xmath57 ) than those of run dhi ( and also than all the runs of types dh and d without winds ) as shown in table 4 .
the gas disk in this run thus dissipates more slowly than the two previous runs as seen in both the snapshots and its surface density profile ( fig . 4 ) .
fewer number of gas particles originating from the disk are heated and included into the halo ( cyan circles ; @xmath161 ) .
the gas halo keeps its initial mass ( @xmath31 ) for at least a couple of gyr ( table 4 ) .
models dhn and dhn - f5 have a gas halo following the nfw density profile ( eq .
2 ) with @xmath86 and 0.05 , respectively . fig .
8 shows the four snapshots from a run of dhn .
compared to the run with the fiducial model dhi , more stars are added into the disk from the disk and the halo gas ( blue and red dots ; see also table 4 ) .
this is led by the higher density of the inner part of the gas halo in model dhn , as seen in the spherically averaged density profiles ( fig .
4 ) , although the initial halo gas fraction and thus the initial total mass of the hot gas are identical in both models .
compared to the run of dhi - f5 having a heavier gas halo , fewer disk gas particles turn into the stars ( blue dots ; except at @xmath162 = 0.1 gyr as shown in table 4 ) .
this is in part because the less massive gas halo in run dhn exerts overall weaker compression to the gas disk .
note that , however , more halo gas particles become the disk gas ( magenta dots ) in run dhn compared to run dhi - f5 and subsequently turns into stars ( red dots ) , due to the higher central density of the gas halo ( fig . 4 ) , even though the initial total mass of the gas halo is five times less than that of model dhi - f5 .
in this run dhn , the total mass of the star disk increases from @xmath46 = 4.399 @xmath42 @xmath101 at the initial time to 4.657 @xmath42 @xmath101 at @xmath162 = 2 gyr ( table 4 ) . the gas disk and the gas halo decrease in mass from @xmath57 = 0.6 @xmath42 @xmath101 at @xmath162 = 0 to 0.3646 @xmath42 @xmath101 after 2 gyr , and from @xmath31 = 1.2 @xmath42 @xmath101 at @xmath162 = 0 to 1.177 @xmath42 @xmath101 after 2 gyr , respectively .
9 shows the the four snapshots from the simulation of model dhn - f5 .
its gas halo has the highest central density among all our models ( fig . 4 ) .
star formation in the disk ( blue and red dots in fig .
9 ) occurs most actively in this model run , ending up with the most massive star disk of @xmath46 = 5.474 @xmath42 @xmath101 at @xmath162 = 2 gyr which is 124% of its initial mass ( table 4 ) .
the surface density of the star disk in the central part becomes higher than all the other runs ( fig .
the gas disk also dissipates very quickly in this run .
most of the original disk gas has been consumed by @xmath162 = 2 gyr ( black dots in the middle two rows in fig . 9 ; @xmath158 in table
4 ) ; the gas particles seen in the disk at later times are mostly those accreted from the hot halo ( magenta dots ; @xmath159 ) .
the total mass of the gas disk decreases to @xmath57 = 0.2252 @xmath42 @xmath101 at @xmath162 = 2 gyr , which is 38% of its initial mass ( table 4 ) .
the total mass of the gas halo also decreases to 5.298 @xmath42 @xmath101 at @xmath162 = 2 gyr , 88% of the initial mass .
the four snapshots from run d is presented in fig .
the galaxy model d does not have a gas halo at the set - up .
new stars are added into the disk as some gas in the disk turn into stars ( blue dots ) .
none of the gas particles initially set as the disk gas become the halo gas according to the @xmath126-@xmath127 criterion throughout the simulation ( left three panels of the bottom row in fig .
the star disk grows in mass from the initial value of @xmath46 = 4.399 @xmath42 @xmath101 to 4.518 @xmath42 @xmath101 at @xmath162 = 2 gyr ( table 4 ) .
the gas disk mass @xmath57 decreases by the same amount through star formation .
the radial extent of the gas disk is maintained well compared to all runs of type dh models which possess a gas halo , as seen in the snapshots ( fig .
10 ) and in the surface density profile ( fig . 4 ) .
11 shows the snapshots from run hi .
this model is initialized with the gas halo component following the isothermal density profile with the halo gas fraction @xmath86 but does not have a gas disk .
over the course of the simulation , no gas particles initially set as the hot halo gas cool down to the disk according to the @xmath126-@xmath127 criterion ( leftmost panel and two rightmost panels in the bottom row in fig . 3 ) .
consequently , no new stars are formed in the star disk ; the masses of the star disk and the gas halo remain unchanged from the initial values ( table 4 ) .
the surface density profile of the star disk deviates from the initial exponential profile due to the central bar and the spiral patterns developed in the star disk , as seen in fig .
it is also seen that the spherically averaged density profile of the gas halo increases in the central part .
the evolution of the four wind test runs listed in table 3 are discussed here . as noted earlier , galactic winds driven by supernova
are included in these runs .
the temperature @xmath126 of the gas components in the models with respect to the local density @xmath127 are shown in fig .
12 , and the evolution of the density profiles of the star and gas disks and the gas halo in fig . 13 .
the particle distributions of the three components in each run at four epochs are presented in figs 14@xmath16917 . the total mass and number of particles of the components
are given in table 4 .
the galaxy model dhir is used for these three runs .
the initial model contains a gas halo with the isothermal density profile and the halo gas fraction of @xmath86 as in the fiducial model dhi but has an additional rotation for the gas halo .
the three runs are thus identical to one another at the start of the runs , and also to run dhir without winds in sec .
however , they are subject to different evolution due to the different wind parameters . the snapshots of run dhir - wa at @xmath162 = 0.1 , 0.5 , 1 , and 2 gyr are presented in fig .
this run accounts for supernova - driven galactic winds in axial mode with the wind efficiency of 2 as summarized in table 3 . compared to run dhir without winds , new stars are formed less actively in the star disk from the gas disk and the gas halo particles ( blue and red dots in the first two rows ) , as some gas are blown away by the winds .
13 shows that the surface density of the disk at the center does not rise as high as that of run dhir at later times .
the total mass of the star disk grows more slowly than in run dhir , and becomes @xmath46 = 4.449 @xmath42 @xmath101 at @xmath162 = 2 gyr which is only a 1% increase from the initial mass ( table 4 ) .
however , the mass of the gas disk is reduced more than in run dhir .
this is because , although fewer gas particles in the disk turn into stars , more gas is expelled from the disk by the energetic winds ( as seen cyan circles in the bottom two rows ) and also fewer gas particles from the halo settle down onto the disk ( magenta dots in the middle two rows ) .
the phenomena can be seen in the @xmath126-@xmath127 diagrams ( fig .
12 ) as well . in the first row of fig .
12 , more gas particles originating from the disk ( cyan dots ) appear above the criterion line at both @xmath162 = 1 and 2 gyr ( first and third panels ; see also @xmath161 in table 4 ) compared to those of run dhir , and less gas originating from the halo ( magenta dots ) lie below the black line ( second and forth panels ; @xmath159 in table 4 ) . the surface density of the gas disk and the spherically averaged density of the gas halo ( fig .
13 ) show less density in the central region at later times than those of run dhi , due to the gas removal from the center by the winds .
the total mass of the gas halo increases through the gas ejected from the disk while less amount of gas is lost to the disk ; it becomes @xmath31 = 1.305 @xmath42 @xmath101 at @xmath162 = 2 gyr , which is a 9% increase from the initial mass , while that of the run dhir is almost unchanged for a few gyr .
four snapshots of run dhir - wa - e1 are presented in fig .
15 . in this simulation ,
the axial mode of stellar winds is chosen as in the previous run dhir - wa , but with half the strength of the wind efficiency @xmath138 = 1 for comparison . mostly , star formation occurs more actively in this run than in run dhir - wa ; the total star disk mass increases to @xmath46 = 4.469 @xmath42 @xmath101 at @xmath162 = 2 gyr , which is a bit greater than that of run dhir - wa .
the total mass of the gas disk decreases less slowly than that in run dhir - wa , because fewer disk gas are expelled from the disk by the weaker winds ( cyan circles in the snapshots ; refer to @xmath161 in table 4 ; see also the @xmath126-@xmath127 plots in fig .
12 ) , although more disk gas turns into stars than in run dhir - wa ( blue dots ; @xmath156 ) .
the total mass of the gas halo increases to @xmath31 = 1.271 @xmath42 @xmath101 at @xmath162 = 2 gyr , which is smaller than that of run dhir - wa . the surface density of the gas disk ( fig .
13 ) rises at the center at later times than in run dhir - wa . the spherically averaged density of the gas halo changes mainly in the inner part .
four snapshots of run dhir - wi are presented in fig . 16 .
this run adopts the isotropic mode for stellar winds .
the wind efficiency is chosen to @xmath138 = 2 as in run dhir - wa .
the number of newly formed stars in the disk is generally similar as that of run dhir - wa .
the total masses of the star and the gas disks at 2 gyr are also close to those of run dhir - wa . in the snapshots ,
the expelled gas from the disk by the winds shows weaker bipolar pattern ( cyan circles ; also refer to the @xmath126-@xmath127 plots in fig .
12 ) than in run dhir - wa , due to the different wind modes used in the two simulations .
the evolution of surface density profile of the gas disk and the spherically averaged density profile of the gas halo ( fig .
13 ) are not identical to those of run dhir - wa , however , the amount of the ejected gas as well as the mass of each component are close to those of run dhir - wa ( table 4 ) .
this run uses the galaxy model d , which does not have a gas halo . to compare with run d in sec .
3.2.3 without winds , run d - wa accounts for galactic winds throughout the simulation
. axial winds with the wind efficiency @xmath138 = 2 are chosen in this run as in run dhir - wa ( table 3 ) .
the snapshots of run d - wa at @xmath162 = 0.1 , 0.5 , 1 , and 2 gyr are presented in fig .
17 . compared to run d without winds , fewer number of stars are added onto the star disk from the gas disk ( blue dots ; also refer to @xmath170 in table 4 ) .
the gas flow caused by the winds is seen as a bipolar pattern in the snapshots .
some of the gas particles ejected from the plane of the gas disk are classified as the gas disk particles in the snapshots , and the others as the gas halo particles ( cyan circles ) , according to the @xmath126-@xmath127 criterion . at later times ,
most of the ejected particles are defined as the halo gas .
the @xmath126-@xmath127 diagrams ( bottom row in fig .
12 ) show that some of the gas particles originally set as the gas disk are located above the criterion line due to the wind effects ; this is in contrast to the case of run d where no gas particles cross the black line . as in the other runs with winds , the winds prevent gas particles from gathering at the center resulting in the smaller surface density at the center of the disk compared to run d , as seen in fig . 13 .
the total mass of the star disk increases from the initial value @xmath46 = 4.399 @xmath42 @xmath101 to @xmath46 = 4.455 @xmath42 @xmath101 at @xmath162 = 2 gyr ( table 4 ) , and that of the gas disk decreases from the initial value @xmath57 = 0.6 @xmath42 @xmath101 to @xmath46 = 0.4824 @xmath42 @xmath101 at @xmath162 = 2 gyr .
the number of the gas halo particles at @xmath162 = 2 gyr is @xmath74 = 2071 .
18 shows the sfrs of all of our simulations for 3 gyr .
the left column is for the runs without winds , and the right column is for the wind test runs . in run d ,
where the galaxy model starts with a single gas component of the gas disk and does not account for the effects of galactic winds , the sfr ( bottom - left panel ) increases at the beginning for about half gyr and decreases continuously afterwards .
the maximum rate is 1.14 @xmath171 at 0.36 gyr .
the star formation decreases afterwards because the cold and dense gas becomes less available as they are continually consumed by star formation ( refer to table 4 ) . in run hi , where the initial galaxy model has only a hot gas halo , the sfr is zero during the entire run .
the sfrs in the runs of type dh models without winds ( top to fifth panels in the left column ) , where the galaxy models are initialized with both gas disk and gas halo , come out quite differently from that of run d without a gas halo . in the fiducial run dhi
( with an isothermal gas halo of @xmath172=0.01 ) , the sfr does not change much throughout the simulation .
the rate reaches 0.837 @xmath171 at 0.024 gyr , decreases slowly to 0.468 @xmath171 at 0.91 gyr , and increases again to 0.745 @xmath171 at 1.49 gyr .
the sfr is kept above 0.65 @xmath171 until about 2 gyr and decreases afterwards .
compared with run d , the sfr in the fiducial run is lower for about the first 1 gyr . however , the sfr becomes higher during the last 2 gyr than in run d. most of the new stars in this run are generated from the gas originating from the gas disk ( table 4 ) . in run dhi - f5 ( with a more massive isothermal gas halo of @xmath172=0.05 ) , the sfr is higher than in the run with the fiducial model dhi at all times .
the sfr in run dhi - f5 increases steeply to the maximum rate 2.658 @xmath171 at 1.48 gyr and decreases later to 0.9 @xmath171 at the end of the run .
the maximum rate is about three times greater than that in the fiducial run .
the time when the rate reaches its maximum is affected by the adopted value of the model parameter for star formation time - scale @xmath148 = 1.47 gyr .
star formation occurs mostly in the gas originally set as the gas disk and only a few percent of the stars are formed out of the initially halo gas ( table 4 ) .
the sfr in run dhir ( with a rotating isothermal gas halo of @xmath172=0.01 ) is maintained relatively constant throughout the simulation than in all the other runs .
the sfr is very close to that in the fiducial run dhi ( with the same but non - rotating gas halo ) for the first 1 gyr but gets lower in the middle of the run from @xmath1731 gyr to @xmath1732.5 gyr by upto @xmath1730.2 @xmath171 .
as listed in table 4 , the number of stars generated from the gas initially set as the gas disk is smaller than that in the fiducial run at each time ; the number of new stars from the gas originally from the gas halo is a few times larger , however their total mass is still insignificant as @xmath174 1% of the total mass of all new stars .
it is implied that the rotation of the gas halo in this run helps some halo gas particles to be accreted down to the gas disk more easily , however , it affects overall to lower star formation activity than in the run without the initial rotation . in run dhn ( with a nfw gas halo of @xmath172=0.01 )
, the sfr is much higher than that in the fiducial run dhi at all times due to the more centrally concentrated nfw gas halo .
the sfr in run dhn is rather more comparable to that in run dhi - f5 than in run dhi , although the maximum rate 1.866 @xmath171 at 1.72 gyr is certainly lower than in run dhi - f5 .
the percentage of the stars formed out of the gas originating from the gas halo becomes significant in run dhn ; it is a few times greater than that in run dhi - f5 ( table 4 ) . in run dhn - f5 ( with a massive nfw gas halo of @xmath172=0.05 ) , the sfr is higher than in all the other runs throughout the simulation .
the sfr increases very rapidly from the start of the run over 5.5 @xmath171 in 0.35 gyr , and reaches the maximum rate 6.94 at 1.66 gyr . after the maximum
, it decreases rapidly to 2.960 @xmath171 at 2.47 gyr and slows down till the end of the run .
the amount of new stars generated from the halo gas at later times is comparable to that from the disk gas , as the gas set as the gas disk are largely consumed by the times while a significant number of gas particles from the centrally concentrated massive gas halo are still cooled down to the gas disk ( table 4 ) . in the wind
test runs , the sfrs come out always lower than those in the corresponding runs without winds , because the winds disturb cold gas from being concentrated . including axial winds with the wind efficiency @xmath138 = 2 , the sfr in run d - wa ( bottom - right ) has its maximum 0.91 @xmath171 at 0.13 gyr , which is 0.23 @xmath171 less than the maximum rate in run d. after the maximum , the sfr decreases more rapidly than in run d down to 0.147 @xmath171 at 1 gyr and 0.022 at 3 gyr . in run dhir - wa with axial winds and the efficiency @xmath138 = 2
, the sfr is also lower than in run dhir without winds . with the reduced wind efficiency @xmath138 = 1 in run dhir - wa - e1 ,
the sfr becomes overall higher than that in run dhir - wa and lower than that in run dhir . in run dhir - wi adopting isotropic winds with the wind efficiency @xmath138 = 2 , the sfr comes out close to that in run dhir - wa with axial winds .
motivated by the existence of hot gaseous halo in spiral galaxies evidenced by both observations and simulations , we have investigated the effects of a gas halo on the evolution of isolated milky way - like galaxies using @xmath0-body / hydrodynamic simulations .
we have also examined the effects of galactic winds driven by supernovae on the evolution . to generate the initial galaxy models
, we used the zeno software package .
we built seven different galaxy models , with or without a gas halo component with varying mass and density profile for comparison ( tables 1 and 2 ) : we tried two different density profiles for the gas halo - an isothermal profile and an nfw profile .
we also tested different amounts of the halo gas fraction - @xmath167 = 0.01 and @xmath167 = 0.05 .
in addition , we studied a case with an initially rotating gas halo .
we set the disk gas fraction @xmath175 = 0.12 equally in all models possessing a gas disk . each galaxy model
was then evolved in isolation using the @xmath0-body / sph code gadget-3 ( an early version ) , first without including stellar winds driven by star formation .
the effects of winds are also studied . in the wind test runs , either axial winds or isotropic winds
were tested , with the wind efficiency of either @xmath138 = 2 or @xmath138 = 1 .
we showed that the effects of the hot gas component on the evolution are significant , particularly in the gas disk dissipation / accretion and in the star formation activity .
the runs with a gas halo show higher sfrs in the middle to last phase of the 3 gyr - simulations than in the run without a gas halo where the sfr reaches its maximum at the early phase and then decreases continuously thereafter .
the star formation activity in the disks is affected by both the mass concentration at the inner part and the total mass of the gas halo .
the more centrally concentrated nfw gas halo leads higher sfrs than in runs with the isothermal gas halo of the same total mass . in runs where the gas halo has higher density at the central halo and/or greater total mass , the sfrs are higher than others and the gas disks dissipate too quickly .
the gas accretion from the halo onto the disk in those runs turns out to occur not much actively , and some portion of the accreted gas are quickly consumed by star formation as well , resulting in unrealistically small gas disks after a few gyr .
it is not clear which physical processes are mainly responsible for the dissipation of gas disks .
models dhi ( the fidicial model ) and dhir ( with the additional rotation of the gas halo to the fiducial model ) show relatively reasonable results in both the gas dissipation and the star formation activity .
the sfr in run dhi is a bit higher than that in run dhir in the middle of the evolution .
the mass of the gas disk is slightly greater in run dhir .
it turned out that the initial rotation of the gas halo leads more gas accretion from the halo to the disk but delays the star formation activity in the disk .
the specific mechanism of how the rotation hinders star formation , however , needs to be further investigated .
with model dhir and model d we performed four more simulations including supernova - driven galactic winds . in these wind test runs ,
the sfrs come out lower than the corresponding runs without winds at all times as expected .
both axial and isotropic winds result in overall similar sfrs .
the run with the reduced wind efficiency shows higher sfr than that with the default value of the wind efficiency . in our initial galaxy models , we kept the total mass of the system as well as the total masses of the halo ( dm + gas ) , disk ( stars + gas ) , and bulge components equally to focus on examining the impacts of the different gas distribution and rotation of the gas halo .
our simulation results , such as in the gas dissipation and accretion and the star formation activities , are restricted by our initial galaxy models and the parameter values in the simulation code .
the wind effects would also be affected by a variety of factors , for instance the total mass of the galaxy model , and be limited by the provided wind model in the code . nonetheless , numerical studies to seek more realistic galaxy models including a gas halo component are demanded since the component as a large reservoir of gas influences the evolution of the galaxies .
adding a gas halo in numerical simulations , various gas dynamical processes that may occur in real systems both isolated or interacting with others , would be better interpreted .
the galaxy models are eventually compared with the ` galaxies ' in cosmological simulations @xcite .
we plan to use our isolated galaxy models to understand the morphology and luminosity transformation of interacting galaxies @xcite found in the sloan digital sky survey @xcite .
we are grateful to joshua e. barnes and volker springel for offering the zeno package and the gadget code , respectively .
h. thanks walter dehnen , juhan kim , and maurice h. p.
m. van putten for helpful discussions .
we thank the korea institute for advanced study for providing computing resources ( kias center for advanced computation linux cluster system ) for this work . here
we describe the procedure for constructing the models possessing both gas disk and gas halo and then disk - only or halo - only models , using several programs provided in the zeno package ( c.f .
@xcite ) . to construct the @xmath0-body realization of the spherical components of the models of type dh ,
we first obtain the cumulative mass profiles of each component by integrating the corresponding mass density profiles with the parameter values shown in table 4 .
for the bulge , we use the program _
gammagsp _ in the zeno package .
this code generates profiles for dehnen models @xcite , including the hernquist profile ( eq . 4 ) with @xmath176 in the input options .
we truncate the mass distribution of the bulge from the radius @xmath34 as indicated in eq . 4 and rescale to the original mass , using the program _
gsptrun_. for the profile of either dm or gas halo following the nfw model ( eq .
2 ) , we run _
halogsp_. the profile of the gas halo following the non - singular isothermal form ( eq . 3 ; type dhi ) is generated by _
isothgsp_. the total spheroid mass profile and the total mass profile are calculated in accordance with the choice of the halo profile .
the disk is represented by an infinitely thin disk at this stage ( see @xcite for detail ) , and the profile of the ` spherical ' ( infinitely thin ) disk is generated by _
expdgsp_. we then combine the mass profiles of the bulge , the dm and the gas halos , and the ` spherical ' disk to get the total cumulative mass profile using _ gspadd _ , and smooth the profile using _ gspsmooth_. finally , we generate each of the spherical components , running two @xmath0-body realization codes , _ gsprealize _ for the bulge and the dm halo and _ gspsphere _ for the gas halo , with the input parameters of the density profile of the component as well as the smoothed total mass profile for the gravity calculation .
the _ gsprealize _ code uses the abel integral to compute the distribution function and generates a configuration of the body in phase - space .
the _ gspshere _ code outputs the initial positions and internal energies of the halo gas particles with zero initial velocities . to make realizations of the star and the gas disks we run _ gspdisk _ twice
this code makes an exponential disk embedded in the smoothed total spheroid mass profile with the disk mass and other parameters given in table 4 . for the initial velocities of the star disk particles , the local circular velocities and the vertical and radial dispersions
are calculated in the way described in @xcite .
for the gas particles , we set zero dispersion .
the procedure for making model dhir ( disk plus halo model with rotation ) is the same as in model dhi ( the fiducial model ) , except an additional step for giving rotation of the gas halo using a simple code we made .
we set the rotation such that the gas halo has the rotation curve with a similar shape but half the amplitude of that of the disk . to build either the disk - only model ( model d ) or the halo - only model ( model hi ) , one can follow the steps described above but skipping the parts related to either the gas halo or the gas disk , respectively .
19 shows four more snapshots from run dhir at later times since those in fig . 7 .
more stars are seen in the star disk than in fig .
the inner part of the gas disk appears largely dissipated as time passes , and the refilling of the gas disk by accreting halo gas does not seem to occur actively . the total mass of the star disk increases to @xmath46 = 4.56 @xmath42 @xmath101 at @xmath162 = 3 gyr , and 4.59 @xmath42 @xmath101 at @xmath162 = 5 gyr .
among the stars in the disk at @xmath162 = 3 gyr , the number of stars added from the gas originally set as the gas disk and the gas halo are @xmath156 = 8600 ( blue dots in the first two rows ) and @xmath157 = 94 ( red dots ) , respectively .
the total mass of the gas disk decreases to @xmath57 = 0.44 @xmath42 @xmath101 at @xmath162 = 3 gyr , and 0.41 @xmath42 @xmath101 at @xmath162 = 5 gyr .
the total mass of the gas halo is maintained to @xmath31 = 1.2 @xmath42 @xmath101 at @xmath162 = 3 and 5 gyr .
anderson , m. e. , & bregman , j. n. 2010 , do hot halos around galaxies contain the missing baryons ?
, , 714 , 320 barnes , j. e. , & hernquist , l. 1992 , dynamics of interacting galaxies , , 30 , 705 barnes , j. e. , & hibbard , j. e. 2009 , identikit 1 : a modeling tool for interacting disk galaxies , , 137 , 3071 blumenthal , g. r. , faber , s. m. , primack , j. r. , & rees , m. j. 1984 , formation of galaxies and large - scale structure with cold dark matter , , 311 , 517 choi , j .- h . , & nagamine , k. 2011 , multicomponent and variable velocity galactic outflow in cosmological hydrodynamic simulations , , 410 , 2579 choi , y .- y . , han , d .- h .
, & kim , s. s. 2010 , korea institute for advanced study value - added galaxy catalog , jkas , 43 , 191 dehnen , w. 1993 , a family of potential - density pairs for spherical galaxies and bulges , , 265 , 250 dekel , a. , & silk , j. 1986 , the origin of dwarf galaxies , cold dark matter , and biased galaxy formation , , 303 , 39 deng , x .- f . , zhang , f. , song , j. , chen , y .- q . , & jiang , p. 2012 , the environmental dependence of the fraction of ` unconventional ' galaxies : faint red and luminous blue , jkas , 45 , 59 hernquist , l. 1990 , an analytical model for spherical galaxies and bulges , , 356 , 359 hernquist , l. 1993 , @xmath0-body realizations of compound galaxies , , 86 , 389 hopkins , p. f. , quataert , e. , & murray , n. 2011 , self - regulated star formation in galaxies via momentum input from massive stars , , 417 , 950 hwang , h. s. , & park , c. 2009 , evidence for morphology and luminosity transformation of galaxies at high redshifts , , 700 , 791 katz , n. , weinberg , d. h. , & hernquist , l. 1996 , cosmological simulations with treesph , , 105 , 19 kazantzidis , s. , magorrian , j. , & moore , b. 2004 , generating equilibrium dark matter halos : inadequacies of the local maxwellian approximation , , 601 , 37 kennicutt , r. c. jr .
1998 , the global schmidt law in star - forming galaxies , , 498 , 541 kim , j. , park , c. , rossi , g. , lee , s. m. , & gott , j. r. 2011 , the new horizon run cosmological @xmath0-body simulations , jkas , 44 , 217 mckee , c. f. , & ostriker , j. p. 1977
, a theory of the interstellar medium : three components regulated by supernova explosions in an inhomogeneous substrate , , 218 , 148 mcmillan , p. j. , & dehnen , w. 2007 , initial conditions for disc galaxies , , 378 , 541 merritt , d. 1985 , distribution functions for spherical galaxies , , 214 , 25p moster , b. p. , macci , a. v. , somerville , r. s. , naab , t. & cox , t. j. 2011 , the effects of a hot gaseous halo in galaxy major mergers , , 415 , 3750 moster , b. p. , macci , a. v. , somerville , r. s. , naab , t. & cox , t. j. 2012 , the effects of a hot gaseous halo on disc thickening in galaxy minor mergers , , 423 , 2045 navarro , j. f. , frenk , c. s. , & white , s. d. m. 1996 , the structure of cold dark matter halos , , 462 , 563 osipkov , l. p. 1979
, spherical systems of gravitating bodies with an ellipsoidal velocity distribution , pisma astr .
, 5 , 77 park , c. , & choi , y .- y .
2009 , combined effects of galaxy interactions and large - scale environment on galaxy properties , , 691 , 1828 park , c. , gott , j. r. , & choi , y .- y .
2008 , transformation of morphology and luminosity classes of the sdss galaxies , , 674 , 784 park , c. , & hwang , h. s. 2009 , interactions of galaxies in the galaxy cluster environment , , 699 , 1595 salpeter , e. e. 1955 , the luminosity function and stellar evolution , , 121 , 161 springel , v. 2005 , the cosmological simulation code gadget-2 , , 364 , 1105 springel , v. , di matteo , t. , & hernquist , l. 2005 , modelling feedback from stars and black holes in galaxy mergers , , 361 , 776 springel , v. , & hernquist , l. 2002 , cosmological smoothed particle hydrodynamics simulations : the entropy equation , , 333 , 649 springel , v. , & hernquist , l. 2003 , cosmological smoothed particle hydrodynamics simulations : a hybrid multiphase model for star formation , , 339 , 289 toomre , a. , & toomre , j. 1972 , galactic bridges and tails , , 178 , 623 | we construct several milky way - like galaxy models containing a gas halo ( as well as gaseous and stellar disks , a dark matter halo , and a stellar bulge ) following either an isothermal or an nfw density profile with varying mass and initial spin .
in addition , galactic winds associated with star formation are tested in some of the simulations .
we evolve these isolated galaxy models using the gadget-3 @xmath0-body / hydrodynamic simulation code , paying particular attention to the effects of the gaseous halo on the evolution .
we find that the evolution of the models is strongly affected by the adopted gas halo component , particularly in the gas dissipation and the star formation activity in the disk .
the model without a gas halo shows an increasing star formation rate ( sfr ) at the beginning of the simulation for some hundreds of millions of years and then a continuously decreasing rate to the end of the run at 3 gyr . whereas the sfrs in the models with a gas halo , depending on the density profile and the total mass of the gas halo , emerge to be either relatively flat throughout the simulations or increasing until the middle of the run over a gigayear , then decreasing to the end .
the models with the more centrally concentrated nfw gas halo show overall higher sfrs than those with the isothermal gas halo of the equal mass .
the gas accretion from the halo onto the disk also occurs more in the models with the nfw gas halo , however , this is shown to take place mostly in the inner part of the disk and not to contribute significantly to the star formation unless the gas halo has very high density at the central part .
the rotation of a gas halo is found to make sfr lower in the model .
the sfrs in the runs including galactic winds are found to be lower than the same runs but without winds .
we conclude that the effects of a hot gaseous halo on the evolution of galaxies are generally too significant to be simply ignored .
we also expect that more hydrodynamical processes in galaxies could be understood through numerical simulations employing both gas disk and gas halo components . |
nuclear dynamics possess various aspects depending on mass regions , excitation energies , and so on . in light - weight nuclei
, it is known that clustering plays a significant role in the features of ground and excited states @xcite . on the other hand , in heavier nuclei ,
the clustering effects are not clear , though many theoretical @xcite and experimental @xcite studies have been conducted . in the @xmath7-shell region , the focus should be on proton - rich @xmath8 nuclei , because such nuclei can have a clustered structure comprising stable nucleus . moreover , proton - rich nuclei have a large radius and stronger coulomb repulsion , and may also derive a cluster structure . in the scope of this research ,
the features of cluster structures in @xmath0ca are a key issue , because they are the heaviest @xmath9 stable nuclei , and many experimental data exist for this nuclei . in this paper
, we have studied the structure of @xmath0ca as a starting point to understand the structures of medium- and heavy - weight @xmath8 nuclei .
@xmath0ca has a typical double closed - shell structure nucleus and has a spherical ground state .
however , it is known that many kinds of deformed band appear in low energy regions . the first @xmath10 band built on the @xmath11 state ( 3.35 mev )
is considered to be a normal - deformed ( nd ) state and the dominant configuration is @xmath12-@xmath13 @xcite . the @xmath14 band built on @xmath15 ( 5.25 mev ) exists just above the @xmath10 band .
it has been suggested that the nd band deforms triaxially and has the @xmath16 side band due to triaxiality @xcite .
the @xmath2-@xmath3ar cluster structure has been studied for a long time , because @xmath0ca is an analogue of @xmath17o , which has an @xmath2-@xmath4c cluster structure in the first @xmath10 band , as a double closed - shell nuclei .
the local potential model @xcite and @xmath2-@xmath3ar orthogonal condition model ( ocm ) @xcite have been performed theoretically .
et al . _ suggested that the first @xmath10 band ( nd ) has an @xmath2-@xmath3ar structure , and predicted that its parity - doublet @xmath18 band and @xmath2-@xmath3ar higher nodal band exist in highly excited states @xcite .
_ obtained the @xmath14 band as well as the @xmath10 and @xmath19 states using the @xmath2-@xmath3ar ocm , and succeeded in reproducing @xmath20 transition strengths @xcite .
experimentally , the @xmath2-@xmath3ar structure is studied through the @xmath3ar(@xmath21li,@xmath22)@xmath0ca reaction @xcite .
the states in these @xmath10 and @xmath19 bands are populated by the @xmath2-transfer reactions and have large @xmath2 spectroscopic factors @xcite . in the experiments we describe , the @xmath2-@xmath3ar higher - nodal states
were also observed @xcite .
it has been suggested that the states in the @xmath10 rotational band built on the @xmath23 state ( 5.21 mev ) have a @xmath24-@xmath25 configuration @xcite , and have been observed during experimental work searching for the @xmath24-@xmath25 states with @xmath26s(@xmath4c , @xmath2)@xmath0ca reactions @xcite . due to the strong population in the multi - nucleon transfer data and the strong @xmath20 transitions @xcite , the @xmath27 ( 5.21 mev ) , @xmath28 ( 5.63 mev ) and @xmath29 ( 6.54 mev ) bands
have been thought to belong to the superdeformed ( sd ) band with the dominant @xmath24-@xmath25 configuration .
recently , by using gammasphere array detectors , the level structure of the deformed bands in @xmath0ca has been explored and many excited states up to high spin have been discovered .
this band was thus confirmed as the sd band @xcite .
motivated by these the experimental observations , many theoretical microscopic studies on deformed states of @xmath0ca have been performed recently with the methods of skyrme - hartree - fock ( shf ) @xcite , shf - bcs + gcm @xcite , spherical - basis amd @xcite and the shell model @xcite .
performed cranked shf calculations without assuming axial symmetry , though energy levels were not calculated @xcite .
performed shf - bcs + gcm calculations @xcite .
although they calculated energy levels and quadrupole transition strengths in the nd and sd bands , they could not study triaxiality nor side bands because they assumed axial symmetry . in these studies ,
the relationship between deformed states and cluster structure was not discussed .
it has been suggested that the sd state forms a @xmath4c-@xmath5si - like cluster structure in spherical - basis amd @xcite . within the spherical - basis amd
, triaxiality does not appear in nd nor sd states .
the purpose of the present study is to understand the clustering and triaxial deformations in the low energy states of @xmath0ca in a unified manner .
we use the framework of amd + gcm .
the basis functions of gcm are obtained by energy variation after parity projection with constraints .
we adopted two kinds of constraints .
one is a constraint on the quadrupole deformation parameter @xmath30 ( @xmath30-constraint ) and the other is on the distance @xmath22 between clusters centers of mass ( @xmath22-constraint ) .
it has already been proven that the @xmath22-constraint is useful for obtaining various kinds of clustering wave function , which are not computed within a simple @xmath30-constraint @xcite .
for example , the @xmath31be(@xmath32)-@xmath4c cluster structure in @xmath33ne is calculated with a @xmath22-constraint but not with the @xmath30-constraint .
also in the case of @xmath0ca , many kinds of cluster structure , for example @xmath2-@xmath3ar , @xmath31be(@xmath32)-@xmath26s and @xmath4c-@xmath5si , can be calculated with the @xmath22-constraint , although no cluster structure is obtained in @xmath0ca with the @xmath30-constraint .
we superposed mean - field - type and cluster - type wave functions calculated with @xmath30- and @xmath22-constraints respectively , and calculated energies and @xmath20 transition strength .
we analyzed the superposed wave functions in order to investigate clustering and triaxial deformations .
this paper is organized as follows . in the next section ( [ sec : formulation ] ) , we explain the framework of this study . the calculated results and discussions are presented in [ sec : results ] , and lastly , we present a summary in [ sec : summary ] .
we used the theoretical framework of amd + gcm . in the present study ,
the amd wave function is a slater determinant of triaxially deformed gaussian wave packets ( deformed - basis amd ) , [ amd ] @xmath34 , \nonumber\\ \\ { | \chi_i \rangle } & = & \alpha_i { | \uparrow \rangle } + \beta_i { | \downarrow \rangle},\\ { | \tau_i \rangle } & = & { | p \rangle}\ { \rm or}\ { | n \rangle}. \end{aligned}\ ] ] here , the complex parameters @xmath35 , which represent the centroids of the gaussian in phase space , take independent values for each single particle wave function .
the width parameters @xmath36 , @xmath37 and @xmath38 are real parameters and take independent values for each of the @xmath39- , @xmath40- and @xmath41-directions , but are common for all nucleons .
the spin part @xmath42 is parametrized by @xmath43 and @xmath44 and the isospin part @xmath45 is fixed as @xmath46 ( proton ) or @xmath47 ( neutron ) .
the quantities @xmath48 are variational parameters and are optimized by energy variation as explained in the next subsection .
the trial wave function in the energy variation with constraints is a parity - projected wave function , @xmath49 where @xmath50 is parity and @xmath51 is the parity operator . in this study
, we will discuss positive parity states .
the hamiltonian is , @xmath52 where @xmath53 and @xmath54 are the kinetic energy and the energy of the center of mass motion respectively , and @xmath55 is the effective nucleon - nucleon interaction .
we have used gogny d1s force ( d1s ) and skyrme sly7 force ( sly7 ) in the present work .
the coulomb force @xmath56 is approximated by a sum of seven gaussians .
we performed energy variation and optimized the variational parameters included in the trial wave function ( eqs .
( [ amd ] ) ) to find the state that minimizes the energy of the system @xmath57 , @xmath58 here , we add the constraint potential @xmath59 to the expectation value of hamiltonian @xmath60 in order to obtain the minimum energy state under the optional constraint condition . in this study , we employed two types of constraint , which are on the quadrupole deformation parameter @xmath30 ( @xmath30-constraint ) and the distance between clusters centers of mass , @xmath22 ( @xmath22-constraint ) by employing the potential @xmath59 , @xmath61 here @xmath30 is the matter quadrupole deformation parameter , which is defined in ref . , and @xmath62 is the distance between the clusters centers of mass @xmath63 and @xmath64 , @xmath65 @xmath66 where @xmath67 is the mass number of cluster @xmath64 and the expression @xmath68 means that the @xmath69th nucleon is contained in cluster c@xmath70 . it should be noted that the @xmath71 component of the spatial center of the single - particle wave function @xmath72 is @xmath73 . when sufficiently large values are chosen for @xmath74 and @xmath75 , the resultant values @xmath30 and @xmath76 of energy variation become @xmath77 , @xmath78 , respectively .
we constrained the @xmath79 and @xmath80values for the @xmath22-constraint . in each calculation of energy variation , we constrained one of @xmath30 , @xmath79 and @xmath80 .
a detailed explanation regarding the @xmath22-constraint may be found in ref . .
the energy variation with the amd wave function is carried out using the frictional cooling method @xcite .
the time evolution equation for the complex parameters @xmath81 and @xmath44 is @xmath82 where @xmath83 is @xmath81 or @xmath44 , and that for the real parameters @xmath84 and @xmath38 is @xmath85 the quantities @xmath86 and @xmath87 are arbitrary positive real numbers .
the energy of the system decreases as time progresses , and after a sufficient number of time steps , we obtain the minimum energy state .
after the constrained energy variation for @xmath88 , we superposed the optimized wave functions employing the quadrupole deformation parameter @xmath30 and the distances between the centers of mass among clusters @xmath89 for @xmath90 configurations as the generator coordinate , @xmath91 where @xmath92 is the parity and total angular momentum projection ( amp ) operator , and @xmath93 and @xmath94 are optimized wave functions with @xmath30- and @xmath89-constraints for the constrained values @xmath95 and @xmath96 respectively .
the integrals over the three euler angles included in @xmath97 are evaluated by numerical integration .
the coefficients @xmath98 and @xmath99 are determined by the hill - wheeler equation , @xmath100 we defined squared overlap @xmath101 to estimate the overlap of a specific model space in the superposed wave function .
suppose the non - orthonormalized wave functions @xmath102 span the functional space @xmath103 , for example , @xmath104 , @xmath105 or @xmath106 .
orthonormalized wave functions @xmath107 are obtained by performing unitary transformations from @xmath102 , @xmath108 using the @xmath109 , squared overlap between @xmath110 and the functional space @xmath103 , @xmath101 is defined as @xmath111
- , @xmath79- and @xmath80-constraints .
the energies are plotted as functions of quadrupole deformation @xmath30 .
the solid ( dotted ) line shows the results for d1s ( sly7 ) .
+ ( b ) : projection of the energy curves onto the @xmath30-@xmath112 plane .
the solid ( dotted ) line shows the results for d1s ( sly7).,scaledwidth=45.0% ] .expectation values of kinetic and potential energies around the ground state , nd and sd minima .
@xmath113 , @xmath114 , @xmath115 and @xmath116 denote kinetic energy , central , spin - orbit and coulomb potentials . [
cols="<,^,^,^,^,^,^,^",options="header " , ] the theoretical and experimental values of quadrupole moments deduced from the @xmath20 strengths for the @xmath117 and @xmath118 bands are listed in table [ tab : comparison ] .
the results of other theoretical studies are also shown in the table .
the intrinsic quadrupole moment @xmath119 is defined by the @xmath6 value as follows .
@xmath120 where @xmath121 is the clebsh - gordan coefficient .
the theoretical quadrupole moments of the nd and sd states are around 130 @xmath122 and 220 @xmath122 respectively . the values are almost unchanged up to high spin .
this means that the nd and sd bands consist of approximately rigid rotor states .
these results are consistent with the rigid rotor - like spin dependence of the excitation energies of the nd and sd states shown in fig .
[ fig : moment ] .
the theoretical @xmath119 moments are consistent with experimental values evaluated from life - time measurements and branching ratios ( i ) except for the @xmath123 transition in the nd state . the small value of the experimental @xmath119 moment for the @xmath124 transition might suggest some structure change in high spin states in the nd band .
we compared our results with those of other theoretical studies .
the @xmath119 values in ( 4 ) @xmath2-@xmath3ar ocm are consistent with the present values for the nd states .
this may indicate that the deformation of the nd states in our calculations is similar to that calculated with the @xmath2-@xmath3ar cluster model . in ( 1 )
the spherical - basis amd and ( 3 ) the shell model , smaller @xmath119 values for the @xmath10 states were obtained by comparison to our model . in ( 2 ) shf - bcs + gcm , the @xmath119 values are also smaller than our results for @xmath10 states .
in particular , the @xmath119 moments for the @xmath125 transition in both the nd and sd states are remarkably small .
this shows that the nd and sd bands obtained in the shf - bcs + gcm calculation are not rigid rotor - like .
this seems to be inconsistent with experimental results regarding the rigid rotor - like property , which has been found in rotational energies and @xmath20 transitions . in recent experimental observations ,
the @xmath119 values have been obtained by globally fitting all transitions .
they ( ( ii)-a and b ) are smaller than the values of the life - time measurement ( i ) and also smaller than the present results .
further analysis requires more detailed measurements of the @xmath20 transition strengths .
_ triaxial rotor _ in this section
we discuss the characteristics of the side bands , @xmath14 , in the nd and sd states . in a naive collective model ,
a @xmath14 band can be described by either a triaxial rotor or a @xmath112 vibration . in order to identify the mode of the side bands , it is useful to analyze the ratio of @xmath20 strength for @xmath14 to @xmath10 transitions , @xmath126 where @xmath127 is the value of quadrupole electric transition strength from a @xmath128 state in the @xmath129 band to a @xmath130 state in the @xmath131 band .
the @xmath132 value is set to the constant value @xmath133 in the @xmath112 vibration mode , while it is a function of the static @xmath112 parameter in the triaxial rotor mode @xcite , @xmath134 \right/ \frac{\sin^2 ( 3\gamma)}{9 - 8\sin^2 ( 3\gamma ) } .\ ] ] strength @xmath132 as a function of @xmath112 value .
the dot - dashed line and the solid line show the limiting values of the @xmath112-vibration and the triaxial rotor respectively .
the @xmath135 and @xmath136 values are the theoretical @xmath132 values for the nd and sd states derived from the @xmath20 transition strengths respectively , while @xmath137 is that of the experiment for the nd state .
the deformation parameters @xmath138 and @xmath139 for the main components of the nd and sd states are also shown .
, scaledwidth=45.0% ] the function @xmath140 is plotted as a function of @xmath112 value in fig .
[ fig : davydov ] .
the value of @xmath141 is 0.7 for the axial symmetric limit , @xmath142 and @xmath143 .
as the triaxiality increases , @xmath141 decreases , going to zero at the peak of triaxiality , @xmath144 .
regarding the nd state , the theoretical value @xmath132 of the @xmath20 strength ratio is nearly equal to 0 , and the experimental value @xmath145 is also close to 0 .
both the values agree well with @xmath146 but are far from that of the @xmath112 vibration limit , 0.7 . in the present results , the @xmath112 value of the main component of the nd state is @xmath147 , which is consistent with @xmath148 for the triaxial rotor derived from the experimental @xmath132 value .
we therefore conclude that the mode of the observed @xmath149 band is a triaxial rotor , not a @xmath112 vibration .
regarding the sd state , the theoretical value of @xmath132 is equal to 0.51 , and the @xmath112 value of the main component of the theoretical sd state is @xmath150 .
these values , @xmath151 can be described by the relation @xmath152 in eq .
( [ r ] ) for the triaxial rotor .
this indicates that the case of @xmath153 in the sd state is the side band of the triaxial rotor .
the @xmath153 band has not yet been observed experimentally . in order to confirm the triaxiality of the sd band ,
observations of the @xmath153 band and the @xmath20 strength for the inter - band transitions are required .
we investigated the ground state and excited states of @xmath0ca in the framework of deformed - basis amd focusing on the aspects of triaxiality and clustering in the deformed states .
superposing mean - field - type and cluster structure - type wave functions obtained with @xmath30- and @xmath22-constraints respectively , we obtained the ground states , nd , sd , oblate and @xmath2-@xmath3ar higher - nodal states .
we found both the nd and sd bands are constructed from triaxially deformed shapes , and we obtained the @xmath14 side bands of these bands .
the theoretical @xmath6 and moments of inertia are consistent with experimental data , although the excitation energies are higher than empirical values . by analyzing quadrupole moments
calculated from the @xmath6 and the @xmath20 strength ratio @xmath132 values , we found that the nd and sd states can be understood by a relatively rigid rotor conception of triaxial deformations . the nd band and it s side band contain approximately 40% of the @xmath2-@xmath3ar cluster structure component , and the sd band and it s side band contain approximately 60% of the @xmath4c-@xmath5si cluster structure component .
the @xmath2-@xmath3ar higher - nodal band was obtained due to the excitation of inter - cluster motion between the @xmath2 and @xmath3ar clusters .
the present results suggested that cluster correlation will be important for deformation and excitation even in medium- and heavy - weight nuclei .
this work has been supported by jsps research fellowships for young scientists .
the numerical calculations were carried out on sx8 at yitp in kyoto university and on sx5 and sx8 at rcnp in osaka university .
the authors would like to thank dr .
inakura and dr .
fujiwara for fruitful discussions .
this work is supported by a grant - in - aid for the 21st century coe `` center for diversity and universality in physics '' from the ministry of education , culture , sports , science and technology of japan ( mext ) .
99 h. horiuchi , k. ikeda and y. suzuki , prog .
52 , 89 ( 1972 ) , and references therein .
y. fujiwara , h. horiuchi , k. ikeda , m. kamimura , k. kat@xmath154 , y. suzuki and e. uegaki , prog .
68 , 29 ( 1980 ) , and references therein .
f. michel , s. ohkubo and g. reidemeister , prog .
132 , 7 ( 1998 ) , and references therein .
t. sakuda and s. ohkubo , prog .
132 , 103 ( 1998 ) , and references therein .
t. yamaya , k. katori , m. fujiwara , s. kato and s. ohkubo , prog .
132 , 73 ( 1998 ) , and references therein .
w. j. gerace and a. m. green , nucl .
phys . * 93 * , 110 ( 1967 ) .
w. j. gerace and a. m. green , nucl
a123 * , 241 ( 1969 ) . w. j. gerace and j. p. mestre , nucl .
a285 * , 253 ( 1977 ) . k. f. pal and r. g. lovas , phys . lett . * b96 * , 19 ( 1980 ) .
s. ohkubo and k. umehara , prog .
. phys . * 80 * , 598 ( 1988 ) .
g. reidemeister , s. ohkubo and f. michel , phys .
c41 , 63 ( 1990 ) .
t. ogawa , y. suzuki and k. ikeda , prog .
. phys . * 57 * , 1072 ( 1977 ) .
t. sakuda and s. ohkubo , phys . rev . * c49 * , 149 ( 1994 ) . t. yamaya , m. saito , m. fujiwara , t. itahashi , k. katori , t. suehiro , s. kato , s. hatori and s. ohkubo , phys . lett . *
b306 * , 1 ( 1993 ) .
t. yamaya , m. saitoh , m. fujiwara , t. itahashi , k. katori , t. suehiro , s. kato , s. hatori and s. ohkubo , nucl . phys . *
a573 * , 154 ( 1994 ) .
r. middleton , j. d. garrett and h. t. fortune , phys .
b39 * , 339 ( 1972 ) .
j. r. macdonald , d. h. wilkinson and d. e. alburger , phys .
c3 * , 219 ( 1971 ) , and references therein .
e. ideguchi _
lett . * 87 * , 222501 ( 2001 ) .
t. inakura , s. mizutori , m. yamagami and k. matsuyanagi , nucl . phys .
* a710 * , 261 ( 2002 ) . m. bender , h. flocard and p. -h .
phys . rev .
* c68 * , 044321 ( 2003 ) .
y. kanada - enyo and m. kimura , phys . rev . *
c72 * , 064322 ( 2005 ) .
e. caurier , j. menendez , f. nowacki , a. poves , phys . rev . *
c75 * , 054317 ( 2007 ) .
y. taniguchi , m. kimura and h. horiuchi , prog .
. phys . * 112 * , 475 ( 2004 ) .
a. dote , h. horiuchi and y. kanada - enyo , phys . rev .
* c56 * , 1844 ( 1997 ) . y. kanada - enyo and h. horiuchi , prog .
. phys . * 93 * , 115 ( 1995 ) .
j. a. cameron and b. singh , nuclear data sheets 102 , 293 ( 2004 ) . c. chiara _ et al .
_ , phys . rev . *
c67 * , 041303 ( 2003 ) .
m. kimura , h. horiuchi , nucl .
a767 , 58 ( 2006 ) .
a. s. davydov and g. f. filippov , nucl
* 8 * , 237 ( 1958 ) . | we have studied the positive - parity states of @xmath0ca using antisymmetrized molecular dynamics ( amd ) and the generator coordinate method ( gcm ) . imposing two different kinds of constraints on the variational calculation
, we have found various kinds of @xmath1 structures such as a deformed - shell structure , as well as @xmath2-@xmath3ar and @xmath4c-@xmath5si cluster structures .
after the gcm calculation , we obtained a normal - deformed band and a superdeformed band together with their side bands associated with triaxial deformation .
the calculated @xmath6 values agreed well with empirical data .
it was also found that the normal - deformed and superdeformed bands have a non - negligible @xmath2-@xmath3ar cluster component and @xmath4c-@xmath5si cluster component , respectively .
this leads to the presence of an @xmath2-@xmath3ar higher - nodal band occurring above the normal - deformed band . |
nuclear water vapor megamasers currently provide the only direct means to map gas in active galactic nuclei ( agn ) on size scales of @xmath20.11 pc .
nearly all of the observational work on h@xmath3o megamasers to date has focused on the @xmath4 rotational transition at 22.235 ghz from the ortho - h@xmath3o molecule .
more than 160 galaxies have been detected in this line so far , the result of some @xmath24000 galaxies surveyed ( e.g. , @xcite ) .
about 130 of the detections are associated with agn , where they are called megamasers because of their large apparent luminosities .
the physical conditions that give rise to maser activity at 22 ghz are also compatible with masing in other transitions of the h@xmath3o molecule , many of which fall in the submillimeter wavelength band ( @xcite ; @xcite ) .
@xcite presented the first observations of h@xmath3o megamaser emission in a transition other than the 22 ghz , detecting maser emission at 183 ghz and ( tentatively ) at 439 ghz towards the galaxy ngc 3079 .
this galaxy had previously been known to host strong 22 ghz masers , with vlbi observations confirming that the 22 ghz emission originates from the galactic nucleus ( @xcite , @xcite ) .
though the signal - to - noise of the ( sub)millimeter detections ( @xmath27@xmath5 for the 183 ghz transition ) was too low to permit detailed study , the maser emission appears to arise from several narrow ( spectrally unresolved ) features spanning a velocity range comparable to that of the 22 ghz emission . the 183 ghz transition was also detected towards arp 220 by @xcite , where it displays a very broad ( @xmath2350 km s@xmath6 ) and almost featureless spectral line structure .
interestingly , this galaxy has not been detected in 22 ghz emission ( e.g. , ) , suggesting that the masing gas has a low density ( @xmath7 @xmath8 ) and temperature ( @xmath9 k ) . from consideration of these physical conditions and the observed line width ,
@xcite interpret the 183 ghz masers in this galaxy as likely originating from a large number ( @xmath210@xmath10 ) of dense molecular cores rather than being associated with the galactic nuclei .
more recently , @xcite used alma to detect 321 ghz h@xmath3o megamaser emission towards the circinus galaxy , another strong 22 ghz nuclear megamaser host ( e.g. , @xcite ) .
the sensitivity of the circinus observation was sufficient to showcase the richness of the high - frequency maser spectrum , opening up for the first time the possibility of using submillimeter masers in ways that had heretofore been restricted to the 22 ghz transition . in this paper
we report the first detection of submillimeter maser emission from ngc 4945 , and we present a new calibration of the maser spectrum for the circinus galaxy .
we note that @xcite offer a parallel analysis of the ngc 4945 data presented here .
the observations and data reduction procedures are described in [ observationsandreduction ] , and in [ discussion ] we discuss the submillimeter emission and compare the 321 ghz masers to those at 22 ghz . throughout this paper
we quote velocities using the optical definition in the heliocentric reference frame .
we have analyzed archival cycle 0 alma observations of five galaxies that are known to have strong ( peak @xmath11 mjy ) 22 ghz water maser emission associated with a central agn : ngc 1068 @xcite , ngc 1386 @xcite , ngc 4945 @xcite , circinus @xcite , and ngc 5793 @xcite .
all targets were observed at a rest - frame frequency of 321.226 ghz ( alma band 7 ) , which corresponds to the @xmath12 rotational transition of ortho - h@xmath3o at an energy of @xmath13 k above ground .
ngc 5793 was further observed at a rest - frame frequency of 325.153 ghz , corresponding to the @xmath14 rotational transition of para - h@xmath3o at an energy of @xmath15 k above ground .
the total bandwidth for each dual - polarization observation was 1.875 ghz , which was split into 3840 channels spaced contiguously every 0.488 mhz ( corresponding to a velocity resolution of @xmath20.5 km s@xmath6 ) .
the longest baselines for these observations were @xmath2360 meters ( corresponding to a typical resolution of @xmath2@xmath16 ) , and there were between 18 and 25 antennas present ( see table [ tab : info ] ) .
we obtained datasets and initial calibration scripts from the alma archive ; all post - processing reduction , imaging , and spectral analysis was done using the common astronomy software applications package ( casa ) .
table [ tab : info ] lists the observing parameters for each galaxy .
we detected and imaged continuum emission for all five sources ( shown in figure [ fig : continuum_contours ] ) , and in ngc 4945 the continuum was strong enough for self - calibration .
two of the galaxies circinus and ngc 4945 also host 321 ghz maser emission ; we self - calibrated the circinus data using the line emission .
lcccccc & & 14:13:09.906 & 13:05:27.279 & 02:42:40.770 & 03:36:46.237 + & & @xmath1765:20:20.468 & @xmath1749:28:04.44 & @xmath1700:00:47.84 & @xmath1735:59:57.39 + & & 434 & 563 & 1137 & 868 + & 2012 jun 01 & 2012 jun 03 & 2012 jun 03 & 2012 jun 03 & 2012 jun 06 & 2012 aug 24 + & 321.226 & 325.153 & 321.226 & 321.226 & 321.226 & 321.226 + & 6.3 & 21.0 & 19.1 & 15.3 & 15.8 & 11.6 + & 1.35 & 0.40 & 0.55 & 0.60 & 0.54 & 0.64 + & 21 & 20 & 18 & 18 & 20 & 25 + & titan & titan & titan & titan & uranus & uranus + & 3c 279 & 3c 279 & 3c 279 & 3c 279 & 3c 454.3 & 3c 454.3 + & j1517243 & j1517243 & j13295608 & j1325430 & j0339017 & j040336 + & @xmath18 & @xmath19 & @xmath20 & @xmath21 & @xmath22 & @xmath23 + & 48 & @xmath24 & @xmath25 & 24 & 32 & 82 + & 9.8 & 7.2 & 12.6@xmath26 & 9.9 & 7.4 & 9.2 + & 0.39 & 0.29 & 0.48 & 3.0 & 0.42 & 0.36 + & 2.5 & 2.5 & 2.0 & 5.0 & 1.5 & 1.0 + & 10.8 & 18.8 & 90.8 & 733 & 47.1 & 4.3 + & 2.1 & 2.4 & 5.6 & 26.7 & 4.5 & 0.36@xmath27 + & @xmath28 & @xmath29 & & @xmath30 & &
+ initial imaging was performed using casa task ` clean ` with natural uv weighting ; after using ` uvcontsub ` ( specifying line - free channels ) to remove the continuum contribution , we separately imaged the line and continuum emission .
we then performed several iterations of phase - only self - calibration , using the @xmath2400 spectral channels with the strongest emission ( @xmath31 mjy , corresponding to the velocity range @xmath2500700 km s@xmath6 ) to determine the phase solutions .
we found that a solution interval of 1 minute ( averaging both polarizations ) was optimal , yielding sufficiently continuous solutions ( i.e. , consecutive phase solution jumps of @xmath32 ) to confidently interpolate the phases .
the calibration solutions were then applied to both the line and continuum data using ` applycal ` , and we stopped iterating self - calibration once there was no noticeable increase in signal - to - noise ratio ( snr ) .
we found that additional amplitude self - calibration did not improve the snr , so we have retained the phase - only calibrations for analysis .
the resulting continuum image is shown in figure [ fig : continuum_contours ] , and the spectrum extracted from the ( spatially unresolved ) line - only data cube is shown in figure [ fig : circinus_321ghz_spectra ] .
the maser emission in ngc 4945 is not sufficiently strong for self - calibration , so we used the continuum emission instead .
because the continuum emission in ngc 4945 is spatially resolved , we only used the longest baselines ( @xmath33 m , corresponding to the unresolved , point - like nuclear component of the emission ) to determine the phase solutions that were then applied to the spectral line data ; no such baseline restrictions were imposed when self - calibrating the continuum image itself .
we used a solution interval of 30 seconds , averaging both polarizations . despite repeated iterations of self - calibration ,
the sensitivity of the continuum image from this snapshot " observation remains dynamic - range limited ( see @xcite ) .
the resulting noise level of @xmath34 mjy beam@xmath6 is thus larger than what one would nominally expect from a sensitivity calculation .
the rest of the reduction procedure matches what was done for circinus ( see [ circinusreduction ] ) . to account for the sizable ( @xmath2@xmath35 ) offset of the emission center from the phase center
, we used ` impbcor ` to apply a primary beam correction prior to extracting a spectrum from the data cube .
the continuum image and spectrum for ngc 4945 are shown in figures [ fig : continuum_contours ] and [ fig : ngc4945_321ghz_spectra ] , respectively .
the continuum structures for ngc 5793 and ngc 4945 are both elongated in one direction ( spanning @xmath24@xmath36 pc in ngc 5793 , and @xmath29@xmath37 pc in ngc 4945 ) , and both appear to have substantial substructure .
both of these galaxies are edge - on spirals , and the elongation axes of the submillimeter continua are aligned with the large - scale optical major axes ( @xcite ; @xcite ) . the continuum in ngc 4945
is also resolved along the minor axis , spanning @xmath21.5@xmath38 pc .
all of this indicates that the continuum emission in these galaxies traces the galactic disks , rather than originating from , e.g. , a molecular torus region around the central agn ( though there may be a contribution to the emission in the centermost regions from such material ) . at these wavelengths
( @xmath39 @xmath40 m ) , the continuum in ngc 5793 and ngc 4945 is likely dominated by optically thin thermal ( i.e. blackbody ) emission from large dust grains ( see , e.g. , ; ) .
the spectral energy distribution ( sed ) of such emission is typically modeled as a modified blackbody function ( e.g. , ) , with the free parameters being the optical depth @xmath41 , the dust temperature @xmath42 , and the power - law index of the dust opacity @xmath43 . with only a single sed point per galaxy
, we must assume fiducial values for two of these parameters ( e.g. , @xmath43 and @xmath42 ) to allow for a measurement of the third ( e.g. , @xmath41 ) .
further assumptions are then necessary to convert the optical depth to , e.g. , a total interstellar medium ( ism ) gas mass , @xmath44 . fortunately , @xcite have developed an empirical calibration of the relationship between dust emission and ism gas mass .
as long as the emission is measured in the rayleigh - jeans tail , the authors found that the calibration is relatively insensitive to whether the ism is dominated by atomic or molecular gas , if the galaxy is normal or undergoing a starburst , or whether the dust lies in the inner or outer regions of the galaxy .
rewritten in a suitable form , the conversion is given by @xmath45 here , @xmath46 is the luminosity distance to the galaxy , @xmath47 is the observing wavelength , @xmath48 is the observed flux density , @xmath49 is the boltzmann constant , @xmath50 is the dust opacity per unit mass of ism , and @xmath42 is the dust temperature .
most of the underlying physics here is contained in @xmath50 , which the authors calibrated using _ planck _ data to be @xmath51 when calculating gas masses , we use the results from to fix @xmath52 , and we adopt a dust temperature of @xmath53 k ( following @xcite ) .
we measure the total continuum flux density for each galaxy using a circular aperture centered on the continuum peak , and we estimate the uncertainty using the dispersion of integrated flux densities measured in 15 non - overlapping , identical apertures that are offset from the continuum emission in the same image .
the results from these measurements are presented in the bottom portion of table [ tab : info ] .
the gas masses estimated from the 321 ghz and 325 ghz observations of ngc 5793 are broadly consistent , while the estimate for ngc 4945 is somewhat lower .
ngc 1068 , ngc 1386 , and circinus all show continuum emission considerably more centrally - concentrated than in ngc 5793 and ngc 4945 , so it is likely that agn contributions to the continua for these galaxies are not negligible . disentangling the thermal ( i.e. , blackbody ) and nonthermal ( e.g. , electron - scattered synchrotron , free - free ) components of the emission is nontrivial , and requires multi - frequency observations ( see , e.g. , @xcite ) . for ngc 1068
, we can compare our observations to those of , who used alma to map the continuum at 349 ghz down to a @xmath54 level of @xmath55 mjy beam@xmath6 .
though this is a factor of @xmath23 more sensitive than the map presented in this paper , we see consistent continuum structure and amplitude in the circumnuclear region ( i.e. , the region containing emission stronger than our sensitivity threshold ) between the two observations .
the 321 ghz maser detection in ngc 4945 which represents the first time such emission has been seen in this galaxy is considerably fainter than in circinus ( figure [ fig : ngc4945_321ghz_spectra ] ) .
individual maser features are detected at the @xmath245@xmath5 level , though the entire complex between 650 km s@xmath6 and 750 km s@xmath6 is detected at @xmath2@xmath56 in integrated intensity .
we have calculated an isotropic luminosity using @xmath57 here , @xmath46 is the distance to the galaxy , @xmath58 is the line rest frequency , and @xmath59 is the flux density as a function of velocity @xmath60 .
scaled to convenient units , this equation becomes @xmath61 adopting a distance to ngc 4945 of 3.7 mpc @xcite , the observed flux of @xmath62 jy km s@xmath6 corresponds to an isotropic luminosity of @xmath63 l@xmath64 .
though the flux density of individual features is down by a factor of @xmath2100 from what is observed at 22 ghz ( e.g. , @xcite ) , the isotropic luminosity is only lower by a factor of @xmath210 .
insofar as we are able to discern spectral structure , we see that it appears to match reasonably well with previous observations of ngc 4945 at 22 ghz ( top panel of figure [ fig : ngc4945_321ghz_spectra ] has been reproduced from @xcite ) .
the increasing feature strength with increasing velocity and the overall appearance of 23 dominant features are both reminiscent of the 22 ghz spectra . however , the 321 ghz features at @xmath2687 km s@xmath6 and @xmath2726 km s@xmath6 ( with possibly a third at @xmath2660 km s@xmath6 ) do nt map one - to - one with regions of 22 ghz emission . rather , and quite intriguingly , the 321 ghz peaks fall precisely where the 22 ghz emission drops off .
unlike with circinus , the origin of the 22 ghz emission from ngc 4945 is not yet well understood .
@xcite made a vlbi map of ngc 4945 at 22 ghz using the southernmost antennas of the vlba , and they found the spatial distribution of the masers to be approximately linear and distributed across @xmath250 mas ( @xmath20.9 pc ) from one end to the other .
this in particular the roughly symmetric location of redshifted and blueshifted emission to either side of the systemic velocity is suggestive of masers situated in an accretion disk .
the limited antennas available for mapping such a low declination source ( @xmath65 ) resulted in the map being rather incomplete ( i.e. , there were several systemic and blueshifted features that were too faint to map ) , but it is the best available for this source .
when measuring the positions of the 321 ghz maser spots , we found them to be spatially coincident ( within the measurement uncertainties ) .
if the intrinsic distribution of the 321 ghz masers matches that of the 22 ghz masers , this is consistent with what we would expect for the @xmath20.5@xmath66 beam and low signal - to - noise of the observations .
working under the assumption that the 321 ghz emission traces material with the same kinematics as the 22 ghz , the observed 321 ghz features correspond only to the redshifted gas in the accretion disk .
if the 321 ghz spectral structure follows that of the 22 ghz emission , then the undetected blue and systemic features would be slightly below our detection threshold .
the low signal - to - noise in the current observations precludes any detailed characterization of this system , which must await higher sensitivity , better angular resolution observations than those presented here .
@xcite discovered the 321 ghz maser in circinus .
here we re - examine the data , using strong maser lines to apply phase self - calibration ( see [ circinusreduction ] ) .
the new calibration improves the snr by a factor of @xmath22 compared the initial analysis
. published 22 ghz spectra of circinus ( e.g. , top panel of figure [ fig : circinus_321ghz_spectra ] , reproduced from @xcite ) show that the bulk of the maser emission occupies velocities between @xmath2250650 km s@xmath6 more or less contiguously , though often with a notable paucity of features near the systemic velocity .
@xcite ( hereafter g03 ) observed circinus between 1997 and 1998 using the australia telescope long baseline array .
they detected two populations of masers , one arising from a warped accretion disk and the other associated with a wide - angle , bipolar outflow .
the 321 ghz masers are weaker in flux density by a factor of @xmath230100 compared to their 22 ghz counterparts .
although the maser flux at 22 ghz is subject to interstellar scintillation @xcite , this effect should be almost completely absent at 321 ghz ( at such a high frequency , the diffractive scale of the turbulence will be much larger than the fresnel scale ; see @xcite ) . at a distance to the galaxy of 4.2 mpc ( measured by @xcite using the tully - fisher relation )
, the observed flux of @xmath67 jy km s@xmath6 corresponds to an isotropic luminosity ( via equation [ eqn : isotropicluminosity ] ) of @xmath2104 l@xmath64 ; this is roughly a factor of four larger than the isotropic luminosity of the 22 ghz masers ( e.g. , @xcite ) .
the 321 ghz and 22 ghz spectra share broad structural similarities .
both have maser emission spanning comparable total velocity ranges and consolidated primarily into two groups located on either side of the systemic velocity , and in both cases the blueshifted group of features is weaker and sparser than the redshifted group .
we can also see that the region around the systemic velocity in the 321 ghz spectrum is devoid of obvious features either because no maser features exist at these velocities , or because they are below our detection threshold which is reminiscent of the same segment of the 22 ghz spectrum .
the vlbi maps from g03 show that the extent of the 22 ghz maser emission in circinus is roughly @xmath68 mas ( @xmath2@xmath69 pc ) , but as with ngc 4945 the 321 ghz maser spots are spatially co - located within our measurement uncertainties .
though the absolute astrometric precision for alma observations is typically limited to @xmath20.05 arcseconds without taking special calibration steps ( @xcite ; ) , the relative uncertainty in point - source position within the same primary beam ( as a fraction of the half - power beam width ) is inversely proportional to the snr ( see , e.g. , ) .
future high - resolution alma observations should thus have little difficulty mapping the 321 ghz masers in circinus . in lieu of a high angular resolution map
, we can use the information contained in the spectrum to glean some understanding of the spatial distribution of the masers . by applying a threshold proximity of 1 mas between any individual maser spot and the disk midline , g03
were able to assign a rough classification to each maser as originating from either the disk or the outflow .
in doing so , they found that the outflow masers dominated the emission between @xmath2300600 km s@xmath6 , and that disk maser emission dominated blueward of @xmath2300 km s@xmath6 and redward of @xmath2600 km s@xmath6 ( see figure 6 in their paper ) . with this picture from g03 as a guideline , we can compare the spectral distribution of the 22 ghz masers to that of the 321 ghz masers .
their similar overall spectral structure suggests that the 321 ghz and 22 ghz masers are tracing roughly the same material .
this is to be expected from consideration of the physical conditions required for strong maser activity in these transitions . @xcite
have performed a thorough exploration of the relevant parameter space ( i.e. , gas density , kinetic temperature , and dust temperature ) and found that the 321 ghz transition shares an optimal gas density ( @xmath70 @xmath8 ) and collisional pumping scheme ( i.e. , low dust temperature ) with the 22 ghz transition , though it prefers a somewhat larger kinetic temperature of @xmath71 k ( compared to @xmath72 k for the 22 ghz transition ) .
this could explain the apparent excess of 321 ghz maser emission between @xmath2650750 km s@xmath6 , which is not typically seen in 22 ghz spectra ( though we note that 22 ghz emission has been seen out to velocities as large as @xmath2900 km s@xmath6 , albeit with a much lower flux density than the bulk of the emission ; see @xcite ) . under this interpretation
, the 321 ghz emission redward of 650 km s@xmath6 originates in the accretion disk at radii interior to where 22 ghz emission is found .
we see no features in the 321 ghz spectrum between @xmath2300500 km s@xmath6 , which is a spectral range dominated by outflow emission at 22 ghz .
either the 321 ghz emission does not trace the outflow at all or the 321 ghz outflow masers in this velocity range are much fainter than their 22 ghz counterparts ( i.e. , down by a larger factor from the higher - velocity emission to either side ) . if the 321 ghz masers in fact only trace the disk emission , then the features detected between @xmath2500600 km s@xmath6 indicate that some of these masers must originate farther out in the disk than the 22 ghz masers .
higher angular resolution observations will be able to discern whether any of the 321 ghz maser originate in the outflow or if they are all associated with the disk .
the possible high - velocity maser features seen at @xmath21070 km s@xmath6 and @xmath21130 km s@xmath6 in the 321 ghz spectrum coincide with an atmospheric line ( see figure [ fig : circinus_321ghz_spectra ] ) and may represent elevated noise .
we present 321 ghz alma observations of ngc 5793 , ngc 1068 , ngc 1386 , ngc 4945 , and the circinus galaxy .
all galaxies are detected in continuum emission , and circinus and ngc 4945 also display h@xmath3o megamaser emission . for ngc 4945
these data represent the first detection of submillimeter megamaser activity , while for circinus we confirm the results of @xcite , with an updated calibration . in both cases
the 321 ghz spectra appear structurally comparable to those of the 22 ghz masers .
the continuum emission in ngc 5793 and ngc 4945 is well - resolved and spatially extended along the optical major axes of these galaxies , which are both edge - on spirals .
this continuum is likely dominated by thermal emission from dust grains in the disk , and we use the observed fluxes to derive approximate ism masses . for the other three galaxies ,
the continuum emission is centrally - concentrated and thus likely contains a substantial non - thermal component from the agn .
though the 22 ghz maser emission in circinus is associated with both the accretion disk and a molecular outflow , it is unclear whether the 321 ghz emission traces both environments or just the disk .
furthermore , we have reason to believe that the 321 ghz masers trace the accretion disk at smaller radial separations from the central smbh than the mapped 22 ghz masers do .
this can be confirmed by future alma observations of circinus , which should seek to obtain a map of the maser features at the highest possible angular resolution .
we thank the anonymous referee for useful comments .
the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc .
this paper makes use of the following alma data : 2011.0.00121.s .
alma is a partnership of eso ( representing its member states ) , nsf ( usa ) and nins ( japan ) , together with nrc ( canada ) and nsc and asiaa ( taiwan ) and kasi ( republic of korea ) , in cooperation with the republic of chile .
the joint alma observatory is operated by eso , aui / nrao and naoj .
this research has made use of the nasa / ipac extragalactic database ( ned ) , which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . | we present 321 ghz observations of five agn from alma cycle 0 archival data : ngc 5793 , ngc 1068 , ngc 1386 , ngc 4945 , and the circinus galaxy .
submillimeter maser emission is detected for the first time towards ngc 4945 , and we present a new analysis of the submillimeter maser system in circinus . none of the other three galaxies show maser emission , though we have detected and imaged the continuum from every galaxy . both ngc 4945 and circinus are known to host strong ( @xmath0 jy ) 22 ghz megamaser emission , and vlbi observations have shown that the masers reside in the innermost @xmath1 parsec of the galaxies .
the peak flux densities of the 321 ghz masers in both systems are substantially weaker ( by a factor of @xmath2100 ) than what is observed at 22 ghz , though the corresponding isotropic luminosities are more closely matched ( within a factor of @xmath210 ) between the two transitions .
we compare the submillimeter spectra presented here to the known 22 ghz spectra in both galaxies , and we argue that while both transitions originate from the gaseous environment near the agn , not all sites are necessarily in common . in circinus , the spectral structure of the 321 ghz masers indicates that they may trace the accretion disk at radii interior to the 22 ghz masers .
the continuum emission in ngc 4945 and ngc 5793 shows a spatial distribution indicative of an origin in the galactic disks ( likely thermal dust emission ) , while for the other three galaxies the emission is centrally concentrated and likely originates from the nucleus . |
we revisit a classical problem : the transmission through a linear array of many identical slabs ( glass plates , plastic transparencies , or the like ) with random separation , as depicted in fig .
[ fig : stack ] .
the transmission probability that stokes derived in 1862 @xcite on the basis of ray - optical arguments ( thereby improving on an earlier attempt by fresnel in 1821 ; see refs .
@xcite and @xcite for the history of the subject ) is not correct because there are crucial interference effects that require a proper wave - optical treatment .
just that was given by berry and klein in 1997 @xcite who found that the average of the logarithm of the transmission probability through @xmath0 slabs is equal to @xmath0 times the logarithm of the single - slab transmission probability , @xmath1 here , @xmath2 is the probability of transmission through a single slab and @xmath3 denotes the transmission probability for @xmath0 slabs .
its implicit dependence on the random phases that originate in the random spacing of the slabs is averaged over , indicated by the @xmath4 notation . as emphasized in ref .
@xcite , the disorder is crucial ; without it , most wavelength components would be transmitted , and the stack should then appear rather transparent , but this is not the case as a simple experiment with a stack of transparencies demonstrates @xcite . identical slabs , each with single - slap transmission probability @xmath2 . the stack as a whole has transmission probability @xmath3 , which depends on the phases that result from the random spacing of the slabs .
we are interested in @xmath5 , the transmission probability of the stack averaged over the @xmath6 phases . ]
it is indeed common to average logarithms because they are known to be `` self averaging '' @xcite , and the exact result ( [ eq : logaver ] ) is truly remarkable .
but one should realize what it tells us about the average transmission probability @xmath7 itself . as a consequence of the inequality @xmath8 the berry
klein relation ( [ eq : logaver ] ) amounts to a lower bound on the average transmission probability , @xmath9 as we shall see below , this bound is not particularly tight because there is a very large range of individual @xmath3 values . in particular , we note that the ray - optics result @xcite @xmath10 is consistent with ( [ eq : bk - lowlim ] ) .
it is the objective of the present contribution to report good wave - optics estimates for @xmath11 and closely related quantities .
in particular , we will improve on the lower bound of ( [ eq : bk - lowlim ] ) and supplement it with an upper bound .
we observe that the upper bound , when used as an approximation for @xmath7 , is unreasonably good and seems to give us the exact asymptotic values of quantities such as @xmath12 or @xmath13 . at present , this coincidence of the upper bound with exact asymptotic values is a poorly understood mystery .
amplitudes on both sides of the @xmath14th slab .
the unitary scattering matrix @xmath15 of ( [ eq : s ] ) relates the incoming amplitudes @xmath16 and @xmath17 to the outgoing amplitudes @xmath18 and @xmath19 , whereas the transfer matrix @xmath20 of ( [ eq : t ] ) connects the amplitudes on the left with the amplitudes on the right . ] for a wave of wavelength @xmath21 , the wave functions to the left and to the right of the @xmath14th slab are @xmath22 where @xmath23 is the position of the left edge and @xmath24 is the thickness of the slab ; see fig .
[ fig : slab ] .
the incoming amplitudes are related to the outgoing amplitudes by the unitary _ scattering matrix _ @xmath15 , @xmath25 where the entries of @xmath15 are restricted by @xmath26 which account for the single - slab transmission probability @xmath2 and the unitary nature of @xmath15 .
the particular values of the complex phases of @xmath27 , @xmath28 , @xmath29 , and @xmath30 are of secondary interest , but we note that we have @xmath31 and @xmath32 for a completely transparent , non - scattering slab , for which @xmath33 . the _ transfer matrix _
@xmath20 is used to express the amplitudes on the right in terms of the amplitudes on the left , @xmath34 the one - to - one relation between @xmath15 and @xmath20 implies that the transfer matrix is of the form @xmath35 0 & { \mathalpha{\mathrm{e}}^{\mbox{\footnotesize$-{\mathalpha{\mathrm{i}}}\beta$}}}\end{array}\right ) \left(\begin{array}{cc } \cosh\theta & \sinh\theta \\[1ex ] \sinh\theta & \cosh\theta \end{array}\right ) \left(\begin{array}{cc}{\mathalpha{\mathrm{e}}^{\mbox{\footnotesize${\mathalpha{\mathrm{i}}}\beta'$ } } } & 0\\ 0 & { \mathalpha{\mathrm{e}}^{\mbox{\footnotesize$-{\mathalpha{\mathrm{i}}}\beta'$}}}\end{array}\right),\ ] ] where @xmath36 and @xmath37 , @xmath38 , @xmath39 are phase factors that have fixed values which , however , are largely irrelevant for what follows .
the transfer matrix for the gap of length @xmath40 between the @xmath14th slab and the @xmath41th slab is the diagonal phase matrix @xmath42 0 & { \mathalpha{\mathrm{e}}^{\mbox{\footnotesize$-{\mathalpha{\mathrm{i}}}kl_n$}}}\end{array}\right)\,.\ ] ] phase matrices of the same structure sandwich the central @xmath43-dependent matrix in ( [ eq : gent ] ) , so that we have @xmath44 \sinh\theta & \cosh\theta \end{array}\right)\ ] ] as a more useful way of writing @xmath20 .
the product of two transfer matrices is another transfer matrix , whereby the relevant observation is the composition law @xmath45 with @xmath46 determined by @xmath47 and the phases @xmath48 and @xmath49 by @xmath50 whereas ( [ eq : gammas ] ) is of no consequence for the following considerations , the rule ( [ eq : newtheta ] ) is of central importance .
we now turn to the situation of fig . [
fig : stack ] , where we have @xmath0 identical slabs separated by gaps @xmath51 , @xmath52 ,
, @xmath53 that are not controlled on the scale set by the wavelength @xmath21 .
therefore , we regard the phase factors @xmath54 as random with a uniform distribution on the unit circle in the complex plane . the over - all transfer matrix @xmath55 is characterized by @xmath56 which is obtained by repeated application of the composition rule ( [ eq : newtheta ] ) , whereby the phases @xmath57 have random values .
each experimental realization of the @xmath0-slab stack of fig .
[ fig : stack ] has different values for these random phases , and the transmission probability @xmath58 varies from one experiment to the next .
we need to average over the @xmath6 random phases to find @xmath7 . regarding the meaning of @xmath59 in ( [ eq : recurrence ] ) and ( [ eq : faver ] ) .
the random phases @xmath60 , @xmath61 , , @xmath62 have already been averaged over , but the averaging over the phases @xmath63 , , @xmath64 is yet to be performed . ]
let us consider a somewhat more general question : what is the average value @xmath65 of a function of @xmath66 , and thus of a function of @xmath3 ?
when the averaging is carried out successively , first averaging over @xmath60 , then over @xmath61 , and so forth , finally over @xmath64 , we have an intermediate value @xmath59 after averaging over the first @xmath67 random phases ; see fig .
[ fig : recursion ] . here
@xmath68 denotes the value of @xmath69 for the stack of slabs @xmath14 to @xmath0 with its dependence on the remaining phases @xmath63 , ,
we then have @xmath70 for the value prior to any averaging , and ( [ eq : newtheta ] ) tells us that we get @xmath71 from @xmath72 by means of @xmath73 and the integration covers any convenient interval of @xmath74 . eventually this takes us to @xmath75 when the recursive averaging over the @xmath6 random phases is completed . for illustration , we take @xmath76 as a first example .
the recurrence relation ( [ eq : recurrence ] ) yields @xmath77 , so that @xmath78 or @xmath79 when stated in terms of transmission probabilities .
a second illustrating example is @xmath80 , for which @xmath81^{n-1}f_1^{\ } ( c')\ , , \nonumber\\ f_n^{\ } ( c)&=&\frac{2}{3}\biggl[\frac{3}{2}f_1^{\ } ( c)\biggr]^n\,,\end{aligned}\ ] ] and @xmath82^n\ ] ] follows .
taken together , ( [ eq:1stex ] ) and ( [ eq:2ndex ] ) tell us that the normalized variance of @xmath66 grows exponentially with the number of slabs , @xmath83^n\nonumber\\ & \approx&\frac{2}{3}\biggl[\frac{3}{2 } -\frac{1}{2}\bigl(\frac{\tau_1^{\ } } { 2-\tau_1^{\ } } \bigr)^2\biggr]^n \quad\mbox{for $ { n\gg1}$.}\end{aligned}\ ] ] the values of @xmath84 cover a correspondingly large range , and so we understand why the two sides of ( [ eq : logineq ] ) differ by much .
this brings us to the much more important @xmath85 case of @xmath86 here , @xmath87 is a manifestation of the `` self - averaging '' of the logarithm ( not _ any _ logarithm though , but this particular one ) , and we get @xmath88 this is the berry
klein result ( [ eq : logaver ] ) , of course .
finally , we turn to calculating @xmath11 .
the first few @xmath59s are @xmath89 giving @xmath90 and it is frustratingly difficult to go beyond @xmath91 .
but it is possible to evaluate the recurrence relation ( [ eq : recurrence ] ) numerically and so determine @xmath92 . in passing ,
we note that @xmath93 and @xmath94 for @xmath95 ; ray optics fails for @xmath96 . average transmission probability for a stack of @xmath0 identical slabs . for @xmath97 and @xmath98 ,
the dotted curve ` a ' shows the values of @xmath99 computed by a numerical evaluation of the recurrence relation ( [ eq : recurrence ] ) , commencing with the small-@xmath14 functions of ( [ eq : firstfew ] ) .
the crosses that follow curve ` a ' are values obtained by a monte carlo calculation that simulated @xmath100 experimental realizations .
the two solid lines are the upper and lower bounds of ( [ eq : upperbound ] ) and ( [ eq : lower4 ] ) , respectively . the dashed line ` b ' is the lower bound ( [ eq : bk - lowlim ] ) derived by berry and klein @xcite . ] for @xmath97
, the outcome of such a computation is shown in the lin - log plot of fig .
[ fig : logplot ] as the dotted curve ` a ' .
the crosses near the curve were obtained by a monte carlo calculation in which @xmath100 experiments were simulated with up to @xmath101 slabs . the straight dashed line ` b ' is the lower bound of ( [ eq : bk - lowlim ] ) .
the solid lines are the upper and lower bounds discussed in the next section .
other values of @xmath2 result in plots with the same general features .
since @xmath102 , we have @xmath103 where the second inequality recognizes that the integral in the first line is a monotonically increasing function of @xmath104 , so that the replacement @xmath105 increases its value .
the integral defining @xmath106 is of elliptic kind and its value is less than @xmath107 if @xmath108 , that is : @xmath109 if @xmath110 .
we conclude by induction that @xmath111 holds for @xmath112 .
the upper bound @xmath113 then follows .
the ray - optics result ( [ eq : rayoptics ] ) is inconsistent with this upper bound .
figure [ fig : bounds ] shows @xmath106 as a function of @xmath2 . upper bound @xmath106 and lower bound @xmath114 on @xmath115 as functions of @xmath2 .
the dashed straight line is the lower bound on @xmath116 of ( [ eq : bk - lowlim ] ) . ]
we derive a lower bound by first observing that @xmath117 with @xmath118 \sqrt{\tau_1^{\ } -\tau_1 ^ 2/4 } & \tau_1^{\ } \leq2-\sqrt{2 } \end{array}\right.\ ] ] and then inferring by induction that @xmath119 holds for @xmath112 .
the lower bound @xmath120 then follows .
the plot of @xmath114 as a function of @xmath2 in fig .
[ fig : bounds ] shows that @xmath121 for @xmath122 and , therefore , this lower bound is more stringent than ( [ eq : bk - lowlim ] ) , but it is not tight either .
we are certain , however , that @xmath11 is bounded exponentially both from above and from below .
figure [ fig : linplot ] illustrates the two bounds @xmath123 for @xmath124 .
the values for curve ` a ' are obtained by the numerical evaluation of the recurrence relation ( [ eq : recurrence ] ) .
clearly all values are well within the two bounds , the horizontal dashed lines .
this figure , and analogous plots for other values of @xmath125 , suggest that @xmath126 the corresponding observation in fig .
[ fig : logplot ] is that , for sufficiently large @xmath0 , line ` a ' there is parallel to the solid line for the upper bound . at present , ( [ eq : conject ] ) is no more than a conjecture that is supported by a body of numerical evidence .
values of @xmath115 for @xmath124 .
the bounds of ( [ eq : bounds ] ) are the two horizontal dashed lines .
curve ` a ' shows the actual values .
the extrapolation explained in the context of ( [ eq : conject ] ) and ( [ eq : extrapol ] ) gives curve ` b ' . ] some of the evidence is curve ` b ' in fig .
[ fig : linplot ] .
its values are obtained by an extrapolation that assumes that @xmath127 for large @xmath0 with @xmath128 and @xmath129 slowly varying with @xmath0 . for two consecutive @xmath0 values of curve ` a ' we can get an estimate of @xmath128 and @xmath129 , and curve ` b ' represents the successive values of @xmath128 thus extrapolated .
the rapid and consistent approach of ` b ' to the horizontal line of the upper limit feeds the expectation that the conjecture ( [ eq : conject ] ) could be true .
we leave the matter at that .
we established the recurrence relation ( [ eq : recurrence ] ) that facilitates the calculation of the average value @xmath130 of any function of @xmath3 , the transmission probability through the stack of @xmath0 identical slabs with random gaps between them .
we observed that the individual values of @xmath3 are spread over a large range and , therefore , @xmath7 exceeds @xmath131 by much .
further , we derived strict upper and lower bounds on @xmath7 , both bounds being exponential functions of @xmath0 .
the ray - optics prediction for @xmath7 is consistent with the lower bound but not with the upper bound .
the upper bound , when used as an approximation for @xmath7 , is of much better accuracy than its derivation suggests and , based on numerical evidence , we conjecture that it is asymptotically exact .
we are grateful for discussions with dominique delande .
centre for quantum technologies is a research centre of excellence funded by ministry of education and national research foundation of singapore . | the transmission through a stack of identical slabs that are separated by gaps with random widths is usually treated by calculating the average of the logarithm of the transmission probability .
we show how to calculate the average of the transmission probability itself with the aid of a recurrence relation and derive analytical upper and lower bounds .
the upper bound , when used as an approximation for the transmission probability , is unreasonably good and we conjecture that it is asymptotically exact . |
in the age of ` big data ' , efficient data representation is an objective of an ever increasing importance .
not only does it simplify the handling of the data due to the reduction of needed storage space or the possible speed - up of processing times .
the knowledge of a ` good ' representation also gives valuable information about the structure of the data itself , simplifying certain processing tasks or even just enabling them in the first place .
as an example we may think of the restoration of corrupted signals or the separation of several superimposed signals of distinct types . often , the data of interest can be modeled in a linear space , for instance a hilbert space as exemplified by the lebesgue spaces of square - integrable functions . in this setting the standard approach for the representation of a signal
is its expansion with respect to a fixed family of basic elements , a so - called dictionary for the data . in practice ,
one usually needs to contend with approximations and therefore resorts to approximation schemes , i.e. , algorithms that deliver for each signal @xmath22 a sequence of approximants @xmath23 converging to the signal . a standard choice here is to use @xmath1-term approximations in the respective dictionary , i.e. , approximants being built from just @xmath1 dictionary elements .
a main goal of approximation theory is the development of approximation schemes with a best possible speed of convergence , commonly quantified by the asymptotic decay of the approximation error @xmath24 as @xmath25 . with regard to @xmath1-term approximations ,
the achievable rate is determined by the utilized dictionary in the background and one aims to find dictionaries providing high approximation rates for the data .
such dictionaries are said to sparsely approximate the corresponding signals and clearly need to be chosen depending on the considered signal class .
for efficient data representation it is therefore essential , first , to be able to precisely specify the type of data under consideration , e.g. , in the form of an appropriate model , and , second , to develop dictionaries , well adapted to the specific data class , providing sparse approximations .
subsequently , we are interested in the sparse approximation of image data . in our investigation , we will always stay in the continuum setting , where images are as usual represented as functions supported on some compact image domain @xmath26 with values containing pixel information at the respective positions , such as e.g. color or brightness information . being compactly supported and bounded , the image data can conveniently be modeled as a subset of the hilbert space @xmath27 , which in turn is considered as a subspace of @xmath28 .
hence , we are in a concrete hilbert space scenario and can resort to the methodology described above , i.e. , we aim for appropriate image models and sparsifying dictionaries . for the space @xmath27 the classic fourier systems constitute an orthonormal basis , providing a straight - forward procedure for representation . however , fourier systems work well only if the functions under consideration are smooth . for general images
such smoothness assumptions are certainly not fulfilled . as another popular representation system
wavelets @xcite come to mind . nowadays
, they are one of the most widely used systems in applied harmonic analysis , with various applications ranging from signal compression ( e.g. jpeg2000 @xcite ) and restoration @xcite to pde solvers @xcite . in particular , they have the ability to sparsely approximate functions , which are smooth apart from isolated point singularities . for general image data , however , such regularity assumptions are still too strict . a characteristic feature of images are edges , leading to curvilinear discontinuities in the data . with respect to such line singularities , wavelet systems do not perform optimally any more .
the isotropy of their scaling prohibits an optimal resolution of these kind of anisotropic structures . with the desire to specifically model the occurrence of edges , the concept of cartoon - like functions emerged
. these are piecewise smooth functions featuring discontinuities along lower - dimensional manifolds , in our case along the @xmath29-dimensional edge curves of the image . based on such functions ,
suitable models for natural images have been conceived and different model classes have been introduced . typically , these classes are characterized by a specific smoothness of the regions and by certain conditions on the separating edges . as examples ,
let us mention the classic cartoons @xcite with @xmath0 regularity of the regions and the discontinuity curves , or the horizon classes considered e.g. in @xcite .
the achievable approximation rate for a class of cartoon - like functions essentially depends on the regularity of the cartoons , including both the smoothness of the edge curves and the smoothness of the regions in between .
it was shown in @xcite that @xmath13 regularity of the regions and the separating edges with @xmath30 allows for an asymptotic rate of order @xmath16 . by information theoretic arguments
, it has further been established that this rate can not be surpassed @xcite , at least in a class - wise sense .
interestingly , the benchmark @xmath16 is the same for the class of so - called binary cartoons , i.e. , cartoon - like functions with constant regions , and it also does not change if one restricts to @xmath13 smooth functions without any edges . with the model of cartoon - like functions at hand , let us turn again to the question of efficient image representation . in the past , a great amount of energy has been devoted to the effort of constructing dictionaries well - suited for cartoon approximation .
thereby , many different paths have been pursued and the developed methods can be divided into two general categories : adaptive and nonadaptive methods .
adaptive methods are by nature more flexible and have the inherent advantage of being able to adjust to the given data . on the downside ,
the increased flexibility typically comes at the cost of higher computational complexity of the employed approximation and reconstruction schemes .
some prominent examples of adaptive methods for image data are based on wedgelet dictionaries @xcite and their higher - order relatives , so - called surflets @xcite .
they have been shown to reach the optimality bound @xmath16 for binary cartoons with @xmath13 regularity @xcite .
other notable dictionaries used for adaptive approximation include beamlets @xcite , platelets @xcite , and derivatives of wedgelets such as multiwedgelets @xcite or smoothlets @xcite .
more recently , new adaptive schemes have emerged that use bases , e.g. , bandelets @xcite , grouplets @xcite , and tetrolets @xcite .
quasi - optimal approximation for @xmath13 cartoons with @xmath30 has been proved e.g. in @xcite for bandelets .
nonadaptive methods are usually much simpler than adaptive schemes , at least from an algorithmic perspective .
mainly , they are based on frames and the corresponding reconstruction formulas
. an easy path to approximation is thus provided by simply thresholding the frame coefficients .
surprisingly , despite the simplicity of such schemes , there exist frames with quasi - optimal approximation performance for certain cartoon classes .
if the edges of the cartoons are straight , different variants of so - called ridgelet frames have been shown to yield quasi - optimal approximation @xcite . originally , the notion of a ridgelet was introduced by cands @xcite in 1998 , who defined them as bivariate ridge functions obtained by tensoring a univariate wavelet with a constant . since these ` pure ridgelets ' are not square - integrable , the concept was later modified in order to obtain frames or bases for @xmath28 . by giving them a slow decay along the ridge ,
donoho constructed an orthonormal basis whose elements are called ` orthonormal ridgelets ' @xcite .
their close relationship to the original concept has been analyzed in @xcite .
another construction , based on directional scaling , goes back to grohs , providing tight frames @xcite .
this kind of construction coincides with the concept of ` @xmath19-curvelets ' presented below . to deal with curved edges
, numerous types of frames have been developed .
an important milestone was the introduction of the first generation of curvelets @xcite by cands and donoho in 1999 .
they represent the first frame to reach the optimal approximation order of @xmath2 for @xmath0 cartoons via simple thresholding . a modification of this system , the second generation of curvelets @xcite , was introduced in 2002 by the same authors .
it is based on a more elegant and simpler construction principle , yet features the same quasi - optimal approximation properties . following this early breakthrough
, other constructions better suited for digital implementation were developed .
let us mention contourlets @xcite by do and vetterli and shearlets , whose construction goes back mainly to guo , kutyniok , labate , lim , and weiss .
the first shearlet construction consisted of band - limited functions and was presented in @xcite .
later , more sophisticated shearlet systems were developed , such as e.g. the well - localized band - limited parseval frame in @xcite or even systems of compactly supported shearlets @xcite . like curvelets ,
those systems provide quasi - optimal approximation for @xmath0 cartoons .
for the classic band - limited shearlets this was established in @xcite , for those with compact support in @xcite .
a common principle underlying the above constructions is parabolic scaling , a type of scaling optimally adapted to @xmath0 singularity curves .
it is essential for the quasi - optimal approximation of @xmath0 cartoons and led to the notion of parabolic molecules @xcite .
this concept unifies various parabolically scaled systems under one roof , in particular the classic curvelet and shearlet systems , and is the predecessor of the more general framework of @xmath3-molecules @xcite . comparing the approximation properties of wavelets , curvelets , and ridgelets ,
a distinct behavior with respect to their ability to resolve edges is characteristic .
ridgelets are optimally adapted to straight edges , curvelets are optimal for @xmath0 line singularities , and wavelets for point singularities .
this distinct behavior is due to the different scaling laws underlying their respective constructions : isotropic scaling for wavelets , parabolic scaling for curvelets , and directional scaling for ridgelets . introducing a parameter @xmath31 and associated @xmath3-scaling matrices @xmath32 one
can interpolate between these different types of scaling and construct corresponding @xmath3-scaled representation systems . incorporating @xmath3-scaling in the original construction of curvelets , for instance , yields so - called @xmath3-curvelets @xcite . for @xmath33 $ ] ,
they constitute a family of systems which encompass ridgelets ( in the sense of @xcite ) for @xmath6 , the classic curvelets for @xmath5 , and wavelets for @xmath4 . in a similar fashion , @xmath3-shearlet systems @xcite
can be obtained by modifying the classic shearlet constructions .
a natural question concerning such @xmath3-scaled systems is how their approximation properties are affected by a change of the parameter @xmath3 . with regard to cartoon approximation
, this question has been pursued in @xcite for @xmath3-curvelet frames and in @xcite for @xmath3-shearlet frames .
it was shown that , if @xmath7 and if the cartoon @xmath22 is of regularity @xmath13 with @xmath34 , simple thresholding of the coefficients yields @xmath1-term approximations @xmath35 with a convergence of @xmath36 which apart from the log - factor is optimal .
later , these results were further extended utilizing the theory of @xmath3-molecules @xcite .
this is a framework providing a unified approach to @xmath3-scaled systems , based solely on assumptions on the time - frequency localization of the respective functions .
it allows to transfer approximation results obtained for one system of @xmath3-molecules to other systems , under certain consistency conditions .
in particular , the rate for @xmath3-curvelets was generalized ( in a weak form ) to other @xmath3-scaled representation systems @xcite , which all achieve a rate of @xmath37 with @xmath38 arbitrarily small . despite these results , many questions concerning @xmath3-scaled representation systems and their ability to approximate
cartoon - like functions remain open , e.g. , their performance in the range @xmath39 or their suitability for the approximation of straight edges . in this research
we want to address these open questions , shedding ( even ) more light on the role of the parameter @xmath3 .
our exposition starts with a short review of @xmath3-scaled systems in section [ sec : curvelets ] , where also a specific construction of an @xmath3-curvelet frame for @xmath28 is presented .
this frame , denoted by @xmath40 , will serve as a prototypical system whose properties have ramifications for other @xmath3-scaled systems , such as for example @xmath3-shearlets , due to the transference principle of the framework of @xmath3-molecules . in the main part of the article , sections [ sec : cartoon ] and [ sec : straight ]
, we analyze the @xmath1-term approximation properties of the frame @xmath40 with regard to different classes of cartoon images . in section [ sec : cartoon ] we start with cartoons with curved edges and first introduce corresponding signal classes of @xmath13 regularity for @xmath41 .
theorem [ thm : benchmark ] recalls @xmath16 as the order of the maximal achievable approximation rate for such @xmath13 cartoons , which can not be surpassed by any polynomial - depth restricted @xmath1-term approximation scheme , independent of the utilized dictionary .
then we recall the quasi - optimal approximation of @xmath3-curvelets , proved in @xcite , if @xmath7 and @xmath34 .
our main findings in section 3 , theorems [ thm : bound1 ] and [ thm : bound2 ] , extend and complement this result .
theorem [ thm : bound1 ] shows that the best possible @xmath1-term approximation rate achievable by @xmath40 for cartoons with curved edges is limited to at most @xmath42 , where @xmath10 and the smoothness of the cartoons is arbitrary .
moreover , according to theorem [ thm : bound2 ] , the achievable rate can not exceed @xmath43 if a simple thresholding scheme is used .
these bounds show that @xmath3-curvelets with @xmath7 can not take advantage of regularity higher than @xmath8 .
furthermore , they prohibit optimal approximation of @xmath13 cartoons if @xmath17 , since decreasing @xmath3 beyond @xmath15 deteriorates the achievable rates compared to the classic curvelets .
hence , with a rate of order @xmath2 , these provide the best performance among all @xmath3-curvelet systems , if the regularity of the cartoons is at least @xmath0 and curved singularities are involved . as a consequence ,
no curvelet system can reach the optimality bound @xmath16 if @xmath17 .
in fact , up to now , no frame construction is known where a nonadaptive approximation scheme can break this @xmath2 barrier and the quest for such frames remains open . in section [ sec : straight ] we consider cartoons featuring only straight edges . for the corresponding classes of regularity @xmath13 the same optimality benchmark @xmath16 holds true as for the cartoons with curved edges .
our main result of section [ sec : straight ] , theorem [ thm : mainappr1 ] , shows that a simple thresholding scheme for the @xmath3-curvelet frame @xmath40 yields approximation rates of order @xmath20 .
hence , here a smaller @xmath3 is beneficial and even ensures quasi - optimal approximation if @xmath44 $ ] .
this finding generalizes earlier results for ridgelets .
we finish with a short discussion of our results in section [ sec : discussion ] . in particular
, we point out some ramifications for other @xmath3-scaled representation systems , utilizing the framework of @xmath3-molecules .
all @xmath3-scaled systems which are frames and in a certain sense consistent with @xmath40 feature similar properties , formulated in theorem [ thm : mol_app ] and corollary [ cor : mol_app ] .
some useful properties of bessel functions needed in section 3 are collected in the appendix .
before we begin , let us fix some general notation .
writing @xmath45 we will refer to the natural numbers without zero , and we let @xmath46 . as usual , @xmath47 , @xmath48 and @xmath49
denote the integer , real and complex numbers .
further , we put @xmath50 and @xmath51 .
we also introduce the ` floor ' and ` ceiling ' of @xmath52 , @xmath53 and @xmath54 . the symbol @xmath55 is used for the torus obtained from the interval @xmath56 $ ] by identifying the endpoints . the unit - circle in @xmath57 is denoted by @xmath58 .
the vector space @xmath59 , @xmath60 , is equipped with the euclidean scalar product @xmath61 and associated norm @xmath62 .
the notation @xmath63 , @xmath64 $ ] , is used for the @xmath65-(quasi-)norms on @xmath59 . for a multi - index @xmath66
, @xmath67 is a differential operator with @xmath68 , @xmath69 , the partial derivative in the @xmath70-th coordinate direction . given a vector @xmath71 , we further define @xmath72 ( with the convention @xmath73 ) .
if @xmath74 holds true for two quantities @xmath75 depending on a set of parameters @xmath76 with a uniform constant @xmath77 , we write @xmath78 or equivalently @xmath79 .
if both , @xmath78 and @xmath80 , hold true , we denote this by @xmath81 . for measurable subsets
@xmath82 we let @xmath83 , @xmath64 $ ] , denote the usual lebesgue spaces with respect to the lebesgue measure .
the corresponding ( quasi-)norms are denoted by @xmath84 , in case @xmath85 we abbreviate @xmath86 .
for the scalar product on @xmath27 the same notation @xmath61 as for the euclidean product on @xmath59 is used .
the lebesgue sequence spaces , for a discrete index set @xmath87 , are denoted by @xmath88 with associated ( quasi-)norms @xmath89 .
the definition of their weak counterparts @xmath90 , equipped with ( quasi-)norms @xmath91 , are recalled in section [ sec : straight ] .
the space @xmath92 , for an integer @xmath93 , shall comprise all continuous real - valued functions on @xmath59 , whose classic derivatives up to order @xmath94 exist .
for @xmath41 we then define @xmath95 where @xmath96 and the hlder constant of exponent @xmath33 $ ] is given by @xmath97 the notation @xmath98 , for some open subset @xmath82 , is used for functions @xmath99 whose support @xmath100 is compact and contained in the closure @xmath101 of @xmath102 .
frequently , we also need to measure functions @xmath103
, @xmath94 , with the following sobolev norms , where @xmath104 $ ] , @xmath105 finally , we will use the following version of the fourier transform . for a schwartz function @xmath106 @xmath107 as usual ,
@xmath108 is extended to the tempered distributions @xmath109 , and we often write @xmath110 for @xmath111 .
directional multi - scale systems based on @xmath3-scaling feature a characteristic tiling of the frequency domain .
the multi - scale structure is reflected by a partition of the fourier plane into dyadic coronae , further divided into wedge - like tiles , where the energy of the system elements is concentrated . in case of inhomogeneous systems ,
a ball around the origin corresponds to the low - frequency base scale .
a prototypical instance of such an @xmath3-scaled system is the frame @xmath40 of @xmath3-curvelets , thoroughly defined in this section .
it is prototypical in the sense that many of its properties transfer via the framework of @xmath3-molecules @xcite to other @xmath3-scaled systems . among these
are other @xmath3-curvelet constructions @xcite , but also band - limited @xcite as well as compactly supported @xcite @xmath3-shearlet systems .
this fact gives the system @xmath40 a special significance for our purpose and motivates its detailed discussion here . before defining @xmath40 , which is similar to the construction of @xmath3-curvelets in @xcite ,
let us first elaborate the geometric aspects of the corresponding frequency tiling . at scales
@xmath112 we have the coronae @xmath113 where @xmath114 is a fixed parameter and @xmath77 is a constant , specified conveniently later .
these coronae are each uniformly divided into an even number of wedges , whose angular width at scale @xmath115 is given by the angle @xmath116 and depends on another parameter @xmath117 $ ] .
the approximate size of the resulting wedges correlates with an @xmath3-scaled rectangle of dimension @xmath118 . by combining opposite wedges to wedge pairs ,
we obtain the tiles for the scales @xmath112 .
there is only one tile associated with the base scale @xmath119 , the low frequency ball @xmath120 . for convenience ,
let us also introduce the angle @xmath121 . according to the above construction , at each scale
@xmath122 the number of tiles @xmath123 is given by @xmath124 in the following , the individual tiles will be denoted by @xmath125 and indexed by the set @xmath126 with @xmath127 and @xmath128 .
hereby we let @xmath129 , and in each corona @xmath130 with @xmath112 the wedge - pair @xmath131 shall be aligned horizontally , i.e. , @xmath132 the remaining tiles @xmath125 , @xmath133 , are obtained via rotations of @xmath131 by integer multiples @xmath134 of the angle @xmath135 defined in .
hence , @xmath136 with rotation matrix @xmath137 the resulting tiling of the fourier domain is schematically depicted in figure [ fig : freq_domain ] ( a ) .
we remark that in contrast to @xcite , where @xmath33 $ ] , we allow @xmath117 $ ] in the @xmath3-curvelet construction .
this range is natural for the considered inhomogeneous systems . if @xmath138 , the number of tiles @xmath123 in each corona decreases with rising scale , and eventually @xmath139 .
thus , at high scales , those systems would behave like isotropically scaled systems with @xmath4 .
let us now turn to the actual construction of the @xmath3-curvelet frame @xmath40 .
to realize the described frequency tiling , smooth functions @xmath140 , @xmath141 , are used , with compact support approximately given by the tiles @xmath142 .
it is convenient to construct them as tensor products of a radial and an angular component .
this allows to realize the desired support separately on the ray @xmath143 and on the circle @xmath144 .
projecting the coronae @xmath130 onto the ray @xmath145 yields the intervals @xmath146 \quad\text{and}\quad\mathcal{i}_{j}:=c\cdot[2^{s(j-1)},2^{s(j+1 ) } ] \ , , \quad j\ge1.\end{aligned}\ ] ] for the radial subdivision , we thus utilize nonnegative smooth functions @xmath147 , @xmath122 , which satisfy the support condition @xmath148 and for @xmath149 @xmath150 more concretely , we assume that the functions @xmath151 , @xmath112 , are generated by a single function @xmath152)$ ] via @xmath153 and that there are @xmath154 such that @xmath155 , & \quad \sqrt{a_1}\le u_0 \le \sqrt{b_1 } \text { on } c\cdot[0,\tau_1 ] , \\
\supp\ , u\subseteq c\cdot[2^{-s}\tau_1,\tau_2 ] , & \quad \sqrt{a_1}\le
u \le\sqrt{b_1 } \text { on } c\cdot[2^{-s}\tau_2,\tau_1 ] .
\end{aligned}\end{aligned}\ ] ] such functions exist and can even be constructed with @xmath156 in . for the angular subdivision , we construct at each scale @xmath122 a smooth partition on the unit circle @xmath144 , reflecting the angular support of the tiles @xmath125 .
we start with a function @xmath157)$ ] with the properties @xmath158 , \quad \sqrt{a_2}\le \widetilde{v } \le\sqrt{b_2 } \text { on } [ -\textstyle{\frac{\pi}{4}},\textstyle{\frac{\pi}{4}}],\quad a_2\le\sum_{k\in{\mathbb{z } } } \widetilde{v}^2(\cdot -k\pi)\le b_2,\end{aligned}\ ] ] where @xmath159 . scaling
then gives rise to the functions @xmath160 ) $ ] for @xmath122 .
via the bijection @xmath161 these functions yield functions @xmath162)$ ] on the unit circle .
we symmetrize @xmath163 and note that @xmath164 on @xmath165 . applying
the rotation then yields functions @xmath166 for every @xmath167 , which satisfy @xmath168 for all @xmath169 .
here we use the notation @xmath170 for @xmath167 . finally , we are ready to define the wedge functions @xmath171 as the polar tensor products @xmath172 these functions are non - negative ` bumps ' approximately supported in the corresponding wedges @xmath173 .
they are symmetric , i.e. , @xmath174 for @xmath175 , and they satisfy @xmath176 let us analyze the support of @xmath177 in more detail .
recall the angular function @xmath178 and note that its support on @xmath58 covers an angle range of @xmath179 with @xmath180 as in .
moreover , @xmath181 on a range of size @xmath182 .
hence , @xmath183 and @xmath184 on @xmath185 for the angular intervals @xmath186 next , recall the functions @xmath151 on the ray with @xmath187 . due to and
their function values are between @xmath188 and @xmath189 on @xmath190 \quad\text{and}\quad \mathcal{i}^{-}_j:=c\cdot[2^{s(j-1 ) } \tau_2,2^{sj}\tau_1],\quad j\ge1,\end{aligned}\ ] ] respectively .
this leads us to the following definition .
for @xmath167 we introduce the wedge pairs @xmath191 the following support properties will be of essential importance later , @xmath192 a geometric illustration is displayed in figure [ fig : freq_domain ] ( b ) .
now we fix @xmath193 in such that each @xmath194 is contained in the respective rectangle @xmath195\times[-2^{js\alpha-1},2^{js\alpha-1}].\end{aligned}\ ] ] the rectangles @xmath196 are of size @xmath197 and hence the fourier system @xmath198 given by @xmath199 constitutes an orthonormal basis for @xmath200 .
consequently , the rotated system @xmath201 of functions @xmath202 is an orthonormal basis for @xmath203 .
after this preparation , we are ready to define the @xmath3-curvelet system @xmath40 .
let @xmath114 , @xmath117 $ ] , and assume that @xmath204 is a family of functions of the form such that holds for @xmath205 .
further , let @xmath206 be the functions defined in .
the curvelet system @xmath207 with associated index set @xmath208 consists of the functions @xmath209 given by @xmath210 note that @xmath211 depends on the utilized family @xmath212 , which is not accounted for in the notation .
the curvelets @xmath213 are real - valued due to the symmetry of @xmath214 .
their @xmath215-norms may vary slightly with scale , however there are constants @xmath216 such that @xmath217 holds true for all @xmath218 .
most importantly , the system @xmath211 is a frame for @xmath28 .
the system @xmath211 given by is a frame for @xmath28 with frame bounds @xmath219 and @xmath220 . the functions @xmath221
satisfy condition wherefore @xmath222 since @xmath223 and since @xmath224 is an orthonormal basis of @xmath225 we have the orthogonal expansion @xmath226 .
the proof is finished by the following equality , @xmath227 the parseval frame @xmath228 is of most interest to us and one might wonder why we did not fix the frame bounds @xmath229 in the beginning .
the reason is that , in the proof of lemma [ lem : induction ] , we need the additional flexibility provided by variable @xmath219 and @xmath220 . subsequently , we will write @xmath230 to refer to the parseval frame @xmath228 .
let us finish this section with a short discussion of the situation in spatial domain . here
the @xmath3-curvelets @xmath231 are translates of the functions @xmath232 .
indeed , since @xmath233 and @xmath234 where @xmath235 is the rotation matrix defined in and @xmath236 is an @xmath3-scaling matrix of the form , we have @xmath237 and hence @xmath238 since @xmath232 is the rotation of @xmath239 by the angle @xmath240 , we arrive at the representation @xmath241 in fact , these systems are instances of @xmath3-molecules , a concept recalled in the definition below .
let @xmath87 be a set and @xmath242 a map , assigning to each @xmath243 a point @xmath244 in the so - called phase - space @xmath245 .
futher , assume that @xmath246 .
a family @xmath247 of functions in @xmath28 is called a _
family of @xmath3-molecules of order @xmath248 with respect to the parametrization @xmath249 _ , if there exist generators @xmath250 such that for all @xmath243 @xmath251 and if for each @xmath252 , @xmath253 , there is a constant @xmath254 such that for all @xmath243 @xmath255 we can deduce from that the @xmath3-curvelets @xmath256 can be represented in the form @xmath257 with respect to the generators @xmath258 since these generators fulfill condition , as shown in lemma [ thm : curvmol ] below , @xmath40 is a system of @xmath3-molecules of arbitrary order , at least in the range @xmath33 $ ] for which the concept was formulated .
the associated parametrization , mapping the curvelet index set @xmath259 into the phase - space @xmath260 , is given by @xmath261 [ thm : curvmol ] let @xmath262 and @xmath263 be fixed .
there is a constant @xmath77 such that for all @xmath122 the generators satisfy the estimate @xmath264 on the fourier side the functions have the form @xmath265 let @xmath122 be arbitrary .
we have @xmath266 and @xmath267\times[-2^{js\alpha-1},2^{js\alpha-1}]=\xi_{j,0},\ ] ] which implies @xmath268\times[-2^{-1},2^{-1}]=\xi_{0,0}.\end{aligned}\ ] ] further , if @xmath269 the function @xmath270 vanishes on the square @xmath271 ^ 2 $ ] .
consequently , @xmath272 vanishes on @xmath273\times \big ( 2^{js(1-\alpha ) } \cdot [ -2^{-2s-5},2^{-2s-5 } ] \big)$ ] .
the mixed derivatives @xmath274 obey uniformly in @xmath122 @xmath275 with the chain rule we deduce @xmath276 due to @xmath277 this estimate together with the support properties of @xmath272 implies . with the machinery of @xmath3-molecules at our disposal
, it is possible to use @xmath40 as an anchor system whose properties have consequences for other @xmath3-scaled systems if they fulfill certain consistency conditions . in particular , approximation properties of @xmath40 are shared by other @xmath3-scaled systems such as e.g. @xmath3-shearlets .
a short discussion of this can be found in section [ sec : discussion ] . for more details on the topic of @xmath3-molecules
we refer to @xcite .
in the two central sections of this article , sections [ sec : cartoon ] and [ sec : straight ] , we study the approximation performance of the @xmath3-curvelet frame @xmath40 with respect to different cartoon classes .
we begin in this section with classes of general cartoons , used e.g. to model natural images . in section
[ sec : straight ] we then turn our focus on cartoons featuring only straight edges . many suitable and well - established models for natural images
are based on the concept of so - called cartoon - like functions . in a nutshell
, such functions can be thought of as a patchwork of smooth regions separated from one another by piecewise - smooth discontinuity curves .
their structure imitates the fact that edges , a typical feature of natural images , are characterized by abrupt changes of color and brightness , whereas changes in the regions in between occur smoothly .
mathematically , models based on this idea can be concretised in different ways .
a classic model @xcite postulates a compact image domain separated into two @xmath0 regions by a closed @xmath0 discontinuity curve .
this model was generalized in various directions , e.g. , to take into account piecewise - smooth edges or to allow more general @xmath13 regularity with @xmath41 .
cartoon classes of this kind have been studied extensively , especially in the range @xmath278 $ ] , e.g. , in @xcite .
another variant are the closely related horizon classes , where the discontinuity is not a closed curve in the image domain but a ( possibly curved ) horizontal or vertical line stretching across .
such classes have been investigated e.g. in @xcite .
let us also mention that there exist extensions to multi - dimensions , see e.g. @xcite .
in particular , the corresponding 3d models have been applied in the investigation of video data .
since we are concerned with image approximation , our attention is restricted to the 2-dimensional setting .
the following definition is a template for different classes of bivariate cartoons , comprising many of those mentioned above .
it provides the flexibility to taylor the model to our particular needs in sections [ sec : cartoon ] and [ sec : straight ] .
[ def : gencart ] let @xmath41 and @xmath279 . given a domain @xmath280 and a set @xmath281 of admissible subsets of @xmath282 ,
the class @xmath283 consists of all functions @xmath284 of the form @xmath285 where @xmath286 and @xmath287 with @xmath288 and @xmath289 .
the class @xmath290 shall be the collection of all ` binary functions ' @xmath291 , where @xmath286 and @xmath292 . for particular choices of @xmath281 many of the classes appearing in the literature
can be retrieved , including classes of horizon - type . in this section
we focus on the class @xmath283 with fixed image domain @xmath293 ^ 2 $ ] and certain @xmath13 domains as admissible sets @xmath281 .
similar to @xcite , we restrict our investigation to star - shaped domains , since those allow a simple parametrization of the boundary curve .
the results obtained however also hold true for more general domains .
let us introduce the collection of admissible sets @xmath294 , @xmath279 , as all translates of sets @xmath295 , whose boundary @xmath296 possesses a parametrization @xmath297 of the form @xmath298\,,\ ] ] where the radius function @xmath299 is a @xmath13 function with @xmath300 where we set @xmath301 .
the condition implies that with @xmath302 we have @xmath303 for every @xmath304 if @xmath305 , and @xmath306 for @xmath307 .
in particular @xmath308 for all @xmath309 .
note , that the set @xmath310 differs from the set of star - shaped domains used in @xcite .
the domains in @xmath310 are not restricted to subsets of @xmath311 ^ 2 $ ] .
in fact , every star - shaped @xmath13 domain with center @xmath19 and @xmath312 is contained in @xmath310 for suitably large @xmath313 .
moreover , the collection @xmath310 is scaling invariant in the sense that for @xmath314 and @xmath315 also @xmath316 , provided @xmath317 .
in addition , with @xmath314 also the complement @xmath318 is contained in @xmath310 . building upon definition [ def :
gencart ] we now define the class of functions which we want to study in this section . we put @xmath293 ^ 2 $ ] and @xmath319 .
further , we assume @xmath41 and @xmath279 . for the resulting class @xmath320 ^ 2;{\text{\sc star}}^\beta(\nu),\nu)$ ]
we simplify the notation @xmath321 ^ 2;\nu):= \mathcal{e}^\beta([-1,1]^2;{\text{\sc star}}^\beta(\nu),\nu).\end{aligned}\ ] ] the associated binary class shall be denoted by @xmath322 ^ 2;\nu):= \mathcal{e}_{bin}^\beta([-1,1]^2;{\text{\sc star}}^\beta(\nu))$ ] .
before we investigate the approximation performance of the @xmath3-curvelet frame @xmath40 with respect to the class @xmath320 ^ 2;\nu)$ ] , let us take a broader stance and aim for best possible @xmath1-term approximation in case we can freely choose the utilized dictionary .
of course , a countable dense subset of @xmath28 would yield arbitrarily good @xmath29-term approximations .
this shows that , without further restrictions , the question of best possible approximation is not well - posed . to cast a realistic scenario , when computing @xmath1-term approximations typically a constraint on the search depth is imposed .
more concretely , given a fixed ordering of the dictionary and some polynomial @xmath323 , it is common to allow only @xmath1-term approximants being built from the first @xmath324 elements of the dictionary . under this so - called polynomial depth search constraint , an upper bound on the maximal achievable approximation rate was first derived by donoho ( * ? ? ?
1 ) for binary @xmath13 cartoons in the range @xmath278 $ ] .
later similar results were proved for more general cartoon classes @xcite .
theorem [ thm : benchmark ] below establishes a bound for the class @xmath320 ^ 2;\nu)$ ] specified in .
[ thm : benchmark ] let @xmath325 and @xmath279 .
assume that there is a constant @xmath77 such that @xmath326 ^ 2;\nu ) } \|f - f_n\|_2 ^ 2 \le c n^{-\gamma } \quad\text { for all } n\in{\mathbb{n}},\ ] ] where @xmath35 denotes the best @xmath1-term approximation of @xmath22 obtained by polynomial depth search in a fixed dictionary . then necessarily @xmath327 . in principle
, this is a known result ( see e.g. @xcite ) .
however , for reasons of completeness , we outline a short proof based on the technique used in @xcite .
it relies on theorem [ thm : upperbound ] below and the fact that the class @xmath328 ^ 2;\nu)$ ] contains a copy of @xmath329 for @xmath330 .
let us recall this notion introduced in @xcite .
a function class @xmath331 is said to contain an _ embedded orthogonal hypercube _ of dimension @xmath332 and side - length @xmath333 if there exist @xmath334 and orthogonal functions @xmath335 , @xmath336 , with @xmath337 such that the collection of hypercube vertices embeds , i.e. , @xmath338 it is said to _
contain a copy of @xmath339 , @xmath340 , _ if it contains a sequence of embedded orthogonal hypercubes , whose associated dimensions @xmath341 and side - lengths @xmath342 satisfy @xmath343 for @xmath344 and with a constant @xmath77 @xmath345 the significance of this notion is due to the following result , which was first obtained in ( * ? ? ?
* thm . 2 ) .
the reformulated version below can be found in ( * ? ? ?
[ thm : upperbound ] suppose , that a class of functions @xmath331 is uniformly @xmath215-bounded and contains a copy of @xmath346 . then , allowing only polynomial depth search in a given dictionary , there is a constant @xmath77 such that for every @xmath347 there is a function @xmath348 and an @xmath349 , @xmath350 such that @xmath351 where @xmath35 denotes the best @xmath1-term approximation under the polynomial depth search constraint .
it remains to investigate for which @xmath340 the class @xmath328 ^ 2;\nu)$ ] contains a copy of @xmath329 . to this end , let us introduce the following subclass of smooth functions for @xmath41 and @xmath279 , @xmath352 ^ 2;\nu):=\big\ { f\in c_0^\beta([-1,1]^2 ) ~:~ \|f\|_{c^\beta}\le\nu \big\}.\end{aligned}\ ] ] note , that the choice @xmath293 ^ 2 $ ] and @xmath353 in definition [ def : gencart ] yields this class . as a consequence , @xmath354 ^ 2;\nu ) \subset \mathcal{e}^{\beta}([-1,1]^2;\nu).\end{aligned}\ ] ] lemma [ lem : copylp ] below is the 2d analogon of the statement of ( * ? ? ?
it shows , in particular , that @xmath355 ^ 2;\nu)$ ] contains a copy of @xmath356 .
hence , as a consequence of , also @xmath328 ^ 2;\nu)$ ] contains a copy of @xmath356 . an application of theorem [ thm : upperbound ] thus yields theorem [ thm : benchmark ] .
[ lem : copylp ] let @xmath279 , @xmath41 , and @xmath357 .
then the following holds true . 1 .
the function class @xmath355 ^ 2;\nu)$ ] contains a copy of @xmath329 .
2 . the class of binary cartoons @xmath358 ^ 2;\nu)$ ] contains a copy of @xmath329 if @xmath359 , otherwise it only contains the zero - function . the proof is a @xmath360d - adaption of the proof of ( * ? ? ?
summarizing , this establishes @xmath16 as an upper bound for the possible order of approximation for general @xmath13 cartoons .
this rate is the benchmark , against which the performance of @xmath40 has to be measured .
we end this paragraph with the following observation .
[ rem : benchmark ] according to lemma [ lem : copylp](i ) , the bound of theorem [ thm : benchmark ] actually holds true for the class @xmath355 ^ 2;\nu)$ ] .
this is a stronger statement due to the inclusion .
further , due to lemma [ lem : copylp](ii ) , a statement analogous to theorem [ thm : benchmark ] holds true for the binary class @xmath358 ^ 2;\nu)$ ] if @xmath359 . according to theorem [ thm : benchmark ] and remark [ rem : benchmark ] the order of the @xmath1-term approximation rate achievable for the classes @xmath358 ^ 2;\nu)$ ] , @xmath359 , and @xmath328 ^ 2;\nu)$ ] , @xmath279 , can not exceed @xmath16 .
this bound is valid for arbitrary dictionaries and independent of the approximation scheme employed , as long as it respects a polynomial depth search condition .
even adaptive approximation schemes can not perform better .
schemes , where these rates are provably achieved , at least up to order , have been developed for binary cartoons based on wedgelets @xcite and surflets @xcite , for general cartoons utilizing bandelets @xcite .
these results show that the optimality benchmark @xmath16 can indeed be realized in practice , at least up to order . however , the utilized schemes are mostly adaptive , only for certain cartoon classes nonadaptive methods with quasi - optimal performance are known . a breakthrough concerning the nonadaptive approximation of @xmath0 cartoons with curved edges was the introduction of curvelets by cands and donoho @xcite . by a simple thresholding scheme ,
curvelet frames achieve an approximation rate matching the class bound @xmath2 up to a log - factor .
the reason for this performance is due to the parabolic scaling employed .
the following argument shall heuristically explain , why this type of scaling is ideal for the representation of @xmath0 edges .
in local cartesian coordinates , a @xmath0 curve can be represented as the graph @xmath361 of a function @xmath362 and one can choose a coordinate system such that @xmath363 .
a taylor expansion then yields approximately @xmath364 , which matches the essential support @xmath365 of parabolically scaled functions .
hence , those can provide optimal resolution of the curve across all scales . a similar heuristic applies to @xmath13 curves if @xmath278 $ ] .
a taylor expansion of @xmath366 yields @xmath367 . the curve is thus contained in a rectangle of size @xmath368 which suggests @xmath3-scaling with @xmath369 for optimal approximation . and
indeed , the classic approximation result by cands and donoho could be extended in ( * ? ? ?
4.1 ) to the range @xmath278 $ ] .
this generalized result is stated below , slightly modified to fit into the setting of this article .
the class @xmath328 ^ 2;\nu)$ ] used here is not fully identical to the class in @xcite .
moreover , only curvelet frames of the type @xmath40 with @xmath370 were considered there .
it is not hard to verify though that the proof carries over to general @xmath114 and that the statement is also valid in our setting .
[ thm : oldcurveappr ] let @xmath278 $ ] , @xmath279 .
for the choice @xmath369 , @xmath114 arbitrary , the frame of @xmath3-curvelets @xmath40 provides almost optimal sparse approximations for the class @xmath328 ^ 2;\nu)$ ] .
more precisely , there exists a constant @xmath77 such that for every @xmath371 ^ 2;\nu)$ ] and @xmath349 @xmath372 where @xmath35 denotes the @xmath1-term approximation of @xmath22 obtained by choosing the @xmath1 largest coefficients .
this theorem naturally raises the question of extendibility beyond the range @xmath278 $ ] , a question pursued in the following subsection .
in particular , we investigate if the choice @xmath369 is still optimal for @xmath17 . obviously , the heuristic consideration from above is not valid any more in this regime . and indeed , we will see that for @xmath17 the optimal choice is not @xmath369 .
in fact , it is still @xmath373 and choosing @xmath374 deteriorates the approximation performance . the main results of this subsection ,
theorems [ thm : bound1 ] and [ thm : bound2 ] , establish bounds on the achievable @xmath1-term approximation rate for the class @xmath375 ^ 2;\nu)}$ ] , @xmath41 , when using the @xmath3-curvelet frame @xmath40 for approximation . unlike the bounds in theorem [ thm : benchmark ] associated with the signal class
the bounds derived here are tied to the particular approximation system @xmath40 . however , via the framework of @xmath3-molecules they are also effective for other @xmath3-scaled systems , such as e.g. @xmath3-shearlets as discussed in section [ sec : discussion ] . in order to establish these bounds
we study the approximability of certain example cartoons . as a suitable object
, we choose the characteristic function of the ball @xmath376 of radius @xmath377 , for which we subsequently use the symbol @xmath378 this function embodies an exceptionally regular cartoon with a closed curved @xmath379-singularity .
it is radial symmetric and binary , contained in @xmath380 ^ 2,\nu)$ ] for arbitrary @xmath41 and @xmath381 .
furthermore , for every @xmath41 and @xmath381 there is @xmath382 such that @xmath383 ^ 2;\nu)}$ ] , wherefore the approximability of @xmath384 has implications for the approximability of these cartoon classes .
the fourier transform of @xmath384 is explicitly known .
let @xmath385 denote the bessel function of order 1 , then according to @xmath386 some properties of @xmath385 and bessel functions in general are collected in the appendix
. at the center of the following investigation is the lemma below , which estimates the energy of @xmath387 contained in the wedges @xmath388 , @xmath141 .
let @xmath389 be a family of functions of the kind with property for @xmath205 .
further , let @xmath390 be the characteristic functions of the sets @xmath391 and @xmath392 defined in . [
lem : thetaest ] there are constants @xmath216 , independent of scale @xmath393 , where @xmath394 is a suitable base scale , such that for all @xmath141 with @xmath395 , where @xmath396 for @xmath167 , @xmath397 let us recall the bessel function @xmath385 of order 1 and its asymptotic behavior . according to there is a constant @xmath77 and a function @xmath398 on @xmath399 satisfying @xmath400 such that @xmath401 this allows to separate terms of higher order from @xmath402 .
we decompose @xmath403 + \big [ \sqrt{\frac{8}{\pi } } \cos(r-\frac{3\pi}{4 } ) r^{-1/2}r_1(r ) + r_1(r)^2 \big]= : t_1(r ) + t_2(r).\end{aligned}\ ] ] for the following argumentation we need the square wave function @xmath404 defined by @xmath405 , \\ 0 & , \ , r\in\bigcup_{k\in{\mathbb{z } } } k\pi + ( 0,\frac{\pi}{2 } ) .
\end{cases}\ ] ] for all @xmath406 it has the property @xmath407
. therefore we can deduce for @xmath408 @xmath409 with @xmath410 \}$ ] . to proceed
, we use the relation @xmath411 which is valid for all @xmath412 and @xmath413 .
we obtain @xmath414 next , we see that with a constant @xmath77 independent of @xmath408 @xmath415 altogether , we conclude that @xmath416 if @xmath417 is fixed , we can deduce for @xmath418 the estimate @xmath419 after this preparation , we can now turn to the actual proof of the assertion .
the relation @xmath420 is a direct consequence of and @xmath421 .
let @xmath422 be the intervals defined in .
further , recall the intervals @xmath423 defined in . using and the definition of @xmath424 we calculate @xmath425 the intervals @xmath426 scale like @xmath427 .
hence , if @xmath428 is chosen large enough by @xmath429 the estimate from above is much easier to establish . if @xmath428 such that @xmath430 we have @xmath431 based on lemma [ lem : thetaest ] we can prove the first main result of this article .
[ thm : bound1 ] let @xmath40 be the @xmath3-curvelet frame constructed in section [ sec : curvelets ] for fixed @xmath432 and @xmath114 .
there exists a constant @xmath77 such that for any given @xmath349 every @xmath1-term approximation @xmath35 of @xmath384 with respect to @xmath40 ( not even subject to a polynomial depth search constraint ) satisfies @xmath433 let @xmath349 be fixed and assume that @xmath434 is a linear combination of @xmath3-curvelets @xmath435 with coefficients @xmath436 .
the curvelets @xmath437 satisfy @xmath438 as recorded in .
it follows @xmath439 where @xmath440 for @xmath441 . using the notation @xmath442 and @xmath443 we get with lemma [ lem : thetaest ] @xmath444 we want to bound the right - hand side from below . by ,
the number of tiles in each corona @xmath130 , @xmath122 , is given by @xmath123 , where @xmath445 and @xmath446 for @xmath112 .
let @xmath447 denote the unique number such that @xmath448 since @xmath449 decreases with rising scale we obtain @xmath450 here we used @xmath451 . since @xmath452 we can finally deduce @xmath453 this result can be strengthened if we restrict to greedy @xmath1-term approximations obtained by thresholding the coefficients .
essential is the following observation , which has also been used in @xcite . due to its importance
we give a rigorous proof here .
[ lem : apriori ] there is a constant @xmath77 such that all curvelets @xmath454 , @xmath218 , satisfy @xmath455 let @xmath456 be the functions from and recall that according to the support of @xmath457 is contained in the unit square @xmath458 for every @xmath122 .
let @xmath459 denote the identity operator .
we have the estimate @xmath460 according to lemma [ thm : curvmol ] the right - hand side is bounded uniformly over all scales .
we conclude that there is a constant @xmath77 , independent of @xmath122 , such that @xmath461 in other words @xmath462 . using the representation
we obtain @xmath463 since @xmath464 the proof is finished .
lemma [ lem : apriori ] allows to deduce a simple a - priori estimate of the curvelet coefficient size , namely latexmath:[\ ] ] further , for each @xmath1070 there is a constant @xmath1075 such that @xmath1072 satisfies the estimate @xmath1076 . for completeness ,
let us finally note that the identity especially holds true in case @xmath1054 , with vanishing @xmath1077 . this is a direct consequence of the definition and the taylor series of the cosine .
the author acknowledges support by the bms ( berlin mathematical school ) and thanks prof .
. gitta kutyniok and anton kolleck for proofreading the manuscript , as well as many helpful comments .
e. j. cands and d. l. donoho .
curvelets a surprisingly effective nonadaptive representation for objects with edges . in c.
rabut , a. cohen , and l. schumaker , editors , _ curves and surfaces _ , pages 105120 .
vanderbilt university press , 2000 .
v. chandrasekaran , m. b. wakin , d. baron , and r. g. baraniuk . compressing piecewise smooth multidimensional functions using surflets : rate - distortion analysis .
technical report , department of electrical and computer engineering , rice university , mar . 2004 .
online available : http://dsp.rice.edu/sites/dsp.rice.edu/files/publications/report/2004/compressin-riceece-2004.pdf .
v. chandrasekaran , m. b. wakin , d. baron , and r. g. baraniuk .
compression of higher dimensional functions containing smooth discontinuities . in _ conference on information sciences and systems _ , princeton , mar .
v. chandrasekaran , m. b. wakin , d. baron , and r. g. baraniuk .
surflets : a sparse representation for multidimensional functions containing smooth discontinuities . in _
ieee symposium on information theory _ ,
chicago , jul .
2004 .
d. l. donoho and x. huo .
beamlet pyramids : a new form of multiresolution analysis suited for extracting lines , curves , and objects from very noisy image data . in _
wavelet applications in signal and image processing viii ( san diego , ca , 2000 ) , proc .
volume 4119 , pages 434444 .
spie , 2000 .
p. grohs and a. obermeier . on the approximation of functions with line singularities by ridgelets .
technical report 2016 - 4 , seminar for applied mathematics , eth zrich , switzerland , 2016 .
online available : http://www.sam.math.ethz.ch/sam_reports/reports_final/reports2016/2016-04_fp.pdf .
k. guo , g. kutyniok , and d. labate .
sparse multidimensional representations using anisotropic dilation and shear operators . in _ wavelets and splines ( athens , ga , 2005 ) _ , pages 189201 .
nashboro press , nashville , tn , 2006 . g. kutyniok , d. labate , w .- q lim , and g. weiss .
sparse multidimensional representation using shearlets . in _
wavelets xi ( san diego , ca , 2005 ) , spie proc .
_ , volume 5914 , pages 254262 .
spie , bellingham , wa , 2005 .
a. lisowska .
multiwedgelets in image denoising . in j. park ,
j. ng , h .-
jeong , and b. waluyo , editors , _ multimedia and ubiquitous engineering : mue 2013 _ , pages 311 .
springer netherlands , dordrecht , 2013 . | the class of cartoon - like functions , classicly defined as piecewise @xmath0 functions consisting of smooth regions separated by @xmath0 discontinuity curves , is a well - established model for image data .
the quest for frames providing optimal approximation for this class has among others led to the development of curvelets , contourlets , and shearlets . due to
parabolic scaling , these systems are able to provide @xmath1-term approximations converging with a quasi - optimal rate of order @xmath2 . replacing parabolic scaling by @xmath3-scaling
, one can construct @xmath3-curvelet and @xmath3-shearlet frames which interpolate between wavelet - type systems for @xmath4 , the classic parabolically scaled systems for @xmath5 , and ridgelet - type systems for @xmath6 .
previous research shows that if @xmath7 they provide quasi - optimal approximation for cartoons of regularity @xmath8 with a rate of order @xmath9 . in this work
we continue the exploration of approximation properties of @xmath3-scaled representation systems , with the aim to better understand the role of the parameter @xmath3 . concerning @xmath3-curvelets with @xmath10
, we prove that the best possible @xmath1-term approximation rate achievable for cartoons with curved edges is limited to at most @xmath11 , independent of the smoothness of the cartoons . the maximal rate that can be obtained by simple thresholding of the frame coefficients is even bounded by @xmath12 .
systems of @xmath3-curvelets thus can not take advantage of regularity higher than @xmath8 if @xmath7 , the rate of @xmath9 can not be surpassed . for @xmath13 cartoons with @xmath14
the classic @xmath15-curvelets provide the best performance with a rate of order @xmath2 , however below the optimal rate of order @xmath16 if @xmath17 . in the range
@xmath18 $ ] the achievable rate can not exceed @xmath11 and deteriorates as @xmath3 approaches @xmath19 .
the approximation performance of @xmath3-curvelets is different if the edges of the cartoons are straight . assuming @xmath13 regularity , we establish an approximation rate of order @xmath20 , which improves as @xmath3 tends to @xmath19 . in the range
@xmath21 $ ] it is even quasi - optimal , generalizing optimality results for ridgelets . by applying the framework of @xmath3-molecules , we finally extend the obtained results to other @xmath3-scaled representation systems , including for instance @xmath3-shearlet frames .
_ keywords : _ cartoon images , nonlinear approximation , wavelets , curvelets , shearlets , ridgelets , anisotropic scaling , @xmath3-molecules . _
msc2000 subject classification : _
41a25 , 41a30 , 42c40 . |
we download twelve noaa files of temperature anomaly @xcite , half land- and half ocean - based .
each set covers the globe in six 30@xmath1 zones of latitude and lists monthly surface temperature anomalies and variances from 1880 through 2016 , 1644 measurements in all .
we arrange and number them in contiguous zones n@xmath6s , 1@xmath66 for land and 7@xmath612 for ocean .
we fit each zone separately using identical technique ( starting values , convergence test @xmath8 ) and afterwards , find the weighted mean of seven with acceptable fits .
( data for the four polar zones are too sparse and zone 3 is discussed later . ) let @xmath9 , @xmath10 , be the time of the @xmath11 measurement in years and @xmath12 its temperature anomaly in @xmath1c .
the time range 1880@xmath62017 is awkward for a polynomial fit because the coefficients vary by orders of magnitude and have rather obscure meanings : the first coefficient , for instance , is the temperature at year 0 .
therefore , before fitting , we transform @xmath13 to a dimensionless variable @xmath14 ranging from -1 to 1 as @xmath13 ranges from 1880 to 2017 . because 1 to any power is 1 ,
the resulting polynomial coefficients are all of the same order of magnitude and easier to interpret .
for instance , the first coefficient is simply the temperature anomaly at midrange , and @xmath15 at the end of range ( @xmath16 ) is simply the sum of the coefficients .
the transformation and its inverse are @xmath17 where in our case @xmath18 the fit function is @xmath19 with @xmath20 , @xmath21 and @xmath22 the gaussian amplitude , mean and @xmath23 respectively
. that form was arrived at after some experimentation ( note unused parameter numbers ) where we found that a three - term ( quadratic ) polynomial was the most the data would support .
the fit improves with more terms , but their error increases sharply , a sure sign of too many adjustable parameters .
we use the standard levenberg - marquardt algorithm to minimize @xmath24 with respect to the six parameters @xmath25 .
the weight @xmath26 of the @xmath11 point is the inverse of the variance @xmath27 given in the noaa tables .
our form of the algorithm , adapted from numerical recipes @xcite allows us ( with negligible effort ) to hold selected parameters fixed at their initial values .
initial values for the gaussian are set by hand in an input file and ( along with all other program constants ) are the same for all 12 data sets to avoid bias .
initial values for the polynomial coefficients are obtained from a preliminary linear least - squares fit without the gaussian .
the usual criterion for a good fit namely @xmath28 is too much to hope for here .
first , non statistical effects are present , for instance the el nio southern oscillation ( enso ) in sveral ocean - based zones .
second , eq.[eqn : csqpdf ] assumes accurate variances , very unlikely in this complicated a problem . the best we achieve for any zone is @xmath29
of course , @xmath30 still tells us whether one fit is better than another .
after convergence , following @xcite , we call marquardt once more with @xmath31 , whereupon the variance in each fitted parameter equals the corresponding diagonal term of the curvature matrix . we deliberately choose conditions to make the program look fairly hard for the @xmath30 minimum : convergence requires @xmath32 improvement in @xmath30 and the initial gaussian amplitude is set at @xmath33 its final value .
the program typically converges in @xmath34 passes , except zone3 which takes 12 passes to get the wrong answer , as we will see .
fitting all twelve zones takes about 1s on an elderly laptop running intel visual fortran .
details , tables and graphs can be found in the long version of this report @xcite . raw land - based ( atmospheric ) data show much more scatter than ocean - based and for both , scatter decreases from poles to equator . weights ( inverse variances ) have the same general pattern , varying by 3@xmath64 orders of magnitude month - to - month at the poles . repeating
the final fit with all @xmath35 shows that our results are insensitive to the weights , but we did use them for the results below .
smoothing the raw data by convolution with a gaussian of @xmath36months reveals distinct ww2 bumps in the peri - equatorial ocean - based zones 9 and 10 , and hints of bumps in some others .
smoothing can be a useful guide to where to look , but by itself leads to no numbers .
after fitting , we studied the gaussian compliance of fit residuals by examining their skewness and kurtosis , both ideally zero , and found some worse than others .
enso , definitely a systematic effect , seems to be quasi random when averaged over many cycles . finally , we replaced the ww2 gaussian by a nearly square pulse also having three free parameters ( amplitude , start and duration ) and found it , too , in all eight zones .
fig.[fig : fitgau10 ] shows the best fit , in zone10 ( zone9 is similar ) .
enso is apparent .
fig.[fig : multiyf ] summarizes all final fits .
left- and right - hand columns are land- and ocean - based respectively , with n at the top .
the ordinate range @xmath37@xmath1c is the same for all panels and the abscissa runs from 1880 through 2016 marked every two decades .
the five zones that missed the final cut are cross - hatched .
all six parameters are free to vary in all zones .
zone3 diverts the three ( gaussian ) degrees of freedom so as to improve the background .
jumping to an unsought nearby minimum of @xmath30 is a common pathology of nonlinear least - squares fits , and one or another zone does this for every set of starting conditions we tried .
fig.[fig : multiyflock ] , however , which is the same fit with the gaussian position and @xmath23 fixed but the amplitude still free to vary , shows that the bump is also present in zone3 with a reasonable amplitude .
space precludes listing fit parameters by zone here ( see @xcite ) but fig.[fig : parms ] shows the trends .
all fits are shown including dropped zones .
error bars are @xmath38 ; note scale factors and a few off - scale points .
starting from the left , the difference between l and o constant terms may result from slightly different definitions of temperature ` anomaly ' .
it has no profound consequences .
more interesting are the next two blocks showing linear and quadratic terms both larger for l than for o. both the rate and the acceleration of surface temperature increase seem to be greater for atmospheric than oceanic measurements .
disequilibrium between the atmosphere and the oceans seems to be increasing with time .
relative errors in those numbers are small except for excluded zone3 .
the gaussian amplitude block confirms the impression ( fig.[fig : multiyf ] ) that the ww2 bump is smaller in land - based measurements , but the final blocks show that the time of occurrence and duration are , within errors , the same .
( we repeat that these were free to vary . )
table[tbl : finalparms ] lists seven - zone weighted means of the fit parameters in two ways : in terms of the scaled time variable @xmath14 used for the fits ( left - hand column ) and in conventional units ( right - hand column ) . to compute the weighted fit for a given year convert to @xmath14 using eq.([eqn : u ] ) and insert values from the left - hand column into eq.([eqn :
f ] ) , or simply read it off the bold line in fig.[fig : predicted ] below .
the final two rows of table[tbl : finalparms ] list two derived quantities , the areas of the weighted gaussian and of the weighted square pulse , in excellent agreement .
because each involves both the amplitude and a measure of duration , their relative errors are larger than in the fitted parameters themselves . since the data do not support a more detailed model of the shape , saying that ww2 was equivalent to a @xmath3@xmath1cyr impulse to global surface temperature is probably the most succinct formulation .
it goes without saying that there are other systematic effects ( enso being the most dramatic ) in the temperature record and there may well be other one - shot events , linked to volcanic eruptions or other phenomena .
we simply have not looked for them here .
fig.[fig : predicted ] encapsulates the present study , showing fitted results for three land - based and four ocean - based time series as well as their weighted average .
we extended it about two decades beyond the present day .
a similar figure in @xcite corresponding to the fits in fig.[fig : multiyflock ] looks cleaner but leads to the same conclusions . according to table[tbl : finalparms ] the amplitude @xmath20 of the ww2 bump differs from zero by @xmath39 . from standard probability tables
the chance of that happening at random is @xmath40 . if we therefore accept that the effect is real , there remain three possibilities .
first , a systematic measuring error might have come and gone coincident with ww2 .
that seems superficially plausible ; after all , ww2 was a time of global upheaval .
however , the systematic aggregation and analysis of surface temperature data cited here did not begin until decades later .
therefore it would have been necessary for a significant fraction of land- and ocean - based stations to change their protocols one way at the beginning of the war and back at the end . the nearest candidate for that
is the well known ` bucket cooling ' effect @xcite .
however , that went in only one direction ( after the war ended ) , it only affected ocean measurements , and ersst.v4 attempts to correct for it @xcite .
second , the bump might be pure coincidence : an unrelated climatic event happened to occur at that time , affecting both land- and ocean - based measurements . logically speaking that argument can never be refuted .
the third possibility is that the ww2 bump was caused by the increase ( enormous relative to that era ) in the combustion of fossil fuels by the world s armies , navies , air forces and factories , let alone the burning of oilfields , cities and the munitions themselves .
indeed , it would seem inconsistent to believe that burning fossil fuels promotes global warming and , simultaneously , that ww2 had no effect .
our study was motivated by curiosity about the ww2 bump , but the background behavior of surface temperature which emerges as a byproduct is of infinitely greater consequence and , mathematically , depends very little on the existence or non - existence of the bump .
the linear coefficient , @xmath41@xmath1c / century ( table[tbl : finalparms ] ) agrees with @xmath42@xmath1c / century from an independent fit to ersst.v4 cited by huang et al . @xcite . also , if in our analysis we omit the the quadratic term and the ww2 bump and simply fit everything with a straight line , the weighted average is @xmath43@xmath1c / century , so that is all consistent .
a positive quadratic coefficient is required by the mere fact that surface temperature exhibited a minimum around 1910 ( as appears in many ocean - based data ) or that it was level from ( say ) 1880@xmath61920 and increased thereafter ( as is more typical of land - based data ) .
all we have done here is to quantify that as objectively as possible .
the positive quadratic coefficient in turn guarantees that the present - day instantaneous rate of change is greater than the long - term average .
unfortunately , looking at fig.[fig : predicted ] , it seems to be far greater : about 2.5@xmath1c / century . in summary ,
the good news is that surface temperature reverted to baseline very quickly after ww2 . in that era at least , surface temperature was stable with respect to small changes .
the bad news is that , because of the quadratic term , the present - day rate of temperature increase is very much larger than the average over the last 137 years .
we thank the harvard university physics department for sustained and generous support . opinions in this report are ours and not those of the university .
david w.j .
thompson , john j. kennedy , john m. wallace and phil d. jones , ` a large discontinuity in the mid - twentieth century in observed global - mean surface temperature , ' nature , * 453 * ( 2008 ) 646 - 649 doi:10.1038/nature06982 .
boyin huang , viva f. banzon , eric freeman , jay lawrimore , wei liu , thomas c. peterson , thomas m. smith , peter w. thorne , scott d. woodruff and huai - min zhang , ` extended reconstructed sea surface temperature version 4 ( ersst.v4 ) .
part i : upgrades and intercomparisons , ' j. climate * 28 * ( 2015 ) 911 - 930 . | using parametric analysis ( curve fitting ) we find a persistent temperature bump , coincident with world war ( ww2 ) , in eight independent time series , four land- and four ocean - based .
we fit the data with a gaussian on a quadratic background .
six parameters ( constant , linear and quadratic background terms and the amplitude , position and width of the gaussian ) are free to vary .
the mean fitted gaussian amplitude is @xmath0@xmath1c , non - zero by @xmath2 and therefore not accidental .
the area is @xmath3@xmath1cyr .
temperature recovered to baseline rather quickly . rather than coincidence , or systematic measuring error synchronized with ww2 , we conjecture the bump is due to human activity , including the greatly increased combustion ( relative to that era ) of fossil and other fuels .
background surface temperature behavior , a byproduct of our study but largely independent of the ww2 bump , is far more consequential nowadays .
the linear term , @xmath4@xmath1c / century , agrees well with other findings , but the present - day rate of increase , @xmath5@xmath1c / century , is far greater because of the quadratic term .
= 1 * for haim goldberg ( 1939@xmath62017 ) * graphs of global surface temperature versus time often show a @xmath7@xmath1c bump coincident with world war ( ww2 ) .
taken at face value , that is direct evidence of the effect of human activity on global temperature .
( there is ample indirect evidence . )
the bump could test of climate models if the excess combustion during ww2 could be quantified .
here we show only that the bump is real .
it is a persistent feature of the surface temperature record , appearing in eight independent time series , four land- and four ocean - based
. a least - squares fit of a gaussian on a quadratic background to each data set yields values and errors for the constant , linear and quadratic coefficients of the background as well as the magnitude , time of occurrence and duration of the bump . |
reactor neutrinos have been used for fundamental research since the discovery of neutrinos . the last decade
have witnessed a remarkable reduction of systematic error in reactor neutrino experiments , by about one of order of magnitude , imposed by the high precision needed to measure @xmath0 by double chooz @xcite ( dc ) , daya bay @xcite ( db ) and reno @xcite , a milestone for the world strategy of neutrino flavour research .
reno has released several updates of the @xmath0 analysis in conferences here disregarded until publications follow .
the reactor measurements are consistent with all measurements sensitive to @xmath0 @xcite obtained via other techniques providing a coherent @xmath0 perspective , as obtained by global fit analyses @xcite .
since the reactor @xmath0 experiments precision is unrivalled , they are expected to dominate the world knowledge on @xmath0 , likely , for a few decades to go .
hence , reactor systematic dominates much of the @xmath0 world knowledge , as experiments reach their final sensitivities .
the measured @xmath0 ( and its uncertainty ) is expected to play a critical role to constrain , or measure , still unknown neutrino oscillation observables , such as cp - violation and the atmospheric mass hierarchy @xcite . to maximise the sensitivity to @xmath0 ,
reactor experiments were forced to conceived experimental setups where flux , detection and background systematics are controlled to the unprecedented level of a few per - mille each contribution ( < 1% total ) .
the statical resolution is boosted by using multi - reactors sites .
the unprecedented precision achieved is experimentally very challenging , therefore the redundancy @xmath0-experiments is critical , specially if their uncertainty budgets are complementary to offer maximal cross - validation . despite some complementary , the reactor experimental setups are unavoidably similar and suffer from similar limitations , hence validation by different techniques would be important , although the precision needed is unattainable today .
the fore - mentioned precision improvement was obtained via multi - detector experimental setups , whereby , at least , two detectors are used for the reduction of the overall systematic budget since correlated systematics among detectors cancel out
. this way , while the _ absolute systematics _ are the same , the _ relative systematics _ are much lower .
the _ absolute systematics _ are still dominant in any single - detector setup , such as dc ( single detector ) , but also all past and , likely , most future reactor experiments .
the systematics reduced by the multi - detector configuration are : detection and flux systematics .
detection systematics benefit from dedicated detector design for them to be _ effectively identical _ , typically so , only upon full calibration , thus implying the same ( or negligibly different ) responses and composition ( cross - section , proton number , etc ) .
flux systematics benefit from the fact that the _ near _ detector(s ) is located closer the reactor(s ) such that the flux modulation originating from neutrino oscillations , @xmath0 in this case , is negligible ( or very small ) compared to the _ far _ detector(s ) located further away .
the far is placed at the expected maximal oscillation deficit driven by the @xmath2 ( atmospheric ) constraint by minos @xcite and t2k @xcite .
the suppression of the flux systematic is highly non - trivial , being the main subject of this publication .
while having effectively identical detectors suffices to reduce detection systematics , from @xmath12.0% @xcite to @xmath10.2% @xcite , just having near and far detectors will not necessarily provide a full cancellation of flux systematics - unlike what was originally thought .
there are three mechanisms leading to possible flux systematic reduction to be elaborated in detail within the paper .
first , multi - reactor uncorrelated uncertainties can benefit from having several identical reactors .
second , the near - far geometry of the experimental setup could enhance ability for the near become an effective _ perfect monitor _ to the far ; i.e. cancelling fully the flux systematic error . and , third
, the nature of the reactor uncertainties ; i.e. whether correlated or uncorrelated among reactors ; might be exploited , as measured by multiple detectors .
reactor systematics suppress from single detector scenario , typically @xmath13%
@xcite to < 1% for multi - detector setups .
dc achieves an impressive @xmath11.7% @xcite for a single - detector setup using the bugey4 data for the mean cross - section per fission normalisation of the fission ; i.e. as an effective _ normalisation - only near_. the most accurate reactor flux anti - neutrino spectrum predictions @xcite rely on the ill uranium and plutonium isotopes input data @xcite .
however , the final flux systematics are a combination of the reactor and spectral systematic errors and depends on depends on each experiment configuration , since the evolution of the fission elements depends on the running configuration of each reactor .
therefore flux systematics are expected , with current knowledge , to be the dominant contribution for some of the reactor @xmath0 experiments , such as db , itself leading the world @xmath0 precision .
this publication develops a framework for the non - trivial propagation of the reactor flux uncertainties and their suppression in the context of multi - detector and multi - reactor experimental setups , providing mechanism for the improvement of the global @xmath0 precision .
most our discussion stays generic on flux systematic propagation ; i.e. no need for the specific experiment error breakdown , allowing easy relative comparison across all experiments .
our study cases , however , inspired on the specific reactor-@xmath0 experiments configurations for maximal pertinence .
the core of our calculations is analytical , but a cross - check was implemented using a dedicated monte - carlo - based analysis , is also presented .
the discussion starts from the simplest single detector configuration , then evolving towards a general formalism applicable to any multi - detector and multi - reactor setups .
the numbers linked to the specific experimental setups are however only guidelines as they are obtained in simplified scenarios , again , to maximise comparability .
so far , the reactor @xmath0-experiments appear to treat the reactor systematics similarly . however , all collaborations are lacking dedicated publications on the topic , so questions about the coherence across collaboration remain .
below , we will briefly summarise the reactor flux systematics information , as provided by the different collaborations .
most experiments rely on pwr reactors , hence the anti - neutrino flux is dominated by four isotopes : @xmath3u , @xmath4pu,@xmath5pu and @xmath6u ( ordered by contribution relevance ) . in general , flux systematics uncertainties are typically broken down into three terms : thermal power , fission fractions and spent fuel , as summarised in table [ tab - uncorpar ] .
the thermal power history is usually provided by the electricity company exploiting the reactor .
the associated error is related to the measurement method , the precision of the installed sensors and the employed sensors calibration @xcite .
each experiment has to precisely study the way how the thermal power is measured in order to estimate the corresponding error and the possible correlations among all the reactors involved , as typically , the same measurement techniques are applied to all identical reactors in a power plant .
the fractional fission rates have to be computed through reactor simulations , while their uncertainty estimation is not simple as they depend on the reactor model and the approximations used for each setup @xcite .
after each cycle , the reactors are stopped for refuelling , typically once per year in a few weeks lasting operation . in this operation , part of the fuel assemblies are exchanged with new ones .
the spent fuel are stored in the dedicated pools , typically located next to the reactor sites with slightly different baselines .
the spent fuel long lived fission products can generate a small fraction of anti - neutrinos above the inverse beta decay threshold @xcite .
therefore , each experiment has to evaluate also the contribution from spent fuel and its uncertainty . .
reactor-@xmath0 experiments breakdown of detector flux 1@xmath7 systematic error @xcite . the daya bay and
reno errors quoted here are correspond to their uncorrelated contributions , as they relied on multi - detector setups .
instead , the double chooz errors includes also the correlated contributions , as quoted from single - detector analysis .
therefore , the double chooz error is expected to be overestimated and will be revised by the collaboration in future publications . thermal power
( @xmath8 ) , fission fractions ( @xmath9 ) and spent fuel are considered . [ cols="<,^,^,^,^ " , ] [ tab - sf ] since both @xmath10 and @xmath11 are given by eq .
[ eq : deffigen ] , we can use directly eq . [ eq_i2d2nr_final ] in order to get the expression for the corresponding sf . inserting the values of the geometrical parameters ,
we show the variation of the sf relative weighting parameter in fig .
[ fig - db_tune ] for fully uncorrelated errors of the reactor fluxes .
the dotted curve represents the variation of the sf when all reactors are operational and the continuous curve represents the mean value between the values obtained when one reactor is off upon reactor refuelling scenario considered in the last section .
the variations in such refuelling scenario are indicated by the grey area in fig .
[ fig - db_tune ] accounting for the different reactors .
there are two interesting cases to be highlighted @xmath12 = 1 : : : this point represents no relative weighting applied on the near sites , for which sf is @xmath10.20 .
the spread of sf when considering the reactor refuelling scenario is @xmath12% .
@xmath12 at minimum : : : the effective sf obtained is about @xmath10.05 .
the spread due to refuelling is about @xmath16% .
our result is consistent with the official result of db @xcite , where they obtained @xmath13 = 0.04 and @xmath14 = 0.3 for the analysed data period , considering that one of the near contains twice the events than the other , we obtained @xmath12 to be 3.38 . using mc
, we also obtained respectively , 0.04 and 0.3 .
our calculation here , thus serves as a replication .
however , due to the difficulties on the physical interpretation of this , the corresponding sf is ignored any further in this publication .
this is consistent with the fact that db discards it for the measurement of @xmath0 .
we have identified , studied and quantified three mechanisms inducing reactor flux uncertainty suppression in the context of multi - detector experiments in multi - reactor sites .
we have quantified the integral error suppression by the _ suppression fraction _ ( sf ) analytically ( cross - checked via mc ) using simplified experimental scenarios to allow coherent relative comparison across experiments .
sf can take values within [ 0,1 ] , where the extreme cases stand , respectively , for total suppression ( sf=0 ) and no suppression(sf=1 ) .
the three mechanisms can be characterised by their respective sf terms , since sf(total ) is defined as sf(total ) @xmath15
sf(@xmath16 ) @xmath17 sf(iso - flux ) @xmath17 sf(correlation ) , where _ i _ )
sf(@xmath16 ) is linked to @xmath18 scaling of the remaining uncorrelated reactor error , _ ii _ ) sf(iso - flux ) is linked to the site iso - flux condition and _ iii _ ) sf(correlation ) is linked to the nature of reactor errors .
both sf(iso - flux ) and sf(correlation ) could lead to total error suppression under specific site conditions .
if those terms are to be exploited , those conditions must be carefully evaluated and demonstrated by each experiments accounting accurately for all pertinent effects , although no experiment have ever done this .
our final results relying simplified refuelling scenarios are summarised in table [ tab - sf ] .
this results are not expected to be used by experiments , but as mere guidelines for more accurate estimations to follow up .
the contribution of sf(correlation ) was not quantified for any specific experimental setup , as it deserves more careful treatment discussed below .
the sf(@xmath16 ) has been actively exploited in publications by experiments to improve our knowledge on @xmath0 under the assumption that the remaining error is is fully reactor uncorrelated .
typically , @xmath16 refers to the number of effectively identical reactors per site . in the case of db and reno ,
this term amounts to @xmath160% suppression ( 6 reactors ) , whereas this yields only @xmath130% suppression for dc - ii .
typically , sf(@xmath16 ) is propagated into the @xmath0 precision as a byproduct of the @xmath19 minimisation .
the estimation of sf(iso - flux ) has not been applied by any experiment so far .
it is likely impractically to be implemented via the @xmath19 minimisation formulation , instead calculations might follows the prescription here presented . once estimated , as demonstrated in this publication , the sf(iso - flux ) is expected to improve the so far published results by db and reno , providing an extra flux error reduction by up to @xmath150% and @xmath140% , respectively on @xmath0 . due to the simplified conditions assumed for our calculations ,
our sf s are expected to be slightly optimistic relative to those to be obtained by dedicated analyses by db and reno eventually . in the case
dc - ii , the sf(iso - flux ) term is expected to yield a dramatic @xmath190% error reduction since the iso - flux condition is almost met .
this makes dc - ii the only @xmath0 experiment likely to benefit from a negligible flux error , compared to other systematics .
dc has officially prospected a conservative 0.1% as flux error for dc - ii @xcite , well within the analysis here presented .
therefore , dc - ii final @xmath0 sensitivity is expected to be dominated its challenging background systematic , thus in maximal complementary to db error budget . since the flux error is dominant for db , the hereby presented error reduction by @xmath150% translates into a significant improvement of the world @xmath0 precision by means of db alone , but also via the envisaged combination by all reactor experiments . ,
is shown for dc - ii , reno and db .
all experiments have assumed , so far , their errors to be reactor uncorrelated ; i.e. @xmath20 .
dc - ii benefits the best sf ( @xmath10.1 ) due to its almost iso - flux site . reno and db benefit mainly from the large number of reactor error suppression , but they also have some partial iso - flux matching hence benefiting from an extra up to @xmath150 error suppression so far neglected in @xmath0 publications . both full - reactor power ( dashed lines ) and a simplified refuelling scenario ( solid lines ) are shown , while the latter is expected to be a more accurate description of the reality . in the refuelling scenario , dc - ii benefits from the null reactor flux error whenever only one reactor is running , while reno and db have the expected opposite trend . ]
the sf(correlation ) term has the potential to provide further flux systematic error suppression , regardless of the reactor site geometry , in the ( unexpected ) limit of full correlation across reactors , sf@xmath210 since the sf(correlation ) term cancels , as illustrated in fig .
[ fig - all_1d ] . assessing
the correlation among reactor errors is a difficult subject , having today no specific common prescription or consensus .
thus the only existing consensus is to adopt the most conservative scenario , implying two distinct cases : single - detector case : : : maximal sf(total ) is obtained if _ total correlation of reactor errors _ is assumed , as shown in fig .
[ fig - dc_0near_2d ] .
this is because both sf(correlation ) and sf(@xmath16 ) terms provide no error suppression ; i.e. sf=1 each .
this the scenario is assumed for all dc - i publications and all single detector experiments , unless otherwise proved , thus affecting all past and and most future experiments .
multi - detector case : : : maximal sf(total ) is obtained if _ total uncorrelation of reactor errors _ is assumed , as shown in fig .
[ fig - dc_1near_2d]-right .
this implies sf(correlation)=1 and sf(@xmath16)=@xmath18 , hence benefiting from maximal sf(@xmath16 ) reduction .
this is the scenario expected to be assumed by db , dc - ii and reno , until otherwise proved . of course , reactor errors are unlikely to be neither totally correlated or totally uncorrelated .
the fact that only those extreme cases are considered in the literate is a mere demonstration of the lack of knowledge for a better handling . beyond the current debate among experts on the subject , the promising exploitation of sf(correlation ) will require strong reactor - type dependences to be accounted and justified carefully by each experiment in dedicated publications
this means that each experiment will have to analyse their respective reactors , thus providing insight evidence of the error correlation behaviour .
the thermal power contribution , typically indicated by @xmath8 , depends mainly on the uncertainty analysis of the internal reactor instrumentation data used for power estimation , as provided by the reactor running company .
instead , the fission fraction evolution , indicated by @xmath22 , depends mainly on the simulation uncertainties analysis , including the assumptions and input parameters ( fuel configuration , etc ) used for the modelling and time evolution .
both the instrumentation and simulations are very specific to each reactor type and , therefore , to the each experiment .
therefore , dedicated analyses are needed by each experiment to justify the delicate exploitation of sf(correlation ) , in the same way that this publication aims to illustrates the realisation of the unprecedented sf(iso - flux ) exploitation . as a consequence of the unsettled complications behind the assessment of sf(correlation ) for each experiment , our description here remains generic in its description , as shown in fig .
[ fig - all_1d ] , nonetheless , we illustrate and quantify the rationale behind for error suppression and its promising exploitation potential .
our approach is also consistent with the fact that none of the @xmath0 experiments have so far provided detailed publications on the non - trivial reactor systematics analyses , including the quantitative justification of all assumptions used so far . in the context of the @xmath0 experiments ,
a few references exist dc @xcite and db@xcite ( nothing yet available for reno ) illustrating a fraction of their reactor flux studies , but not dealing with the inter - reactor error correlation needed for the exploitation of sf(correction ) .
dc is , however , finalising a dedicated publication @xcite on their reactor flux systematics quoted so far .
thus , reactor flux error critical subject , despite its major impact to the final precision on @xmath0 , remains unfortunately somewhat obscure in today s literature .
as time goes , the statistical error of reactor experiments is no longer dominant , the treatment of the systematics should be clearly laid well in advance to maximise the stability and reliability of the @xmath0 measurement , whose impact has critical implication transcending reactor neutrino results , affecting , for example , current searches for neutrino cp - violation .
this publication provides reactor neutrino experiments with a coherent treatment for reactor flux systematic uncertainty for multi - detector experiments in multi - reactor contributions .
we started with careful treatment of the single detector in a single reactor site scenario , being the most relevant case for most reactor experiments beyond @xmath0 .
we have demonstrated that the challenging reactor flux systematic do not trivially cancel by the adoption of that multi - detector experiments .
however , we have identified several means for error suppression in the context of double chooz , daya bay and reno experiments , using simplified scenarios to maximise relative comparison .
we computed an integral error _ suppression fraction _
( sf ) , which can be broken down into three components , defined as sf(total)@xmath15sf(iso - flux)@xmath17sf(@xmath16)@xmath17sf(correlation ) , where sf(@xmath16 ) suppresses the uncorrelated error of identical reactors , sf(iso - flux ) suppresses the total error if the site geometry meets fully , or partially , the iso - flux condition and sf(correlation ) suppresses the error if the reactor errors are fully correlated .
sf(iso - flux ) and sf(correlation ) could lead to total cancellation of the flux error .
however , only sf(@xmath16 ) has been exploited to improve the @xmath0 precision , although total cancellation is impossible .
this publication deals in detail on the calculation for sf(iso - flux ) , thus paving the ground for its exploitation , yielding two important observations .
first , dc , once in its near+far configuration , is the only experiment expected to benefit from a negligible reactor flux error , thanks to the @xmath190% iso - flux error suppression .
second , daya bay and reno could also benefit from their partial iso - flux , thus yielding up to @xmath150% flux error suppression .
thus , this publication embodies a major improvement in the global precision of @xmath0 by improving the precision of all experiments measuring it , including current results .
finally , we have highlighted the potential for a mechanism , currently neglected , for error suppression relying on further reactor error correlation insight , characterised by the sf(correlation ) term , further improving the precision of all multi - detector experiments .
we would like to thank h. de kerret ( in2p3/cnrs - apc , france ) as well as c. buck ( mpik , germany ) , l. camilleri ( columbia university , usa ) , r. carr ( columbia university , usa ) , l. giot ( subatech , france ) and m. ishitsuka ( tokyo institute of technology , japan ) for suggestions and comments on the manuscript .
this work was supported by the ief marie curie programme ( p. novella ) .
99 y. abe _ et al . _ [ double chooz collaboration ] , improved measurements of the neutrino mixing angle @xmath0 with the double chooz detector , _ jhep _ * 10 * ( 2014 ) 086 . f. p.
an _ et al . _ [ daya bay collaboration ] , improved measurement of electron antineutrino disappearance at daya bay , _ chin . phys . c _ * 37 * ( 2013 ) 011001 f. p. an _
et al . _
[ daya bay collaboration ] , spectral measurement of electron antineutrino oscillation amplitude and frequency at daya bay , _ phys .
* 112 * ( 2014 ) 061801 j. k. ahn _ et al . _
[ reno collaboration ] , observation of reactor electron antineutrino disappearance in the reno experiment , _ phys .
* 108 * ( 2012 ) 191802 p. adamson _ et al . _
[ minos collaboration ] , measurement of neutrino and antineutrino oscillations using beam and atmospheric data in minos _ phys .
lett . _ * 110 * ( 2013 ) 251801 k. abe _ et al . _
[ t2k collaboration ] , observation of electron neutrino appearance in a muon neutrino beam , _ phys .
* 112 * ( 2014 ) 061802 m.c .
gonzalez - garcia , m. maltoni , th .
schwetz [ global fit analysis ] , updated fit to three neutrino mixing : status of leptonic cp violation , _ jhep _ * 1411 * ( 2014 ) 052 d.v .
forero , m. tortola , j.w.f .
valle [ global fit analysis ] , neutrino oscillations refitted , _ phys .
d _ * 90 * ( 2014 ) 093006 f. capozzi , g.l .
fogli , e. lisi , a. marrone , d. montanino , a. palazzo [ global fit analysis ] , status of three - neutrino oscillation parameters , circa 2013 , _ phys .
d _ * 89 * ( 2014 ) 093018 y .- f .
li _ et al .
_ , unambiguous determination of the neutrino mass hierarchy using reactor neutrinos , _ phys .
d _ * 88 * ( 2013 ) 013008 th .
mueller _ et al .
_ , improved predictions of reactor antineutrino spectra , _ phys .
c _ * 83 * ( 2011 ) 054615 p. huber , determination of antineutrino spectra from nuclear reactors , _ phys .
c _ * 84 * ( 2011 ) 024617 k. schreckenbach _ et al .
_ , determination of the antineutrino spectrum from @xmath23 thermal neutron fission products up to 9.5 mev _ phys .
lett . _ * 160b * ( 1985 ) 325 a. a. hahn _ et al .
_ , antineutrino spectra from @xmath24 and @xmath25 thermal neutron fission products _ phys .
lett . _ * 218b * ( 1989 ) 365 z. djurcic _
et al . _ , uncertainties in the anti - neutrino production at nuclear reactors , _
j.phys.g : nucl .
* 36 * ( 2009 ) 045002 a. onillon , prediction of the reactor fission rates and the estimation of the associated uncertainties in the frame of the double chooz experiment , phd thesis , university of mines nantes ( 2014 ) [ https://tel.archives-ouvertes.fr/tel-01082405 ] an feng - peng _ et al . _ , systematic impact of spent nuclear fuel on @xmath0 sensitivity at reactor neutrino experiment _
chinese phys .
c _ * 33 * ( 2009 ) 711 m. drosg , dealing with uncertainties - a guide to error analysis , second enlarged edition _
isbn 978 - 3 - 642 - 01383 - 6 , springer - verlag berlin heidelberg _ ( 2009 ) , 152 - 154 .
jones _ et al .
_ , reactor simulation for antineutrino experiments using dragon and mure .
* 86 * 012001 ( 2012 ) .
ma _ et al .
_ , uncertainties analysis of fission fraction for reactor antineutrino experiments using dragon
. arxiv:1405.6807 y. abe _ et al .
_ [ double chooz collaboration ] , reactor flux systematic error for the double chooz experiment . _ in preparation_.
in order to deliver the correlation coefficient for partially correlated uncertainties , we generalise the approach presented in @xcite . keeping the same notations , let us consider a measurement @xmath26 which depends on two variables @xmath27 and @xmath28 having the absolute uncertainties : @xmath29 and @xmath30 .
we split one of them , for example @xmath29 , into its two components @xmath31 and @xmath32 representing respectively the totally uncorrelated and totally correlated both relative to @xmath30 .
since the general error propagation formula is symmetric and the correlation is mutual , one can split whether @xmath29 or @xmath30 .
the fraction between the correlated and uncorrelated components is given characterised by a constant @xmath33 such that @xmath34 . since @xmath35 ,
the uncorrelated and correlated components are expressed as | this publication provides a coherent treatment for the reactor neutrino flux uncertainties suppression , specially focussed on the latest @xmath0 measurement .
the treatment starts with single detector in single reactor site , most relevant for all reactor experiments beyond @xmath0 .
we demonstrate there is no trivial error cancellation , thus the flux systematic error can remain dominant even after the adoption of multi - detector configurations .
however , three mechanisms for flux error suppression have been identified and calculated in the context of double chooz , daya bay and reno sites .
our analysis computes the error _ suppression fraction _ using simplified scenarios to maximise relative comparison among experiments .
we have validated the only mechanism exploited so far by experiments to improve the precision of the published @xmath0 .
the other two newly identified mechanisms could lead to total error flux cancellation under specific conditions and are expected to have major implications on the global @xmath0 knowledge today .
first , double chooz , in its final configuration , is the only experiment benefiting from a negligible reactor flux error due to a @xmath190% geometrical suppression .
second , daya bay and reno could benefit from their partial geometrical cancellation , yielding a potential @xmath150% error suppression , thus significantly improving the global @xmath0 precision today .
and third , we illustrate the rationale behind further error suppression upon the exploitation of the inter - reactor error correlations , so far neglected .
so , our publication is a key step forward in the context of high precision neutrino reactor experiments providing insight on the suppression of their intrinsic flux error uncertainty , thus affecting past and current experimental results , as well as the design of future experiments . |
in this proceedings , we summarize our work on the determination of the topological susceptibility and zero mode size distribution of @xmath1 and @xmath0 gauge fields using the overlap formalism . most of this work has appeared in print @xcite .
we will focus here on new results which are extensions of the published work , namely the topological susceptibility for @xmath1 pure gauge theory , the zero mode size distribution for both @xmath1 and @xmath0 , and the density of zero eigenvalues @xmath2 for @xmath1 and @xmath0 .
a more detailed discussion of the role of the zero eigenvalue density can be found in the proceedings of the plenary talk given by narayanan @xcite .
the overlap formalism for constructing a chiral gauge theory on the lattice @xcite provides a natural definition of the index , @xmath3 , of the associated chiral dirac operator .
the index is equal to half the difference of negative and positive eigenvalues of the hermitian wilson - dirac operator @xmath4 where @xmath5 is the usual wilson
dirac operator ( we will use negative sign for the mass term throughout ) . on the lattice , because of the additive mass renormalization , the crossings of zero occur at positive @xmath6 and spread out in @xmath6 .
it is easy to see , that no eigenvalues of @xmath7 can be zero for @xmath8 . since in the free case the first doublers become massless at @xmath9
we restrict ourselves to @xmath10 . a simple way to compute the index @xmath3 is to compute the lowest eigenvalues of @xmath7 at some suitably small @xmath6 before any crossings of zero occurred .
then @xmath6 is slowly varied and the number and direction of zero crossings are tracked .
the net number at some @xmath11 is the index of the overlap chiral dirac operator .
we have applied this procedure to compute the index , which we take as the definition of topological charge on the lattice , of various gauge field ensembles with gauge group su(3 ) including pure gauge , symanzik improved pure gauge , and two dynamical flavor generated gauge fields @xcite , and recently pure gauge su(2 ) .
we found that the zero crossings start occurring at some , ensemble dependent , @xmath12 and continue occurring for all @xmath6 in @xmath13 in sufficiently large lattice volumes .
hence , the spectral gap is closed in the entire range . to further investigate the zero level crossings , we compute the density of zero eigenvalues @xmath14 for @xmath7 by fitting the integrated density @xmath15 to a line through the origin for some cutoff @xmath16 . by studying the scaling of @xmath14 for quenched @xmath0
, we conclude that the density of zero eigenvalues falls exponentially in the inverse lattice spacing and only vanishes in the continuum limit ( see figure 6 of narayanan s talk @xcite ) .
we will observe latter that these zeros are due to small localized modes .
more discussion can be found in the talk of narayanan @xcite .
using the index of the chiral dirac operator as our ( fermionic ) definition of the topological charge of the gauge field background , we obtain the topological susceptibility @xmath17 as a function of @xmath11 .
all of the gauge ensembles studied have the general characteristic that the susceptibility rises sharply in the region where @xmath14 is peaked , and then it essentially flattens out in the region where @xmath14 is small @xcite .
we define a size of the eigenvector associated with the level crossing zero mode @xmath18 motivated by the thooft zero mode @xmath19 where @xmath20 are the eigenvectors of @xmath21 .
another definition based on the second moment of @xmath22 was used in ref .
we should emphasize that we look only at the sizes of eigenmodes that cross , and only close to the crossing point . only then
can we expect to get a good estimate of the localization size inspired by the t hooft zero mode .
we show in figure [ b6.0_all ] a detailed study of the @xmath23 , @xmath24 pure gauge ensemble . on the first line
is shown the density of zero eigenvalues @xmath14 and the number of crossings in each mass bin .
since there are a nonzero number of crossings , we see that @xmath14 does indeed measure zero eigenvalues , and not just small eigenvalues near zero .
we also see that @xmath14 rises sharply in @xmath6 , then falls to a nonzero value where there is a small number of zero level crossings . on the second line of figure [ b6.0_all ]
, we show the size of the zero modes @xmath25 .
the modes are large near @xmath26 where @xmath14 is large , then @xmath27 drops to about @xmath28 or @xmath29 lattice spacings up to @xmath9 .
we see that the corresponding @xmath17 rises sharply when @xmath27 is large for @xmath6 near @xmath26 , then is quite stable when @xmath27 is small .
this result show that while the index , @xmath3 , of the field is @xmath6 dependent , @xmath17 ( a physical quantity ) is independent of the contribution from the small modes for @xmath30 . to further clarify the relative contribution of the zero modes , in the last line of figure [ b6.0_all ] the zero mode size distribution
is shown which peaks for @xmath31 . in the adjacent graph ,
@xmath17 , here defined by the contribution of zero modes of size @xmath27 and larger , is stable when @xmath31 .
hence , the small modes do not affect the estimate of @xmath17 .
our estimates of @xmath17 are shown in table [ tab : results ] where we use the string tension value @xmath32(mev ) to set the scale .
our results are in rough agreement with other groups @xcite .
in figure [ dn_all ] we plot the zero mode size distribution for several @xmath33 s for @xmath0 and @xmath1 which is of interest for the instanton liquid model .
we find the distributions always peaks for sizes about @xmath28 - @xmath29 lattice spacings . for @xmath0 , we see small secondary peaks at @xmath34fm for @xmath23 and @xmath35 .
however , there is only a shoulder for @xmath36 at @xmath37fm . for @xmath1 which are all the same lattice size , we find the secondary peaks shift to smaller @xmath27 for increasing @xmath33 .
these secondary peaks do not occur at any general physical radius and are consistent with a finite volume effect .
edwards , u.m heller , r. narayanan , r.l .
singleton , jr .
phys . b518
( 1998 ) 319336 ; r.g .
edwards , u.m .
heller and r. narayanan , nucl . phys . b522 ( 1998 ) 285300 ; hep - lat/9802016 , to appear in nucl .
b ; hep - lat/9806011 , to appear in phys . lett . b. | we use the overlap formalism to define a topological index on the lattice .
we study the spectral flow of the hermitian wilson - dirac operator and identify zero crossings with topological objects .
we determine the topological susceptibility and zero mode size distribution , and we comment on the stability of our results .
# 1to 0pt#1 3@xmath0 # 1*[#1 ] * |
consider a pair @xmath2 , where @xmath3 is a compact metric space with metric @xmath4 and @xmath5 is a continuous map .
we call @xmath6 a _ flow _ or a _ dynamical system _ because we will consider iterations @xmath7
. a sequence of complex numbers @xmath8 , @xmath9 , is said to be _ observed _ in @xmath6 if there is a continuous function @xmath10 and a point @xmath11 such that @xmath12 . given a sequence of complex numbers
@xmath13 , @xmath14 .
we say that it is _ linearly disjoint _ from the flow @xmath6 if we have @xmath15 for any observable @xmath8 in @xmath6 .
an interesting sequence is the mbius function @xmath16 .
recall that by definition @xmath17 , @xmath18 if @xmath19 is a product of @xmath20 distinct primes , and @xmath21 if @xmath19 is not square - free .
sarnak has conjectured that the mbius function @xmath22 is linearly disjoint from all flows with zero topological entropy ( see @xcite ) .
this is now called sarnak s conjecture , which remains open in its generality although it is proved in several special cases @xcite . as pointed out in @xcite
, this conjecture has a connection with many important problems in number theory , for example , the riemann hypothesis .
more recently , in @xcite , ge studied sarnak s conjecture in association with a @xmath23 algebra and its maximal ideal space and in @xcite , which contains a good survey of the subject and a rather complete list of refereces , el . abdalaoui et al . considered the sarnak conjecture for sequences of numbers in @xmath24 by comparing several natural generalizations
an important property satisfied by the mbius function is @xmath25 which is equivalent to the prime number theorem that the number of prime numbers less than or equal to @xmath26 is approximately @xmath27 . a stronger statement ( conjecture )
is that for @xmath28 @xmath29 which is equivalent to the riemann hypothesis .
another property satisfied by the mbius function is davenport s theorem @xcite that @xmath30 for any real number @xmath31 and hua s theorem @xcite @xmath32 for any real number @xmath31 and any integer @xmath33 .
the reader can refer to @xcite for more delicate estimates for these sums for @xmath34 . in this paper
, we will not only consider the mbius function but also a sequence of complex numbers @xmath35 satisfying the oscillating condition @xmath36 and the growth condition @xmath37 for some @xmath38 ( such a sequence @xmath39 will be called an oscillating sequence ) .
there are many arithmetic functions , including the mbius function and the liouville function , which are oscillating sequences ( see @xcite ) .
recall that the liouville function @xmath40 is defined as @xmath41 where @xmath42 is the number of prime factors of @xmath19 , counted with multiplicity . on the other hand
, we will define a minimally mean attractable flow and a minimally mean - l - stable flow .
we will prove that all oscillating sequences are linearly disjoint from all minimally mean attractable and minimally mean - l - stable flows .
we will prove that @xmath0-adic polynomial flows , @xmath0-adic rational flows with good reduction , automorphism flows with zero topological entropy on @xmath1-torus , all diagonalized affine flows of @xmath1-torus with zero topological entropy , all feigenbaum zero topological entropy flows and all orientation - preserving circle homeomorphism flows are all minimally mean attractable and minimally mean - l - stable flows .
we organize our paper as follows . in section 2 , we define oscillating sequences ( definition [ os ] ) . in section 3 ,
we define minimally mean - l - stable flows ( definition [ mmls ] ) and minimal mean attractable flows ( definition [ mma ] ) .
we will also review mean - l - stable flows and its equivalent notion , mean - equicontinuous flows . in section 4 , we prove that all oscillating sequences are linearly disjoint from all minimal mean attractable and minimally mean - l - stable flows ( theorem [ main1 ] ) .
in particular , as an example , we have that the mbius function is linearly disjoint from all minimal mean attractable and minimally mean - l - stable flows ( corollary [ scmmammls ] ) .
as another example , we have that all oscillating sequences are linearly disjoint from all equicontinuous flows ( corollary [ m2 ] ) . in section 5
, we discuss @xmath0-adic polynomial flows and @xmath0-adic rational flows .
all polynomials with @xmath0-adic integral coefficients define equicontinuous flows on the ring of @xmath0-adic integers .
thus they are minimal mean attractable and minimally mean - l - stable flows .
all @xmath0-adic rational maps with good reduction define minimal mean attractable and minimally mean - l - stable flows on the projective line @xmath43 . therefore
, all oscillating sequences are linearly disjoint from all these @xmath0-adic polynomial flows and @xmath0-adic rational flows ( corollary [ pp ] and corollary [ rp ] ) . in section 6
, we explain how to use our method to study affine maps and automorphisms of @xmath1-torus with zero topological entropy .
we first prove that in the diagonalized case , all oscillating sequences are linearly disjoint from all flows defined by affine maps of @xmath1-torus with zero topological entropy ( corollary [ diagcoro ] ) since these flows are equicontinuous ( proposition [ diag ] ) .
we then prove that all flows defined by automorphisms of @xmath1-torus with zero topological entropy are minimal mean attractable and minimally mean - l - stable flows [ nondiagthm ] ( they are not equicontinuous ) .
thus all oscillating sequences are linearly disjoint from all flows defined by automorphisms of @xmath1-torus with zero topological entropy ( corollary [ nondiagcoro ] ) .
furthermore , we give an affine map of @xmath1-torus with zero topological entropy and an oscillating sequence such that this oscillating sequence is not linearly disjoint from the flow defined by this affine map .
notice that liu and sarnak has proved that the mbius function is linearly disjoint from this example in @xcite .
we further note that in order to be linearly disjoint from all flows with zero topological entropy , an oscillating sequence should not only be oscillating in the first order but also be oscillating for any higher order ( remark [ higher ] ) and the mbius function has this property as shown by hua in @xcite ( see equation ( [ hua ] ) ) . in section 7
, we prove that all feigenbaum zero topological entropy flows are minimal mean attractable and minimally mean - l - stable flows .
thus all oscillating sequences are linearly disjoint from all feigenbaum zero topological entropy flows ( theorem [ main2 ] ) . in section 8 , we prove that denjoy counter - examples are minimal mean attractable but not equicontinuous even when restricted to their non - wandering sets , which are minimal subsets .
however , we prove that denjoy counter - examples are minimally mean - l - stable ( theorem [ ceq ] ) . in the proof of this theorem , we first prove that the flow defined by a denjoy counter - example is minimally mean - l - stable when its non - wandering set has zero lebesgue measure , and then we show that every denjoy counter - example is conjugate to a denjoy counter - example whose non - wandering set has zero lebesgue measure ( an obseravtion pointed out to us by davit karagulyan ) .
thus all oscillating sequences are linearly disjoint from all flows defined by denjoy counter - examples ( corollary [ scdenjoy ] ) .
the case of mbius sequence was already considered by karagulyan @xcite and now is one of the consequences of theorems [ main1 ] and [ ceq ] ( corollary [ kar ] ) .
_ acknowledgement .
_ we would like to thank professors hedi daboussi , liming ge , davit karagulyan , jianya liu , jrg schmeling , jie wu , xiangdong ye and enrique pujals for sharing information and having many interesting discussions with us .
we would also like to thank professor peter sarnak for his encouragement comments on our initial version of this paper .
the work was started when the first author was visiting lund university and he would like to thank the knut and alice wallenberg foundation for its support . both authors would like to thank the academy of mathematics and systems science at the chinese academy of sciences for its hospitality when they visited there and worked on this project .
in classical analysis ( see @xcite ) , a sequence of complex numbers @xmath44 is called a generalized almost periodic sequence if for any @xmath45 the following limit exists @xmath46 almost periodic and generalized almost periodic sequence were studied by bohr , hartman , et al .
the spectrum of a generalized almost periodic sequence is the set of @xmath47 such that @xmath48 . in general
, the spectrum is countable .
we are interested in generalized almost periodic sequence with empty spectrum defined on positive integers .
thus we introduce the following definition . [ os ]
let @xmath49 , @xmath14 , be a sequence of complex numbers .
we say that @xmath39 is an _ oscillating sequence _ if the cesaro means @xmath50 converge to zero as @xmath26 tends to the infinity for every @xmath51 .
all oscillating sequences in this paper are also assumed satisfying the growth condition ( [ growth ] ) .
if we only assume that @xmath52 exists but is not necessarily zero for every @xmath53 , we denote @xmath54 so , the statement that @xmath39 is an oscillating sequence means @xmath55 the complementary of @xmath56 is defined to be the spectrum of * c*. let us give several examples of oscillating sequences .
[ ex1 ] the sequence of complex numbers @xmath57 for a fixed @xmath31 is not an oscillating sequence , because @xmath58 note that the spectrum is one - point set @xmath59 .
[ ex2 ] the sequence of complex numbers @xmath60 is an oscillating sequence because the cesaro means @xmath61 uniformly on @xmath51 ( ( * ? ? ?
* vol . 1 ,
p. 197 ) ) .
[ ex3 ] the sequence of complex numbers @xmath62 for any fixed irrational @xmath63 is an oscillating sequence because the sequence @xmath64 is uniformly distributed on the unit interval @xmath65 $ ] for any fixed @xmath51 .
so the cesaro means @xmath66 [ n2 ] consider the sequence of complex numbers @xmath67 where @xmath68 is a rational number with @xmath69 .
the spectrum of @xmath39 is the set of rational numbers @xmath70 such that @xmath71 and @xmath72 we have @xmath73 if @xmath74 is irrational or if @xmath75 with @xmath76 , the last sum is bounded and the above limit is zero .
then such a @xmath74 is not in the spectrum . if @xmath71 , we have @xmath77 it is easy to check that @xmath78 @xmath79 @xmath80 the following example is due to davenport @xcite ( see also @xcite ) .
[ ex4 ] the mbius function @xmath81 is an oscillating sequence .
actually a stronger result holds for the cesaro means @xmath82 for any @xmath83 where the estimate is uniform on @xmath51 .
another example can be obtained from the following proposition .
let @xmath84 be a sequence of independent and identically distributed real random variables such that @xmath85 such a sequence is called a subnormal sequence .
[ ex5 ] suppose @xmath86 ) is a subnormal sequence .
let @xmath87 be a sequence of real numbers such that @xmath88 for some @xmath89 .
then almost surely the random sequence @xmath90 is an oscillating sequence . actually , as a consequence of the salem - zygmund inequality on the uniform estimate of random trigonometric polynomials and as we refer to @xcite , where the rademacher random sequence was considered but the proof for subnormal sequences is the same , we have almost surely the cesaro means @xmath91 where the estimate is uniform on @xmath51 .
this proves the proposition . from @xcite we can get other oscillating sequences similar to those in proposition [ ex5 ] .
the next example is due to daboussi - delange @xcite .
[ ex6 ] let @xmath35 be a multiplicative arithmetical function such that @xmath92 for all @xmath19 .
then @xmath39 is an oscillating sequence if and only if for any dirichlet character @xmath93 and and real number @xmath94 , we have @xmath95 where the sum is taken over prime numbers @xmath0 .
this classical result of daboussi and delange is generalized to asymptotic orthogonality of multiplicative functions to `` irrational '' nilsequences by n. frantzikinakis and b. host @xcite .
the oscillating sequences shares the following property that arithmetic subsequences of an oscillating sequence are oscillating .
suppose that @xmath96 is an oscillating sequence .
then for any @xmath97 and any @xmath98 we have @xmath99 the idea comes from @xcite .
the proof is based on the fact that @xmath100 according to @xmath101 or @xmath102 ( @xmath103 ) .
it follows that @xmath104 as @xmath105 , since @xmath39 is an oscillating sequence .
the proof shows that for ( [ c - mod - q ] ) to hold for a fixed @xmath74 it suffices that @xmath106 for all @xmath107 . in @xcite , kahane and saias
studied completely multiplicative functions with zero sum .
these functions share another kind of oscillating property .
in this section we define a class of flows from which all oscillating sequences will be proved to be linearly disjoint .
we start with recalling the classical definition of equicontinuous flows .
[ eqcn ] a flow @xmath108 is said to be _ equicontinuous _ if the family @xmath109 is equicontinuous .
that is to say , for every @xmath110 there is a @xmath111 such that whenever @xmath112 with @xmath113 , we have @xmath114 for @xmath115 .
it is well known that a flow @xmath108 with @xmath116 being surjective is equicontinuous if and only if there exists a compatible metric @xmath117 on @xmath3 such that @xmath116 acts on @xmath3 as an isometry , i.e. , @xmath118 for all @xmath119 .
thus , if @xmath116 is surjective and equicontinuous , then @xmath116 must be a homeomorphism .
moreover , it is also known that for an equicontinuous flow @xmath6 , if @xmath116 is a homeomorphism , then @xmath116 acts as a minimal system on the closure @xmath120 of the forward orbit @xmath121 of any @xmath122 ( see @xcite ) .
thus , for an equicontinuous flow @xmath6 , when @xmath116 is a transitive homeomorphism , it is always conjugate to a minimal rotation on a compact abelian metric group . in this case
, let @xmath123 be the unique haar probability measure on @xmath3 , then the measurable dynamical system @xmath124 has discrete spectrum . an equicontinuous flow @xmath6 such that @xmath116 is a homeomorphism can be decomposed into minimal subsystems .
sarnak s conjecture holds for such a flow .
actually , we will prove that sarnak s conjecture holds for a much larger class of flows which we will call minimally mean attractable and minimally mean - l - stable flows . we give a more detailed description below .
we first weaken the equicontinuity condition following fomin in @xcite .
suppose @xmath125 is the set of natural numbers .
let @xmath126 be a subset of @xmath127 .
the upper density of @xmath126 is , by definition , @xmath128)}{n}.\ ] ] [ mma ] a flow @xmath108 is said to be _ mean - l - stable _ ( * mls * for short ) ( here @xmath129 recalls the sense of lyapunov ) if for every @xmath110 , there is a @xmath111 such that @xmath113 implies @xmath130 for all @xmath131 except a subset of natural numbers with the upper density less than @xmath132 .
fomin proved that a minimal and * mls * flow is uniquely ergodic in @xcite .
oxtoby proved that a transitive and * mls * flow is uniquely ergodic in @xcite .
li , tu and ye further proved that every ergodic invariant measure on a * mls * flow has discrete spectrum .
any flow @xmath6 admits a minimal sub - flow , for example , the restriction on the @xmath133-limit set of any point .
[ mmls ] we say that a flow @xmath134 is _ minimally * mls * _ ( * mmls * for short ) if for every minimal subset @xmath135 , the sub - flow @xmath136 is * mls*. following @xcite , we say the flow @xmath137 is _ mean - equicontinuous _ at a point @xmath11 if for every @xmath110 , there is @xmath111 such that for every @xmath138 with @xmath139 we have @xmath140 we say that the flow @xmath137 is mean - equicontinuous if it is mean - equicontinuous at every point @xmath122 .
that is , by the compactness of @xmath3 , the above inequality ( [ mec ] ) holds for all @xmath141 such that @xmath139 .
the following proposition is proved by li , tu , and ye in @xcite [ eq ] a flow @xmath137 is mean - equicontinuous if and only if it is * mls*. let @xmath142 be the space of all continuous functions @xmath10 with maximum norm @xmath143 [ meaneq ] suppose the flow @xmath137 is * mls*. suppose @xmath35 is a sequence satisfying the growth condition ( [ growth ] ) .
then for any continuous function @xmath144 , @xmath145 is an equicontinuous sequence in @xmath142 .
suppose @xmath146 , otherwise it is trivial .
take a @xmath147 satisfying ( [ growth ] ) .
suppose @xmath148 be the number such that @xmath149 there is a positive constant @xmath150 ( we can take @xmath150 to be the supremum of @xmath151 ) such that for any @xmath141 , we have @xmath152 this is a direct consequence of the hlder inequality and ( [ growth ] ) . by the uniform continuity of @xmath153 , for any @xmath28 there exists @xmath154 such that @xmath155 we can take @xmath156 . since @xmath6 is * mls * , there is a @xmath157 such that @xmath141 with @xmath158 implies @xmath159 for all positive integer @xmath19 except a set , denoted @xmath126 , of upper density less than @xmath160 .
there exists an integer @xmath161 such that @xmath162)}{n } < \eta , \qquad \forall n \ge n^*\ ] ] thus by ( [ holder1 ] ) , if @xmath163 and if @xmath164 we have @xmath165 in the above inequality , we first split the sum in ( [ holder1 ] ) into two sums according to @xmath166 or @xmath167
. the finite many family @xmath168 being equicontinuous , there exists @xmath169 such that @xmath170 finally we conclude the proposition from ( [ difference * ] ) and ( [ difference * * ] ) .
in @xcite , auslander had decomposed a * mls * flow into a star closed decomposition . in this decomposition
each component contains precisely one minimal set and all invariant measures concentrated on minimal sets .
this implies that from measure - theoretical point of view , the union of these minimal sets is an attractor : all points outside the union of these minimal sets are attracted into minimal sets eventually .
we will develop this idea into a new concept called the minimal mean attractablity .
[ mma ] suppose @xmath108 is a flow .
suppose @xmath171 is a closed subset of @xmath3 and @xmath172 is minimal .
we say @xmath122 is _ mean attracted _ to @xmath171 if for any @xmath28 there is a point @xmath173 ( depending on @xmath174 and @xmath132 ) such that @xmath175 the _ basin of attraction _ of @xmath171 , denoted @xmath176 , is defined to be the set of all points @xmath174 which are mean attracted to @xmath171 .
it is trivial that @xmath177 .
we call @xmath6 _ minimally mean attractable _ ( * mma * for short ) if @xmath178 where @xmath171 varies among all minimal subsets of @xmath3 .
recall that a point @xmath122 is attracted to @xmath171 if @xmath179 in general , this can not imply to that @xmath174 is mean attracted to @xmath171 . however ,
if @xmath171 is a periodic cycle , it does .
we have even the following proposition .
let @xmath122 .
suppose that @xmath180 is a periodic cycle of period @xmath34 .
then for any @xmath28 sufficiently small , there exists an integer @xmath181 and @xmath182 such that @xmath183 let @xmath184 .
take any @xmath28 sufficient small so that the balls @xmath185 , @xmath186 , are disjoint .
for any @xmath187 and any @xmath181 , consider the subset of the natural numbers @xmath188 for any fixed @xmath189 , @xmath190 by the continuity of @xmath116 , we have @xmath191 this implies that we can find a @xmath192 such that @xmath193 let @xmath194 .
take @xmath195 .
then @xmath196 and from ( [ recurrence ] ) , @xmath197 we take @xmath198 such that @xmath199 , then we have that @xmath200 [ 1ex ] the flow @xmath201 , t(x)=-x^{2})\ ] ] is * mma * and * mmls*. the point @xmath202 and @xmath203 are only two fixed points . since @xmath204 and @xmath205 , @xmath202 is the repelling fixed point and @xmath203 is the attracting fixed point .
we have that @xmath206 thus we have @xmath207= { \rm basin}(\{-1\})\cup { \rm basin}(\{0\}).\ ] ] this implies that @xmath6 is * mma * and * mmls*. this flow is actually equicontinuous on any closed interval @xmath208\subset ( -1,1)$ ] .
the next example is different ( see remark [ neq ] ) .
let @xmath209 be the standard 2-torus .
consider the automorphism @xmath210 for a fixed integer @xmath74 .
[ 2ex ] the flow @xmath211 is * mma * and * mmls*. actually we have that all automorphisms on @xmath212 with zero topological entropy are * mma * and * mmls*. let @xmath213 be the circle .
every circle @xmath214 ( @xmath215 being fixed ) is @xmath216-invariant .
the restriction of @xmath216 on each @xmath217 is a one - dimensional rotation with the rotation number @xmath218 .
if @xmath218 is irrational , then @xmath219 is minimal .
if @xmath218 is rational , then every point @xmath220 is a periodic point .
then the restriction of @xmath216 on each of these periodic orbits is minimal . in this case ,
furthermore , we have that the minimal decompsition @xmath222 so this flow is * mma*. since @xmath216 on each @xmath217 is isometric , so it is equicontinuous on each of its minimal subset .
so it is * mmls*. as we shall see in section 6 , any automorphism @xmath116 on @xmath212 with zero topological entropy is ( linearly ) topologically conjugate to a @xmath216 for some integer @xmath74 .
so the flow @xmath223 is also * mma * and * mmls*. [ neq ] the flow @xmath224 is not mean - equicontinuous ( equivalently , not * mls * ) because the rotation number of @xmath225 varies with @xmath215 .
recall that @xmath108 is a flow if @xmath3 is a compact metric space with metric @xmath226 and @xmath227 is a continuous map .
a sequence @xmath228 is said to be _
observable _ in @xmath6 if there exists a continuous function @xmath144 and a point @xmath122 such that @xmath12 for all @xmath229 .
following the idea of sarnak @xcite , we say that a given sequence of complex numbers @xmath35 is _ linearly disjoint _ from @xmath6 if for any observable @xmath228 in @xmath6 , we have @xmath230 the first main result in this paper is the following one .
[ main1 ] any oscillating sequence @xmath35 satisfying the growth condition ( [ growth ] ) is linearly disjoint from all * mma * and * mmls * flows @xmath134 .
more precisely , for any continuous function @xmath144 , we have that @xmath231 for every @xmath122 and the limit is uniform on each minimal subset . by the hypothesis , @xmath3 has the decomposition in equality ( [ decomp ] ) .
so , we need only prove that for any minimal subset @xmath232 , we have @xmath233 for this purpose , we define @xmath234 we first prove that the sequence @xmath235 converges to @xmath203 uniformly on @xmath171 . the sequence @xmath236 is uniformly bounded in @xmath237 by @xmath238 for some constant @xmath239 and is equicontinuous by proposition [ meaneq ] .
so , it is precompact in @xmath237 by the arzela - ascoli theorem .
therefore , every subsequence of @xmath236 has a convergent subsequence .
what we have to prove is the claim that every convergent subsequence converges uniformly on @xmath171 to zero .
let us prove this claim by contradiction .
suppose that some subsequence @xmath240 converges uniformly to a continuous function @xmath241 which is not identically zero .
assume @xmath242 for some @xmath243 and we can actually assume that @xmath244 for some @xmath245 by the continuity of @xmath246 . on the other hand , we consider the probability measures on @xmath171 : @xmath247 where @xmath248 denote the dirac measure concentrated at @xmath215 .
let @xmath249 be a weak limit of a subsequence of the sequence @xmath250 .
then it is @xmath116-invariant . since @xmath171 is minimal , by the gottschalk theorem ( see ( * ? ? ?
* lemma 4 ) ) , @xmath251 is almost periodic in the sense that for every @xmath28 , the set of @xmath19 such that @xmath252 is relatively dense .
this implies that @xmath253 for any @xmath28 ( the support of @xmath249 is actually whole minimal set @xmath171 ) .
now consider the measure - preserving dynamical system @xmath254 where @xmath255 is the borel @xmath256-field of @xmath171 .
let @xmath257 be the spectral measure of @xmath153 with respect to the @xmath116-invariant probability measure @xmath249 on @xmath171 , @xmath153 being considered as in @xmath258 and @xmath257 being defined on the unite circle @xmath259 by the spectral lemma ( see @xcite ) , we have @xmath260 since @xmath261 is an oscillating sequence , by the lebesgue dominated convergence theorem , the right hand of the above equation converges to @xmath203 .
this implies that @xmath262 converges to zero in @xmath258-norm .
in particular , @xmath263 converges to zero in @xmath258-norm .
consequently , there is a subsequence @xmath264 of @xmath265 such that @xmath266 converges to zero @xmath249-almost everywhere , which contradicts ( [ contradiction2 ] ) and the fact that @xmath267 .
the contradiction implies that the sequence @xmath236 converges uniformly on @xmath171 to @xmath203 .
we have just proved the limit ( [ limsup - k ] ) for @xmath268 .
now we are going to prove ( [ limsup - k ] ) for a general point @xmath269 .
let @xmath154 be an arbitrarily small number . by the uniform continuity of @xmath153 ,
there is a @xmath28 such that @xmath270 implies @xmath271 .
we assume that @xmath272 . for @xmath273 ,
by the definition of @xmath274 , there exists @xmath275 such that @xmath276 which implies that @xmath277 where @xmath278 , because @xmath279 ) \le \sum_{n=1}^{n } d(t^n x , t^n z).\ ] ] write @xmath280 as @xmath198 , we have proved that @xmath281 tends to zero . on the other hand , following the exact same proof of proposition [ meaneq ] , we show that @xmath282 since @xmath283 is arbitrary , we have thus proved ( [ limsup - k ] ) for every @xmath273 .
a special case of theorem [ main1 ] is that [ scmmammls ] the mbius function is linearly disjoint from all * mma * and * mmls * flows .
this is because the mbius function gives an oscillating sequence due to davenport s theorem ( see example [ ex4 ] ) .
another consequence of theorem [ main1 ] is that [ m2 ] any oscillating sequence satisfying the growth condition ( [ growth ] ) is linearly disjoint from all equicontinuous flows .
let @xmath134 be an equicontinuous flow .
the sequence of compact sets @xmath284 decrease to a non - empty compact set , which we denote by @xmath285 .
the sub - flow @xmath286 is surjective and equicontinuous and thus a homeomorphism .
the set @xmath285 is decomposed into minimal subsets .
we claim that each point @xmath122 is mean attracted to some minimal set .
in fact , the sequence @xmath287 admits a subsequence @xmath288 converging to a limit point @xmath289 , which must belong to @xmath285 .
recall the definition of equicontinuity for @xmath116 : for any @xmath28 there exists @xmath154 such that @xmath290 take @xmath291 sufficiently large such that @xmath292 . as @xmath286 is a homeomorphism , there exist @xmath293 such that
@xmath294 so @xmath295 by the equicontinuity , we have @xmath296 it follows that @xmath297 thus we have proved that @xmath174 is mean attracted to @xmath298 where @xmath299 means the forward orbit .
so we have proved that this flow is * mma*. but it is equicontinuous on each minimal subset , so it is also * mmls*. [ r1b ] as one can see in the proof of theorem [ main1 ] , we do nt really need that the spectrum of the sequence @xmath261 is empty .
what we really need is that the spectrum of @xmath39 as a subset in the circle @xmath300 is disjoint from the support of @xmath257 in the circle @xmath300 .
in other words , the oscillating condition on @xmath39 in theorem [ main1 ] can be relaxed to that @xmath301 as a direct consequence of the proof of theorem [ main1 ] and remark [ r1b ] , we have that [ m2 ] suppose @xmath302 is the unit circle .
let @xmath303 be the rigid rotation where @xmath304 is irrational .
any sequence @xmath261 satisfying the growth condition ( [ growth ] ) such that @xmath305 is linearly disjoint from the flow @xmath306 .
we use the notation in the proof of theorem [ main1 ] .
the rotation @xmath307 admits the lebesgue measure as the unique invariant measure and the support of @xmath257 is contained in @xmath308 .
actually it is easy to check that @xmath309 [ r1 ] the idea of using the spectral lemma in the proof of theorem [ main1 ] comes from @xcite where the ergodic hilbert transforms associated to irrational rotations are studied
. we will apply theorem [ main1 ] to several different flows . in section 5 , we apply to all @xmath0-adic polynomial flows and @xmath0-adic rational flows . in section 6 , we apply to all automorphisms and some affine maps of @xmath1-torus with zero topological entropy . in section 7
, we apply to all feigenbaum zero topological entropy flows . in section 8 , we apply to all orientation - preserving irrational circle homeomorphisms .
we give here some examples of equicontinuous flows in the fields of @xmath0-adic numbers .
these flows share a very nice minimal decomposition ( @xcite , @xcite ) .
we can compare this decomposition with auslander s star closed decomposition @xcite .
let @xmath310 be a prime number and let @xmath311 be the ring of @xmath0-adic integers .
let @xmath312 $ ] be the ring of all polynomials with coefficients in @xmath311 .
then every polynomial @xmath313 $ ] defines a flow @xmath314 .
it is trivial that @xmath315 is @xmath316-lipschitz function , i.e. @xmath317 where @xmath318 is the @xmath0-adic norm on @xmath311 .
so , the flow @xmath314 is equicontinuous . in particular , the adding machine @xmath319 $ ] is a special case and it is actually an equicontinuous homeomorphism .
corollary [ m2 ] gives us the following result .
[ pp ] any oscillating sequence @xmath320 satisfying the growth condition ( [ growth ] ) is linearly disjoint from all @xmath0-adic polynomial flows @xmath314 for @xmath313 $ ] . it was proved in @xcite that for a polynomial flow @xmath314 as in corollary [ pp ] , we have the minimal decomposition @xmath321 where @xmath322 is the finite set consisting of all periodic points of @xmath315 , @xmath323 is the union of all ( at most countably many ) clopen invariant sets such that each @xmath324 is a finite union of balls and each subsystem @xmath325 is minimal , and points in @xmath326 lie in the attracting basin of a periodic orbit or of a minimal sub - flow . as we proved in corollary [ m2 ] , points in @xmath326 are mean attracted to a periodic cycle or a minimal set @xmath327 .
more generally , we can consider @xmath0-adic rational flow in @xmath0-adic field .
let @xmath328 be the field of @xmath0-adic numbers .
any point in the projective line @xmath329 may be given in homogeneous coordinates by a pair @xmath330 $ ] of points in @xmath331 which are not both zero .
two such pairs @xmath330 $ ] and @xmath332 $ ] with nonzero factor @xmath333 are identified .
the field @xmath331 may be identified with the subset of @xmath329 given by @xmath334 \in \mathbb p^1(\qp ) \mid x \in \qp\right\}.\ ] ] this subset covers all points in @xmath329 except one : the point of infinity , which may be given as @xmath335.$ ] the spherical metric defined on @xmath329 is analogous to the standard spherical metric on the riemann sphere . if @xmath336 $ ] and @xmath337 $ ] are two points in @xmath329 , we define @xmath338 or , viewing @xmath339 as @xmath340 , for @xmath341 we define @xmath342 and @xmath343 remark that the restriction of the spherical metric on the ring @xmath344 of @xmath0-adic integers is same to the metric induced by the absolute value @xmath318 .
a rational map @xmath345 induces a map on @xmath329 , which we still denote as @xmath346 .
rational maps are always lipschitz continuous on @xmath329 with respect to the spherical metric ( ( * ? ? ?
* theorem 2.14 ) ) .
rational maps with good reduction are @xmath316-lipschitz continuous ( @xcite ) in the sense that @xmath347 [ rp ] any oscillating sequence @xmath261 satisfying the growth condition ( [ growth ] ) is linearly disjoint from all @xmath0-adic rational flows @xmath348 , where @xmath349 is a rational map with good reduction .
it was proved in @xcite that for a rational function with good reduction of order at least @xmath1 , we have a minimal decomposition , @xmath350 which is similar to the case of @xmath0-adic polynomial flow .
one special case of corollary [ rp ] is that [ psarnak ] the mbius function is linearly disjoint from all @xmath0-adic polynomial flows @xmath351 and all @xmath0-adic rational flow with good reduction @xmath352 .
let @xmath353 be the @xmath1-torus .
let @xmath354 be the space of all matrices @xmath355 of integral entries such that @xmath356 .
then all @xmath357 of @xmath212 for @xmath358 represent all automorphisms of @xmath359 .
affine maps on @xmath359 are of the form @xmath360 where @xmath355 is an automorphism of @xmath212 and @xmath361 is an element in @xmath212 . considered @xmath116 as a flow on @xmath212 , the entropy is equal to the logarithm of the sum of the modulus of all eigenvalues of @xmath355 of modulus strictly larger than @xmath316 .
thus @xmath116 is of zero topological entropy if and only if all eigenvalues of @xmath355 are of modulus @xmath316 .
let @xmath362 be the euclidean norm on @xmath363 .
recall that the induced metric on @xmath212 is defined by @xmath364 also recall that the hermitian norm on @xmath365 on @xmath366 is defined by @xmath367 the space @xmath363 is considered as a subspace of @xmath363 .
it is clear that for @xmath368 , we have @xmath369 the following proposition actually hods in higher dimension , but we just state in the dimension @xmath1 .
[ diag ] let @xmath370 .
suppose that @xmath355 is diagonalizable in the field of complex numbers and have all its eigenvalues of modulus @xmath316 .
then there exists a constant @xmath150 such that for all @xmath371 and all @xmath372 we have @xmath373 thus the flow @xmath374 for every @xmath375 is equicontinuous .
let @xmath376 where @xmath377 is the diagonal matrix with the eigenvalues @xmath378 of @xmath355 on the diagonal and @xmath315 is an invertible complex matrix .
then @xmath379 where @xmath380 .
thus @xmath382 let @xmath383 such that @xmath384 .
replace @xmath385 by @xmath386 in the above inequality , we get @xmath387 since @xmath388 for all @xmath389 , finally we get for @xmath390 , @xmath391 combining this proposition and corollary [ m2 ] , we have that [ diagcoro ] suppose @xmath358 is diagonalizable in complex field . and
all its eigenvalues have modulus @xmath316 .
then any oscillating sequence @xmath261 satisfying the growth condition ( [ growth ] ) is linearly disjoint from the flow @xmath392 for every @xmath375 .
now suppose @xmath355 is not diagonalizable . in this case
, @xmath355 has a double eigenvalue @xmath316 or @xmath202 and @xmath393 .
a @xmath394-square matrix is called a _ module matrix _ if all entries are integers and its determinant is @xmath316 .
[ nondiag ] let @xmath395 be a modular matrix with @xmath396 as a double eigenvalue .
there exists a modular matrix @xmath315 such that @xmath397 where @xmath216 is a modular matrix of the form @xmath398 if the double eigenvalue of @xmath395 is @xmath202 .
then @xmath399 has a double eigenvalue @xmath316 .
so we only need to prove the case that the double eigenvalue of @xmath395 is @xmath316 . in this case
the trace of @xmath395 is @xmath1 .
so we can write @xmath400 the fact @xmath401 means @xmath402
if @xmath403 , we must have @xmath404 by ( [ det ] ) .
so @xmath405 thus we are done .
if @xmath406 , we must have @xmath404 and there is nothing to do .
assume @xmath407 and @xmath408 . by ( [ det ] ) , we have @xmath409 and @xmath410 .
we are looking for @xmath411 which is an eigenvector associated to @xmath316 , i.e. @xmath412 .
this equation is equivalent to @xmath413 let @xmath226 be the gcd of @xmath414 and @xmath415 .
the above two equations become one equation @xmath416 we choose an integral solution to the first equation of ( [ ee1 ] ) : @xmath417 then we are looking for a generalized eigenvector @xmath418 such that @xmath419 with @xmath74 an integer to be determined . the first line of the last equation is equivalent to @xmath420
we claim that @xmath421 .
in fact , @xmath422 .
let us factorize @xmath423 into primes : @xmath424 then we can write @xmath425 so @xmath426 where @xmath427 .
then @xmath421 is equivalent to say that @xmath428 for all @xmath429 , which is true because @xmath430 .
take @xmath431 .
then the equation ( [ ee2 ] ) becomes @xmath432 by the bzout theorem , the equation ( [ ee3 ] ) admits a solution @xmath433 in @xmath434 because @xmath435 .
let @xmath436 be the @xmath394 matrix with columns @xmath385 and @xmath437 .
let @xmath438 and @xmath439 be the two rows of @xmath315 .
the first equation of ( [ ee1 ] ) and ( [ ee3 ] ) can be rewritten as @xmath440 then @xmath441 where the relation ( [ ee4 ] ) is used for the third equality .
this implies that @xmath315 ( as well as @xmath442 ) is a positive modular matrix and @xmath443 for some @xmath444 .
following proposition [ diag ] , proposition [ nondiag ] , example [ 2ex ] and theorem [ main1 ] , we have that [ nondiagthm ] suppose @xmath358 and suppose all its eigenvalues have modulus @xmath316 . then the flow @xmath445 is * mma * and * mmls*. a consequence of theorem [ main1 ] and theorem [ nondiagthm ] is the corollary .
[ nondiagcoro ] suppose @xmath358 and suppose all its eigenvalues have modulus @xmath316 .
any oscillating sequence @xmath261 satisfying the growth condition ( [ growth ] ) is linearly disjoint from the flow @xmath446 .
the disjointness of the mbius function from the flow in corollary [ nondiagcoro ] was proved by liu and sarnak in @xcite .
let us make a remark about the number @xmath447 in proposition [ nondiag ] .
it may not be unique but can not also arbitrary .
suppose we have another @xmath448 such that proposition [ nondiag ] holds .
then we have a modular matrix @xmath315 such that @xmath449 let @xmath450 be the first row of @xmath315 and @xmath451 be the second row of @xmath315 .
then we have that @xmath452 this implies that @xmath453 if and only if @xmath454 . if @xmath455 and @xmath456 , then @xmath457 and @xmath458 . since @xmath459
, this implies that @xmath460 .
that is @xmath461 must be the square of an integer .
this implies that @xmath216 and @xmath462 are ( linearly ) conjugate dynamical systems on @xmath359 if and only if @xmath461 is the square of an integer .
for example , @xmath463 are not conjugate flows on @xmath359 .
furthermore , let us give an example to which proposition [ nondiag ] applies : consider the modular matrix @xmath464 then we have @xmath465 when @xmath355 is not diagonalizable , we show in the following example that there is an oscillating sequence and an affine map @xmath466 on @xmath1-tours such that this oscillating sequence is not linearly disjoint from the flow defined by this affine map .
let @xmath467 where @xmath468 is an irrational number .
define @xmath469 the sequence @xmath470 is an oscillating sequence ( refer to example 3 ) satisfying the growth condition ( [ growth ] ) .
but it is not linearly disjoint from the flow @xmath471 ( whose topological entropy is zero ) .
let @xmath472 .
observe that @xmath473 if we take the continuous function @xmath474 , then we have @xmath475 which does not tend to zero . as a skew product of an irrational rotation ,
the map @xmath476 is uniquely ergodic with the lebesque measure as the invariant measure having full support .
this implies that @xmath476 is minimal , so * mma*. but it is not * mmls * ( equivalently * mls * ) .
otherwise , theorem [ main1 ] would imply that the sequence @xmath39 defined by @xmath477 is linearly disjoint from the flow @xmath478 .
[ higher ] in @xcite , liu and sarnak has proved that the mbius function @xmath479 is linearly disjoint from all affine flows @xmath480 whose topological entropy is zero , where @xmath358 and @xmath375 ( the result holds in high dimensions ) .
this result and the above example indicate that in order for an oscillating sequence @xmath261 to be linearly disjoint from all zero entropy flows , it may not be oscillating only in the first order but also oscillating in any higher order , that is , @xmath481 for all real polynomials @xmath315 .
the mbius function and the von mangoldt function , which is defined as @xmath482 if @xmath483 for some prime number and integer @xmath34 and @xmath203 otherwise , share the above oscillating property for any high order as shown in @xcite .
a famous zero topological entropy flow has been studied extensively for last fourty years is the so - called feigenbaum quadratic - like map .
it is a non - linear dynamical system and does not have the mean - equicontinuous property on the whole space . in this section , we will prove that a feigenbaum zero topological entropy flow is minimally mean attractable ( * mma * ) and minimally mean - l - stable ( * mmls * ) .
thus sarnak s conjecture is true for such a dynamical system .
we use @xmath484 to denote the topological entropy of a dynamical system @xmath5 . for the definition of the topological entropy
, the reader can refer to @xcite .
let @xmath485 $ ] .
suppose @xmath486 it is a family of quadratic polynomials with parameter @xmath487 .
then @xmath202 is a fixed point of @xmath216 and @xmath216 maps @xmath316 to @xmath202 .
the map @xmath216 has a unique critical point @xmath203 and it is non - degenerate , that is , @xmath488 and @xmath489 . when @xmath490 , @xmath491 has @xmath202 as a parabolic fixed point , that is , @xmath492 ( see figure 1 ) . and @xmath491 has no other fixed point .
one can see easily that @xmath493 so @xmath494 and in this case , @xmath495 is the only minimal set and @xmath496 $ ] .
thus the flow @xmath497 , t_{t_{0}})$ ] is * mma * and * mmls*. [ fig1 ] when @xmath498 but close to @xmath499 , @xmath202 becomes a repelling fixed point of @xmath216 , that is , @xmath500 , but @xmath216 has another fixed point @xmath501 which is an attracting fixed point , that is , @xmath502 and @xmath503 ( see figure 2 ) .
furthermore , we have that @xmath504 [ fig2 ] this phenomenon keeps until @xmath501 becomes a parabolic fixed point at @xmath505 , that is , @xmath506 .
so , for @xmath507 , we have that @xmath508 and @xmath216 has two minimal subsets @xmath495 and @xmath509 and @xmath510 and @xmath511 .
this proves that the flow @xmath497 , t_{t})$ ] is * mma * and * mmls*. when @xmath512 but close to @xmath505 , @xmath501 becomes a repelling fixed point of @xmath216 , that is , @xmath502 and @xmath513 , but there is a periodic cycle of period @xmath1 appearing near @xmath501 , which we denote as @xmath514 .
this cycle is attractive , that is , @xmath515 and @xmath516 ( see figure 2 ) .
moreover , @xmath517 this phenomenon keeps until the cycle @xmath514 becomes a parabolic periodic cycle of period @xmath1 at @xmath518 , that is , @xmath519 so , for @xmath520 , we have that @xmath508 and @xmath216 has three minimal subsets @xmath495 and @xmath521 and @xmath522 with @xmath523 and @xmath524 we have proved that the flow @xmath497 , t_{t})$ ] is * mma * and * mmls * for @xmath525 .
[ fig3 ] to see how the periodic cycle of period @xmath1 was born , we can use the period doubling operator @xmath526 as follows .
when @xmath527 , let @xmath528 .
consider the map @xmath529 it is not a quadratic polynomial but has a graph similar to @xmath216 for @xmath530 .
so when @xmath531 , @xmath532 has a parabolic fixed point @xmath202 and when @xmath533 , @xmath532 has @xmath202 as a repelling fixed point and there is an attractive fixed point @xmath534 .
when @xmath535 , @xmath532 has @xmath202 as a repelling fixed point and @xmath534 becomes a parabolic fixed point .
note that @xmath536 become a cycle of periodic points of period @xmath1 for @xmath216 . in general ,
repeat the above process , we have the following dichotomy .
there is an increasing sequence @xmath537 for any @xmath538 , @xmath216 has a repelling periodic cycle @xmath539 of period @xmath540 for any @xmath541 and an attractive ( or parabolic but semi - attractive ) periodic cycle @xmath542 such that @xmath543 where @xmath544 is a countable subset of @xmath545 .
so we have that @xmath508 for @xmath546 , where @xmath547 in the case @xmath548 , @xmath216 has @xmath549 minimal subset sets , @xmath550 , @xmath551 , and @xmath552 with @xmath510 and @xmath553 and @xmath554\setminus \cup_{i=0}^{\infty } \hbox{\rm basin}(k_{i}).\ ] ] thus we have proved that the flow @xmath497 , t_{t})$ ] is * mma * and * mmls*. concluding from the above , we have that [ easyone ] for any @xmath555 , the flow @xmath556 is * mma * and * mmls*. [ prefeig ] any oscillating sequence @xmath261 satisfying the growth condition ( [ growth ] ) is linearly disjoint from @xmath557 for any @xmath555 a special case in corollary [ prefeig ] is the mbius function .
the quadratic polynomial @xmath558 is called the _
feigenbaum quadratic polynomial_. this polynomial has been studied for the past 40 years since it was discovered by feigenbaum @xcite and , independently , by coullet and tresser @xcite , in 1970s .
the following theorem can be found in the literature @xcite .
[ fei ] the quadratic polynomial @xmath558 on @xmath545 as a dynamical system has two repelling fixed points @xmath202 and @xmath559 and a repelling periodic cycle @xmath560 of period @xmath561 for every @xmath389 .
there is a cantor set @xmath562 such that @xmath563 where @xmath564 is a countable subset of @xmath545 .
moreover , the topological entropy @xmath565 in the rest of this section , we will give a rigorous proof of that @xmath558 on @xmath566 is * mls * ( that is , mean - equicontinuous ) .
actually , we will prove all feigenbaum zero topological entropy dynamical systems are * mls * ( that is , mean - equicontinuous ) on their corresponding set .
our strategy in the proof is as follows .
first we will give a rigorous proof of that the restriction of a feigenbaum zero topological entropy flow to its cantor set attractor is topologically conjugate to the adding machine on the @xmath1-adic symbolic flow ( refer to section 5 ) .
this fact was first observed by feigenbaum in @xcite when he proposed his famous universality theory in chaos ( and , independently , by coullet and tresser @xcite ) .
then it was emphasized by sullivan in his many lectures attended by the second author in 1986 - 1989 and in our paper and book @xcite .
our rigorous proof is based on two facts .
the first one is the existence of the fixed point of the periodic doubling operator which was hot topic during the past 40 years in the research of complex dynamics and chaos .
the existence of the fixed point was first observed by feigenbaum in @xcite and used to explain his university theory in chaos .
then was proved by lanford in @xcite with computer - assistant ( but still a legitimate mathematical proof ) .
furthermore , a conceptual understanding and more rigorous mathematical proof by using teichmller theory and following idea of moscow rigidity theorem was given by very recent work of many mathematicians ( refer to @xcite .
the other is the connection between cantor set attractor of the feigenbaum fixed point and the non - escaping set of an induced expanding flow as studied in our previous paper @xcite . to state our main result
, we first define the space of all feigenbaum quadratic - like maps .
suppose @xmath567 is a @xmath568-map .
we call it a _ quadratic - like map _ if 1
. @xmath569 ; 2 .
@xmath205 and @xmath570 ; 3 .
@xmath571 for @xmath572 and @xmath573 for @xmath574 $ ] ; 4 .
the schwarzian derivative @xmath575 let @xmath576 be the space of all quadratic - like maps .
then the topological entropy @xmath577 defines a functional . using the theory of kneading sequences of milnor - thurston @xcite
, the space @xmath576 is divided into two parts , @xmath578 and @xmath579 the boundary of both spaces @xmath580 is called the _ space of feigenbaum quadratic - like maps_. the space of feigenbaum quadratic - like maps can be defined much more general . for examples
, we can assume that the type of critical point is @xmath581 for some @xmath582 and that the smoothness is @xmath583 where @xmath584 means the zygmund modulus of continuity .
see @xcite for a more general definition of feigenbaum folding maps . to avoid to lost the reader from our main purpose , we use an easy definition of a feigenbaum map .
what we will prove in theorem [ main2 ] works for any feigenbaum folding map but the argument will be more complicate
. we would like to note that the space @xmath585 is all quadratic - like maps with the same feigenbaum kneading sequence which is the kneading sequence of @xmath558 .
moreover , we have the following result from the literature @xcite .
[ fei1 ] any @xmath586 has two fixed points @xmath202 and @xmath587 and a repelling periodic cycle @xmath588 of period @xmath561 for every @xmath389 . there is a cantor set @xmath589 such that @xmath590 where @xmath591 is a countable subset of @xmath545 .
furthermore , there is a homeomorphism @xmath592 of @xmath545 such that @xmath593 thus the topological entropy @xmath594 we call @xmath595 for @xmath586 a feigenbaum zero topological entropy flow . [ main2 ] any feigenbaum zero topological entropy flow @xmath6 is * mma * and * mmls*. we first define the period doubling operator on @xmath585 as follows .
any @xmath586 has two fixed points @xmath596 and @xmath597 .
define @xmath598 then @xmath599 .
thus @xmath600 defines an operator which is called the period doubling operator . as we have remarked in the paragraph after theorem [ fei ] , the period doubling operator has a unique fixed point @xmath601 , that is , @xmath602 with the existence of this fixed point , we continue our proof as follows .
consider the critical value @xmath603 and @xmath604 , @xmath605 , @xmath606 .
then we have that @xmath607 let @xmath608 $ ] and let @xmath609 $ ] and @xmath610 $ ] .
[ fig4 ] we now define a @xmath568 expanding map @xmath611 .
let @xmath597 be the fixed point of @xmath612 .
define @xmath613 it has been shown in @xcite that @xmath611 is expanding , that is , @xmath614 .
therefore , the non - escaping set @xmath615 is a cantor set ( see ( * ? ? ?
* chapter 1 ) for the detailed proof of this fact ) .
let @xmath616 be the set of all post - critical points for @xmath612 .
then the attracting cantor set for @xmath612 satisfies that @xmath617 since @xmath618 , we can see that endpoints of all intervals in @xmath619 are contained in @xmath620 .
this implies that @xmath621 since @xmath611 is an expanding map of degree @xmath1 , we can conjugate @xmath622 to the shift map @xmath256 of the symbolic space ( the @xmath1-adic space in section 5 ) @xmath623 as follows .
consider the initial partition @xmath624 we can pull back it to get the @xmath625-partition @xmath626 where @xmath627 .
we label intervals in @xmath628 inductively as that @xmath629 where @xmath630 . from this labelling
, we have that @xmath631 thus for any @xmath632 , we have that the set @xmath633 contains one point . set @xmath634 .
suppose the cantor set @xmath635 is a metric space with the metric induced from the lebesgue metric on @xmath545 and suppose @xmath636 is a metric space with the metric defined as @xmath637 if @xmath638 and @xmath639 and if @xmath640 .
then we have that @xmath641 is a homeomorphism such that @xmath642 now we define the add machining @xmath643 on @xmath636 as follows . for any @xmath644 , if @xmath645 , then @xmath646 and if @xmath647 , then the first digit of @xmath648 is @xmath203 and consider @xmath649 , etc .
from the construction of @xmath650 , we have that @xmath651 this implies that @xmath652 and , furthermore , @xmath653 for @xmath654 . since the adding machine @xmath655 is isometric then equicontinuous and @xmath656 is a conjugacy , we have that @xmath657 is equicontinuous .
now for any @xmath586 , we have a homeomorphism @xmath658 such that @xmath659 both @xmath660 and @xmath661 are uniformly continuous .
so we have that @xmath662 is equicontinuous .
now what we have proved is that @xmath116 has countably many minimal subsets @xmath495 , @xmath663 , @xmath664 for all @xmath389 , and @xmath665 . since @xmath666 or @xmath667 contains only one fixed point , it is easy to see that @xmath668 similarly , since @xmath669 contains only one periodic cycle of period @xmath561 for @xmath389 , @xmath670 now we claim that @xmath671\setminus \cup_{n=0}^{\infty } \hbox{\rm basin } ( k_{n}).\ ] ] to prove this claim , we consider @xmath612 and the induced expanding map @xmath126 .
let @xmath628 be the @xmath625-partition in ( [ part ] ) .
since @xmath126 is expanding , we have a constant @xmath672 such that @xmath673 the map @xmath612 permutes intervals in @xmath628 ( see fig .
4 ) . for any @xmath674\setminus \cup_{n=0}^{\infty } \hbox{\rm basin}_{g } ( k_{n}),\ ] ] we know that @xmath675 as @xmath676 .
note that @xmath677 therefore , for any @xmath28 , we have an integer @xmath678 such that @xmath679 .
we also have an integer @xmath680 such that @xmath681 is in some interval @xmath682 .
take a point @xmath683 .
then we have that @xmath684 since @xmath685 is a homeomorphism , we have a point @xmath686 for @xmath687 such that @xmath688 ( refer to fig .
this implies that @xmath689 since @xmath612 permutes intervals in @xmath690 , for any @xmath691 , we have that @xmath692 and @xmath693 are in the same interval in @xmath694 , this further implies that @xmath695 thus we have that @xmath696 this implies that @xmath697\setminus \cup_{n=0}^{\infty } \hbox{\rm basin}_{g } ( k_{n}).\ ] ] we proved the claim for @xmath612 . since @xmath698 and since both @xmath592 and @xmath699 are uniformly continuous on @xmath545 , we proved also the claim for @xmath116 .
now we knew that @xmath700 is minimal and equicontinuous .
for each @xmath701 , the map @xmath702 is minimal and equicontinuous because it is just a periodic cycle .
thus we finally proved that @xmath703 is * mma * and * mmls*. one consequence of theorem [ main1 ] and theorem [ main2 ] is the following corollary with the mbius function as a special example .
[ scfeig ] any oscillating sequence is linearly disjoint from all feigenbaum zero topological entropy flows . after this paper
was completed , we learnt that karagulyan @xcite has proved that the mbius function is linearly disjoint from all continuous interval maps with zero topological entropy . it can be proved that any oscillating sequence is linear disjoint from all continuous interval maps with zero topological entropy and the special property of mbius function is not needed in karagulyan s proof , but only the oscillating property .
we expect that all continuous interval maps with zero topological entropy are * mma * and * mmls*.
in this section , we discuss flows of zero topological entropy on the unit circle which are * mma * and * mmls * but not equicontinuous on its minimal set .
* @xmath707 and for every @xmath708 , @xmath709 as @xmath676 , where @xmath126 is the fixed point set of @xmath710 , * @xmath711 and @xmath116 is conjugate to a rigid rotation @xmath712 for some @xmath713 , or * @xmath714 and @xmath710 is the identity .
the above three cases are simple and theorem [ main1 ] applies .
moreover , the flow defined by an orientation - preserving circle homeomorphism is equicontinuous if and only if it is conjugate to a rigid rotation .
therefore , we know that all oscillating sequences are linearly disjoint from all orientation - preserving rational circle homeomorphisms and all orientation - preserving irrational circle homeomorphisms which are not denjoy counter - example .
we are now concerning that if theorem [ main1 ] applies to all denjoy counter - examples of orientation - preserving irrational circle homeomorphisms .
a denjoy counter - example is an orientation - preserving circle homeomorphism which is only semi - conjugate to a rigid rotation @xmath715 for an irrational rotation number @xmath716 .
that is , we have a continuous onto ( but not 1 - 1 ) map @xmath717 such that @xmath718 on @xmath300 .
thus @xmath706 is not equicontinuous according to @xcite .
let us first state the poincar theorem on orientation - preserving circle homeomorphisms with irrational rotation numbers .
[ pt ] suppose @xmath705 is an orientation - preserving circle homeomorphism with irrational rotation number @xmath117 . then there is a continuous onto map @xmath717 satisfying ( [ conj ] ) .
moreover , the @xmath133-limit set @xmath719 for a point @xmath720 in the circle is either a cantor set @xmath562 , which is independent of @xmath720 , or the whole circle @xmath721 . in the case
@xmath722 , @xmath592 is a homeomorphism . in the case
@xmath723 , a cantor set , @xmath724 is the union of all wandering intervals . here
@xmath562 is also called the non - wandering set for @xmath116 .
denjoy constucted an example @xmath116 of an orientation - preserving circle homeomorphism with irrational rotation number such that for any @xmath725 , @xmath723 is a cantor set . in this case
, @xmath726 is minimal .
we call such a map a denjoy counter - example .
[ ceq ] suppose @xmath729 is a denjoy counter - example .
the flow @xmath730 is * mma * but the flow @xmath731 is not equicontinuous .
however , the flow @xmath731 is * mls * , thus , the flow @xmath730 is * mma * and * mmls*. for any @xmath732 close to each other , let @xmath733 $ ] be the shortest arc with endpoints @xmath734 $ ] in the circle .
the arc length of @xmath733 $ ] is @xmath735 .
assume the total length of @xmath721 is @xmath316 .
for any @xmath738 , it is in a wandering interval @xmath739 , where @xmath740
. let @xmath741 , @xmath742 .
then we have that @xmath743 thus @xmath744 as @xmath676 .
since @xmath745 and @xmath746 , we have that @xmath747 of all wandering intervals , which is countable .
we partition it into full orbits @xmath748 where @xmath129 is the number of full orbits of wandering intervals . here
@xmath129 is either a positive integer or @xmath749 .
since @xmath562 is a cantor set , we have @xmath750 let @xmath592 be the semi - conjugacy between @xmath116 and the rigid rotation @xmath751
. see ( [ conj ] ) .
then the image @xmath752 is a singleton @xmath753 and we write @xmath754 . by ( [ conj ] )
, @xmath755 is the full orbit of @xmath756 under iterations of @xmath751 for any fixed @xmath757 .
also notice that @xmath592 is monotone increasing .
so , if @xmath174 and @xmath215 are two points in the cantor set @xmath562 we have @xmath758 \quad\mbox{\rm iff}\quad x_{m , n } \in [ h(x ) , h(y)].\ ] ] the rigid rotation @xmath751 is uniquely ergodic with the lebesgue measure as the unique invariant measure .
this fact , together with ( [ hh1 ] ) , implies that for any @xmath759 , @xmath760 , \ ; -n \leq k\leq 0\})}{n } = |z - w|.\ ] ] this allows us to prove that @xmath728 is not equicontinuous .
in fact , given any @xmath761 , by ( [ hh2 ] ) and ( [ hh3 ] ) we have that @xmath762 , \ ; -n \leq k\leq 0\})}{n } \\ & = & \lim_{n\to \infty } \frac{\#(\ { k\;|\ ; x_{0,k}\in [ h(x ) , h(y ) ] , \ ;
-n \leq k\leq 0\})}{n } = |h(x)-h(y)| > 0.\end{aligned}\ ] ] thus there are infinitely many negative integers @xmath763 such that @xmath764 $ ] . for such @xmath763 s
we have @xmath765\ ] ] so that @xmath766 for infinitely many positive integers @xmath767 . therefore , @xmath768 is not equicontinuous .
next we prove that @xmath768 is * mls*. we first assume that the non - wandering set @xmath562 has zero lebesgue measure . under this assumption , we have that @xmath769 moreover , for any two points @xmath761 , @xmath770 , consider the rectangle @xmath771 where @xmath772 . for any @xmath28 , we have a large integer @xmath773 such that @xmath774 since @xmath592 is uniformly continuous , we can find a @xmath775 such that for any @xmath761 with @xmath776 , @xmath777 let @xmath778 where @xmath779\}.\ ] ] for any fixed @xmath780 , we have @xmath781)}{n } & = & \lim_{n\to \infty } \frac{\#(\ { k\;|\ ; i_{m , k+l}\subset [ x , y ] , \ ; -n \leq k\leq 0\})}{n } \\ & = & \lim_{n\to \infty } \frac{\#(\ { k\;|\ ; x_{m , k+l}\in [ h(x ) , h(y ) ] , \ ; -n \leq k\leq 0\})}{n } \\ & = & |h(x)-h(y)| > 0.\end{aligned}\ ] ] therefore the density of @xmath126 exists and is equal to @xmath782)}{n } = ( n_{0}\wedge(l-1)+1 ) ( 2n_{0}+1 ) |h(x)-h(y)| < \epsilon\ ] ] for any @xmath783 , the interval @xmath784 $ ] contains only wandering intervals from @xmath785 from ( [ zeq ] ) , we have that @xmath786 this implies that @xmath728 is * mls*. thus it is mean - equicontinuous .
since the flow @xmath706 has only one mean minimal attractor @xmath562 , so it is * mmls*. when the non - wandering set @xmath562 has a positive lebesgue measure , there is another denjoy counter - example @xmath787 whose non - wandering set @xmath788 has zero lebesgue measure such that both of them are semi - conjugate to the same rigidity rotation @xmath751 .
that is , we have two continuous onto maps @xmath789 such that @xmath790 now we can define @xmath791 which formally can be written as @xmath792 .
it maps every wandering interval to the corresponding wandering interval bijectively and preserves orders of wandering intervals . and
it maps @xmath562 to @xmath788 bijectively .
thus it is a homeomorphism of the circle such that @xmath793 since @xmath787 is * mmls * and since @xmath116 is topologically conjugate to @xmath787 , we get that @xmath116 is * mmls*. j. bourgain , p. sarnak , and ziegler , _ disjointness of mbius from horocycle flows_. from fourier analysis and number theory to radon transforms and geometry , 67 - 83 , dev .
28 , springer , new york , 2013 .
p. coullet and c. tresser , _ itration dndomorphismes et groupe de renormalisation_. c. r. acad . sci .
paris ser . , a - b * 287 * , 1978 , pp
. a577-a580 .
( j. phys . coll .
39:c5 ( 1978 ) , 25 - 28 ; supplment au 39:8 ) .
d. sullivan , _ bounds , quadratic differentials , and renormalization conjectures_. american mathematical society centennial publications , volume 2 : mathematics into the twenty - first century , ams , providence , ri , 1992 , pp .
417 - 466 . | we define oscillating sequences which include the mbius function in the number theory .
we also define minimally mean attractable flows and minimally mean - l - stable flows .
it is proved that all oscillating sequences are linearly disjoint from minimally mean attractable and minimally mean - l - stable flows . in particular , that is the case for the mbius function .
several minimally mean attractable and minimally mean - l - stable flows are examined .
these flows include the ones defined by all @xmath0-adic polynomials , all @xmath0-adic rational maps with good reduction , all automorphisms of @xmath1-torus with zero topological entropy , all diagonalized affine maps of @xmath1-torus with zero topological entropy , all feigenbaum zero topological entropy flows , and all orientation - preserving circle homeomorphisms . . the second author is partially supported by the collaboration grant from the simons foundation [ grant number 199837 ] and the cuny collaborative incentive research grants [ grant number 2013 ] and awards from psc - cuny and a grant from nsfc [ grant number 11171121 ] . |
we consider a bose - condensed atomic cloud in a quasi-1d geometry in which the transverse motion is frozen by the confining potential of an optical or magnetic atomic waveguide @xcite .
gravity is made immaterial either by placing the waveguide axis along the horizontal plane or by counterbalancing the gravitational field with a suitable magnetic field gradient .
, width=288 ] a periodic potential is created along the waveguide axis by means of a pair of far - off resonance laser beams of frequency @xmath0 and wave vector @xmath1 crossing the waveguide at an angle @xmath2 as in fig.[fig : setup ] @xcite : denoting with the ( slowly varying ) single beam rabi frequency and with @xmath3 the atomic transition frequency , the optical potential experienced by the atoms is given by @xmath4 , with @xmath5 and @xmath6 .
for a red- or blue - detuned laser field , the optical potential is respectively attractive or repulsive ; the lattice period @xmath7 can be tuned by varying the angle @xmath2 .
the longitudinal envelope @xmath8 of the lattice potential is determined by the profile of the laser beam waist and is assumed to smoothly vary on a length scale significantly longer than the lattice period @xmath9 ; unless differently specified , @xmath8 will be taken as a gaussian @xmath10 of height @xmath11 and spatial length @xmath12 .
if both the kinetic and the interaction energy are much smaller than the transverse level spacing of the waveguide , this latter can be considered as being a single - mode one and the condensate wave function can be written in the factorized form @xmath13 where @xmath14 is the ground - state eigenfunction of the transverse confining potential with the appropriate @xmath15 normalization . within this approximation ,
the dynamics of the condensed atomic cloud can be described by a one - dimensional gross - pitaevskii equation @xmath16 where @xmath17 is the atomic mass and the renormalized 1d effective interaction @xmath18 is written in terms of the usual 3d scattering length @xmath19 as @xcite @xmath20
as it happens to electrons in crystalline solids @xcite and light in periodic dielectric systems such as photonic band gap crystals @xcite , the atomic dispersion in an infinite periodic potential is characterized at linear regime ( i.e. in the non - interacting case ) by allowed bands and forbidden gaps .
when the depth @xmath11 of the lattice potential is weak with respect to the bragg energy @xmath21 the lowest part of the band structure can be accurately described within a _ nearly - free atom _
approximation @xcite in which only two coupled modes @xcite are taken into account @xmath22 in this restricted @xmath23 basis , the linear regime hamiltonian has the following form @xmath24 and the eigenenergies @xmath25 are equal to @xmath26 { \label{eq : bands}}\end{gathered}\ ] ] where we have set @xmath27 ; the positive sign holds for the upper , _ conduction band _ and the minus sign for the lower , _ valence band_. no states are present between @xmath28 and @xmath29 : this is the lowest _ forbidden gap _ in which the matter waves can not propagate through the lattice .
the group velocity is given by @xmath30^{-1/2}\right\};\end{gathered}\ ] ] close to band edge ( @xmath31 ) , the group velocity @xmath32 is much reduced with respect to the bragg velocity @xmath33 ; further away ( @xmath34 ) , it recovers the free - space value @xmath35 .
the effective mass of the atom is given by @xmath36^{-3/2}\right\}.\end{gathered}\ ] ] at band edge ( @xmath37 ) , @xmath38 is much smaller in absolute value than the free - space atomic mass @xmath17 ; as usual , it is positive at the conduction band edge , while it is negative at the valence band edge .
further away from the band edge , @xmath38 coincides with the positive free - space value @xmath17 @xcite . around the band edge , the weights of the forward and backward traveling waves are comparable @xmath39 for both valence and conduction bands : the density profile of the bloch eigenfunction thus has a standing wave shape with a spatial period equal to the lattice period @xmath9 .
further away from the gap one component dominates the other : the bloch eigenfunction has then a running wave character and the density profile is uniform over the unit cell of the lattice .
we now consider a coherent matter pulse ( e.g. extracted from a bose - einstein condensate ) which is moving along the waveguide with a uniform velocity @xmath40 close to the bragg velocity @xmath41 . initially the atomic pulse is far outside the lattice .
the initial density distribution of atoms in the cloud is taken as having a gaussian shape @xmath42 the carrier momentum is @xmath43 and the carrier kinetic energy @xmath44 ; since the wave packet is finite in space , its fourier transform @xmath45 has a finite momentum spread @xmath46 around @xmath47 . ) optical lattice with gaussian profile ( @xmath48 ) .
lower panels : linear regime time - evolution of gaussian incident pulses ( @xmath49 ) centered at respectively @xmath50 ( b ) ( transmitted ) and @xmath51 ( c ) ( reflected ) .
the spatial extension of the lattice is indicated by the vertical dashed lines .
[ fig : linear],width=288 ] if the density is sufficiently low and the interactions sufficiently weak , the nonlinear term in the gross - pitaevskii equation ( [ eq : gp ] ) can be neglected and the time - evolution of the matter field is accurately described by a linear schrdinger equation . in this case , the superposition principle holds and the transmission of the pulse can be well described in momentum space in terms of the wave vector - dependent transmission amplitude @xmath52 @xmath53 if the whole of the incident wave packet is contained in either a transmitting or a reflecting region of the spectrum ( fig.[fig : linear]a ) , it will be respectively transmitted ( fig.[fig : linear]b ) or reflected ( fig.[fig : linear]c ) without sensible reshaping . as it has been discussed in full detail in previous papers @xcite
, complete transmission occurs whenever propagating states are available at all spatial positions for the frequencies of interest .
given the smooth envelope of the lattice , interface reflections do not occur and the matter wave adiabatically follows the shape of the corresponding bloch state ; the typical sinusoidal shape of band edge bloch wave functions @xcite multiplying the broad pulse envelope can be clearly recognized while the wave packet is crossing the lattice ( fig.[fig : linear]b , inset ) . on the other hand ,
if the wave packet frequency falls inside the forbidden gap at some spatial positions , the matter wave is not able to cross that region and is then nearly completely reflected back ( fig.[fig : linear]c ) ; the inversion point is located at the beginning of the forbidden region , i.e. at the point at which the carrier frequency @xmath54 enters the forbidden gap . )
gaussian pulse close to resonance with a quasi - bound mode at @xmath55 ( see fig.[fig : linear]a ) ; lattice parameters as in figs.[fig : setup ] and [ fig : linear ] .
the spatial extension of the lattice ( @xmath48 ) is indicated by the vertical dashed lines .
upper panels : ( a ) expanded view of the transmitted pulse shape for the pulse in ( c ) ( solid line ) and for a longer @xmath56 pulse ( dashed line ) ; ( b ) incident ( dashed ) and transmitted ( solid ) spectra for the incident pulse in ( c ) .
[ fig : linearres],width=288 ] when one or more resonance peaks due to quasi - bound modes are contained within the spectrum of the incident wave packet , the pulse will be partially reflected and partially transmitted ( fig.[fig : linearres]b - c ) ; the strong frequency dependence of the transmission amplitude leads to a profound reshaping of the spatial profile of the pulse ( fig.[fig : linearres]a ) .
an incident pulse of linewidth much wider than the quasi - bound mode linewidth has in fact a time duration much shorter than the characteristic decay time of the mode ; this latter can therefore be considered as being impulsively excited by the incident pulse and than exponentially decaying on a much longer time scale . the exponential tail shown by the transmitted pulse profile when a single quasi - bound mode is excited ( dashed line in fig.[fig : linearres]a ) is a clear signature of transmission occurring via a single resonant quasi - bound mode @xcite .
if several isolated modes are instead present in the frequency interval encompassed by the incident spectrum , the spectrum of the transmitted pulse will contain several peaks ( fig.[fig : linearres]c ) and a complex shape will result from the interference of the different frequency components .
for example , if the incident pulse has a duration comparable to the dephasing time of a pair of neighboring modes ( i.e. the inverse of their frequency difference ) , both of them will be impulsively excited and the profile of the transmitted pulse will show oscillations following the time evolution of the relative phase ( solid line in fig.[fig : linearres]a ) ; these oscillations can be interpreted as _ quantum beatings _ between the two coherently excited quasi - bound modes coupled to the same continuum of propagating modes outside the lattice
for higher values of the atomic density and the coupling constant @xmath57 , the effect of the atom - atom interactions is no longer negligible and the mean - field nonlinear term @xmath58 in ( [ eq : gp ] ) starts playing an important role in the dynamics ; in the following , we shall focus on the experimentally most relevant case of a positive scattering length @xmath59 , which means a repulsive interaction among atoms in free space . in @xcite we discussed the case of a continuous wave ( cw ) beam of coherent atoms incident on a finite lattice ; depending on the detuning of the incident beam with respect to the frequency of a quasi - bound mode of the lattice , atom optical limiting or bistability
was predicted for an incident beam respectively on or above the resonance frequency .
here we shall consider the more realistic case of a finite atomic wave packet incident on a finite optical lattice ; its central frequency @xmath54 is taken to be close to a quasi - bound mode of the lattice and its temporal duration much longer than the lifetime of the mode so that the frequency spread is narrower than the mode linewidth . above
( b ) resonance with the quasi - bound mode of the lattice at @xmath60 ( solid lines ) . for comparison ,
same calculations neglecting the interaction term ( dashed lines ) and in the absence of the lattice ( dot - dashed lines ) .
same lattice parameters as in figs.[fig : linear ] and [ fig : linearres ] ; the pulse starts at @xmath61 with a gaussian width equal to @xmath62 .
[ fig : nonlinear],width=288 ] ( black lines ) . for comparison ,
same calculation neglecting the interaction term ( grey lines ) .
[ fig : nonlrefl],width=288 ] first we consider the case of an incident wave packet with a center - of - mass kinetic energy @xmath63 exactly on resonance with a quasi - bound mode ( fig.[fig : nonlinear]a ) : in the absence of interactions ( dashed line ) , the wave packet is nearly completely transmitted without any shape distortion ; only a small fraction of the pulse is reflected ( fig.[fig : nonlrefl]a ) .
thanks to the resonance condition with the quasi - bound mode , the atomic density inside the lattice is significantly larger than the one in the incident wave packet . in the presence of interaction , the main effect of nonlinearity is to blue - shift the quasi - bound mode frequency @xcite and push it out of resonance with the incident beam ; this negative feedback effect , already present in the cw treatment @xcite , gives not only a saturation of the transmission as a function of the incident density , but also a significant reshaping of the wave packet : the higher density part is cut and the resulting wave packet is flattened .
the negative feedback in the transmission obviously corresponds to a positive feedback in the reflection . in the opposite case of a wave packet incident with a kinetic energy higher than the quasi - mode frequency ,
a low density wave packet is nearly completely reflected ( fig.[fig : nonlinear]b , dashed line ) . at higher densities ,
the cw calculations in @xcite have predicted the appearance of atom optical bistability effects : several stationary states with different transmitted intensities can indeed be found for the same value of the incident intensity . on such grounds , we are able to put forward a physical interpretation of the transmitted pulse shape in the presence of interactions shown as a solid line in fig.[fig : nonlinear]b : only a small fraction of the leading part of the pulse is transmitted since the incident frequency is out of resonance with the empty quasi - bound mode ; moreover , this part of the incident wave packet has been accelerated by the repulsive mean - field interactions before reaching the lattice and thus pushed even further off resonance .
for the same reason , the trailing part of the pulse has been instead slowed down with respect to the central velocity and thus results closer to resonance to the quasi - bound mode ; when this trailing part of the pulse reaches the lattice , the quasi - bound mode starts to be effectively populated .
the main effect of the interactions among the atoms in the quasi - bound mode is to push its frequency closer to resonance with the incident wave packet : the sudden increase in the density of the transmitted pulse which is apparent in fig.[fig : nonlinear]b is a direct consequence of this positive feedback .
this behaviour is analog to the jump from the lower to the upper branch of the bistability loop which occur in the cw case for incident intensities growing beyond the switch - on threshold .
the shape of the reflected pulse is complementary to the one of the transmitted one : in the presence of the nonlinearity , the reflected pulse shows a dip corresponding to the switch - on of the transmission ( fig.[fig : nonlrefl]b ) . provided
the interaction energy is much smaller than the spacing of the different quasi - bound modes , the transmission dynamics is mostly determined by a single resonant mode and the shape of the matter field inside the lattice is fixed by the eigenfunction of the mode ; this guarantees that no _ modulational instability _ @xcite can take place even in the presence of an effectively attractive interaction such as the one which occurs in the case of negative mass @xmath64 valence band atoms for which the sign of the effective interaction is reversed with respect to free space . in the absence of spatial confinement
, a spatially extended wave packet of coherent valence band atoms is instead subjected to modulational instability : consider for example an attractive lattice and an incident wave packet with a kinetic energy just below the lower edge of the gap . in the present paper
we shall limit ourselves to the case of a sufficiently weak nonlinearity @xmath65 in order for the band structure of the atomic dispersion not to be washed out by the interaction term . since propagating valence band states are available at all spatial positions ,
the pulse is able to penetrate inside the lattice without any reflection .
as soon as the pulse is in the lattice , the modulational instability sets in and the initially uniform envelope starts being modulated with an amplitude which grows exponentially in time ; at the end , a train of short and intense pulses is found ( fig.[fig : modinstab ] ) .
the seed for this modulational instability is automatically provided by the density modulations which inevitably arise while the pulse is entering the non - uniform lattice . in order to make the discussion the simplest ,
this has been assumed to have an uniform depth in its central part and wings at its ends ; after the initial preparation phase , the pulses therefore propagate through an effectively uniform system .
, @xmath66 , @xmath67 ) across an attractive lattice ( @xmath68 ) .
the lattice has a flat profile in the central @xmath69 region and gaussian ( @xmath70 ) wings ; its spatial extension is indicated by the vertical dashed lines .
[ fig : modinstab],width=288 ] a similar modulational instability is well - known to occur in nonlinear bragg fiber optics @xcite ; a cw laser beam traveling in the conduction ( valence ) photonic band of a bragg lattice with a focusing ( defocusing ) nonlinear refractive index is subjected to self - pulsation and therefore converted into a train of pulses .
since the nonlinear refraction index can be interpreted as an effective photon - photon interaction @xcite , the self - pulsing effect is easily interpreted as the modulational instability following the presence of an effectively attractive interaction .
a regular train of short optical pulses with a well defined period and intensity has been theoretically shown to be produced if the modulational instability of a cw laser beam propagating in a lattice is seeded with a weak periodic intensity modulation @xcite .
the shape of each pulse is stable during propagation since the group velocity dispersion is counterbalanced by the optical nonlinearity . following the literature
, we shall call these pulses _ gap solitons _ @xcite , even if from a strictly mathematical point of view they could be classified only as _ solitary waves _ , since the wave equation ( [ eq : gp ] ) in a periodic potential is not exactly integrable @xcite and collisions between two such solitons do not exactly preserve the pulse shape @xcite ; as we shall see in the following , the expression _ gap soliton _ is however physically justified by the fact that the pulse distortion following a collision is generally small . since the first observation of optical gap solitons in 1995 , intense experimental activity is actually in progress for the generation and characterization of gap solitons and bragg modulational instabilities in optical fibers @xcite . in very recent years ,
solitonic excitations are beginning to be investigated also in the context of nonlinear atom optics @xcite : dark solitons in the form of stable density dips in a otherwise uniform condensate have been recently observed @xcite ; bright gap solitons @xcite and modulational instabilities @xcite are actually under intense investigation ; despite the positive atom - atom scattering length , such stable atomic pulses can exist inside an optical lattice thanks to the negative effective mass of valence band atoms . in the present section we present a new method to generate narrow bright gap solitons starting from a wide atomic condensate incident on an optical lattice
consider a long coherent matter wave pulse ( i.e. a bose - condensed atomic cloud ) incident on the same attractive lattice as in the previous section with a kinetic energy just above the lower edge of the gap . at linear regime ,
such a wave packet is nearly completely reflected so that the interference of the incident and reflected waves creates a standing wave pattern in front of the reflection point . far outside the lattice ,
the period of the pattern is fixed by the wave vector @xmath47 of the incident wave packet and is therefore of the order of the lattice period @xmath9 . inside the lattice ,
the standing wave pattern originates instead from the interference of a forward and a backward bloch waves : the amplitude of the fast oscillations follows from the density modulation of each bloch wave function with a periodicity @xmath71 , while the local period of the slower modulation which originates from the interference of the two bloch waves @xmath72 is equal to @xmath73 .
as we approach the reflection point , the bloch waves approach to the band edge , so that @xmath74 and the period of the modulation is strongly increased with respect to the lattice period , although it still remains much shorter than the size of incident wave ; this effect is apparent in the leftmost snapshot of fig.[fig : solit ] . , @xmath66 , @xmath75 ; same lattice parameters as in fig.[fig : modinstab ] ) : pulse shape snapshots at different times .
vertical dashed lines indicate the spatial extension of the lattice ; the dot - dashed line indicate the reflection point at linear regime .
[ fig : solit],width=288 ] provided the nonlinear term is sufficiently small , the interactions do not wash out the standing wave interference pattern @xcite but simply blue - shift the local band edge at the spatial position of the rightmost antinodes ; in this way , the wave packet frequency is pushed out from the local gap and valence band states result available for propagation .
the resulting short pulses , stabilized by the effective attractive interaction , can therefore propagate along the lattice as solitonic objects .
the higher the density , the larger the number of pulses for which this mechanism is effective and which are then able to propagate along all the lattice without being reflected . by carefully choosing the density ,
a single soliton can be launched along the lattice ( fig.[fig : solit ] ) ; unfortunately , given the complexity of the nucleation process , it is not physically evident how the parameters of the soliton ( e.g. the group velocity @xmath76 and the peak density ) can be controlled by acting on the parameters of the wide incident bose - condensate .
once the pulse has crossed the flat region of the lattice and has got to its opposite end , the effective mass of the atoms becomes positive again and the pulse is immediately broadened under the combined effect of group velocity dispersion and mean field repulsion ( see the two last snapshots in fig.[fig : solit ] ) .
a proof of the solitonic nature of the generated pulse is obtained from a study of its collisional dynamics ( fig.[fig : collis ] ) : a pair of such pulses symmetrically generated at the two ends of the lattice collide in the middle of the lattice .
their solitonic properties result clearly from the fact that their shapes as well as the number of atoms contained in each of them are only weakly affected by the collision process .
the small broadening of the pulses that can be observed in fig.[fig : collis ] is a signature of the fact that the nonlinear wave equation in a periodic potential is integrable only in an approximate way @xcite . and
propagating along the lattice with opposite velocities .
all the collisional dynamics takes place in the region @xmath69 where the lattice profile is flat .
[ fig : collis],width=288 ] thanks to the effective mass @xmath38 much smaller than the free space mass @xmath17 ( for the parameters of fig.[fig : solit ] @xmath77 ) , the solitonic width is significantly larger than the free space healing length at the same value of the density ; for a typical value of the lattice period of the order of @xmath78 , the solitonic length turns out of the order of @xmath79 which is well within the capabilities of actual detection systems .
the mean - field interaction energies required for the observation of solitonic effects are in general smaller or of the order of one tenth of the recoil energy @xmath80 : for the most relevant case of @xmath81rb atoms and @xmath82na , this value corresponds to reasonable densities of the order of @xmath83 .
the characteristic time for the modulational instability and soliton formation processes described in the previous sections is of the order of @xmath84 , which means @xmath85 for @xmath82rb atoms and @xmath86 for @xmath81na .
the method here described for the generation of bright gap solitons is significantly different from previous proposals @xcite : in our approach the soliton pulse shape originates not from the whole bec cloud , but only from the much shorter density bump corresponding to an antinode of the standing matter wave .
this fact allows one to obtain short solitons from a wide bec without the need for a dramatic pulse compression under the effect of effectively attractive interactions .
provided the gap soliton is wide enough , only a narrow group of bloch states around a central wave vector @xmath87 are populated and an accurate description can be analytically obtained within the so - called envelope function framework : denoting with @xmath88 the bloch eigenfunction at @xmath89 , we write the wave function @xmath90 of the coherent matter pulse as the product of the slowly varying envelope @xmath91 and the fastly oscillating bloch eigenfunction @xmath88 @xmath92 in the following , the bloch eigenfunction @xmath93 is assumed to be normalized according to @xmath94 which corresponds to @xmath95 in the @xmath96 basis of ( [ eq : twomodes ] ) . if its variations of the envelope @xmath91 are slow enough , it can be shown @xcite that @xmath91 obeys a simple integrable nonlinear schrdinger equation ( nlse ) @xmath97 in which the effective mass @xmath38 and the effective coupling @xmath98 depend on the central wave vector @xmath87 ; obviously , the stability of the solitonic pulse requires @xmath99 , i.e. @xmath87 close to the upper edge of the valence band at @xmath100 .
while an explicit expression for @xmath101 has been already given in ( [ eq : meff ] ) , the effective coupling @xmath102 turns out to be expressed in terms of the bloch wave function by @xmath103 within the two - mode ansatz ( [ eq : twomodes ] ) , this quantity can be rewritten in the simple form @xmath104 where @xmath105 are the projections of the bloch eigenfunction on the forward and backward propagating waves .
notice that the density modulation of the bloch wavefunction at gap edge ( @xmath106 ) makes the effective coupling a factor @xmath107 larger than the one far from the gap , i.e. the free - space one . in fig.[fig :
solit ] with the analytical prediction ( [ eq : solitanal ] ) for the envelope ( solid line ) .
approximate prediction obtained using the band - edge values ( @xmath108 for @xmath98 and @xmath38 .
[ fig : solit2],width=288 ] under these assumptions , the envelope @xmath109 of the solitonic wave packet has the simple expression @xcite @xmath110 with the width @xmath111 given by @xmath112 as expected , the size @xmath111 of the soliton is of the order of the healing length @xmath113 .
the accuracy of this approximate description is apparent in fig.[fig : solit2 ] where we compare the numerically obtained wave packet with the analytical prediction ( [ eq : solitanal ] ) for the envelope ; the central wave vector @xmath87 of the wave packet has been determined from the group velocity by means of ( [ eq : vg ] ) , the envelope amplitude @xmath114 from the peak density of the pulse .
in this paper we have theoretically investigated the transmission dynamics of coherent matter pulses ( such as those that can be extracted from bose - einstein condensates ) which incide on finite optical lattices ; such systems are matter wave analogs of the nonlinear bragg fibers currently studied in nonlinear optics . at linear regime ( i.e. in the low - density or weak - interaction limit ) , we have characterized the dependence of the intensity and shape of the transmitted pulse on the velocity and size of the incident condensate in terms of the dispersion of matter waves inside the lattice ; as in the case of light waves in periodic dielectric structures or in the case of electrons in crystalline solids , the dispersion of matter waves in the periodic potential of optical lattices is in fact characterized by allowed bands and forbidden gaps . the dynamics in the presence of interactions is found to be even richer : an interpretation of the numerically predicted effects is put forward in terms of familiar concepts from nonlinear optics , such as optical limiting , optical bistability , and modulational instability . in particular , we have investigated a possible way of generating narrow bright gap solitons from a wide incident bose - condensate : the modulational instability is seeded from the strongly modulated density profile of the standing wave which is formed in front of the finite optical lattice by the interference of the incident and reflected matter waves .
the solitonic nature of the generated pulses has been checked from their shape which is in excellent agreement with a simple analytical model as well as from their dynamical and collisional properties .
finally , we have verified that the range of physical parameters that is required for the observation of the effects predicted in the present paper falls well within the possibilities of actual experimental setups .
i.c . acknowledges financial support from the eu ( `` marie curie '' fellowship of programme `` improving human research potential and the socio - economic knowledge base '' under contract number hpmf - ct-2000 - 00901 ) .
i.c . is grateful to yvan castin , lincoln carr and anna minguzzi for useful discussions .
laboratoire kastler brossel is a unit de recherche de lcole normale suprieure et de luniversit paris 6 , associe au cnrs . c. s. adams and e. riis , prog .
quantum electron .
* 21 * , 1 ( 1997 ) ; e. peik , m. ben dahan , i. bouchoule , y. castin , and c. salomon phys . rev . a 55 , 2989 - 3001 ( 1997 ) ; q. niu , x .-
g . zhao , g. a. georgakis , and m. g. raizen , phys .
lett . * 76 * , 4504 ( 1996 ) ; s. r. wilkinson , c. f. bharucha , k. w. madison , q. niu , and m. g. raizen , phys .
lett . * 76 * , 4512 ( 1996 ) b. p.
anderson and m. kasevich , science * 282 * , 1686 ( 1998 ) ; s. burger _
et al . _ ,
lett . * 86 * , 4447 ( 2001 ) ; o. morsch , j. h. mller , m. cristiani , d. ciampini , and e. arimondo , phys .
87 * , 140402 ( 2001 ) ; m. greiner , i. bloch , o. mandel , t. w. hnsch , and t. esslinger , phys .
lett . * 87 * , 160405 ( 2001 ) o. zobay , s. ptting , p. meystre , and e. m. wright , phys .
a * 59 * , 643 ( 1999 ) ; s. ptting , o. zobay , p. meystre , e.m .
wright , j. mod . opt . * 47 * , 2653 ( 2000 ) ; a. trombettoni and a. smerzi , phys .
* 86 * , 2353 ( 2001 ) k. berg - sorensen and k. molmer , phys .
a * 58 * , 1480 ( 1998 ) ; d. jaksch , c. bruder , j. i. cirac , c. w. gardiner , and p. zoller , phys . rev . lett . * 81 * , 3108 ( 1998 ) ; s. potting , m. cramer , c. h. schwalb , h. pu , p. meystre , phys . rev . a * 64 * , 023604 ( 2001 ) b. j. eggleton , r. e. slusher , c. m. de sterke , p. a. krug , and j. e. sipe , phys . rev
* 76 * , 1627 ( 1996 ) ; b. j. eggleton , c. m. de sterke , and r. e. slusher , j. opt
. soc . am .
b * 16 * , 587 ( 1999 ) ; s. pitois , m. haelterman , and g. millot , opt .
lett , * 26 * , 780 ( 2001 ) c. m. de sterke and j. e. sipe , phys
a * 42 * , 2858 ( 1990 ) ; c. m. de sterke , phys . rev .
a * 45 * , 8252 ( 1992 ) ; h. g. winful , r. zamir and s. feldman , appl .
* 58 * , 1001 ( 1991 ) ; c. m. de sterke , phys . rev .
a * 45 * , 8252 ( 1992 ) ; a. b. aceves , c. de angelis , and s. wabnitz , opt . lett . * 17 * , 1566 ( 1992 ) ; n. m. litchinitser , g. p. agrawal , b. j. eggleton , and g. lenz , optics express * 3 * , 411 ( 1998 ) i. carusotto and g. c. la rocca , phys . rev . lett . * 84 * , 399 ( 2000 ) ; i.carusotto and g.c.la rocca in _ bose - einstein condensates and atom lasers _ eds . s. martellucci , a. n. chester , a. aspect , and m. inguscio ( kluwer academic / plenum publishers , new york , 2000 ) , pag.153 w. p. reinhardt and c. w. clark , j. phys .
b * 30 * , l785-l789 ( 1997 ) ; a. d. jackson , g. m. kavoulakis , and c. j. pethick , phys . rev .
a * 58 * , 2417 ( 1998 ) ; th .
busch and j. r. anglin , phys .
lett . * 84 * , 2298 ( 2000 ) ; th .
busch and j. r. anglin , phys .
lett . * 87 * , 010401 ( 2001 ) | we theoretically investigate the transmission dynamics of coherent matter wave pulses across finite optical lattices in both the linear and the nonlinear regimes .
the shape and the intensity of the transmitted pulse are found to strongly depend on the parameters of the incident pulse , in particular its velocity and density : a clear physical picture for the main features observed in the numerical simulations is given in terms of the atomic band dispersion in the periodic potential of the optical lattice .
signatures of nonlinear effects due the atom - atom interaction are discussed in detail , such as atom optical limiting and atom optical bistability . for positive scattering lengths ,
matter waves propagating close to the top of the valence band are shown to be subject to modulational instability .
a new scheme for the experimental generation of narrow bright gap solitons from a wide bose - einstein condensate is proposed : the modulational instability is seeded in a controlled way starting from the strongly modulated density profile of a standing matter wave and the solitonic nature of the generated pulses is checked from their shape and their collisional properties . in recent years
, a great interest has been devoted to theoretical as well as experimental studies of the propagation of matter waves in the periodic potential of optical lattices .
the first experiments were carried out using ultracold atomic samples @xcite ; later on , the realization of atomic bose - einstein condensates ( becs ) @xcite and their coherent loading into optical lattices @xcite have opened the possibility of investigating features which follow from the coherent nature of the bose - condensed atomic sample @xcite . at the same time , the propagation of light waves in linear and nonlinear periodic dielectric structures has been a very active field of research : global photonic band gaps ( pbg ) have been observed @xcite and one - dimensional nonlinear periodic systems such as nonlinear bragg fibers @xcite are actually under intense investigation given the wealth of different phenomena including optical bistability , modulational instability and solitonic propagation that can be observed @xcite . given the very close analogy between the behavior of coherent matter waves and nonlinear optics , we expect that the concepts currently used to study the physics of nonlinear bragg fibers can be fruitfully extended to the physics of coherent matter waves in optical lattices : the optical potential of the lattice plays in fact the role of the periodic refractive index , the atom - atom interactions are the atom - optical analog of a kerr - like nonlinear refractive index and the gross - pitaevskii equation of mean - field theory corresponds to maxwell s equation with a nonlinear polarization term @xcite . in the present paper , we shall report a theoretical investigation of the transmission dynamics of coherent matter pulses which incide onto a finite optical lattice with a velocity of the order of the bragg velocity . in this velocity range ,
bragg reflection processes are most effective and the atomic dispersion in the lattice is completely different from the free - space one .
depending on the value of the density , spatial size and velocity of the incident atomic cloud , as well as on the depth and length of the lattice , a number of different behaviors are predicted by numerical simulations ; here , we shall focus our attention on the shape of the transmitted pulse just after it has crossed the lattice as well as while it is still propagating in the lattice . in particular
, we shall discuss a new mechanism which can be used to generate narrow bright atomic gap solitons propagating along the lattice starting from a wide incident bose - einstein condensate . for more details on the continuous - wave transmission and reflection spectra at linear regime
, the reader can refer to @xcite ; some aspects of the linear pulse dynamics are discussed in @xcite .
the geometry considered in the present paper as well as in @xcite is significantly different from the one usually considered in recent experimental works @xcite of bec dynamics in optical lattices in which an optical lattice is switched on and superimposed to a stationary condensate ; the dynamics of the condensate inside the lattice is then studied in response to some external force such as gravity , an acceleration of the lattice , or a spatial translation of the magnetic potential . the present paper is organized as follows : sec.[sec : system ] describes the physical system under examination . in sec.[sec : bands ] we review some basic concepts on the dispersion of matter waves in the periodic potential of an infinite optical lattice and present some simple analytical calculations which accurately reproduce the dispersion of the atomic bands in the neighborhood of the first forbidden gap .
the linear regime propagation of coherent matter wave pulses across finite lattices is the subject of sec.[sec : linear ] .
the intensity as well as the shape of the transmitted pulse are found to strongly depend on the properties of the weak incident pulse , in particular its velocity and spatial size ; a simple interpretation of the observed phenomena in terms of allowed bands and forbidden gaps is provided . in sec.[sec : nonlinear ] , we discuss the effect of atom - atom interactions on the propagation of the pulse in the different cases : simple explanations for the observed behavior are put forward in terms of familiar concepts of nonlinear optics such as optical limiting , optical bistability , or modulational instability ; in particular , modulational instability is shown to occur for positive scattering length atoms when the matter wave is propagating at the top of the valence band ; at this point of the dispersion , the effective mass is in fact negative . in sec.[sec : soliton ] , we propose a new scheme to control the modulational instability of valence band atoms in order to obtain a narrow bright gap soliton from a wide condensate : as an initial seed for the instability , the standing matter wave pattern created by the interference of the incident and reflected waves is used .
the solitonic nature of the generated pulses is verified by looking at their dynamical and collisional properties as well as by comparing the pulse shape with the analytical predictions of the envelope - function approximation discussed in sec.[sec : solitanal ] .
conclusions are finally drawn in sec.[sec : conclu ] . |
pulsar wind nebulae ( pwne ) are the bubbles inflated by the relativistic winds of neutron stars as they interact with their surroundings .
their radio to x - ray emission is believed to result from synchrotron radiation of the high - energy particles injected by a rotation - powered neutron star in the presence of a strong magnetic field .
as such , pwne represent an ideal laboratory to study the physics of neutron stars and particle acceleration .
they are also good indicators for the presence of an active rotation - powered pulsar ; see @xcite for a detailed review on the structure and evolution of pwne . while the crab nebula has been viewed to represent the prototype pwn , a growing list of pwne show properties unlike the crab : namely , different spectral properties ( spectral index , spectral break in the radio ) and different morphology .
these are referred to in the literature as `` plerions of the second kind '' or non - crab - like plerions ( see * ? ? ?
* ) , a class that we have been targeting with combined radio and x - ray studies ; see @xcite for a recent review .
this paper addresses , using x - ray observations , the nature of ctb 87 ; the least studied object ( at least in x - rays ) among this growing class of pwne . in a subsequent paper ( kothes et al . in preparation ) , we present a dedicated study of the remnant at radio wavelengths . ctb 87 ( g74.9 + 1.2)@xmath10 has been classified as a plerionic type snr based on its centrally - filled morphology , lack of evidence of a shell , and linearly polarized radio flux @xcite . in the radio ,
ctb 87 has a size of 8@xmath36 and a flux density of 9 jy at 1 ghz @xcite .
early x - ray observations with the _ einstein _ satellite also identified ctb 87 as a plerionic snr , with an x - ray luminosity nearly 100 times weaker than the crab nebula in the @xmath11 kev band @xcite .
newer sensitive radio studies indicated that the central bright structure has a relatively steep radio spectrum , with a spectral index @xmath12 = 0.5 ( @xmath13 ) , which is unusual for a plerionic snr .
in addition , larger diffuse emission , @xmath016 in size , has been identified and characterized by a much flatter spectrum which is more consistent of a plerionic snr ( kothes et al .
2003 ; kothes et al . in preparation ) .
@xcite suggested that ctb 87 is located in the cygnus arm , and detected a very faint optical shell with 20 diameter in the vicinity of ctb 87 .
this optical shell has not been however confirmed and the author concludes that the small amount of data prevents them from drawing a firm conclusion .
locating ctb 87 in the cygnus arm puts the remnant at a much smaller distance than that estimated using hi data from the canadian galactic plane survey ( @xmath14 kpc , @xcite ) , a distance that implies a location within the perseus spiral arm .
egret _ source has been suggested to be possibly associated with ctb 87 , although @xcite argued that the snr is too weak and far away ( using an old distance estimate of 12 kpc ) to be the most likely candidate .
the revised kinematic distance to ctb 87 of 6.1 kpc weakened this argument , further increasing interest in ctb 87 as a potential @xmath15-ray source .
other @xmath15-ray studies of the area near ctb 87 with _ milagro
_ led to the detection of an unresolved source , mgro j2019 + 37 @xcite , which was further studied and resolved with _
veritas_. the _ veritas _ source , ver j2016 + 372 , showed an absence of variability throughout the observations and a spectrum similar to other pwne previously detected in the vhe band @xcite .
finally , _ fermi _ detected a high - energy source , 2fgl 2015.6 + 3709 , which was found to be likely variable and therefore a physical association with ctb 87 was deemed improbable @xcite .
here we present a detailed study of the x - ray observations of ctb 87 .
preliminary results have been published in @xcite and @xcite .
this paper is focused on the observations acquired with the _ chandra x - ray observatory _
, with a total exposure time of @xmath087 ks . we also briefly present an unpublished archival observation with the _ advanced satellite for cosmology and astrophysics ( asca ) _ , primarily to compare the overall spectrum of the nebula with that obtained with _
chandra_. section [ section : obs ] discusses the observations and data processing . in section [ section : images ] , we present the observed x - ray morphology and show the first evidence for a compact nebula , with a torus / jet - like structure , as well as a diffuse nebula that is offset from the radio nebula .
these structures surround a point source , the putative neutron star powering ctb 87 .
section [ section : spectra ] presents the spectral analysis , including the first spatially resolved spectroscopic analysis .
we show that the spectra from the point source and the compact and diffuse nebulae are well described by a power law with a hard photon index , and that the index generally steepens away from the point source . in section
[ section : disc ] we discuss the pwn morphology , properties , and stage of evolution , and infer the properties of the putative neutron star powering the pwn .
we propose that ctb 87 is most likely an evolved pwn , with overall multi - wavelength properties similar to the pwne in the snrs vela and [email protected] .
( g74.9 + 1.2 ) was observed with _
chandra _ on two occasions .
an 18 ks observation was performed with acis - s3 ( 2001 july 8 , obsid 1037 ) and a 70 ks observation ( 2010 january 16 , obsid 11092 ) was performed with acis - i3 .
the _ chandra _
data was processed with ciao v4.3 and caldb v4.4.1 @xcite .
the level 1 event files were reprocessed to new level 2 event files , using _
acis_process_events _ ( correct for ccd charge transfer inefficiencies , remove bad pixels ) and _ dmcopy _ ( remove bad grades and status , filter for good time intervals and energy ) .
the effective exposures after filtering were 17.8 ks ( obs 1037 ) and 69.3 ks ( obs 11092 ) , yielding a total effective exposure of 87.1 ks . for image creation
, the two observations were merged after applying the aspect solution to each observation ( _ reproject_events _ ) .
_ wavdetect _ was used to detect point sources and _ dmfilth _ was used to remove the sources and fill the holes with a poisson distribution calculated from a background region surrounding each source . _
asphist _ , _ mkinstmap _ , and _ mkexpmap _ were used to create exposure maps for each ccd , which were then summed .
_ dmimgthresh _ was used to remove pixels with very low exposure times and the image was divided by the exposure map to produce the final exposure corrected image .
this procedure was repeated for various energy bands and the resulting images are presented in section [ section : images ] .
spectra were extracted from each observation separately , and simultaneously fit using _ xspec _ v12.6.0 .
_ specextract _ was used to extract spectra and response files from regions of interest .
the regions used and resulting spectra are discussed in section [ section : spectra ] . when discussing diffuse emission , each larger region excludes events contained in the interior regions , as well as any counts from point sources detected within the diffuse emission .
spectra were grouped using _ dmgroup _
, with the minimum number of counts per bin listed in table 2 .
all uncertainties quoted on model parameters are to 90% confidence level . for calculations requiring a distance estimate ,
we assume @xmath16 kpc @xcite . ctb 87 was also observed on 1995
may 29 with _ asca _
( sequence # 53041000 ) .
it was imaged with the gis for 53 ks but was out of the field of view of the sis .
data was processed with _ xselect _ ( heasoft v5.3.1 ) .
we present the _ asca _
results only for the purpose of comparing the spectral result for the pwn to the global fit with _ chandra_. the x - ray observations of ctb 87 presented in this paper are summarized in table [ table : observations ] .
figure [ figure : pwn ] ( left panel ) shows the combined _ chandra _ observations 1037 and 11092 , after removing point sources , filling the holes , and exposure correcting the image , as described in section [ section : obs ] . the energy range covered is [email protected] kev
( events below 0.3 kev and above 7.0 kev were excluded as the background dominates at these energies ) and the net exposure time is 87.1 ks . the image was smoothed with a 2d gaussian ( @xmath17 = 3 ) and 21 cm contours are overlaid , using data acquired with the _ dominion radio astrophysical observatory _ as part of the canadian galactic plane survey @xcite .
the entire image is 525 @xmath3 525 .
the pwn is @xmath0200 @xmath3 300 overall with brighter emission to the southeast .
the total observed x - ray flux of the nebula is 1.8 @xmath3 10@xmath18 erg @xmath5 s@xmath7 ( @xmath00.3 cts s@xmath7 ) .
the peak of the x - ray emission is seen in the southeast of the nebula , with a point source clearly visible at the peak ( section [ section : ptimage ] ) .
the x - ray peak is located near the southeastern edge of the bright radio nebula , offset from the center of the nebula , with the diffuse emission extending 100 , 100 , 250 , and 60 from the peak to the se , ne , nw , and sw , respectively .
we observe arcs at the edges of the x - ray nebula ( highlighted on figure [ figure : pwn ] ) , extending from the southeast near the point source , back toward the radio peak .
the nature of these features is discussed in section [ section : disc ] .
figure [ figure : pwn ] ( right panel ) is a color image created by dividing the data into three bands , exposure correcting each , and smoothing ( @xmath17 = 8) each before combining .
the red , green and blue colors correspond to the [email protected] kev , [email protected] kev , and [email protected] kev bands , respectively .
the contrast of each image was adjusted to reduce the background and bring out the pwn emission .
we see the hardest ( bluest ) emission near the bright point source ( the putative pulsar , section [ section : ptimage ] ) .
the arcs along the edge of the pwn ( highlighted on figure [ figure : pwn ] , left panel ) appear as possible bow shock structures and are real : the signal - to - noise ratio of the eastern arc is 12.8 while the signal - to - noise ratio of the western arc is 11.5 . in section [ section : disc ]
we discuss the nature of these features .
thanks to _ chandra _ s superb resolution , a point source - the putative pulsar - is found at the southeastern edge of the bright radio nebula with the following coordinates : @xmath12(2000)=20@xmath1916@xmath20092 , @xmath21(2000)=+3711105 .
we refer to this new _ chandra _ point source as cxou j201609.2 + 371110 .
a point spread function ( psf ) was created using _
mkpsf _ for an energy of 3 kev ( characteristic of the source s energy histogram ) and an off - axis angle of 238 ( the location of the point source in observation 11092 ) and found to have a fwhm of 115 .
we note that this is the positional uncertainty of the point source .
figure [ figure : pointsource ] ( left ) shows a 20 @xmath3 20 region , centered on the point source and combining both _
chandra _ observations ( as for figure 1 ) . this exposure corrected image was smoothed using a gaussian with @xmath17 = 1and covers the energy range [email protected] kev .
we see evidence of an arc - like ( torus ) structure to the northeast of the point source with @xmath010 diameter , and jet - like features extending @xmath05 to the northeast and southwest .
such high - resolution features are expected to form close to the termination shock , due to the deposition of the neutron star s wind energy into its surroundings ( see e.g. * ? ? ?
figure [ figure : pointsource ] ( right ) shows a radial profile created from the observational data , along with a radial profile of the psf .
the data is consistent with a point source to a radius of @xmath222 , beyond which the data clearly has a higher surface brightness than a point source .
this excess emission we term the ` compact nebula ' and is comprised of the candidate torus and jets , as seen in the left panel of figure [ figure : pointsource ] .
we also tested the source extent using _
srcextent_. the source observed size was 205 , the psf observed size was 115 , and the estimated intrinsic size was 169 , confirming that there is an extended source surrounding the point source .
the psf described above was then used as a convolution kernel in @xmath23 to model the point source emission and determine the morphology of the excess emission .
figure [ figure : residuals ] shows the data , the model , and the residuals resulting from the fit .
the model consisted of a constant background level plus a 2d gaussian with a fwhm of 11 and an amplitude of 117 counts / pixel .
we see a large excess in the residual map to the northeast of the point source location , seeming again to have the torus - jet morphology as suggested in figure [ figure : pointsource ] .
to extract the spectrum of the point source , we used a circular region centered at cxou j201609.2 + 371110 and with a radius = 2 .
the background spectrum was extracted using an annulus surrounding the source with radius = 2@xmath14 .
given the small number of counts ( see table 2 ) , we freeze the column density , @xmath2 , to the best value of 1.4 @xmath3 10@xmath4 @xmath5 obtained from fitting the total diffuse emission ( see section [ section : nebulaspectra ] ) .
we attempted fitting with an absorbed power law model and an absorbed blackbody model , as would be expected from a neutron star .
fitting with the absorbed power law model yields a hard photon index of @xmath24 = 1.1 ( [email protected] ) and an absorbed flux of @xmath25 = 5.2 @xmath3 10@xmath26 erg @xmath5 s@xmath7 ( @xmath27 = 0.91 ; see table 2 ) .
the corresponding x - ray luminosity is @xmath28 = 3.0 @xmath3 10@xmath29 @xmath30 erg s@xmath7 or @xmath31 = 2.5 @xmath3 10@xmath29 @xmath30 erg s@xmath7 .
the left panel of figure [ figure : pointspectrum ] shows the data and fitted model , with the spectra from the two observations grouped together for plotting purposes only .
we note that allowing the column density to vary gives a column density of @xmath2 = 1.6 ( [email protected] ) @xmath3 10@xmath4 @xmath5 ( @xmath27 = 0.93 ) , consistent with that of the diffuse nebula , but due to the low number of counts the parameters are much less constrained .
fitting with an absorbed blackbody model we find a similarly suitable fit ( @xmath27 = 0.99 , figure [ figure : pointspectrum ] , right ) with a temperature @xmath32 = 1.1 ( [email protected] ) kev , an absorbed flux of @xmath25 = 3.2 @xmath3 10@xmath26 erg @xmath5 s@xmath7 , and a luminosity @xmath28 = 1.8 @xmath3 10@xmath29 @xmath30 erg s@xmath7 or @xmath31 = 1.6 @xmath3 10@xmath29 @xmath30 erg s@xmath7 .
however , the temperature is unreasonably high and , when combined with the luminosity , implies a neutron star radius of @xmath00.03 km which is unreasonably small .
attempting a two component model ( power law plus blackbody ) was not possible given the poor statistics .
therefore we favor the power law model fit , which is also consistent with the x - ray emission seen from other neutron stars powering pwne , such as psr j2022 + 3842 in g76.9 + 1.0 ( @xmath24 = 1.0 @xmath33 0.2 , * ? ? ?
* ) , and psr j1833 - 1034 in g21.5 - 0.9
( @xmath24 = 1.14 @xmath33 0.07 , * ? ? ?
we then studied the emission from the compact nebula by extracting a spectrum from an annulus with radius 2@xmath110 , using a larger annulus with radius 10@xmath120 as the background spectrum .
this compact nebula , like the point source , has a hard spectrum with @xmath24 = 1.2 ( [email protected] ) and an x - ray luminosity of @xmath28 = 8.7 @xmath3 10@xmath29 @xmath30 erg s@xmath7 or @xmath31 = 6.5 @xmath3 10@xmath29 @xmath30 erg s@xmath7 .
the best - fit spectral parameters are summarized in table [ table : spectra ] and the compact nebula spectrum is shown in figure [ figure : compactnebspectrum ] . to study the properties of the extended diffuse emission , we extracted a spectrum of the extended nebula shown in figure [ figure : pwn ] ( right image ) , omitting the inner 10 radius compact nebula .
a power law model provides a good fit , as would be expected from a pwn , with the following best - fit parameters ( see table 2 ) : @xmath2 = 1.38 ( [email protected] ) @xmath3 10@xmath4 @xmath5 , @xmath24 = 1.68 ( [email protected] ) , @xmath27 = 1.01 , @xmath25 = 2.1 @xmath3 10@xmath18 erg @xmath5 s@xmath7 , @xmath28 = 1.6 @xmath3 10@xmath6 @xmath30 erg s@xmath7 , and @xmath31 = 9.6 @xmath3 10@xmath34 @xmath30 erg s@xmath7 .
the above fit is consistent with the results we find from archived _ asca _ gis data .
the point source in ctb 87 was not resolved in the _ asca _ observation and so we fit the entire nebula , a circular region with a radius of 34 , with a background spectrum extracted from the same detector , outside the diffuse emission . using gis-2 and gis-3 data (
ctb 87 was not visible in the sis data ) , we accumulate 3213 counts after background subtraction ( compared to 10136 net counts with _
chandra _ ) .
fitting gis-2 and gis-3 data simultaneously with a power law model in the [email protected] kev energy band , we find that @xmath2 = 1.14 ( [email protected] ) @xmath3 10@xmath4 @xmath5 , @xmath24 = 1.77 ( [email protected] ) , @xmath27 = 0.95 , @xmath35 = 2.2@xmath3 10@xmath18 erg @xmath5 s@xmath7 , @xmath28 = 1.4 @xmath3 10@xmath6 @xmath30 erg s@xmath7 , and @xmath31 = 1.1 @xmath3 10@xmath6 @xmath30 erg s@xmath7 ( see figure [ figure : pwnspectrum ] , left ) .
next , we perform a spatially resolved spectroscopic study of the extended nebula . to look for variations in the photon index without an artificially varying column density while fitting smaller regions with fewer counts
, we freeze @xmath2 to the value from the overall diffuse fit .
table [ table : spectra ] shows the results of fitting a power law model to the regions shown in figure [ figure : pwn ] ( right ) and figure [ figure : pwnspectrum ] ( right ) shows the power law fit to each region ( obs .
11092 and 1037 were fit simultaneously and are grouped together for plotting purposes only ) .
beginning at the point source and moving outward across the pwn , we observe an overall increase ( or steepening ) in the photon index .
this spectral steepening is expected as we move away from a neutron star due to expansion and synchrotron losses @xcite .
we also explored background regions within the diffuse emission for the inner and mid regions .
we find a similar trend in the photon index , however , with a lower net flux and with the parameters being less constrained .
the brightening along the edges of the diffuse pwn is well fit by a power law with @xmath24 = 1.5 ( 1.21.9 ) for the eastern arc and @xmath24 = 1.7 ( 1.32.1 ) for the western arc .
we have searched for thermal emission that would arise from hot plasma in the snr or interaction with the reverse shock .
we do not find evidence for any significant emission .
in particular , the residuals ( lower panels of figure [ figure : pwnspectrum ] ) do not show any significant excess at low energies that may indicate the presence of thermal emission in the remnant
. however , this may be due to the limited sensitivity of the observation and field coverage .
a deeper , more sensitive , x - ray observation covering the entire radio nebula is needed to search for any faint thermal x - ray emission .
this is further discussed below .
the snr s age is normally determined using the thermal emission properties of the x - ray emitting plasma associated with the snr shell .
in particular , knowledge of the shock velocity ( @xmath36 ) and the actual snr size ( @xmath37 ) yields a direct measurement of the age as : @xmath38 = @xmath39 @xmath40 , where @xmath39 is a factor that depends on the evolutionary phase of expansion of the snr . for ctb 87 ,
no snr shell has been detected ( as mentioned above ) . in the absence of a snr shell detection
, we refer next to the offset between the radio and x - ray emission from the pwn , which can be attributed to the neutron star s motion since its birth following the supernova explosion .
the peak of the 21 cm emission is located at @xmath12(2000)=20@xmath1916@xmath20044 , @xmath21(2000)=+3712316 , @xmath0100 from the x - ray peak ( or cxou j201609.2 + 371110 ) . at a distance of @xmath16
kpc @xcite , this separation translates to @xmath41 pc perpendicular to the line of sight .
4b of @xcite shows the majority of pulsars with an age @xmath223 myr have a two - dimensional speed in the range @xmath42 km s@xmath7 . from this range
, we estimate an age of @xmath05@xmath128 kyr for ctb 87 , assuming the pulsar was born at the location of the radio peak and is now located at the x - ray peak .
this is a reasonable assumption particularly since the x - ray emission from a pwn is associated with synchrotron radiation from the freshly injected particles , while the radio emission characterizes the older population of synchrotron emitting electrons over the neutron star s lifetime . using the average 2d speed of 307 km s@xmath7 for pulsars with an age @xmath223 myr @xcite , we estimate an age of @xmath09 kyr for ctb 87 .
these age estimates are comparable to the 11 kyr age estimated for vela x @xcite and the 18 kyr age estimated for [email protected] @xcite , pwne which we compare ctb 87 to further below . the advanced age for ctb 87 ( at least an order of magnitude higher than young pwne like the crab and [email protected] ) is also hinted at by the large radio size .
the central 8@xmath43@xmath36@xmath43 radio component implies a physical size of @xmath0(14 @xmath3 11)@xmath44 pc . the extended 16@xmath43 diffuse component ( kothes et al .
2003 ; kothes et al . in preparation ) implies an even larger pulsar wind nebula of @xmath028@xmath44 pc in extent .
this is slightly smaller , but comparable in size to the unusual pwn g76.9 + 1.0 ( 29 pc @xmath3 35 pc ) which has been recently argued to be an evolved pwn with an age @xmath22 36 kyr and which harbors a newly discovered energetic pulsar with a 9 kyr characteristic age ( arzoumanian et al .
2011 ) . in the remainder of this discussion
, we will assume an age of @xmath45 kyr for ctb 87 where calculations require an age estimate , keeping in mind that the age is uncertain given the estimates above ( see also an estimate of the putative pulsar s characteristic age in section [ section : pulsarproperties ] ) .
moreover , if we take into account the even larger offset ( an additional 30@xmath46 ) between the peak of the _ extended _ radio emission and the peak of the x - ray emission , the age would be 30% larger for the same assumed velocity .
this will be further discussed in the forthcoming radio paper ( kothes et al . in preparation ) . additionally ,
if the offset is due to a reverse shock ( section [ section : reverseshock ] ) and not related to the proper motion , this estimate may not reflect the age of the system .
we should note that there is accumulating evidence for pulsars spin axes to be roughly aligned with the direction of their motion , albeit with a small number of systems for which 3d information is available @xcite .
@xcite provide further evidence that alignment between projected spin - axes and pulsar proper motion is more likely than orthogonality . in figure
[ figure : pointsource ] ( left ) , we see the possible jets ( believed to define the spin - axis ) are nearly perpendicular to the apparent proposed proper motion of the putative pulsar . if our assumed pulsar s direction of motion and jets direction are correct , then ctb 87 would be a system that does nt follow the expected alignment from the above - mentioned studies .
a deeper _ chandra _ observation and proper motion studies of the x - ray source will confirm this . to the best of our knowledge ,
no pulsations have been reported from ctb 87 @xcite .
we can however infer the properties of the neutron star powering ctb 87 from the x - ray luminosity of the nebula .
the unabsorbed flux of ctb 87 in the @xmath47 kev band is @xmath48 erg @xmath5 s@xmath7 , implying a luminosity of @xmath49 erg s@xmath7 .
we can estimate the spin - down energy loss of the neutron star powering the pwn from the pwn luminosity using the empirical relationship @xmath50 @xcite .
this yields an estimate of @xmath51 erg s@xmath7 , consistent with the estimate of @xmath52 erg s@xmath7 obtained with the empirical relationship @xmath53 @xcite .
the efficiency in converting the spin - down luminosity to synchrotron emission is then @xmath54 , which is within the range observed for pwne powered by rotation - powered pulsars .
the pwn luminosity can also be used to estimate the characteristic age ( @xmath55 ) of the neutron star using @xmath56 @xcite .
we estimate an age of @xmath57 years . assuming an age of @xmath58 kyr ( keeping in mind the large uncertainty on this estimate , noted in section [ section : age ] ) , @xmath59 erg s@xmath7 , a braking index of @xmath60=3 ( magneto - dipole braking ) , a moment of inertia of @xmath61 = 10@xmath62 g @xmath63 , and a neutron star radius of @xmath64 = 10@xmath65 km
, we can estimate the properties of the putative pulsar ( equations described for e.g. in * ? ? ?
the period of the pulsar is estimated as @xmath66 @xmath67 seconds .
the corresponding period derivative is @xmath68 and the corresponding equatorial surface dipole magnetic field is @xmath69 g , within the range of the field expected for pulsars with prominent pwne ( @xmath70 g , * ? ? ? * ) .
we expect to observe a termination shock around a rotation - powered neutron star , at a radius @xmath71 ; where @xmath72 is the nebular pressure . figure [ figure : pointsource ] suggests that the arc seen at a radius of @xmath05 corresponds to the termination shock . at an assumed distance of 6.1 kpc
, this corresponds to a distance of 0.15 pc from the point source .
this estimate is in agreement with the typical scale of 0.1 pc observed in other pwne @xcite . using the spin - down energy inferred above , and equating the expression of the termination shock site to 5
, we estimate the pressure in the pwn to be @xmath73 dyne @xmath5 .
we can then estimate the magnetic field in the nebula as @xmath74 g .
this estimate is in agreement with the preliminary value inferred from the radio study assuming equipartition between particle energy and magnetic field energy ( kothes et al . in preparation ) .
the absence of a snr shell detection implies expansion into a low - density medium . to infer the upper limit on the density of the medium that the pulsar wind nebula is expanding into ( which would be the shocked ejecta for a vela - like snr ) , we estimate next the maximum thermal contribution to the diffuse emission allowed by the data by adding a thermal component to the power law model .
we freeze the best fit power law parameters ( table [ table : spectra ] , ` total diffuse ' region ) , and add an astrophysical plasma emission code ( apec ) thermal component @xcite .
we increase the emission measure ( @xmath75 ) until the model level under 1 kev exceeds the data by 2@xmath17 , finding a conservative upper limit for the normalization on the thermal component of 9 @xmath3 10@xmath76 @xmath77 ( for @xmath32 = 0.5 kev ) .
this implies that @xmath78 = 10@xmath79 @xmath80 @xmath22 4.0 @xmath3 10@xmath81 @xmath82 @xmath9 , where @xmath83 is the volume filling factor , @xmath84 is the post - shock electron density , and @xmath85 . using a volume of 5.6 @xmath3 10@xmath86 @xmath87 @xmath88 for the pwn , @xmath89 @xmath9 .
if a snr shell exists beyond the detected pwn , the shocked ism must have a density lower than that estimated above for the ejecta , @xmath90 @xmath9 = @xmath91 @xmath9 ( for a volume filling factor @xmath92 ) .
this then implies an upper limit on the upstream ambient density of @xmath8 @xmath9 ( @xmath93 g @xmath9 where @xmath94 includes only hydrogen ) , assuming a compression ratio of 4 .
the inferred low density supports the lack of detection of the snr shell or x - ray emitting ejecta , and is suggestive of expansion into a stellar wind - blown bubble @xcite .
this upper limit is consistent with estimates obtained by extracting the background spectrum ( outside the pwn ) and assuming that the thermal emission from a putative shell is buried underneath the background . from these background spectra
, we estimate an upper limit of @xmath95 erg @xmath5 s@xmath7 arcmin@xmath96 on the observed x - ray flux of the putative snr shell .
the above density upper limit is similar to density estimates for g21.5 - 0.9 @xcite , whose shell was revealed after accumulating @xmath00.5 msec using _
chandra_. a deep x - ray observation of ctb 87 to search for the missing snr shell is therefore warranted .
additionally , future studies of the surrounding environment with co or hi data may help to further constrain the density .
this scenario will be revisited and further discussed in the follow - up radio paper ( kothes et al . in preparation ) .
assuming that the remnant is evolving in a uniform medium , the reverse shock will reach the center of the snr in a time @xmath97 kyr .
for an ejected mass of a few solar masses , an explosion energy of @xmath98 ergs , and a number density of @xmath99 @xmath9 , this is greater than 4 kyr for ctb 87 , consistent with our estimate above for the age ( see section 5.1 ) . for a remnant in the sedov stage ,
the snr radius can be estimated as @xmath100 @xcite . again assuming an explosion energy of @xmath101 erg , an ism density of @xmath102 @xmath9 , and an age of @xmath45 kyr ( large uncertainty noted in section [ section : age ] )
, we estimate the snr radius to be @xmath103 pc , which is larger than the revised size of the pwn ( @xmath028 pc diameter at 6.1 kpc ; see section 5.1 ) .
we conclude that all of the above mentioned arguments point to ctb 87 being an evolved pwn expanding into a low density medium , likely due to the stellar wind bubble blown by the progenitor star .
in section 5.2 we have interpreted the compact @xmath05 nebula shown in figure 2 as a wind - blown bubble associated with the deposition of the neutron star s wind energy into the surroundings . as shown in figure 1 , in addition to this compact nebula , the x - ray emission from ctb 87 is characterized by a cometary - like structure extending @xmath0250 northwest from cxou j201609.2 + 371110 and bridging the x - ray and radio emission peaks .
as mentioned earlier , this extended emission has some brightening along its edges ( see figure 1 ) .
this morphology may be interpreted in two ways .
first , these arcs may be a bow shock structure due to the motion of the pulsar to the southeast through the snr interior , leading to the formation of a pwn with a hybrid morphology ( i.e. a wind - blown bubble and a bow shock nebula ) .
alternatively , the reverse snr shock may be running into , and crushing , the pwn .
we discuss next these two scenarios . a pulsar given a kick at birth that has sufficient velocity will move out of the pwn and create a bow shock . as the pulsar moves outward in the snr , away from the supernova explosion site , the sound speed in the shocked ejecta decreases and the pulsar s motion becomes supersonic .
the ram - pressure from the pulsar s motion confines the nebula and the pwn evolves to a cometary appearance , with the bow shock forming around the head of the pwn after the pulsar has crossed @xmath068% of the snr shell , at half of the crossing time @xcite .
the pulsar wind in a bow shock pwn decelerates at a termination shock where the pressure balance is due to ram pressure from the neutron star s motion . in the direction of the star s motion ,
the termination shock radius , @xmath104 , is defined by @xmath105 ( @xmath106 for an isotropic wind ) , and is located at approximately half of the observed bow shock radius @xcite .
ram - pressure confinement in this hybrid morphology scenario requires that the ram pressure due to the pulsar s proper motion in the snr environment , @xmath107 , exceeds the nebular pressure , @xmath108 , estimated in section 5.2 ( @xmath109 is the density of the medium in which the psr is moving ) . assuming that the brightening along the pwn edge in figure [ figure : pwn ] is due to a bow shock , @xmath110 cm . using our inferred value for @xmath111 ( section 5.2 )
, we find that @xmath112 dyne @xmath5 .
this is only @xmath07% of the pressure @xmath113 dyne @xmath5 estimated from the termination shock of the more compact nebula ( section 5.2 ) , a ratio that is consistent with @xmath114 .
furthermore , for @xmath107 to exceed this nebular pressure of @xmath010@xmath115 dyne @xmath5 , @xmath116 should exceed @xmath0170 km s@xmath7 ( for an assumed ambient density of 0.2 @xmath9 , section 5.3 ) . as the formation of a prominent bow shock requires the nebular pressure to be exceeded by a large amount , the above suggests a high pulsar velocity
is needed for creation of a bow shock and argues against the cometary morphology being caused by ram - pressure confinement . in the following ,
we discuss the observed morphology in the light of an evolved pwn encountering the snr reverse shock .
this is motivated by the age inferred above ( section 5.1 ) and the offset observed between the radio and x - ray nebulae , as seen in other evolved pwne further described below .
a detailed description of the evolution of a pwn in a snr can be found in @xcite , @xcite , and @xcite .
we here briefly summarize .
as the expanding snr sweeps up sufficient mass from the ism , it evolves into a sedov - taylor phase . a reverse shock develops , running back into the supernova ejecta and eventually crushing the pwn , causing it to rebound and oscillate several times .
the timescale for the reverse shock to interact with the entire pwn is a significant fraction of the pwn life @xcite .
after these oscillations fade , the pwn again expands , now into hot , shocked ejecta .
the pulsar may have traveled a distance comparable or larger than the equivalent spherical pwn around a stationary pulsar .
the pulsar leaves its original wind bubble and generates a new smaller pwn at its current location , off - center within the pwn .
therefore , after the passage of the reverse shock the pwn consists of ( i ) a _ relic _ pwn from the initial energetic stage of the pulsar wind , appearing as a central ( possibly distorted ) radio pwn with little corresponding x - ray emission , and ( ii ) a head containing the pulsar , directed toward the snr shell , and emitting in x - rays .
express an upper limit on the age , @xmath117 , at which the reverse shock collides with the entire pwn surface as @xmath118 yr where @xmath119 is the total mechanical energy of the snr in units of @xmath98 erg , @xmath120 is the ejected mass , and @xmath94 is the ambient hydrogen number density . for ctb 87
we assume ( as above ) the explosion energy is @xmath98 erg , that the ejected mass is a few solar masses , and that the ambient density is @xmath220.2 @xmath9 .
we find that @xmath121 kyr for ctb 87 , at the low end of the estimated age ( section 5.1 ) .
this suggests that the reverse shock has potentially encountered the pwn , and is the cause of the observed morphology .
the supernova explosion may also have been asymmetric , or the ambient density may be non - uniform , leading to a more rapid evolution in one direction .
this is to be further discussed in the forthcoming radio paper ( kothes et al .
, in preparation ) .
this conclusion is further supported by the tev emission detected from ctb 87 @xcite .
focusing on _ chandra_-detected pwne near extended tev sources , @xcite find that faint extensions of compact pwne point toward offset centers of tev emission , supporting the notion that these tev sources are crushed pwne .
recent _ veritas _ studies of ctb 87 indicate that the tev emission overlaps with the radio emission , further supporting the picture of a ` relic ' pwn ( e. aliu and m. roberts 2012 , private communication ) .
the lack of x - ray emission from the snr ejecta or snr shell remains however a question to be addressed with deeper observations .
[email protected] is a composite remnant with a symmetric radio shell and off - center non - thermal component ( pwn ) , similar in structure to ctb 87 .
there is a radio finger extending from the pwn , at the end of which is a compact x - ray source that may be the pulsar @xcite .
_ chandra _ images show a compact source embedded in a cometary structure that trails back toward the radio pwn , as well as two prong - like structures extending into a large bubble in the opposite direction from the pwn @xcite .
we see a similar structure in ctb 87 ( figure 1 ) with the x - ray nebula offset from the radio nebula and most of the x - ray nebula trailing back toward the radio nebula .
we also see emission ahead of the point source in ctb 87 , extending to the southeast away from the radio nebula .
@xcite similarly favor a reverse shock interaction scenario over a bow shock scenario .
[email protected] however has more detailed spectra available , and thermal x - ray emission has been detected with _
xmm - newton_. the snr shell detection in [email protected] also allows further constraints to be placed on the snr properties .
tev emission from [email protected] has been detected by _
_ , showing the @xmath15-ray centroid is located between the radio centroid and the pulsar candidate @xcite .
vela - x ( associated with psr b0833 - 45 ) is another example of a pwn whose morphology has been affected by its interaction with the reverse shock ( blondin et al .
the radio nebula extends south of the pulsar , the northern portion possibly having been disturbed by the reverse shock .
vela - x has a complicated filamentary structure in x - rays , with thermal emission and enhanced o , ne , and mg abundances detected , signifying ejecta that has been mixed into the nebula by the propagation of the reverse shock @xcite .
the explanation of vela - x as a relic pwn was confirmed by @xcite with the detection of tev emission that is offset from the pulsar .
we have presented the first detailed x - ray study of ctb 87 , a large radio nebula classified as a pwn .
thanks to _ chandra _ s unprecedented resolution , we have detected a point source , cxou j201609.2 + 371110 , the likely putative pulsar powering ctb 87 , inside a compact x - ray nebula that is @xmath0100 offset from the peak of the radio nebula .
the compact @xmath05-radius nebula is characterized by a torus / jet - like structure indicating its origin as the putative pulsar s wind - blown bubble .
in addition to the x - ray source and compact nebula , a more extended nebula was found to trail back toward the peak of the radio emission .
a spatially resolved spectroscopic study of the nebula shows that the pwn is well fit by a power law model , with the overall spectrum steepening away from the point source .
we also observed brightening at the edges of the pwn , and discussed two possible interpretations : a bow shock caused by the motion of the pulsar or a pwn crushed by the snr s reverse shock . based on the observed multi - wavelength morphology , timescale for the reverse shock to interact with the pwn , and estimated age of the pwn
, the reverse shock interaction scenario seems more plausible .
a deep x - ray observation is needed to search for thermal emission associated with the shocked ejecta and/or the snr shell , as well as to confirm the high - resolution structures detected in the compact nebula . timing studies in the radio and x - ray
will allow the search for pulsations from cxou j201609.2 + 371110 .
finally , in the long term , a proper motion study of the putative pulsar is needed to shed light on its motion within the radio nebula .
we thank the referee for useful comments on the manuscript .
the research presented in this paper has used data from the canadian galactic plane survey , a canadian project with international partners , supported by the natural sciences and engineering research council ( nserc ) , and made use of the nasa s astrophysics data system and the high - energy astrophysics science archive research maintained at nasa s goddard space flight center .
this research was primarily supported by nserc .
ssh acknowledges support from an nserc discovery grant , the canada research chairs program , the canadian space agency , the canadian institute for theoretical astrophysics , and canada s foundation for innovation .
abdo , a.a . ,
allen , b. , berley , d. , et al .
2007 , , 658 , l33 acero , f. , djannati - ata , a. , frster , a. , gallant , y. , renaud , m. , for the h.e.s.s .
collaboration 2011 , icrc , 7 , 184 aharonian , f. akhperjanian , a.g .
, bazer - bachi , a.r . ,
2006 , , 448 , l43 aliu , e. for the _ veritas _ collaboration 2011 , icrc , 7 , 227 arzoumanian , z. , gotthelf , e.v . , ransom , s.m . ,
2011 , , 739 , 39 blondin , j. , m. chevalier , r. a. , & frierson , d. m. 2001 , , 563 , 806 biggs , j.d . , &
lyne , a.g .
1996 , , 282 , 691 dickel , j.r . , & denoyer , l.k . 1975 , , 80 , 437 duin , r.m . ,
israel , f.p . ,
dickel , j.r . , & seaquist , e.r .
1975 , , 38 , 461 ferrand , g. , & safi - harb , s. 2012 , adspr , 49 , 9 fruscione , a. , mcdowell , j.c . ,
allen , g.e . ,
2006 , , 6270 gaensler , b.m . , & slane , p.o .
2006 , , 44 , 17 gorham , p.w . , ray , p.s . ,
anderson , s.b . ,
kulkarni , s.r .
, & prince , t.a .
1996 , , 458 , 257 green , d.a .
2009 , basi , 37 , 45 halpern , j.p . ,
eracleous , m. , mukherjee , r. , & gotthelf , e.v .
2001 , , 551 , 1016 hobbs , g. , lorimer , d.r . ,
lyne , a.g . , & kramer , m. 2005 , , 360 , 974 kargaltsev , o. , & pavlov , g.g .
2008 , in aip conf .
983 , 40 years of pulsars : millisecond pulsars , magnetars and more , ed . c. bassa , z. wang , a. cumming , & v. m. kaspi (
melville , ny : aip ) , 171 kothes , r. , reich , w. , foster , t. , & byun , d. 2003 , , 588 , 852 la massa , s.m . ,
slane , p.o . , & de jager , o.c .
2008 , , 689 , l121 li , x .- h . ,
lu , f .- j .
, & li , z. 2008 , , 682 , 1166 lorimer , d.r . , lyne , a.g . , & camilo , f. 1998 , , 331 , 1002 matheson , h. & safi - harb , s. 2010 , , 724 , 572 ng , c .- y . &
romani , r.w .
2007 , , 660 , 1357 nolan , p.l . ,
abdo , a.a . ,
ackermann , m. , et al .
2012 , , 199 , 31 noutsos , a. , kramer , m. , carr , p. & johnston , s. 2012 , , 423 , 2736 possenti , a. , cerutti , r. , colpi , m. , & mereghetti , s. 2002 , , 387 , 993 safi - harb , s. , harrus , i.m . , petre , r. , et al .
2001 , , 561 , 308 safi - harb , s. , matheson , h. , & kothes , r. 2011 , 218th aas meeting , boston , ma safi - harb , s. 2012 , in aip conf .
ser . 1505 ,
high energy gamma - ray astronomy : 5th international meeting on high energy gamma - ray astronomy , ed .
f. a. aharonian , w. hofmann , & f. m. rieger , 13 ( arxiv:1210.5406 ) sitnik , t.g .
2010 , astronomy reports , 54 , 317 smith , r. k. , brickhouse , n. s. , liedahl , d. a. , & raymond , j. c. 2001 , , 556 , l91 taylor , a. r. , gibson , s. j. , peracaula , m. , et al .
2003 , , 125 , 3145 temim , t. , slane , p. , gaensler , b.m . ,
hughes , j.p .
, & van der swaluw , e. 2009 , , 691 , 895 van der swaluw , e. , achterberg , a. , gallant , y.a . , downes , t.p . , & keppens , r. 2003 , , 397 , 913 van der swaluw , e. , downes , t.p . , & keegan , r. 2004 , , 420 , 937 wilson , a.s .
1980 , , 241 , l19 weaver , r. , mccray , r. , castor , j. , shapiro , p. , & moore , r. 1977 , , 218 , 377 woltjer , l. , salvati , m. , pacini , f. , & bandiera , r. 1997 , , 325 , 295 = 1 . *
middle : * 2d model of the point source emission , convolved with the psf . *
right : * residuals after subtracting the model from the data , smoothed with @xmath17 = 1 .
all three panels cover the same area .
the cross marks the peak of the x - ray emission and is at the same location in all images .
see section [ section : ptimage ] for details of psf creation , model , and the fitted parameters.,title="fig:",width=201 ] = 1 .
* middle : * 2d model of the point source emission , convolved with the psf . *
right : * residuals after subtracting the model from the data , smoothed with @xmath17 = 1 .
all three panels cover the same area .
the cross marks the peak of the x - ray emission and is at the same location in all images .
see section [ section : ptimage ] for details of psf creation , model , and the fitted parameters.,title="fig:",width=201 ] = 1
. * middle : * 2d model of the point source emission , convolved with the psf . *
right : * residuals after subtracting the model from the data , smoothed with @xmath17 = 1 .
all three panels cover the same area .
the cross marks the peak of the x - ray emission and is at the same location in all images .
see section [ section : ptimage ] for details of psf creation , model , and the fitted parameters.,title="fig:",width=201 ] llrcr @xmath122 ( gis ) & 1995 may 29 & 53041000 & gehrels & 53.1 + @xmath123 ( acis - s ) & 2001 july 8 & 1037 & garmire & 17.8 + @xmath123 ( acis - i ) & 2010 january 16 & 11092 & safi - harb & 69.3 [ table : observations ] lcccccc # observed counts ( [email protected] kev ) & 321 & 822 & 4390 & 4400 & 13 677 & 22473 + # counts after bkgd subtraction & 202 & 578 & 3195 & 2439 & 3767 & 9356 + minimum # counts per bin & 5 & 10 & 50 & 50 & 50 & 50 + region area ( arcsec@xmath124 ) & 12.6 & 301.6 & 6.31e3 & 1.03e4 & 5.23e4 & 6.92e4 + @xmath2 ( 10@xmath4 @xmath5 ) & 1.4 ( frozen ) & 1.4 ( frozen ) & 1.4 ( frozen ) & 1.4 ( frozen ) & 1.4 ( frozen ) & 1.38 ( 1.21 @xmath1 1.57 ) + photon index , @xmath24 & 1.1 ( 0.7 @xmath1 1.6 ) & 1.2 ( 0.9 @xmath1 1.4 ) & 1.57 ( 1.50 @xmath1 1.65 ) & 1.72 ( 1.63 @xmath1 1.82 ) & 1.83 ( 1.69 @xmath1 1.96 ) & 1.68 ( 1.54 @xmath1 1.84 ) + norm . ( photons kev@xmath7
@xmath5 s@xmath7 ) & 5.2 ( [email protected] ) e-6 & 1.6 ( [email protected] ) e-5 & 1.58 ( [email protected] ) e-4 & 1.39 ( [email protected] ) e-4 & 2.3 ( [email protected] ) e-4 & 5.1 ( 4.2 @xmath1 6.3 )
e-4 + @xmath27 & 0.91 & 1.10 & 0.92 & 1.22 & 1.04 & 1.01 + @xmath125 & 55 & 70 & 78 & 67 & 161 & 314 + abs . flux ( erg @xmath5 s@xmath7 , [email protected] kev ) & 5.2e-14 & 1.5e-13 & 7.7e-13 & 5.4e-13 & 7.4e-13 & 2.1e-12 + luminosity ( @xmath30 erg s@xmath7 , [email protected] kev ) & 3.0e32 & 8.7e32 & 5.5e33 & 4.2e33 & 6.3e33 & 1.6e34 + luminosity ( @xmath30 erg s@xmath7 , [email protected] kev ) & 2.5e32 & 6.5e32 & 3.5e33 & 2.5e33 & 3.4e33 & 9.6e33 [ table : spectra ] | pulsar wind nebulae ( pwne ) studies with the _ chandra x - ray observatory _ have opened a new window to address the physics of pulsar winds , zoom on their interaction with their hosting supernova remnant ( snr ) and interstellar medium , and identify their powering engines .
we here present a new 70 ks , plus an archived 18 ks , _ chandra _ acis observation of the snr ctb 87 ( g74.9 + 1.2 ) , classified as a pwn with unusual radio properties and poorly studied in x - rays .
we find that the peak of the x - ray emission is clearly offset from the peak of the radio emission by @xmath0100 and located at the southeastern edge of the radio nebula .
we detect a point source - the putative pulsar - at the peak of the x - ray emission and study its spectrum separately from the pwn .
this new point source , cxou j201609.2 + 371110 , is surrounded by a compact nebula displaying a torus - like structure and possibly a jet .
a more extended diffuse nebula is offset from the radio nebula , extending from the point source to the northwest for @xmath0250 .
the spectra of the point source , compact nebula and extended diffuse nebula are all well described by a power law model with a photon index of 1.1 ( [email protected] ) , 1.2 ( [email protected] ) and 1.7 ( [email protected] ) , respectively , for a column density @xmath2 = 1.38 ( [email protected] ) @xmath3 10@xmath4 @xmath5 ( 90% confidence ) .
the total x - ray luminosity of the source is @xmath01.6 @xmath3 10@xmath6 erg s@xmath7 at an assumed distance of 6.1 kpc , with @xmath0 2% and 6% contribution from the point source and compact nebula , respectively .
the observed properties suggest that ctb 87 is an evolved ( @xmath05@xmath128 kyr ) pwn , with the extended radio emission likely a ` relic ' pwn , as in vela - x and [email protected] . to date , however , there is no evidence for thermal x - ray emission from this snr , and the snr shell is still missing , suggesting expansion into a low - density medium ( @xmath8 @xmath9 ) , likely caused by a stellar wind bubble blown by the progenitor star . |
a standard gaussian random variable @xmath0 is very unlikely to be large : @xmath1 ( cf .
* lemma a.3 ) ) , but it is relatively likely to be small : by direct computation , @xmath2 in fact , for every finite collection of iid standard gaussian random variables @xmath3 , + analysis of the density of the @xmath4 distribution shows that @xmath5 where @xmath6 is the gamma function .
similarly , for _ finitely _ many gaussian random variables , the asymptotic of @xmath7 is always algebraic in @xmath8 , as @xmath9 . on the other hand , for a standard @xmath10-dimensional
brownian motion @xmath11 , @xmath12 , @xmath13 that is , @xmath14 cf .
* theorem 6.3 and corollary 3.1 ) or corollary [ cor : nbm ] below .
transition from a finite to an infinite number of gaussian random variables typically leads to a qualitative change of behavior of small ball ( or lower tail ) probabilities : if @xmath15 then , as @xmath16 , the decay of @xmath17 is usually faster that polynomial in @xmath18 the logarithmic asymptotic is rather robust : if @xmath19 is the solution of the linear equation @xmath20 with a positive - definite matrix @xmath21 , then @xmath22 regardless of the matrix @xmath23 ; cf .
* theorem 4.5 ) . in one - dimensional case , continues to hold even with some time - dependent drifts @xcite . a possible infinite - dimensional generalization of is the stochastic wave equation @xmath24 where @xmath25 is a two - parameter brownian sheet and @xmath26 is the corresponding space - time gaussian white noise .
indeed , a change of variables reduces to @xmath27 with a different brownian sheet @xmath28 ; cf .
* theorem 3.1 ) .
equation can thus be considered an extension of to two independent variables in the spirit of ( * ? ? ?
* section 7.4.2 ) ; according to @xcite , the small ball probabilities for @xmath29 and @xmath30 have similar asymtotics .
so far , the paper @xcite appears to be the only work addressing the question of small ball probabilities for stochastic partial differential equations .
the objective of the current paper is to investigate the asymptotic behavior of @xmath31 @xmath16 , where @xmath29 is the solution of the stochastic parabolic equation @xmath32 @xmath33 is a positive self - adjoint elliptic operator on a bounded domain @xmath34 , @xmath35 is space - time gaussian white noise , and @xmath36 is the scale of sobolev space generated by @xmath33 .
an expansion of the solution of in eigenfunctions of @xmath33 leads to an infinite system of ordinary differential equations , making an infinite - dimensional version of .
the results can be summarized as follows : for both @xmath37 and @xmath38 , as @xmath9 , @xmath39 where @xmath40 , @xmath41 , and @xmath42 are suitable numbers .
for example , @xmath43 and @xmath44 .
in particular , if @xmath45 , then the result is very similar to the finite - dimensional case .
the details are below in theorem [ prop : cbm - sb ] ( @xmath38 ) and theorem [ th : main ] @xmath46 . throughout the paper , for @xmath47 , @xmath48 ,
the notation @xmath49 means @xmath50 @xmath51 means @xmath52 and @xmath53 means @xmath54 .
the variable @xmath55 can be discrete or continuous and the limiting value @xmath56 finite or infinite .
we also fix @xmath57 , a stochastic basis satisfying the usual assumptions .
let @xmath58 be independent identically distributed standard gaussian random variables and let @xmath59 be positive real numbers such that @xmath60 . by direct computation , @xmath61 and then tauberian theorems make it possible to connect the asymptotic of the right - hand side of as @xmath62 with the asymptotic of @xmath63 as @xmath9 .
the most general result in this direction was obtained in @xcite : @xmath64 the function @xmath65 is defined implicitly by the relation @xmath66 and this implicit dependence on @xmath8 is the main drawback of in concrete applications .
less precise but more explicit bounds are possible using exponential tauberian theorems , such as theorems [ thm : tauberian ] and [ thm : tauberian1 ] below ; they are modifications of ( * ? ? ?
* theorem 3.5 ) ( which , in turn , is a modification of ( * ? ? ? * theorem 4.12.9 ) ) .
[ thm : tauberian ] let @xmath67 be a non - negative random variable .
then @xmath68 holds if and only if @xmath69 while is only _ logarithmic _ asymptotic of the probability and is not as strong as , it is usually more explicit than and is good enough in many applications .
when holds , we say that the random variable @xmath67 has the small ball rate @xmath70 and the small ball constant @xmath71 occasionally , a more refined version of theorem [ thm : tauberian ] is necessary .
[ thm : tauberian1 ] let @xmath67 be a non - negative random variable
. then @xmath72 holds if and only if @xmath73 let @xmath74 be the solution of the equation @xmath75 with @xmath76 and @xmath77 , and assume that the initial condition @xmath78 is independent of the brownian motion @xmath79 and is a gaussian random variable with mean @xmath80 and variance @xmath81 .
then @xmath82 where @xmath83 this follows by direct computation using ( * ? ? ?
* theorem 3 ) ; see also ( * ? ? ? * lemma 17.3 ) when @xmath84 . for the standard brownian motion , with @xmath85 , and @xmath86 ,
equality becomes the well - known cameron - martin formula : @xmath87 as an illustration of theorem [ thm : tauberian ] , let us confirm .
[ cor : nbm ] if @xmath11 , @xmath88 is an @xmath10-dimensional standard brownian motion , then @xmath89 by and independence of the components of @xmath90 , @xmath91 then follows from theorem [ thm : tauberian ] with @xmath92 and @xmath93 .
let @xmath33 be a positive - definite self - adjoint elliptic operator of order @xmath94 on a bounded domain @xmath34 with sufficiently smooth boundary ; alternatively , @xmath95 can be a smooth closed @xmath96-dimensional manifold with smooth measure @xmath97 .
denote by @xmath98 the eigenvalues of @xmath33 , and by @xmath99 the corresponding normalized eigenfunctions .
our main assumption is that the weyl - type asymptotic holds for @xmath100 : @xmath101 with constant @xmath102 depending only on the region @xmath95 ; see ( * ? ? ?
* theorem 1.2.1 ) .
for example , if @xmath103 on @xmath104 with zero boundary conditions , and @xmath105 is the lebesgue measure of @xmath95 , then @xmath106 and @xmath107 for @xmath108 [ that is , @xmath109 is a smooth compactly supported real - valued function on @xmath95 ] and @xmath111 , define @xmath112 then define the space @xmath113 as the closure of @xmath114 with respect to the norm @xmath115 . in particular ,
@xmath116 we also define @xmath117 and identify @xmath118 with a possibly divergent series @xmath119 note that @xmath120 for all @xmath121 .
a cylindrical brownian motion on @xmath122 is a gaussian process @xmath123 , indexed by @xmath124 $ ] and @xmath125 , such that @xmath126 [ prop : wn ] define @xmath127 and @xmath128 then , for @xmath129 , @xmath130 and @xmath131 definition of @xmath132 implies that @xmath133 are independent standard brownian motions . then both and follow by direct computation .
in particular , by , @xmath134 if and only if @xmath135 .
moreover , @xmath136 so that extends to @xmath125 , and , by kolmogorov s criterion , @xmath30 has a modification in @xmath137 , @xmath129 .
we then get the following analogue of .
[ prop : cbm - sb ] as @xmath9 , @xmath138 where @xmath139 by and , @xmath140 and by , @xmath141 if @xmath142 , then the series @xmath143 converges , and the dominated convergence theorem implies @xmath144 or @xmath145 so that the first relation in follows from theorem [ thm : tauberian ] .
if @xmath146 , then we establish the asymptotic of by comparison with a suitable integral .
note that @xmath147 and the function @xmath148 satisfies @xmath149 with @xmath150 .
let @xmath151 . by direct computation , as @xmath62 , @xmath152 after that , theorems [ thm : tauberian1 ] and [ thm : tauberian ] imply the remaining relations in .
we now use the process @xmath30 to construct an infinite - dimensional analogue of .
given @xmath153 , consider the equation @xmath154 for example , with @xmath155,\\ & { \mathrm{a}}=-\partial^2/\partial x^2 , \
{ \rm \ zero \ boundary \ conditions},\\ & \lambda_k = k^2,\ \varphi_k(x)=(2/\pi)^{1/2 } \sin(kx ) . \end{split}\ ] ] and @xmath156 , equation becomes @xmath157 the solution of the equation is a mapping from @xmath158 $ ] to @xmath159 with the following properties : 1 .
there exists a @xmath111 such that @xmath160 2 . for every @xmath161 , the process @xmath162 , @xmath12 ,
is @xmath163-adapted .
3 . for every @xmath164 ,
the equality @xmath165 holds in @xmath166 [ prop : he ] equation has a unique solution @xmath167 .
moreover , for every @xmath168 , @xmath169 the result can be derived from general existence and uniqueness theorems for stochastic evolution equations , such as ( * ? ? ?
* theorem 5.4 ) or ( * ? ? ?
* theorem 3.1.1 ) ; below is an outline of a direct proof . taking @xmath170 in we find @xmath171 that is , @xmath172 then @xmath173 and @xmath174 by , @xmath175 if @xmath129 , then @xmath176 , and follows .
similarly , @xmath177 and then kolmogorov s criterion implies that @xmath29 has a modification in @xmath178 . to establish uniqueness , note that the difference @xmath179 of two solutions satisfies the deterministic equation @xmath180 with zero initial condition .
[ th : main ] as @xmath9 , @xmath181 where @xmath182 define @xmath183 by and , @xmath184 where @xmath185 the goal is to show that , as @xmath186 , @xmath187 and @xmath188 where @xmath189 after that , relations will follow from theorems [ thm : tauberian ] and [ thm : tauberian1 ] .
we start by establishing .
we then show that @xmath190 and @xmath191 are of lower order compared to @xmath192 : @xmath193 note that @xmath194 and recall that holds . if @xmath142 , then @xmath195 since , for every @xmath121 , @xmath196 the dominated convergence theorem implies @xmath197 therefore , for @xmath142 , @xmath198 when @xmath199 , define the function @xmath200 by , @xmath201 whereas implies @xmath202 with @xmath203 , uniformly in @xmath204 .
define @xmath205 we will now show that @xmath206 to begin , let us establish the asymptotic of @xmath207 . with the notations @xmath208 @xmath209 if @xmath210 , then implies @xmath211 by lhospital s rule , for every @xmath212 @xmath213 therefore , as @xmath186 , @xmath214 if @xmath215 , then implies @xmath216 in particular , @xmath217 to establish , write @xmath218 where @xmath219 then will follow from @xmath220 we have @xmath221 because , for fixed @xmath204 , @xmath222 , @xmath223 , and the function @xmath109 has at most one critical point .
if @xmath224 , then @xmath225 , and follows from .
if @xmath226 [ which is possible when @xmath227 , then @xmath228 [ by balancing @xmath229 and @xmath230 , so that , with @xmath231 , @xmath232 to get a bound on @xmath233 , note that @xmath234 where @xmath235 is a suitable constant independent of @xmath204 . together with , inequality implies @xmath236 and the constant @xmath237 does not depend on @xmath204 . by integral comparison , @xmath238 and ,
similar to the derivation of , @xmath239 which implies .
the asymptotic of @xmath192 is now proved ; a more compact form of is @xmath240 it remains to establish .
recall that @xmath241 cf .
, , and . by definition , @xmath242 , which means @xmath243 because the function @xmath244 is decreasing for @xmath245 $ ] and @xmath246 .
next , @xmath247 and therefore @xmath248 because @xmath249 using , inequality becomes @xmath250 where @xmath251 @xmath252 and @xmath253 is a suitable constant independent of @xmath204 .
then , by integral comparison , @xmath254 finally , by direct computation , @xmath255 note that @xmath256 if @xmath257 . comparing with
, we see that @xmath258 for all @xmath257 . to show that @xmath259 , note that and inequality @xmath260 imply @xmath261 _ this completes the proof of theorem [ th : main ] .
_ comparing and , we see that , for @xmath142 , the small ball behavior of both @xmath30 and @xmath29 is the same and , in a certain sense , similar to the finite - dimensional case . for @xmath215 ,
the small ball rate for @xmath30 is bigger than the small ball rate for @xmath29 ; in fact , in this range of @xmath262 , the small ball rate is a decreasing function of @xmath263 . at @xmath210 , a kind of a phase transition takes place . for equation
, we have @xmath264 then @xmath265 where @xmath266 similarly , @xmath267 where @xmath268 in particular , taking @xmath269 , we get a rather explicit logarithmic asymptotic @xmath270 where @xmath271 figure 1 presents the small ball rates @xmath42 when @xmath272 for the solution of ( bold curve ) and for the underlying noise @xmath30 .
[ fig1 ] $ ] : small ball rate in @xmath273 , @xmath272 , for the solution ( bold curve ) and the noise.,title="fig : " ]
if @xmath274 , is a sequence of real numbers such that @xmath275 for some @xmath276 and all @xmath277 , then , by , @xmath278 defines an equivalent norm on @xmath113 .
when @xmath142 , we get immediate analogues of theorems [ prop : cbm - sb ] and [ th : main ] . let @xmath279 denote either @xmath29 or @xmath30 . for @xmath142 , @xmath280 direct computations show that @xmath281 either by [ when @xmath37 ] or by [ when @xmath38 ] , and then follows from theorem [ thm : tauberian ] . when @xmath284 , analogues of theorems [ prop : cbm - sb ] and [ th : main ] exist under an additional assumption about the numbers @xmath285 .
[ prop : eqn ] assume that @xmath286 then , as @xmath16 , @xmath287 where @xmath288 for the solution @xmath167 of equation , @xmath289 where @xmath290 by and , @xmath291 to derive , it remains to replace @xmath102 in with @xmath292 . after a slightly more detailed analysis ,
follows from in a similar way .
note that if @xmath293 , then becomes . in the special case @xmath294 ,
an alternative proof of proposition [ prop : eqn ] is possible using the results from ( * ? ? ?
* example 2 ) ; for technical reasons , such a proof is usually not possible under the general assumption . without ( that is , assuming only ) , a precise logarithmic asymptotic of the small ball probabilities may not exist when @xmath284 , but the corresponding upper and lower bounds can still be derived .
a generalization of @xmath132 is @xmath295-cylindrical brownian motion @xmath296 defined by @xmath297 where @xmath295 is a non - negative symmetric operator on @xmath122 . if @xmath298 then , similar to , we set @xmath299 and consider the equation @xmath300 if furthermore @xmath301 then the question about the asymptotic of @xmath302 @xmath9 , is reduced to the corresponding question for the solution of the equation @xmath303 where @xmath304 .
for example , if @xmath305 , that is , @xmath306 , @xmath307 , then @xmath308 assume that the initial condition @xmath309 in is independent of @xmath30 and has the form @xmath310 where @xmath311 are independent gaussian random variables with mean @xmath312 and variance @xmath313 .
to ensure , condition @xmath314 must hold for all @xmath315 .
in the finite - dimensional case , it is known ( * ? ? ?
* theorem 4.5 ) that the initial condition may affect the small ball constant but not the small ball rate : if @xmath316 is a gaussian random vector independent of @xmath90 , then @xmath317 where @xmath40 may depend on the mean and covariance of @xmath316 . in particular , if the covariance matrix of @xmath316 is non - singular , then @xmath318 , that is , the initial condition does not change the small ball asymptotic at the logarithmic level .
the corresponding results in the infinite - dimensional case are as follows .
[ prop : inf - dim - ic ] assume that holds . 1 .
if @xmath319 then holds .
2 . if @xmath142 and @xmath320 for all @xmath121 , then @xmath321 the idea is to trace the contributions of the initial condition throughout the proof of theorem [ th : main ] . in particular , our objective is the asymptotic of @xmath322 as @xmath186 .
by , the initial condition contributes an extra multiplicative term @xmath323 where @xmath324 also recall that , with zero initial condition , the dominant term is @xmath325 under , @xmath326 similarly , @xmath327 in other words , if the variance of the initial condition is strictly non - degenerate , then the initial condition does not affect the asymptotic of @xmath322 as @xmath186 .
if @xmath142 and @xmath320 , then implies @xmath328 and , similar to the proof of , @xmath329 after that , theorem [ thm : tauberian ] implies .
the stationary case requires special consideration ; cf .
@xcite for one - dimensional ou process .
assume that @xmath330 and @xmath331 . then holds .
even though direct application of proposition [ prop : inf - dim - ic](1 ) is not possible because now @xmath332 and fails , very little changes in the actual proof : with only the @xmath333 term present , we see that @xmath334 still holds , that is , the initial condition does not affect the asymptotic of @xmath322 .
if @xmath284 , then initial condition can affect the small ball rate .
for example , assume that the initial condition in is non - random [ @xmath320 ] and @xmath336 , so that holds .
using , , and , @xmath337 similar to derivation of , @xmath338 that is , @xmath339 with @xmath340 from . by theorem [ thm : tauberian1 ] with @xmath341 , @xmath342 , @xmath343 , @xmath344 which is very different from : the rate has an additional logarithmic term and the constant does not depend on @xmath345 .
consider a linear operator @xmath33 on a separable hilbert space @xmath346 . if @xmath33 is symmetric and has a pure point spectrum @xmath347 and the corresponding eigenfunctions @xmath348 form an orthonormal basis in @xmath346 , then all the constructions from section [ sec : dspe ] can be repeated , and an analog of theorem [ th : main ] can be stated and proved for the evolution equation @xmath349 where @xmath35 is a cylindrical brownian motion on @xmath346 .
the details depend on the asymptotic behavior of @xmath100 as @xmath350 .
for example , consider the equation @xmath351 then @xmath352 and @xmath353 cf .
* section 1.4 ) . since operator @xmath33 has order 2 , we define @xmath354 recall that the norm in the traditional sobolev space on @xmath355 is @xmath356\!|_{\gamma}^2=\int_{-\infty}^{+\infty } fourier transform of @xmath109 .
in particular , it follows that @xmath357\!|_{\gamma}^2=\infty\ ] ] for every @xmath111 and every @xmath358 [ roughly speaking , because @xmath359 and each @xmath348 is an eigenfunction of the fourier transform ] , and consequently the solution of @xmath360 , @xmath361 , does not belong to any traditional sobolev space on @xmath355 . on the other hand ,
similar to propositions [ prop : wn ] and [ prop : he ] , @xmath362 @xmath29 is the solution of .
the corresponding small ball asymptotics can also be derived .
the following relations hold as @xmath9 : @xmath363 where @xmath364 and @xmath365 where @xmath366 the case of @xmath30 follows from the asymptotic of @xmath367 similar to the proof of theorem [ prop : cbm - sb ] .
the case of @xmath29 follows the same steps as the proof of theorem [ th : main ] .
consider the equation @xmath368 with @xmath153 , and assume that the positive - definite operators @xmath33 and @xmath295 commute , have purely point spectrum and act in a scale @xmath113 of hilbert spaces , and @xmath369 where @xmath370 is the embedding operator .
if @xmath371 , then the logarithmic asymptotic of the small ball probabilities is similar to finite - dimensional case : @xmath372 for both @xmath37 and @xmath373 .
infinite - dimensional effects appear when @xmath374 : the small ball rate now depends on @xmath262 and can be arbitrarily large , whereas the small ball constant depends on the operator @xmath33 .
w. v. li and q .- m .
gaussian processes : inequalities , small ball probabilities and applications . in d.
n. shanbhag and c. r. rao , editors , _
stochastic processes : theory and methods _ , volume 19 of _ handbook of statist .
_ , pages 533597 .
north - holland , amsterdam , 2001 .
j. b. walsh .
an introduction to stochastic partial differential equations . in p.
l. hennequin , editor , _
ecole dt de probabilits de saint - flour , xiv , lecture notes in mathematics _ , volume 1180 , pages 265439 , springer , 1984 . | while small ball , or lower tail , asymptotic for gaussian measures generated by solutions of stochastic ordinary differential equations is relatively well understood , a lot less is known in the case of stochastic partial differential equations .
the paper presents exact logarithmic asymptotics of the small ball probabilities in a scale of sobolev spaces when the gaussian measure is generated by the solution of a diagonalizable stochastic parabolic equation .
compared to the finite - dimensional case , new effects appear in a certain range of the sobolev exponents . |
the free electron lasers ( fels ) now beginning to come online produce radiation many orders of magnitude brighter than than any existing source , and enable experiments previously the domain only of science fiction .
one such proposed experiment @xcite envisages reconstructing the 3d structure of a microscopic entity such as a virus from many ultrashort diffraction patterns of many identical copies of the particles in random orientations from single pulses of fel radiation .
although the particles will undoubtedly suffer catastrophic radiation damage , the ultrashort nature of fel radiation is expected to produce diffraction patterns of the particles before significant disintegration .
an experiment on individual mimivirus particles was reported recently @xcite .
the paper illustrates convincing diffraction patterns of the virus particle in two different orientations , from which 2d projections of the particles are reconstructed using an iterative phasing algorithm .
although such particles are known to be largely icosahedral , little evidence of the icosahedral shape is evident in the reconstructed projections .
several algorithms have been proposed for reconstructing a full 3d image of the particle from an ensemble of many such diffraction patterns from randomly oriented particles .
the methodology followed by some of these approaches @xcite is to find the likely orientation of the measured diffraction patterns in the 3d reciprocal space of the particle .
another approach @xcite dispenses with finding the likely orientations of the individual diffraction patterns by integrating over orientations , in an attempt to find the spherical harmonic representation of the 3d diffraction volume of a single particle from the averages of the angular correlations of the intensities on the measured diffraction patterns .
this method of analysis is even applicable to individual diffraction patterns from multiple identical particles @xcite .
the particles need to be frozen in space or time while the scattering is taking place .
if the scattering is from a single fel pulse of radiation , the particles will be essentially frozen in time for the duration of the scattering even if not frozen in space .
this opens this method to the analysis of scattering from particles in random orientations within a droplet .
with such an approach , the `` hit rate '' in an experiment with a fel can become 100% , whereas a low hit rate is to be expected when attempting to hit submicron particles with a submicron pulsed laser beam .
the signal - to - noise ratio from such snapshot patterns is independent of the number of particles per shot , but increases with the square root of the number of shots @xcite .
this approach also has the advantage that it operates on a compressed version of the voluminous data produced by a fel .
we point out here another advantage of this approach : it is easily amenable to simplifications resulting from any known point - group symmetry of the particles under study .
this is a powerful advantage for the study of virus structure , which is dominated by that of its protein coat which encloses the genetic material , dna or rna , which contain the instructions for the replication of the virus . to quote from caspar and klug @xcite
`` there are only a limited number of efficient designs possible for a biological container which can be constructed from a large number of identical protein molecules , the two basic designs are helical tubes and icosahedral shells '' .
viruses have regular shapes since they are formed by the self assembly of identical protein subunits which are coded by the limited quantity of genetic material capable of being stored within the small volume enclosed by its protein coat .
an icosahedron , for example can be formed by the self assembly of at least 60 identical subunits .
the genetic material needs to code for just one of these subunits , a factor of at least 60 smaller than the entire structure .
the first aim of this approach is to find the spherical harmonic representation of the intensity distribution of any resolution shell in the reciprocal space of a single particle .
any prior information about the nature of this distribution may be incorporated by limiting the set of spherical harmonics over which the summation is performed and by any relationship amongst the amplitudes of the different spherical harmonics which are a consequence of any known point - group symmetry .
an obvious restriction of the form of the intensity distribution @xmath0 is its known inversion ( or friedel ) symmetry . since @xmath1 it follows that a spherical harmonic expansion of an intensity distribution may contain only even values of the angular momentum quantum number @xmath2 .
the fact that the intensity distribution is real , allows the restriction to a summation over just the so - called real spherical harmonics ( rshs ) @xmath3 defined by the combinations of spherical harmonics : @xmath4 & m>0 \\
y_{l}^{0}(\theta,\phi ) & m=0 \\ \frac{1}{i\sqrt{2 } } \left [ y_l^m(\theta,\phi ) - ( -1)^m y_l^{-m}(\theta,\phi ) \right ] & m < 0 \end{array } \right.\ ] ] where the set of rsh s with @xmath5 form a set , whose @xmath6 dependence is of the form @xmath7 , and the set with @xmath8 likewise a set with @xmath6 dependence of the form @xmath9 . if the reconstructed intensity distribution has a mirror plane , this may be chosen to be the @xmath10 plane , or the plane for which @xmath11 .
then ( [ genexp ] ) may be replaced by a summation over only the subset of rshs for which @xmath5 , and we may take @xmath12 since both the right hand side ( rhs ) and the left hand side ( lhs ) of the above equation are real , the coefficients @xmath13 may also be taken as real . 0,1 , and 2 .
the plots are made with matlab by an adaptation of software by denise l. chan ( avilable from the mathworks web site).,width=529 ] the 3d polar plots of fig .
[ sphrm ] display the familiar forms of the rshs for the values of @xmath14 0,1,2 .
further point group symmetries of @xmath15 result in still further restrictions on allowed terms of the general expansion ( [ genexp ] ) above .
when the 3d intensity distribution has icosahedral symmetry , for example , ( [ genexp ] ) may be replaced by @xmath16 where the quantities @xmath17 are known as icosahedral harmonics ( ihs ) , specified up to and including @xmath1430 by only the angular momentum quantum number @xmath2 . since the orientation of our reconstructed 3d intensity distribution in the frame of reference of the particle may be chosen arbitrarily , the @xmath10 plane may be chosen to be the mirror plane , allowing the ihs ( [ icoexp ] ) to be constructed from just the rshs of positive @xmath18 , i.e. @xmath19 where the coefficients @xmath20 are the real numbers for normalized rshs tabulated by e.g. jack and harrison ( 1975 ) @xcite , the ones for the lowest allowed even values of @xmath2 being reproduced in table [ table1 ] . since the ihs involve a sum over the magnetic quantum number , at least up to @xmath21 , they depend on the quantum number @xmath2 only
. the forms of the icosahedral harmonics of lowest even degree , @xmath2=0,6,10,12 , and 16 are illustrated in fig .
[ icoharm ] using the same 3d polar plots .
.expansion coefficients of the lowest even degree icosahedral harmonics with z - axis chosen to be the 5-fold rotation axis . for the list up to @xmath2=30 , see e.g. @xcite .
note the rows are characterized by @xmath2 and the columns by @xmath18 . [ cols="^,^,^,^,^,^",options="header " , ] [ table1 ] 0,6,10,12 and 16 .
each is a linear combination of the rsh s of the magnetic quantum numbers indicated .
visualization by the same software as fig .
[ sphrm],width=453 ] note that since the rsh s @xmath22 are orthonormal with respect to integrations over spherical shell , the icosahedral harmonics @xmath17 will also be orthonormal with respect to the same integration provided @xmath23 this condition is clearly satisfied by the coefficients in table 1 .
the average angular correlations amongst the resolution rings of the different measured diffraction patterns contain information about the 3d diffraction volume of a single particle .
such angular correlations are defined by @xmath24 where @xmath25 is the intensity on diffraction pattern @xmath26 , @xmath27 is the number of available diffraction patterns from random orientations of the particle , @xmath28 is the @xmath29-th of @xmath30 discrete values of @xmath6 , and @xmath31 the angular fourier tranform of each resolution ring of the @xmath26-th diffraction pattern . indeed the fastest way to calculate the average angular correlation @xmath32 is to exploit the cross - correlation theorem by performing the angular fourier transform ( [ eyem ] ) of each individual diffraction pattern , take the product of the fourier transform and its complex conjugate followed by the inverse transform , and to average the results over the diffraction patterns ( ( [ eyem ] ) plus the second equality of ( [ c2 ] ) ) .
it has been shown @xcite that if the data from enough diffraction patterns of randomly oriented identical particles are averaged , @xmath33 where @xmath34\ ] ] where @xmath35 is a legendre polynomial of order @xmath2 , @xmath36 @xmath37 is the wavenumber of the incident beam , and @xmath38 since in eq.([bl_eqn ] ) , the lhs may be found from experiment , and @xmath39 is a known mathematical function , @xmath40 may be found by solving this equation . due to its form , @xmath40 contains information about the 3d diffraction volume of the particle via the spherical harmonic expansion coefficients @xmath41 . because the rshs are related to the regular spherical harmonics by a unitary transformation ,
one may also write @xmath42 this is a more convenient form since all quantities in this equation are real .
if the @xmath13 coefficients may be found from deduced values of @xmath40 , the expression ( [ realexp ] ) is just as convenient for reconstructing the 3d diffraction volume as ( [ genexp ] ) .
however , finding the correct @xmath13 s from known @xmath40 s is still a formidable task since it involves taking a matrix square root @xcite .
such a square root is necessarily ambiguous by an orthogonal matrix which can not be found from the @xmath40 s alone . in principle
, such an orthogonal matrix may be found from the so - called angular triple correlations @xcite or by an iterative phasing algorithm that alternately satisfies constraints to the meaured angular correlations and in the 3d space of the reconstructed intensity disrtibution @xcite . when the particle under study is known to have a high degree of symmetry , like an icosahedral virus , this problem is greatly simplified . comparing ( [ realexp ] ) and ( [ icoexp ] ) with definition ( [ icodef ] ) , one can deduce that , for a diffraction volume with icosahedral symmetry , @xmath43 substituting ( [ icos ] ) into ( [ blr ] ) we see that one may write @xmath44 and using ( [ norm ] ) this may be simplified further to @xmath45 we see here the great advantage of using ih s rather than rshs for this problem of icosahedral symmetry .
the sum over @xmath18 in the rhs of ( [ blr ] ) has disappeared completely in the rhs of ( [ bg ] ) !
the rhs of this equation is just the product of two scalars .
a diffraction volume of icosahedral symmetry may be reconstructed via ( [ icoexp ] ) if the coefficients @xmath46 are known .
since the other quantities in ( [ icos ] ) are real , it is clear that @xmath46 may be chosen to be real .
the magnitudes of the @xmath46 coefficients may be found from the diagonal quantities @xmath47 deduced from the intensity autocorrelations on resolution ring @xmath48 via @xmath49 thus , the only remaining task in determining the coefficients @xmath46 is determing the signs of these real numbers .
a simple way is to notice that the expression ( [ icoexp ] ) for the intensity of a resolution shell of radius @xmath48 in the 3d diffraction volume may be rewritten @xmath50 j_l(\theta,\phi ) \label{ico2}\ ] ] where the only unknown quantities in the rhs are the signs of @xmath46 . since the only permitted values of the quantum number @xmath2 of the icosahedral harmonic coefficients @xmath46 of a diffraction volume are the even permitted values up to @xmath2=30 , namely @xmath2=0 , 6 , 10 , 12 , 16 , 18 , 20 , 22 , 24 , 26 , 28 , and 30 , we attempted to determine these signs by an exhaustive search over the 2@xmath51 combinations of signs by finding the combination that minimized @xmath52 where @xmath53 are the negative values of @xmath54 , for a chosen resolutiuon shell @xmath48 .
the physical basis of this is simply that @xmath15 has to be a positive definite quantity , and our best approximation to this is a function with a minimum sum of the magnitudes of negative values . as subsequent results show , this easily implemented prescription seemed accurate enough to find a good enough approximation to the correct signs of these coefficients for the chosen reference resolution ring . to maximixe the number of non - negligible magnitudes @xmath55 we chose a high - resolution resolution ring . in order to avoid almost all values @xmath55 being very small , and thus subject to significant rounding - off errors
, we found the best compromise to choose the reference ring to be one for which @xmath56 , where @xmath57 is the value of @xmath48 for the outermost resolution shell . from eq .
( [ bg ] ) , we see that the icosahedral harmonic expansion coefficients of the same quantum number @xmath2 , corresponding to a different resolution shell @xmath58 are related to the now known ones of resolution shell @xmath48 by the simple quotient @xmath59 thus , having found the coefficients @xmath46 for a paticular shell @xmath48 , those of the other shells @xmath58 were determined from this simple quotient , involving the quantities @xmath40 directly calculable from the average intensity cross correlations between different resolution rings on the measured diffraction patterns .
thus the exhaustive search though all 2@xmath60 combination of signs needs to be performed only for a single resolution ring @xmath48 . a knowledge of the expansion coefficients for all the resolution shells should enable a reconstruction of the 3d diffraction volume via ( [ icoexp ] ) .
if this intensity distribution is interpolated onto an oversampled @xcite 3d cartesian reciprocal - space grid , @xmath61 , say , an iterative phasing algorithm @xcite may be applied to reconstruct the 3d electron density of the scattering particle .
a central thesis of this paper is that the the scattered intensity from an icosahedral particle may be represented by a sum of icosahedral harmonics .
we first sought to verify this proposition by calculating first the spherical harmonic expansion coefficients of a simple icosahedral particle via the expresion @xmath62 where @xmath63 is the form factor of the jth atom , @xmath64 is its coordinate , and @xmath65 is a spherical bessel function of order @xmath2 . for our initial tests we simulated the scattering from an artificial molecule of identical atoms at the vertices of a regular icosahedron ( fig .
[ regico ] ) of edge length 2 @xcite ( which we take to be to be in units , with cartesian coordinates ( also assumed to be in ) : + + ( 0 , @xmath661 , @xmath67 ) + ( @xmath66 1 , @xmath67 , 0 ) + ( @xmath67 , 0 , @xmath66 1 ) + + where @xmath68 is the golden ratio ( 1+@xmath69)/2 . coefficients ( arbitrarily taking @xmath70 @xmath71 ) assuming 12 identical atoms at the vertices of a regular icosahedron .
the first two entries in each column in each line are the @xmath2 and @xmath18 values .
the next two are the real and imaginary parts of @xmath72 .
it will be seen that all coefficients are zero except those for which @xmath2=0 or 6 , and that all non - zero coefficients are real.,width=264 ] the resulting calculated values of the amplitudes @xmath73 ( arbitrarily taking @xmath74 ) for all possible values of of @xmath2 and @xmath18 are listed in fig .
[ alm_coeff ] .
note that the amplitudes @xmath73 are all real , and that , for the values listed , they are non - zero only for @xmath2=0 and 6 . the values for @xmath2=1,2,3,4,5 , and 7 are all seen to be zero , corresponding to non - existing icosahedral harmonics for these values of @xmath2
. here too , all coefficients are zero except those for which @xmath2=0 or 6 , and all non - zero coefficients are real .
note that this result will be true for any icosahedral orientation since the a rotation matrix ( wigner d - matrix ) will mix only amplitudes of different magnetic quantum number corresponding to the same angular momentum @xmath2 .
the z - axis of the simple icosahedron used for this test is a 2-fold rotation axis , not 5-fold , unlike e.g. fig .
[ icoharm ] above , or table 1 below .
this is why the amplitudes corresponding to every other value of @xmath18 are non zero for @xmath2=6 rather than every integer multiple of 5 , when @xmath75 is chosen to be a 5-fold axis . , except for values of the @xmath76 coefficients calculated by eq .
( [ clebsch ] ) from the @xmath72 values in fig .
[ alm_coeff].,width=264 ] of greater interest for our method are the allowed values of @xmath77 for the coefficients , @xmath78 , of the spherical harmonic expansions of the scattered intensity . since @xmath79
it must follow that if @xmath80 has icosahedral symmetry , so must @xmath81 .
however , this is not entirely obvious from the relationship between the two sets of coefficients @xmath82 where @xmath83 is a clebsch - gordan coefficient @xcite . according to the usual theory of the vector addition of angular momenta , the allowed values of @xmath77 are all integers in the range from @xmath84 to @xmath85 , with no obvious indication that @xmath77=1,2,3,4,5 , and 7 , for instance , are forbidden .
however , a straightforward evaluation of the @xmath76 coefficients via ( [ clebsch ] ) reveals this to be the case , as is seen by the tabulated values of these coefficients in fig .
[ ilm_coeff ] . .
[ stnv ] we next tested this on a realistic model of the small icosahedral virus , satellite tobacco necrosis virus ( stnv ) whose atomic coordinates are deposited in the protein data bank under entry 2buk ( fig .
[ stnv ] ) .
we calculated @xmath80 from the usual structure factor expression @xmath86 and constructed the diffraction volume from ( [ eye ] ) . by integrating over spherical shells of @xmath87 we evaluated the spherical harmonic expansion coefficients of the 3d diffraction volume of stnv from @xmath88 where @xmath89 is the unit vector @xmath90 , with this integration conveniently performed by gaussian qudrature @xcite .
plots of the real and imaginary parts of @xmath91 in fig .
[ ilm ] clearly show the same trend of vanishing components corresponding to @xmath92 and 7 and in addition vanishing components for @xmath93 and 15 , exactly consitent with the tablulated values of icosahedral expansion coefficients in table 1 .
what is more , it was found that @xmath94 the precise condition for the reality of the @xmath13 coefficients of the rshs , and hence of the icosahedral harmonic expansion coefficients @xmath46 via ( [ icos ] ) .
coefficients calculated from the computed diffraction volume of stnv .
each dot represents a value of the ( lm ) pair .
note that these coefficients are largely absent for @xmath2=1,2,3,4,5,7,8,9,11,13,14,15.,width=264 ] since this result is a consequence of the icosahedral symmetry of the diffraction volume @xmath81 , it is to be expected of the diffraction volume of all icosahedral viruses ( assuming the protein coat to be the dominant scatterer ) . in view of ( [ bl ] ) this must mean that the @xmath40 coefficients computed from the data of diffraction patterns of random orientations of all icosahedral particles must all have vanishing values for @xmath95 , thus providing a very simple test of whether the diffraction patterns measured in a `` diffract and destroy '' experiment with a fel are from an icosahedral particle . assuming this is indeed found to be approximately true in practice ( even the so - called icosahedral viruses may have appendages which break the icosahedral symmetry of the protein coat , and of course the genetic material inside the protein coat
would not be expected to have this symmetry .
however , if the bulk of the material of the virus may be assumed to constitute the protein coat , this must be approximately the case ) .
the icosahedral structure of the protein coat may be found by an analysis of the large @xmath96 @xmath40 coefficients extactable from the average angular correlations of the diffraction data .
we next attemped a reconstruction of satellite tobacco necrosis virus ( stnv ) from diffraction patterns simulated for directions of incidence on a single particle from a uniform angular distribution in so(3 ) @xcite . for the model of stnv we took the data of the bological assembly of stnv from pdb entry 2buk .
due to the large number of atoms in this biological assembly ( @xmath97 100,000 ) , the most convenient way to do this was to take slices through a precalculated 3d diffraction volume of this structure .
average angular correlations of these simulated diffraction patterns were calculated by the formulae ( [ c2 ] ) and ( [ eyem ] ) . and the @xmath40 coefficients were calculated from these by inverting eq .
( [ bl_eqn ] ) . for the 10,000 simulated diffraction patterns in our test ,
this process took about a quarter of an hour on a single processor on a desktop computer . in a real experiment
, one may have to deal with perhaps 100 times as many diffraction patterns , with more pixels per pattern , so the processing time could be several orders of magnitude greater .
however , the bulk of the time will be spent in generating the average angular correlations @xmath32 ( [ c2 ] ) , a process which easily lends itself to parallelization , since subsets of the diffraction patterns may be averaged by separate computer processors , and the averages themselves subsequently averaged .
nevertheless , this process of reduction of terabytes ( tb ) of measured experimental data is probably the most computer - resource intensive part of our method .
having thus reduced our data to a set of @xmath40 coefficients for a set of 30 values of @xmath2 , and 61 values of @xmath48 ( and @xmath58 ) , we were left with a set of 30@xmath9861@xmath9861 real numbers which formed the input to our reconstruction algorithm .
this required about a mb of storage / memory . in a real experiment also ,
our method requires the million - fold reduction of the tb of data to a mb of floating - point ( real ) numbers that form the input to our reconstruction algorithm .
it is recommended that this data reduction be performed at the site of the data to reduce by a million - fold or so the quantity of data that needs to be transmitted over the internet to the site where the image reconstruction is performed . at current rates ,
the transmission of terabytes of data over the internet could take several weeks , whereas the time for the transmission of a mb of data could be measured in seconds .
in addition this process of data reduction is expected to result in considerable noise - reduction of the raw data though averaging @xcite .
since the @xmath40 coefficients are related to the expansion coefficients @xmath13 of the real spherical harmonics ( which satisfy the same selection rule on @xmath2 as do the expansion coefficients @xmath41 of the regular spherical harmonics ) , it would be expected that the @xmath99 elements of these coefficients are dominant .
this was found to be the case for our simulations of stnv .
some of the larger , predominantly icosahedral , viruses may have appendages like the unique vertex and `` hair '' of the mimivirus @xcite , or the spike fom a unique vertex of the chlorella virus @xcite
. indeed , with values of these coefficients extracted from experimental single - particle diffraction patterns from an unknown particle , the satisfaction of this selection rule would be an excellent test of the degree to which the particle is icosahedral .
inclusion of only the large @xmath96 of the @xmath40 coefficients in the reconstruction algorithm consistent with icosahedral symmetery is equivalent to finding the closest icosahedral approximation to the structure . , @xmath57 , of about 4.7 nm@xmath100 , implying a resolution of @xmath97 1.3 nm .
both the outer and inner surfaces of the virus capsid are apparent in this representation .
a ribbon diagram of the structure in pdb entry 2buk is seen to fit within this capsid.,width=264 ] the procedure described in section 3 was then followed to reconstruct a 3d diffraction volume , consisting of set of scattered intensities @xmath101 over 3d reciprocal space as a function of the reciprocal - space coordinate @xmath102 . in our simulations
, we took this to be a 61@xmath9861@xmath9861 array of real numbers .
the computer time for this process was almost ridiculously short , amounting to no more than a few seconds on a single - processor desktop computer . except that the diffraction volume was reconstucted from the average of angular correlations on 10,000 diffraction patterns of stnv from uniformly distributed directions over so(3 ) .
the reconstructed electron density is seen to be remarkably similar to that in fig .
[ direct].,width=264 ] except that image displayed is a cut perpendicular to the 5-fold axis of the virus .
the 5-fold symmetry of both the external and internal surfaces of the capsid in this projection are clearly visible.,width=264 ] the final step is the recovery of a 3d electron density of the particle . this may be done by a standard iterative phasing algorithm .
we used the `` charge flipping '' algorithm of oszlnyi and sto @xcite . in order to judge the accuracy of the our algorithm in recovering the 3d diffraction volume
, we performed this recovery of the 3d electron density from both the diffraction volume @xmath81 calculated directly from the stnv structure factors ( fig .
[ direct ] ) , and also by our algorithm from the @xmath40 coefficients ( fig .
[ reconstr ] ) , which may be computed from the measured data of the fel diffraction patterns from random particle orientations .
the similarity of the reconstructed images of figs .
[ direct ] and [ reconstr ] was a further indication of the validity of the method of image reconstruction from the quantities @xmath40 derivable from the average angular correlations .
the fact that the reconstructed image consists of a thin protein shell is also seen from the slice perpendicular to a 5-fold rotation axis through the reconstructed image of fig .
[ reconstr ] depicted in fig .
[ cutout ] . in the case of all three figures , a ribbon representation of the structure from the biological assembly of the stnv virus from the same pdb structure used to simulate the diffraction patterns
is superimposed on the semi - transparent electron density to show the excellence of the reconstruction .
it should be emphasized that nowhere in our theory is it assumed that the structure consists of a thin protein shell , unlike the so - called shell model that has been used in the saxs analysis of virus capsids @xcite . in our case ,
the existence of a shell is deduced by an iterative phasing algorithm from the anaysis of data from diffraction patterns of random particle orientations without any assumptions on our part .
satellite tobacco necrosis virus ( stnv ) is an example of a virus with a perfectly icosahedral protein coat @xcite .
a host cell gets access to the genetic material of this virus by ingesting it whole and dissolving its protein coat .
many of the larger viruses are only approximately icosahedral : they often have appendages , such as a neck sticking out of the coat that is used to inject the genetic material inside the coat into a host cell whose protein making capability is hijacked by the virus dna or rna .
an ultimate reconstruction algorithm should be able to reconstruct these non - icosahedral parts of the structure in addition to the icosahedral part .
the above procedure has determined the icosahedral harmonic expansion coefficients @xmath46 that best fit the measured quantities @xmath40 .
any deviations from these values are due to the non - icosahedral parts of the structure .
any differences between the experimental values of @xmath40 and @xmath103 may be written @xmath104 in terms of @xmath105 , the extra contribution to the rsh expansion coefficients due to deviations from icosahedral symmetry .
note that for @xmath106 combinations not associated with icosahedral harmonics , e.g. those for which there is no entry in a list like table 1 , the terms @xmath20 will be zero , and only the quadratic terms in @xmath107 will survive in ( [ delta_b ] ) .
determination of the @xmath105 coefficients which optimize the agreement the theoretical expression ( [ delta_b ] ) and the measured values will enable the construction of a better estimate of a single - particle diffraction volume via @xmath108 the presence of the correction terms @xmath105 , which have no symmetry restrictions ( apart from friedel symmetry ) will allow the diffraction volume calculated by this formula to include deviations from icosahedral symmetry
. application of an interative phasing algorithm to an oversampled diffraction volume calculated by this expression will enable the determination of the full structure of the virus , including any appendages that break the approximate icosahedral symmetry .
the remarkable similarity of the reconstructed electron densities of figs .
[ direct ] and [ reconstr ] , and the fit of the latter to the model of stnv from the pdb file , are indications of the correctness of the method of reconstruction of the 3d diffraction volume from the average angular correlations of the 10,000 simulated diffraction patterns of stnv .
we calculated from these the @xmath40 coefficients for all values of @xmath2 from 0 to @xmath109 .
we found good agreement with the selection rule on the @xmath2 coefficients in which the sizes of the @xmath110 coefficients for all odd values of @xmath2 were small ( due to friedel , or inversion , symmetry ) and in addition the even values @xmath2=2,4,8 , and 14 were also small , due to the icosahedral symmetry of the 3d diffraction volume of a single particle .
we included @xmath46 coefficients for the non - negilible @xmath40 coefficients up to @xmath2=30 ( up to which value the icosahedral harmonic expansion coefficients depend on the @xmath2 quantum number only ) .
if @xmath57 is the maximum value of the reciprocal - space coordinate @xmath48 up to which the reconstruction is valid , conventional wisdom @xcite suggests that @xmath111 and @xmath57 should be related by @xmath112 where @xmath113 is the radius of the particle .
taking @xmath114 where @xmath115 is the resolution . substituting ( [ qmax ] ) into ( [ limit ] ) and rearranging ,
we find that @xmath116 stnv has a radius of @xmath97 100 suggesting a resolution of about 20 . in practice we found that increasing @xmath57 a further 50% or so , while keeping @xmath111 fixed at 30 seemed to improve the quality of the reconstructed image .
presumably because up to about 1.5@xmath57 the spherical harmonic expansion coefficients of @xmath2 greater than 30 remain small
. it should be emphasized this is not necessarily an absolute limit of the resolution obtainable with the use of icosahedral harmonics .
the higher order harmonics , at least up to @xmath2=44 , have been tabulated by zheng _
at least up to this value , the degeneracy of the icosahedral harmonics characterized by a particular value of @xmath2 is no more than two .
although the algorithm for recovering the expansion coefficients of such degenerate hamonics from the experimental data is a little more complicated , it seems far from an insuperable problem . the images in figs .
[ direct ] to [ reconstr ] were computed by an iterative phasing algorithm @xcite from a reciprocal - space distribution of intensities oversampled @xcite by a factor of @xmath97 2 with respect to the size of stnv , up to a @xmath57 value of @xmath97 0.47 @xmath100 ( a 61@xmath9861@xmath9861 array ) , implying a resolution of about 13 , and a @xmath117 ratio closer to @xmath118 .
further , the images of fig .
[ direct]-[cutout ] reveal this coat to be hollow .
the slice ( fig .
[ cutout ] ) through the reconstructed image perpendicular to the 5-fold axis reveals both external and internal surfaces of 5-fold rotational symmetry .
the revelation of the hollow nature of the protein coat is of course an extra feature contained in the 3d intensity distribution above and beyond the assumed icosahedral symmetry .
it is revealed by the iterative reconstruction algorithm used @xcite due to the paricular variation of the @xmath40 coefficients with the radial reciprocal - space coordinates @xmath48 and @xmath58 .
some the advantages of this method of analysis of single particle diffraction patterns from unknown particle orientations compared with other proposed algorithms @xcite should be pointed out . since it has been shown @xcite , that the angular correlations of multiple identical particles in arbitrary orientations are essentially identical to those from a single particle , the method we have described is equally applicable to droplets containing multiple particles injected into the xfel @xcite as to the injection of single particles in random orientations .
thus there is no need to discard diffraction patterns from multiple particle hits . since the inputs to our algorithm are not the direct photon counts , but rather the average of the angular correlations between intensities of the _ same _ diffraction patterns , it is insensitive to shot - to - shot fluctuations between the diffraction patterns , as may be caused by intensity variations of the incident x - ray beam or , for example , by the number of particles scattering a particular x - ray pulse .
the raw experimental data is likely to consist of @xmath97 10@xmath119 diffraction patterns , each of @xmath97 10@xmath119 pixels .
thus the raw experimental data will require tb of storage .
of course , this is very noisy data , and the structural information content is much less than this .
the averaging of the angular correlations that we perform may be regarded as a form of data averaging that results in infomormation concentration and noise reduction .
even if the number of values of q chosen is , say , 61 , and these coefficeints are evaluated for , say , 30 values of @xmath2 , the total number of these ( real ) coefficients will be only of the order of 100,000 , requiring less than a mb of storage .
this data reduction is best performed at the site of the data to allow the tranference of a million times less data over the internet to the site of image reconstruction .
the reconstruction of 3d images of the quality of figs . 8 - 10 from a properly constructed set of @xmath40 coefficients is extraordinarily rapid . in our calculations ,
reconstruction of an array of @xmath101 values representing a 3d diffraction volume at reciprocal - space coordinates @xmath102 on a 61@xmath9861@xmath9861 cartesian grid took just a few seconds on a single intel q6600 processor , using an intel fortran compiler .
the reconstruction of real - space images of the quality of figs . 8 - 10 from this array by means of a `` charge flipping '' algorithm @xcite took a further 4 minutes for 200 iterations on a laptop pc .
of course , the averaging of the data from the different diffraction pattterns assumes they all arise from copies of the particle in different orientations ( as does the technique of small angle x - ray scattering , saxs , for example ) . in order to distinguish between different conformations of the individual molecules
, it may be necessary to operate on the entire ensemble of all the measured diffraction patterns one of the disadvantages of such methods is the need to operate on single - particle diffraction patterns and thus , unlike with our method , diffraction patterns from multiple hits need to be removed .
such methods also face the problem of the uncertain normalization of incident intensities between successive pulses of incident radiation .
also , in contrast to our method , such techniques may require the tranferance of perhaps tb of data over the internet to the site of the data performing the analysis , which needs to be equipped with a cluster of computers performing parallel computations .
when reconstructing the structure of a virus from `` diffract and destroy '' type single - particle diffraction experiments proposed for the free electron laser @xcite , one may exploit the dictum of caspar and klug @xcite that `` there are only a limited number of efficient designs possible for a biological container which can be constructed from a large number of identical protein molecules , the two basic designs are helical tubes and icosahedral shells '' .
we offer here a solution for the case of icosahedral viruses . for those viruses which are substantially , though not completely icosahedral
, the method proposed is expected to be useful nontheless for initially reconstructing the approximate icosahedral structure .
the deviations from this structure can then be found by a perturbation theory which does not impose this symmetry .
the input to the algorithm is data from diffraction patterns of randomly oriented identical particles ( where the particle orientations are unknown ) in the form of the average of the angular correlations . as a consequence of any approximate icosahedral symmetry of the scattering particles ,
the angular momentum decomposition of the angular correlations contains only a few dominant contibutions from low values of the angular momenta .
this immediately suggests a simple test of whether the experimentally measured data are from the scattering by an icosahedral object .
if so , the components of the quantities @xmath40 , derivable from the angular correlations , should have much smaller values for @xmath2=1,2,3,4 , and 5 than for @xmath2=0 and @xmath2=6 , for example .
what is more , as we have shown in this paper , the coefficients @xmath46 of the icosahedral harmonic expansion of the 3d diffraction volume of the particle may be derived from the @xmath40 data and a positivity condition on the intensities of the 3d diffraction volume
. this will be the case even if the individual diffraction patterns are a result of scattering from more than one particle , so there will be no need to discard the diffraction patterns from multiple particles .
having obtained the coefficients of an icosahedral harmonic expansion , the 3d diffraction volume may be reconstructed as a sum over these icosahedral harmonics . by definition , the resulting diffraction volume will have icosahedral symmetry .
if this is constructed at a grid that is oversampled by a factor of 2 in each dimension , we have shown that a `` charge flipping '' algorithm with no fixed support contraint is able to reconstruct a 3d image of the particle .
we find that this procedure not only reconstructs an icosahedral shape for the particle , in simulations for the satellite tobacco necrosis virus ( stnv ) it even reveals the hollow nature of the protein coat . | the first experimental data from single - particle scattering experiments from free electron lasers ( fels ) are now becoming available .
the first such experiments are being performed on relatively large objects such as viruses , which produce relatively low - resolution , low - noise diffraction patterns in so - called `` diffract - and - destroy '' experiments .
we describe a very simple test on the angular correlations of measured diffraction data to determine if the scattering is from an icosahedral particle .
if this is confirmed , the efficient algorithm proposed can then combine diffraction data from multiple shots of particles in random unknown orientations to generate a full 3d image of the icosahedral particle .
we demonstrate this with a simulation for the satellite tobacco necrosis virus ( stnv ) , the atomic coordinates of whose asymmetric unit is given in protein data bank entry 2buk .
+ 99 r. neutze , r. wouts , d. van der spoel , e. weckert , and j. hajdu , `` potential for biomolecular imaging with femtosecond x - ray pulses '' , nature * 406 * , 752 - 757 ( 2000 ) .
m. m. seubert et al . , `` single mimivirus particles intercepted and imaged with an x - ray laser , '' nature * 470 , * 7881 ( 2011 ) .
v. l. shneerson , a. ourmazd , and d. k. saldin , `` crystallography without crystals i : the common - line method for assembling a 3d diffraction volume from single - particle scattering '' , acta cryst . a * 64 * , 303 - 315 ( 2008 ) .
r. fung , v. l. shneerson , d. k. saldin , and a ourmazd , `` structure from fleeting illumination of faint spinning objects in flight '' , nature physics * 5 * , 64 - 67 ( 2009 ) .
n .- t .
d. loh and v. elser , `` reconstruction algorithm for single - particle diffraction imaging experiments '' , phys .
rev .
e * 80 * , 026705 ( 2009 ) .
d. k. saldin , v. l. shneerson , r. fung , and a. ourmazd , `` structure of isolated biomolecules from ultrashort x - ray pulses : exploiting the symmetry of random orientations '' , j. phys : condens .
matter ( 2009 ) .
z. kam , `` determination of macromolecular structure in solution by spatial correlation of scattering fluctuations '' , macromolecules * 10 * , 927 - 934 ( 1977 ) .
r. a. kirian , k. e. schmidt , and j. c. h. spence , `` signal , noise , and resolution in correlated fluctuations from snapshot saxs '' , submitted for publication ( 2011 ) .
d. l. d. caspar and a. klug , `` physical principles in the construction of regular viruses '' , cold spring harbor symp .
. biol .
* 27 * , 1 - 24 ( 1962 ) .
a. jack and s. c. harrison , `` on the interpretation of small - angle x - ray solution scattering from spherical viruses '' , j. mol .
biol . * 99 * , 15 - 25 ( 1975 ) .
z. kam , `` the reconstruction of structure from electron micrographs of randomly oriented particles '' , j. theor .
biol . * 82 * , 15 - 39 ( 1980 ) .
d. k. saldin , h. c. poon , v. l. shneerson , m. howells , h. n. chapman , r. kirian , k. e. schmidt , and j. c. h. spence , `` beyond small angle x - ray scattering : exploiting angular correlations '' , phys .
rev .
b * 81 * , 174105 ( 2010 ) .
j. miao , p. charalambous , j. kirz , and d. sayre , `` extending the methodology of x - ray crystallohgraphy to allow imaging of mikcrometer - sized non - crstalline specimens '' , nature * 400 * , 342 - 344 ( 1999 ) .
s. marchesini , `` a unified evaluation of iterative projection algorithms for phase retrieval '' , rev .
sci .
instrum . * 78 * , 049901 ( 2007 ) .
see e.g. the wikipedia entry at http://en.wikipedia.org/wiki/icosahedron d. a. varshalovich , a. n. moskalev , and v. k. khersonskii , _ quantum theory of angular momentum _
, ( world scientific , 1988 ) , p.148 . p. j. davis and p. rabinowitz , _ methods of numerical integration _ , 2nd ed .
( dover publications , 1984 ) .
l. lovisolo and e. a. b. da silva , `` uniform distribution of points on a hyper - sphere with applications to vector bit - plane encoding '' , ieee proc .
vis . image signal process .
* 148 * , 187 - 193 ( 2001 ) .
h. c. poon and d. k. saldin , `` beyond the crystallization paradigm : structure determination from diffraction patterns of ensembles of randomly oriented particles '' , ultramicrosc . * 111 * , 798 - 806 ( 2011 ) . c. xiao , p. r. chipman , a. j. battisti , v. d. bowman , p. renesto , d. raoult , and m. g. rossmann , `` cryo - electron microscopy of the giant mimivirus '' , j. mol .
biol . *
353 * , i493 - 496 ( 2005 ) .
m. v. cherrier , c. a. xiao , v. d. bowman , a. j. battisti , x. d. yan , p. r. chipman , t. s. baker , j. l. van etten , and m. g. rossmann , `` an icosahedral algal virus has a complex unique vertex decorated by a spike '' , pnas * 106 * , 11085 - 11089 ( 2009 ) .
g. oszlyi and a. sto , `` _ ab initio _ structure solution by charge flipping '' , acta cryst .
a*60 * , 134 - 141 ( 2004 ) .
g. oszlyi and a. sto , `` _ ab initio _ structure solution by charge flipping .
ii . use of weak reflections '' , acta cryst .
a*61 * , 147 - 152 ( 2005 ) .
y. zheng , p. c. doerschuk , and j. e. johnson , `` determination of three - dimensional low - resolution viral structure from solution x - ray scattering data '' , biophys .
j. * 69 * , 619 - 639 ( 1995 ) .
viperdb , a database for icosahedral virus capsid structures may be accessed at http://viperdb.scripps.edu . as of may 5 , 2011
, it contained 298 icosahedral viruses , of which the structures of 265 has been determined by x - ray diffraction , and 33 by cryo electron microscopy .
j. b. pendry , _ low energy electron diffraction _ , ( academic press , 1974 ) .
d. k. saldin , v. l. shneerson , m. howells , s. marchesini , h. n. chapman , r. kirian , k. e. schmidt , and j. c. h. spence , `` structure of a single particle from scattering of many particles randomly oriented about an axis : towards structure solution without crystallization ? '' , new j. phys .
* 12 * , 035014 ( 2010 ) .
d. k. saldin , h. c. poon , m. bogan , s. marchesini , d. a. shapiro , r. a. kirian , u. weierstall , and j. c. h. spence , `` new light on disordered ensembles : ab - inito structure determination of one particle from scattering fluctuations of many copies '' , phys .
. lett .
* 106 * , 115501 ( 2011 ) .
d. de ponte , r. b. doak , p. fromme , g. hembree , m. hunter , s. marchesini , k. e. schmidt , u. weierstall , and j. c. h. spence , j. synch .
rad .
* 15 * , 593 - 599 ( 2008 ) . |
the scenario of formation of low mass stars ( m @xmath3 8 m@xmath4 ) , through disk accretion controlled by magnetic field is already well understood @xcite . for high masses ( m @xmath5 8 m@xmath4 ) , however , there are theoretical problems for applying the same scenario . from the observational point of view , there are too few bona fide candidates to test the models .
this is due to the fact that massive young stellar objects ( mysos ) are short lived and occur inside large molecular clouds , subject to huge reddening .
the relevant phases of massive stars formation are not seen directly at wavelengths shorter than the mid - infrared ( mir ) and also to wavelengths longer than radio - mm .
shocked molecular outflows are `` smoking guns '' of the accretion process and can be traced outside the central heavily obscured zone .
the aim of this work is to use the h@xmath0 @xmath1=1 - 0 s(1 ) emission to constrain the possible models . even after decades of progress both theoretical and observational fronts ,
formation of massive stars remains an open question .
the short timescale for myso formation indicates that high rates of mass accretion are required .
two scenarios currently dominate the discussions , @xmath6 in the first , high mass stars are formed by accretion through a disk @xcite ; @xmath7 in the second , high mass stars are formed via coalescence of low mass stars @xcite . in the last decade , a number of observations have been presented favoring the first scenario , among them @xcite and @xcite .
low and intermediate mass stars are formed by the gravitational collapse of the parental giant molecular cloud ( gmc ) , followed by the accretion process @xcite . during the accretion phase
, material is ejected as well via collimated bipolar jets . however ,
when a yso reaches 8 m@xmath4 , the radiative flux becomes so intense ( using @xmath8 , the ratio between the radiative fluxes of an o5 and a b3 star masses of @xmath9 and @xmath10 m@xmath4 , respectively is @xmath11 ) that it may interrupt the accretion flow .
a process that constrains the outcoming radiation field to narrower angles may leave some room for the accretion process to continues in some directions .
this seems to be the case for the outflows driven by young stars from a very broad mass range , as previous reported by several authors @xcite .
outflows associated with high - mass objects are expected to be more energetic than the outflows observed in lower mass ysos @xcite , with velocities greater than @xmath12 km s@xmath13 @xcite .
some authors have found evidences that outflows associated with massive stars are scaled up versions of their low mass counterparts @xcite while other works have reported that no well - collimated outflows have been found toward mysos @xcite .
massive yso outflows mapped in high - velocity co lines have collimation factors r = length / width @xmath2 [email protected] as compared to r @xmath2 [email protected] for low mass stars @xcite , indicating a weak tendency that outflows associated with massive stars are less collimated than those from low mass stars as previously thought @xcite . besides the degree of collimation
, these massive outflows would work removing mass from the plane of the accretion disk , lowering the density on the plane and , therefore , facilitating the accretion flow to reach the stellar core as shown in the recent 3-d simulations presented by @xcite .
although these authors have not included the outflow activity on their simulations , they argue that the presence of outflows would decrease the star formation efficiency from 70% ( considering purely radiation effects ) to 50% . in the second scenario ,
massive stars are formed by collision and coalescence of low mass stars in dense clusters @xcite .
the low mass stars are formed after the gravitational collapse of the gmc and through an accretion disk , interact with each other , colliding to form stars with greater masses . in this case
, the massive stars would be formed in the highest density parts of clusters , and the low mass stars would be the dominant fraction of the entire cluster mass @xcite .
moreover , no massive stars would be expected in small stellar clusters .
while there is no observational evidence for stellar merges in clusters , a growing number of observations supports the accretion disk scenario by answering questions such as the duration of accretion period @xcite , the required accretion rates to overcome the radiation pressure @xcite and the confirmation of molecular jets ejected from these massive objects @xcite .
the existence of outflows driven by stars from a large mass range is a potentially interesting aspect , since stellar mergers can not produce stable outflows ; therefore , the observation of such flows can be used to discern between the two scenarios of massive star formation .
bipolar molecular jets associated with myso candidates indicate that massive stars are formed by accretion .
however , previous samples were contaminated by low mass objects , mostly due to inaccurate photometry . recently , myso candidates were observed in the h@xmath0 @xmath1=1 - 0 s(1 ) transition and collimated jets were identified , suggesting accreting discs around these objects @xcite .
although this work presents observational evidence for the accretion scenario , few mysos were confirmed in that sample . to identify bonafide myso candidates , accurate
photometry and well developed selection criteria are necessary to avoid contamination by normal h regions , proto - planetary nebulae ( ppne ) and cool stars .
the red msx source ( rms ) survey @xcite contains a large list of objects measured in 2mass , msx , spitzer , and reprocessed iras / igal images for objects along the galactic plane , including ultra - compact ( uc ) h regions and a large sample of mysos .
additional observations were used to separate the mysos from uchs , such as water masers @xcite and centimetric emission @xcite .
recently , the bolometric fluxes of those sources were derived by @xcite using ir and submillimetric data and adopting the kinematic distances derived by co observations from @xcite and a spectral energy distribution ( sed ) model .
they found there are few mysos ( @xmath15 ) with l @xmath16 10@xmath17 l@xmath4 , and many mysos ( @xmath18 ) with 10@xmath19 l@xmath4 @xmath20 l @xmath20 10@xmath17 l@xmath4 .
there are a few examples of well documented accretion disks around massive ( 10 - 30 m@xmath4 ) forming stars @xcite using keplerian motion in the co lines .
disks have been identified in a few cases @xcite through profile fitting of emission features seen in high spectral resolution data , however , the accretion process might have already ceased and the disks are now passively evolving .
although the scenario of an accretion disk may apply for all massive stars , the details are lacking .
instead of doing a detailed study of a small number of potential candidates that might harbor a disk , we are moving toward a large statistical study which will point to accretion signatures ( or not ) of a well selected sample of mysos , selected by the rms survey .
a smaller survey based on the present motivation was conducted by @xcite .
these authors found that @xmath21 of the targets display h@xmath0 emission , which is indicative of shocks , and in @xmath22 of the sample the h@xmath0 features have some symmetry , suggesting collimated outflows and so accretion disks . such signatures are more frequent in pre - uch than in uch phase .
the most luminous ysos in that sample correspond to early b or late o star progenitors .
the lack of more massive stars may be due to several causes such as : insufficient number of targets ( 50 ) ; underestimation of luminosities due to errors in the kinematic distances or in the method of magnitude extraction .
problems with the kinematic distances of star forming regions have been pointed out by authors who derive spectrophotometric and trigonometric distances ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* and references therein ) .
kinematic distances are half the magnitude of the kinematic ones .
as discussed in @xcite , the kinematic distances are larger than their non - kinematic counterparts mostly due to two reasons : a ) breaking the ambiguity in the far / near distance is not always a reliable procedure for the kinematic distance method , b ) in directions close to the galactic center , where radial velocities are small , local perturbations in the gas produces velocity components of the same magnitude as the galactic rotation . if this result is reliable , than the luminosity of those stars would be @xmath23 4 times smaller than predicted by the original data
unfortunately , mysos in pre - uch phases have not been reported for massive star clusters , where spectrophotometric distances are accurate . in this paper , we present the results of our survey , which is organized as follows . in section 2 ,
the observations and data reduction are described . in section 3 we present overall results from the survey . in section 4
we discuss the results .
finally , our conclusions are summarized in section 5 .
the images of the sources presenting extended h@xmath0 emission are shown in the appendix . for those sources that do not display h@xmath0 emission ,
the images are found in the vizier catalogue .
the sample of 353 myso candidates comprises 135 sources located in the northern hemisphere ( @xmath24 ) and 219 from the southern hemisphere ( @xmath25 ) .
most of the northern subsample was observed with the canada - france - hawaii telescope ( cfht ) and the southern subsample was observed using the southern astrophysical research telescope ( soar ) .
details of the observing setup used on each telescope are described below .
the data were obtained with the cfht , mauna kea , hawaii , using the wide - field infrared camera ( wircam ) between 2010 august and 2012 august .
the wircam detector is a mosaic of four 2048 @xmath26 2048 rockwell scientific hawaii-2rg hgcdte arrays , with a @xmath27 gap between detectors .
the scale is @xmath28/pixel so that a single exposure covers a field of view ( fov ) of 21 @xmath26 21 arcmin@xmath29 .
the median seeing was @xmath30 and the airmass of each pointing was @xmath31 .
wircam is described by @xcite .
the sample was observed through the h@xmath0 ( @xmath32 m , @xmath33 m ) and _ k_-band continuum ( @xmath34 m , @xmath35 m ) narrow band filters . each target
was observed using a four point dither pattern , defining a @xmath36 square on the sky , with exposure time of @xmath37 seconds per frame , totalling 600 seconds per filter .
the dither pattern was employed to cover the 45@xmath38 gap between the detectors and also to facilitate the identification and the suppression of bad pixels and cosmic rays . using the observational setup described above
, we reached an average 3@xmath39 sensitivity of [email protected] and [email protected] on point sources in the h@xmath0 and continuum filters , respectively .
the equivalent 3@xmath39 sensitivity per pixel is @xmath40 wm@xmath41arcsec@xmath29 and @xmath42wm@xmath41arcsec@xmath29 for the h@xmath0 and continuum filters , respectively .
the southern sample have been observed with the southern astrophysical research telescope ( soar ) , cerro pachn , chile , using the ohio state imager / spectrometer ( osiris ) camera between 2011 march and 2011 june and the spartan camera between 2011 august and 2012 december .
the osiris camera has a 1024 @xmath26 1024 rockewell scientific hawaii hgcdte detector with a pixel scale of 0.139@xmath38 .
a single exposure covers a fov of approximately 1.5 @xmath26 1.5arcmin@xmath29 .
osiris is described by @xcite .
the narrow - band h@xmath0 ( @xmath43 m , @xmath44 m ) and continuum ( @xmath45 m , @xmath46 m ) filters were used for the osiris observations .
the data taken with osiris were observed using a three point dither pattern , separated by 20 arcseconds in the e - w direction , with exposure times of 180 s per frame , totaling 540 s per filter . using the observational setup for osiris
, we reached an average 3@xmath39 sensitivity of [email protected] and [email protected] on point sources in the narrow h@xmath0 and continuum filters , respectively .
the equivalent 3@xmath39 sensitivity per pixel is @xmath47wm@xmath41arcsec@xmath29 and @xmath48wm@xmath41arcsec@xmath29 for the h@xmath0 and continuum filters , respectively .
the spartan infrared camera @xcite has four hawaii
ii arrays deployed in a 2 @xmath26 2 mosaic .
the pixel scale is 0.067@xmath38 such that a single exposure covers a fov of approximately 5 @xmath26 5arcmin@xmath29 .
the narrow - band h@xmath0 ( @xmath49 m , @xmath50 m ) and continuum filters from the spartan s filter set ( @xmath51 m , @xmath52 m ) were used .
the targets observed with spartan used a four point dither pattern , defining a @xmath36 square on the sky , with exposure time of 150 seconds per frame , totaling 600 seconds per filter . using the observational setup described above
, we reached an average 3@xmath39 sensitivity of [email protected] and [email protected] on point sources in the narrow h@xmath0 and continuum filters , respectively .
the equivalent 3@xmath39 sensitivity per pixel was @xmath53wm@xmath41arcsec@xmath29 and @xmath54wm@xmath41arcsec@xmath29 for the h@xmath0 and continuum filters , respectively .
+ the central coordinates of the first 20 objects in the complete sample are listed in table [ table : observations ] ( the full version of this table containing 353 entries is available online at the cds ) .
each h@xmath0 processed image ( see below ) was continuum - subtracted by an associated _ k_-band continuum image scaled to the same effective bandwidth .
the usage of narrow filters avoided contamination by nebular features such as brackett-@xmath55 that would be present in a broad - band filter and so might mask h@xmath0 emission . | we present the results from a survey , designed to investigate the accretion process of massive young stellar objects ( mysos ) through near infrared narrow band imaging using the h@xmath0 @xmath1=1 - 0 s(1 ) transition filter .
a sample of 353 massive young stellar object ( myso ) candidates was selected from the red msx source survey using photometric criteria at longer wavelengths ( infrared and submillimeter ) and chosen with positions throughout the galactic plane .
our survey was carried out at the soar telescope in chile and cfht in hawaii covering both hemispheres .
the data reveal that extended h@xmath0 emission is a good tracer of outflow activity , which is a signpost of accretion process on young massive stars .
almost half of the sample exhibit extended h@xmath0 emission and 74 sources ( 21% ) have polar morphology , suggesting collimated outflows .
the polar - like structures are more likely to appear on radio - quiet sources , indicating these structures occur during the pre - uch phase
. we also found an important fraction of sources associated with fluorescent h@xmath0 diffuse emission that could be due to a more evolved phase .
the images also indicate only @xmath223% ( 80 ) of the sample is associated with extant ( young ) stellar clusters .
these results support the scenario in which massive stars are formed by accretion disks , since the merging of low mass stars would not produce outflow structures .
= 1 [ firstpage ] infrared : stars stars : formation stars : early - type stars : massive stars : pre - main sequence
ism : jets and outflows . |
it is well - known that a wide variety of social networks have the so - called _ small world _
property ; networks of size @xmath1 have diameter @xmath2 , meaning that between any two nodes there exists a path of size @xmath2 .
this property is also shared by sparse random graphs ; however , random graphs lack several other important properties of real - world networks , such as clustering and heavy - tailed degree distributions ( see @xcite for a review ) .
another property of social networks which distinguishes them from random graphs , and which has received somewhat less attention , is their _
navigability_. that is , not only do short paths exist , but it is easy to find them using only local information .
the pioneering work of milgram in the 1960s @xcite showed that people can , at least some of the time , find short paths to other distant people by pursuing the greedy strategy where they pass the message to whichever of their own acquaintances they feel is `` closest '' to the target recipient .
this experiment has recently been repeated using e - mail networks @xcite . in a random graph , although a short path exists , a local algorithm must be lucky to find it as it can do little better than a random walk on the network .
kleinberg @xcite addressed the issue of navigability by considering a small - world model consisting of a @xmath3-dimensional lattice with long - range links added to it . unlike the watts - strogatz model @xcite in which the long - range links are uniformly distributed , in the kleinberg model pairs of nodes a distance @xmath4 apart are connected with some probability @xmath5 , where @xmath6 and the number of outgoing links per node is fixed . while a finite - dimensional lattice is obviously a gross oversimplification of the social spaces in which we live , or the conceptual spaces in which one web page seems `` closer '' to another , this model captures the essential features of the navigation problem : how can we minimize the routing time given local information about some metric ?
kleinberg studied the performance of the greedy algorithm , in which each node passes the message to whichever of its neighbors , either local or long - range , that is closest to the destination on the underlying lattice .
he showed that if @xmath7 this algorithm achieves a routing time of @xmath0 on lattices of size @xmath1 .
however , if @xmath8 , the routing time grows as @xmath9 for some @xmath10 . note that if when @xmath7 , we integrate over spheres of radius @xmath4 , we find that the distribution of link lengths is @xmath11 , with a cutoff when @xmath4 reaches the system size ; this distribution provides the right mix of long- , medium- , and short - range links for the greedy algorithm to quickly `` zero in '' on its destination .
similar results have been obtained for the case where the underlying graph is a tree , representing a hierarchical structure , or an overlapping set of trees representing multiple affiliations in society @xcite . while kleinberg s work largely characterizes the
properties a network needs to have to be navigable , it leaves open the question of how or why networks might evolve these properties . indeed , the topology of a social network is constantly being modified by its members . if , whenever these members find it difficult to search the network , they modify their own connections in an effort to make future searches easier , this dynamical process should make the network more navigable over time . in this paper , we consider a dynamic network model inspired by surfers on the web , each of whom controls the outgoing links from their home page .
we start at a source node @xmath12 , and choose a random destination node @xmath13 . based on the ( metric ) distance between @xmath12 and @xmath13
, we set a threshold @xmath14 on the number of ( topological ) steps we feel the journey should take .
if after @xmath14 steps we have not reached our destination , we stop searching , and rewire @xmath12 s long - range link to the place where we gave up . in the human - friendship network
, this might correspond to gaining new acquaintances in the course of a search ; on the web , it corresponds to creating a bookmark / linking to the relevant pages , which other surfers can then use @xcite .
our main result is that this process does indeed cause the distribution of link lengths to converge to a power law @xmath15 .
the precise value of @xmath16 varies with lattice size , and differs somewhat from kleinberg s prediction @xmath17 .
we believe this is due to finite - size effects ; to support this belief , we directly construct networks with power - law link length distributions ( as opposed to rewiring them ) , measure the exponent @xmath18 that minimizes the routing time , and find that it converges rather slowly to @xmath3 as @xmath19 , roughly as @xmath20 .
however , even though the exponents differ somewhat , our rewiring process produces networks whose routing times match or improve those of kleinberg s optimum .
other types of dynamical models have been studied ( e.g. @xcite ) .
however , the model discussed here appears to be the first to optimize for navigability .
one restriction of kleinberg s model of a social space consists of a @xmath3-dimensional lattice of size @xmath1 , where each node @xmath21 is connected to its @xmath22 nearest neighbors and has a single long - range link to a node @xmath23 , chosen with @xmath6 where @xmath24 is the manhattan distance ( the @xmath25 norm ) .
such networks are denoted @xmath26 in @xcite ; here the two @xmath27s indicate the radius of the local and the number of long - range links per node .
our goal is to show that our rewiring process causes networks with a range of initial link length distributions to converge to this form , where @xmath28 and is close to kleinberg s prediction @xmath17 .
we report here on where @xmath29 , for which we can feasibly study networks of size up to @xmath30 ; our results for @xmath31 are qualitatively similar . in each round of the rewiring process , we choose the source node @xmath12 uniformly , and choose the destination @xmath13 according to a _ demand distribution _ @xmath32 .
one method of selection is to choose a distance @xmath4 according to a distribution @xmath33 , and then choose @xmath13 randomly from among the nodes that satisfy @xmath34 . here
, we take @xmath33 to be uniform , which for @xmath29 means that @xmath13 is simply a uniformly random node ; we comment in the conclusions on the effect of other demand distributions .
starting at @xmath12 , the greedy algorithm produces a path @xmath35 where @xmath36 is the routing time , and where for each @xmath37 , @xmath38 is the outgoing neighbor of @xmath39 that minimizes @xmath40 ( with ties broken randomly ) .
the rewiring process works as follows : if @xmath41 , i.e. , if the greedy algorithm reaches some @xmath39 with @xmath42 , then we discontinue the search and change @xmath12 s outgoing link to point to @xmath39 . in our experiments ,
we choose the threshold @xmath14 uniformly at random from the interval @xmath43 $ ] .
such a naive selection avoids making assumptions about the network s size , performance or topology .
in particular , it avoids assuming the @xmath0 routing time which both the rewired networks and kleinberg s networks achieve . for our initial conditions ,
the link length distributions have exponent @xmath44 ranging from @xmath45 , a uniform distribution as in the watts - strogatz model , to @xmath46 , where all the `` long - range links '' are simply self - loops .
we run @xmath47 rounds per node of the rewiring process ; that is , @xmath48 times we choose nodes @xmath49 , run the greedy algorithm from @xmath12 to @xmath13 and rewire if @xmath50 .
we then compare the final link length distribution to kleinberg s optimum @xmath11 . after running the rewiring process until @xmath51 .
the initial graphs had @xmath52 , i.e. , only self - loops . ] for a range of initial link length exponents @xmath44 , where @xmath53 . at @xmath47 increases
, @xmath16 converges to @xmath54 regardless of the value of @xmath44 . ]
figure [ fig : logbinned ] illustrates the resulting distributions for networks of size @xmath55 and @xmath52 .
we see that after @xmath48 rounds ( we describe our choice of @xmath47 in the next section ) the rewiring process has built a power - law distribution of link lengths , @xmath15 with @xmath56 respectively .
figure [ fig : variousa ] shows that a range of initial distributions converge to the same final distribution .
starting with initial exponents @xmath44 ranging from @xmath45 to @xmath57 , the exponent of the rewired link length distributions all converge to @xmath58 ( here @xmath59 ) as the rewiring continues .
while our rewiring process produces a power - law link length distribution , the exponent is noticeably different from the value @xmath60 ( more generally , @xmath17 ) which kleinberg proved in the limit @xmath19 .
notably , figure [ fig : variousa ] shows that networks with an initial exponent @xmath61 actually move away from this value as they are rewired .
this deflection appears to be due to finite - size effects , and in fact the exponent @xmath18 that minimizes the routing time on finite lattices turns out to be rather different from kleinberg s value even at reasonably large @xmath1 .
we examined system sizes over six orders of magnitude , @xmath62 . for each size , we constructed networks with @xmath63 ranging from @xmath45 to @xmath27 , and measured the mean routing time @xmath64 over @xmath65 trials for each @xmath63 .
we estimated @xmath18 as the minimum of the best quadratic fit of @xmath64 .
figure [ fig : static - egfit ] illustrates @xmath64 for several values of @xmath1 and clearly shows that @xmath66 .
the dashed line illustrates @xmath18 s slow approach to @xmath27 as @xmath1 increases .
figure [ fig : static - rminasymptote ] shows the dependence of @xmath18 on @xmath1 ; it is fit rather closely by the form @xmath67 for @xmath68 .
the fact that @xmath18 converges polylogarithmically , as opposed to polynomially , indicates that finite - size effects are quite severe .
mean routing time @xmath64 for several network sizes .
the dashed line shows the path @xmath18 , defined as the minimum of the best quadratic fit , follows as @xmath1 increases . ]
finite - size dependence of @xmath18 for @xmath1 up to @xmath69 .
the solid line shows a fit to @xmath70 , showing that @xmath18 converges polylogarithmically slowly to @xmath3 as @xmath19 . ]
mean routing time @xmath36 for networks with @xmath71 .
shown is a fit to @xmath72 of the form @xmath73 . ] however , while @xmath16 differs from @xmath3 ( and from @xmath18 ) our rewired networks achieve a mean routing time equivalent to networks with exponent @xmath18 , and better than those with exponent @xmath3 .
[ fig : static - meanrouting ] compares the mean routing times for the three types of networks .
all three are closely fit with a curve of the form @xmath74 , just as kleinberg s analysis predicts @xcite . of course , the routing time of the network depends on the number of rewiring rounds . in figure
[ fig : relaxtime ] , we show the number of rounds per node @xmath47 required to achieve a mean routing time @xmath75 which is only @xmath76 greater than @xmath77 , i.e. , the routing time of a network with exponent @xmath18 .
this _ rewiring time _ grows as @xmath78 rounds per node , or @xmath79 rounds total . ,
i.e. , the number of rounds per node the rewiring process needs to achieve a routing time @xmath80 , within @xmath76 of that of a network with exponent @xmath18 .
the solid line is a fit of the form @xmath81 . ]
kleinberg @xcite explored the navigability of small - world networks built on an underlying @xmath3-dimensional space . in the limit of infinite size , he found the mean routing time in minimized by a power - law of link lengths with exponent @xmath7 @xcite . here , we have explained how might develop such a distribution over time , by a decentralized rewiring process that relies only on local information ( and is even ignorant of the size of the network ) .
this process has a natural interpretation : the topology of a social network is constantly being modified by its members , who update their personal connections as they explore and navigate the network .
if a becomes frustrated because the journey to a destination takes too long , they can be expected to change their to make similar journeys more quickly in the future .
our results show that rewiring causes a wide range of initial topologies to converge to a power - law distribution of link lengths , very similar to kleinberg s .
the exponent @xmath16 we obtain differs significantly from @xmath3 , its optimal value on infinite lattices .
we attribute this deflection to finite size effects which cause the optimal exponent @xmath18 to converge polylogarithmically as @xmath19 .
however , the rewired network achieves the same mean routing time as power - law networks with exponent @xmath18 , and better routing times than those with @xmath17 . specifically , the mean routing time as a function of system size is @xmath0 as predicted by kleinberg s analysis .
the number of rounds of the rewiring process needed to achieve this routing time grows as a low - degree polynomial of @xmath1 .
in addition to creating and maintaining a power - law distribution of link lengths , we believe this rewiring process to be _
adaptive_. for instance , we conjecture that if new nodes are added to the network , or if certain nodes or links are removed , it will dynamically optimize for these new situations .
we also believe , based on preliminary results , that if the demand distribution @xmath32 is not uniform , e.g. if certain destinations are more popular than others , or if the source and the destination are correlated ( both of which are true in any real network ) , that it will optimize routing times for the source - destination pairs that appear more frequently .
these adaptive properties would be particular useful in networks where nodes are constantly coming on- and off - line and where the demand of each destination rises and falls over time , such as live peer - to - peer networks , distributed sensor networks or massively parallel computers .
our discussion contemplates a `` social space '' consisting of a finite - dimensional lattice , an obviously poor model for the complex social spaces we routinely navigate .
an interesting study would be an analogous rewiring process for networks whose underlying structure is hierarchical @xcite , involves multiple group affiliations @xcite or is otherwise structured , as in the peer - to - peer network freenet @xcite with the modifications described in @xcite .
we leave these as directions for future work .
i. clarke , et al , `` freenet : a distributed anonymous information storage and retrieval system , '' in _ designing privacy enhancing technologies _ , lncs 2009 , h. federrath , ed .
, springer - verlag , berlin ( 2001 ) . | networks created and maintained by social processes , such as the human friendship network and the world wide web , appear to exhibit the property of _ navigability _ : namely , not only do short paths exist between any pair of nodes , but such paths can easily be found using only local information .
it has been shown that for networks with an underlying metric , algorithms using only local information perform extremely well if there is a power - law distribution of link lengths . however , it is not clear why or how real networks might develop this distribution . in this paper
we define a decentralized `` rewiring '' process , inspired by surfers on the web , in which each surfer attempts to travel from their home page to a random destination , and updates the outgoing link from their home page if this journey takes too long .
we show that this process does indeed cause the link length distribution to converge to a power law , achieving a routing time of @xmath0 on networks of size @xmath1 .
we also study finite - size effects on the optimal exponent , and show that it converges polylogarithmically slowly as the lattice size goes to infinity . |
there are various indications of the ultrastrong magnetic fields , @xmath1 g , from astrophysical observations and terrestrial accelerator experiments and there have been continuous efforts in finding their definite evidence in nature . among the examples considered
so far are strong magnetic fields captured on magnetar and gamma ray bursts(grb ) central engine and also created in noncentral heavy - ion collisions .
magnetars@xcite are considered to be neutron stars , where strong magnetic fields of typically @xmath2 g are the main source of energy .
even the stronger field strength is expected inside magnetars .
one of the viable models of grb central engine is to make use the idea of tapping the rotational energy of black hole by the strong magnetic field@xcite .
it is estimated to be @xmath3 g particularly for long and energetic bursts . in noncentral heavy - ion collisions , strong magnetic fields can also be created by the two electric currents in opposite directions generated by two colliding nuclei .
it is expected that the magnetic fields in rhic au + au collisions and lhc pb + pb collisions can be as large as @xmath4 g and @xmath5 g respectively@xcite .
these field strengths are much stronger than the critical magnetic field and offer an interesting opportunity to study the effect of super - strong electromagnetic fields beyond the classical electrodynamics and beyond the standard model of electro - weak interaction .
one of the interesting questions with given such strong electromagnetic fields is the stability of the field configuration or equivalently whether there is any instability , which leads to the particle production .
vacuum instability of the strong electromagnetic field has been discussed since long time ago .
the instability of the strong electric field due to the creation of electron pairs is one of the examples , which is known as schwinger process@xcite .
what matters are the coupling of particles to the electromagnetic field and the mass of the particle to be produced .
the critical electric field for electron is @xmath0 .
if it is possible to imagine a charged particle lighter than the electron , the critical field can be lowered down to the field strength produced in the future high intensity laser and will be subjected to be tested in the laboratory .
the pair production of fermions in a purely magnetic field configuration is shown to be absent@xcite .
therefore , the pair production of minimally interacting particles is considered to be a purely electric effect . however , while the minimal coupling derived by the local gauge invariance is of fundamental nature , there appear also non - minimal couplings as well in the form of the effective theory .
pauli introduced a non - minimal coupling of spin-1/2 particles with electromagnetic fields , which can be interpreted as an effective interaction of fermions with an anomalous magnetic moment@xcite .
hence for the neutral fermions , which have no minimal coupling to electromagnetic fields , the nonvanishing magnetic moments may be the primary window through which the electromagnetic interaction of neutral fermions can be probed with the pauli interaction .
it is known the spatial inhomogeneity of the magnetic field exerts force on the magnetic dipole moment through the pauli interaction .
it plays a similar role analogous to the electric field for the creation of charged particle pairs with the minimal coupling .
the possibility of pair production of the neutral fermions in a purely magnetic field configuration with spatial inhomogeneity has been demonstrated in 2 + 1 dimension@xcite .
the production rate in 3 + 1 dimension has been calculated explicitly for the magnetic fields with a spatial inhomogeneity@xcite , which can be approximated as m^4 e^-a m^2/|b| analogues to the schwinger process .
the instability of the particular vacuum under the strong magnetic field can be formulated as the emergence of imaginary parts of the effective potential . for uniform magnetic fields which interact with spin-1/2 fermions through the pauli interaction@xcite
, it is found that the non - vanishing imaginary part develops for a magnetic field stronger than the critical field @xmath6 , whose strength is the ratio of the fermion mass to its magnetic moment , @xmath7 , ( v_eff ) = ( -1)^3 ( + 3 ) . in section 2 , the calculations of the effective potential and vacuum decay rates for the neutral fermion with pauli coupling to electromagnetic field will be reviewed . the implication of non - vanishing magnetic dipole moment of neutrinos , the instability of the strong magnetic field and the neutrino production through the pauli coupling in high energy collider experiment , is discussed in section 3 .
the instability of the electro - weak vacuum for a strong magnetic field was discussed long time ago@xcite .
the one - loop effective potential considering the weak - boson(@xmath8 ) loop is found to have imaginary part under the pure magnetic background , ( v_eff ) = b^2(1-)(b - ) [ imvw ] where @xmath9 is the mass of @xmath8-boson . in the limit of @xmath10 , eq.([imvw ] )
agrees with the effective potential given by nielson and olesen @xcite .
it was argued that the instability could be avoided if the condensation of @xmath8 and @xmath11 bosons@xcite appears with the strong magnetic field .
the basic reason for the emergence of an imaginary part for @xmath12 is that the energy eigenvalue crosses zero at @xmath13 because of the anomalous magnetic moment of @xmath8 boson .
one can see that there is also a level crossing of energy eigenvalue of a neutrino with pauli interaction for a strong enough magnetic field . for a uniform magnetic field ,
the energy eigenvalues of the hamiltonian are given by e = [ epauli ] where @xmath14 and @xmath15 are respectively the longitudinal and the transversal momentum to the magnetic field direction .
one can see that for a magnetic field stronger than the critical field @xmath16 , the ground state with @xmath17 crosses the zero energy state .
this indicates the possible instability of the magnetic field configuration beyond critical field strength as in electro - weak instability @xcite . for a uniform magnetic field ,
the imaginary part of the one - loop effective action is calculated explicitly@xcite ( v_eff ) = ( -1)^3 ( + 3)(|b| - m ) . which takes the similar form as in eq .
( [ imvw ] ) .
it is interesting to note that the development of the imaginary part associated with a level crossing has been also demonstrated in the different contexts@xcite .
the state occupied in the negative energy sea becomes a particle state and the vacant positive energy state plunges into the negative sea to make an antiparticle state .
the instability can be interpreted as the pair creation at the expense of the magnetic field strength .
then the particle production rate is given approximately by ~2 i(v_eff ) = ( -1)^3 ( + 3)(|b| - m).[rate ] for a nonuniform magnetic field , the inhomogeneity of the magnetic field coupled directly to the magnetic dipole moment plays an interesting role analogous to the electric field for a charged particle : the non - zero gradient of the magnetic field can exert a force on a magnetic dipole moment .
then the vacuum production of neutral fermions with a non - zero magnetic moment in an inhomogeneous magnetic field is possible . as a simple example , a static magnetic field configuration with a constant gradient , @xmath18 , along x - direction , b_z(x ) = b_0 + bx . has been considered .
it is not necessary to consider an infinitely extended ever - increasing magnetic field to meet the linear magnetic field configuration .
because the particle production rate density is a local quantity , it is sufficient to have a uniform gradient magnetic field in the compton wavelength scale of the particle .
the imaginary part of the one - loop effective potential is calculated in the integral form@xcite , ( v_eff ) & = & - ^_0 [ f ( v , v ) - v - v ] + & & - ^_0 [ ( v v)^3/2- 1 - ] v , where v & = & s b , = , = , + f(a , b ) & = & ^1_0 d(1-)(a^2-b ) numerically it is found that it can be fitted to an analytic expression@xcite , ( v_eff ) m^4 e^- b , which is decreasing exponentially with respect to the inverse of the field gradient .
the qualitative feature of this result can be compared with the expression obtained recently by gavrilov and gitman@xcite .
neutrinos are the lightest particles(if not photon or gluon ) in the standard model " , of which electromagnetic property is poorly known experimentally . moreover
the observation of neutrino oscillation shows the necessity of neutrino masses .
it implies that the standard model is subjected to be modified such that non - trivial electromagnetic structure of neutrino should be reconsidered although they are assigned to be neutral .
while there is no observation which is not consistent with the electric - charge neutrality of neutrinos , the possibility of anomalous electromagnetic form factor is an open question theoretically and experimentally@xcite .
so far there is no experimental evidence of the magnetic dipole moment of the neutrinos but one can not simply rule out the magnetic moment just because of the electrical neutrality .
for example , in a minimal extension of the standard model to incorporate the neutrino mass the anomalous magnetic moment of a neutrino is known to be developed in one loop calculation @xcite , @xmath19 . here
one can notice that the non - zero mass is essential to get a non - vanishing magnetic moment .
there are strong evidences from the neutrino oscillation observations that neutrinos have nonzero - masses @xcite .
however , the simple extensions of the standard model can give only much smaller magnetic moment , @xmath20 if we take the neutrino mass as 1 ev , than the experimental bound .
gemma collaboration@xcite observed that the neutrino magnetic moment is bounded from above by the value @xmath21 and the bound obtained from borexino@xcite data gave @xmath22 .
since a wide range of neutrino magnetic moments is possible up to the current laboratory upper limit beyond standard model , it is interesting to investigate the phenomenological consequences of the nonvanishing magnetic moment up to the current bounds , which may constrain the bound tighter than the direct measurements of the neutrino magnetic dipole moment . for the vacuum instability , the typical field strength(or critical field strength )
is given by @xmath23 .
taking the possible magnetic moment to be as large as the experimental upper bound @xmath24 and the mass of the neutrino @xmath25 constrained by the neutrino oscillations , the critical field strength is estimated to be @xmath26 g , which is not far from the field strength inferred on the magnetar surface and might be possible inside the magnetar . with vacuum instability , the particle production rate per unit time per unit volume is typically of order @xmath27 for the critical field strength . as a possible environment ,
let us consider the pair creation of neutrinos with non - zero magnetic dipole moments from the very strongly magnetized compact objects with @xmath28 and radius @xmath29 . with @xmath30 as an effective volume of the magnetosphere , the production rate can be estimated@xcite as _ ~10 ^ 39 ( ) ^4 ( ) ^4()^3 / s. using the currently estimated masses of neutrinos the corresponding luminosity is estimated to be @xmath31 erg / s assuming @xmath32 , which is much weaker than the x - ray luminosity of magnetars , @xmath33erg / s .
however , since the time scale is much longer than the schwinger process by a factor of @xmath34 , it can be a continuous source of neutrinos .
a larger luminosity for observation can be easily obtained with a massive particle but the magnetic moment should be increasing to keep the critical field strength , which significantly constrains models for particles involved .
for the vacuum instability beyond critical field strength in case of the uniform magnetic field , the interpretation of vacuum instability as the neutrino pair production may not be the whole story .
it is because there may be a possibility that the instability could be avoided if a nontrivial condensation@xcite develops along with the critical magnetic field , which is subjected to further discussion
. the pair production of neutrinos in collider experiments can be enhanced if endowed with the nonvanishing magnetic dipole moment .
the pair production of neutrinos are expected from the annihilation of charged particles in colliding experiments through photon channel@xcite in addition to the weak boson channel in the standard model .
one of the interesting features of pauli coupling is that the angular distribution of pair creation peaks at @xmath35 .
it is compared to the angular distribution of the standard model process , minimum at @xmath35 but maximum for @xmath36 and @xmath37 . for the majorana neutrinos
, the pairs with different flavors can be produced due to the transit magnetic moment couplings .
the the angular distribution peaks at @xmath38 with respect to the beam direction , which is the enhancement of the neutrino production at right angle in contrast to weak z - channel .
it is interesting to note that the flavor mixing angle can be inferred if the flavor correlation in pairs can be measured in the future experiments@xcite .
the detectors located around the right angle to the beam direction can measure the back - to - back correlations in the pair production , where the production rate is supposed to be maximum .
most of the detectors are using electromagnetic triggers at the end stations , which implies corresponding charged leptons(@xmath39 and @xmath40 ) , for example electrons or muons , are those to be detected finally .
the majorana pairs in mass eigenstates , @xmath41 and @xmath42 , are produced through the pauli coupling and interact weakly with other particles to produce charged leptons to be detected .
to simplify the situation let us consider only two mass eigenstates which are mixed states of weak eigenstates , @xmath43 @xmath44 , _ 1 = _ + _ , _ 2 = - _ + _ where @xmath45 is the mixing angle .
now consider two targets a and b placed at the opposite side of the center of mass at the right angle to the beam direction . at the targets the neutrinos can produce the corresponding charged leptons via the charged current weak interaction(w - boson ) .
the back - to - back correlation , r , defined by the product of the detection rates of @xmath46 type leptons at the target a and @xmath47 type leptons at the target b , can be given by r = ( 1 + ^2 2 ) .
since the correlation @xmath48 turns out to be dependent only on the mixing angle , @xmath45 , it may serve as a clean measurement of mixing angles in the collider experiments .
for example the maximum value , @xmath49 , is obtained for the cases of no - mixing(@xmath50 ) , @xmath51 for the maximal mixing(@xmath52 ) .
these features are quite different aspect from standard model , which can be clearly distinguished in the high energy collider experiment@xcite . sine the strength of pauli coupling to the photon is determined by the neutrino magnetic dipole moment , which is very small , the pair creation cross section through photon channel by the pauli coupling is smaller that that of weak boson(@xmath11 ) channel until the center of mass energy of the colliding particle is sufficiently high .
to get some idea of the energy scale for which the pauli coupling becomes dominant , we can define the energy scale , @xmath53 , for which the total cross section becomes @xmath54 of the standard model , e_0.1 = 10 ^ 3 ( ) tev .
it is a bit higher than lhc energy , @xmath55 tev and also higher than the gzk cut - off energy of ultra high energy cosmic rays , @xmath56 tev .
so to see the distinguished feature from standard model , the higher energy collider experiments than present lhc seem to be needed .
the magnetic field is expected be created by the two electric currents in opposite directions generated by two colliding nuclei .
it turns out the field strength may be very large as @xmath57 g at rhic and lhc .
it gives @xmath58 times stronger pauli coupling than expected at the surface of the magnetars .
the magnetic vacuum instability due to magnetic dipole moment of neutrino can be much more enhanced . how to identify experimental observables corresponding to the vacuum instability and particle productions that are sensitive to the neutrino magnetic dipole moment
may be an interesting and challenging subject to be discussed .
the author would like to thank dmitry gitman , sang pyo kim and yongsung yoon for useful comments and discussions .
0 s.mereghetti , j.pons , a. melaos , space science review , 191 , 1315(2015 ) ; mcgill catalogue , http://www.physics.mcgill.ca/@xmath59 pulsar / magnetar / main.html h.k .
lee , r. wijers and g.e .
brown , phys .
325 , 83 ( 1999 ) .
huang , rep .
79 , 076302 ( 2016 ) .
j. schwinger , phys .
82 , 664(1951 ) .
see for example , g. dunne and t.m .
hall , phys .
b419 , 322(1998 ) ; phys .
d60 , 065002(1999 ) .
w. pauli , rev .
13 , 203(1941 ) .
lin , j. phys .
g. 25 , 1793(1999 ) .
lee and y.s .
yoon , jhep 0603 , 078(2006 ) .
s.p . gavrilov and d.m .
gitman phys . rev .
d 87 , 125025 ( 2013 ) .
kim , h.k .
lee and y.s .
yoon , phys.rev.d82,025015(2010 ) and references therein .
lee and y.s .
yoon , jhep 0703,086 ( 2007 ) .
koh , h.k .
lee , w - g .
paeng and y. yoon , j. korean phys .
56 , 1884 ( 2010 ) . c. giunti and a. studenkin , rev .
87 , 531(2015 ) .
h.k.lee , phys .
d84 , 077302 ( 2011 ) .
j. ambjorn , r.j .
hughes and n.k .
nielson , ann .
phys.(ny ) , 150 , 92(1983):j . ambjorn and p. olesen , nucl.phys.b 330 , 193(1990 ) and references therein .
nielson and p. olesen , nucl .
b 144 , 376(1978 ) . n. graham and r.l .
jaffe , phys.lett .
b435 , 145 ( 1998 ) .
lee and y. yoon , mod .
a 22 , 2081(2007 ) .
k. fujikawa and r. shrock , phys .
45 , 963 ( 1980 ) .
see for example , d. forero , m. tortola , j. valle , phys .
d 90 , 093006(2014 ) .
beda , v.b .
brudanin , v.g.egorov , d.v .
medvedev et al . , adv .
high energy phys .
2012 , 350150(2012 ) .
canas , o.g .
miranda , a .
parada , m. tortola and j.w.f .
valle , physics letters b 753 , 191(2016 ) h.k .
lee and y. yoon , mod .
a 23 , 17(2008 ) . | vacuum instability of the strong electromagnetic field has been discussed since long time ago .
the instability of the strong electric field due to creation of electron pairs is one of the examples , which is known as schwinger process .
what matters are the coupling of particles to the electromagnetic field and the mass of the particle to be produced .
the critical electric field for electrons in the minimal coupling is @xmath0 .
spin 1/2 neutral particles but with magnetic dipole moments can interact with the electromagnetic field through pauli coupling .
the instability of the particular vacuum under the strong magnetic field can be formulated as the emergence of imaginary parts of the effective potential . in this talk ,
the development of the imaginary part in the effective potential as a function of the magnetic field strength is discussed for the configurations of the uniform magnetic field and the inhomogeneous magnetic field .
neutrinos are the lightest particle(if not photon or gluon ) in the standard model " , of which electromagnetic property is poorly known experimentally .
recently the observation of neutrino oscillation shows the necessity of neutrino masses .
it implies that the standard model is subjected to be modified such that non - trivial electromagnetic structure of neutrino should be reconsidered although they are assigned to be neutral . and
the possibility of anomalous electromagnetic form factor is an open question theoretically and experimentally . in this talk ,
the implication of non - vanishing magnetic dipole moment of neutrinos is also discussed : the instability of the strong magnetic field and the enhancement of neutrino production in high energy collider experiments . |
the standard model of particle physics , described by the gauge group su(3)@xmath7su(2)@xmath7u(1 ) is one of the most successful physical theories ever . nonetheless , the standard model is theoretically unsatisfying for a number of reasons .
for instance , the instability of the weak scale against radiative corrections has motivated the study of numerous theories , including technicolor , supersymmetry and theories with additional dimensions . while the gauge hierarchy may be the most compelling motivation for new physics , there are additional reasons to consider theories with further structure .
it was long ago realized that the gauge group of the standard model could be embedded into a simple group @xcite .
the appeal of this idea was substantiated by measurements of electroweak observables which suggested that the values of the standard model gauge couplings unify at a scale @xmath8 @xcite .
as data became increasingly precise , it became apparent that the simplest grand unified theory ( gut ) , minimal su(5 ) , predicted a value for @xmath9 that was incompatible with observation .
however , the combination of supersymmetry and grand unification @xcite has met with great success in predicting @xmath9 @xcite . for these supersymmetric theories , the gut symmetry must be broken at a somewhat higher scale , @xmath10 .
one smoking gun signal of grand unification is proton decay , which arises from dimension six operators generated by exchange of the @xmath1 and @xmath2 gauge bosons , and , in supersymmetric theories , from dimension five operators generated by colored higgsino exchange @xcite . in supersymmetric guts , the dimension six operators typically do not lead to observable proton decay , but the dimension five operators do , and place strong limits on these theories @xcite .
recently ref .
@xcite studied the dimension five proton decay and found that even large masses for the first - two generation sparticles can not save minimal supersymmetric su(5 ) .
they found that successful gauge coupling unification requires that the higgs triplet mass fall in the range @xmath11 .
however , even with decoupling of the superpartners , the limit from proton decay was shown to require @xmath12 . while additional field content may generate threshold effects to rectify this , in this paper we propose an alternative possibility .
we will embed a unified gauge theory into five dimensions in which gauge and higgs fields propagate in the additional dimension .
we will show that if the gut is broken on a brane at a high scale ( _ i.e. _ , @xmath0 , the planck scale of the five - dimensional theory ) , the wave functions of the gauge fields and higgs triplets can develop an approximate node on the brane , suppressing proton decay operators arising from their exchange .
we will show that the running above the compactification scale is not only consistent with , but is in fact equivalent to ordinary four dimensional logarithmic unification with the compactification scale identified as the mass of the higgs triplet . in doing so , we reformulate the question of the @xmath13 hierarchy because the breaking occurs at the fundamental scale of the theory ( which , in general , is still larger than @xmath14 ) . in its place
we require a somewhat large hierarchy @xmath15 between the planck and compactification scales .
the traditional gut scale , @xmath16 , is a derived scale in this framework : there is no new physics there .
because of the @xmath4 breaking of the grand unified group on the brane , we will also easily understand deviations from su(5 ) predicted relationships among yukawa couplings .
we consider a 5d supersymmetric su(5 ) theory with the fifth dimension compactified on an @xmath17 orbifold ( in section [ sec : so(10 ) ] we will consider the possibility of larger gauge groups ) .
the bulk exhibits n=1 supersymmetry in five dimensions , which translates into an n=2 supersymmetry in four dimensions .
the fields that propagate in the bulk are contained in a gauge supermultiplet and two higgs hypermultiplets transforming as @xmath18 and @xmath19 under su(5 ) . under the 4d n=1 supersymmetry preserved after the orbifold compactification ,
the gauge supermultiplet decomposes into a vector superfield @xmath20 and an associated chiral adjoint @xmath21 .
the two hypermultiplets yield four chiral multiplets @xmath22 , @xmath23 , @xmath24 and @xmath25 .
( @xmath22 and @xmath25 transform as @xmath18 under the su(5 ) , while @xmath23 and @xmath24 as @xmath26 . ) under the orbifold @xmath27 , which flips the sign of the fifth coordinate ( @xmath28 ) , the various superfields transform as @xmath29 which leaves zero modes only for the mssm fields and their gut counterparts .
the @xmath17 orbifold has fixed points at @xmath30 and @xmath31 .
we take chiral matter to be localized to the @xmath31 fixed point , along with an adjoint chiral superfield @xmath32 .
the su(5 ) symmetry is broken by the vacuum expectation value ( vev ) of @xmath32 , which we assume to have the form @xmath33 we imagine that this breaking occurs at or near the five - dimensional planck scale @xmath0 , which we assume to be much larger than the conventional gut scale of @xmath16 .
we will refer to the @xmath31 fixed point as the gut - breaking brane .
bulk fields even under the orbifold @xmath27 can couple directly to the fields on the gut - breaking brane .
for instance , the higgs fields can couple to the chiral matter fields @xmath34 and @xmath35 as @xmath36 in the superpotential .
we also take there to be a bare higgs mass term , @xmath37 , as well as higgs coupling to the gut breaking , @xmath38 , in the superpotential . while it would be interesting to study standard solutions to the doublet - triplet splitting problem in this framework
, we will not endeavor to do so here .
rather , we will simply tune these two contributions against one another so that the su(2 ) doublets @xmath39 @xmath40 and @xmath41 @xmath42 are nearly massless , while the colored triplets @xmath43 @xmath40 and @xmath44 @xmath42 end up with a large mass term @xmath45 .
note that @xmath46 is a dimensionless parameter since @xmath43 and @xmath44 are bulk fields and have dimension @xmath47 .
here , @xmath46 arises as a sum of terms ( with higher order terms potentially relevant in the strongly coupled theory ) .
in addition to the bare mass term , the @xmath48 term contributes .
the natural size of this operator is restricted only by perturbativity , with an upper bound @xmath49 suggested by naive dimensional analysis in higher dimensions @xcite .
once the gauge symmetry is broken , the @xmath1 and @xmath2 bosons acquire brane - localized masses through the interaction @xmath50 where @xmath51 indexes the broken generators of the group and @xmath52 is the 5d gauge coupling , with mass dimension @xmath53 .
if we were to decompose each @xmath1 and @xmath2 bulk vector field in a naive kaluza - klein ( kk ) basis , we would estimate that the zero modes have mass @xmath54 .
such an estimate incorrectly assumes that the gauge boson wave functions do not change appreciably in the presence of the large localized mass term . to find the correct spectrum we must solve the differential equation @xmath55 which leads to kk masses @xmath56 given by @xcite @xmath57 where @xmath58 .
for @xmath59 this equation gives a spectrum whose low - lying levels are approximately @xmath60 ( with @xmath61 ) , whereas the usual kk spectrum is @xmath62 .
we see that the kk tower has been shifted up one half unit .
this is easy to understand intuitively .
the gauge field picks up a mass of @xmath63 if it does not avoid the brane .
if it avoids the brane entirely , it does not `` see '' the mass term , and picks up a smaller mass of @xmath64 from a nontrivial profile of the wave function in the extra dimension .
the localized mass term acts merely to dynamically assign the boundary condition that the wave functions should vanish at the gut - breaking brane .
the boundary condition is not absolute , of course , so that the spectrum is not precisely spaced according to @xmath65 .
the value of the wave function at the gut - breaking brane goes as @xmath66 , and there are corrections to the mass which go like @xmath67 . modes with masses near or above the scale @xmath68 become essentially degenerate with the ordinary kk tower , @xmath69 .
we have not broken supersymmetry , and thus the gauginos corresponding to the broken su(5 ) generators have an identical spectrum .
the kk towers for the gauge fields are illustrated in fig .
[ fig : spectrum]a . for each unbroken su(3)@xmath7su(2)@xmath7u(1 ) ( 3 - 2 - 1 ) generator there is a massless vector multiplet and massive vector multiplets at @xmath70 , @xmath71 , @xmath72 , and so on . for each broken ( @xmath73 ) generator
there is a tower of massive vector multiplets whose low levels are offset from those of the 3 - 2 - 1 tower by @xmath74 , but whose higher levels relax into alignment with the 3 - 2 - 1 tower .
although in fig .
[ fig : spectrum]a this transition to degeneracy has already been achieved at @xmath75 , for the parameters of interest in our model , the relaxation will actually occur much more gradually , over the span of @xmath76 kk levels .
note that for each @xmath73 massive vector multiplet but one there is a corresponding 3 - 2 - 1 massive vector multiplet .
the remaining one is paired with the _ massless _ 3 - 2 - 1 vector multiplet , indicating a larger total number of states in the @xmath73 tower relative to the 3 - 2 - 1 tower .
these excess states are the eaten goldstone degrees of freedom from the @xmath32 multiplet .
this means that above the scale @xmath77 the kk towers of gauge fields contributing to the renormalization group ( rg ) running are completely su(5 ) symmetric except that one chiral adjoint for each 3 - 2 - 1 vector multiplet is missing .
these missing degrees of freedom are provided by the physical @xmath32 fields which remain after the goldstone components of @xmath32 are eaten by the @xmath73 gauge multiplets .
therefore , above the scale @xmath78 and @xmath79 ( the mass of physical @xmath32 fields ) , the spectrum contributing to the running is completely su(5 ) symmetric .
( 400,460)(-20,-60 ) ( -20,400)[b](a ) ( 5,240)(400,240 ) ( 10,230)(10,400 ) ( 0,240)[r]@xmath80 ( 8,255)(12,255 ) ( 8,270)(12,270 ) ( 6,270)[r]@xmath81 ( 8,285)(12,285 ) ( 8,300)(12,300 ) ( 6,300)[r]@xmath82 ( 8,315)(12,315 ) ( 8,330)(12,330 ) ( 6,330)[r]@xmath83 ( 8,345)(12,345 ) ( 8,360)(12,360 ) ( 6,360)[r]@xmath84 ( 8,375)(12,375 ) ( 8,390)(12,390 ) ( 6,390)[r]@xmath85 ( 60,215)[b]@xmath86 ( 60,190)[b]@xmath87 ( 35,240)(85,240 ) ( 60,240)3 ( 190,215)[b]@xmath88(190,190)[b]@xmath89 ( 165,270)(215,270 ) ( 190,270)3 ( 165,300)(215,300 ) ( 190,300)3 ( 165,330)(215,330 ) ( 190,330)3 ( 165,360)(215,360 ) ( 190,360)3 ( 165,390)(215,390 ) ( 190,390)3 ( 340,215)[b]@xmath90(340,190)[b]@xmath91 ( 315,255)(365,255 ) ( 340,255)3 ( 315,282)(365,282 ) ( 340,282)3 ( 315,309)(365,309 ) ( 340,309)3 ( 315,336)(365,336 ) ( 340,336)3 ( 315,363)(365,363 ) ( 340,363)3 ( 315,390)(365,390 ) ( 340,390)3 ( -20,160)[b](b ) ( 5,0)(400,0 ) ( 10,-10)(10,160 ) ( 0,0)[r]@xmath80 ( 8,15)(12,15 ) ( 8,30)(12,30 ) ( 6,30)[r]@xmath81 ( 8,45)(12,45 ) ( 8,60)(12,60 ) ( 6,60)[r]@xmath82 ( 8,75)(12,75 ) ( 8,90)(12,90 ) ( 6,90)[r]@xmath83 ( 8,105)(12,105 ) ( 8,120)(12,120 ) ( 6,120)[r]@xmath84 ( 8,135)(12,135 ) ( 8,150)(12,150 ) ( 6,150)[r]@xmath85 ( 60,-25)[b]@xmath92 ( 60,-50)[b]@xmath93 ( 35,0)(85,0 ) ( 60,0)3 ( 190,-25)[b]@xmath94 , @xmath95 ( 190,-50)[b]@xmath96 ( 165,30)(215,30 ) ( 190,30)3 ( 165,60)(215,60 ) ( 190,60)3 ( 165,90)(215,90 ) ( 190,90)3 ( 165,120)(215,120 ) ( 190,120)3 ( 165,150)(215,150 ) ( 190,150)3 ( 340,-25)[b]@xmath97 , @xmath98 ( 340,-50)[b]@xmath99 ( 315,15)(365,15 ) ( 340,15)3 ( 315,45)(365,45 ) ( 340,45)3 ( 315,75)(365,75 ) ( 340,75)3 ( 315,105)(365,105 ) ( 340,105)3 ( 315,135)(365,135 ) ( 340,135)3 the spectrum analysis for the higgs triplet fields @xmath43 and @xmath44 is somewhat different , because the brane coupling @xmath46 is dimensionless , and hence the naive mass that the zero modes would pick up by not avoiding the brane is just @xmath100 which is comparable to @xmath101 for order one @xmath46 . for our purposes , we will take @xmath46 to be a large parameter , roughly @xmath102 , so that the higgs wave functions will in fact be strongly repelled from the brane .
this size for @xmath46 is quite consistent with the general features of the theory , which must be somewhat strongly coupled in order to achieve order one top yukawa and gauge couplings in spite of the large radius .
the equations that determine the colored higgsino spectrum are @xmath103 @xmath104 @xmath105 @xmath106 we can analyze these equations by writing the solutions as @xmath107 where @xmath108 and @xmath109 are even under @xmath28 , while @xmath110 and @xmath111 are odd . if we integrate the first two of these equations in a region @xmath112 where @xmath113 , we obtain two constraints , @xmath114 @xmath115 here , two constraints must be satisfied mode by mode .
let us define @xmath116 where @xmath117 take @xmath118 , and identify both @xmath119 and @xmath120 .
this allows us to rewrite eqs .
( [ eq : higgsinoeqs1 ] ) ( [ eq : higgsinoeqs4 ] ) as ( neglecting singular terms ) @xmath121 @xmath122 @xmath123 @xmath124 which give solutions @xmath125 with @xmath126 defined by @xmath127 thus , the kk masses for the colored higgs are given by @xmath128 where @xmath129 runs from negative infinity to positive infinity ( the physical masses are given by the absolute value of @xmath126 ) .
for large @xmath46 , @xmath130 , so @xmath131 , and the masses fall into nearly degenerate pairs , @xmath132 . note that , in contrast to the case of the gauge fields , even for high values of @xmath129 , the triplet higgsinos do not become degenerate with the doublet higgsinos .
supersymmetry fixes the masses and couplings of the higgs scalars to be the same as for the higgsinos .
the kk towers for the higgs fields are shown in fig .
[ fig : spectrum]b .
the doublet spectrum consists of a pair of massless @xmath133 chiral multiplets , and pairs of exactly degenerate hypermultiplets at @xmath70 , @xmath71 , @xmath72 , and so on . for large @xmath46 ,
the triplet spectrum consists of nearly degenerate pairs of hypermultiplets at @xmath74 , @xmath134 , @xmath135 , and so on . in fig .
[ fig : spectrum]b , the non - degeneracy between these pairs is not resolved .
a crucial question in our model is how the gauge couplings evolve above the compactification scale . to address this question , it is useful to focus on the `` differential running '' , _
i.e. _ , the non - uniform evolution of the gauge couplings . above @xmath74
a whole tower of modes contributes to the evolution of @xmath136 and @xmath137 .
however , we should emphasize that _ this scenario does not employ power law unification .
_ in contrast with ref .
@xcite , the overwhelming power - law contribution to the running above @xmath138 is su(5 ) universal , so it is useful to focus only on the quantities which distinguish the couplings and lead to non - uniform evolution above @xmath74 .
for instance , the matter fields fall into complete su(5 ) multiplets so their effects only change the coupling value at unification , but do not influence whether and at what scale the couplings unify .
in contrast , the presence of the higgs doublets leads to non - uniform running .
we will see shortly that above @xmath138 , the differential running arises from a sum of a large number of threshold effects .
let us define the scale - dependent quantities @xmath139 with @xmath140 vanishing trivially .
the evolution of @xmath141 above @xmath74 ( but below @xmath142 ) takes the form @xmath143 where @xmath144 represents contributions arising from higgs loops and @xmath145 represents contributions arising from gauge loops .
contributions from matter loops vanish because they contribute universally to the running . using the known spectra for the various kk towers and their beta function contributions ( shown in fig .
[ fig : spectrum ] ) , we can calculate these quantities . for the higgs contributions we find @xmath146 @xmath147 here the @xmath126 are the colored higgs masses given in eq .
( [ eq : higgsspectrum ] ) .
recall that there are actually two slightly non - degenerate @xmath148 near each explicitly shown colored higgs level in fig .
[ fig : spectrum]b . for the gauge contributions we have @xmath149 @xmath150 where @xmath151 are the @xmath73 vector multiplet masses , given by the solutions to eq .
( [ eq : gaugespectrum ] ) .
the running of @xmath152 is given by @xmath153 let us consider the higgs and gauge contributions to this running separately , starting with the higgs loops first .
we have @xmath154,\ ] ] and @xmath155.\ ] ] here we mean the number of doublet and triplet chiral superfields : there are two massless doublets , four triplets with mass @xmath156 , four doublets with mass @xmath70 , and so on .
both of the above expressions have the expected @xmath157 characteristic of logarithmic running , but have coefficients which _ average out to zero_. below @xmath74 only the massless @xmath39 and @xmath41 doublets contribute , so @xmath158 ( @xmath159 ) runs in the positive ( negative ) direction . at @xmath74
we encounter four triplet chiral superfields and @xmath158 ( @xmath159 ) now runs in the negative ( positive ) direction .
the directions reverse again when we gain four doublets at @xmath70 , and again at @xmath134 when we gain another four triplets .
these threshold effects continue , but become increasingly negligible . taking the triplet masses to be @xmath160 , the threshold effects from the states between @xmath161 and @xmath162 are proportional to @xmath163 which vanishes as @xmath164 for large @xmath129
. function as in ref .
@xcite , we would still find the high mass threshold effects to be negligible . ]
we conclude that the differential running due to the higgs fields dies off quickly above the compactification scale .
thus the total higgs contribution to the running between @xmath74 and a high scale such as @xmath142 or @xmath78 is essentially independent of the high scale . taking the pairs of triplet hypermultiplets to be exactly degenerate
, we find that for high scales , @xmath165 and similarly , @xmath166 note that the ratio @xmath167 is the same as one would have calculated by including the contributions from the massless doublets alone .
this fact will be important when we compare with the running in 4d minimal su(5 ) in section [ subsection : uamsu5 ] .
in contrast to the higgs threshold effects , those arising from gauge field loops do not quickly die off .
in fact , we will see that these effects add up to an effective logarithmic running all the way up to @xmath142 or @xmath78 .
the definitions we have used for @xmath168 are very convenient for making connection to ordinary , four - dimensional running .
in particular , @xmath169 receives no gauge contribution below the gut scale in four - dimensional theories , and likewise here @xmath170 will not receive contributions to differential running .
@xmath171 and @xmath172 , on the other hand , _ will _ receive corrections above the compactification scale . referring to eqs .
( [ eq : r2 g ] ) and ( [ eq : r3 g ] ) , we see that the quantities @xmath173 are proportional to the quadratic casimir coefficients @xmath174 of the gauge groups .
thus , we can write the gauge contributions as @xmath175 where @xmath174 is 2 for su(2 ) and 3 for su(3 ) .
furthermore , we have @xmath176.\ ] ] the notation here is as in fig .
[ fig : spectrum]a .
note that in contrast to the case for the higgs multiplets , here the coefficient multiplying @xmath157 does not average to zero .
rather , for low modes ( such that @xmath177 ) , the coefficient is 3 up to @xmath74 , 1 between @xmath74 and @xmath70 , then 3 again until @xmath134 , and so on . for
the higher mass modes , such that the @xmath73 and 3 - 2 - 1 multiplets are nearly degenerate , the coefficient becomes fixed at 1 . if the running were coming entirely from the massless 3 - 2 - 1 vector multiplet , the coefficient multiplying @xmath157 would be 3 , so we see that above the compactification scale , the differential running due to the gauge loops is slowed somewhat relative to the ordinary logarithmic evolution in 4d .
note that even once the @xmath73 and 3 - 2 - 1 multiplets become degenerate , the differential running continues . in this regime ,
the running is due entirely to the eaten goldstone states contained in the @xmath73 kk tower .
the differential running stops completely only once we reach the scale @xmath142 , when the rest of the @xmath32 degrees of freedom begin to propagate in the loops . happens to be below the the scale at which the @xmath73 and 3 - 2 - 1 towers become degenerate ( determined by @xmath78 ) , the differential running due to the goldstone degrees of freedom ceases above @xmath142 , but the total differential running does not die off until a scale @xmath178 . ]
the qualitative picture for gauge coupling unification in our model is depicted in fig .
[ fig : unification ] . above the compactification scale ,
the non - uniform evolution of the gauge couplings slows , so that unification occurs at a larger scale than usual
. the unification scale will essentially be the larger of @xmath142 and @xmath179 .
next we will verify that this picture is in fact correct , and we will make more precise the connection to unification in ordinary minimal su(5 ) in 4d .
( 320,200)(-20,-40 ) ( 0,150)(290,150 ) ( 200,136.7)(240,150)3 ( 200,125)(240,150)3 ( 0,70)(200,136.7 ) ( 0,0)(200,125 ) ( 200,136.7)(290,150 ) ( 200,125)(290,150 ) ( -10,-5)[b]@xmath180 ( -10,65)[b]@xmath181 ( -10,145)[b]@xmath140 ( 290,150)(315,150 ) ( -5,-20)(320,-20 ) ( 320,-35)[b]@xmath182 ( 0,-24)(0,-16 ) ( 200,-24)(200,-16 ) ( 240,-24)(240,-16 ) ( 290,-24)(290,-16 ) ( 0,-38)[b]@xmath183 ( 200,-38)[b]@xmath70 ( 240,-38)[b]@xmath184 ( 290,-38)[b]@xmath78 up to this point we have discussed the differential evolution of the couplings above the compactification scale , but have not explicitly demonstrated that the couplings in our model unify in a manner consistent with electroweak scale measurements of the 3 - 2 - 1 couplings . here we will demonstrate that , at one loop , the successful prediction of @xmath185 of minimal su(5 ) with @xmath186 @xcite is equivalent to the successful prediction of @xmath185 in this model with @xmath187 .
that is , the fact that minimal su(5 ) with a triplet higgs below the gut scale can give @xmath185 correctly _ guarantees _ that this model can do as well . to see this
we must reexamine the running of gauge couplings in four - dimensional , minimal su(5 ) . above the higgs triplet mass ,
the gauge couplings evolve as @xmath188 to make connection with our discussion of differential running in the five - dimensional theory , we subtract off a universal piece @xmath189 from each gauge coupling , and we find @xmath190 where @xmath184 is the conventional gut scale @xmath191 gev . below the scale @xmath192 ,
the running of the gauge couplings in the 5d theory is the same as for the 4d theory .
thus , if we can demonstrate that @xmath193 has the same form as eq .
( [ eq:4dunif ] ) , we will have shown that successful unification in ordinary minimal su(5 ) implies successful unification in our model .
here the superscript in @xmath194 is to emphasize that it is the colored higgs mass that gives successful unification in the 4d theory that is meant , rather than the lightest kk mode of the colored higgs in the 5d theory .
the higgs contributions to @xmath193 are @xmath195 where @xmath196 .
the first term comes from the ordinary running due to @xmath39 and @xmath41 below @xmath74 , and the second term is the finite running above @xmath74 found in eqs .
( [ eq : higgsrun2 ] ) and ( [ eq : higgsrun3 ] ) .
since the two terms are both proportional to @xmath197 , we see that the higgs contribution to the differential running in 4d ( which takes place entirely below @xmath192 ) is emulated in 5d with the proper choice of compactification scale . had the differential running due to the higgs multiplets turned off right at @xmath74 , we would simply identify @xmath198 . because we find a small amount of differential running above @xmath74 , we instead identify @xmath199 .
we expect small additional corrections to this quantity from further finite one - loop effects , but the precise value @xmath200 is not important : what matters is that , given any value of @xmath192 that leads to a successful prediction of @xmath9 in the 4d theory , there is a value for @xmath70 ( @xmath201 ) such that the total differential running due to higgs loops in the 5d theory will be the same as in the 4d theory .
having chosen this value for @xmath70 , @xmath193 arises entirely from loops of gauge and physical @xmath32 fields : @xmath202 where the first term comes from the ordinary 4d running below @xmath74 , and two terms in the curly bracket represent the contributions from gauge kk towers and physical @xmath32 fields , respectively ( for the parameterization of the gauge contribution , see eq .
( [ eq : delta ] ) ) .
we find that this has the same form as eq .
( [ eq:4dunif ] ) , with the required identification @xmath203 note that @xmath204 under @xmath205 , where @xmath206 is some mass scale which is a function of @xmath70 and @xmath78 , so that the quantity in the left - hand side has a well - defined value which is a function of @xmath207 and @xmath142 .
( @xmath78 is the scale at which the differential running almost ceases in the 5d theory . )
we see that _ the form of differential running above the higgs triplet mass in minimal su(5 ) is precisely what arises from threshold effects in the case of the five dimensional theory .
_ this holds irrespective of whether the step - function approximation of fields in the rg running is very good or not .
for instance , had we used the full one - loop calculation as in ref .
@xcite , all gauge contributions would still be proportional to @xmath208 . at one loop ,
the structure of unification is equivalent to that of minimal su(5 ) . that this model can successfully yield @xmath185 is no more nor less remarkable than in four - dimensional unified theories .
the remarkable feature is that the grand unified group is broken `` maximally '' , at a scale well above the masses of the lightest @xmath1 and @xmath2 gauge bosons .
this feature changes the predictions of the theory , as we will see in sections [ sec : protondecay ] and [ sec : yukawas ] .
let us make one final comment regarding the differential running of the couplings . in a sense , its logarithmic ( as opposed to power law ) behavior is natural .
the power law evolution is characteristic of the bulk ( _ i.e. _ , five dimensional ) nature of the gauge coupling .
in contrast , the su(5 ) breaking occurs strictly on a brane . simply by lorentz invariance
, the brane breaking can not contribute to the bulk coupling , but instead should just generate a brane contribution to the gauge kinetic term .
consequently , we would not _ a priori _ have expected the effects of this su(5 ) breaking to exhibit a power law behavior .
now that we have shown the direct correspondence between the differential running in our model to that of ordinary minimal su(5 ) , we can determine the scale at which successful unification occurs .
in other words , given a compactification scale @xmath101 , we can find the values of @xmath142 and @xmath209 that give @xmath9 correctly .
the allowed compactification scales are already known : these are related to the allowed values of @xmath210 in minimal su(5 ) by @xmath211 .
then we need only to find the values of @xmath142 and @xmath209 for which eq .
( [ eq : connect ] ) holds .
these values are indicated in fig .
[ fig : allowedscales ] for compactification scales between @xmath212 and @xmath213 ( corresponding to @xmath214 gev and @xmath215 gev ) . for these compactification scales ,
the 5d planck scale @xmath0 is @xmath216 gev , and fig .
[ fig : allowedscales ] shows that the gut - breaking scale can be at or near this scale and the unification of three gauge couplings can be attained below @xmath0 depending on the parameters of the model . or @xmath78 is above the 5d planck scale , the gauge unification is not completed in a field theoretic regime . in this case ,
our field theoretic treatment is not fully trustable above @xmath0 and would have some uncertainties coming from the cutoff scale physics . ] in the parameter region where the field theoretic unification works , the ratio of the cutoff to the compactification scales is @xmath217 , and no gauge or yukawa couplings become nonperturbative below the cutoff scale .
thus , our one - loop treatment of the running is well justified .
we also find that @xmath209 must be somewhat smaller than @xmath0 in this parameter region .
this is consistent with the observation that the theory is more or less strongly coupled at the scale @xmath0 to have @xmath4 yukawa and gauge couplings in 4d ; if the theory is strongly coupled at @xmath0 , @xmath209 is naturally @xmath218 , since the superpotential giving the vev of @xmath32 scales like @xmath219 .
even then , however , the @xmath32 couples to the other fields in the combination of @xmath220 , so that various operators feel order one gut breaking , since @xmath221 .
therefore , we treat as if @xmath32 has a vev of order @xmath0 in the following discussions , although all the arguments also apply in the strongly coupled case , @xmath218 .
we finally comment on effects from brane operators like @xmath222 where @xmath223 and @xmath224 is the field strength superfield .
although it is possible that these operators are somehow suppressed at the cut - off scale , from effective field theory point of view it is generically expected that they are present with order one coefficients .
they give tree - level splitting of three gauge couplings for su(3 ) , su(2 ) and u(1 ) , and could affect the previous analysis of the gauge coupling unification . in particular , since we are considering @xmath225 , one might think that they give @xmath4 correction to @xmath9 .
however , we can expect that the effect of these operators are actually smaller by making the following observations .
as an example , let us first consider the extreme case where all the interactions are strongly coupled at the scale @xmath0 .
then , the operators involving the field strength superfield scale as @xmath226,\ ] ] where @xmath227 is an @xmath4 quantity .
( strictly speaking , @xmath227 must be somewhat smaller than 1 so that all higher dimensional operators involving @xmath228 do not equally contribute and make the theory unpredictable . ) on integrating over @xmath229 , the zero - mode 4d gauge couplings @xmath230 are given as @xmath231 here , the first and second terms are su(5)-preserving and su(5)-violating contributions , respectively .
since we know that @xmath232 , we have to take @xmath233 in this strongly coupled case .
this shows that the su(5)-violating contribution coming from brane operators is suppressed by a factor of @xmath234 in this case .
in fact , the theory is not truly strongly coupled at the scale @xmath0 in the realistic case discussed in previous sections so that the one - loop treatment of gauge coupling evolutions is reliable .
nevertheless , the above argument applies more generically ; the su(5)-violating brane contribution is small relative to the su(5)-preserving bulk contribution due to the large volume factor @xmath235 .
thus , the correction to @xmath9 is expected to be small .
we will not discuss possible effects of the brane operators further , and assume that they are negligible in the subsequent discussions .
one of the key signals of grand unification is proton decay . @xmath1 and @xmath2 gauge boson exchange generates dimension six proton decay operators in the low energy theory , and higgsino triplet exchange generates dimension five operators .
one might expect that since the particles appear at a scale @xmath236 , this will be the suppression scale of the proton decay operators .
one interesting possibility offered by this framework is that _ the rate of proton decay is not constrained by gauge invariance to be related to the strength of gauge and yukawa interactions in the usual way . _
we have already noted that the dangerous particles have wave functions at the @xmath31 fixed point that are small compared with those of their 3 - 2 - 1 or doublet counterparts . on the other hand
, there is a tower of states that can mediate proton decay , and we must sum each contribution . in this section ,
we investigate proton decay operators generated by exchanges of @xmath237 gauge bosons and colored higgsinos , and show that the present model can satisfy the constraints coming from experimental lower bounds on the proton lifetime .
dimension six proton decay operators arise from the exchange of @xmath1 and @xmath2 gauge bosons .
the coupling of these bosons to fields on the brane is suppressed by a factor @xmath238 .
comparison with four - dimensional theories can be made by replacing @xmath239 where @xmath151 are the masses of the @xmath1 and @xmath2 gauge boson kk modes .
these masses satisfy eq .
( [ eq : gaugespectrum ] ) , which allows us to approximate the sum as @xmath240 thus , the `` effective mass '' of the @xmath1 and @xmath2 bosons is @xmath241 . for @xmath209 at or near the five dimensional planck scale , this will typically be larger than @xmath184 , although it is not required to satisfy experimental constraints .
in any case , despite the fact that the lightest @xmath1 and @xmath2 gauge bosons have masses of only @xmath242 gev , it is easy to satisfy proton decay constraints coming from dimension six operators in this model . dimension five operators come from integrating out the higgsino triplets .
again , we make connection with the four dimensional theory by finding an effective triplet higgsino mass by taking the whole sum over kk modes . here
we make the replacement @xmath243 where @xmath126 are the masses of the colored higgs kk modes , given in eq .
( [ eq : higgsspectrum ] ) . using this spectrum
, we obtain the sum @xmath244 thus , proton decay from dimension five operators is suppressed by a factor @xmath245 compared to the compactification scale . in ordinary minimal su(5 ) , proton decay limits require @xmath246 gev , so for a compactification scale of @xmath247 , we would need a @xmath46 parameter of ten to adequately suppress proton decay operators .
this leaves open the possibility that proton decay could be observed in future experiments .
however , for larger @xmath46 detection becomes increasingly unlikely .
in addition to the ordinary yukawa couplings between the higgs and matter fields on the brane , there can be derivative couplings of the conjugate higgs fields to the matter fields as well .
these operators such as @xmath248 can lead to proton decay .
of course , coefficients @xmath249 and @xmath250 of these operators are not related to the usual yukawa couplings , @xmath251 and @xmath252 , by su(5 ) , so that their actual significance is unknown .
furthermore , there is an ambiguity in what we mean by the derivative of the conjugate field , which is not differentiable at the point @xmath31 .
however , using eqs .
( [ eq : higgsinoeqs1 ] ) and ( [ eq : higgsinoeqs2 ] ) , we can rewrite the coupling as @xmath253 where the summation over @xmath254 arises from rewriting the @xmath255-function as a sum of the kk mode . while we can compute proton decay diagrams involving these vertices , it is now apparent that such diagrams have a strong dependence on how we cut off the sum of the kk mode . if the cutoff is done near the fundamental scale , these diagrams can , at least in principle , give comparable contribution to those involving the usual yukawa couplings . however , it is not entirely clear what the cutoff for these diagrams should be . in the case where the brane is dynamical ,
the summation in the above operators are cut off at the scale of the brane tension @xcite . for couplings at the orbifold fixed point , which is not dynamical and can not fluctuate
, it is unclear whether the cutoff is the fundamental scale or a lower scale , such as the radion mass . for our purposes here
, we will not address these issues further . since we can not _ a priori _ know the size of the couplings of these operators and their flavor structure , estimating the resulting proton decay rate is already very uncertain .
however , it is possible that such operators may provide an opportunity for detectable dimension five proton decay in the near future .
so far , we have a framework that looks like ordinary su(5 ) except with suppressed proton decay .
while the breaking occurs at a scale near the 5d planck scale @xmath0 , the @xmath237 gauge bosons are much lighter .
still , in a very real sense , the breaking of su(5 ) is `` maximal '' , which can manifest itself in deviations from su(5 ) expectations .
we here consider the effects of such maximal gut breaking on the fermion yukawa couplings . in the minimal su(5 ) , one important prediction is the unification of the yukawa couplings
. a successful prediction of the theory is the unification of the bottom and @xmath256 yukawa couplings at the gut scale @xcite .
however , it is well known that the su(5 ) relations fail in the lighter first - two generations . for instance , an su(5 ) relation @xmath257 fails by a factor of ten
. terms to the yukawa couplings @xcite . ] in ordinary 4d su(5 ) gut , it has been suggested that the operators involving the @xmath32 field can correct this discrepancy @xcite .
however , this mechanism does not work in most theories where the fermion mass hierarchy is explained by the froggatt - nielsen mechanism @xcite . in this mechanism ,
the matter fields carry generation dependent flavor u(1 ) charges and the various yukawa couplings are generated through the u(1 ) breaking spurion .
this is an attractive mechanism in that it not only suppresses the first - two generation yukawa couplings but also suppresses dangerous tree - level dimension five proton decay operators .
however , having employed this mechanism , the gut - breaking operators involving @xmath209 can modify su(5 ) mass relations only by an amount suppressed by a factors of @xmath13 , which is too small to accommodate order one deviations suggested by the observed quark and lepton masses .
field @xcite . ] in the present model , on the other hand , there is no suppression of higher dimensional operators involving gut - breaking effects , since @xmath209 is near the cut - off scale @xmath0 . in this sense , fields living on the gut - breaking brane
see the gut broken maximally , and their yukawa couplings need not respect the su(5 ) symmetry .
thus , the failure of su(5 ) to describe @xmath258 seems quite natural . in this framework , however , the success of the @xmath259-@xmath260 unification must be viewed as an accident , unless there is a reason why the coupling of @xmath32 to the third generation is somewhat suppressed .
up to this point , we have restricted our attention to a scenario with su(5 ) gauge group , but it is interesting to consider larger gauge groups .
one possible difference comes from additional matter fields required by larger gauge groups .
for instance , in so(10 ) there is an additional state , right - handed neutrino .
if this additional field acquires a mass around the cut - off scale @xmath261 from so(10 ) breaking , then there will be a running effect between @xmath101 and @xmath261 which is not so(10 ) universal .
it is , however , important that this does not affect our previous analyses ; the 3 - 2 - 1 unification still works as in section [ sec : diffrunning ] even in the case of so(10 ) .
the additional higgs states do not contribute to differential running above @xmath101 , and the threshold effects are proportional to 3 - 2 - 1 @xmath262-functions .
this means that we can have larger gauge groups broken all the way to 3 - 2 - 1 around the cut - off scale . in the so(10 ) case , this may result in too large right - handed neutrino masses to give an appropriate mass scale for atmospheric oscillations @xcite through see - saw mechanism @xcite . then , we may need somewhat small coefficient in front of the operator which gives right - handed neutrino masses or have to resort to other ways to generate neutrino masses within supersymmetric models , for instance , though @xmath263-parity violation @xcite or supersymmetry breaking @xcite .
the possibility that the standard model gauge group is merely a subgroup of a larger , simple group is an attractive one .
unfortunately , the simplest version of supersymmetric su(5 ) predicts proton decay at rates incompatible with experiment .
we have demonstrated that the incorporation of just one new ingredient an additional dimension in which gauge and higgs fields propagate brings about crucial changes relative to ordinary guts .
the couplings of the lightest @xmath1 and @xmath2 bosons are no longer directly linked through gauge symmetry to those of the lightest 3 - 2 - 1 bosons , and the couplings of the lightest higgs triplets are no longer directly linked to those of the higgs doublets .
the strong breaking of the gut symmetry `` pushes away '' the wave functions of these states , suppressing the generated proton decay below experimental limits , all while retaining _
ordinary , logarithmic unification_. such features seem special to a five dimensional theory , incapable of reproduction in simple four dimensional theories .
the presence of extra dimensions of this size has been motivated previously as a means to resolve the supersymmetric flavor problem @xcite .
our gut - breaking picture can nicely fit into this framework of supersymmetry - breaking mediation .
for instance , if supersymmetry is broken at @xmath30 fixed point by @xmath264-term vev of singlet or non - singlet field , it naturally realizes the scenarios of refs .
@xcite and @xcite , respectively .
the incorporation of our gut picture within these scenarios may give interesting signatures , since we now have one more piece of information about parameters of the model ; the compactification radius is determined by the gauge coupling unification .
we leave an investigation of detailed phenomenology for future work
. to summarize , the framework we have described is very simple , but can describe various observed features such as gauge coupling unification , lack of unification in yukawa couplings , the absence of proton decay , and so on
. it will be interesting to add non - minimal structure to the model as a means of deriving experimental signatures .
we thank m.l .
graesser , l.j . hall , d.b .
kaplan , d.e .
kaplan , a.e .
nelson , m.e . peskin and l. randall for useful discussions .
h. georgi and s. l. glashow , phys .
lett . * 32 * , 438 ( 1974 ) .
h. georgi , h. r. quinn and s. weinberg , phys .
lett . * 33 * , 451 ( 1974 ) .
s. dimopoulos and h. georgi , nucl .
b * 193 * , 150 ( 1981 ) ; + n. sakai , z. phys .
c * 11 * , 153 ( 1981 ) .
s. dimopoulos , s. raby and f. wilczek , phys .
d * 24 * , 1681 ( 1981 ) ; + l. e. ibanez and g. g. ross , phys . lett .
b * 105 * , 439 ( 1981 ) .
n. sakai and t. yanagida , nucl .
b * 197 * , 533 ( 1982 ) ; + s. weinberg , phys .
d * 26 * , 287 ( 1982 ) .
j. hisano , h. murayama and t. yanagida , nucl .
b * 402 * , 46 ( 1993 ) [ hep - ph/9207279 ] ; + for a recent analysis , see t. goto and t. nihei , phys .
d * 59 * , 115009 ( 1999 ) [ hep - ph/9808255 ] . h. murayama and a. pierce , to appear . z. chacko , m. a. luty and e. ponton , jhep*0007 * , 036 ( 2000 ) [ hep - ph/9909248 ] . n. arkani - hamed , l. hall , y. nomura , d. smith and n. weiner , hep - ph/0102090 . k. r. dienes , e. dudas and t. gherghetta , phys . lett .
b * 436 * , 55 ( 1998 ) [ hep - ph/9803466 ] ; nucl .
b * 537 * , 47 ( 1999 ) [ hep - ph/9806292 ] .
d. ghilencea and g. g. ross , phys .
b * 442 * , 165 ( 1998 ) [ hep - ph/9809217 ] ; nucl .
b * 569 * , 391 ( 2000 ) [ hep - ph/9908369 ] .
l. hall and y. nomura , hep - ph/0103125 .
j. hisano , h. murayama and t. yanagida , phys .
lett . * 69 * , 1014 ( 1992 ) .
m. bando , t. kugo , t. noguchi and k. yoshioka , phys .
lett . * 83 * , 3601 ( 1999 ) [ hep - ph/9906549 ] .
m. s. chanowitz , j. ellis and m. k. gaillard , nucl .
b * 128 * , 506 ( 1977 ) .
j. l. diaz - cruz , h. murayama and a. pierce , hep - ph/0012275 .
j. ellis and m. k. gaillard , phys .
b * 88 * , 315 ( 1979 ) . c. d. froggatt and h. b. nielsen , nucl .
b * 147 * , 277 ( 1979 ) .
k. i. izawa , k. kurosawa , y. nomura and t. yanagida , phys .
d * 60 * , 115016 ( 1999 ) [ hep - ph/9904303 ] .
y. fukuda _ et al .
_ [ super - kamiokande collaboration ] , phys .
81 * , 1562 ( 1998 ) [ hep - ex/9807003 ] .
t. yanagida , in _ proc . of the workshop on the unified theory and baryon number in the universe _ , ed
. o. sawada and a. sugamoto ( kek report 79 - 18 , 1979 ) , p. 95 ; m. gell - mann , p. ramond and r. slansky , in _
p. van nieuwenhuizen and d.z .
freedman ( north holland , amsterdam , 1979 ) , p. 315 . l. j. hall and m. suzuki , nucl .
b * 231 * , 419 ( 1984 ) ; + i. lee , nucl .
b * 246 * , 120 ( 1984 ) .
n. arkani - hamed , l. hall , h. murayama , d. smith and n. weiner , hep - ph/0006312 ; hep - ph/0007001 ; + f. borzumati and y. nomura , hep - ph/0007018 .
l. randall and r. sundrum , nucl .
b * 557 * , 79 ( 1999 ) [ hep - th/9810155 ] .
d. e. kaplan , g. d. kribs and m. schmaltz , phys .
d * 62 * , 035010 ( 2000 ) [ hep - ph/9911293 ] ; + z. chacko , m. a. luty , a. e. nelson and e. ponton , jhep*0001 * , 003 ( 2000 ) [ hep - ph/9911323 ] .
d. e. kaplan and g. d. kribs , jhep*0009 * , 048 ( 2000 ) [ hep - ph/0009195 ] . | we present a five - dimensional supersymmetric su(5 ) theory in which the gauge symmetry is broken maximally ( _ i.e. _ at the 5d planck scale @xmath0 ) on the same 4d brane where chiral matter is localized .
masses of the lightest kaluza - klein modes for the colored higgs and @xmath1 and @xmath2 gauge fields are determined by the compactification scale of the fifth dimension , @xmath3 gev , rather than by @xmath0 .
these fields wave functions are repelled from the gut - breaking brane , so that proton decay rates are suppressed below experimental limits . above the compactification scale ,
the _ differences _ between the standard model gauge couplings evolve logarithmically , so that ordinary logarithmic gauge coupling unification is preserved .
the maximal breaking of the grand unified group can also lead to other effects , such as @xmath4 deviations from su(5 ) predictions of yukawa couplings , even in models utilizing the froggatt - nielsen mechanism . #
1#2 uw - pt-01/09 + ucb - pth-01/12 + lbnl-47677 + .5 in * gut breaking on the brane * .40 in yasunori nomura@xmath5 , david smith@xmath5 and neal weiner@xmath6 .20 in @xmath5_department of physics , _ + and + _ theoretical physics group , lawrence berkeley national laboratory , + university of california , berkeley , california 94720 _ 0.08 in @xmath6_department of physics , + university of washington , seattle , washington 98195 , usa _
.5 in |
a classical problem in ergodic theory is to classify measure preserving transformations . since the celebrated results of ornstein in 1970 ( @xcite )
much research has been done on finding invariants for invertible maps . for non - invertible systems
there are some methods for comparing one - sided markov shifts on a finite alphabet ( see @xcite , @xcite and @xcite ) , but in general not much is known . in this article we are interested in isomorphisms between certain piecewise linear expanding interval maps , called positive and negative @xmath2-transformations .
as they are not invertible and they can not in general be described by a finite state markov shift , these maps do not fit in the framework of @xcite , @xcite and @xcite .
the @xmath2-transformation @xmath3 ( mod 1 ) with @xmath4 is a simple example of a chaotic dynamical system . there is a close link between the transformation @xmath5 and @xmath2-expansions of real numbers , i.e. , expressions of the form @xmath6 , where @xmath7 for all @xmath8 .
initiated by the works of rnyi ( @xcite ) and parry ( @xcite ) , this relation was extensively studied .
dynamical properties of the map can be used to study combinatorial properties of the expansions and vice versa . in the past few years
this link has lead to research on negative @xmath2-transformations .
recent results in theoretical computer science have shown that negative @xmath2-expansions , i.e. , expansions of the form @xmath9 , have better robustness properties when used for encoding purposes ( @xcite ) . in @xcite ito and
sadahiro studied the relation between negative @xmath2-expansions and a specific piecewise linear map with constant slope @xmath1 defined by @xmath10 where @xmath11 indicates the largest integer not exceeding @xmath12 .
this transformation maps the interval @xmath13 to itself .
ito and sadahiro used this map to show that the positive and negative @xmath2-expansions share many properties . since then , many others have found similarities between positive and negative @xmath2-expansions ( see @xcite , @xcite , @xcite and @xcite for example ) .
piecewise linear expanding interval maps with constant negative slope in general and the map @xmath14 in particular have also generated a lot of attention .
the papers @xcite , @xcite , @xcite , @xcite and @xcite deal , among other things , with invariant measures , transitivity and attractors of the maps in question .
it seems natural to ask exactly how different the positive and negative slope maps really are . in this paper
we let @xmath15 and we compare two maps : the positive and negative @xmath2-transformations @xmath16 and @xmath17 \to [ 0,1]$ ] given by @xmath18 .
\end{array}\right.\ ] ] the map @xmath19 is measurably isomorphic to the map @xmath14 with isomorphism @xmath20 . as in @xcite
we choose to work with @xmath19 instead of @xmath14 since s is defined on the unit interval , it has the same critical point as @xmath5 and , as for @xmath5 , its dynamical behaviour depends mainly on the orbit of the point 1 .
figure [ f : goldenmean ] shows both maps for a specific @xmath2 .
note that for all @xmath21 the map @xmath5 has only one fixed point , while @xmath19 has two .
hence , @xmath5 and @xmath19 can not be topologically conjugated , i.e. , there are no homeomorphisms @xmath22 \to [ 0,1]$ ] that preserve the dynamics by @xmath23 .
that is why we study the possibility of @xmath5 and @xmath19 being measurably isomorphic .
this article is organised as follows .
we first describe the dynamical properties of the maps @xmath5 and @xmath19 that can be found in the literature . in the two next sections we prove the two main results .
theorem [ t : multinacci ] states that for a specific set of @xmath2 s , the multinacci numbers , the maps @xmath5 and @xmath19 are measurably isomorphic .
it is proved by showing that @xmath5 and @xmath19 are both isomorphic to the same markov shift .
the second main result , theorem [ t : noniso ] , states that for all other values of @xmath2 the maps @xmath5 and @xmath19 are not isomorphic .
showing that two maps are not isomorphic involves showing that no map can qualify as an isomorphism .
therefore the proof of theorem [ t : noniso ] needed a novel approach .
we consider the number of preimages that points in @xmath24 $ ] can possibly have under @xmath25 and @xmath26 for an appropriate choice of @xmath27 .
we then show that these numbers do not coincide on sets of positive measure .
for the proof of theorems [ t : multinacci ] and [ t : noniso ] we need a precise description of the dynamical behaviour of the iterates of @xmath5 and @xmath19 .
the key result is lemma [ l : nbonaccis ] .
.2 cm * acknowledgement : * the author worked on this project in various locations .
the research was started in the eu fp6 marie curie research training network cody ( mrtn 2006 035651 ) at warwick university , then continued under the research grant fwf s6913 at the university of vienna and finally finished under the nwo veni grant 639.031.140 at leiden university .
the author would like to thank a number people for fruitful discussions and valuable suggestions : karma dajani , sebastian van strien , henk bruin and tom kempton .
let @xmath15 and let @xmath5 and @xmath19 be the transformations defined in ( [ q : tands ] ) .
let @xmath28 be the borel @xmath29-algebra on @xmath24 $ ] and @xmath30 the lebesgue measure on @xmath31 , \mathcal b)$ ] .
it is well known that there exists a unique invariant measure @xmath32 for @xmath5 equivalent to @xmath30 .
this measure is ergodic and it is the unique measure of maximal entropy with entropy @xmath33 .
for @xmath19 it is known that a unique invariant measure , absolutely continuous with respect to @xmath30 exists and that this is measure is ergodic ( see @xcite and @xcite ) .
we call this measure @xmath34 . combined results from @xcite and @xcite give that @xmath34 is the unique measure of maximal entropy of @xmath19 with @xmath35 .
the dynamics of both maps are largely determined by the _ orbits _ of the point 1 , i.e. , by the sets @xmath36 and @xmath37 .
the focus of this article is on the existence of isomorphisms between @xmath5 and @xmath19 . in general
we call two dynamical systems @xmath38 and @xmath39 ( or the two transformations @xmath40 and @xmath41 ) _ ( measurably ) isomorphic _ if there are sets @xmath42 and @xmath43 with @xmath44 and if there exists a map @xmath45 with the following properties : * @xmath46 is a measurable bijection , * @xmath46 is measure preserving , i.e. , for all @xmath47 , @xmath48 , * @xmath46 preserves the dynamics , i.e. , @xmath49 . as was mentioned in the introduction
, we will prove theorem [ t : multinacci ] by constructing a suitable markov shift .
a one - sided markov shift on a finite alphabet is defined as follows .
let @xmath50 be the set of states or the alphabet and let @xmath51 be a stochastic matrix . define the set @xmath52 let @xmath53 denote the left shift , i.e. , @xmath54 with @xmath55 for all @xmath56 .
the pair @xmath57 is called the _ symbolic system _ for the matrix @xmath58 .
note that a symbolic system can also be given by a ( 0,1)-matrix instead of a stochastic matrix .
let @xmath59 be the @xmath29-algebra on @xmath60 generated by the cylinders , i.e. , the sets obtained by specifying the values of the first @xmath61 elements of the sequence , @xmath8 .
let @xmath62 be a positive probability vector such that @xmath63 .
the markov measure @xmath64 associated to @xmath58 is defined on the cylinders by @xmath65 the system @xmath66 is called a one - sided _ markov shift_.
in this section we show that the positive and negative @xmath2-transformations are measurably isomorphic when @xmath2 is a multinacci number .
multinacci numbers are specific pisot numbers . a _
pisot number _ is a real algebraic integer bigger than 1 with all its galois conjugates less than 1 in modulus . for @xmath67 , the @xmath68-th _ multinacci number _
@xmath69 is the pisot number with minimal polynomial @xmath70 then @xmath71 is the golden ratio and @xmath72 is usually referred to as the tribonacci number .
it is well - known that the sequence @xmath73 is strictly increasing and @xmath74 .
we use the multinacci numbers to give a detailed description of the dynamical behaviour of the maps @xmath5 and @xmath19 . for the results in this section , we need results from @xcite on piecewise linear markov maps .
a transformation @xmath75 \to [ 0,1]$ ] is called a _ piecewise linear markov map _ if there exists points @xmath76 such that for all @xmath77 , @xmath40 is a linear homeomorphism on the interval @xmath78 with @xmath79 \quad \text{for some } \ , 0\le j < k \le 1.\ ] ] the partition @xmath80 of the interval @xmath24 $ ] given by @xmath81 is called a _ markov partition _ for @xmath40 .
note that when we say _ partition _ , we mean a finite collections of sets with positive @xmath30-measure that are pairwise disjoint and cover the whole interval @xmath24 $ ] upto a set of @xmath30-measure zero .
the map @xmath40 is associated to a ( 0,1)-matrix @xmath82 that says which transitions between intervals @xmath83 are possible . to obtain @xmath82 , we label these intervals with a number : @xmath84 for some @xmath85 .
( we could of course simply take @xmath86 , but later we want to have the opportunity to choose @xmath87 )
. then @xmath88
if @xmath89 and @xmath90 otherwise . let @xmath91 be the symbolic system associated to @xmath82 .
below are some results from @xcite adapted to our setting . *
( proposition 1 of @xcite ) suppose @xmath82 is an irreducible @xmath92 ( 0,1)-matrix with spectral radius @xmath93 and of which the 1 s in each row are not separated by 0 s .
suppose that @xmath16 ( or @xmath94 ) is a piecewise linear markov map with matrix @xmath82 .
then there is a countable set @xmath95 $ ] and a homeomorphism @xmath96\backslash c$ ] such that @xmath97 ( or @xmath98 ) . *
( proposition 2 of @xcite ) under the same conditions as in ( r1 ) , the symbolic system @xmath91 given by @xmath82 admits a unique maximal markov measure @xmath99 . *
( theorem 2 of @xcite ) under the same conditions as in ( r1 ) , the measure @xmath100 is the unique probability measure for @xmath5 ( or @xmath19 ) that is invariant for @xmath5 ( or @xmath19 ) , maximizes entropy and is equivalent to @xmath30 .
we will show that for @xmath101 both @xmath5 and @xmath19 are piecewise linear markov maps with the same ( 0,1)-matrix @xmath82 .
since @xmath32 and @xmath34 are the unique measures of maximal entropy for @xmath5 and @xmath19 respectively , theorem 2 from @xcite then gives that @xmath5 and @xmath19 are isomorphic , if @xmath82 satisfies the conditions from ( r1 ) . as an illustration
we first show how this works for @xmath71 . in figure
[ f : goldenmean ] we see both maps @xmath5 and @xmath19 for @xmath102 . -.3 cm note that @xmath103 . set @xmath104 and @xmath105
. then @xmath106 and @xmath107 .
hence , @xmath5 is a piecewise linear markov map on the markov partition @xmath108 with matrix @xmath109 let @xmath91 denote the symbolic system given by the matrix @xmath82 . note that @xmath82 is irreducible and that the 1 s in each row are touching . moreover ,
the characteristic polynomial of @xmath82 is equal to the minimal polynomial of @xmath71 .
hence , the spectral radius of @xmath82 is @xmath71 and @xmath5 and @xmath82 satisfy all the conditions of ( r1 ) .
so , @xmath31 , \mathcal b , \mu , t)$ ] and @xmath110 are isomorphic .
for @xmath19 , note that @xmath111 and that @xmath112.\ ] ] hence , if we set @xmath113 and @xmath114 , then we see that @xmath82 is also the appropriate ( 0,1)-matrix for @xmath19 .
hence , @xmath31 , \mathcal b , \nu , s)$ ] is also isomorphic to @xmath110 and therefore to @xmath31 , \mathcal b , \mu , t)$ ] .
the map @xmath5 behaves very regularly for each multinacci number @xmath69 .
note that for each @xmath115 , @xmath116 and hence @xmath117 the points in ( [ q : markovt ] ) give a markov partition for @xmath5 .
set @xmath118 and for @xmath119 , @xmath120 .
then @xmath5 is a piecewise linear markov map with associated ( 0,1)-matrix @xmath121 in figure [ f : graphbn ] we see the corresponding graph .
let @xmath91 denote the symbolic system given by @xmath82 .
note that @xmath82 is irreducible and that the 1 s in each row are contiguous . since the characteristic polynomial of @xmath82 is the minimal polynomial of @xmath69 ,
the spectral radius of @xmath82 is @xmath69 .
hence , all the requirements of ( r1 ) are fulfilled and @xmath31 , \mathcal b , \mu , t)$ ] and @xmath110 are isomorphic .
.2 cm for the map @xmath19 we have to distinguish between even and odd @xmath68 .
note that the small and the large fixed points of the map are given by @xmath122 and @xmath123 respectively .
the following lemma on the orbit of 1 under @xmath19 is a key result in this paper .
[ l : nbonaccis ] for all @xmath67 , if @xmath124 and @xmath115 , then @xmath125 moreover , @xmath126 with equality if and only if @xmath101 . to prove the lemma
, we also prove the following claim .
+ * claim : * if @xmath124 , then for all @xmath127 , @xmath128 .2 cm we prove all statements by induction .
it is clear that for all @xmath93 , @xmath129 .
so the claim is clear for @xmath130 .
then @xmath131 recall from ( [ q : polynomial ] ) that @xmath132 is the minimal polynomial of @xmath71 . since @xmath71 is a pisot number , we know that @xmath71 is the only root of this polynomial in the interval @xmath133 .
hence @xmath134 if and only if @xmath135 and equality holds if and only if @xmath136 .
this proves the lemma for @xmath130 .
.2 cm for @xmath137 , let @xmath138
. then @xmath139 and thus the claim is true for @xmath137 . to prove the statement of the lemma , @xmath140 note that the left hand side of the last inequality is @xmath141 , the minimal polynomial of the pisot number @xmath72 . hence @xmath142 if and only if @xmath143 with equality if and only if @xmath144 .
.2 cm now suppose that the lemma and the claim are true for some @xmath67 and let @xmath145 .
recall that the sequence @xmath146 is strictly increasing towards 2 , so that we only need to prove the claim and the last statement of the lemma for @xmath147 .
if @xmath68 is even , then by the induction hypothesis we know that @xmath148 hence , @xmath149 which proves the claim in this case .
moreover , @xmath150 if and only if @xmath151 since the left hand side is the minimal polynomial of the pisot number @xmath152 , we have that @xmath153 if and only if @xmath154 with equality if and only if @xmath155 .
similarly , if @xmath68 is odd , then the induction hypothesis gives @xmath156 and thus , @xmath157 this shows that the claim is correct .
then , @xmath158 if and only if @xmath159 which proves the lemma as above .
[ r : orbit0]the map @xmath19 is linear with slope @xmath1 on the intervals @xmath160 $ ] and @xmath161 $ ] .
therefore , lemma [ l : nbonaccis ] gives us even more information on the orbit of 1 under @xmath19 .
for @xmath67 and @xmath162 , the order of the points @xmath163 , @xmath164 , is completely determined as shown in figure [ f : points ] .
recall that if @xmath165 , then @xmath166 or @xmath167 depending on the parity of @xmath68 .
-.3 cm remark [ r : orbit0 ] and figure [ f : points ] give us the markov partition for @xmath19 when @xmath101 . if @xmath68 is even , then @xmath168 and @xmath19 is a piecewise linear markov map with partition @xmath169 we label the partition elements as follows .
set @xmath170 .
note that @xmath171.\ ] ] hence , all other states are mapped into @xmath172 .
let @xmath173 and note that then @xmath174 .
next set @xmath175 , so that @xmath176 , and continue in this manner until we have @xmath177 .
then @xmath82 from ( [ q : mt ] ) is the ( 0,1)-matrix corresponding to @xmath19 . .2 cm if @xmath68 is odd , then the markov partition is given by @xmath178 the labeling of the state goes in the same way as for the even case with @xmath179 . then again
@xmath82 from ( [ q : mt ] ) is the ( 0,1)-matrix corresponding to the piecewise linear markov map @xmath19 . since @xmath82 satisfies all the conditions from ( r1 ) , we have the following theorem .
[ t : multinacci ] for @xmath67 , let @xmath69 denote the @xmath68-th multinacci number , i.e. , the pisot number with minimal polynomial @xmath180 . then the positive @xmath69-transformation @xmath5 and the negative @xmath69-transformation @xmath19 are measurably isomorphic .
next we show that for all non - multinacci numbers @xmath2 the transformations @xmath5 and @xmath19 are * not * isomorphic .
we do this by studying the iterates of the maps @xmath5 and @xmath19 . recall that if @xmath46 is an isomorphism , then @xmath181 and thus @xmath182 for all @xmath56 . hence ,
if we can show that for some iterate @xmath25 there is an interval ( or set of positive @xmath32-measure ) , such that each point in this interval has @xmath61 preimages under @xmath25 and that the set of points with @xmath61 preimages under the map @xmath26 has @xmath34-measure zero , then @xmath5 and @xmath19 can not be isomorphic .
we make this more precise .
for @xmath56 define the sequences of maps @xmath183 \to \mathbb n$ ] by @xmath184 \ , : \ , t^n y = x \ } \quad \text{and } \quad \psi^-_n(x ) = \ # \ { y \in [ 0,1 ] \ , : \ ,
s^n y = x \}.\ ] ] these maps count the number of preimages that points have under the iterates of @xmath5 and @xmath19 . for @xmath185 define the sets @xmath186 \ , : \ , \psi_n^+(x)=k\ } \quad \text{and } \quad i_n^-(k ) = \ { x \in [ 0,1 ] \ , : \ , \psi_n^-(x)=k\}.\ ] ] note that the range of the functions @xmath187 and @xmath188 and the shapes of the sets @xmath189 and @xmath190 depend on the images of the intervals of monotonicity of @xmath25 and @xmath26 respectively .
we call a non - empty , open interval @xmath191 an _ interval of monotonicity _ for @xmath25 if @xmath192 is continuous and linear and if there is no bigger open interval @xmath193 with the same property .
a similar definition holds for @xmath19 .
let @xmath191 be an interval of monotonicity of @xmath25 .
then clearly @xmath194 for some @xmath195 .
hence , the sets @xmath189 are either empty or of the form @xmath196 for some @xmath197 , @xmath87 . if @xmath191 is an interval of monotonicity for the map @xmath26 , then @xmath198 or @xmath199 for some @xmath197 , @xmath87 .
thus , the sets @xmath190 are either empty or they contain an interval of the form @xmath200 . if @xmath46 is an isomorphism from @xmath5 to @xmath19 , then @xmath201 and thus @xmath202 for all @xmath203 .
what we will do is show that if @xmath204 , then there is a @xmath8 , such that @xmath205 .
we do this by showing that one of the sets has measure zero , while the other has positive measure . to illustrate the general approach , we first consider @xmath206 and @xmath207 .
. then @xmath209 .
moreover @xmath210 if and only if @xmath211 , where @xmath212 is the real root of the polynomial @xmath213 .
we see the first three iterates of the map @xmath5 in the top row of figure [ f : smallbetas ] .
note that for @xmath211 , @xmath214 and for @xmath215 , @xmath216 .
hence , @xmath217 for all @xmath208 .
recall that @xmath19 has two fixed points and that @xmath218 is the smaller of the two .
if @xmath219 , then @xmath220 .
the first three iterates of @xmath19 are shown in the bottom row of figure [ f : smallbetas ] . from figure
[ f : smallbetas ] we see that for all @xmath206 , @xmath221 .
hence , @xmath222 and the maps @xmath5 and @xmath19 can not be isomorphic . if @xmath207 , then @xmath223 , @xmath224 and @xmath225 .
this allows us to calculate the range of @xmath187 for @xmath226 . for @xmath19 , lemma [ l : nbonaccis ] implies that @xmath227 and @xmath228 . to accurately describe the first three iterates of @xmath19
, we have to know if @xmath229 or not . therefore , let @xmath230 denote the real root of the polynomial @xmath231 . then @xmath232
if @xmath233 and @xmath234 if @xmath235 .
this allows us to calculate @xmath236 for @xmath226 and all @xmath237 . in figure
[ f : biggerbetas ] we show the graphs of @xmath238 and @xmath239 .
there we see that for all @xmath207 , @xmath240 and @xmath241 .
these are the coloured areas in the top line of figure [ f : biggerbetas ] . on the other hand , for @xmath239 we have @xmath242 if @xmath243 ( see figure [ f : biggerbetas](d ) and ( e ) ) and @xmath244 if @xmath245 ( see figure [ f : biggerbetas](e ) and ( f ) ) . hence , @xmath5 and @xmath19 can not be isomorphic for these @xmath2 s . to prove that @xmath5 and @xmath19 can not be isomorphic if @xmath247 , we follow the approach from the previous section .
we will look at the ranges of @xmath248 and @xmath249 and therefore we study @xmath25 and @xmath26 .
we begin with @xmath5 .
[ [ the - positive - beta - transformation - between - beta_n-1-and - beta_n ] ] the positive @xmath2-transformation between @xmath246 and @xmath69 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ for @xmath162 , @xmath250 , we know that @xmath251 and @xmath252 .
we are interested in the range of @xmath187 . to determine this exactly
, we will take a closer look at the images of the ( open ) intervals of monotonicity of the iterates of @xmath5 and we will assign a type to each of them .
the map @xmath5 has two intervals of monotonicity : @xmath253 and @xmath254 .
the images of these intervals are @xmath255 and @xmath256 , hence there are two different types of images and both types occur exactly once for @xmath5 .
we call an interval of monotonicity with image @xmath255 type 0 and the one with image @xmath257 type 1 .
this gives that @xmath258 since we are only interested in the @xmath32-measure of sets , it is enough to consider open intervals here .
@xmath259 has four intervals of monotonicity , but only three different types : @xmath260 there are two times type 0 , one type 1 and one new type , @xmath261 , which we call type 2 .
moreover , we see that a type 0 interval splits into a type 0 and a type 1 for the next iteration and a type 1 interval splits in a type 0 and a type 2 .
this gives @xmath262 going one step further , we see that for @xmath238 the type 2 interval splits into a type 0 interval and a type 3 interval with image @xmath263 . since @xmath264 this pattern continues up to @xmath265 and we can recursively determine the numbers with which each of the types of intervals occur for @xmath265 .
let @xmath266 denote the number of type @xmath267 intervals that @xmath268 has .
we have the following lemma .
fix some @xmath250 .
first note that for @xmath269 , @xmath268 can only have intervals of type @xmath278 , which proves the middle part of the lemma .
we prove the first statement by induction on @xmath99 .
we have already proved the statement for @xmath279 .
now suppose that the statement is true for some @xmath280 .
each type @xmath267 interval , @xmath281 , has image @xmath282 under @xmath268 .
hence , by ( [ q : pointst ] ) each type @xmath267 interval splits into a type 0 interval and a type @xmath283 interval for @xmath284 .
thus , @xmath285 moreover , @xmath286 for all @xmath287 and the first statement of the lemma is proved . for the last part , first note that @xmath288 implies that the type @xmath289 interval of @xmath265
does not split into two intervals of monotonicity for @xmath25 .
hence , the type @xmath289 interval only produces a type @xmath68 interval for @xmath25 .
this implies that @xmath290 for all @xmath85 we still have @xmath291 , which gives the lemma .
the points @xmath292 partition the interval @xmath24 $ ] into intervals on which @xmath187 is constant . by adding the appropriate numbers @xmath293 we get the values of @xmath248 . note that @xmath248 is a monotone decreasing function on @xmath294 .
the next lemma gives some information about the range of @xmath248 .
let @xmath299 be such that @xmath300 .
in figure [ f : psi+ ] we can see which numbers @xmath293 we need to add in order to get the values @xmath301 can take on each interval . for all @xmath302 on the interval @xmath303
we have @xmath304 .
also , @xmath305 on the interval @xmath306 .
the last part of lemma [ l : kappas ] implies that these numbers are odd .
furthermore , @xmath307 on @xmath308 , which is even , and for @xmath309 we have the even value @xmath310 on @xmath303 .
on @xmath311 we have @xmath312 , which is odd .
[ r : values](i ) by lemma [ l : values ] we know a few values of @xmath187 explicitly .
for example , @xmath313 on @xmath314 , which also the maximum of @xmath187 . since @xmath315 we also know that @xmath316 on @xmath317 .
the minimum of @xmath187 is attained on the interval @xmath318 and here @xmath319 .
since the exact position of @xmath320 is unknown , we can not say more .
+ ( ii ) lemma [
l : values ] also implies that all the odd values in the range of @xmath248 other than the maximum are smaller than the even values .
[ [ the - negative - beta - transformation - between - beta_n-1-and - beta_n ] ] the negative @xmath2-transformation between @xmath246 and @xmath69 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ for the map
@xmath19 the situation is more complicated .
we will describe the structure of the images of the intervals of monotonicity of the iterates of @xmath19 , just as we did for @xmath5 .
recall that figure [ f : points ] gives the relative positions of the points @xmath163 .
note that the point @xmath321 can be either to the left or to the right of @xmath322 .
this means that for all @xmath162 the maps @xmath323 , @xmath324 have the same structure and the map @xmath26 can behave in two different ways , according to where the point @xmath321 lies relative to @xmath322 .
consider the maps @xmath323 , @xmath127 .
as before we assign a type to each of the ( open ) intervals of monotonicity of @xmath323 according to the shape of their images under @xmath323 .
we are interested in determining the numbers of intervals of each type . for @xmath19
, we have two types , @xmath325 and @xmath326 , and both occur once .
we call the one with image @xmath255 type 0 and the other one type 1 .
then , @xmath327 again we are not bothered with the endpoints of the intervals , since we are only interested in the measures of the sets . for @xmath328 the intervals of monotonicity are @xmath329 , @xmath330 , @xmath331 and @xmath332 .
so , the type 0 interval for @xmath19 splits into a type 0 and a type 1 interval for @xmath328 and the type 1 interval for @xmath19 splits into a type 1 interval and an interval with image @xmath334 , which we call type 2 for @xmath328 .
hence , for @xmath328 we have one type 0 , two times type 1 and one type 1 .
this gives @xmath335 for @xmath239 we only have to determine how the type 2 interval splits .
since @xmath336 , the interval becomes a type 0 interval and a type 3 interval with image @xmath337 .
if @xmath338 , then @xmath339 and the type 3 interval splits into a type 1 and a type 4 for @xmath340 . we can continue in this manner upto @xmath341 , see figure [ f : types ] .
we can now inductively determine the number of intervals of monotonicity of each type for @xmath341 and then also for @xmath26 .
let @xmath342 denote the number of intervals of type @xmath267 that @xmath343 has . hence , we know that @xmath344 and @xmath345 and @xmath346 . * if @xmath366 , then @xmath367 , @xmath368 , @xmath369 and for all @xmath370 , @xmath371 .
moreover , the type @xmath68 interval has image @xmath372 under @xmath26 . * if @xmath373 , then @xmath367 , @xmath368 , @xmath374 and for all @xmath370 , @xmath371 . * if @xmath375 , then @xmath367 , @xmath376 , @xmath368 , @xmath369 and for all @xmath377 , @xmath371 .
moreover , the type @xmath68 interval has image @xmath378 under @xmath26 . *
if @xmath366 , then @xmath379 , @xmath368 , @xmath369 and for all @xmath370 , @xmath371 . moreover , the type @xmath68 interval has image @xmath380 under @xmath26 . * if @xmath373 , then @xmath367 , @xmath368 , @xmath374 and for all @xmath370 , @xmath371 . *
if @xmath375 , then @xmath367 , @xmath368 , @xmath369 and for all @xmath370 , @xmath371 .
moreover , the type @xmath68 interval has image @xmath380 under @xmath26 . by lemma [ l : iotas ] we know that @xmath381 , @xmath382 , @xmath383 and for all @xmath349 , @xmath384 .
first assume that @xmath68 is even .
recall that the image of the type ( @xmath385 ) interval under @xmath341 is @xmath386 .
we treat each case separately .
.2 cm ( 1 ) suppose @xmath366 . then the type ( @xmath385 ) interval splits into an interval with image @xmath387 , i.e. , a type 1 interval , and a type @xmath68 interval with image @xmath372 under @xmath26 .
hence , the recursions ( [ q : rec1 ] ) and ( [ q : rec2 ] ) still hold and this gives the result in this case .
\(3 ) suppose @xmath375
. then @xmath389 .
thus , the type ( @xmath385 ) interval does not split and becomes a type @xmath68 interval for @xmath26 .
this implies that there is one type 1 interval less than there would be according to ( [ q : rec1 ] ) .
this gives the lemma in the even case .
.2 cm ( 1 * ) suppose @xmath366
. then @xmath391 .
thus , the type ( @xmath385 ) does not split and becomes a type @xmath68 interval for @xmath26 .
therefore , there is one type 0 interval less than there would be according to ( [ q : rec1 ] ) . for all other values of @xmath267 , the formulas ( [ q : rec1 ] ) and ( [ q : rec2 ] ) still hold .
( 3 * ) suppose @xmath375 .
then the type ( @xmath385 ) interval splits into a type 0 interval and a type @xmath68 interval with image @xmath380 under @xmath26 .
hence , the recursions ( [ q : rec1 ] ) and ( [ q : rec2 ] ) still hold and this gives the lemma .
[ t : noniso ] for @xmath67 , let @xmath69 be the @xmath68-th multinacci number , i.e. , the pisot number with minimal polynomial @xmath180 .
fix @xmath68 and let @xmath162 . then the positive @xmath2-transformation @xmath5 and the negative @xmath2-transformation @xmath19 are not measurably isomorphic . for the proof of this theorem
we calculate some values in the range of @xmath188 and compare them to the information about @xmath248 obtained in lemma [ l : values ] and remark [ r : values ] .
our goal is to show that in all cases @xmath187 takes a value that does not occur for @xmath188 or the other way around . just as for @xmath5 ,
the points @xmath163 determine a partition of the interval @xmath255 into intervals on which @xmath249 is constant . to determine the values of @xmath188 we need to add the numbers @xmath364 for the right @xmath267 . for
even @xmath68 the function @xmath188 takes its maximum on the interval @xmath393 and for odd @xmath68 the maximum is attained on the interval @xmath394 , see figure [ f : psi- ] .
\(1 ) if @xmath395 , then @xmath396 . for @xmath397 ,
we have @xmath398 hence , according to remark [ r : values](i ) , the maxima of @xmath187 and @xmath188 are equal .
for @xmath399 it holds that @xmath400 . note that @xmath401 is an odd number smaller than the maximum of @xmath187 . on the other hand , on @xmath402 , @xmath403 recall that by lemma [ l : values ] , all the odd values in the range of @xmath248 other than the maximum must be smaller than the even values . since @xmath404 is an even number and @xmath405 , either @xmath406 or @xmath407 . on the other hand , @xmath408 and @xmath409 .
this implies that @xmath5 and @xmath19 can not be isomorphic transformations .
.1 cm ( 2 ) if @xmath373 , the maximal value of @xmath188 can be calculated just as for ( a ) and is equal to @xmath410 . then @xmath411 and @xmath412 for all other @xmath413 $ ] .
this means that @xmath188 does not take the value @xmath414 , i.e. , @xmath415 . by remark [ r : values](i ) , we know that @xmath416 , showing that @xmath5 and @xmath19 can not be isomorphic in this case .
\(3 ) if @xmath417 , then @xmath418 . for @xmath397 ,
we have @xmath419 since @xmath420 , we have @xmath421 and @xmath412 for all other @xmath413 $ ] . again
@xmath188 does not take the value @xmath414 and thus by remark [ r : values](i ) @xmath5 and @xmath19 can not be isomorphic .
this finishes the proof for even @xmath68 . for odd @xmath68
the proof goes more or less along the same lines .
recall figure [ f : points](b ) and recall that the function @xmath188 takes its maximum on the interval @xmath422 .
similar to above , it can be shown that also for odd @xmath68 this maximum is equal to @xmath410 in all cases .
we will use lemma [ l : cases ] to determine other values of @xmath188 as above and therefore we treat the three cases ( 1 * ) , ( 2 * ) and ( 3 * ) separately .
.1 cm ( 1 * ) suppose that @xmath395 .
then @xmath423 , which implies that @xmath424 and @xmath412 for all other @xmath413 $ ] .
this means that @xmath188 does not take the value @xmath414 , i.e. , @xmath415 . by remark [ r : values](i ) , we know that @xmath416 , showing that @xmath5 and @xmath19 can not be isomorphic in this case .
.1 cm ( 2 * ) if @xmath373 , then @xmath425 and @xmath412 for all other @xmath413 $ ] .
again @xmath188 does not take the value @xmath414 and thus by remark [ r : values](i ) @xmath5 and @xmath19 can not be isomorphic .
.1 cm ( 3 * ) if @xmath417 , then @xmath426 . according to remark [ r : values ] , the maxima of @xmath187 and @xmath188 are equal . for @xmath427 , we have @xmath428 , which is an odd number smaller than the maximum of @xmath187 . on the other hand , on @xmath429 , @xmath430 which is an even number and bigger than @xmath401 . by remark [ r :
values](ii ) , either @xmath431 or @xmath406 , while @xmath432 and @xmath409 .
this implies that @xmath5 and @xmath19 can not be isomorphic transformations .
note that the previous theorem gives an illustration of the well known fact that two one - sided markov shifts on a finite alphabet with the same topological entropy do not necessarily have to be isomorphic .
for all pisot numbers @xmath2 the maps @xmath5 and @xmath19 are isomorphic to a one - sided finite alphabet markov shift with topological entropy equal to @xmath433 .
since pisot numbers lie dense in the interval @xmath434 , there are many pisot numbers in this interval for which the maps @xmath5 and @xmath19 are not isomorphic .
hence , the same holds for the associated markov shifts .
one can still wonder how different the dynamics of positive slope and negative slope piecewise linear maps really is .
for example , in case @xmath435 is not a multinacci number , does there exists any piecewise linear map with constant slope @xmath2 that is isomorphic to the map @xmath19 ? if the map @xmath19 is a piecewise linear markov map for the @xmath2 in question , then one can always construct an isomorphic piecewise linear map with constant slope @xmath2 based on the markov partition and the corresponding ( 0,1)-matrix .
but what if @xmath19 is not a piecewise linear markov map ? theorems [ t : multinacci ] and [ t : noniso ] will most likely extend to @xmath2-transformations with more than 2 digits , i.e. , with @xmath436 .
we then define @xmath437 and @xmath438 .
the two maps @xmath439 and @xmath440 are probably isomorphic if and only if @xmath2 is an integer or a ` generalised multinacci number ' .
the generalised multinacci numbers are the numbers satisfying @xmath441 with @xmath442 and @xmath56 . for @xmath21
this leaves only the multinacci numbers . | we compare a piecewise linear map with constant slope @xmath0 and a piecewise linear map with constant slope @xmath1 .
these maps are called the positive and negative @xmath2-transformations .
we show that for a certain set of @xmath2 s , the multinacci numbers , there exists a measurable isomorphism between these two maps .
we further show that for for all other values of @xmath2 between 1 and 2 the two maps can not be isomorphic .
[ section ] [ section ] [ section ] [ section ] [ section ] [ section ] [ section ] |
suppose that we observe equispaced samples ( at the nyquist sampling rate ) of a number of @xmath6 sinusoidal signals : denoted by matrix @xmath7 , on the index set @xmath8 , where @xmath2 is the number of equispaced samples per sinusoidal signal , @xmath9 , and @xmath10 .
that is , we have @xmath6 measurement vectors corresponding to the @xmath6 columns of @xmath11 . here
@xmath12 indexes the entries of @xmath11 , @xmath13 , @xmath14 $ ] denotes the @xmath15th normalized frequency ( the starting point 0 and the ending point 1 are identical ) , @xmath16 is the ( complex ) amplitude of the @xmath15th frequency component composing the @xmath17th sinusoidal signal , and @xmath5 is the number of the components which is unknown but typically small .
moreover , we let @xmath18 be the sample size of each measurement vector .
following from the literature of spectral analysis @xcite , the observed data @xmath19 are called complete if @xmath20 ( i.e. , @xmath21 or @xmath22 ) and incomplete otherwise . in the latter case , the unobserved data on the complementary index set @xmath23 , @xmath24 , are called missing data .
let @xmath25 denote the set of the frequencies .
the problem concerned in this paper is to recover @xmath26 given the observed data , which is referred to as the problem of joint sparse frequency recovery in the sense that the multiple measurement vectors ( mmvs ) ( i.e. , the @xmath6 columns of @xmath27 ) share the same @xmath5 frequencies .
after @xmath26 is obtained , the amplitudes @xmath28 and the missing data can be easily obtained by a simple least - squares method . in the single measurement vector ( smv )
case where @xmath29 , this frequency recovery problem is known as line spectral estimation and has wide applications in communications , radar , sonar , seismology , astronomy and so on @xcite . with complete data
, prony s method can recover the frequencies once @xmath30 by root - finding regardless of the values of the frequencies ( see , e.g. , @xcite ) .
however , it does not work in the presence of missing data and is sensitive to measurement noise . with the development of sparse signal representation and later the compressed sensing ( cs ) concept @xcite , which studies the recovery of a sparse signal from a number of linear measurements much less than its ambient dimension ,
sparse methods for frequency recovery have been popular in the past decade .
but unfortunately , the frequencies of interest are usually assumed to lie on a fixed grid of the frequency domain @xmath31 because the development of cs so far has been focused on signals that can be sparsely represented under a finite discrete dictionary . under the aforementioned assumption the observation model ( [ formu : observmodel1 ] )
can be written into an underdetermined system of linear equations and then sparse methods are applied to solve the sparse signal involved whose support corresponds to the frequency set @xmath26 .
typical sparse methods include combinatorial optimization or @xmath0 ( pseudo-)norm minimization , its convex relaxation or @xmath1 norm minimization , and greed methods such as orthogonal matching pursuit ( omp ) @xcite .
the @xmath0 minimization has the best theoretical guarantee , typically ensuring exact recovery once @xmath32 , however , it is np - hard and can not be practically solved . in conventional wisdom ,
the maximal @xmath5 allowed in the @xmath1 minimization and omp for guaranteed exact recovery is inversely proportional to a metric called coherence which , however , increases dramatically as the grid gets fine . on the other hand , due to the grid selection
, basis mismatches become a major problem of cs - based methods and many modifications have been proposed to alleviate this drawback ( see , e.g. , @xcite ) . a breakthrough came up recently .
cands and fernandez - granda @xcite deal directly with the continuous frequency recovery problem and therefore completely eliminate the basis mismatches . in particular
, they consider the complete data case and show that the frequencies can be exactly recovered via convex optimization once any two frequencies are separate by at least @xmath3 .
that means , up to @xmath33 frequencies can be recovered within a polynomial time under the frequency separation condition .
the convex optimization is based on the so - called total variation norm or atomic norm which extends the @xmath1 norm to the continuous frequency setting and is formulated as semidefinite programming ( sdp ) @xcite .
inspired by @xcite , tang _
_ @xcite study the problem of continuous frequency recovery from partial observations ( i.e. , incomplete data ) based on the atomic norm minimization . under the same frequency separation condition
, they show that a number of @xmath34 randomly located measurements is sufficient to guarantee exact recovery with high probability .
several subsequent works on this topic have been done and an incomplete list includes @xcite . in the mmv case , an example at hand is direction of arrival ( doa ) estimation in array processing @xcite .
in particular , suppose that @xmath5 farfield , narrowband sources impinge on an array of sensors and one wants to know their directions .
the output of the sensor array can be modeled by ( [ formu : observmodel1 ] ) , where each frequency corresponds to one source direction .
the sampling index set @xmath35 therein represents the geometry of the sensor array . to be specific
, @xmath36 in the complete data case corresponds to an @xmath2-element uniform linear array ( ula ) while @xmath37 denotes a sparse linear array ( sla ) .
each measurement vector is composed of the output of the sensor array at one snapshot , and the mmvs are obtained by taking measurements at multiple snapshots , where the source directions are assumed constant during the time window .
music @xcite is prominent for this joint sparse frequency recovery problem , however , it is sensitive to correlations of the sources and requires a sufficient number of snapshots such that the sample variance can capture the whole signal subspace . in lieu of sparse methods in the smv case ,
joint sparse recovery techniques have been widely studied in the mmv case which , besides the sparse property , exploit the prior knowledge that the mmvs share the same sparsity profile , known as joint sparsity @xcite .
it has been vastly reported that the performance of joint sparse recovery can be generally improved by increasing the number of measurement vectors .
theoretical results include @xcite on the @xmath38 norm under a mild condition , and @xcite on greed methods and @xcite on the @xmath1 norm under the assumption that the joint sparse signals are randomly drawn , in particular , the rows of the source signals @xmath39 in ( [ formu : observmodel1 ] ) are at general positions .
however , it is worth noting that the theoretical guarantee of any joint sparse recovery technique can not be improved without additional assumptions of the sparse signals of interest . to see this ,
suppose that all the columns of @xmath40 are identical up to some scale factors ( any two sources / rows of @xmath40 are identical up to a scale factor as well and usually said to be coherent in array processing ) and therefore , the mmvs are simply scaled replica of a smv and do not provide additional information for the recovery . in this respect , the results of @xcite and @xcite are referred to as the _ average case _ analysis while those accounting for the aforementioned extreme case are called _ worst case _ analysis . similarly to the smv case , the joint sparse recovery methods rely on discretization / gridding of the frequency domain and suffer from basis mismatches .
unlike the smv case in which the continuous frequency recovery methods have been recently studied , results are rare on the joint sparse frequency recovery concerned in this paper . to the best of our knowledge
, the only known discretization - free / gridless method is introduced in our previous work @xcite based on a statistical perspective and utilizing a weighted covariance fitting ( wcf ) criterion in @xcite . in the main context of this paper
we will show that this wcf technique is related to the mmv atomic norm method to be proposed in this paper . in this paper
, we present optimization methods for joint sparse frequency recovery and theoretically analyze their performances in the noiseless setting .
we firstly consider a continuous @xmath0 norm formulation , referred to as the atomic @xmath0 norm , and present its theoretical guarantees for exact frequency recovery , which extends the conventional discrete problem formulation , concept and result to the continuous setting .
we next consider its convex relaxation , referred to as the ( mmv ) atomic norm , and investigate its theoretical guarantees with complete and incomplete data separately .
in particular , given the complete data we prove that the frequencies can be exactly recovered by solving a convex optimization problem once they are separate by at least @xmath3 . under the same frequency separation condition ,
a number of @xmath4 randomly located samples per measurement vector is sufficient to guarantee that the frequencies and the missing data can be exactly recovered with high probability by solving an atomic norm minimization problem . as a result , our analysis with the atomic norm extends the results of @xcite to the mmv case .
since no or very mild assumptions are made for the source signals in our analysis , the _ worst case _
theoretical guarantees above do not improve as the number of measurement vectors increases . moreover , we formulate the atomic @xmath0 norm and the atomic norm problems as rank minimization and sdp , respectively .
extensive numerical simulations are carried out which validate our theoretical results and further demonstrate that the frequency separation condition required for exact recovery can be relaxed in general as the number of measurement vectors increases but can not in the worst case .
our results provide theoretical guidance for the practical array processing applications and will inspire further studies such as the average case analysis .
notations used in this paper are as follows .
@xmath41 and @xmath42 denote the sets of real and complex numbers respectively .
@xmath31 denotes the unit circle @xmath43 $ ] by identifying the beginning and ending points .
boldface letters are reserved for vectors and matrices .
for an integer @xmath2 , @xmath44\triangleq\lbra{1,\cdots , n}$ ] .
@xmath45 denotes the amplitude of a scalar or cardinality of a set .
@xmath46 , @xmath47 and @xmath48 denote the @xmath1 , @xmath49 and frobenius norms respectively . @xmath50 and @xmath51 are the matrix transpose and conjugate transpose of @xmath52 respectively .
@xmath53 is the @xmath54th entry of a vector @xmath55 , and @xmath56 denotes the @xmath54th row of a matrix @xmath52 . unless otherwise stated , @xmath57 and @xmath58 respectively reserve the entries of @xmath55 and the rows of @xmath52 in the index set @xmath35 .
for a vector @xmath55 , @xmath59 is a diagonal matrix with @xmath55 being its diagonal .
@xmath60 means @xmath61 for all @xmath54 .
@xmath62 denotes the rank of a matrix @xmath52 and @xmath63 the trace . for positive semidefinite matrices @xmath52 and @xmath64 , @xmath65 means that @xmath66 is positive semidefinite .
@xmath67 denotes the expectation and @xmath68 the probability of an event .
@xmath69 is an estimator of @xmath70 . for notational simplicity , a random variable and its numerical value
will not be distinguished .
the rest of the paper is organized as follows .
section [ sec : atomell0min ] presents the atomic @xmath0 norm method and studies its theoretical guarantees .
section [ sec : convexrelaxation ] turns to the convex relaxation method and its theoretical results .
section [ sec : connection ] discusses connections of our methods to prior arts .
sections [ sec : proofcompletedata ] and [ sec : proofincompletedata ] present proofs of two main theorems in section [ sec : convexrelaxation ] .
section [ sec : simulation ] provides numerical simulations and finally section [ sec : conclusion ] concludes this paper .
we exploit the sparsity to solve the problem of joint sparse frequency recovery . in particular , we seek a set of frequencies of the minimum length among the infinitely many candidates which can express the observed data @xmath27 .
to state it formally , we denote @xmath71 and @xmath72 .
then ( [ formu : observmodel1 ] ) can be written as where @xmath73 and @xmath74 with @xmath75 .
let @xmath76 denote the unit complex @xmath77-sphere or real @xmath78-sphere .
we define the continuous dictionary or the set of atoms we see by ( [ formu : observmodel ] ) that @xmath11 is a linear combination of a number of @xmath5 atoms in @xmath79 .
in particular , we say that a decomposition of @xmath11 as in ( [ formu : observmodel ] ) is an atomic decomposition of order @xmath5 if @xmath80 and the frequencies @xmath81 are distinct . for @xmath82 , we define the smallest number of atoms that can express it as its atomic @xmath0 ( pseudo-)norm : so , we propose to recover the frequencies by minimizing the atomic @xmath0 norm of some @xmath83 which is consistent with the observed data on @xmath8 , i.e. , to solve the following optimization problem : for @xmath84 , denote by @xmath85 a ( hermitian ) toeplitz matrix with where @xmath86 is the @xmath54th entry of @xmath87 .
we provide a finite dimensional formulation of @xmath88 in the following proposition .
@xmath88 defined in ( [ formu : al0 ] ) equals the optimal value of the following rank minimization problem : [ prop : al0_rankmin ] the proof of proposition [ prop : al0_rankmin ] is based on the classical vandermonde decomposition lemma stated as follows .
any positive semidefinite toeplitz matrix @xmath85 of rank @xmath89 has an order-@xmath89 vandermonde decomposition : where @xmath90 and @xmath91 with @xmath92 . moreover , the decomposition is unique if @xmath93 . @xmath94
[ lem : toeplitz ] the vandermonde decomposition is not unique if @xmath95 has full rank .
in fact , we can arbitrarily choose @xmath96 and let @xmath97 .
it follows that the residue @xmath98 is of rank @xmath99 and remains a positive semidefinite toeplitz matrix , which admits a unique vandermonde decomposition of order @xmath99 and further results in a decomposition of @xmath95 of order @xmath2 .
[ rem : vanddecompfullrank ] _ proof of proposition [ prop : al0_rankmin ] : _ let @xmath100 and @xmath101 , where @xmath102 denotes an optimal solution to the rank minimization problem ( [ formu : al0_rankmin ] ) .
we need to show that @xmath103 . on one hand , since @xmath104 we have @xmath105 and @xmath106 .
it follows from lemma [ lem : toeplitz ] that @xmath107 , where @xmath108 , @xmath91 and @xmath92 .
moreover , @xmath83 lies in the range space of @xmath109 and therefore , there exists @xmath110 such that @xmath111 , where @xmath112 denotes the @xmath54th row of @xmath113 . by the definition of the atomic @xmath0 norm in ( [ formu : al0 ] ) we have @xmath114 . on the other hand ,
let @xmath115 be an atomic decomposition of @xmath83 , where @xmath52 is similarly defined as before and @xmath116 .
let @xmath117 and @xmath118 for arbitrary @xmath92 , @xmath119 .
then as a result , @xmath120 defines a feasible solution of the rank minimization problem ( [ formu : al0_rankmin ] ) , and we have that @xmath121 . ' '' '' proposition [ prop : al0_rankmin ] presents a rank minimization problem to characterize the atomic @xmath0 norm .
it follows that ( [ formu : al0min ] ) can be formulated as follows : in ( [ formu : al0_rankmin1 ] ) we need to complete a ( structured positive semidefinite ) low rank matrix @xmath122 with partial access to its entries .
therefore , we establish a link between the frequency recovery problem and low rank matrix completion @xcite .
note that a similar rank minimization problem is presented in @xcite in the smv case , where the rank is put on the matrix @xmath95 rather than the full matrix @xmath122 .
later we will see that ( [ formu : al0_rankmin1 ] ) has a clearer connection to the convex relaxation method presented in section [ sec : convexrelaxation ] .
suppose we could solve the rank minimization problem in ( [ formu : al0_rankmin1 ] ) . according to the proof of proposition [ prop : al0_rankmin ]
, we can obtain the solution of the frequencies by the vandermonde decomposition of @xmath109 , where @xmath123 is an optimal solution .
while lemma [ lem : toeplitz ] states the existence of the decomposition , readers are referred to ( * ? ? ?
* appendix a ) for a detailed approach to its realization .
we next investigate theoretical guarantees of the proposed atomic @xmath0 norm method .
to analyze the atomic @xmath0 norm minimization problem ( [ formu : al0min ] ) or equivalently ( [ formu : al0_rankmin1 ] ) , we generalize the concept of spark in @xcite to the case of continuous dictionary .
we define the following continuous dictionary with respect to the index set @xmath35 similarly to ( [ formu : atomset ] ) : given the continuous dictionary @xmath124 , the quantity spark of @xmath124 , denoted by @xmath125 , is the smallest number of atoms of @xmath124 which are linearly dependent . in general ,
it is np - hard to compute @xmath125 given some sampling index set @xmath35 .
some preliminary results are presented in the following proposition .
the following results hold about @xmath125 with @xmath126 : 1 .
@xmath127 , 2 .
@xmath128 if and only if the elements of is not coprime , and 3 .
@xmath129 if @xmath35 consists of @xmath130 consecutive integers .
[ prop : fullspark ] \1 ) @xmath131 since all atoms of @xmath124 are nonzero .
@xmath132 because @xmath126 and any @xmath133 atoms of @xmath124 are linearly dependent .
\2 ) @xmath134 if and only if there exist @xmath135 such that @xmath136 and @xmath137 are linearly dependent .
we next prove the equivalence between the linear dependence and the non - coprimality .
_ sufficiency : _
if @xmath136 and @xmath137 are linearly dependent , i.e. , @xmath138 with @xmath139 and @xmath140 , then it holds for any @xmath141 that and thus @xmath142 is an integer for any @xmath143 , i.e. , @xmath144 is an integer for any @xmath145 .
it follows that @xmath146 is a rational number .
let @xmath147 , where @xmath148 are coprime positive integers with @xmath149 .
consequently , @xmath150 is a common divisor of the elements of @xmath151 .
_ necessity : _ suppose integer @xmath152 is a common divisor of the elements of @xmath151 .
consider any two @xmath153 satisfying that @xmath154 .
then , for any @xmath141 , the last equality holds since @xmath155 evenly divides @xmath156 .
it follows that @xmath136 and @xmath137 are linearly dependent .
\3 ) if @xmath35 consists of @xmath130 consecutive integers , say , @xmath157 with @xmath158 , then @xmath159 , @xmath160 , are linearly independent for any @xmath130 distinct @xmath161 since the determinant as a result , any @xmath130 atoms of @xmath124 are linearly independent , which together with @xmath132 concludes that @xmath129 .
proposition [ prop : fullspark ] presents the range of @xmath125 with respect to the sampling index set @xmath35 .
a necessary and sufficient condition is provided under which @xmath125 achieves the lower bound 2 , where there exist two atoms that are linearly dependent . in this case , in fact
, for any atom in @xmath124 we can always find one such that they are linearly dependent following from the proof above .
however , such @xmath35 is rare . to see this ,
suppose that @xmath35 is chosen uniformly at random .
then it satisfies the condition with probability no greater than where @xmath162 denotes the number of @xmath130-combinations given an @xmath2-element set .
it is clear that the probability equals 0 when @xmath163 .
numerically , the probability is less than @xmath164 when @xmath165 and @xmath166 respectively .
a sufficient ( but unnecessary ) condition is also provided in the third part under which @xmath124 achieves the upper bound @xmath133 .
we say that @xmath124 is a full spark dictionary if @xmath129 following from the discrete setting in @xcite .
we will see the benefits of a full spark dictionary for frequency recovery in the ensuing subsection .
we provide theoretical guarantees of the atomic @xmath0 norm minimization in ( [ formu : al0min ] ) or the rank minimization in ( [ formu : al0_rankmin1 ] ) for frequency recovery in this subsection .
in particular , we have the following result , which can be considered as a continuous version of ( * ? ? ?
* theorem 2.4 ) .
@xmath167 is the unique optimizer to ( [ formu : al0min ] ) or ( [ formu : al0_rankmin1 ] ) if moreover , the atomic decomposition above is the unique one satisfying that @xmath168 .
[ thm : al0_guanrantee ] suppose that there exists @xmath169 satisfying that @xmath170 and @xmath171 , and @xmath172 is an atomic decomposition of order @xmath173 .
let @xmath174 ( the matrix consisting of those @xmath175 , @xmath176 ) , @xmath177 and @xmath178 .
in addition , let @xmath179 and @xmath180 .
then we have where @xmath181 , @xmath182 , @xmath183 and @xmath184 are matrices of proper dimensions .
it follows that and thus on the other hand , it follows from @xmath170 that note that @xmath58 is composed of atoms in @xmath124 , and has a nontrivial null space by ( [ formu : diffnot0 ] ) .
then by the definition of spark we have moreover , for the nullity ( dimension of the null space ) of @xmath58 it holds that consequently , the equality together with ( [ formu : rankbound ] ) and ( [ formu : nullitybound ] ) yields that @xmath185 , and then which contradicts the condition in ( [ kbound ] ) . to show the uniqueness of the atomic decomposition
, we observe that the condition in ( [ kbound ] ) implies that @xmath186 since @xmath187 .
according to the definition of spark , any @xmath5 atoms in @xmath124 are linearly independent .
therefore , the atomic decomposition is unique given the set of frequencies @xmath188 .
now suppose there exists another decomposition @xmath189 with @xmath173 and a different set of frequencies @xmath190 ( i.e. , there exists some @xmath191 ) .
note that we have repetitively used the same notations for notational simplicity and we similarly define the other notations .
once again we have ( [ formu : diffnot0 ] ) since @xmath192 is nonempty followed by @xmath193 .
the rest of the proof follows from the same arguments .
it is generally difficult to compute @xmath125 given @xmath35 , which makes the condition ( [ kbound ] ) hard to check in a particular scenario .
however , it is checkable in the special complete data case , which is shown in the following corollary .
note that the rank minimization problem ( [ formu : al0_rankmin1 ] ) can still help to recover the frequencies though ( [ formu : al0min ] ) admits a trivial solution .
@xmath167 is the unique atomic decomposition satisfying that @xmath168 if [ cor : al0_guanrantee ] the result follows from @xmath194 given @xmath36 , which holds according to the third part of proposition [ prop : fullspark ] .
theorem [ thm : al0_guanrantee ] shows that the frequencies can be exactly recovered by minimizing the atomic @xmath0 norm or equivalently solving the rank minimization problem ( [ formu : al0_rankmin1 ] ) provided that the sparsity level @xmath5 is sufficiently small .
in particular , the upper bound of @xmath5 depends on a particular sampling index set @xmath35 and the observed data @xmath27 . in the smv case where @xmath195 and therefore , the condition becomes @xmath196 ( or @xmath197 in the complete data case ) .
as we take more measurement vectors , we have a chance to recover more complex signals by potentially increasing @xmath198 ( or @xmath199 ) , which is practically relevant in array processing applications .
when @xmath198 achieves the maximum value @xmath5 , the sparsity level @xmath5 can be as large as @xmath200 ( or @xmath99 in the complete data case ) , which is consistent with the result in array processing ( see e.g. , @xcite ) .
[ rem : l0_coherent ] an unpleasant scenario is in the presence of coherent sources . in an extreme case where all the sources are coherent , i.e. , all the rows of @xmath40 are identical up to some scale factors , it holds that @xmath201 as in the smv case and taking more measurement vectors does not improve the number of frequencies recoverable .
in fact , we can easily verify that all the measurement vectors are identical up to scale factors as well and thus provide the same amount of information for frequency recovery as a smv . in this case
, it is easy to verify that the optimal solution of @xmath87 to ( [ formu : al0_rankmin1 ] ) remains constant when we add or delete columns of @xmath83 . since the upper bound increases with @xmath125 , an interesting topic in the future is to study selection of the sampling index set @xmath35 , which in array processing corresponds to design of the geometry of the sla , such that the spark is maximized .
while preliminary results have been obtained in the discrete frequency setting ( the frequencies are fixed on a uniform grid ) in @xcite , it is worth noting that the continuous setting concerned here is more practical and challenging .
the atomic @xmath0 norm exploits sparsity directly , however , it is nonconvex and the formulated rank minimization problem can not be globally solved with a practical algorithm . to avoid the nonconvexity
, we utilize convex relaxation to relax the atomic @xmath0 norm .
in particular , it can be relaxed in two ways from two different perspectives .
one is to relax the atomic @xmath0 norm to the atomic @xmath1 norm , or simply the atomic norm , which is defined as the gauge function of @xmath202 , the convex hull of @xmath79 @xcite : the atomic norm @xmath203 induced by @xmath79 defined in ( [ formu : atomset ] ) is a norm and thus convex by the property of the gauge function .
the concept of atomic norm is firstly introduced in @xcite and the authors argue that the atomic norm minimization problem is the best convex heuristic for recovering simple models with respect to a given atomic set .
the other way of convex relaxation is based on a perspective of rank minimization illustrated by ( [ formu : al0_rankmin ] ) and to relax the pseudo rank norm to the nuclear norm or equivalently the trace norm for a positive semidefinite matrix , i.e. , to replace @xmath204 by @xmath205 in ( [ formu : al0_rankmin ] ) .
it is worthy noting that the nuclear norm has been extensively studied for low rank matrix completion and recovery @xcite .
interestingly enough , the two ways of convex relaxation are equivalent , which is shown in the following result .
@xmath206 defined in ( [ formu : atomicnorm ] ) equals the optimal value of the following sdp : [ thm : an_sdp ] before presenting the proof of theorem [ thm : an_sdp ] , we clarify some notations first .
we use the following identity whenever @xmath207 : it follows that @xmath208 is finite if and only if @xmath209 is in the range space of @xmath210 .
in fact , ( [ formu : yry ] ) is equivalent to defining @xmath211 when @xmath210 loses rank .
the following result will be used to prove theorem [ thm : an_sdp ] . given @xmath212 , it holds that @xmath213 .
[ lem : lem1 ] we need only to show that for any @xmath214 satisfying @xmath215 it holds that @xmath216 , or equivalently , @xmath217 .
the conclusion follows from that _ proof of theorem [ thm : an_sdp ] : _ provided @xmath218 , we have equivalently @xmath219 and @xmath220 .
so , we need to show that where @xmath221 is the first entry of @xmath87 as before . by the vandermonde decomposition we have @xmath222 , where @xmath223 is composed of columns @xmath224 , and @xmath225 with @xmath92 .
moreover , we can verify that @xmath226 .
for the @xmath17th column of @xmath83 , say @xmath227 , it holds by lemma [ lem : lem1 ] that therefore , we complete the proof via the following equalities : the second last equality holds by the substitutions @xmath228 and @xmath229 , followed by @xmath230 , where @xmath231 denotes the @xmath54th row of @xmath113 .
the last equality follows from the definition of the atomic norm . ' '' ''
the semidefinite formulation ( [ formu : an_sdp ] ) of the atomic norm was firstly reported in our conference paper @xcite .
when preparing this paper , we found that the same result was also independently obtained in @xcite .
other optimization methods can be studied in the future for joint sparse frequency recovery by borrowing ideas in low rank matrix recovery ( see , e.g. , @xcite ) . as an example , the ( nonconvex ) log - determinant criterion @xmath232 , where @xmath233 is a small number , has been studied in @xcite and implemented by a series of sdps via a majorization - maximization ( mm ) process .
theorem [ thm : an_sdp ] shows that the atomic norm can be formulated as an sdp and thus can be globally solved using standard sdp solvers such as sdpt3 @xcite .
note that the semidefinite formulation ( [ formu : an_sdp ] ) generalizes the results in the smv case in @xcite . as a result , by the convex relaxation we propose the following optimization problem for frequency recovery : or equivalently , for a matrix @xmath234 , the dual norm of the atomic norm is defined as where @xmath235 for two matrices ( or column / row vectors ) @xmath52 and @xmath64 of proper dimensions and @xmath236 takes the real part of a complex argument .
the dual problem of ( [ formu : anmin ] ) is thus following from a standard lagrangian analysis @xcite .
on the other hand , the dual problem of ( [ formu : anmin ] ) has the following semidefinite formulation based on the semidefinite formulation ( [ formu : an_sdp ] ) and a standard lagrangian analysis .
that means , the first two constraints in ( [ formu : dual ] ) can equivalently characterize @xmath237 .
we empirically find that the dual problem ( [ formu : dual ] ) can be solved more efficiently than the primal problem ( [ formu : an_sdp ] ) using , for example , sdpt3 , while most solvers can return an optimal solution of ( [ formu : an_sdp ] ) ( as the dual of ( [ formu : dual ] ) ) for free when solving ( [ formu : dual ] ) .
given an optimal solution @xmath238 of ( [ formu : an_sdp ] ) , we can retrieve the frequencies of interest and obtain corresponding atomic decomposition(s ) of @xmath239 as follows .
consider first the case where @xmath109 is rank deficient , i.e. , @xmath240 .
then we can obtain a unique vandermonde decomposition @xmath241 of order @xmath242 by lemma [ lem : toeplitz ] . since @xmath239 is in the range space of @xmath109 , there exist unique @xmath243 , @xmath244 and @xmath245 corresponding to the variables in the proof of theorem [ thm : an_sdp ] such that is an atomic decomposition of @xmath239 with that means , we have obtained a unique atomic decomposition that achieves the atomic norm . on the other hand ,
when @xmath109 has full rank , there exist infinitely many vandermonde decompositions of @xmath109 of order @xmath2 according to remark [ rem : vanddecompfullrank ] .
consequently , we can obtain infinitely many atomic decompositions of @xmath239 of order @xmath2 with each achieving the atomic norm .
therefore , it is impossible to exactly recover the frequencies .
to guarantee exact frequency recovery with the convex relaxation , we require that the frequencies of interest be appropriately separate . to quantify the separation ,
we define _ minimum separation _ of a set of frequencies as follows . for a finite subset @xmath246 ,
the minimum separation of @xmath26 is defined as the closest wrap - around distance between any two elements , for example , the distance between @xmath247 and @xmath248 equals @xmath249 .
inspired by @xcite and @xcite , we study the cases of complete data and incomplete data separately .
our analysis in the former is deterministic and applicable once @xmath250 while that in the latter is based on non - asymptotic analysis of random matrices @xcite .
since the rows of the source signals @xmath40 are allowed to be correlated or even coherent in our analysis , it is _ worst case _ analysis as opposed to the _ average case _ analysis in @xcite . given the complete data , the proposed atomic norm minimization problem ( [ formu : anmin ] ) has a trivial solution , however , the semidefinite formulation ( [ formu : an_sdp ] ) can be used to recover the frequencies .
its theoretical guarantee is presented in the following result , which generalizes that in the smv case in @xcite .
@xmath167 is the unique atomic decomposition satisfying that @xmath251 if @xmath252 and @xmath250 . is more like a technical requirement but not an obstacle in practice ( see numerical simulations in section [ sec : simulation ] ) . ] [ thm : completedata ] see section [ sec : proofcompletedata ] .
theorem [ thm : completedata ] presents a result of uniform recovery in the sense that it holds for all source signals @xmath40 once the frequency separation condition @xmath253 is satisfied . consider the observation model ( [ formu : observmodel1 ] ) which contains @xmath254 real variables ( @xmath255 for @xmath40 and @xmath5 for @xmath256 ) and @xmath257 real measurements . as a result , for guaranteed uniform recovery
, a necessary condition is that @xmath258 or equivalently @xmath259 . on the other hand ,
given the sparsity level @xmath5 , we can always select a set of equispaced frequencies which has a minimum separation of @xmath260 .
therefore , a necessary condition of the minimum separation for uniform recovery is @xmath261 .
[ rem : necessaryseparation ] we do not impose any assumption on the source signals @xmath40 in theorem [ thm : completedata ] . therefore , it can be applied uniformly to all kinds of source signals including correlated or even coherent sources .
when all the sources are coherent as discussed in remark [ rem : l0_coherent ] , taking more measurement vectors does not increase the information for frequency recovery . in this case , in fact , it is easy to show that the proposed atomic norm minimization problem ( [ formu : an_sdp ] ) produces the same solution of @xmath87 up to a positive scale factor and thus the same frequency solution when increasing / decreasing the number of measurement vectors . as a result
, we can not expect a better theoretical guarantee than that in the smv case .
our contribution by theorem [ thm : completedata ] is showing that in the presence of mmvs we can confidently recover the frequencies via a single convex optimization problem by exploiting the joint sparsity therein .
though our _ worst case _ analysis does not improve over that in @xcite on the smv case , it will be shown in section [ sec : simulation ] via numerical simulations that the proposed joint sparse frequency recovery method improves the recovery performance significantly when the source signals are at general positions .
we pose this _ average case _ analysis as a future work .
it is difficult to analyze the atomic norm minimization problem ( [ formu : anmin ] ) or ( [ formu : an_sdp ] ) with a specific sampling index set @xmath35 .
instead , we assume that it is selected uniformly at random .
our result is presented in the following theorem , which generalizes that in the smv case in @xcite with modifications .
suppose we observe @xmath167 on the index set @xmath8 , where @xmath262 is of size @xmath130 and selected uniformly at random .
assume that @xmath263 are independent random variables with @xmath264 . if @xmath252 , then there exists a numerical constant @xmath265 such that is sufficient to guarantee that , with probability at least @xmath266 , @xmath11 is the unique optimizer to ( [ formu : anmin ] ) or ( [ formu : an_sdp ] ) and @xmath167 is the unique atomic decomposition satisfying that @xmath251 .
[ thm : incompletedata ] see section [ sec : proofincompletedata ] .
we emphasize that the dependence of the sample size @xmath130 per snapshot on @xmath6 is for controlling the probability of successful recovery . to make it clear
, we consider the case where we seek to recover the columns of @xmath11 independently via the smv atomic norm minimization .
then with @xmath130 satisfying ( [ formu : an_bound ] ) at @xmath29 as in @xcite , each column of @xmath11 can be recovered with probability @xmath266 .
it follows that @xmath11 can be exactly recovered with probability at least @xmath267 .
however , if we attempt to recover @xmath11 via a single convex optimization problem that we propose , then with the same number of measurements the success probability is improved to @xmath268 ( replacing @xmath269 in ( [ formu : an_bound ] ) by @xmath270 ) . 1 .
compared to @xcite on the smv case , theorem [ thm : incompletedata ] relaxes the assumption of @xmath271 . in particular , the phases are assumed in the former drawn i.i.d . from the uniform distribution on the unit sphere .
in contrast , they are only required to be independent and have zero means
the relaxation of the assumption has significant impact on the practical doa estimation problem in two aspects . on one hand
, each @xmath272 corresponds to one source and therefore , they do not necessarily obey an identical distribution . on the other hand , under the assumption of theorem [ thm : incompletedata ]
the sources are allowed to be ( spatially ) coherent or temporally correlated , which can be encountered in practical scenarios .
for example ,
the source signals @xmath40 are temporarily ( column - wise ) correlated if the rows of @xmath40 are i.i.d .
gaussian with mean zero and non - diagonal covariance .
two sources ( or two rows of @xmath40 ) are certain to be coherent ( identical up to a scale factor ) if they are i.i.d .
gaussian with mean zero and covariance of rank one , where the resulting @xmath272 s satisfy the assumptions in theorem [ thm : incompletedata ] .
in contrast , any two sources with @xmath272 drawn i.i.d . from the uniform distribution on @xmath273 are uncorrelated .
theorem [ thm : incompletedata ] shows that , if we fix the number of measurement vectors @xmath6 , then a number of @xmath4 samples per measurement vector are sufficient to recover the frequencies and the missing data by solving the convex optimization problem ( [ formu : an_sdp ] ) provided that the frequencies are sufficiently separate .
when applied to array processing , it means that a number of @xmath4 sensors are sufficient to exactly determine the directions of @xmath5 sources .
our proofs of theorems [ thm : completedata ] and [ thm : incompletedata ] , which are deferred to sections [ sec : proofcompletedata ] and [ sec : proofincompletedata ] respectively , are inspired by those in @xcite and @xcite and follow similar procedures .
the main challenge of our proofs is dealing with vector - valued dual polynomials induced by the mmv problem .
the problem of joint sparse frequency recovery concerned in this paper has been widely studied within the framework of cs , typically under the topic of doa estimation . since
cs has been focused on signals that can be sparsely represented under a finite discrete dictionary , discretization / gridding of the frequency domain has become standard , or equivalently , the frequencies are assumed to lie on a fixed grid . now recall the atomic @xmath274 norm , @xmath275 , in ( [ formu : al0 ] ) and ( [ formu : atomicnorm ] ) .
it is easy to show that , for @xmath275 , where @xmath276 . as a result , the atomic @xmath0 norm ( or the atomic norm ) generalizes the @xmath277 norm ( or the @xmath278 norm ) in grid - based joint sparse recovery methods ( see , e.g. , @xcite ) to the continuous frequency setting .
it is worth noting that , since the grid - based methods rely on the ideal assumption that the true frequencies lie exactly on a fixed grid , we can not expect exact frequency recovery when it fails unlike the continuous recovery technique studied in this paper .
moreover , even if the assumption holds , existing coherence or rip - based analysis for the @xmath278-based joint sparse recovery is very conservative , compared to the results in this paper , due to the high coherence in the presence of a dense grid .
readers are referred to @xcite for detailed discussions in the smv case . to the best of our knowledge , the only discretization - free / gridless technique for joint sparse frequency recovery is introduced in @xcite prior to this work , termed as the sparse and parametric approach ( spa ) . differently from the atomic norm technique proposed in this paper , spa is from a statistical perspective and based on a weighted covariance fitting ( wcf ) criterion .
but we show next that the two methods are connected .
given the complete data case as an example . in the limiting noiseless case , spa is equivalent to the following optimization problem : where @xmath279 denotes the sample covariance . let @xmath280
. then we have the following equalities / equivalences : where the third equality follows from the semidefinite formulation of the atomic norm in theorem [ thm : an_sdp ] . as a result , spa is equivalent to computing the atomic norm of @xmath281 .
note by ( [ formu : observmodel ] ) that so , spa is the atomic norm method proposed in this paper with modifications of the source signals . in the special smv case where @xmath29 and @xmath282 is a positive scalar ,
the two techniques are exactly equivalent .
we note that a similar result holds with incomplete data and will be omitted .
we have considered the general case where @xmath283 in this paper . for technical reasons ,
our proof is mainly focused on the symmetric case where @xmath284 with @xmath285 , in which all our previous results hold as well with the substitution @xmath286 .
we complete the proof by showing that the same result holds in the general case .
the following proposition presents the so - called _ dual certificate _ which guarantees the optimality of a certain atomic decomposition .
suppose that the atomic set @xmath79 is composed of atoms defined by @xmath287 whose rows are indexed by the set @xmath288 being either @xmath289 or @xmath290 .
then @xmath291 is the unique atomic decomposition satisfying that @xmath292 if @xmath293 and there exists a vector - valued dual polynomial @xmath294 satisfying that where the coefficient matrix @xmath295 . and @xmath296 . ]
[ prop : dualpoly1 ] for a vector - valued polynomial @xmath297 satisfying ( [ formu : cons1 ] ) and ( [ formu : cons2 ] ) , we have by ( [ formu : dualan ] ) that based on ( [ formu : cons1 ] ) and ( [ formu : cons2 ] ) it is easy to show that the last inequality follows from the definition of the atomic norm . on the other hand , it holds that @xmath298 , implying that @xmath299 .
we next show the uniqueness .
suppose that there exists another decomposition @xmath300 satisfying also that @xmath301 .
there must exist some @xmath302 due to linear independence between @xmath303 , @xmath293 , for both the settings of @xmath288 , otherwise the two decompositions will be identical .
we complete the proof by the following contradiction : the rest of the proof is focused on construction of a dual polynomial as in proposition [ prop : dualpoly1 ] , for which the condition @xmath293 naturally holds .
we revisit the proof in @xcite for the smv case in the ensuing subsection and present our proof for the general mmv case after that .
we consider only the case where @xmath284 .
the reason will be clear in the next subsection .
cands and fernandez - granda @xcite consider a polynomial @xmath304 of the following form : where @xmath305 is the squared fejr kernel with coefficients obeying that @xmath306 , @xmath307 .
denote by @xmath308 , @xmath309 and @xmath310 the first three derivatives of @xmath311 . to obtain the coefficients @xmath312 and @xmath313 such that @xmath314 satisfies the constraints of the dual polynomial , they require for @xmath315 that the equality ( [ formu : dualpolycons11 ] ) ensures that @xmath316 satisfies the interpolation condition ( [ formu : cons1 ] ) , while ( [ formu : dualpolycons21 ] ) is used to guarantee the inequality ( [ formu : cons2 ] ) .
( [ formu : dualpolycons11 ] ) and ( [ formu : dualpolycons21 ] ) can be written more compactly as where @xmath317 , @xmath318 , @xmath319 , @xmath320 , @xmath321 is the vector by stacking @xmath322 together and similarly for @xmath323 .
the system of linear equations above is rescaled following from @xcite such that the coefficient matrix of ( [ formu : linsys1 ] ) is symmetric , positive definite , and very close to identity .
then we have where @xmath324 is the schur complement , and the solution @xmath325 is shown to be close to @xmath326 with @xmath327 and @xmath328 upper bounded . to prove that the dual polynomial @xmath316 with the specified coefficients satisfies the condition ( [ formu : cons2 ] ) , the authors divide the the frequency domain @xmath31 into two regions near to and far from the frequencies @xmath26 : with @xmath329 .
@xmath330 is shown directly on @xmath331 , and on each continuous interval of @xmath332 , @xmath333 is shown to be strictly concave and obtains the maximum 1 at each element in @xmath26 .
we summarize some results in @xcite which will be used later .
assume @xmath334 and @xmath335 . 1 . 2
. for @xmath336 ( or @xmath337 ) , 3 . for @xmath338 , for @xmath339 , @xmath315 , and for @xmath340 , theorem [ thm
: completedata ] is a direct result of the proposition below .
under the assumptions of theorem [ thm : completedata ] , there exists a low - frequency vector - valued polynomial as in ( [ formu : dualpoly1 ] ) satisfying ( [ formu : cons1 ] ) and ( [ formu : cons2 ] ) for @xmath288 being either @xmath289 or @xmath290 , where @xmath293 naturally holds .
[ prop : dualpoly ] to prove proposition [ prop : dualpoly ] , we first show that once we can construct such a dual polynomial in the symmetric case , then a similar polynomial can be obtained in the general case .
assume @xmath341 with @xmath285 and @xmath342 .
suppose that @xmath11 has a decomposition in the general case moreover , suppose that we can construct a dual polynomial @xmath343 in the symmetric case satisfying that and @xmath344
. now define @xmath234 with and it can be readily verified that that means , we have constructed a dual polynomial @xmath345 for the general case as in ( [ formu : dualpoly1 ] ) satisfying ( [ formu : cons1 ] ) and ( [ formu : cons2 ] ) with @xmath346 .
we prove proposition [ prop : dualpoly ] in the symmetric case in the rest of this subsection and thus complete the proof of theorem [ thm : completedata ] .
inspired by @xcite , we construct the following vector - valued dual polynomial : where for each @xmath15 , @xmath347 are vector - valued coefficients . again , for @xmath315 , we impose the following equalities : or equivalently , where we denote @xmath348 , @xmath349 and similarly for @xmath350 .
now we have @xmath255 equations with @xmath255 variables in total .
therefore , we can uniquely determine @xmath351 and @xmath352 as in ( [ formu : alphabeta1 ] ) following from ( [ formu : linsysalpbet ] ) . note that the coefficient matrix in ( [ formu : linsysalpbet ] ) is close to identity and hence , @xmath325 should be close to @xmath326 . differently from that in the smv case
, however , @xmath353 and @xmath354 are vector - valued .
the difficulty of the remaining proof is to provide tight bounds for them .
to do this , we introduce the concept of @xmath355 matrix norm and its induced operator norm as follows .
we define the @xmath355 norm of a matrix @xmath356 as and its induced norm of a linear operator @xmath357 as where @xmath358 , @xmath359 and @xmath360 are positive integers . by the definition
, we have @xmath361 and expect to bound @xmath362 and @xmath363 using the induced norm of the operators @xmath364 , @xmath365 .
so we need to calculate the induced norm first .
interestingly , the induced @xmath355 norm is identical to the @xmath366 norm , which is given in the following result .
@xmath367 for any linear operator @xmath368 defined by a matrix @xmath369 such that @xmath370 for any @xmath371 of proper dimension .
[ lem : twoinfinitynorm ] according to the definition of the induced @xmath366 norm , we have @xmath372 , where @xmath373 is the @xmath54th row of @xmath369 .
we first show that @xmath374 .
denote by @xmath375 and @xmath376 respectively the first column and the rest of @xmath371 .
note that if @xmath377 , then @xmath378 and @xmath379 .
hence , we next show that @xmath380 to complete the proof .
note that where @xmath381 .
it is easy to show that if @xmath382 , then @xmath207 and @xmath383 , and vise versa , where @xmath383 means @xmath384 for all the entries @xmath385 of @xmath210 .
that means , the two constraints @xmath207 and @xmath386 are equivalent to @xmath382 . as a result , where @xmath387 denotes the vectorized form of a matrix argument , i.e. , by stacking all its columns as a column vector .
the last inequality follows from that @xmath388 provided @xmath386 , and @xmath389 . following from lemma [ lem : twoinfinitynorm ] , we use only @xmath390 rather than @xmath391 to denote the induced @xmath355 norm hereafter for notational simplicity , which also avoids ambiguities between the matrix norm and the induced operator norm of a matrix ( a matrix can also denote a linear operator ) .
the following lemma provides upper bounds for @xmath351 and @xmath352 . under the assumptions of proposition
[ prop : dualpoly ] , the matrices @xmath392 determined by ( [ formu : linsysalpbet ] ) satisfy that it follows that @xmath393 , @xmath394 . as in ( [ formu : alphabeta1 ] ) we have it follows that @xmath395 , and the interval of @xmath396 is a direct result of the inequality provided @xmath397 , @xmath119 . utilizing @xmath398 and @xmath399 , we have we now complete the proof of proposition [ prop : dualpoly ] by showing that the polynomial @xmath345 constructed above satisfies ( [ formu : cons2 ] ) .
on @xmath331 , on each interval @xmath400 , @xmath315 , since @xmath401 and @xmath402 , we need only to show that @xmath403 is a concave function , i.e. , we observe that moreover , and it follows that finally , we conclude that for incomplete data , we also begin with the dual certificate stated in the proposition below , which generalizes proposition [ prop : dualpoly1 ] for complete data .
suppose that the atomic set @xmath79 is composed of atoms defined by @xmath287 whose rows are indexed by the set @xmath288 being either @xmath289 or @xmath290 .
then @xmath291 is the unique optimizer to ( [ formu : anmin ] ) or ( [ formu : an_sdp ] ) if @xmath404 are linearly independent and there exists a vector - valued dual polynomial @xmath294 satisfying that where the coefficient matrix @xmath295 .
moreover , @xmath291 is the unique atomic decomposition satisfying that @xmath292 .
[ prop : dualpoly_inc ] the proof is similar to that of proposition [ prop : dualpoly1 ] .
note that the dual problem of ( [ formu : anmin ] ) or ( [ formu : an_sdp ] ) is ( [ formu : dual1 ] ) , and strong duality holds between them .
we can similarly show that @xmath237 and @xmath405 . since @xmath406 is primal - dual feasible , we conclude that @xmath11 is a primal optimal solution due to the strong duality .
suppose that there exists another optimal solution @xmath407 with an atomic decomposition @xmath408 satisfying that @xmath409 .
since @xmath410 and that @xmath404 are linearly independent , there must exist some @xmath302 , otherwise the two decompositions will be identical and @xmath411 . consequently , we have the following contradiction : showing that @xmath11 is the unique optimizer . with similar arguments , we can conclude the uniqueness of the atomic decomposition of @xmath11 .
the condition that @xmath404 are linearly independent is used to prove the uniqueness of the optimizer to ( [ formu : anmin ] ) or ( [ formu : an_sdp ] ) . from this point of view , proposition ii.4 of @xcite or at least its proof for the special smv case is flawed , where this condition is omitted .
however , we will show later that it can be satisfied for free when we construct the dual polynomial @xmath345 .
similarly to the proof of theorem [ thm : completedata ] , we first revisit the smv case in @xcite , where a scalar - valued dual polynomial is constructed inspired by @xcite and proven to satisfy the three constraints listed in proposition [ prop : dualpoly_inc ] .
while the uniform sampling model is difficult to analyze directly , an equivalent bernoulli observation model is studied instead following from @xcite , where we say `` equivalent '' in the sense that the probability that ( [ formu : anmin ] ) or ( [ formu : an_sdp ] ) fails to recover the original signal @xmath11 under the uniform model is at most twice of that under the bernoulli model .
assume in the bernoulli model that the samples indexed by @xmath284 are observed independently with probability @xmath412 , i.e. , we observe about @xmath130 entries on average . in mathematics ,
we let @xmath413 be i.i.d .
bernoulli random variables such that where @xmath414 or @xmath247 indicates whether we observe the @xmath54th entry in @xmath288 .
it follows that the sampling index set @xmath415 . to deal with the randomness of the observation model
, @xcite considers a random dual polynomial where @xmath416 , a random analog of @xmath305 , denotes a random kernel as follows : with @xmath417 defined in ( [ formu : gnj ] ) .
it is clear that @xmath418 and similarly for its derivatives .
the proof of @xcite is mainly based on that the random kernel @xmath416 is concentrated tightly around its expectation @xmath419 as the sample size is large enough . as in the complete data case ,
let us denote @xmath420 , @xmath421 and @xmath422 , and then obtain the system of linear equations as in ( [ formu : linsys1 ] ) by imposing conditions as in ( [ formu : dualpolycons11 ] ) and ( [ formu : dualpolycons21 ] ) , where @xmath423 , @xmath351 , @xmath352 and @xmath424 are defined in the same manner .
the authors of @xcite show that the coefficient matrix @xmath425 in ( [ formu : linsys2 ] ) is concentrated tightly around its expectation @xmath426 , where @xmath427 is the deterministic analog of @xmath428 .
that means , @xmath429 is close to @xmath430 with high probability , and hence , @xmath325 is well defined by ( [ formu : linsys2 ] ) and close to @xmath431 with high probability .
according to ( [ formu : randpoly1 ] ) and ( [ formu : linsys2 ] ) , @xmath432 satisfies the interpolation condition ( [ formu : cons1_inc ] ) and the support condition ( [ formu : cons3_inc ] ) . to further prove that it obeys ( [ formu : cons2_inc ] ) , the authors show that the random perturbation between @xmath432 ( and its derivatives ) and its deterministic analog @xmath316 ( and its derivatives ) in section [ sec : detourcomplete ] can be arbitrarily small provided the sample size @xmath433 is sufficiently large , firstly on a set of grid points and then extended to the continuous domain @xmath31 . the proof is completed by showing that @xmath434 ( or its second derivative ) can be arbitrarily close to @xmath333 ( or its second derivative ) on @xmath331 ( or @xmath332 ) .
we list some useful results in @xcite which are also necessary in our later proof .
we start with some notations . for @xmath435 $ ] ,
define the event on @xmath436 , @xmath429 is guaranteed to be invertible . then we introduce the partitions @xmath437 and @xmath438 , where @xmath439 , @xmath440 , @xmath441 and @xmath210 are of the same dimension .
let and similarly define its deterministic analog @xmath442 , where @xmath443 denotes the @xmath444th derivative of @xmath445 and @xmath315 .
let @xmath269 be a small positive number and @xmath265 a constant which may vary from instance to instance .
assume @xmath446 .
we have @xmath447 if [ lem : invertibility ] assume @xmath446 .
let @xmath435 $ ] and consider a finite set @xmath448 .
then , we have for some constant @xmath449 , where @xmath450 and @xmath451 [ lem : i1bound ] assume @xmath446 . on the event @xmath436
, we have for some constant @xmath452 .
[ lem : i2bound ] theorem [ thm : incompletedata ] is a direct result of the following proposition . under the assumptions of theorem [ thm : incompletedata ]
, then there exists a numerical constant @xmath265 such that is sufficient to guarantee that with probability at least @xmath266 there exists a vector - valued polynomial as in ( [ formu : dualpoly1_inc ] ) satisfying ( [ formu : cons1_inc])([formu : cons3_inc ] ) for @xmath288 being either @xmath289 or @xmath290 , where @xmath404 are linearly independent .
[ prop : dualpoly1_inc ] similarly to the case of complete data , we first argue that we can consider only the symmetric case where @xmath284 . in particular , we attempt to show that , under the assumptions of proposition [ prop : dualpoly1_inc ] , if we can construct a vector - valued dual polynomial in the symmetric case , then a similar polynomial can be obtained in the general case .
our arguments are similar to those in subsection [ sec : proofcompdata ] for complete data .
we omit the details but emphasize some important features . for a fixed set of frequencies @xmath26 , @xmath453 defined in ( [ formu : yotwodecompositions ] )
are mutually independent , with @xmath454 and @xmath455 under the same assumptions of @xmath456 .
moreover , the linear independence of @xmath457 remains unchanged .
the rest of the proof is to show the existence of a dual polynomial as in proposition [ prop : dualpoly1_inc ] in the symmetric case .
based on lemma [ lem : invertibility ] we can easily obtain the following result , which shows the linear independence of @xmath457 .
assume @xmath446 .
then there exists a numerical constant @xmath265 such that @xmath458 is sufficient to guarantee that , with probability at least @xmath266 , @xmath457 are linearly independent .
[ lem : linearindependence ] according to ( [ formu : ckbar ] ) , we have where @xmath459 is a diagonal matrix composed of the entries @xmath460 , @xmath461 , and thus has full rank .
denote @xmath462 , i.e. , the matrix formed by @xmath457 .
it follows that on the event @xmath436 , since @xmath427 , as well as @xmath463 , is invertible , @xmath58 must have full column rank provided @xmath464 , i , e . , @xmath457 are linearly independent .
let us fix @xmath465 .
then the bound of @xmath130 follows directly from lemma [ lem : invertibility ] . for a specific sampling index set @xmath35 ,
it is generally infeasible to compute the quantity @xmath125 as mentioned in section [ sec : spark ] .
here we show that if @xmath35 is selected uniformly at random , then @xmath466 is sufficient to guarantee that with high probability any @xmath5 atoms in @xmath124 are linearly independent under a frequency separation condition .
define the _ conditional spark _ of @xmath124 as then we have @xmath467 .
this will be of independent interest .
inspired by @xcite , we construct the following vector - valued polynomial where @xmath445 is the random kernel , and @xmath468 and @xmath469 are vector - valued coefficients as in the complete data case .
it is clear by ( [ formu : ckbar ] ) that @xmath470 above satisfies the support condition ( [ formu : cons3_inc ] ) . to make it satisfy ( [ formu : cons1_inc ] ) and ( [ formu : cons2_inc ] )
, we impose again that i.e. , where @xmath349 and similarly for @xmath350 .
it follows that , on the event @xmath436 , @xmath471 , and where @xmath472 is defined in ( [ formu : vl ] ) . as in @xcite
, we decompose @xmath473 into three parts : which results in a decomposition on @xmath474 : where @xmath475 is the dual polynomial ( [ formu : dualpoly ] ) in the complete data case with @xmath476 . here
we have defined denote the event we attempt to prove that @xmath477 happens , i.e. , @xmath474 and @xmath478 are close to each other on a fixed grid , with high probability provided @xmath130 is sufficiently large , which is summarized in the following proposition .
suppose @xmath479 is a finite set of grid points .
under the assumptions of proposition [ prop : dualpoly1_inc ] , there exists a numerical constant @xmath265 such that if then [ prop : boundgrid ] to prove proposition [ prop : boundgrid ] , we need to show that @xmath480 and @xmath481 are small on @xmath482 . differently from the smv case in @xcite , however , both @xmath480 and @xmath481 are vector - valued and the difficulty is to provide tight bounds on their norms . to do this , we present the following lemma which corresponds to the vector - form heoffding s inequality .
let the rows of @xmath483 be sampled independently on the complex hyper - sphere @xmath273 with zeros means . then , for all @xmath484 , @xmath485 , and @xmath486 , [ lem : vectorheoffding ] see appendix [ sec : appendheoffding ] .
the vector - form heoffding s inequality generalizes the common scenario with @xmath29 which is used in @xcite to bound the scalars @xmath487 and @xmath488 by utilizing lemma [ lem : i1bound ] and lemma [ lem : i2bound ] , respectively . following the same procedures as in @xcite with minor modifications of the coefficients , we can bound @xmath487 and @xmath488 on @xmath482 with high probability , which is summarized in the ensuing lemma [ lem : i1 ] and lemma [ lem : i2 ] respectively .
we omit their proofs and interested readers are referred to the proofs of lemmas iv.8 and iv.9 in @xcite .
under the assumptions of proposition [ prop : dualpoly1_inc ] , there exists a numerical constant @xmath265 such that if then we have [ lem : i1 ] under the assumptions of proposition [ prop : dualpoly1_inc ] , there exists a numerical constant @xmath265 such that if then we have [ lem : i2 ] now we can conclude proposition [ prop : boundgrid ] by combining ( [ formu : decomposition ] ) , lemma [ lem : i1 ] and lemma [ lem : i2 ] with suitable redefinition of @xmath265 , @xmath489 and @xmath269 .
we have shown in proposition [ prop : boundgrid ] that @xmath474 and @xmath478 are close to each other on a fixed grid @xmath490 with high probability .
we extend the result to the whole circle @xmath31 in the following proposition .
under the assumptions of proposition [ prop : dualpoly1_inc ] , there exists a numerical constant @xmath265 such that if then with probability @xmath266 , we have [ prop : appendcontinuous ] see appendix [ sec : appendcontinuous ] .
now we are ready to show that the vector - valued polynomial @xmath470 constructed above satisfies ( [ formu : cons2_inc ] ) , and so complete the proof together with lemma [ lem : linearindependence ] . under the assumptions of proposition [ prop : dualpoly1_inc ] ,
there exists constant @xmath265 such that if then with probability @xmath266 we have the proof is based on proposition [ prop : appendcontinuous ] and the results in the complete data case in section [ sec : proofcompletedata1 ] . on @xmath331 , we have so , it suffices to choose @xmath491 . on each interval @xmath400 , @xmath315
, we also accomplish the proof by showing that according to the results in section [ sec : proofcompletedata1 ] , it is easy to verify that for a numerical constant @xmath492 .
let @xmath493 .
by proposition [ prop : appendcontinuous ] we have moreover , with simple derivations we have it follows that recall that @xmath494 as @xmath495 .
so we have as @xmath489 is a sufficiently small numerical value . with this choice of @xmath489
, we obtain from ( [ formu : mboundcontinu ] ) the following ( slightly stronger ) condition
we consider the complete data case and test the frequency recovery performance of the proposed atomic norm method with respect to the frequency separation condition . in particular , we consider two types of frequencies , equispaced and random , and two types of source signals , uncorrelated and coherent .
we fix @xmath496 and vary @xmath497 ( a lower bound of the minimum separation of frequencies ) from @xmath498 ( or @xmath499 for random frequencies ) to @xmath500 at a step of @xmath501 . in the case of equispaced frequencies , for each @xmath497 we select a set of frequencies @xmath26 of the maximal cardinality @xmath502 with the frequency separation @xmath503 . in the case of random frequencies ,
we generate the frequency set @xmath26 , @xmath504 , by repetitively adding new frequencies ( generated uniformly at random ) till no more can be added .
therefore , any two adjacent frequencies in @xmath26 are separate by a value in the interval @xmath505 .
it follows that @xmath506 $ ] .
we empirically find that @xmath507 which is the mid - point of the interval above .
we first consider uncorrelated sources , where the source signals @xmath508 in ( [ formu : observmodel1 ] ) are drawn i.i.d . from a standard complex gaussian distribution .
moreover , we consider the number of measurement vectors @xmath509 . for each value of @xmath497 and each type of frequencies ,
we carry out 20 monte carlo runs and calculate the success rate of frequency recovery . in each run
, we generate a set of frequencies @xmath26 and @xmath510 and obtain the complete data @xmath11 . for each value of @xmath6
, we attempt to recover the frequencies by the proposed atomic norm method , implemented by sdpt3 @xcite in matlab , based on the first @xmath6 columns of @xmath11 . given the frequency solution @xmath511 and _ mean amplitude _ @xmath512 ( see section [ sec : freqretriev ] ) , we denote by @xmath513 the vector of the largest @xmath5 amplitudes ( the rest by @xmath514 ) and by @xmath515 the corresponding set of frequencies
. we may consider the rest of the frequencies in @xmath516 are spurious peaks with amplitudes @xmath514 .
the recovery is considered successful if @xmath517 , root mean squared error ( rmse ) of frequency @xmath518 and maximum absolute error of amplitude @xmath519 , where @xmath520 denotes the true vector of amplitudes .
our simulation results are presented in figs .
[ fig : complete_l_uniformfreq ] and [ fig : complete_l_nonuniformfreq ] , which verify the conclusion of theorem [ thm : completedata ] that the frequencies can be exactly recovered by the proposed atomic norm method under a frequency separation condition . moreover , when we take more measurement vectors , the performance of recovery improves and it seems that a weaker frequency separation condition is sufficient to guarantee exact frequency recovery in this case of uncorrelated sources . by comparing fig .
[ fig : complete_l_uniformfreq ] and fig .
[ fig : complete_l_nonuniformfreq ] , we also observe that a stronger frequency separation condition is required in the case of equispaced frequencies where more frequencies are present and located more closely .
it is worthy noting that with the equispaced frequencies the minimum separation at which the atomic norm method starts to succeed is close to the necessary separation condition in remark [ rem : necessaryseparation ] that @xmath261 .
we next consider coherent sources . in this simulation , we fix @xmath521 and consider different percentages , denoted by @xmath522 , of the @xmath5 source signals which are coherent ( identical up to a complex scale factor ) .
that is , @xmath523 refers exactly to the case of uncorrelated sources considered previously .
@xmath524 means that all the sources signals are coherent and the mmv case is equivalent to the smv case . for each type of frequencies ,
we consider five values of @xmath522 ranging from @xmath525 to @xmath526 and calculate each success rate over 20 monte carlo runs .
our simulation results are presented in figs .
[ fig : complete_cohsource_uniformfreq ] and [ fig : complete_cohsource_nonuniformfreq ] .
it is shown that , as the percentage of coherent sources increases , the success rate decreases and a stronger frequency separation condition is required for exact frequency recovery .
as @xmath522 equals the extreme value @xmath526 , the curves of success rate approximately match those at @xmath29 in figs .
[ fig : complete_l_uniformfreq ] and [ fig : complete_l_nonuniformfreq ] , verifying that taking more measurement vectors does not necessarily improve the performance of frequency recovery . finally , we report the computational speed of the proposed atomic norm method .
it takes about 11s to solve one sdp on average and the cpu times differ slightly at the three values of @xmath6 .
about 22 hours are used in total to produce the data used in fig .
[ fig : completedata ] . for incomplete data , we study the so - called phase transition phenomenon in the @xmath527 plane .
in particular , we fix @xmath496 , @xmath521 and @xmath528 , and study the performance of our proposed atomic norm minimization method in signal and frequency recovery with different settings of the source signal .
the frequency set @xmath26 is randomly generated with @xmath504 and @xmath529 ( differently from that in the last subsection , the process of adding frequencies is terminated as @xmath529 ) . in our simulation
, we vary @xmath530 and at each @xmath130 , @xmath531 since it is difficult to generate a set of frequencies with @xmath532 under the aforementioned frequency separation condition . in this simulation , we consider temporarily correlated sources . in particular
, suppose that each row of source signal @xmath113 has a toeplitz covariance matrix @xmath533 ( up to a positive scale factor ) .
therefore , @xmath534 means that the source signals at different snapshots are uncorrelated while @xmath535 means completely correlated . in our simulation
, we first generate @xmath536 from an i.i.d .
standard complex gaussian distribution and then let @xmath537 , where we consider @xmath538 . for each combination @xmath527 , we carry out 20 monte carlo runs and calculate the rate of successful recovery with respect to each @xmath89 .
the recovery is considered successful if @xmath517 , the relative rmse of data recovery @xmath539 , the rmse of frequency recovery @xmath540 and the maximum absolute error of amplitude recovery @xmath541 , where @xmath242 and the last two metrics are defined as in the previous simulation , and @xmath239 denotes the solution of @xmath83 .
our simulation results are presented in fig .
[ fig : phasetransition ] , where a transition from perfect recovery to complete failure can be observed in each subfigure .
more frequencies can be recovered when more samples are observed .
moreover , when the correlations between the mmvs , indicated by @xmath89 , increase , the phase of successful recovery decreases , and at the same time the phase transition becomes less clear . in the extreme case
where @xmath542 , all the measurement vectors are completed correlated , which in fact is equivalent to the smv case . therefore , by comparing fig .
[ fig : pt100 ] and the first three subfigures , we can conclude that the performance of frequency recovery improves in general when more measurement vectors are observed .
it is also worth noting that @xmath543 corresponds to the complete data case considered in the previous simulation .
we also plot the line @xmath544 in figs .
[ fig : pt00]-[fig : pt90 ] and @xmath545 in fig .
[ fig : pt100 ] ( straight gray lines ) which are upper bounds of the sufficient condition in theorem [ thm : al0_guanrantee ] for the atomic @xmath0 norm minimization .
we see that successful recoveries can be obtained even above the lines , indicating good performance of the proposed atomic norm minimization method .
again , we report the computational speed .
it takes about 13s on average to solve each problem and almost 200 hours in total to generate the whole data set used in fig .
[ fig : phasetransition ] . while this paper has been focused on the noiseless case ,
one naturally wonders the performance of the proposed method in the practical noisy case .
we provide a simple simulation as follows to show this .
we set @xmath546 , @xmath547 with @xmath35 randomly generated , @xmath548 sources with frequencies of @xmath549 , @xmath550 and @xmath551 and powers of 2 , 3 and 1 respectively , and @xmath521 .
the signals of each source are generated with constant amplitude and random phases .
complex white gaussian noise is added to the observed samples with noise variance @xmath552 .
we attempt to denoise the observed noisy signal @xmath553 and recover the frequency components by solving the following optimization problem : where @xmath489 , set to @xmath554 ( mean + twice standard deviation ) , upper bounds the frobenius norm of the noise with large probability .
our simulation results of one monte carlo run are presented in fig .
[ fig : noisycase ] .
the smv case is studied in fig .
[ fig : noisycase1 ] , where only the first measurement vector is used for frequency recovery .
it is shown that the three frequency components are correctly identified by the atomic norm minimization method while music fails .
the mmv case is studied in fig .
[ fig : noisycase2 ] with uncorrelated sources , where both atomic norm minimization and music succeed to identify the three frequency components .
the coherent source case is presented in fig .
[ fig : noisycase3 ] , where we modify source 3 in fig .
[ fig : noisycase2 ] such that it is coherent with source 1 .
music fails to detect the coherent sources as expected while the proposed method still performs well .
all of the three subfigures show that spurious frequency components can be present using the atomic norm minimization method , however , their powers are insignificant . to be specific , the spurious components have about @xmath555 of the total powers in fig .
[ fig : noisycase1 ] , and this number is on the magnitude of @xmath556 in the latter two subfigures . since the numerical results imply that the proposed method is robust to noise , a theoretical analysis should be investigated in future studies .
the proposed method takes about @xmath557s in each scenario .
finally , it is worthy noting that the proposed method requires to know the noise level while music needs the source number .
we studied the joint sparse frequency recovery problem in this paper which arises in practical array processing applications .
we presented nonconvex and convex optimization methods for its solution and analyzed their theoretical guarantees under no or very weak assumptions of the source signals .
our results extend the smv atomic norm methods and their theoretical guarantees in @xcite to the mmv case , extend the existing discrete joint sparse recovery framework to the continuous dictionary setting , and provide theoretical guidance for the array processing applications .
while this paper is focused on the worst case analysis , it will be interesting to investigate in the future the average case under stronger assumptions of the source signals , in which numerical simulations of this paper suggest that a weaker frequency separation condition is sufficient for exact recovery as the number of measurement vectors increases .
the vector - form heoffding s inequality in lemma [ lem : vectorheoffding ] is a corollary of the following matrix - form heoffding s inequality in @xcite .
consider a finite sequence @xmath558 of independent , random , self - adjoint matrices with dimension @xmath559 , and let @xmath560 be a sequence of fixed self - adjoint matrices .
assume that each random matrix satisfies then , for all @xmath486 , where @xmath561 and @xmath562 denotes the largest eigenvalue .
we now prove lemma [ lem : vectorheoffding ] based on the _ dilation _ technique @xcite .
in particular , the dilation of a vector @xmath563 is a self - adjoint matrix @xmath564 .
it is not difficult to show that @xmath565 and @xmath566 .
we let @xmath567 and @xmath568 , where @xmath569 is the @xmath15th row of @xmath424 with @xmath570 .
it follows that @xmath558 are independent matrices of dimension @xmath571 with @xmath572 according to the assumptions of @xmath573 . by the aforementioned properties of the dilation
, we have that @xmath574 and @xmath575 .
so , we can choose @xmath576 and it follows that @xmath577 .
we complete the proof by applying the matrix - form heoffding s inequality to @xmath558 :
the proof is based on the following bernstein s polynomial inequality .
let @xmath578 be any polynomial of degree @xmath579 on complex numbers with derivative @xmath580 . then
, [ lem : bernstein ] on the event @xmath581 , we make the following decomposition for some @xmath582 and @xmath583 : the second term has been bounded in proposition [ prop : boundgrid ] .
we next provide a upper bound for @xmath584 , while the same bound is applicable to @xmath585 under similar arguments . since @xmath586 is vector - valued and @xmath587 for any @xmath563 .
we attempt to bound @xmath588 elementwise .
denote @xmath589 the @xmath54th entry of @xmath590 .
then we have for some constant @xmath265 following from @xcite by noticing that @xmath591 , where @xmath592 denotes the @xmath54th column of @xmath424 . viewing @xmath593 as a polynomial of @xmath594 of degree @xmath595 , we get by applying the bernstein s polynomial inequality that it follows that then we can select @xmath482 satisfying @xmath596 such that for any @xmath582 there exists a point @xmath583 satisfying that @xmath597 .
consequently , we have @xmath598 , which together with ( [ formu : decomp ] ) gives that on @xmath581 finally , ( [ formu : mboundcontinu ] ) is a direct consequence of ( [ formu : mboundgrid ] ) by inserting that @xmath596 .
when ( [ formu : mboundcontinu ] ) is satisfied , we have @xmath599 by proposition [ prop : boundgrid ] and lemma [ lem : invertibility ] .
z. yang would like to thank gongguo tang for fruitful discussions .
z. yang and l. xie , `` continuous compressed sensing with a single or multiple measurement vectors , '' in _ 2014 ieee workshop on statistical signal processing ( ssp ) , gold coast , australia , available online at https://dl.dropboxusercontent.com/u/34897711/ssp14.pdf_ , june 2014 .
e. cands , j. romberg , and t. tao , `` robust uncertainty principles : exact signal reconstruction from highly incomplete frequency information , '' _ ieee transactions on information theory _ ,
52 , no . 2 ,
489509 , 2006 .
d. l. donoho and m. elad , `` optimally sparse representation in general ( nonorthogonal ) dictionaries via @xmath600 minimization , '' _ proceedings of the national academy of sciences _ , vol .
100 , no . 5 , pp . 21972202 , 2003 .
l. hu , z. shi , j. zhou , and q. fu , `` compressed sensing of complex sinusoids : an approach based on dictionary refinement , '' _ ieee transactions on signal processing _ , vol . 60 , no . 7 , pp .
38093822 , 2012 .
z. yang , c. zhang , and l. xie , `` robustly stable signal recovery in compressed sensing with structured matrix perturbation , '' _ ieee transactions on signal processing _ , vol .
60 , no . 9 ,
pp . 46584671 , 2012 .
d. malioutov , m. cetin , and a. willsky , `` a sparse signal reconstruction perspective for source localization with sensor arrays , '' _ ieee transactions on signal processing _ ,
53 , no . 8 , pp . 30103022 , 2005 .
s. f. cotter , b. d. rao , k. engan , and k. kreutz - delgado , `` sparse solutions to linear inverse problems with multiple measurement vectors , '' _ ieee transactions on signal processing _
53 , no . 7 , pp . 24772488 , 2005 .
r. gribonval , h. rauhut , k. schnass , and p. vandergheynst , `` atoms of all channels , unite ! average case analysis of multi - channel sparse recovery using greedy algorithms , '' _ journal of fourier analysis and applications _ , vol .
14 , no . 5 - 6 , pp . 655687 , 2008 .
j. m. kim , o. k. lee , and j. c. ye , `` compressive music : revisiting the link between compressive sensing and array signal processing , '' _ ieee transactions on information theory _ , vol .
58 , no . 1 ,
pp . 278301 , 2012 .
z. yang , l. xie , and c. zhang , `` a discretization - free sparse and parametric approach for linear array signal processing , '' _ ieee transactions on signal processing , accepted with madatory minor revisions ( aq ) , available at http://arxiv.org/abs/1312.7695_ , 2014 .
j. b. kruskal , `` three - way arrays : rank and uniqueness of trilinear decompositions , with application to arithmetic complexity and statistics , '' _ linear algebra and its applications _ , vol .
18 , no . 2 ,
pp . 95138 , 1977 .
m. malek - mohammadi , m. babaie - zadeh , a. amini , and c. jutten , `` recovery of low - rank matrices under affine constraints via a smoothed rank function , '' _ ieee transactions on signal processing _ , vol .
62 , no . 4 , pp .
981992 , 2014 . | frequency recovery / estimation from samples of superimposed sinusoidal signals is a classical problem in statistical signal processing .
its research has been recently advanced by atomic norm techniques which deal with continuous - valued frequencies and completely eliminate basis mismatches of existing compressed sensing methods .
this work investigates the frequency recovery problem in the presence of multiple measurement vectors ( mmvs ) which share the same frequency components , termed as joint sparse frequency recovery and arising naturally from array processing applications . @xmath0- and @xmath1-norm - like formulations , referred to as atomic @xmath0 norm and the atomic norm ,
are proposed to recover the frequencies and cast as ( nonconvex ) rank minimization and ( convex ) semidefinite programming , respectively .
their guarantees for exact recovery are theoretically analyzed which extend existing results with a single measurement vector ( smv ) to the mmv case and meanwhile generalize the existing joint sparse compressed sensing framework to the continuous dictionary setting .
in particular , given a set of @xmath2 regularly spaced samples per measurement vector it is shown that the frequencies can be exactly recovered via solving a convex optimization problem once they are separate by at least ( approximately ) @xmath3 . under the same frequency separation condition , a random subset of @xmath2 regularly spaced samples of size @xmath4 per measurement vector is sufficient to guarantee exact recovery of the @xmath5 frequencies and missing samples with high probability via similar convex optimization .
extensive numerical simulations are provided to validate our analysis and demonstrate the effectiveness of the proposed method .
* keywords : * array processing , atomic norm , compressed sensing , direction of arrival ( doa ) estimation , frequency recovery / estimation , joint sparsity , low rank matrix completion , multiple measurement vector ( mmv ) . |
ngc 6888 is an emission nebula associated with the wolf - rayet star wr 136 with spectral type wn6(h ) , and it is a typical example of a ring nebula resulting from the continuous interaction between a strong stellar wind and the interstellar medium .
projected on the sky , this nebula seems to be a filamentary and clumpy gaseous ellipse emitting mainly in h@xmath4 and optical [ n ii ] lines , with an axis - size of @xmath5 arcmin @xcite , and it is surrounded by a homogeneous sphere which emits mainly in [ o iii ] 5007 @xcite .
the combination of its filamentary morphology and its nitrogen enhanced abundance suggests that the ejections , coming from a previous evolutionary stage of the star wr 136 , have contributed in an important way to the formation of the ring nebula that we see today @xcite .
both the nebula ngc 6888 and its central star wr 136 have been already studied in several wavelength ranges from x - rays to radio emission .
for example , the stellar parameters , such as the stellar temperature , mass - loss rate and terminal velocity , have been calculated by @xcite and @xcite by using stellar atmosphere codes and comparing the results with observed ultraviolet , optical and infrared stellar spectra ; while the wind velocity law and clumpiness were studied by @xcite .
@xcite showed that wr 136 emits in x - ray energies .
in particular , these authors found soft emission ( @xmath6 kev ) coming from outside the wind and hard emission ( @xmath7 kev ) coming from the gas inside the wind .
the complete hot bubble in the interior of ngc 6888 was observed with the _ space telescope suzaku _ @xcite , confirming the existence of nitrogen enriched x - ray emitting plasma which consists of two thermal components , one emiting soft x - rays and the other one , hard x - rays .
based on this , the authors proposed a new x - ray emission mechanism based on the thermal interaction between heavy particles , given that the known mechanisms are not capable to explain simultaneously these two x - ray emitting components .
however , as these authors explain , it is necessary to have better observational restrictions to confirm this idea .
on the other hand , complete hydrodynamical simulations performed by @xcite , attempting to reproduce the x - ray observations , indicate that the nebular morphology plays a key role in the presence of x - rays in ring nebulae .
these authors conclude that the x - ray emission is higher in bubbles with shell - like morphology than the one present in clumpy filamentary bubbles . however , as the authors indicate , the stellar evolutionary models are not able to reproduce simultaneously the observed stellar parameters and the spectral signatures of the observed x - ray luminosity .
the physical properties in different zones of ngc 6888 , such as electronic temperature and density , expansion velocity and chemical composition , have been investigated by several authors since @xcite to recently @xcite .
ev92 made a chemodynamical study and they found that ngc 6888 has a chemical composition revealing the material processed in the interior of a star with an initial mass between 25 m@xmath8 and 40 m@xmath8 .
these results were confirmed by m - d14 who detected for the first time the c ii 4267 recombination line , allowing to estimate the c abundance . in this way they presented a complete study of the trace of cno cycle in the nebula .
fm12 made a very complete study of ngc 6888 through integral field spectroscopy observations performing 1d and 2d analyses .
their results show indeed that ngc 6888 has a complex structure with shells and filaments moving with different velocities .
in particular they proposed an onion - like " morphology composed mainly by three shells . concerning the plasma properties ,
they found that the electron density ranges from less than 100 to 360 @xmath9 and the electron temperature presents variations from @xmath107 700 to @xmath1010 200 k. in addition , they reported important variations in the nitrogen abundance across the nebula , the n / o ratio being much higher in the inner region .
recently @xcite derived n / o values from the infrared [ n iii ] 57-@xmath11 m and [ o iii]-88 @xmath11 m lines and their results would indicate that the edge region of ngc 6888 , over the major axis , has lower n abundance with respect to the inner region near the star .
all the previous works focus only on the modelling of the star or the nebula but not both simultaneously .
further , when a single study of these systems is made ( star or nebula ) it is difficult to assign a value to the physical parameters only based on generic data .
this is why we focus in a semiempirical method based in the self - consistent modelling of the star and its associated nebula .
it is logical to proceed in this way because the nebula associated to the star is naturally an additional restriction to the stellar atmosphere model , because the nebula can tell us about the ionizing ultraviolet photons coming from the star and then reprocessed and reemitted with low energy through recombination and collisional processes .
in this way we can realize how well a stellar model can produce the ionization stage of the nebula and for this purpose we build a photoionization model of ngc 6888 based on spectroscopic observations where the modelled stellar spectral energy distribution ( sed ) is used as the ionization source .
this methodology will give us the advantage to obtain more constrained and coherent physical values for all the components of the entire system than those resulting from a single study of each component as it was made until now .
the results from this work will help to better constraint other kind of studies as models of stellar evolution , astrophysical bubble formation and , in particular , the presence of x - rays in the bubbles since this hot gaseous component imposes an additional and important restriction to understand the whole system , given that is there where the stellar wind energy is stored .
this paper is structured as follows : in section 2 , observations and data reduction process are described .
3 contains the information of the line flux measurements and the error estimates . in sect .
4 we describe the stellar and nebular models as well as our methodology to constraint the free parameters .
a discussion about our best self - consistent model is presented in sect .
5 . finally , in sect . 6 , we present our conclusions .
[ cols="<,^,^,^,^,^ " , ] the comparison between the model predictions and the observed dereddened line fluxes is presented in table [ tab : res1 ] and table [ tab : res2 ] corresponding to the photionization model results by using _ star1 _ and _ star3 _ sed s as ionizing source , respectively .
parameter , in these tables , shows the goodness of our model and is defined following @xcite as : + @xmath12 where @xmath13 is the value returned by the model , @xmath14 is the observed flux and @xmath15 is the accepted tolerance in dex of this observable .
we define the tolerance to be + @xmath16 where @xmath17 is the measured error .
when a line flux is well reproduced the _
value is between @xmath181.0 and 1.0 .
the results in table [ tab : res1 ] show that , under the assumptions made for this model and using _
star1 _ as ionizing source , the observed line fluxes are not well reproduced .
in particular the lines from highly ionized elements ( o@xmath19 , ar@xmath19 ) are extremely overpredicted .
the best case is for the values in position a where only half of the whole line fluxes ( the low excitation ones ) detected in this region are well reproduced .
the model results are worse for the other positions .
the fact that the ionization of the nebula is strongly overpredicted could be due to : 1 ) a too high ionization parameter @xmath20 where @xmath21 is the ionizing photon rate , @xmath22 the hydrogen number density and @xmath23 the distance to the ionizing source .
] , 2 ) a too high stellar effective temperature .
notice that this model was obtained using only the high density component ( @xmath24 ) . adding any contribution of the low density component would made the situation even worse .
as @xmath23 , @xmath21 and @xmath25 are well constrained through the _ hipparcos _ distance to the object , its observed angular size , the observed stellar luminosity and the nebular density coming from the plasma diagnostic , then it follows that @xmath20 is also well constrained .
the only remaining way to reproduce the observed low value of the [ o iii]/[o ii ] line ratio is to consider a star with a temperature lower than the value derived for model _ star1 _ , i.e. , we must not deal with the number of ionization photons but rather with their energy .
the fit improves very much when we use the model _
star3 _ which has @xmath26 70 000 k. the ionization of the dense gas is even lower than observed , leading to the full use of our hypothesis of the 2-components model .
the predicted emission lines from the corresponding photoionization model are compared with observations in table [ tab : res2 ] .
it can be seen that almost all the fluxes are well reproduced , indicating that the star wr 136 emits the uv ionizing photons corresponding to a star with this low effective temperature .
as it can be seen in figure [ fig : stelspec ] and it was discussed in 4.1 , at least three stellar models , ( _ star1 , star2 _ and _ star3 _ ) with different @xmath27 and therefore a different @xmath28 fit similarly well the stellar observations , in particular the stellar continuum and several emission lines .
the differences among these models appear mainly in the he lines , where the he ii / he i line ratio is well reproduced when considering an effective temperature of 110 000 k but it is understimated when the effective temperature is reduced , i.e. , at a lower stellar temperature the stellar he i lines are stronger and the stellar he ii lines become weaker than observed . as we discussed before , it seems that there is not a unique solution for the stellar model . in figure
[ fig : spectrauv ] we show a comparison between the uv region of the stellar models from table [ tab : w136 ] .
we also overplot the uv behaviour of two blackbodies with temperatures of 70 000 and 46 000 k which correspond to the temperatures @xmath27 and @xmath29 of model _
we can see that the key difference between the stellar models comes from the far uv region of the spectra , whose uv photons are absorbed by the nebula ( between 1.0 and @xmath10 7.0 ryd ) .
this difference is well appreciated when we compare the models with the blackbodies . from this figure
it is clear that a blackbody with the same temperature than the modelled spectra , 70 000 or even 46 000 k , shows an excess of uv photons which would lead to a much more ionized nebula .
it is also evident the high absorption of the uv photons emitted by the star , due to the presence of the stellar wind .
owing to there are not stellar observations in the far uv to discriminate which stellar model is reproducing the real spectral energy distribution of the star wr 136 , it is necessary to add an external restriction . in this case ,
such additional restriction is the associated nebula ngc 6888 that is highly sensitive to this uv ionizing radiation .
in addition , the combination of the temperature and mass - loss effects in the stellar wind , considered for this model , can reproduce the line profiles coming from recombination process . therefore
, if we are close to a good value for @xmath27 , it is the same for the mass - loss rate , @xmath30 .
this is important because @xmath30 represents a good constrain for the hydrodynamical simulations .
as in this study we have found a well constrained value for @xmath30 ( log @xmath30 = @xmath184.95 ) , which is similar to the values found by other authors ( table [ tab : w136 ] ) , we can rule out the assumption made recently by @xcite where they conclude that it is necessary a @xmath30 3 - 4 times lower than observed for this type of stars to reproduce the x - ray flux of the emitting hot bubble . . the black solid line and the dashed line correspond to blackbodies of 70 000 and 46 000 k , respectively ,
according to the temperatures @xmath27 and @xmath29 of model _
star3_.,scaledwidth=50.0% ] the presence of a very fast stellar wind coming from the star wr 136 could imply that part of the ionization of the gas is due to shocks .
diagrams made by @xcite and @xcite are usually employed to determine if a line emitting region is mainly ionized by ob stars ( like h ii regions ) or by other source ( agn , liners , shocks ) .
see for example @xcite for the case of shocks . in figure
[ fig : bpt ] we show an [ oiii]5007/h@xmath31 vs. [ nii ] 6584/h@xmath4 diagram .
the solid line shows the separation between global photoionization models for star forming regions ( nsf galaxies ) and agns as defined by @xcite .
the dashed line represents the same separation , but shifted on the log ( [ n ii ] 6584/h@xmath4 ) axis by 0.96 dex , which is the difference between the n / o adopted by @xcite and the n / o of ngc 6888 deduced in this work .
the three black diamonds correspond to observations for positions a , b and c and are obviously located to the right side of the solid line , but well inside the photoionized zone with n / o ratio similar to ngc 6888 value . the effect of a limited aperture size , which is not including the whole nebula is illustrated by the position of the color points in our fig .
[ fig : bpt ] .
they show the emission line ratios of every spaxel of our 2d images obtained from our models .
the colors correspond to different values of @xmath32 ( the weight of the dense component ) , the blue symbols are obtained for the case of considering only the low - density component , and the red ones for the case of considering only the high - density component , the green symbols being an intermediate situation .
as we can see , the points spread in an ample region of the diagram .
the black square symbol is the position in the diagram of the global model , as if it were fully included in the slit .
therefore if just a small portion of a whole nebula is observed , one must be careful in using diagnostic diagrams to decide if the emission of a nebula is caused by the presence of shocks , because as we demonstrated in fig .
[ fig : bpt ] only photoionization and an enhanced n abundance are sufficient to obtain points located out of the regions represented by global photoionized models ( solid line ) . to deeper investigate our assumption of chemical homogeneity of ngc 6888 , in fig .
[ fig : bpt ] we added the observations by ev92 , m - d14 .
it is found that they lie inside or close to our model points .
also we added the observations by fm12 ( marked as black dots and stars ) resulting from their 1d analysis of different regions of ngc 6888 .
we can see that their points show a large spread in this diagram . to explain the behaviour of their observations and to derive values of the n and o abundances
, fm12 performed a grid of photoionization models using the code cloudy varying the ionization parameter and the n / o ratio . as they used global models for each observation without taking into account that they were observing only a part of the nebula they concluded that there are important differences in the n / o ratio across ngc 6888 .
however , our figure shows that this is not necessarily the case . in particular ,
let us focus on the data of fm12 marked as black dots , which correspond to their regions inside ngc 6888 with large nitrogen enrichment .
we can see that all these points , but the point labeled x2 in their paper , are close to our model points and are following their trend .
the rest of the points ( stars ) show a large dispersion compared to our model points and it is not expected for the model to reproduce their positions because they belong to regions located out from the edge of ngc 6888 with low ( solar ) nitrogen abundance . therefore , figure [ fig : bpt ] illustrates the fact that all the observed points _ within _ the nebula can be reproduced by a rather simple chemically homogeneous model . /o@xmath3
ratio from a quadrant of our best photoionization model which is chemically homogeneous .
the star is assumed to be at the origin of the map .
three different values of _ w _ were used to show how the values of the icf changes over the whole structure of nebula .
our positions and the position of fm12 are shown . ] to continue the discussion about the chemical homogeneity in ngc 6888 , we analyse the behaviour of the icf used to determine the n / o ratio .
let us remember that it is generally assumed that n@xmath3/o@xmath3 is equal to n / o . from our photoionization model , we are able to determine in each region of the nebula the icf needed for the n@xmath3/o@xmath3 ratio to obtain the n / o value .
we show in figure [ fig : no ] a map of this icf for three different values of _ w _ : 10@xmath33 , 10@xmath34 and 10@xmath35 .
it is evident that for a given _ w _ , this icf varies over the nebula showing a difference as large as a factor of two . in adittion
, we can see that the value of this icf changes as the contribution of the high - density component is modified .
for example , in the case when _
w _ = 10@xmath35 , the icf for position a is around 1.0 . in general
, we can see in figure [ fig : no ] that for positions closer to the star the icf is around 1.5 or larger .
therefore , in the zones near to the star an icf greater than one is needed to estimate the n / o ratio .
hence , this indicates that n / o could be understimated when it is simply assumed that n@xmath3/o@xmath3 is equal to n / o .
in this way , to calculate the nitrogen abundance for position b2 of fm12 and for ev92 a better value for this icf is 1.5 . once this icf is used for these positions the n / o ratio increases relative to the values in table [ tab : par ] , to 0.47 for ev92 and @xmath180.15 for b2 fm12 .
therefore these values are still discrepant .
in particular , for the position b2 the n / o ratio reaches a value of @xmath180.15 lower than the others for the rim , which is contrary to the predictions of stellar evolution models .
the differences seen of the icf in different zones of the map are due to the combination of two effects : the ionization structure of the nebula and the relative contribution to the fluxes of the high density clumpy region .
a similar problem with the icfs is affecting the results of @xcite who used _ herschel _ observations of the [ o iii ] 88 @xmath11 m and [ n iii ] 57 @xmath11 m lines , emitted in two positions of ngc 6888 , to determine the n / o abundance ratio under the assumption that n@xmath19/o@xmath19 is equal to n / o .
for ngc 6888 this is a very rough approximation because only a small fraction of the nebula was included in their slits and @xcite did not take into account the nebular ionization structure , which indicates that the ion n@xmath19 is only dominant over n@xmath3 in the zones near to the star .
the previous discussion illustrates very well that detailed modelling ( even the simplest as we do here : 2 almost roundish homogeneous components ) is a very powerful way to identify effects that , if not considered , can easily drive the observer to wrong conclusions .
we computed several atmosphere models using the code cmfgen to reproduce the stellar observations of wr 136 .
the characteristics of the star such as @xmath36 , @xmath30 and @xmath37 agree with the values found by other authors , i.e. , hgl06 and @xcite .
we also derived the stellar abundances of he , c , n , o , si , p , s and fe ( listed in table [ tab : abun ] ) which agree with hgl06 s model . we find that models with very different effective temperature , @xmath27 from 70 000 k to 110 000 k , can reproduce fairly well the stellar continuum and several emission lines , although the he lines , indicative of temperature , are worse reproduced by the cooler models .
therefore there is a degeneracy in atmosphere models , than apparently , as pointed out by @xcite , is due to our misunderstanding of the structure of the wind velocity field . to decide which stellar model produces the correct uv amount of ionizing photons coming from wr 136
, we computed a photoionization model for the associated ring nebula ngc 6888 , using the sed produced by the stellar models .
this led us to conclude that wr 136 follows better the spectral behaviour of a star with a @xmath27 of 70 kk .
the photoionization model was built assuming chemical homogeneity and including two components : a low - density region enclosing a denser region .
when these two gaseous components are considered and assuming only photoionization , we were capable to reproduce the observed ionization structure in both shells , contrary to the presence of shocks as the probable mechanism producing the observed ionization structure of ngc 6888 .
we can not immediatly rule out the importance of shock effects , but given that some diagnostic diagram such as the [ o iii ] 5007/h@xmath31 vs. [ n ii ] 6584/h@xmath4 indicate emission of h ii type , and that we can reproduce several emission features of the nebula by only assuming photoionization , we established that the photoionization heating is more important than the presence of shocks .
this latter idea was already mentionned by ev92 combining the diagnostic diagrams of @xcite with their observations in a zone close to the central star and shown in figure [ fig : slit ] .
the nebular abundances derived from our photoionization model are similar to the values derived from our position a and the values reported by ev92 and m - d14 .
it is worth to notice that both the stellar atmosphere of wr 136 and the nebular plasma present a large n- and c - enhancement and an o depletion , as it can be seen in table [ tab : abun ] .
this contributes to the idea that the observed gas was ejected by the central star through the mass - loss driven by the stellar wind .
considering only photoionization and a chemically homogeneous nebula , we were able to explain the pretended differences found in the n / o ratio ( derived from n@xmath3/o@xmath3 or n@xmath19/o@xmath19 in different zones of the nebula ) by fm12 and @xcite , who proposed a chemically inhomogeneous nebula .
these pretended differences are due to effects of the ionization structure within the nebula . with our simple model
we show that the icfs for global nebulae should not be use for observations of small zones observed in extended nebulae . from our results
we have learned that there are several stellar models with different set of values that could reproduce the observables , like the models _
star1 _ and _ star3 _ presented in this study
. therefore external constraints are crucial to determine which atmosphere models represent the real stellar properties .
the nebulae around stars can help to deal with the degeneracy mentioned above .
finally , in this work we show the advantages of modelling simultaneously the stellar and nebular spectra of a wolf - rayet star plus nebula system .
we verify whether a stellar atmosphere model of the star wr 136 provides a good description of the stellar observations and is able to satisfactorily reproduce the nebular characteristics such as the ionization degree .
this allow us to better constraint all the physical parameters of both the star and the nebula , including their physical conditions and their chemical abundances .
this work received partial finantial support from unam papiit in109614 and conacyt cb-2010/153985 . the fuse and
iue data presented in this paper were obtained from the multimission archive at the space telescope science institute ( mast ) .
stsci is operated by the association of universities for research in astronomy , inc .
, under nasa contract nas5 - 26555 .
support for mast for non - hst data is provided by the nasa office of space science via grant nag5 - 7584 and by other grants and contracts .
we thank g. koenigsberger , a. peimbert and j. garca - rojas for their helpful comments .
99 aggarwal k. m. , keenan f. p. 1999
, apjs , 123 , 311 allen mark g. , groves brent a. , dopita michael a. , sutherland ralph s. , kewley lisa j. , 2008 , apjs , 178 , 20 baldwin j. a. , phillips m. m. , terlevich r. , 1981 , pasp , 93 , 5 cardelli jason a. , clayton geoffrey c. , mathis john s. , 1989 , apj , 345,245 chu y .- h . , treffers r. r. , kwitter k. b. 1983 , apjs , 53 , 937 crowther p. a. 2007 , ara&a , 45 , 177 crowther p. a. , smith l. j. 1996 , a&a , 305 , 541 eenens p.r.j . , williams p. m. , 1994 , mnras , 269 , 108 esteban c. , vlchez j. m. 1992 , apj , 390 , 536 , ( ev92 ) ferland g. j. , korista k. t. , verner d. a. , ferguson j. w. , kingdon j. b. , verner e. m. , 1998 , pasp , 110 , 761 ferland g. j. , porter r. l. , van hoof p. a. m. , williams r. j. r. , abel n. p. , lykins m. l. , shaw g. , henney w. j. , stancil p. c. 2013 , rmaa , 49 , 13 fernndez - martn a. , martn - gordon d. , vlchez j. m. , prez montero e. , riera a. , snchez s. f. , 2012 , a&a , 541 , 119 , ( fm12 ) galavis m.e . , mendoza c. , zeippen c. j. 1995 , a&as 111 , 347 galavis m. e. , mendoza c. , zeippen c. j. 1997 , a&as , 123 , 159 garca - segura g. , langer n. , mac low m .- m .
1996 , a&a , 316 , 133 gray r. o. 1999 , ascl.soft10002g gruendl r. a. , chu y .- h . ,
dunne b. c. , points s. d. 2000 , aj , 120 , 2670 hamann w .-
, grafener g. , liermann a. , 2006 , a&a , 457 , 1015 , ( hgl06 ) hillier d. j. , 1991 , iaus , 143 , 59 hillier d. j. , miller d. l. 1998 , apj , 496 , 407 hillier d. j. , lanz t. , heap s. r. , hubeny i. , smith l. j. , evans c. j. , lennon d. j. , bouret j. c. 2003 , apj , 588 , 1039 ignace r. , cassinelli j. p. , quigley m. , babler b. 2001 , apj , 558 , 771 ignace r. , quigley m. f. , cassinelli j. p. , 2003 ,
apj , 596 , 538 kauffmann g. , heckman t. m. , tremonti c. , brinchmann j. , charlot s. , white s. d. m. , ridgway s. e. , brinkmann j. , fukugita m. , hall p. b. , zeljko i. , gordon t. r. , donald p. s. 2003 , mnras , 346 , 1055 kaufman v. , sugar j. 1986 , jpcrd , 15 , 321 kingsburgh r. l. , barlow m. j. 1994 , mnras , 271 , 25 kwitter k. b. , 1981 , apj , 245 , 154 lamers h. j. g. l. m. , cassinelli j. p. introduction to stellar winds / cambridge u press , 1999 levine s. , chakrabarty d. 1994 , tevhnical report mu-94 - 04 , instituto de astronoma , universidad nacional autnoma de mxico .
luridiana v. , morisset c. , shaw r. a. , 2015 , a&a , 573,42 mendoza c. 1983 , i.a.u .
103 , p. 143
mesa - delgado a. , esteban , c. , garca - rojas j. reyes - prez j. morisset c. , bresolin f. , 2014 , apj , 785 , 100 , ( m - d14 ) moore brian d. , hester j. jeff , scowen paul a. , 2000 , aj , 119 , 2991 morisset c. 2013 .
sites.google.com/site/cloudy3d morisset c. , georgiev l. , 2009 , a&a , 507 , 1517 prez - montero e. , contini t. 2009 , mnras , 398 , 949 podobedova l. i. , kelleher d. e. , wiese w. l. 2009 , jpcrd , 38,171 porter r. l. , ferland g. j. , storey p. j. , detisch m. j. 2013 , mnras 433 , 89 pradhan a. k. , montenegro m. , nahar s. n. , eissner w. 2006 , mnras , 366 , 6 prinja raman k. , barlow m. j. , howarth ian d. 1990 , apj , 361,607 sabbadin f. , minello s. , bianchini a. 1977 , a&a , 60 , 147 schmutz w. , hamann w .-
, wessolowski u. , 1989 , a&a , 210,236 skinner stephen l. , zhekov svetozar a. , gdel manuel , schmutz werner , sokal kimberly r. , 2010 , aj , 139 , 825 stock d. j. , barlow m. j. , 2014 , mnras , 441 , 3065 storey p.j .
, zeippen c.j .
, mnras , 312 , 813 , 2000 tayal s. s. , 2007 , apjs , 171 , 331 tayal s. s. 2011 , apjs , 195 , 12 tayal s. s. , zatsarinny , o. 2010 , apjs , 188 , 32 toal j. a. , arthur s. j. 2011 .
apj , 737 , 100 van der hucht k.a . , 2001 ,
new astron .
, 45 , 135 van der hucht k.a .
, cassinelli j.p . , williams p.m. , 1986 , a&a , 168 , 111 van leeuwen f. , 2007 , a&a , 474 , 653 .
van marle a. j. , langer n. , & garca - segura g. 2005 , a&a , 444 , 837 veilleux s. , osterbrock d. , e. , 1987 , apjs , 63 , 295 wiese w. l. , fuhr j. r. , deters t. m. , jpcrd monograph(7 ) pp .
1 - 522 ( 1996 ) zeippen c. j. 1982 , mnras , 198 , 111 zhekov s. a. , park s. 2011 , apj , 728 , 135 | we analyse whether a stellar atmosphere model computed with the code cmfgen provides an optimal description of the stellar observations of wr 136 and simultaneously reproduces the nebular observations of ngc 6888 , such as the ionization degree , which is modelled with the pycloudy code .
all the observational material available ( far and near uv and optical spectra ) were used to constrain such models .
we found that even when the stellar luminosity and the mass - loss rate were well constrained , the stellar temperature @xmath0 at @xmath1 = 20 , can be in a range between 70 000 and 110 000 k. when using the nebula as an additional restriction we found that the stellar models with @xmath2 70 000 k represent the best solution for both , the star and the nebula .
results from the photoionization model show that if we consider a chemically homogeneous nebula , the observed n@xmath3/o@xmath3 ratios found in different nebular zones can be reproduced , therefore it is not necessary to assume a chemical inhomogeneous nebula .
our work shows the importance of calculating coherent models including stellar and nebular constraints .
this allowed us to determine , in a consistent way , all the physical parameters of both the star and its associated nebula .
the chemical abundances derived are 12 + log(n / h ) = 9.95 , 12 + log(c / h ) = 7.84 and 12 + log(o / h ) = 8.76 for the star and 12 + log(n / h ) = 8.40 , 12 + log(c / h ) = 8.86 and 12 + log(o / h ) = 8.20 .
thus the star and the nebula are largely n- and c- enriched and o - depleted . [ firstpage ] circumstellar matter stars : wolf - rayet stars . |
it has been recognized that the relation between classical and quantum integrable systems is useful for studying non - perturbative properties of supersymmetric gauge theories and the ads / cft correspondence @xcite . the ode / im correspondence @xcite provides an interesting example of this classical / quantum correspondence , which relates the spectral determinants of certain ordinary differential equations to the bethe ansatz equations in the massless limit of certain integrable models .
it is an interesting problem to make a complete list of this ode / im correspondence .
the ordinary differential equations for the integrable models related to classical lie algebras have been proposed in @xcite .
the wronskian of the solutions obeys the functional relations called the @xmath1-system , which leads to the bethe ansatz equations of the related quantum integrable system .
the @xmath1-system for classical lie algebra has been reformulated in the form of the matrix valued linear differential equations @xcite , where the bethe ansatz equations of the integrable models associated with the untwisted affine lie algebra @xmath3 of a classical lie algebra @xmath4 are related to the linear differential equations associated with the langlands dual @xmath5 . the ode / im correspondence has been generalized to massive integrable models .
it was found that for the classical sinh - gordon equation modified by a conformal transformation , the spectral problem for the associated linear problem leads to the functional relations of the quantum sine - gordon model @xcite . by taking the conformal limit
, it reduces to the ode / im correspondence for the schrdinger type differential equation @xcite .
recently the massive ode / im correspondence has been generalized to a class of modified affine toda field equations @xcite . in particular , locke and
one of the present authors studied the modified affine toda equations for affine lie algebra @xmath6 , where @xmath7 is an untwisted affine lie algebra including exceptional type @xcite .
it has been shown that from their associated linear problems one obtains the @xmath1-system which leads to the bethe ansatz equations for the affine lie algebra @xmath8 .
it would be an interesting problem to explore the modified affine toda field equation with an affine lie algebra @xmath8 which is not of the form of the langlands dual of an untwisted one , where the corresponding integrable models are not identified yet . in this paper
we will work with the modified affine toda field equation associated with the affine lie algebra @xmath0 ( or @xmath9 ) , which provides the simplest and nontrivial example .
this equation also appears in the study of the area of minimal surface with a null - polygonal boundary in @xmath10 spacetime @xcite , which is dual to the gluon scattering amplitudes with specific momentum configurations .
the equation of motion of strings is described by the @xmath0 affine toda field equation modified by the conformal transformation .
the stokes problem of the associated linear system determines the functional equations for the cross - rations of external momenta .
these functional relations are known to be the same as the y - system of the homogeneous sine - gordon model @xcite and the free energy of the y - system determines the area of the minimal surface .
the purpose of this paper is to apply the massive ode / im correspondence to the modified @xmath0 affine toda field equation and to investigate the functional relations for the stokes coefficients of the linear problem , which include the bethe ansatz equations , the t - q relations , the t - system and the y - system .
we study the boundary condition of the t - system arising from the nontrivial monodromy of the linear problem solution around the origin .
this monodromy condition also appears in the study of the form factors via the ads / cft correspondence @xcite .
this paper is organized as follows : in sect .
2 , we introduce modified @xmath0 affine toda equation and the associated linear problem . in sect .
3 , we discuss the @xmath1-system and derive the bethe ansatz equations . in sect .
4 , we discuss the spinor representation of @xmath11 in detail and study the quantum wronskian and the t - q relations . in sect . 5 , we argue the t - system and y - system and their boundary conditions which come from the monodromy of the solution of the linear system around the origin . in sect . 6 , we investigate the high - temperature limit of the y - system in the presence of monodromy .
7 is devoted for conclusions and discussion . in the appendix
, we summarize the auxiliary t - functions and their functional relations used in this paper .
the lie algebra @xmath12 has simple roots @xmath13 and @xmath14 , where @xmath15 ( @xmath16 ) is an orthonormal basis of @xmath17 .
we denote the highest root by @xmath18 , which is given by @xmath19 , and define the extended root @xmath20 . @xmath21 and @xmath22 are the fundamental weights satisfying @xmath23 .
let @xmath24 ( @xmath16 , @xmath25 ) be the chevalley basis of @xmath11 , where @xmath26 is the set of roots .
let @xmath27 be the two - component scalar field on the complex plane with coordinates @xmath28 .
we define the modified affine toda field equation for @xmath0 by @xmath29 where @xmath30 , @xmath31 , @xmath32 is a mass parameter and @xmath33 is a coupling .
@xmath34 is a holomorphic function of @xmath35 and is chosen as @xmath36 with @xmath37 and @xmath38 is a complex parameter .
this equation is obtained by the conformal transformation @xmath39 with @xmath40 and the field redefinition @xmath41 , where @xmath42 is the co - weyl vector .
, which is obtained by the same conformal transformation but a different field redefinition @xmath43 .
this modified equation is related to ( [ eq : mafe1 ] ) by a field redefinition . ]
note that the coxeter number of @xmath11 is 4 .
( [ eq : mafe1 ] ) can be written in the form of the compatibility condition @xmath44=0 $ ] of the linear differential equations defined in a @xmath11-module : @xmath45 where the connections are defined by @xmath46 here @xmath47 is the spectral parameter .
we are interested in the special class of solutions of ( [ eq : mafe1 ] ) , which satisfy the periodicity condition @xmath48 and the boundary conditions at infinity and the origin of the complex plane : @xmath49 where we have introduced the polar coordinate @xmath50 by @xmath51 and and @xmath52 is a 2-vector satisfying @xmath53 ( @xmath54 ) . due to the special form ( [ eq : pz1 ] ) of @xmath34 , ( [ eq : mafe1 ] ) and the linear problem are invariant under the symanzik rotation @xmath55 for an integer @xmath56 .
this also acts on the solution @xmath57 , which is denoted as @xmath58 .
the linear problem is also invariant under the transformation : @xmath59 where @xmath60 .
we now consider the solutions of the linear differential equations ( [ eq : linear1 ] ) in the basic @xmath0-module @xmath61 ( @xmath62 ) associated with the highest weight @xmath63 .
let @xmath64 be the orthonormal basis of @xmath61 with @xmath65 eigenvalue @xmath66 , where @xmath67 . for the lie algebra @xmath11 , @xmath68 is 5-dimensional vector representation , whose matrix representation is given by @xmath69 and @xmath70 .
here @xmath71 denotes the matrix whose @xmath72-element is @xmath73 .
similarly , @xmath74 is a 4-dimensional spinor representation .
its matrix representation is given by @xmath75 and @xmath76 .
we are interested in the small ( or subdominant ) solution @xmath77 , which decays fastest along the real axis .
this was studied in @xcite for @xmath6 for an untwisted affine lie algebra @xmath78 .
in general , the small solution @xmath77 at large @xmath79 is given by @xmath80 with @xmath81 being a normalization constant . here
@xmath82 and @xmath83 denote the eigenvector and its eigenvalue of the matrix @xmath84 with the eigenvalue of the largest real part . applying the symanzik rotation @xmath85 ( @xmath86 ) ,
one obtains the small solution @xmath87 in the stokes sector @xmath88 for the vector representation ( [ eq : vectrep1 ] ) , the eigenvalues of @xmath89 are @xmath90 ( @xmath91 ) and @xmath92 . for the spinor representation ( [ eq : spinrep1 ] ) , they are @xmath93 and @xmath94 . for @xmath68 ,
one has two eigenvalues with the largest real part and the corresponding solutions in @xmath68 are not subdominant along the real axis .
so we introduce the @xmath95-rotated symanzik solution @xmath96 .
this is a solution of the linear problem with the @xmath95-rotated connection @xmath97 and @xmath98 which is obtained by replacing @xmath99 in ( [ eq : conn1 ] ) .
then the @xmath96 behaves along the real positive axis as ( [ eq : subdom1 ] ) with @xmath100 and @xmath101 .
we define the basis of the solutions around @xmath102 behaves as @xmath103 : @xmath104 which are invariant under @xmath85 @xcite .
the small solution @xmath105 and @xmath106 can be expanded in this basis as @xmath107 we call @xmath108 the q - functions . from the relation @xmath109 ,
the coefficients @xmath110 satisfy the quasi - periodicity condition : @xmath111
the linear problem in the basic @xmath0-modules @xmath61 can be also defined in other @xmath11-modules corresponding to the ( anti-)symmetrized tensor product of @xmath61 s .
the inclusion maps between the modules induce the relation between the small solutions , which is called the @xmath1-system @xcite .
for example , we consider the inclusion map @xmath112 by these maps the highest weight state @xmath113 is mapped to @xmath114 and @xmath115 to @xmath116 .
we use this map to relate the solutions of the linear problem defined on the different modules .
@xmath117 is a solution of the linear problem ( [ eq : linear1 ] ) on @xmath118 due to invariance of ( [ eq : linear1 ] ) under the symanzik rotation @xmath119 .
this solution is mapped into the module @xmath120 by @xmath121 .
now @xmath122 is the unique solution in @xmath120 with the same asymptotic behavior at large @xmath79 . in a similar way we can identify @xmath123 with @xmath105 .
thus we obtain the @xmath1-system : @xmath124 expanding the small solutions in the basis @xmath125 and substituting them into the @xmath1-system , one obtains the functional relation for the q - functions @xmath126 and @xmath127 : @xmath128 denoting the zeros of the q - functions @xmath129 by @xmath130 ( @xmath131 ) , one obtains the bethe ansatz equations @xmath132 note that these differ from those of the integrable model based on the @xmath133 @xcite , which is expected from the langlands duality between @xmath134 and @xmath0 .
the bethe ansatz equations for @xmath135 do not include the squared q - functions
. it would be interesting to study the solutions of the bethe ansatz equations ( [ eq : bae1 ] ) in the conformal limit and explore the corresponding integrable model .
the @xmath1-system in the previous section has been obtained by investigating the asymptotic solution in a single stokes sector , @xmath136 for example .
now we consider the solutions of the linear problem in the whole complex plane .
we focus on @xmath74 because this is the minimal dimensional representation and the solution in the vector representation can be constructed via the inclusion map @xmath137 .
since we are considering a @xmath138 spinor , it is natural to introduce the charge conjugation . associated with the linear problem ( [ eq : linear1 ] ) in the spinor representation , we define the transposed linear problem : @xmath139 the solution @xmath140 of these equations are related to @xmath141 by the charge conjugation : @xmath142 where @xmath143 this is a @xmath144 symmetry of the linear problem .
note that @xmath145 .
one can define the inner product @xmath146 between @xmath147 and @xmath148 .
the inner product is independent of @xmath35 and @xmath149 when @xmath150 ( @xmath151 ) is a solution of the ( transposed ) linear problem .
the wronskian of any four linearly independent solutions @xmath152 ( @xmath153 ) @xmath154 is also independent of @xmath35 and @xmath149 .
we define the @xmath155-rotated solution @xmath156 in the module @xmath74 .
this is the subdominant solution in the stokes sector @xmath157 but it gives a divergent solution in the sectors @xmath158 and @xmath159 .
one can choose @xmath160 as a basis of the solutions .
we normalize the solution @xmath161 such that @xmath162 by choosing the normalization constant @xmath163 in ( [ eq : subdom1 ] ) as @xmath164 . from the asymptotic behavior of @xmath161 and @xmath165 at large @xmath79 , we find @xmath166 and @xmath167 . then from the condition ( [ eq : nor1 ] ) we find @xmath168 we write it in the form @xmath169 .
since the basis @xmath170 is orthonormal , we can fix the normalization of @xmath171 as @xmath172 which simplifies the functional relations described below .
now we take @xmath173 as the basis of the solutions of the linear system .
we introduce a set of functions @xmath174 ( @xmath175 , @xmath176 ) by @xmath177},\\ { \cal t } _ { 2,m}(\lambda)&=&\langle s_{-2},s_{-1},s_{1},s_{m+1}\rangle^{[-m]},\\ { \cal t}_{3,m}(\lambda)&=&\langle s_{-2},s_{0},s_{1},s_{m+1}\rangle^{[-m]},\end{aligned}\ ] ] where @xmath178}(\lambda)\equiv f(\lambda+{m\over2}{2\pi i\over 4m})$ ] .
a solution @xmath161 ( @xmath86 ) is expanded in terms of this basis as @xmath179}_{1,k-2}s_{-2}+{\cal t}^{[k-1]}_{3,k-1}s_{-1}-{\cal t}^{[k-1]}_{2,k-1}s_0 + { \cal t}^{[k-1]}_{1,k-1}s_1.\end{aligned}\ ] ] the coefficients of @xmath180 , @xmath181 and @xmath182 follow from the definition of @xmath183 directly .
the coefficient of @xmath184 is evaluated as @xmath185 . using the the identity : @xmath186 } & = \langle s_{i_1 + 1 } , s_{i_2 + 1 } , s_{i_3 + 1 } , s_{i_4 + 1}\rangle,\end{aligned}\ ] ] which follows from the symanzik rotation
, it is shown to be equal to @xmath187}=-{\cal t}^{[k]}_{1,k-2}$ ] .
we expand @xmath188 in terms of the basis @xmath171 : @xmath189 where @xmath190 .
the exterior product @xmath191 in @xmath192 is also expanded in the basis @xmath193 .
the coefficient of the highest weight vector is evaluated as @xmath194 where we introduce the determinant @xmath195 for @xmath196 we have @xmath197}$ ] . for @xmath198
, we obtain @xmath199 .
in particular , from the normalization condition ( [ eq : normalization1 ] ) we find that @xmath200 the relation ( [ eq : qw1 ] ) can be regarded as the quantum wronskian relation @xcite .
let us consider two more examples .
for @xmath201 with @xmath202 and @xmath203 , using the @xmath1-system ( [ eq : psi - system2 ] ) , we find @xmath204 for @xmath205 , using ( [ eq : barsk2 ] ) , we have @xmath206 which becomes @xmath207 by the formula ( [ eq : charge1 ] ) .
we then get @xmath208 we note that the determinants ( [ eq : minor1 ] ) satisfy the plcker relations @xmath209 in particular one finds @xmath210 from these equations , we can solve @xmath211 as @xmath212 this equation is the t - q relation of the @xmath213-type quantum integrable models @xcite . now using ( [ eq : w201 ] ) and ( [ eq : w3m101 ] ) , ( [ eq : t - q1 ] )
becomes @xmath214}q_{1}^{(2)}{q}_{1}^{(1)[-1]}q_{1}^{(2)[-2 ] } & = q_{1}^{(2)[2]}{q}_{1}^{(1)[-1]}q_{1}^{(2)[-2 ] } + q_{1}^{(2)[-4]}{q}_{1}^{(1)[1]}q_{1}^{(2)[-2 ] } + { q}_{1}^{(1)[-3]}q_{1}^{(2)}q_{1}^{(2)}\nonumber\\ & + q_{1}^{(2)[-4]}q_{1}^{(2)}{q}_{1}^{(1)[-1]}. \label{eq : t - qrel1}\end{aligned}\ ] ] this is the t - q relation for the @xmath2-type . from this relation
we obtain the bethe equations , which was also derived in the previous section by using the @xmath1-system .
one can also derive a set of the relations : @xmath215 from these equations @xmath216 is solved as @xmath217 from ( [ eq : w201 ] ) , ( [ eq : w3m101 ] ) and @xmath218}$ ] , we get the t - q relation for @xmath219 : @xmath220}&=&\frac{q_{1 } ^{(1 ) [ -1]}}{q_{1}^{(1 ) [ -3]}}+\frac{q_{1}^{(2 ) [ -2]}}{q_{1}^{(2 ) [ -4]}}(\frac{q_{1}^{(2 ) [ -2]}q_{1}^{(1 ) [ -5]}}{q_{1}^{(2 ) [ -4]}q_{1}^{(1 ) [ -3]}}+\frac{q_{1}^{(2 ) [ -6]}}{q_{1}^{(2 ) [ -4]}})\nonumber\\&&+\frac{q_{1}^{(2 ) [ -6]}}{q_{1}^{(2 ) [ -4]}}(\frac{q_{1}^{(2 ) [ -2]}}{q_{1}^{(2 ) [ -4]}}+\frac{q_{1}^{(2 ) [ -6]}q_{1}^{(1 ) [ -3]}}{q_{1}^{(1 ) [ -5]}q_{1}^{(2 ) [ -4]}})+\frac{q_{1}^{(1 ) [ -7]}}{q_{1}^{(1 ) [ -5]}}.\end{aligned}\ ] ] from this equation , one obtains the bethe equations ( [ eq : bae1 ] ) again .
now we study the functional relations which are satisfied by @xmath183 .
first we calculate the product of @xmath221 and @xmath222 . from the plcker relation
@xmath223 we get the identities @xmath224}{\cal t}_{1,m-1}^{[m+1]}&=&{\cal t}_{1,m}^{[m]}+{\cal t}_{2,m-1}^{[m+1]}.\nonumber\\ { \cal t}_{2,1}^{[+1]}{\cal t}_{1,m-1}^{[m+1]}&=&{\cal t}_{2,m}^{[m]}+{\cal t}_{3,m-1}^{[m+1]},\nonumber\\ { \cal t}_{3,1}^{[+1]}{\cal t}_{1,m-1}^{[m+1 ] } & = & { \cal t}_{3,m}^{[m]}+{\cal t}_{1,m-2}^{[m+2 ] } , \label{eq : t - t_a_3 - -1}\end{aligned}\ ] ] these relations are a generalization of the fusion relation of the modified sinh - gordon equation @xcite to the modified @xmath0 affine toda field equation .
however , one finds @xmath225}{\cal t}_{1,m}^{[-1]}&=&{\cal t}_{1,m+1}{\cal t}_{1,m-1}+\langle s_{-1},s_{0},s_{m+1},s_{m+2}\rangle^{[-m-1]},\end{aligned}\ ] ] where the second term in the r.h.s .
is not the form of the @xmath183 functions .
we add this function to a member of the t - functions and define @xmath226},\\ t_{2,m}(\lambda)&=&\langle s_{-1},s_{0},s_{m+1},s_{m+2}\rangle^{[-m-1]},\label{eq : t - funct1}\end{aligned}\ ] ] for @xmath176 .
the new function @xmath227 satisfies the identity @xmath228}t_{2,m}^{[-1 ] } & = t_{2,m-1}t_{2,m+1}+\langle s_{-1 } , s_m , s_{m+1 } , s_{m+1}\rangle^{[-m ] } t_{1,m}.\end{aligned}\ ] ] we then introduce @xmath229}.\end{aligned}\ ] ] but this is not new . using ( [ eq : charge1 ] ) and ( [ eq : barsk2 ] ) , we can show that @xmath230 .
finally we obtain the t - system of @xmath2 type : @xmath231}t_{1,m}^{[-1]}&=t_{1,m-1}t_{1,m+1}+t_{2,m}\nonumber \\ t_{2,m}^{[+1]}t_{2,m}^{[-1]}&=t_{2,m+1}t_{2,m-1}+t_{1,m}t_{1,m } , \label{eq : t - system1}\end{aligned}\ ] ] which is obtained by the reduction of @xmath213 t - system with the identification @xmath232 .
other functions @xmath233 , @xmath234 can be expressed in terms of @xmath235 by using @xmath236}$ ] , @xmath237}$ ] and ( [ eq : t - t_a_31 ] ) .
they also satisfy the identities : @xmath238}{\cal t}_{1,m}^{[2]}-t_{2,m } , \nonumber\\ { \cal t}_{2,m+1}^{[+1]}{\cal t}_{1,m}&= { \cal t}_{1,m+1}^{[+1]}{\cal t}_{2,m}+t_{2,m+1}. \label{eq : t - rec1}\end{aligned}\ ] ] we next introduce the y - functions by @xmath239 they satisfy the y - system of @xmath2 type @xmath240}_{a , m}y^{[+1]}_{4-a , m}}{y_{a+1,m}y_{a-1,m}}=\frac{(1+y_{a , m+1})(1+y_{4-a , m-1})}{(1+y_{a+1,m})(1+y_{a-1,m } ) } , \label{eq : y - system2}\end{aligned}\ ] ] where @xmath241 and @xmath242 .
the t - system ( [ eq : t - system1 ] ) and the y - system ( [ eq : y - system2 ] ) imply that the langlands duality between the modified @xmath0 affine toda equation and the functional equations of the @xmath2 quantum integrable system .
we now discuss the boundary condition of the t - system and the y - system .
it is easy to see that @xmath243 and @xmath244 . in order to determine the boundary conditions @xmath245 for large @xmath32
, we need to study the small solutions @xmath246 in the whole complex plane . when @xmath247 is not a rational number , the stokes sectors cover the complex plane infinitely many times .
so the t - functions @xmath235 are defined independently for arbitrary positive integer @xmath32 . in this paper
we will consider the case @xmath248 with @xmath249 being a positive integer in detail is a rational number , we can do similar arguments .
but it is not discussed in this paper . ] . in this case
there are @xmath250 stokes sectors in the complex plane .
when we go around the origin , the solution @xmath251 is defined in the sector @xmath252 , which is the same as @xmath157 .
then the small solution @xmath253 is proportional to @xmath251 : @xmath254 for @xmath255 , the linear system has no simple pole at the origin .
the solution has no monodromy around it .
then we have @xmath256 which implies @xmath257 the condition ( [ eq : monodromy0 ] ) leads to the boundary conditions for the t-/y - functions : @xmath258 and @xmath259 .
the truncated t-/y - system becomes the same as the one for the @xmath250-point gluon scattering amplitudes in ads@xmath260 at strong coupling @xcite . for @xmath261 , the solutions of the linear system have monodromy around the origin .
we introduce a monodromy matrix @xmath262 by @xmath263 from the normalization condition ( [ eq : nor1 ] ) we find @xmath264 .
we also introduce the proportionality factor @xmath265 in ( [ eq : monodromy0 ] ) for @xmath266 by @xmath267 let us expand the solution @xmath268 in the basis @xmath269 whose coefficient has been defined as @xmath270
. then we substitute its symanzik rotation into ( [ eq : monodromy3 ] ) . in the basis @xmath271
, the monodromy matrix becomes diagonal and takes the form @xmath272 .
moreover from the quasi - periodicity condition ( [ eq : period1 ] ) one finds that @xmath273 . plugging ( [ eq : monodromy3 ] ) into ( [ eq : monodromy2 ] )
, we get the relation @xmath274 which generalizes the condition ( [ eq : monodromy0 ] ) and determines the boundary condition for the t - system .
it is convenient to use the ( multi-)trace of the monodromy matrix @xmath275 : @xmath276 and @xmath277 , which are basis independent quantities .
these traces can be also expressed using the wronskians : @xmath278 here the r.h.s . of these equations
are expressed by @xmath233 , @xmath234 and the auxiliary t - functions @xmath279 , @xmath280 , @xmath281 defined in @xcite , in addition to the t - functions ( [ eq : t - funct1 ] ) . in appendix
a , we will summarize these auxiliary t - functions and their recursion relations . in the diagonal basis , they are evaluated as @xmath282\cos[\beta(g_1+g_2)\pi]\end{aligned}\ ] ] where @xmath283 ( @xmath16 ) . for @xmath255 ,
one finds that @xmath284 and @xmath285 .
the monodromy conditions ( [ eq : monodromy4 ] ) determine @xmath286 ( @xmath62 ) .
then the t - system extends up to @xmath287 and the t - functions @xmath235 for @xmath288 are determined by the t - system and the monodromy conditions . concerning the y - system ( [ eq : y - system2 ] )
, it also extends up to @xmath289 .
it is convenient to introduce new y - functions @xmath290 ( @xmath291 ) by @xmath292 whose functional relations are given by @xmath293 } \bar{y}_{4-a}^{[-1]}\over \bar{y}_{a+1}\bar{y}_{a-1 } } & = { 1+y_{a , n-2}\over ( 1+y_{a+1,n-1})(1+y_{a-1,n-1})}. \label{eq : y - system3}\end{aligned}\ ] ] the y - system ( [ eq : y - system2 ] ) for @xmath289 and ( [ eq : y - system3 ] ) contains @xmath294 in the r.h.s .
of the equations .
@xmath294 are expressed as @xmath295 and @xmath286 are expressed in terms of the lower t - functions . for the @xmath296 ( @xmath297 ) case ,
they are also expressed in terms of the lower y - functions by solving ( [ eq : y - system1 ] ) .
then ( [ eq : y - system2 ] ) and ( [ eq : y - system3 ] ) with ( [ eq : y - system4 ] ) become the closed functional relations .
note that the present t- and y - systems are the same as those of form factors in ads@xmath298 @xcite .
however the function @xmath34 has different pole structure from the present one . in the case of even @xmath250 and @xmath299 ( or @xmath300 )
, one can consider the limit to the modified sinh - gordon equation @xcite ( or gluon scattering amplitudes in ads@xmath301 @xcite ) , where in this limit the @xmath138 spinor is decomposed into left and right - handed spinors . in this reduction
the t - functions @xmath235 reduce to the functions @xmath302 ( @xmath303 ) , which are defined by the inner product of the left - handed spinors .
they satisfy @xmath304},\quad t_{2,2k}=t_k^{[3]}t_k^{[+1]},\quad t_{2,2k+1}=-t_k^{[2]}t_{k+1}^{[2 ] } , \nonumber\\ & \langle s_{-2},s_0,s_1,s_{n-1}\rangle = t_{\frac{n}{2}-2}^{[n+2]},~~\langle s_{-2},s_{-1},s_1,s_{n}\rangle = t_{\frac{n}{2}}^{[n]}.\end{aligned}\ ] ] here @xmath305 obey the functional relations @xmath306}t_{k}^{[-2]}&=1+t_{k-1}t_{k+1}.\end{aligned}\ ] ] using ( [ eq : t - reduce ] ) , we can rewrite @xmath307 in terms of the left - handed part and right - handed part . decomposing these two parts we obtain @xmath308 , which is the trace of monodromy in left - handed part .
the y - functions @xmath309 and @xmath310 reduce to @xmath311 ( @xmath312 ) and @xmath313 as @xmath314 } , \quad \bar{y}_2=\infty .
\label{eq : ads3-y1}\end{aligned}\ ] ] here the y - functions @xmath315 and @xmath316 satisfy the @xmath317-type y - system @xcite @xmath318}y_k^{[-2]}&=(1+y_{k-1})(1+y_{k+2 } ) , \quad ( k=1,\ldots , { n\over2}-3 ) , \nonumber\\ \bar{y}^{[2]}\bar{y}^{[-2]}&=1+y_{{n\over2}-2 } , \nonumber\\ y_{{n\over2}-2}^{[2]}y_{{n\over2}-2}^{[-2]}&=
( 1+y_{{n\over2}-3})(1 - 2 \cos\pi \beta g_2 \bar{y}+\bar{y}^2 ) .\end{aligned}\ ] ]
in the previous section we have seen that the t-/y - system becomes the extended one in the presence of monodromy .
the standard approach to analyze the ( extended ) y - system is to derive the thermodynamic bethe ansatz ( tba ) equations and investigate their free energy .
the ir ( or low - temperature ) limit of the tba system are characterized by the wkb approximation , whereas in the uv ( or high - temperature ) limit is characterized by the spectral parameter independent y - functions and the free energy is determined by the dilog formulas @xcite . since the present y - system is very complicated , we leave the detailed tba analysis to the subsequent paper .
instead we will study the high - temperature limit of the y - system and their solutions explicitly for the simplest case @xmath319 . for @xmath319 .
using the t - system , @xmath235 ( @xmath62 , @xmath320 ) are solved in terms of @xmath321 and @xmath322 using ( [ eq : t - system1 ] ) . substituting them into the monodromy conditions we obtain two equations for @xmath323 and @xmath324 @xmath325 \cos[\beta g_2 \pi ] = 0 , \nonumber\\ & -2 x^6 + y^2 ( -3 + y^2)^2 + 3 x^4 ( 8 - 8 y + 3 y^2 ) - 6 x^2 ( 3 - 2 y^3 + y^4)\nonumber\\ & - 4 \cos[\beta ( g_1 - g_2 ) \pi ] \cos[\beta ( g_1 + g_2 ) \pi ] = 0 .
\label{eq : monodromy5}\end{aligned}\ ] ] for @xmath326 , _
i.e. _ when there is no monodromy around the origin , we find the solutions of the above algebraic equations are given by @xmath327 and @xmath328 . for @xmath329 , we get @xmath330 , @xmath331 , @xmath332 ( @xmath333 ) , @xmath334 and @xmath335 .
this solution corresponds to the @xmath336 limit of the y - system . for @xmath337
, we get @xmath338 , @xmath339 and @xmath340 , which corresponds to the constant y - system of the 6-point amplitudes @xcite . for other solutions
, we do not find any corresponding physical quantities .
now we turn on @xmath341 with keeping @xmath299 .
we find the solutions of ( [ eq : monodromy5 ] ) with @xmath342 are given by @xmath343 this gives @xmath344 which turns out to be a constant solution .
this corresponds to a constant solution of the @xmath345 form factor with @xmath319 gluons@xcite . note that @xmath346}=\bar{y}_1y_{2,4}$ ] in ( [ eq : ads3-y1 ] ) for @xmath319 . as for the solutions starting from @xmath347
, we solve eqs .
( [ eq : monodromy5 ] ) numerically . the graphs of @xmath348 and @xmath349 are shown in fig .
[ fig:1 ] and the corresponding y - functions @xmath350 and @xmath351 are shown in fig .
[ fig:2 ] .
there arise four branches from the point @xmath352 , which arrive at the solutions @xmath353 , @xmath354 , @xmath355 and @xmath356 at @xmath357 .
these solutions provide a deformation of the constant t-/y - functions by the monodromy parameter @xmath341 .
in this paper , we studied the massive ode / im correspondence for modified @xmath0 affine toda field equation . by investigating the solutions of the linear problem associated with the modified affine toda equation
, we derived the @xmath1-system .
this leads to the bethe ansatz equations corresponding to the integrable model which is not identified yet .
we also derived the same bethe ansatz equations from the t - q relations .
we constructed the t - system and y - system from the wronskians of the solutions of the linear problem .
these systems have non - trivial boundary conditions due to the presence of monodromy around the origin .
it would be interesting to generalize the present approach to modified affine toda field equations associated with other affine lie algebras which are not of the langlands dual of an untwisted affine lie algebra@xcite .
it is also interesting to study the massless limit of this linear problem and investigate the description by using the free field realization of conformal field theory @xcite . for the linear system
associated the null - polygonal minimal surface in ads@xmath260 , we have seen that the corresponding integrable system is the homogeneous sine - gordon model @xcite .
when the solution has monodromy around the origin , we have seen the t - system and y - system are extended and they take the form that appears in the strong - coupling limit of the form factor in @xmath358 super yang - mills theory . for a general polynomial @xmath34 and the appropriate boundary conditions for the solutions of the linear problem , one can describe the minimal surface problem using the massive ode / im correspondence . in particular
it is interesting to explore the ode / im correspondence for the minimal surface in @xmath359 , where the corresponding quantum integrable model is not known yet .
we would like to thank yuji satoh and christopher locke for valuable discussion and collaboration in an early stage of this work .
this work is supported in part by grant - in - aid for scientific research 15k05043 from japan society for the promotion of science ( jsps ) .
in this appendix we summarize the auxiliary t - functions and their recursion relations @xcite . from these relations
we express @xmath276 and @xmath360 in terms of @xmath235 ( @xmath361 ) .
furthermore we can express @xmath294 in the lower y - functions and get a closed y - system .
we define the functions @xmath362 , @xmath363 , @xmath279 , @xmath280 @xmath281 ( @xmath176 ) by @xmath364},\label{eq : u-1}\quad v_{1,m}=\langle s_{-2 } , s_0 , s_{m+1 } , s_{m+2}\rangle^{[-m]},\\ w_{1,m}&=\langle s_{-2 } , s_0 , s_{m+1 } , s_{m+3}\rangle^{[-m-1]},\\ w_{2,m}&=\langle s_{-1 } , s_0 , s_{m+1 } , s_{m+4}\rangle^{[-m-2]},\quad \bar{w}_{2,m}=\langle s_{-2 } , s_1 , s_{m+2 } , s_{m+3}\rangle^{[-m-2]}.\label{eq : w-1}\end{aligned}\ ] ] from the plcker relation ( [ eq : plucker2 ] ) , we can show that these auxiliary t - functions satisfy @xmath365}_{1,m-1}t_{2,m+1}+t^{[+1]}_{1,m+1}t^{[-1]}_{2,m } , \\
v_{1,m}t_{3,m}&=t^{[+1]}_{1,m-1}t_{2,m+1}+t^{[-1]}_{3,m+1}t^{[+1]}_{2,m } , \\
w_{1,m}t_{2,m}&=v^{[-1]}_{1,m}u^{[+1]}_{1,m}-t^{[-1]}_{1,m}t^{[+1]}_{3,m } , \\ w_{2,m}t_{2,m+1}&= u^{[+1]}_{1,m+1}u_{1,m}-t^{[-1]}_{2,m}t^{[+1]}_{2,m+2 } , \\
\bar{w}_{2,m}t_{2,m+1}&= v^{[-1]}_{1,m+1}v_{1,m}-t^{[+1]}_{2,m}t^{[-1]}_{2,m+2}.\end{aligned}\ ] ] then @xmath362 , @xmath363 , @xmath279 , @xmath280 and @xmath281 are expressed in terms of @xmath366 .
the ( symmetrized ) trace of the monodromy matrix becomes @xmath367}-{\cal t}_{2,n-1}^{[n-1 ] } + { \cal t}_{3,n-2}^{[n-2]}-t_{1,n-4}^{[n-2]}\right),\label{eq : tr-1 } \\
{ \rm tr}^{(2)}\omega & = t_{2,n}^{[n-1]}+t_{2,n-4}^{[n-1 ] } -w_{1,n-2}^{[n-1]}+w_{2,n-3}^{[n-1 ] } + \bar{w}_{2,n-3}^{[n-1]}-w_{1,n-4}^{[n-1]}.\label{eq : tr-2}\end{aligned}\ ] ] using @xmath368 } , { \cal t}_{2,1}=t_{2,1}^{[-1]}$ ] and the identities ( [ eq : t - t_a_31 ] ) , we get @xmath369}t_{1,n-3}^{[+1]}-t_{1,n-4}^{[+2]}\label{eq : t3 - 1}\\ { \cal t}_{2,n-1}&= & t_{2,1}^{[-n+1]}t_{1,n-2}^{[+1]}-\{t_{1,1}^{[-n+1]}t_{1,n-3}^{[+1]}-t_{1,n-4}^{[+2]}\}.\label{eq : t2 - 1}\end{aligned}\ ] ] the ( symmetrized ) traces are expressed in terms of @xmath366 as @xmath370}-t_{1,n-4}^{[n-2]}-\left(t_{2,1}^{[-n+1]}t_{1,n-2}^{[+1]}-(t_{1,1}t_{1,n-3}^{[n]}-t_{1,n-4}^{[n+1]})\right)+(t_{1,1}^{[-1]}t_{1,n-3}^{[n-1]}-t_{1,n-4}^{[n]})\right ) , \nonumber\\ { \rm tr}^{(2)}\omega&=&t_{2,n}^{[n-1]}+t_{2,n-4}^{[n-1]}\nonumber\\&&-\left(-\frac{t_{3,n-2}^{[n]}t_{1,n-2}^{[n-2]}}{t_{2,n-2}^{[n-1]}}+\frac{1}{t_{2,n-2}^{[n-1]}}\frac{t_{1,n-1}^{[n+1]}t_{2,n-2}^{[n-1]}+t_{2,n-1}^{[n]}t_{1,n-3}^{[n-1]}}{t_{3,n-2}^{[n]}}\frac{t_{2,n-2}^{[n-1]}t_{1,n-1}^{[n-3]}+t_{1,n-3}^{[n-1]}t_{2,n-1}^{[n-2]}}{t_{3,n-2}^{[n-2]}}\right)\nonumber\\&&+\left(-\frac{t_{2,n-1}^{[n]}t_{2,n-3}^{[n-2]}}{t_{2,n-2}^{[n-1]}}+\frac{1}{t_{2,n-2}^{[n-1]}}\frac{t_{1,n-3}^{[n-1]}t_{2,n-1}^{[n]}+t_{1,n-1}^{[n+1]}t_{2,n-2}^{[n-1]}}{t_{1,n-2}^{[n]}}\frac{t_{1,n-4}^{[n-2]}t_{2,n-2}^{[n-1]}+t_{1,n-2}^{[n]}t_{2,n-3}^{[n-2]}}{t_{1,n-3}^{[n-1]}}\right)\nonumber\\&&+\left(-\frac{t_{2,n-3}^{[n]}t_{2,n-1}^{[n-2]}}{t_{2,n-2}^{[n-1]}}+\frac{1}{t_{2,n-2}^{[n-1]}}\frac{t_{2,n-3}^{[n]}t_{1,n-2}^{[n-2]}+t_{1,n-4}^{[n]}t_{2,n-2}^{[n-1]}}{t_{3,n-3}^{[n-1]}}\frac{t_{2,n-2}^{[n-1]}t_{1,n-1}^{[n-3]}+t_{1,n-3}^{[n-1]}t_{2,n-1}^{[n-2]}}{t_{3,n-2}^{[n-2]}}\right)\nonumber\\ & & -\left(-\frac{t_{3,n-4}^{[n]}t_{1,n-4}^{[n-2]}}{t_{2,n-4}^{[n-1]}}+\frac{1}{t_{2,n-4}^{[n-1]}}\frac{t_{1,n-3}^{[n+1]}t_{2,n-4}^{[n-1]}+t_{2,n-3}^{[n]}t_{1,n-5}^{[n-1]}}{t_{3,n-4}^{[n]}}\frac{t_{2,n-4}^{[n-1]}t_{1,n-3}^{[n-3]}+t_{1,n-5}^{[n-1]}t_{2,n-3}^{[n-2]}}{t_{3,n-4}^{[n-2]}}\right).\nonumber\\\end{aligned}\ ] ] from these equations we can write @xmath371 and @xmath372 in terms of lower t - functions . in the case of @xmath296 with @xmath373 , the t - functions can be further expressed in terms of the y - functions by solving ( [ eq : y - system1 ] ) .
we then obtain a closed y - system . b. a. burrington , jhep * 1109 * , 002 ( 2011 ) [ arxiv:1105.3227 [ hep - th ] ] .
l. f. alday , j. maldacena , a. sever and p. vieira , j. phys .
a * 43 * , 485401 ( 2010 ) [ arxiv:1002.2459 [ hep - th ] ] .
y. hatsuda , k. ito , k. sakai and y. satoh , jhep * 1004 * , 108 ( 2010 ) [ arxiv:1002.2941 [ hep - th ] ] .
y. hatsuda , k. ito , k. sakai and y. satoh , jhep * 1104 * , 100 ( 2011 ) [ arxiv:1102.2477 [ hep - th ] ] .
j. maldacena and a. zhiboedov , jhep * 1011 * , 104 ( 2010 ) [ arxiv:1009.1139 [ hep - th ] ] .
a. kuniba , t. nakanishi and j. suzuki , int .
j. mod .
phys . a * 9 * , 5215 ( 1994 ) [ hep - th/9309137 ] .
a. kuniba , t. nakanishi and j. suzuki , int .
j. mod .
phys . a * 9 * , 5267 ( 1994 ) [ hep - th/9310060 ] . c. locke ,
`` ode / im correspondence and affine toda field equations '' , doctral thesis , tokyo institute of technology , 2015 v. v. bazhanov , s. l. lukyanov and a. b. zamolodchikov , commun . math .
phys . * 177 * , 381 ( 1996 ) [ hep - th/9412229 ] .
v. v. bazhanov , s. l. lukyanov and a. b. zamolodchikov , ion , commun .
phys . * 190 * , 247 ( 1997 ) [ hep - th/9604044 ] . | we study the massive ode / im correspondence for modified @xmath0 affine toda field equation . based on the @xmath1-system for the solutions of the associated linear problem
, we obtain the bethe ansatz equations .
we also discuss the t - q relations , the t - system and the y - system , which are shown to be related to those of the @xmath2 integrable system .
we consider the case that the solution of the linear problem has a monodromy around the origin , which imposes nontrivial boundary conditions for the t-/y - system .
the high - temperature limit of the t- and y - system and their monodromy dependence are studied numerically .
tit / hep-654
+ may , 2016 * ode / im correspondence for modified @xmath0 affine toda field equation * .75em 2.5 cm katsushi ito and hongfei shu 2.5em _ department of physics , + tokyo institute of technology + tokyo , 152 - 8551 , japan _ 3.0em = 0.7 cm |
it is well known that a random walker making short - range hops , namely a brownian particle in a thermal environment , has an average incremental displacement which goes as square root of the time difference .
although this property is rather robust , deviation from this diffusive behaviour has been observed in a wide variety of physical situations @xcite .
super - diffusion is well known in the context of turbulence and l@xmath8vy walks @xcite . yet another way to see anomalous diffusion is to subject the random walker to a quenched velocity field .
a concrete example of this is the matheron - de marsily ( mdm ) model , which was introduced to study the hydrodynamic dispersion of solute particles in sedimentary layered porous rock formation @xcite . in the mdm model
a single particle diffuses in a @xmath9-dimensional layered medium , such that the disorder is anisotropic .
the motion of the particle is purely brownian along the transverse @xmath10 direction , while along the longitudinal @xmath1 direction , added to the thermal noise , the particle is driven by a quenched drift velocity @xmath11 that is a random function of only the transverse coordinate @xmath10 .
the coupling between diffusion and convection due to the spatially random but temporally static velocity field , generates a typical bias in the @xmath1 direction giving rise to a super - diffusive longitudinal transport for large @xmath6 @xcite .
the temporal behaviour is quantitatively described by the standard two - time dependent correlation function , and the persistence probability .
intuitively , persistence of a system concerns the property of the system to stay in some given state .
the persistence probability @xmath12 which is related to the first - passage probability @xmath13 , as @xmath14 is simply the probability that the particle does not cross a given point up to time @xmath6 . in systems with super - diffusion
, @xmath12 is expected to get enhanced .
persistence properties in disordered systems are harder to find , and thus results known for such models like the sinai model @xcite and the mdm model @xcite are of great importance .
recently for a @xmath15 dimensional version of mdm model the persistence probability was analytically derived in @xcite . in earlier studies of the mdm model @xcite ,
only short range correlated @xmath11 had been considered . however , since originally the model was motivated @xcite by general flow in certain types of fractured rocks with parallel fractures which allow the propagation of dissolved species , it is possible that the flows parallel to the bedding might have velocity correlations which are spread over long distances .
it may be mentioned that different aspects of dynamics in mdm fields with algebraic correlations , have been studied @xcite .
generally long range correlated disordered problems are hard to solve . for isotropic disorder with power law correlations , various results
are known @xcite .
the mdm model being anisotropic allows for exact anatytic treatment as compared to field theoretic treatment done in other cases . in this paper
we have been able to analytically derive the incremental correlation function and persistence probability , for the mdm model with long range _ power law _ correlated @xmath11 .
we find that long range correlation incorporates new temporal behaviour , as compared to the short range mdm model .
the relevant exponent of @xmath12 becomes dependent on the power law exponent @xmath7 and the transverse dimension @xmath0 .
the method that we follow is similar to @xcite and was used even earlier in @xcite .
we first show that @xmath5 is a fractional brownian motion ( fbm ) , and then use its known first passage property to solve for the persistence probability .
a stochastic process @xmath5 ( with zero mean @xmath16 ) is called a fbm if its incremental two - time correlation function @xmath17 is stationary , i.e. , depends only on the difference @xmath18 and moreover , grows asymptotically as a power law , @xmath19 the parameter @xmath20 is called the hurst exponent that characterizes the fbm@xcite and @xmath21 denotes the expectation value over all realizations of the process .
for example , the ordinary brownian motion is a fbm with @xmath22 .
the zero crossing properties of fbm has been studied before@xcite and it is known , both analytically and numerically , that the probability that a fbm does not cross zero up to time @xmath6 has a power law decay , @xmath23 for large @xmath6 with @xmath24@xcite .
we use this property to derive @xmath25 from @xmath20 .
we consider a generalized version of the mdm model in @xmath26 transverse plus @xmath27 longitudinal spatial dimensions .
the transverse coordinates of the particle , denoted by @xmath28 s with @xmath29 , undergo ordinary brownian motion , @xmath30 where @xmath31 s are zero mean gaussian white noises with correlators , @xmath32 .
equivalently we can use a single equation involving @xmath0-dimensional vectors @xmath33 and @xmath34 : @xmath35 the longitudinal coordinate @xmath5 , in contrast , is driven by a random drift @xmath36 $ ] that depends only on the transverse coordinates @xmath37 , @xmath38 + \eta_x(t ) , \label{xc}\ ] ] where @xmath39 is again a zero mean gaussian white noise with @xmath40 .
the velocity field @xmath41 $ ] is quenched , i.e. , constant in time .
operationally this means that for a given realization of the function @xmath42 , one has to evolve @xmath5 over ` thermal histories ' in eq .
( [ xc ] ) and then one needs to ` disorder average ' ( denoted by @xmath43 ) over different realizations of the random function @xmath42 . in the original mdm model and in the latter works on slightly modified versions of it @xcite , the random function
@xmath44 has been considered to be a gaussian with zero mean ( @xmath45 ) , with a correlator @xmath46 , which is either @xmath47 or is short ranged namely @xmath48^{d/2 } } } \exp\left[- { { \left(\vec y_1-\vec y_2\right)^2}/{2 d a^2}}\right]$ ] ( where @xmath49 is a small cutoff ) .
in contrast to the above , here we consider the random function @xmath42 to have long - range power - law correlation : & = & 0 + & = & 1 .
[ vv ] the exponent @xmath50 denotes the strength of the correlation function , and the short distance cutoff scale ` @xmath49 ' ensures that there is no divergence as @xmath51 .
this model is analytically treated in section [ ana ] and numerically in section [ num ] .
we will see that the role played by the scale @xmath49 is very interesting . for @xmath52 ,
it is irrelevant and can be set to zero , but for @xmath53 , it has to be finite to ensure a well - behaved asymptotic behaviour .
our main result is that the thermal and disorder averaged two - time correlation function of the @xmath1-coordinate , @xmath54 .
thus the process @xmath5 is an fbm .
the hurst exponent @xmath55 $ ] .
this means that for @xmath56 , the exponent will change from @xmath57 for @xmath52 , to @xmath58 for @xmath53 . on the other hand , for @xmath59 , @xmath60 for @xmath61 , and @xmath62 for @xmath63 .
it follows from this that the persistence probability of @xmath64 , with @xmath65 $ ] .
we have also found the lines in @xmath0 and @xmath7 plane along which there is marginal behaviour .
the analytical treatment that we present here is similar to that of @xcite up to the point that the explicit power law form of velocity correlation is substituted . nevertheless , for completeness
, we give the steps briefly . in order to solve for correlation functions of @xmath5 from eq .
[ xc ] , we need knowledge of correlation functions of @xmath66 . by integrating eq .
[ yc1 ] we get @xmath67 from the above linear relation we conclude that since @xmath68 is a gaussian random variable , so is @xmath2 . moreover another vector @xmath69 which will be useful in the following calculations is also a gaussian random variable by virtue of the linear relation . using eq .
[ y1 ] , @xmath70 and it follows that @xmath71 and variance @xmath72 , since @xmath32 .
thus explicitly the distribution @xmath73^{-d/2}\exp\left[{-{\vec y}^2/{2d|\tau_1-\tau_2|}}\right ] .
\label{gauss}\ ] ] with the above solution of the transverse coordinates @xmath66 , we now proceed to treat the longitudinal coordinate @xmath5 .
we note that the averages so far over the transverse coordinates , namely @xmath74 were thermal averages .
the coordinate @xmath5 depends on both a thermal noise @xmath39 and a quenched random noise @xmath42 .
so we need to do a thermal as well as quenched disorder average , which we will denote by @xmath75 , where in particular the @xmath76 denotes the quenched disorder average . integrating eq .
( [ xc ] ) we get , @xmath77d\tau .
\label{xc1}\ ] ] from eq .
( [ xc1 ] ) we get @xmath78 , while using @xmath40 , we get @xmath79 where the integral @xmath80 is given by @xmath81v[{\vec y}(\tau_2 ) ] } } \rangle d\tau_1d\tau_2 \nonumber \\ & = & \int_0^{t_1}\int_0^{t_2 } \left\langle { 1 \over { ( a + |\vec y|)^{\alpha } } } \right\rangle d\tau_1 d\tau_2 . \label{idef}\end{aligned}\ ] ] in the above , we have used the correlator of @xmath42 from eq .
[ vv ] and @xmath82 is the same as in eq .
[ y1 ] . since we know the probability distribution @xmath83 explicitly from eq .
[ gauss ] , we can proceed from eq .
[ idef ] to evaluate i(t_1,t_2 ) & = & _ 0^t_1 _ 0^t_2 _ -^ d_1 d_2 1 ( |y| + a)^ e^-|y|^2 d^dy + & = & 2 ( d/2 ) _ 0^t_1 _ 0^t_2 _ 0^ d_1 d_2 e^-z^2 z^d-1 dz + & = & 2 ( d/2 ) _ 0^ e^-z^2 z^d-1 j(z , t_1,t_2 ) dz . [ it1t2 ] in eq . [ it1t2 ] we have substituted @xmath84 in going from the first line to the second , and then performed the two time integrals exactly to get explicitly the function j(z , t_1,t_2 ) & = & _ 0^t_1 _ 0^t_2 d_1 d_2 + [ jdef ] & = & 12 a^2- min(t_1,t_2 ) + ( a+z)^4- + ( a+z)^4- - a^4- - ( a+z)^4- + & + & ( 3 - 2 ) + & + & a^2(a+z)^2- + a^2(a+z)^2- - a^4- - a^2(a+z)^2- . [ jfull ]
although eq .
[ jfull ] has many terms , the two time - correlation function @xmath85 has a dependence only on @xmath86 , and has the following comparatively simpler expression : c(t ) = t + 4 ( d/2 ) _ 0^ dz e^-z^2 z^d-1 \ { a t z^2 + 1 z^4 & [ & 2b(a+z)^4- - 2 c a(a+z)^3- + & + & 2d a^2 ( a+z)^2- - ( 2b-2c+2d)a^4- ] } [ c ] in eq .
[ c ] we have used the symbols @xmath87 $ ] , @xmath88 $ ] , @xmath89 $ ] and @xmath90 $ ] .
further , eq . [ c ] can be written as @xmath91 , where @xmath92 is the integral which has to be analyzed to get the leading temporal behaviour . defining a new variable of integration @xmath93 and substituting in @xmath92 , we get |c = 4 t^2-d/2 _ 0^ dy^ e^-y^^2/2d t y^^d-1 \ { a 2d ^2 + 4d^2 ^4 & [ & 2b(a+y^)^4- - 2 c a(a+y^)^3- + & + & 2d a^2 ( a+y^)^2- - ( 2b-2c+2d)a^4- ] } [ cbar ] it turns out that one has to be very careful to extract the asymptotic behaviour of the above integral , because the correlations at scales @xmath94 and @xmath95 contribute different power laws , and they compete with each other .
it is best to split the integral @xmath96 in eq .
[ cbar ] into two parts @xmath97 and @xmath98 .
accordingly @xmath99 . in @xmath100
we expand the terms of the integrand in powers of @xmath101 ( with @xmath94 ) , while in @xmath102 we expand the terms in powers of @xmath103 ( with @xmath95 ) .
the @xmath100 integral , with its integrand expanded in a series yields : c_1 & = & 4 t^2-d/2 _ 0^a dy^ e^-y^^2/2d t y^^d-1 a^- [ 1 - 415 y^ a + o(y^^2 a^2 ) ] + & & b_1 a^d- t^2-d/2 [ c1 ] where @xmath104 $ ] . in obtaining eq .
[ c1 ] we have used the fact that for small @xmath105 and very large @xmath6 , we may approximate @xmath106 inside the integral . in eq .
[ c1 ] the factor @xmath107 smoothly goes to zero if @xmath108 for @xmath52 .
if we wish , by taking smaller and smaller cutoff @xmath49 , we can actually ignore the contribution of of @xmath100 with respect to @xmath102 . on the other hand ,
if @xmath53 , @xmath108 will lead to a divergence , and hence we have to keep @xmath49 finite , and thus the contribution of @xmath100 can not be ignored .
now we turn to the integral @xmath102 and it goes as follows : c_2 & = & 4 t^2-d/2 _
dy^ e^-y^^2/2d t y^^d-1 \{a 2d ^2 + 4d^2 ^ [ 2b + ( 2b ( 4 - ) - 2c ) a + o(a^2 ^2 ) - ( 2b - 2c + 2d ) a^4- ^4- ] } + & & a_0 t + a_1 t^2-/2 + o(t^3/2-/2 ) + b_2 a^d - t^2-d/2 [ c2 ] where the newly defined constants are @xmath109 $ ] , @xmath110 $ ] , @xmath111 - { { 8bd^2 } /(d-\alpha ) } + o(a ) + { b_4}\}$ ] , and @xmath112 $ ] . in going from the first line to the second line in eq .
[ c2 ] we have made the approximation of replacing the upper limit @xmath113 by @xmath114 , as the exponential factor @xmath115 sharply cuts off the integral beyond that scale anyway .
after that we have set @xmath106 and then done the integrals .
we see that in eq .
[ c2 ] , @xmath102 has three dominant power laws with powers equal to @xmath27 , @xmath116 , and @xmath117 . combining eqs .
[ c1 ] and [ c2 ] , we get finally the two - time incremental correlation function from eq .
[ c ] as c(t_1,t_2 ) = c(t ) b_1 t + a_1 t^2 - /2 + b_2 t^2 - d/2 , [ cfinal ] where @xmath118 , and @xmath119 .
this implies that asymptotically , at long time @xmath6 , @xmath120 .
thus @xmath5 is a fbm process and the associated hurst exponent @xmath20 is given by h = max[1/2 , ( 1 - /4 ) , ( 1 - d/4 ) ] [ hurst2 ] more explicitly , it means that for @xmath56 , for @xmath52 , @xmath121 while for @xmath53 , @xmath122 . on the other hand , for @xmath59 , for @xmath61 , @xmath123 while for @xmath63 , @xmath62 .
this scenario is shown in fig .
[ fig4 ] .
we have analysed the cases of marginal behaviour separately . for short - range disordered mdm model , @xmath124
was shown @xcite to have marginal behaviour .
for the current model , we have done explicit calculations ( similar as above ) for the cases @xmath125 , @xmath124 and @xmath126 @xcite .
@xmath127 for @xmath125 , for large @xmath6 , @xmath128 ( for @xmath129 ) , @xmath130 ( for @xmath56 ) , and @xmath131 ( for @xmath124 ) .
this means that the semi - infinite line @xmath125 with @xmath59 ( see fig .
[ fig4 ] ) has marginal behaviour .
@xmath132 for @xmath133 , @xmath134 ( for @xmath63 ) and @xmath135 ( for @xmath61 ) .
hence the semi - infinite line @xmath124 with @xmath63 ( fig .
[ fig4 ] ) exhibits marginality .
@xmath136 finally we also find logarithmic correction , namely @xmath137 , along the line @xmath126 , for @xmath7 and @xmath0 both @xmath138 . from the known first - passage property of fbm mentioned earlier , the disorder and thermal averaged probability that the process @xmath5 does not cross zero up to time @xmath6 decays as a power law , @xmath64 with @xmath24 . using the results for @xmath20 in eq .
( [ hurst2 ] ) , we get = min[1/2 , /4 , d/4 ] .
[ theta ] this means that for @xmath56 , if @xmath52 , @xmath139 while if @xmath53 , @xmath140 . for @xmath59 , if @xmath61 , @xmath139 , while if @xmath63 , @xmath141 .
although for @xmath56 and @xmath53 , we expect analytically @xmath142 for large @xmath6 , we find numerically ( see section [ num ] ) that it is rather tricky to find this power law .
the latter is a consequence of the eq .
[ c1 ] , which tells us that this long time behaviour arises , predominantly , as a contribution of correlations at distance scales @xmath94 . on the other hand eq .
[ c2 ] shows that the power law @xmath143 arises out of a contribution due to correlations at scale @xmath95 .
so if a finite cutoff @xmath49 is not kept , and the distance scales @xmath144 are not probed carefully , numerically one may find a wrong asymptotic power namely @xmath145 instead of the expected power @xmath146 .
we will discuss this further in the next section [ num ] .
since the exponents in eq . [ hurst2 ] followed from eq .
[ cfinal ] which is an approximate asymptotic form of the exact eq .
[ c ] , we first checked by numerically integrating eq .
[ c ] that the predicted powers are indeed true .
while doing this , we realized that in order to see the power @xmath147 in the regime @xmath148 ( for @xmath149 ) one should be very careful to choose the numerically finite integration measure @xmath150 ( where @xmath151 approximates infinitesimal @xmath152 ) , in order to see the asymptotics .
this is related to the comment made earlier that we have to probe @xmath144 ( where @xmath93 as mentioned earlier ) to get the asymptotic contribution , in this regime .
we now turn to verifying the analytical predictions of the previous section , namely @xmath20 as in eq .
[ hurst2 ] and the corresponding exponent @xmath25 as in eq .
[ theta ] , by doing numerical simulations .
we use the time discretised version of eqs .
[ xc ] and [ yc ] , @xmath153 where @xmath154 , with @xmath155 being an integer .
we choose @xmath156 in our simulations so that the stability is guaranteed @xcite .
the variables @xmath157 and @xmath158 are independent gaussian variables for all @xmath159 and each is distributed with zero mean and unit variance .
moreover our simulations here are all for @xmath160 , i.e. we have a single transverse coordinate @xmath10 .
hence the subscript @xmath161 assumes a single value equal to @xmath27 . unlike @xcite
, here we had to numerically create quenched power law correlated velocity @xmath11 of the form given by eq .
we note that if we have a delta - correlated velocity correlation in fourier @xmath162-space with a @xmath162-dependent prefactor as follows = c(ka ) |k_1|^-1 ( k_1 + k_2 ) , [ vv_k ] with constant @xmath163 $ ] , in the real @xmath10-space we have @xmath164 given by eq .
[ vv ] up to an constant factor .
the factor @xmath165 in eq .
[ vv_k ] is the square of the variance of the distribution of @xmath166 in @xmath162-space . in our numerics , we first generate a realization of gaussian distributed random numbers @xmath167 in @xmath162-space with zero mean and variance proportional to @xmath168 , and then fourier transform back to real space .
the latter realization of random @xmath11 are thus guaranteed to have the required power law correlation in real @xmath10-space .
[ fig1 ] we note that to see the correct asymptotic behaviour , one has to explore distance scales @xmath169 .
we consider a grid along the @xmath10-direction with grid spacing @xmath170 , extending from @xmath171 to @xmath172 . at each point of this grid
we choose @xmath11 by a discrete fourier transform of the random numbers @xmath173 obtained for all @xmath174 ( with @xmath175 ) .
once a set of @xmath176 is thus chosen , they remain fixed at all times during different thermal histories .
this set @xmath176 constitutes a particular realization of disorder .
finally one performs the disorder average @xmath177 by averaging over various realizations of the set @xmath176 . in all our simulations
we have averaged over @xmath178 realizations of disorder using @xmath179 thermal realizations for each set of disorder .
there are further , three important things to be noted .
firstly , for @xmath52 , the role of cutoff @xmath49 is unimportant and hence it may be safely ignored .
we have verified for @xmath160 that the results are unchanged in comparison to finite @xmath49 , by taking @xmath108 , which is equivalent to setting @xmath180 in eq .
however , the role of @xmath49 in the @xmath181 regime can not be overemphasized .
the cutoff @xmath49 has to be chosen such that @xmath182 .
we have used @xmath183 in our simulations . for low values of the ratio @xmath184 , the expected behaviour of @xmath185 is not seen .
the second point is that the value @xmath186 leads to numerical difficulties for @xmath187 .
the latter was taken care of by using a very small finite value of @xmath162 in place of @xmath186 .
finally we note that the two integrals in the constant @xmath188 are rapidly decaying integrals and better results are obtained by using very small grids of integration . in figure 2
, we see that for different @xmath189 , the incremental correlation function @xmath120 with exponent @xmath190 , as predicted by eq .
[ hurst2 ] . in figure 3
, we see that for different @xmath191 , again @xmath120 , but now with @xmath192 .
we directly measured the averaged persistence probability @xmath12 in our simulations .
the procedure is described clearly in @xcite .
our results are shown in figure 3 .
we have shown @xmath12 for two different @xmath7 values : for @xmath193 , the power law exponent @xmath139 as expected , while for @xmath194 we have @xmath195 .
in this paper , we have derived the incremental two - time correlation function @xmath196 and persistence probability @xmath12 of the longitudinal @xmath1-coordinate of a random walker subjected to a quenched random _
power law _ correlated velocity field .
this problem is a variant of the original mdm model , now with long range velocity correlation .
the main result is shown in figure [ fig4 ] .
there are three distinct regions in the parameter space of @xmath0 versus @xmath7 .
@xmath127 for low dimensions ( @xmath56 ) and large power law exponents ( @xmath53 ) , the behaviour of @xmath196 and @xmath12 are same as that of the short - range correlated mdm model @xcite .
thus this regime is _ short - range disorder _ dominated .
@xmath132 on the other hand for strong long - range disorder i.e. @xmath197 $ ] , the temporal behaviour of @xmath196 and @xmath12 depend on @xmath7 continuously , so that long - range correlation really has a new effect .
the latter regime is _ long - range disorder _ dominated .
@xmath136 finally for @xmath59 and @xmath198 , the hurst exponent is @xmath199 , i.e. , the same as that of a simple random walker ( srw ) .
quenched disorder is irrelevant , and this regime is thus _ thermal noise _ dominated .
further , at the boundaries in the fig . [ fig4 ] denoted by the dotted lines , there are logarithmic corrections to the power law expressions for @xmath196 and @xmath12 .
we note that the the quenched random velocity field is having an effect on the temporal behaviour of a srw , only when the disorder is sufficiently strong and transverse dimension @xmath0 is small . for short - range correlated mdm model
@xcite the @xmath20 changes only when the number of transverse directions @xmath56 .
it is as if the velocity kicks in the @xmath1-direction fail to have much impact when the particle has many more transverse directions to explore . in this paper
we find that for @xmath197 $ ] , i.e. for stronger correlated or longer ranged disorder , the short - range correlated mdm model exponents give way to new exponents .
thus even if transverse dimensions are large , sufficiently long - range disorder makes the random walker move super - diffusively . for high dimension ( @xmath129 ) and weak disorder ( @xmath63 ) , the quenched noise looses out to the thermal noise , and the asymptotic temporal properties are that of simple brownian motion .
we also note that the short - distance cutoff scale @xmath49 of the power law , plays a crucial role in this problem . for the case
@xmath53 , numerically , one has to be very careful to explore distance scales much smaller compared to the power - law cutoff scale @xmath49 .
it is one of the most important aspects of this regime that very short distance correlations seem to control the long time behaviour .
for the regime of @xmath52 , since correlations at scale @xmath95 determine the asymptotic behaviour rather than @xmath94 , the cutoff @xmath49 may be ignored .
we are thankful to s.n .
majumdar for reading the manuscript carefully and giving the very important suggestion of analyzing the case of @xmath53 with a finite cutoff .
it led to the correct understanding of the exponents in that regime .
we are also thankful to a. sain for some helpful suggestions .
d. das acknowledges the seed grant no .
@xmath200 of i.i.t .
bombay , india , for financial support .
a. comtet and d.s .
dean , j. phys .
a * 31 * , 8595 ( 1998 ) ; f. igloi and h. rieger , phys .
e * 58 * , 4238 ( 1998 ) ; d.s .
fisher , p. le doussal , and c. monthus , phys .
lett . * 80 * , 3539 ( 1998 ) ; s.n .
majumdar and a. comtet , phys .
e * 66 * , 061105 ( 2002 ) .
b. gaveau and l. schulman , j. stat .
66 , 375 ( 1992 ) ; b. gaveau and a. meritet , lett .
15 , 351 ( 1988 ) ; g. oshanin and a. blumen , phys .
e 49 , 4185 ( 1994 ) ; macromol .
theory simul . 4 , 87 ( 1995 ) | we study the motion of a random walker in one longitudinal and @xmath0 transverse dimensions with a quenched power law correlated velocity field in the longitudinal @xmath1-direction .
the model is a modification of the matheron - de marsily ( mdm ) model , with long - range velocity correlation . for a velocity correlation function , dependent on transverse co - ordinates @xmath2 as @xmath3
, we analytically calculate the two - time correlation function of the @xmath1-coordinate .
we find that the motion of the @xmath1-coordinate is a fractional brownian motion ( fbm ) , with a hurst exponent @xmath4 $ ] . from this and known properties of fbm
, we calculate the disorder averaged persistence probability of @xmath5 up to time @xmath6 .
we also find the lines in the parameter space of @xmath0 and @xmath7 along which there is marginal behaviour .
we present results of simulations which support our analytical calculation .
pacs numbers : 02.50.ey , 05.40.-a , 05.60.-k |
polarization effects @xcite arising from the visible light propagation in the anisotropic or gyrotropic medium are well - known .
the theory @xcite makes prediction about the analogous effects for @xmath0-quanta with energy @xmath1 gev propagating in single crystals , which present the anisotropic medium .
the main absorption process of @xmath0-quanta in the single crystals is the electron - positron pair production .
the cross section of this process depends on the linear polarization of @xmath0-beam with respect to crystallographic planes . as a result , the monochromatic linearly polarized @xmath0-beam
may be presented as a superposition of two electromagnetic waves with different refractive indices due to which the transformation of linear ( circular ) polarization to circular ( linear ) one takes place . on the other hand ,
the process of @xmath2-pair production in single crystals is similar to the same process in a linearly polarized laser wave @xcite .
a possibility to use a bunch of linearly polarized laser photons as a `` single crystal '' is pointed in @xcite , but the concrete calculations of this possibility are not given .
in the recent paper @xcite it has been shown that the polarization effects such as birefringence and rotation of polarization plane for @xmath0- beams with energies of tens gev or more take place for short ( about some picoseconds ) laser bunches and parameters of lasers , which may be provided by real techniques . in the cited paper the differential equations which determine the variation of stokes parameters and intensity of @xmath0-quanta traversing a bunch of arbitrary polarized laser photons are obtained . for these calculations
the well - known scattering amplitudes for the process of elastic scattering light by light @xcite were used .
then in paper @xcite the process of @xmath0-beam propagation in the field of a laser wave was investigated with the use of traditional optical methods .
bisides , the transformation of @xmath0-beam linear polarization into circular in the anisotropic medium was discussed in @xcite .
the case of @xmath0-quanta propagation in single crystals described in @xcite is a special case of such a process .
the general case of @xmath0-propagation in single crystals oriented in the angle region corresponding to process of the coherent pair production was considered in ref.(@xcite ) . in this paper
it was shown that the propagating @xmath0-beam is a superposition of the two elliptically polarized waves but the description of variation of the @xmath0-beam polarization is unavailaible . in the present paper
we study the general case of high energy @xmath0-quanta propagation through the anisotropic medium .
the anisotropic medium is determined as such a medium , which optical properties may be described using a symmetric permittivity tensor @xcite . as will be indicated
the simplest example of the anisotropic medium of the general type is a superposition of the two linearly polarized laser waves with different frequencies moving in the same direction .
we will study in detail the @xmath0-quanta propagation in such a combined laser wave with the goal of better understanding of this process in more complicated cases , as in single crystals .
we write the equations of the electromagnetic field in a medium ( @xmath0-beam propagating in a laser wave , single crystal , and the like ) in the following form @xcite @xmath3 @xmath4 - is the intensity of electric field and @xmath5 and @xmath6 are the electric and magnetic induction vectors , t is the time , c is the speed of light .
all the properties of the medium are reflected in the relation between @xmath7 and @xmath5 .
eqs.([1 ] ) would suffice to describe the @xmath0-beam propagation in a medium and such a property as the intensity of magnetic field is not needed @xcite .
we represent the relation between @xmath5 and @xmath8 in the form @xmath9 where @xmath10 is the complex permittivity tensor and @xmath11 is the frequency of @xmath0-quanta . by way of example of the anisotropic medium
let us consider the superposition of the two linearly polarized laser waves , moving in the same direction .
the frequencies of these waves ( photon energies ) are different . in order to determine the permittivity tensor in the case of a monochromatic field ( high energy @xmath0-beam ) @xmath12 propagating in above - mentioned laser medium , where @xmath13 is the wave vector of the @xmath0-quanta , we find the average energy lost by the electromagnetic wave per unit volume @xmath14 and per unit time @xcite @xmath15 the mechanism by which the wave loses energy is @xmath16-pair production in the field of the laser wave @xcite .
the process is determined primarily by the transverse part of the permittivity tensor , while the longitudinal components of the tensor are higher - order infinitesimals in the interaction constant @xmath17 @xcite . taking this into account , and in the coordinate system one axis of which is oriented parallel to the wave vector of @xmath0-quanta , we have from ( [ 3 ] ) @xmath18 where @xmath19 are stokes parameters of @xmath0-beam . on the other hand , knowing the cross section @xmath20 of the pair production process , we can write @xmath21 where @xmath22 is the number of photons per volume unit of the laser wave with the linear polarization equals @xmath23 , and @xmath24 is the similar number for second wave with the linear polarization equals @xmath25 , @xmath26 and @xmath27 are corresponding cross sections for @xmath16 -pair production in @xmath28 -interactions , @xmath29 and @xmath30 are the stokes parameters of the laser waves ( @xmath31 ) . factor 2 in this formula is due to the counter - motion of the @xmath0-beam and laser wave .
note , that the eq.([5 ] ) is true , when the intensity of laser wave are not high ( see below ) .
we can write the cross section of @xmath16-pair production in the following form @xcite @xmath32 where @xmath33 is the invariant variable , @xmath34 is the @xmath0-quantum energy , @xmath35 and @xmath36 are the mass and classical radius of electron , @xmath37 and @xmath38 are the energy and stokes parameters of the laser photon .
it is well known that the pair production is a threshold process and , because of this , the laser wave is a transparent medium for @xmath0-beam , when @xmath39 or @xmath40 .
there are two different photon energies @xmath41 and @xmath42 in the case of dichromatic laser wave . because of this , it is convenient to use the two corresponding invariant variable @xmath43 and @xmath44 .
it is evident that @xmath45 . comparing eqs.([4 ] ) and ( [ 5 ] ) we can find the components of permittivity tensor caused by @xmath0-quanta absorption . then we can determine the other components of the tensor with the help of the following dispersion relations @xcite @xmath46 @xmath47 where @xmath48 is the kronecker @xmath49-function . comparing eqs.([4 ] ) and ( [ 5 ] ) , we get @xmath50 it easy to verify that @xmath51 .
the same result is evident from the theory of generalized receptivity @xcite .
the other components of the permittivity tensor can be calculated with the help of relations ( [ 9 ] ) -([10 ] ) .
the results of calculations of the components @xmath52 for the arbitrary coordinate system , one axis of which is oriented parallel to the wave vector of the @xmath0-quanta , are presented below @xmath53 @xmath54 @xmath55 @xmath56 @xmath57 @xmath58 where @xmath59 is the mean square of intensity electric field for every laser wave and the functions @xmath60 are equal to : @xmath61 ^ 2+[\sqrt{1+z}-{z\over 2}l_+]^2 - { { \pi^2z^2}\over { 4}},\ ; 0<z \le 1 , \cr -[\sqrt{z-1 } -z \arcctg\sqrt{z-1}]^2+[\sqrt{1+z}- { z\over 2}l_+]^2 , \ ; z>1 .
\cr } \label{21}\ ] ] @xmath62 @xmath63 @xmath64 the function @xmath65 is equal to : @xmath66 the constant @xmath67 is the critical field of quantum electrodynamics .
the presented here data completely determine the permittivity tensor for high energy @xmath0-quanta traversing a dichromatic linearly polarized bunch of laser photons . in a number of problems in optics
it is more convenient to employ the tensor @xmath68 , which is inverse of the tensor @xmath52 . when @xmath69 , these tensors are related by @xmath70 it should be noted the following peculiarities of the permittivity tensor : \1 ) our description can be extended to the case when the laser bunch is superposition more then two linearly polarized waves .
it is obvious that the analogous terms should be added in the tensor components in these cases .
\2 ) let us that @xmath71 .
then the laser bunch is a transparent medium at @xmath72 . in this case
the all components @xmath73 are equal to zero .
\3 ) in the general case the symmetric complex tensor @xmath73 does not reduce to principal axes @xcite
( i.e. , there does not exist a coordinate system in which the tensors @xmath74 and @xmath73 are simultaneously diagonal ) .
4)in the case , when the two waves have the same direction of linear polarization or their polarizations are orthogonal , the tensor @xmath52 can be reduced to a diagonal form .
the permittivity tensor for monochromatic linearly polarized wave can be reduced to the diagonal form for the all time .
everything said above is also true for tensor @xmath68 .
the main problem of optic of anisotropic medium is to investigate the propagation of monochromatic plane waves , characterized by definite values of the frequency @xmath11 and wave vector @xmath75 .
such waves , satisfying a homogeneous wave equation , are called normal electromagnetic waves @xcite , and they have the form @xmath76 where @xmath77 is the complex vector , independent of coordinates @xmath78 and time , @xmath79 is the complex index of refraction and @xmath80 is a real unit vector .
the vectors @xmath5 and @xmath6 have the same form . from maxwell s equations ( 1 )
we obtain the wave equation @xmath81 taking into account the relation between @xmath5 and @xmath8 in a system of coordinates in which the axis @xmath82 is oriented parallel to the wave vector , we obtain @xmath83 for a monochromatic plane wave it follows from these equations that @xmath84 from the condition that the two homogeneous equations are compatible , we find the index of refraction of the @xmath0-quanta @xmath85 where @xmath86 and @xmath87 are , respectively , the trace and determinant of the matrix @xmath68 . thus ,
in the general case the @xmath0-beam propagates through the laser wave as the superposition of two electromagnetic waves with different refractive indices .
note , that the two roots of ( [ 30 ] ) which have form @xmath88 , are superfluous .
they correspond to the @xmath0-quanta motion in the reverse direction . in the general case the refractive indices are complex values .
however , they are real values , when the laser bunch is a transparent medium for @xmath0-quantum ( all components of the permittivity tensor are real number in this case ) .
fig.1 illustrates the refractive indices as functions of the invariant variable @xmath89 ( the laser wave parameters are in the caption ) .
here we consider the polarization properties one of two normal electromagnetic waves . from dispersion equations ( [ 29 ] ) we find the ratios of the components of the vector @xmath5 @xmath90 where @xmath49 is the phase shift between @xmath91 and @xmath92 .
this ratio @xmath93 can be reduced to zero or to the form @xmath94 ( since @xmath95 , where @xmath96 and @xmath97 are the semiaxes of the ellipse and @xmath98 @xcite ) by the rotation of the coordinate system around the wave vector of @xmath0-quanta ( the @xmath82-axis is constantly aligned with the wave vector ) .
the first case corresponds to the propagation of a linearly polarized wave and the second case corresponds to an elliptically polarized wave ; in addition , @xmath99 corresponds to left ( right ) - hand polarization of @xmath0-quanta .
the different cases of @xmath0-beam propagation in anisotropic medium , which optical properties described by the symmetric tensor , were considered in paper @xcite . in the case when permittivity tensor may be reduce to principal axes ( i.e. , there is a coordinate system in which @xmath100 ) the normal electromagnetic waves are linearly polarized . in general case the permittivity tensor does not reduce to principal axes and the normal waves are elliptically polarized .
the propagation of @xmath0-beam , which is a superposition of the two linearly polarized waves , was considered in a number papers @xcite . because of this , in the following we will consider the case when @xmath0-beam is the superposition of the two elliptically polarized normal waves . in the case under consideration
the refractive indices are the complex values . because of this , the value @xmath93 is also complex and we get the following relation between two normal waves @xmath101 where the indices in parentheses refer to the waves with refractive indices @xmath102 and @xmath103 . in
what follows we will use only one of two values , namely , the @xmath104 ( without pointing any indices ) . in our case
one can obtain @xmath105 where the indices in parentheses refer to waves with refractive indices @xmath102 and @xmath103 .
from here we can see that @xmath106 and @xmath107 vectors are orthogonal but @xmath106 and @xmath108 vectors are not orthogonal if the value @xmath109 is not equal to zero .
let us name the stokes parameters of the normal wave with the refractive indices @xmath102 and @xmath103 respectively as @xmath110 and @xmath111 .
then we get @xmath112 we have also the following relations @xmath113 . the angle of ellipse turn @xmath114 is found from relation @xmath115 .
fig.2 illustrates the results of calculations of @xmath116 as functions of the invariant variable @xmath117 under various conditions .
now we can find the relations describing the variations of intensity and stokes parameters of @xmath0-quanta propagating in the uniform ( @xmath118 ) laser wave . then representing the @xmath0-beam as the superposition of two normal waves corresponding to the polarization state of a laser wave we get the following relations @xmath119 where @xmath120 are the intensity and stokes parameters of @xmath0-quanta on the laser bunch thickness equal to x. the partial intensities @xmath121 have the following form ( the physical sense of these values is easy to understand , if the relation @xmath122 is written in the component - wise form ) @xmath123 the initial partial intensities are defined from the following relations @xmath124 the relations between @xmath125 and @xmath126 values were used , because of this the @xmath126-values are absent in eqs.([45])-([48 ] ) .
besides , we assume that @xmath127 .
the parameters @xmath128 have the following form : @xmath129 eqs.([37])-([40 ] ) describe the general case of @xmath0-beam propagation , when the variations intensity and stokes parameters are determined by the imaginary values of refractive indices and the difference of their real quantities . however , these equations do not described such cases , when the relation @xmath130 takes place . in the case @xmath130
, one can use the known relations from papers @xcite or one can find limits of the obtained here eqs.([37])-([40 ] ) .
for example , one can offer @xmath131 and @xmath49 direct to zero .
figs.3,4 illustrate the variations of stokes parameters and intensity of an initially unpolarized @xmath0-beam moving in the dichromatic laser wave .
the influence of the laser wave intensity on the @xmath16-pair production was studied in a number of papers ( see ref.@xcite and literature therein ) .
the degree of intensity of a dichromatic laser wave can be characterized by the dimensionless parameter @xcite @xmath132 where @xmath133 and @xmath134 . here
we have considered the case of the relatively not high intensity of a laser wave , when @xmath135 .
the results of papers @xcite are allowed one to write down the components of permittivity tensor taking into account the first terms of the expansion of the intensity in taylor series .
thus we have found that the already obtained components of the tensor should be transformed with the help of the following simple rules .
firstly , the variables @xmath136 are substituted by the variables @xmath137 . secondly , the value @xmath138 ( the critical field ) in eqs.([15])-([20 ] ) is substituted by the value @xmath139 .
thirdly , the functions @xmath140 are substituted by the functions @xmath141 where @xmath142 .
the new condition for the pair production threshold follows from these rules .
it is @xmath143 , where i is index of wave with the most photon energy .
it means that the threshold energy of @xmath0-beam enhances ( at the fixed frequency of laser photons ) . in the strict sense
the field of application of these more refined relations satisfies the condition @xmath135 .
nevertheless , we can receive the important information in this case @xcite . in the case ,
when the parameter @xmath144 , the pair production process is similar to an analogous process in the constant electromagnetic field .
the permittivity tensor for these fields was found in ref.@xcite and some particular calculations of @xmath0-quanta propagation are in @xcite .
the optical properties of an anisotropic medium can be described by the use of the symmetric permittivity tensor . in general case
the tensor components are complex values .
it means , broadly speaking , that the permittivity tensor does not reduce to principal axes with the result that the normal electromagnetic waves ( the eigenfunctions of the problem ) present two elliptically polarized waves . because of
this some peculiarities in @xmath0-quanta propagation in anisotropic medium are appeared .
the simplest example of such a medium of the general type is the dichromatic laser wave involving two linearly polarized waves with different frequencies .
other example is a single crystal oriented in region of the coherent pair production process @xcite .
the permittivity tensor and polarization characteristics of normal waves in single crystals were obtained in paper @xcite , and it was shown that orientation regions are in single crystals , where the circular polarization of normal waves is high ( @xmath145 in maximum ) .
however in single crystals the components of permittivity tensor are a sum of sufficiently large number of terms that make the investigation more difficult .
note that eqs.([37]-[40 ] ) for variations of @xmath0-beam intensity and stokes parameters are true for an arbitrary anisotropic medium , when @xmath146 we made some calculations of @xmath0-beam propagation in the field of dichromatic wave in the case , when the frequency ratio @xmath147 is equal to 2 ( see figs.(1 - 4 ) ) .
we take the values @xmath148 for the polarization state of the dichromatic wave .
it means that the angle between directions of these two polarizations is equal to 45 degree .
one can see from fig.2 that value @xmath149 depends on ratio of the electric field intensities @xmath150 .
the value @xmath151 , when r=0.658 or 4.12 ( curves 1 and 4 on fig.2 ) .
the refractive indices for r=0.658 are shown on fig.1 .
one can see that the real and imaginary parts of refractive indices of two normal electromagnetic waves are equal in magnitude at @xmath152 .
this is the so - called in classical crystal optics case of singular axis @xcite .
now we can make a conclusion that in single crystals the high degree of circular polarization of normal waves @xcite is due to the common action of the two `` strong '' crystallographic planes with the 45 degree angle between them .
the ( 110 ) and ( 010 ) planes are responsible for the effect at the conditions of the cited paper . fig.3 illustrates the variations of stokes parameters as functions of the laser bunch thickness . the behavior of these curves it is easily to understand .
as already noted , the @xmath0-beam in a medium can be presented as a superposition of two normal electromagnetic waves with the different refractive indices .
because of this , one normal wave is adsorbed more then another wave and after propagation of some thickness @xmath82 only this wave would then be left behind and @xmath153 or @xmath154 . referring to fig.3
it will be observed that such a thickness is enough large when @xmath155 .
the reason is that the difference of imaginary parts of refractive indices is a small value in the pointed region of the variable @xmath117(see fig.1 ) .
the dichromatic wave or single crystals @xcite are sensitive to the circular polarization of @xmath0-beam ( see figs.3,4 ) .
so , initially an unpolarized @xmath0-beam moving in an anisotropic medium become circularly polarized one .
this case is differ from the known case @xcite of the @xmath0-beam propagation in the anisotropic medium . as is shown in ref.@xcite the only linearly polarized beam
is transformed in circularly polarized one . in principle , the single crystal sensitivity to a circular polarization can be used for determination of polarization of high energy ( in tens gev and more ) @xmath0-quanta and electrons .
it is believed that a @xmath0-beam propagation in the linearly polarized monochromatic laser wave moving in the magnetic field ( normally to it direction ) is similar to such a propagation in a dichromatic wave .
the another analogous example is two monochromatic laser waves with the equal frequencies moving at an nonzero angle between them .
on the other hand we would expect that high energy @xmath0-quanta , propagating over cosmological distances , can be polarized due to their interaction with magnetic fields and/or extragalactic starlight photons @xcite . with this point of view
one can explain the anisotropy in electromagnetic propagation over cosmological distances @xcite , as might appear at first sight .
however this explanation is required a high electromagnetic energy density in the cosmic space ( @xmath156 , if take the birefringence scale of order @xmath157 cm @xcite ) .
this estimate follows from ref.@xcite where asymptotic values of the refractive indices are calculated . the propagation of @xmath0-quanta through a laser wave ( when @xmath135 ) is similar to the same process in single crystals for the region of coherent pair production .
for example , the permittivity tensor in single crystals @xcite is determined by the functions @xmath158 as in a laser wave . however , the existence of some frequencies of pseudophotons and incoherent pair production in single crystals is the main difference between these two cases .
note that the permittivity tensor components ( see also ref.@xcite ) can be presented as a linear combinations of the invariant helicity amplitudes for the forward light by light scattering @xcite .
note that there are no experiments yet in support of the transformation of @xmath0-beam polarization in single crystals and laser waves .
nevertheless , a number of proposals on the investigation and utilization of this phenomenon @xcite is available .
landau and e.m.lifshitz , electrodynamics of continuous media , pergamon , oxford ( 1960 ) .
v.m . agranovich and v.l .
ginzburg , spatial dispersion in crystal optics and the theory of excitons , wiley , new york ( 1967 ) .
n.cabibbo , g. da prato , g. de franceschi , and u. mosco phys .
lett , 1962 .
v.9 p.435 .
baier , v.m.katkov , and v.m .
strakhovenko , electromagnetic processes at high energies in oriented single crystals [ in russian ] , nauka , novosibirsk , 1989 . v.a .
maisheev , v.l .
mikhalev , a.m. frolov , jetp , v.101 , p.1376 , ( 1992 ) ; sov.phys .
jetp v.74 , p.740 , ( 1992 ) .
kotkin and v.g .
serbo , preprint hep - ph/9611345 ( submitted to phys
lett . ) b.de tollis , il nuovo chimento 32 , 754 ( 1964 ) and 35 , 1182 ( 1965 ) .
v.b.berestetskii , e.m .
lifshitz , and l.p .
pitaevskii , quantum electrodynamics , pergamon , oxford ( 1982 ) .
maisheev , preprint ihep 97 - 25 , protvino , 1997 .
r.o.avakian , s.m.darbinian , k.a.ispirian , m.k.ispirian , nucl.instr . and meth . , 1986 ,
b 115 , p.387 .
nikishov , v.i .
ritus jetp v.52 , p. 1707
v.g . baryshevskii and v.v .
tikhomirov , usp .
nauk v.154 , p.529 , ( 1989 ) sov .
v. 32 , p. 1013
landau , and e.m .
lifshitz statistical physics , moscow , nayka , 1964 .
landau , and e.m .
lifshitz , the classical theory of fields , pergamon , oxford ( 1975 ) .
maisheev , v.l .
mikhalev , a.m. frolov , preprint ihep 91 - 30 , protvino , 1991 .
v.n.baier , v.m .
katkov , and v.s.fadin , radiation from relativistic electrons [ in russian ] , atomizdat , moscow , 1973 .
a.m. frolov , v.a .
maisheev , v.l .
mikhaljov nucl .
instr . and meth
. 1987 v. a254 , p. 549 .
m.h.salamon and f.w.stecker .
preprint : astro - ph/9708182 b.nodland , j.p.ralston .
preprint : astro - ph/9704196 | propagation of @xmath0-beam in the anisotropic medium is considered .
the simplest example of such a medium of the general type is a combination of the two linearly polarized monochromatic laser waves with different frequencies ( dichromatic wave ) .
the optical properties of this combination are described with the use of the permittivity tensor .
the refractive indices and polarization characteristics of normal electromagnetic waves propagating in the anisotropic medium are found .
the relations , describing variations of @xmath0 -beam intensity and stokes parameters as functions of the propagation length , are obtained .
the influence of laser wave intensity on the propagation process is calculated .
the @xmath0-beam intensity losses in the dichromatic wave depend on the initial circular polarization of @xmath0-quanta .
this effect is also applied to single crystals which are oriented in some regions of coherent pair production . in principle
, the single crystal sensitivity to a circular polarization can be used for determination of polarization of high energy ( in tens gev and more ) @xmath0-quanta and electrons .
-2.3 cm 25 cm |
rabin s tree theorem states , via an automata - theoretic proof , the decidability of the monadic second order ( short : @xmath0 ) theory of the complete binary tree .
it allows to derive the decidability of seemingly very different theories ( e.g. , the @xmath0-theory of the real line where set quantification is restricted to closed sets @xcite ) .
its importance is stressed by seese s result that any class of graphs of bounded degree with a decidable @xmath0-theory has bounded tree - width ( i.e. , is `` tree - like '' ) @xcite . in @xcite ,
shelah reports a generalization of rabin s tree theorem that was proved by shelah and stupp .
the idea is to start with a structure @xmath1 and to consider the tree whose nodes are the finite words over the universe of @xmath1 together with the prefix order on these words .
then the immediate successors of any node in this tree can naturally be identified with the elements of the structure @xmath1 hence they carry the relations of @xmath1 .
the resulting tree with additional relations is called _ shelah - stupp - iteration_. the above mentioned result of shelah and stupp states that the @xmath0-theory of the shelah - stupp - iteration can be reduced to the @xmath0-theory of the base structure @xmath1 .
if @xmath1 is the two - elements set , then rabin s tree theorem follows .
a further extension is attributed to muchnik @xcite who added a unary clone predicate to shelah and stupp s iteration resulting in the _ muchnik - iteration_. this clone predicate states that the last two letters of a word are the same .
this allows , e.g. , to define the unfolding of a rooted graph in its muchnik - iteration @xcite .
muchnik s theorem then gives a reduction of the mso - theory of the muchnik - iteration to the mso - theory of the base structure .
the proof was not published by muchnik himself , but , using automata - theoretic methods , walukiewicz showed that the reduction in muchnik s theorem is even uniform ( i.e. , independent from the concrete base structure ) @xcite .
since , as mentioned above , the unfolding of a rooted graph can be defined in the muchnik - iteration , the mso - theory of this unfolding can be reduced to that of the graph @xcite .
this result forms the basis for caucal s hierarchy @xcite of infinite graphs with a decidable mso - theory .
walukiewicz s automata - theoretic proof ideas have been shown to work for the muchnik - iteration and stronger logics like courcelle s counting mso and guarded second - order logic by blumensath & kreutzer @xcite . in @xcite , we asked for a first - order version of muchnik s result and failed .
more precisely , we constructed structures with a decidable first - order theory whose muchnik - iteration has an undecidable first - order theory . as it turns out , the only culprit is muchnik s clone predicate since , on the positive side , we were able to uniformly reduce the first - order theory ( and even the monadic chain theory where set variables range over chains , only ) of the shelah - stupp - iteration to the first - order theory of the base structure .
the aim of this paper is to clarify the role of weak monadic second order logic @xmath2 in the context of shelah - stupp- and muchnik - iteration .
we first define infinitary versions of these iterations that contain , in addition to the finite words , also @xmath3-words . on the positive side , we prove a rather satisfactory relation between the theories of the infinitary shelah - stupp - iteration and the base structure .
more precisely , the shelah - stupp result together with some techniques from @xcite allows to uniformly reduce the @xmath4-theory of the infinitary shelah - stupp - iteration ( where set quantification is restricted to closed sets ) to the mso - theory of the base set .
our result from @xcite ensures that shelah - stupp - iteration is @xmath5-compatible in the sense of courcelle ( i.e. , the @xmath5-theory of the infinitary shelah - stupp - iteration can be reduced uniformly to the @xmath5-theory of the base structure ) .
our new positive result states that shelah - stupp - iteration is also @xmath2-compatible . to obtain this result
, one first observes that the finiteness of a set in the shelah - stupp - iteration is definable in @xmath6 ( where quantification is restricted to finite unions of chains ) , hence the @xmath2-theory of the shelah - stupp - iteration can be reduced to its @xmath6-theory . for this logic , we then prove a result analogous to rabin s basis theorem : any consistent @xmath6-property in the shelah - stupp - iteration of a finite union of chains ( i.e. , of a certain set of words over the base structure ) has a witness that can be accepted by a small automaton . but an automaton over a fixed set of states can be identified with its transition matrix , i.e. , with a fixed number of finite sets in the base structure .
we then prove that @xmath6-properties of the language of an automaton can effectively be translated into @xmath2-properties of the transition matrix . on the negative side
, we prove that infinitary muchnik - iteration is not @xmath2-compatible .
namely , there is a tree @xmath7 with decidable @xmath2-theory such that for any set @xmath8 of natural numbers , there exists an @xmath2-equivalent tree @xmath9 such that @xmath8 can be reduced to the @xmath2-theory of the infinitary muchnik - iteration of @xmath9 .
this proof uses the fact that the existence of an infinite branch in a tree is not expressible in @xmath2 , but it is a first - order ( and therefore a @xmath2- ) property of the infinitary muchnik - iteration .
a _ ( relational ) signature _ @xmath10 consists of finitely many constant and relation symbols ( together with the arity of the latter ) ; a _ purely relational signature _ does not contain any constant symbols .
formulas use _ individual _ and _ set variables _ , usually denoted by small and capital , resp . , letters from the end of the alphabet . _ atomic formulas _
are @xmath11 , @xmath12 , and @xmath13 where @xmath14 is an @xmath15-ary relation symbol from @xmath10 , @xmath16 are individual variables or constant symbols , and @xmath17 is a set variable .
_ formulas _ are obtained from atomic formulas by conjunction , negation , and quantification @xmath18 for @xmath19 an individual or a set variable .
a _ sentence _ is a formula without free variables .
the satisfaction relation @xmath20 between a @xmath10-structure @xmath1 and formulas is defined as usual . for two @xmath10-structures @xmath1 and @xmath21
, we write @xmath22 if , for any sentence @xmath23 of quantifier depth at most @xmath24 , we have @xmath25 iff @xmath26 . if @xmath1 and @xmath21 agree on all first - order formulas ( i.e. , formulas without set quantification ) of quantifier depth at most @xmath24 , then we write @xmath27 .
let @xmath28 be a partially ordered set .
a set @xmath29 is a _ chain _
if @xmath30 is linearly ordered , it is a _ multichain _ if @xmath8 is a finite union of chains .
an element @xmath31 is a _ branching point _ if @xmath32 is nonempty and does not have a least element .
we will also consider different restrictions of the satisfaction relation @xmath20 where set variables range over certain subsets , only .
in particular , we will meet the following restrictions .
* set quantification can be restricted to finite sets , i.e. , we will discuss weak monadic second order logic .
the resulting satisfaction relation is denoted @xmath33 and the equivalence of structures @xmath34 .
* set quantification can be restricted to chains ( where we assume a designated binary relation symbol @xmath35 in @xmath10 ) which results in @xmath36 and @xmath37 , cf .
thomas @xcite .
* @xmath38 etc .
refer to the restriction of set quantification to multichains .
* the superscript @xmath39 denotes that set variables range over closed sets , only ( where we associate a natural topology to any @xmath10-structure ) , cf .
rabin @xcite .
let @xmath40 be some transformation of @xmath10-structures into @xmath41-structures , e.g. , transitive closure .
a very strong relation between the @xmath42-theory of @xmath1 and the @xmath43-theory of @xmath44 is the existence of _ a single _ computable function @xmath45 that reduces the @xmath43-theory of @xmath44 to the @xmath42-theory of @xmath1 _ for any @xmath10-structure @xmath1_. as shorthand for this fact , we say `` the transformation @xmath40 is @xmath46-compatible '' or , slightly less precise `` the @xmath43-theory of @xmath44 is _ uniformly reducible _ to the @xmath42-theory of @xmath1 . ''
@xmath47-compatible transformations are simply called @xmath43-compatible .
any @xmath0-transduction is @xmath0-compatible @xcite and finite set interpretations are @xmath48-compatible @xcite .
feferman & vaught showed that any generalized product is fo - compatible @xcite .
finally , any generalized sum is @xmath0-compatible by shelah @xcite .
let @xmath49 be a ( not necessarily finite ) alphabet . with @xmath50
we denote the set of all finite words over @xmath49 , @xmath51 is the set of infinite words , and @xmath52 .
the prefix relation on finite and infinite words is @xmath35 .
the set of finite prefixes of a word @xmath53 is denoted @xmath54 , if @xmath55 , then @xmath56 . for @xmath57 and @xmath58
let @xmath59 denote the left - quotient of @xmath60 with respect to @xmath61 .
let @xmath10 be a relational signature and let @xmath62 be a structure over the signature @xmath10 .
the _ infinitary shelah - stupp - iteration @xmath63 of @xmath1 _ is the structure @xmath64 where , for @xmath65 , @xmath66 the _ ( finitary ) shelah - stupp - iteration @xmath67 _ is the restriction of @xmath63 to the set of finite words @xmath50 .
[ e - tree ] suppose the structure @xmath1 has two elements @xmath68 and @xmath69 and two unary relations @xmath70 and @xmath71
. then @xmath72 and @xmath73 .
hence the finitary shelah - stupp - iteration @xmath67 can be visualized as a complete binary tree with unary predicates telling whether the current node is the first or the second son of its father .
in addition , the root @xmath74 is a constant of the shelah - stupp - iteration @xmath67 .
furthermore , the infinitary shelah - stupp - iteration @xmath63 adds leaves to this tree at the end of any branch .
since this allows to define @xmath75 in @xmath63 , the unrestricted mso - theory of @xmath63 is undecidable .
[ e - rsm ] a _ recursive state machine _ @xmath76 ( cf .
@xcite ) over a finite alphabet @xmath77 is a tuple @xmath78 where each _ component state machine _ @xmath79 consists of the following pieces * a set @xmath80 of _ nodes _ and a set @xmath81 of _ boxes _ ( the only essential difference between this definition and the original one from @xcite is that these sets are not required to be finite ) . * a labeling @xmath82 that assigns to every box the index of one of the component machines @xmath83 . * a tuple of sets of _ interface nodes _ @xmath84 . *
a transition relation @xmath85 for some fixed finite alphabet @xmath77 . in the following , we assume the sets @xmath77 , @xmath81 and @xmath86 mutually disjoint .
then @xmath87 is the set of boxes of @xmath76 , @xmath88 the set of nodes , @xmath89 the set of @xmath90 interface nodes , and @xmath91 is a function from @xmath92 into @xmath93 .
finally , let @xmath94 denote the union of all the local transition relations .
a _ state _ of @xmath76 is a tuple @xmath95 where @xmath96 and @xmath97 are boxes and @xmath98 is a node .
the set of all states is denoted @xmath99 .
when the recursive state machine @xmath76 is in state @xmath100 , then the control is with the component machine no .
global moves of the machine come in four forms : ( a ) the component machine @xmath102 moves from one node to another , ( b ) it moves from a node to a box @xmath103 which transfers control to the component machine no .
@xmath104 , ( c ) it returns control to the calling machine no .
@xmath105 which moves into some node , or ( c ) it returns control to the calling machine no .
@xmath105 that makes a move into another box @xmath103 and therefore transfers control to @xmath106 .
this defines a labeled transition system @xmath107 , the _ unfolding of @xmath76_. this unfolding is a relational structure that we want to define in the shelah - stupp - iteration of the recursive state machine @xmath76 . for this to work
, we first have to transform @xmath76 into a relational structure over some signature @xmath10 . for @xmath108 , we consider the following sets : @xmath109 then the following relational structure @xmath1 encodes the recursive state machine @xmath76 : @xmath110 then the unfolding of @xmath76 can be defined in the shelah - stupp - iteration of @xmath1 .
for this , one constructs @xmath5-formulas that ( 1 ) hold true of a word @xmath111 iff @xmath112 is a state and ( 2 ) hold true of a pair of states @xmath113 iff @xmath114 .
[ l - monoids ] * ( cf .
@xcite ) * the shelah - stupp iteration allows to reduce the cayley graph of a free product to the cayley graphs of the factors .
let @xmath115 be monoids finitely generated by @xmath116 for @xmath117 and let @xmath118 denote the rooted cayley graph of @xmath119 .
then the cayley graph @xmath120 of the free product @xmath121 of these monoids can be defined in the shelah - stupp iteration of the disjoint union of the cayley graphs @xmath122 . for this to work ,
let @xmath123 be the disjoint union of the monoids @xmath119 and consider the structure @xmath124 where @xmath125 is the set of units .
then a word @xmath126 belongs to the direct product @xmath127 iff the following holds in the shelah - stupp iteration of @xmath76 : @xmath128 where @xmath129 denotes the immediate successor relation of the partial order @xmath35 . for @xmath130 and @xmath131 ,
we have @xmath132 ( i.e. , @xmath133 ) iff the shelah - stupp iteration satisfies @xmath134 muchnik introduced the additional unary _ clone predicate _ @xmath135 .
the extension of the shelah - stupp - iterations by this clone predicate will be called _ finitary and infinitary muchnik - iteration _ @xmath136 and @xmath137 , resp .
courcelle and walukiewicz @xcite showed that the unfolding of a directed rooted graph @xmath138 can be defined in the muchnik iteration @xmath139 of @xmath138 . to simplify notation
, we will occasionally omit the word `` finitary '' and just speak of the shelah - stupp- and muchnik - iteration .
rabin s tree theorem @xcite states the decidability of the monadic second order theory of the complete binary tree . as a _
corollary _ of his proof technique by tree automata , one obtains rabin s basis theorem ( * ? ? ?
* theorem 26 ) : let @xmath23 be a formula with free variables @xmath140 and let @xmath141 be regular languages such that the binary tree satisfies @xmath142
. then it satisfies @xmath143 where @xmath144 is obtained from @xmath23 by restricting all quantifications to regular sets . to obtain this basis theorem
, it suffices to show that validity of @xmath145 implies the existence of a regular set @xmath146 such that @xmath147 holds true in the binary tree .
this is precisely what this section shows in our context of @xmath6 and the shelah - stupp - iteration @xmath67 .
even more , we will not only show that the set @xmath146 can be chosen regular , but we will also bound the size of the automaton accepting it . _ throughout this section , @xmath10 denotes some purely relational signature . _ for @xmath148 ,
let @xmath149 be the extension of the signature @xmath150 by @xmath151 constants and @xmath152 unary relations .
using hintikka - formulas ( see @xcite for the definition and properties of these formulas ) one can show that for any of the signatures @xmath149 and @xmath153 , there are only finitely many equivalence classes of @xmath154 . an upper bound @xmath155 for the number of equivalence classes of @xmath154 on formulas over the signature @xmath156 can be computed effectively .
now let @xmath157 be some @xmath10-structure .
for @xmath58 , let @xmath158 denote the @xmath159-structure @xmath160 where * the relation @xmath161 is the restriction of @xmath35 to @xmath162 and * @xmath163 is the restriction of @xmath164 to @xmath165 . for any @xmath166
, the mapping @xmath167 with @xmath168 is an isomorphism this is the reason to consider @xmath163 and not the restriction of @xmath164 to @xmath162 .
similarly , the @xmath169-structure @xmath170 is defined for @xmath171 with @xmath172 . here , again , @xmath163 is the restriction of @xmath164 to @xmath173 .
frequently , we will consider the structure @xmath67 together with some additional unary predicates @xmath174 .
as for the plain structure @xmath67 , we will also meet the restriction of @xmath175 to the set @xmath162 , i.e. , the structure @xmath176 . to simplify notation ,
this will be denoted @xmath177 ; the structure @xmath178 is to be understood similarly .
* example [ e - tree ] ( continued ) . * in the case of example [ e - tree ] , @xmath158 is just the subtree rooted at the node @xmath61 . on the other hand
, @xmath179 is obtained from @xmath158 by deleting all descendants of @xmath180 and marking the node @xmath180 as a constant .
thus , we can think of @xmath179 as a tree with a marked leaf .
these _ special trees _ are fundamental in the work of gurevich & shelah @xcite and of thomas @xcite . in the following ,
fix some @xmath181 .
we then define the operations of product and infinite product of @xmath149-structures : if @xmath182 is a @xmath156-structure and @xmath183 a disjoint @xmath149-structure with @xmath184 , then their _ product _
@xmath185 is a @xmath149-structure .
it is obtained from the structure @xmath186 by identifying @xmath187 and @xmath188 , taking the transitive closure of the partial orders , and extending the resulting structure by the list of constants @xmath189 .
now let @xmath190 be disjoint @xmath156-structures with constants @xmath191 and @xmath192 for @xmath193 .
then the _ infinite product _
@xmath194 is a @xmath195-structure .
it is obtained from the disjoint union of the structures @xmath190 by identifying @xmath192 and @xmath196 for any @xmath193 .
the only constant of this infinite product is @xmath197 .
if @xmath198 for all @xmath193 , then we write simply @xmath199 for the infinite product of the structures @xmath190 .
standard applications of ehrenfeucht - frass - games ( see @xcite ) yield : [ p - ef ] let @xmath200 , @xmath201 be @xmath156-structures for @xmath193 and let @xmath202 be some @xmath203-structures such that @xmath204 for @xmath193 and @xmath205 . then @xmath206 [ r - typical - use ] we sketch a typical use of the above proposition in this section .
let @xmath207 be some sufficiently long word .
since @xmath154 has only finitely many equivalence classes , there exist words @xmath208 with @xmath209 and @xmath210 such that @xmath211 .
hence we obtain @xmath212 ( this proves that every consistent property of a single element of @xmath67 is witnessed by some `` short '' word . )
the last isomorphism does not hold for the muchnik - iteration since the clone predicate allows to express that the last letter of @xmath61 and the first letter of @xmath180 are connected by some edge in the graph @xmath1 .
we consider complete deterministic finite automata @xmath213 , called _ automata _ for short .
its language is denoted @xmath214 .
we will also write @xmath215 for @xmath216 .
the _ transition matrix _ of @xmath217 is the tuple @xmath218 with @xmath219 .
an _ automaton _ is a tuple @xmath213 where * @xmath99 is a finite set of states , @xmath220 is the initial state , * @xmath92 is a finite alphabet , * @xmath221 is a transition function , and * @xmath222 is a set of accepting states .
the _ language @xmath214 accepted by @xmath217 _ is the set of words @xmath223 with @xmath224 ( where we extended the function @xmath225 to a function from @xmath226 to @xmath99 in the usual way ) . as explained above
, we will use automata to describe subsets of the shelah - stupp iteration @xmath67 , i.e. , the alphabet @xmath92 will always be a finite subset of the universe of @xmath1 .
these regular subsets have the following nice property whose proof is obvious . let @xmath1 be a @xmath10-structure with universe @xmath49 and let @xmath213 be an automaton with alphabet @xmath227 .
then , for any @xmath228 with @xmath229 , the mapping @xmath230 is an isomorphism from @xmath231 onto @xmath232 . as a consequence ,
the number of isomorphism classes of structures @xmath232 is finite .
this fails in the muchnik - iteration even for @xmath233 : with @xmath234 and @xmath235 , we have @xmath236 iff @xmath237 since the structure @xmath238 can be defined in @xmath239 .
while multichains in the shelah - stupp - iteration can be rather complicated , this section shows that , up to logical equivalence , we can restrict attention to `` simple '' multichains . here
, `` simple '' means that they are regular and , even more , can be accepted by a `` small '' automaton .
_ for the rest of this section , let @xmath157 be some fixed @xmath10-structure and @xmath240 . for @xmath241 , let @xmath242 be automata with @xmath243 such that @xmath244 is a multichain in the shelah - stupp iteration @xmath67 .
write @xmath245 for the tuple of multichains @xmath246 . _ [ p - regular - chain - exists ] let @xmath247 be a chain
. then there exist @xmath166 , @xmath248 , and @xmath249 such that @xmath250 for all @xmath251 and @xmath252 with @xmath253 .
one shows the existence of @xmath254 such that @xmath255 is a chain , @xmath256 for all @xmath251 , @xmath257 , and @xmath258 .
this uses arguments similar to those in remark [ r - typical - use ] and ramsey s theorem .
the result follows with @xmath259 , @xmath260 , @xmath261 , and @xmath262 .
the above proposition shows that every consistent property of a chain is witnessed by some regular chain @xmath263 . using the pigeonhole principle and arguments as in remark [ r - typical - use ]
, one can bound the lengths of @xmath61 and @xmath180 to obtain [ p - chain - quantification ] let @xmath247 be a chain .
then there exists an automaton @xmath264 with at most @xmath265 states such that @xmath266 is a chain and @xmath267 .
it is our aim to prove a similar result for arbitrary multichains in place of the chain @xmath268 in the proposition above .
certainly , in order to get a small automaton for a multichain , the branching points of this set have to be short words . again using arguments as in remark [ r - typical - use ] , one obtains [ l - small - set-1 ] let @xmath269 be a multichain .
then there exists a multichain @xmath270 such that * @xmath271 and * any branching point of @xmath272 has length at most @xmath273 .
[ l - small - set-2 ] let @xmath8 be a multichain such that all branching points of @xmath8 have length at most @xmath274 .
then there exists an automaton @xmath264 with at most @xmath275 many states such that @xmath266 is a multichain and @xmath276 .
let @xmath277 and @xmath278 .
the lemma is shown by induction on @xmath279 .
if @xmath280 , then @xmath8 is a chain , i.e. , the result follows from prop .
[ p - chain - quantification ] .
now let @xmath8 be a multichain such that any branching point has length at most @xmath281 .
by the induction hypothesis , for every @xmath282 , there exists an automaton @xmath283 with at most @xmath284 many states such that @xmath285 is a multichain and @xmath286 let @xmath287 be the equivalence relation on @xmath49 with @xmath288 if and only if 1 .
@xmath289 for all @xmath290 and 2 .
@xmath291 .
let @xmath292 contain precisely one element @xmath293 from any @xmath287-equivalence class .
then the set @xmath294 is a multichain and can be accepted by some automaton @xmath264 with the right number of states .
then @xmath295 is obtained from @xmath296 by replacing any subtree @xmath297 with the equivalent structure @xmath298 for @xmath299 .
hence , by prop .
[ p - ef ] , @xmath300 .
putting these two lemmas together , we obtain that , indeed , every consistent property of a multichain @xmath8 is witnessed by some multichain that can be accepted by some `` small '' automaton : [ p - multichain ] let @xmath269 be some multichain .
then there exists an automaton @xmath264 with at most @xmath301 many states ( where @xmath302 , @xmath277 ) such that @xmath266 is a multichain and @xmath303 .
now a result analogous to rabin s basis theorem follows immediately let @xmath1 be a @xmath10-structure , let @xmath23 be an @xmath6-formula in the language of the shelah - stupp - iteration @xmath67 with free variables @xmath140 and let @xmath304 be regular languages such that @xmath305 .
then @xmath306 where @xmath307 denotes that set quantification is restricted to regular multichains .
recall that rabin s basis theorem follows from his tree theorem whose proof , in turn , uses the effective complementation of rabin tree automata .
while the above theorem is an analogue of rabin s basis theorem , the proof is more direct and does in particular not rest on any complementation of automata .
the results of the previous section , as explained at the beginning , imply that quantification in an @xmath6-sentence can be restricted to regular sets that are accepted by `` small '' automata . in this section , we will use this insight to reduce the @xmath6-theory of the shelah - stupp - iteration to the @xmath2-theory of the base structure . fix some @xmath10-structure @xmath1 with universe @xmath49 , some finite set of states @xmath99 , some initial state @xmath309 , and some set of final states @xmath222 . then , for any automaton @xmath213 with @xmath310 , the language @xmath214 is a set in the shelah - stupp - iteration @xmath67 while its transition matrix is a tuple of finite sets in the base structure @xmath1 .
_ the idea of our reduction is that @xmath6-properties of the set @xmath214 in the shelah - stupp - iteration @xmath67 can ( effectively ) be translated into @xmath2-properties of the transition matrix @xmath311 in the base structure @xmath1 . _ in precisely this spirit
, the following lemma expresses simple properties of the automaton @xmath217 and of the language @xmath214 in terms of @xmath5-properties of @xmath312 .
[ l - simple - formulas ] let @xmath222 be finite sets and @xmath220 . there exist formulas @xmath313 for @xmath314 and @xmath315 of @xmath5 with free variables @xmath316 for @xmath314 such that for any @xmath10-structure @xmath1 and any automaton @xmath213 with transition matrix @xmath311 : 1 .
@xmath317 iff there exists a word @xmath318 with @xmath319 .
2 . @xmath320 iff @xmath214 is a multichain . the proof is based on the observation that ( 1 ) one only needs to search for a path of length at most @xmath321 and ( 2 ) that @xmath214 is a multichain iff no branching point belongs to some cycle .
so far , we showed that simple properties of @xmath214 are actually @xmath5- ( and therefore @xmath2- ) properties of the transition matrix of @xmath217 .
we now push this idea further and consider arbitrary @xmath6-properties of a tuple of languages @xmath322 .
in addition , it is convenient not to consider formulas as before , but to code individual variables by set variables . in the following proof
, we will therefore use other formulas for the logic @xmath6 : atomic formulas are @xmath323 , @xmath324 and @xmath325 for @xmath326 a constant symbol , @xmath327 set variables , and @xmath14 an @xmath15-ary relation symbol .
we define @xmath328 iff there are @xmath329 with @xmath330 .
quantification now ranges over all multichains in the shelah - stupp - iteration @xmath67 .
[ t - reduction ] there is an algorithm with the following specification + input : * @xmath181 , * finite sets @xmath331 and states @xmath332 for @xmath251 , * and a formula @xmath333 with free variables among @xmath174 in the language of the shelah - stupp - iteration @xmath67 .
output : a formula @xmath334 in the language of @xmath1 with free variables among @xmath335 for @xmath336 and @xmath251 with the following property : if @xmath1 is a @xmath10-structure and @xmath337 are automata with @xmath338 for @xmath251 , then @xmath339 the proof proceeds by induction on the construction of the formula @xmath333 , we only sketch the most interesting part @xmath340 .
set @xmath277 , @xmath341 , and @xmath342 .
let @xmath1 be a @xmath10-structure and let @xmath343 be automata with @xmath243 and transition matrix @xmath344 .
then , by prop
. [ p - multichain ] , @xmath345 iff there exists an automaton @xmath264 with @xmath151 states such that @xmath346 using the induction hypothesis on @xmath347 and @xmath348 , this is the case if and only if there exist finite sets @xmath349 for @xmath350=\{1,2,\dots , k\}$ ] such that * @xmath351 forms the transition matrix of some automaton with alphabet @xmath92 * for some @xmath352 $ ] , the automaton @xmath353,b,1,t^{\ell+1},f)$ ] * * accepts a multichain @xmath8 ( i.e. , @xmath354,1,f)}$ ] ) and * * this multichain satisfies @xmath347 ( i.e. , @xmath355),({\overline}\iota,1),({\overline}f , f))}$ ] ) .
since all these properties can be expressed in @xmath2 , the construction of @xmath334 is complete . as an immediate consequence
, we get a uniform version of shelah and stupp s theorem for the logics @xmath2 and @xmath6 : [ t - stupp - iteration ] finitary shelah - stupp - iteration is @xmath308-compatible .
@xmath356-compatibility of shelah - stupp - iteration @xcite can alternatively be shown along the same lines : one allows incomplete automata and proves an analogue of prop .
[ p - chain - quantification ] for the logic @xmath357
. then theorem [ t - stupp - iteration ] can be shown for the pair of logics @xmath356 .
our argument goes as follows : from a set @xmath359 , we construct a tree @xmath9
. the @xmath2-theory of this tree will be independent from @xmath8 and @xmath8 will be @xmath5-definable in the infinitary muchnik - iteration @xmath360 .
assuming @xmath358-compatibility of the infinitary muchnik - iteration , the set @xmath8 will be reduced uniformly to the @xmath2-theory of @xmath9 . for @xmath361 ,
this yields a contradiction .
a _ tree _ is a structure @xmath362 where @xmath35 is a partial order on @xmath363 such that , for any @xmath364 , @xmath365 is a finite linear order and @xmath366 for all @xmath364 .
we will consider the set @xmath367 of sequences in @xmath368 whose second components decrease .
this set , together with the prefix relation @xmath35 , forms a tree @xmath369 with root @xmath74 that we also denote @xmath7 .
nodes of the form @xmath370 are leaves of @xmath7 .
any inner node of @xmath7 has infinitely many children ( among them , there are infinitely many leaves ) . furthermore , all the branches of @xmath7 are finite . even more , if @xmath371 is a node different from the root , then the branches passing through @xmath371 have bounded length .
we will also consider the set @xmath372 where @xmath68 is an arbitrary symbol .
together with the prefix relation , this yields another tree @xmath373 that we denote @xmath374 . differently from @xmath7
, it has an infinite branch , namely the set of all nodes @xmath375 for @xmath193 . for two trees @xmath376 and
@xmath311 and a node @xmath180 of @xmath376 , let @xmath377 denote the tree obtained from the disjoint union of @xmath376 and @xmath311 by identifying @xmath180 with the root of @xmath311 ( i.e. , the node @xmath180 gets additional children , namely the children of the root in @xmath311 ) .
it is important for our later arguments that this operation transforms trees equivalent wrt .
@xmath378 into equivalent structures .
more precisely [ p - ef2 ] let @xmath376 , @xmath311 , and @xmath379 be trees and @xmath380 such that @xmath381 .
then @xmath382 for any node @xmath180 of @xmath376 . with @xmath383 , the set
@xmath384 together with the prefix relation and the root , is considered as a tree that we denote @xmath385 .
[ p - equivalent ] for any @xmath380 , we have @xmath386 . the statement is shown by induction on @xmath151 where the base case @xmath387 is trivial . to show @xmath388 , it suffices to prove for any formula @xmath389 of quantifier - depth at most @xmath151 @xmath390 assuming @xmath391 , there exist @xmath193 and @xmath392 finite with @xmath393 .
hence we have @xmath394 hence @xmath395 and therefore @xmath396 . using @xmath397
( see complete paper for the proof ) , we obtain @xmath398 .
conversely , one can argue similarly again using @xmath397 .
this proves that the existence of an infinite path can not be expressed in weak monadic second order logic since @xmath374 has such a path and @xmath7 does not .
using an idea from @xcite , the existence of an infinite path is a first - order property of the infinitary muchnik - iteration .
the following lemma pushes this idea a bit further : [ l - u ] let @xmath399 be a tree and let @xmath400 be the union of all infinite branches of @xmath311 .
then the @xmath2-theory of @xmath401 is uniformly reducible to the @xmath2-theory of the infinitary muchnik - iteration @xmath402 of the tree @xmath403 without the extra predicate . for @xmath359 ,
let @xmath404 and @xmath405 .
then @xmath9 is obtained from the linear order @xmath406 by attaching the tree @xmath374 to elements from @xmath8 and the tree @xmath7 to the remaining numbers .
[ t - complicated ] for @xmath359 , we have @xmath407 for all @xmath380 , and @xmath8 can be reduced to the @xmath5-theory of the infinitary muchnik - iteration @xmath408 . using ehrenfeucht - frass - games and prop .
[ p - equivalent ] , one obtains @xmath409 for the second statement , it suffices , by lemma [ l - u ] , to reduce @xmath8 to the first - order theory of @xmath410 where @xmath411 is the set of nodes of the tree @xmath9 that belong to some infinite branch .
if a transformation @xmath40 is @xmath358-compatible , then for any structure @xmath1 , the @xmath5-theory of @xmath44 can be reduced to the the @xmath2-theory of @xmath1 .
contrary to this , the above theorem states that the @xmath5-theory of the infinitary muchnik - iteration can be arbitrarily more complicated than the @xmath2-theory of the base structure .
hence we obtain infinitary muchnik - iteration is not @xmath358-compatible .
table [ tab : sum ] summarizes our knowledge about the compatibility of muchnik s and shelah & stupp s iteration .
it consists of four subtables dealing with finitary and infinitary muchnik - iteration and with finitary and infinitary shelah - stupp - iteration .
the sign + in cell @xmath46 of a subtable denotes that the respective iteration is @xmath46-compatible , denotes the opposite .
minus - signs without further marking hold since the base structure can be defined in any of its iterations .
capital letters denote references : ( a ) is @xcite , ( b ) @xcite , ( c ) ( * ? ? ? * prop .
3.4 ) , ( d ) ( * ? ? ?
* thm . 4.10 ) ,
( e ) theorem [ t - stupp - iteration ] , ( f ) theorem [ t - complicated ] , and ( g ) since the base structure is definable in its iteration and finiteness of a set is no mso - property .
small letters denote that the result follows from theorem [ t - fin - inf ] below and some further `` simple '' arguments from the result marked by the corresponding capital letter .
[ t - fin - inf ] let @xmath46 be any of the pairs of logics @xmath412 , @xmath413 , or @xmath414 .
there exists a computable function @xmath45 such that , for any @xmath10-structure @xmath1 , @xmath45 reduces the @xmath43-theory of @xmath137 to the @xmath415-theory of @xmath136 .
note the main difference between muchnik- and shelah - stupp - iteration : the latter is @xmath43-compatible for all relevant logics while only @xmath0 behaves that nicely with respect to ( infinitary ) muchnik - iteration a referee proposed to also consider the variant of mso where set quantification is restricted to countable sets . as to
whether muchnik iteration is compatible with this logic is not clear at the moment . | we investigate the relation between the theory of the iterations in the sense of shelah - stupp and of muchnik , resp . , and the theory of the base structure for several logics .
these logics are obtained from the restriction of set quantification in monadic second order logic to certain subsets like , e.g. , finite sets , chains , and finite unions of chains .
we show that these theories of the shelah - stupp iteration can be reduced to corresponding theories of the base structure .
this fails for muchnik s iteration .
dietrich kuske |
the authors thank dr joonsuk huh for helpful discussion and implementation of stochastic schrdinger equations and thomas markovich for the calculation of spectral densities .
t.f thanks prof .
yuji mochizuki for discussion on the electronic - structure calculations and prof .
kazuya watanabe and mr .
shunsuke tanaka for useful discussions on their experiments .
the majority of our computation was carried out on harvard university s odyssey cluster , supported by the research computing group of the fas division of science .
t.f acknowledge support from jsps grant - in - aid for scientific research on innovative areas `` dynamical ordering of biomolecular systems for creation of integrated functions '' ( grant no
. 25102002 ) .
t. f. , n. s. and a. a .-
acknowledge support from the center for excitonics , an energy frontier research center funded by the us department of energy , office of science and office of basic energy sciences under award de - sc0001088 .
the electronic hamiltonian in one - electron one - hole basis is given as follows : @xmath12 the parameters for this electronic hamiltonian were obtained based on the method by fujimoto @xcite .
state energies of localized frenkel and ct states can be calculated as follows : @xmath39 @xmath40 here , @xmath41 is a local excitation energy , @xmath42 is an electron affinity , and @xmath43 is an ionization potential . the electron affinities and ionization potentials are defined as orbital energies or energy difference between neutral and charged molecules . electronic interactions among frenkel states are described as coulomb interactions between transition densities of localized excitations of @xmath7th and @xmath9th molecules , @xmath44 those among frenkel and ct states are transfer integrals , @xmath45 in addition to the transfer integrals describing single electron or hole transfer steps , we adopt the homo - lumo amplitude of the excited state , @xmath46 , to take account of the sequential process of the localized excitation of @xmath7th molecule and subsequent electron ( hole ) transfer from @xmath7th to @xmath9th molecules @xcite .
we consider single electron or hole transfer steps for interactions among ct states : @xmath47 this term is nonzero unless @xmath48 or @xmath49 . due to the symmetry of the molecular crystal ,
we adopt same values for @xmath50 , @xmath42 , and @xmath43 .
the transfer integrals were considered for nearest - neighbor pairs as they are short - range interactions .
the long - range coulomb interactions , @xmath14 and @xmath15 , are approximated as sum of pairwise interactions of atomic point charges , @xmath51 @xmath52 where @xmath53 denotes a nuclear position , and @xmath54 are transition atomic charges@xcite , @xmath55 are electron atomic charges , and @xmath56 are hole atomic charges .
those atomic point charges were determined in such a way that they reproduce electrostatic potentials of transition ( @xmath54 ) , lumo ( @xmath55 ) , or homo densities ( @xmath56 ) .
@xmath50 , @xmath42 , and @xmath43 were calculated at the b3lyp / aug - cc - pvdz level embedded in the presence of external point charges of surrounding molecules .
these single - molecular properties are obtained by using q - chem @xcite .
the atomic charges describing the electrostatic potential of a dntt molecule were calculated by the resp method at the hartree - fock/6 - 31 g * level .
transition , electron , and hole charges are obtained by a multi - layer fragment molecular orbital method at the configuration interaction singles/6 - 31 g * level @xcite .
the transfer integrals were calculated by the projective method developed by kirkpatrick @xcite .
fragment molecular orbital calculations were performed by using abinit - mp @xcite .
spectral densities ( sds ) describe the frequency - dependent coupling strength between electron and phonon @xcite .
the coupling strength of ct states with a phonon bath would depend on the electron
hole distance : in small e
h separation , electron and hole interact together with a phonon bath as a frenkel state . on the other hand , in large e
h separation , the electron and hole interact with a phonon bath individually .
we need to treat @xmath57 state - specific sds for ct states , which is computationally expensive to calculate all of them from mixed quantum / classical simulations @xcite .
instead , we define an e
h distance dependent sd , @xmath58 , from frenkel ( @xmath59 ) , electron ( @xmath60 ) , and hole ( @xmath61 ) sds .
the sd of a ct state , the e
h distance of which is @xmath62 , is defined as @xmath58 .
the asymptotic behavior of @xmath58 is @xmath63 for @xmath64 and @xmath65 for @xmath66 .
we interpolate these limits with one decay parameter @xmath6 , assuming the @xmath67 dependence as follows , @xmath68 a 4 x 4 x 3 supercell was first created based on the experimentally determined dntt unit cell @xcite .
molecular dynamics simulations were carried out with a time step of 2 fs in the nvt ensemble for @xmath73 = 298 k with periodic boundary conditions .
other simulation conditions were same as that of ref @xcite except that the non - bonded interactions were cut off at 12 . after an equilibrium run of 5 ps ,
snapshots were collected at every 2 fs time steps along a production run of 10 ps .
tddft b3lyp/6 - 31 g * calculations were performed for those snapshots to obtain @xmath74 , @xmath75 and @xmath76 . @xmath75 and
@xmath76 were approximated as the homo and lumo kohn - sham orbital energies within koopmans theorem .
we adopt a superresolution method to extract sds from the autocorrelation functions @xcite .
electron and hole sds were scaled such that the reorganization energies of electron and hole became 101 and 65 mev @xcite , respectively .
the calculated sds for the frenkel state and charge - separated state ( @xmath77 ) are shown in fig .
the reorganization energy of the frenkel sd was calculated to be 53 mev .
the sds of ct states are defined with @xmath6 being 5 , which is close to the center of mass distance between nearest - neighbor molecules . | charge transfer states in organic semiconductors play crucial roles in processes such as singlet fission and exciton dissociation at donor / acceptor interfaces . recently , a time - resolved spectroscopy study of dinaphtho[2,3-b:2@xmath03@xmath0-f]thieno[3,2-b]-thiophene ( dntt ) thin films provided evidence for the formation of mixed frenkel and charge - transfer excitons after the photoexcitation . here
we investigate optical properties and excitation dynamics of the dntt thin films by combining ab initio calculations and a stochastic schrdinger equation .
our theory predicts that the low - energy frenkel exciton band consists of 8 to 47% ct character .
the quantum dynamics simulations show coherent dynamics of frenkel and ct states in 50 fs after the optical excitation .
we demonstrate the role of charge delocalization and localization in the mixing of ct states with frenkel excitons as well as the role of their decoherence .
organic semiconductors ( oscs ) are widely investigated as candidates for inexpensive and flexible materials for photovoltaics and other optoelectronic applications .
@xcite . in recent years
, new oscs have been designed and improved through computational modeling @xcite and virtual high - throughput screening @xcite .
modeling energy and charge transport properties is also essential for understanding structure - property relationships and thus for rationally designing novel oscs @xcite .
the low - energy optical excitations in oscs lead to the formation of a bound electron hole ( e h ) pair , a frenkel exciton .
excitonic properties of organic crystals are substantially different from those of isolated molecules , owing to excitonic couplings , near - field interactions between electronic transitions @xcite .
interactions with a charge - transfer ( ct ) state a state in which the electron and hole are located on spatially separated regions can also affect the optical properties of an exciton , as has been demonstrated by several theoretical studies @xcite .
experimentally , the degree of ct character has been studied by momentum - dependent electron - loss spectroscopy @xcite or electroabsorption spectroscopy @xcite .
ct states can act as precursors for interfacial ct states @xcite ; mixing of ct states with frenkel excitons would facilitate charge separation in photovoltaic materials .
they have also gained recent attention due to their relevance to singlet fission @xcite , in which singlet to triplet conversion can proceed via sequential ct steps @xcite .
here we focus on dinaphtho[2,3-b:2@xmath03@xmath0-f]thieno[3,2-b]-thiophene ( dntt ) , a @xmath1-type osc originally introduced by takimiya and co - workers @xcite .
dntt and its derivatives @xcite have gained attention due to their high hole mobility values and air stability . a recent time - resolved spectroscopy study by ishino et al .
@xcite concludes that mixed frenkel and ct excitons are formed after the optical excitation .
although the degree of ct character in excited states has been reported as described in the earlier paragraph , its role in excitation dynamics has remained unclear .
furthermore , there has been growing interest in the effects of excited - state delocalizations on charge photogeneration processes @xcite .
therefore , we studied the excited - state dynamics in dntt as a model system to provide insight into the role of the ct states in oscs . in this letter
, we present a theoretical study of optical properties and dissipative excitonic dynamics of dntt .
the optical absorption spectra and associated ct character are computed by introducing a tight - binding hamiltonian that treats frenkel and ct states on the same footing .
the excitation dynamics of the dntt thin films was simulated by a stochastic schrdinger equation with spectral densities derived from molecular dynamics simulations and excited state quantum chemistry calculations .
our simulations show coherent dynamics of the frenkel and ct excitons in 50 fs after optical excitation .
the importance of charge delocalization in the mixing of ct states will be discussed .
we adapt a tight - binding model of a system comprising of one electron and one hole in a molecular aggregate @xcite .
each molecule has two frontier orbitals : the highest occupied molecular orbital ( homo ) and the lowest unoccupied molecular orbital ( lumo ) .
a diabatic state of one electron and one hole , with localized wavefunctions , is obtained from creation operators as follows : @xmath2 , where @xmath3 is an electronic ground state of the molecular aggregate , the operators @xmath4 and @xmath5 create an electron of @xmath6 spin on @xmath7th molecule ( site ) and a hole of @xmath8 spin on @xmath9th molecule , respectively .
@xmath10 represents a localized frenkel state on @xmath7th molecule , while @xmath11 denotes a ct state of an electron and a hole being localized on @xmath7th and @xmath9th molecules , respectively .
the electronic hamiltonian in this one - electron one - hole basis can be written as follows : @xmath12 @xmath13 are composed of electron / hole transfer integrals for off - diagonal elements and electron affinities / ionization potentials for diagonals .
@xmath14 are excitonic and @xmath15 are e h coulomb interactions .
to calculate the electronic hamiltonian , we follow fujimoto @xcite and adapt that methodology to the fragment molecular orbital methods @xcite .
representative state energies and electronic couplings are shown in table i and ii .
the transfer integrals and excitonic interactions were calculated from fragment molecular orbital calculations at the hartree - fock/6 - 31 g * level .
the localized excitation energy , ionization potential , and electron affinity were obtained from density functional theory at the b3lyp / aug - cc - pvdz level .
further details of the electronic structure calculations are shown in the appendix a. .frenkel and ct energies in ev ( see figure 1(b ) for site labeling ) . [ cols="^,>,>,>",options="header " , ] + by diagonalizing the electronic hamiltonian , we get an adiabatic wave function , @xmath16 , where @xmath17 .
we obtained the absorption spectrum from the corresponding transition dipole moments , @xmath18 , where @xmath19 denotes a transition dipole moment of the @xmath7th molecule .
we use a 7 @xmath20 7 @xmath20 7 supercell containing 686 molecules for the model structure for the dntt crystal . each energy level
is broadened by a convolution of a lorentz function of one mev .
we also calculated ct populations of the @xmath21th adiabatic wave function , @xmath22 , which quantifies the probability that the electron and hole are found on different molecules .
figure 1(d ) shows the computed optical absorption spectrum of the dntt crystal in comparison with the measurement from ref@xcite .
the absorption spectrum is composed of the main excitation with the peak position of 2.83 ev , and relatively weak excitation at around 3.3 and 3.6 ev .
the lower energy bands consist of frenkel states partially mixed with the ct states .
the latter two weak bands have mostly ct character , the intensities of which are borrowed from the lower frenkel energy bands .
the ct character of the brightest state is 22% , and that of largest frenkel band is 47% .
the red edge of the main excitation has lower ct character due to the energy difference with the ct states .
our model predicts that the low energy exciton band has 8 to 47% of ct character .
the computed absorption spectrum is in reasonable agreement with the experiment , while the peak at around 2.94 ev can not be reproduced .
this may be ascribed to vibrational progressions @xcite .
however , it lies outside the scope of this letter to incorporate the vibrational degrees of freedom into the absorption spectra . to investigate the quantum dynamics of the frenkel
ct excitons in the dntt , we apply the stochastic schrdinger equation formulated by zhao and coworkers @xcite . in this formalism ,
time evolution of an excitonic wave function coupled to a phonon bath is obtained as follows : @xmath23\bigl| \psi(t ) \rangle , \label{sse}\end{aligned}\ ] ] where @xmath24 . here
effects of the phonon bath are incorporated through a stochastic force @xmath25 and a zero - temperature bath correlation function @xmath26 .
both of them are calculated from a spectral density @xmath27 for the @xmath28th diabatic state @xcite .
the spectral densities @xcite have been obtained by combing molecular dynamics simulations for the dntt crystal with excited - state calculations using time - dependent density functional theory at b3lyp/6 - 31 g * level . the details for calculating spectral densities are presented in the appendix b. an ensemble average of stochastic realization of eq ( 2 ) yields the two - body electron
hole density matrix , @xmath29 a system of 18 dntt molecules in the @xmath30 plane was taken as a model of thin films . as illustrative examples ,
localized frenkel ( @xmath31 ) and ct ( @xmath32 ) states are used for initial conditions for the simulations of dissipative quantum dynamics .
see figure 2(a ) and ( b ) for the structure and site labeling .
the state energies of the frenkel and ct states are 2.85 and 3.49 ev , respectively .
the runge - kutta method was used for numerical propagation with a time of 0.1 fs .
the excitonic density matrix and physical properties are averaged over 5,000 trajectories at the temperature of 300 k. to see dissipative dynamics of the 18 dntt molecular system , frenkel and ct populations , and electronic energies defined by @xmath33 are shown in figure 2(c ) and ( d ) .
thermal equilibration is achieved within 1.5 and 2.7 ps from the initial frenkel and ct states , respectively , as they are converged to the similar electronic energy of 2.75 ev . from the initial ct state ,
the ct populations follow almost the same dynamics as the electronic energies ; these plots reflect the relaxation dynamics from the higher ct state toward lower frenkel states . on the other hand , from the initial frenkel state , the ct populations
is first increased and then decreased toward the equilibration value of around 0.1 .
as shown in figure 2(d ) , the ct states are strongly mixed with the frenkel states in the initial 50 fs a mixed frenkel
ct exciton is formed . in what follows
, we investigate the ultrafast time scale mixing of the ct states .
in addition to the frenkel and ct populations , e h separation , inverse participation ratios ( iprs ) of the electron and hole density matrices , and coherence with the initial frenkel state are shown in figure 3 ( a)(d ) .
the e h separation of @xmath34 is defined as the center of mass distance between @xmath7th and @xmath9th molecules . to calculate the iprs , one - body electron or hole density matrices
were first calculated by tracing out the hole or electron basis : @xmath35 or @xmath36 .
the iprs of electron or hole density matrix is given by @xmath37 these measures describe the delocalization of the electron or hole density matrix @xcite . in figure 3(a ) , the ct populations are increased and show oscillatory behavior which implies coherence among the frenkel and ct states . the corresponding picture are also observed from the e h separation : the e h separation is first increased and subsequently decreased to the equilibrated value of around 0.5 . because the hole transfer integrals are higher than those for the electrons , there is an increase in the e h separation which is subsequently moderated by e
h coulomb attractions .
this competition between the higher hole transfer and e h coulomb attraction results in the oscillation .
the effect of higher hole transfer is also seen by comparing electron and hole iprs as represented in figure 3(c ) , indicating that the hole ipr is greater than that of electron .
the delocalization of the hole wave function is also confirmed by coherence among diabatic states , @xmath38 , as presented in figure 3(d ) .
this delocalization can effectively weaken the e h coulomb attraction and thus become an initial driving force for increasing e
h separation and the mixing of ct states .
localization of wave functions due to the nuclear vibrations works toward decreasing the e h separations .
our finding is in line with the observations by tamura and burghardt @xcite that charge separation at the donor / acceptor interface can be enhanced by charge delocalization . in the time - resolved spectroscopy experiment by ishino et al .
@xcite , the pump pulses excite to exciton manifolds at 3.1 ev , which leads to the optical excitation of the blue edge of the absorption maximum .
the subsequent relxation to the lowest exciton state occurs within about 2.1 ps .
our simulations shows that the relaxation time from the localized frenkel state , which is resonant to the absorption maximum , to the lowest exciton state is about 1.5 ps .
the relaxation time of 1.5 ps would qualitatively correspond to that observed by the experiments .
however , simulating more extended systems would be necessary for quantitative comparison between the simulation and experiments .
the mixing of ct states in optical excitations of the dntt has been proposed based on a derivative - like feature of transient absorption spectra @xcite : upon the excitation pulse of 3.1 ev the transition absorption spectra exhibit derivative - like features which are similar to stark spectra .
this result suggests the existence of transient charged species exerting electric fields on surrounding molecules .
it follows that the ct excitons are formed after photoexcitation , the local electric fields of which induce stark shifts of surrounding molecules . relating our simulations with the experiments , we argue that the derivative - like features would appear after frenkel ct decoherence .
if the optically - excited state is a superposition of frenkel and ct states , the absorption of them can not be disentangled .
in other words , stark effects from ct states occur after the frenkel - ct decoherence time .
our simulations predict that the derivative - like feature may appear about 50 fs after the optical excitation .
a more rigorous comparison to the experimental observations has to be done by directly calculating the time - resolved spectroscopy signals @xcite , which will be studied elsewhere .
we have presented a theoretical study on the optical properties and excitation dynamics of the dntt thin films .
the tight - binding hamiltonian combined with ab initio calculations reasonably reproduces the experimental absorption spectrum and predicts that the low - energy frenkel exciton band consist of 8 to 47% ct character .
the quantum dynamics simulations show the coherent dynamics of frenkel and ct states about 50 fs after photoexcitation .
the oscillation and lifetime of the e h separations are also analyzed in terms of charge delocalization / localization and the e
h coulomb attractions . combined with large - scale quantum dynamics simulations @xcite ,
the present approach can be applied to more extended systems such as organic bulk heterojunction solar cells .
such a device - level simulation will be useful to design and improve novel organic electronic materials . |
this is the second of a series of papers in which we try to classify four dimensional @xmath0 superconformal field theories ( scfts ) using classification of singularity .
this program has several interesting features : * the classification of field theory is reduced to the classification of singularities , which in many cases are much simpler than the classification using field theory tools . *
many highly non - trivial physical questions such as coulomb branch spectrum and the seiberg - witten solution @xcite can be easily found by studying the mini - versal deformation of the singularity . in @xcite , we conjecture that any three dimensional rational gorenstein graded isolated singularity should define a @xmath0 scft .
a complete list of hypersurface singularities was obtained in @xcite , and this immediately gives us a large number of new four dimensional @xmath0 scfts .
the natural next step is to classify three dimensional rational weighted homogeneous isolated complete intersection singularities ( icis ) . to our surprise
, the space of such singularities is also very rich , and we succeed in giving a complete classification .
let s summarize our major findings : * the number of polynomials defining icis is two , i.e. the singularity is defined as @xmath1 . * we find a total of 303 class of singularities , and some of them consist only finite number of models , but we do get many infinite sequences .
our classification gives many new interesting 4d @xmath0 scfts .
some of these singularities describe the familiar gauge theory , i.e. the singularity @xmath2 describes the affine @xmath3 quiver gauge theory with su type gauge group , see figure .
the major purpose of this paper is to describe the classification , and more detailed study of the corresponding scfts will appear in a different publication .
scft described by the singularity @xmath2 . here
the gauge group is @xmath4.,width=192 ] this paper is organized as follows : section two reviews the connection between the physics of 4d @xmath0 scft and the property of icis ; section iii reviews some preliminary facts about the general property of singularities ; section iv proves that the maximal embedding dimension of 3d rational icis is five , and section v classifies the general weighted homogeneous icis and we also compute the milnor number and the monomial basis of the mini - versal deformations .
four dimensional @xmath0 scft has an important @xmath5 r symmetry .
these theories have half bps scalar operators @xmath6 which can get expectation value and parameterize the coulomb branch , here @xmath7 is the @xmath8 charge and this operator is a singlet under @xmath9 symmetry .
the scaling dimension of operator @xmath6 is @xmath10 , and if @xmath11 , one can turn on the relevant deformation @xmath12 the scaling dimension of the coupling constant is determined by the relation @xmath13 + [ { \cal e}_{r,(0,0)}]=2 $ ] .
the coupling constants do not lift the coulomb branch , but will change the infrared physics , so we need to include those coupling constants besides the expectation values of @xmath6 to parameterize the coulomb branch .
we also have the dimension one mass parameters @xmath14 which can also change the ir physics , which should also be included as the parameters of coulomb branch . to solve the coulomb branch of a @xmath0 scft , we need to achieve the following two goals : * determine the set of rational numbers which include the scaling dimension of the coupling constants @xmath15 , coulomb branch operators @xmath6 , and mass parameters @xmath14 . @xmath16
since the scaling dimension of the coupling constant is paired with that of coulomb branch operator , this set is symmetric with respect to identity .
* once we find out the parameters on the coulomb branch , we want to write down a seiberg - witten curve which describes the low energy effective theory on the coulomb branch @xmath17 here @xmath18 includes all the parameters discussed above .
these two questions are central in understanding coulomb branch of a @xmath0 scft , and in general are quite hard to answer .
if our 4d scft is engineered using 3-fold singularity , the above two questions can be found from the miniversal deformation of the singularity .
the formulae for the hypersurface case has been given in @xcite . in the following , we will review the relevant formulas for icis case , which is first derived in @xcite .
consider a three dimensional icis defined by two polynomials @xmath19 , where@xmath20 is the map @xmath21 .
we require the defining polynomials to have a manifest @xmath22 action , which is proportional to the @xmath8 symmetry of the field theory .
we normalize the @xmath22 action so that the weights of the coordinates @xmath23 are @xmath24 , and the degree of @xmath25 is one , the degree of @xmath26 is d : @xmath27 this singularity has a distinguished @xmath28 form : @xmath29 which has charge @xmath30 under the @xmath22 action . to define a sensible 4d scft
, we require this charge to be positive , which means that @xmath31 we conjecture that this condition is necessary and sufficient to define a scft .
such singularity is called rational singularity , see section iii for the definition .
the sw solution is described by the mini - versal deformation of the singularity : @xmath32 here @xmath33 is the monomial basis of the jacobi module of @xmath20 , and @xmath34 is the milnor number . the coefficient @xmath35 is identified with the parameters on coulomb branch .
the scaling dimension of @xmath35 is determined by requiring @xmath36 to have dimension one as its integration over the middle homology cycle of milnor fibration gives the mass of bps particle , and we have @xmath37:~~[\lambda_{\alpha}]={1-q_{\alpha}\over \sum w_i-1-d } , \nonumber\\ & \phi_{\alpha}=[0,\phi_{\alpha}]:~~[\lambda_{\alpha}]={d - q_{\alpha}\over \sum w_i-1-d}. \label{spectrum}\end{aligned}\ ] ] here @xmath38 is the @xmath22 charge of the monomial @xmath33 .
the spectrum is classified into following categories : * coulomb branch operator @xmath6 if @xmath39>1 $ ] . * mass parameters if @xmath39=1 $ ] .
* coupling constants for relevant deformations if @xmath40<1 $ ] .
* exact marginal deformations if @xmath39=0 $ ] .
these deformations are related to the moduli of the singularity . *
irrelevant deformations if @xmath39<0 $ ] .
the spectrum is paired and is symmetric with respect to one , which is in perfect agreement with the field theory expectation .
* example * : consider the singularity @xmath41 .
the weights and degrees of two polynomials are @xmath42 .
the jacobi module @xmath43 has the basis @xmath44,~\phi_2=[0,z_4
^ 2],~\phi_3=[0,z_5],~\phi_4=[0,z_4],~\phi_5=[0,z_3 ] , \nonumber\\ & \phi_6=[0,z_2],~\phi_7=[0,z_1],~\phi_8=[0,1],~\phi_9=[1,0].\end{aligned}\ ] ] using the formula ( [ spectrum ] ) , we find the scaling dimension of the coefficients @xmath45 in mini - versal deformation : @xmath46=[\lambda_2]=0,~[\lambda_3]=[\lambda_4]=[\lambda_5]=[\lambda_6]=[\lambda_7]=1,~[\lambda_8]=[\lambda_9]=2;\ ] ] so this theory has two exact marginal deformations , two coulomb branch operators with dimension 2 , and five mass parameters .
the corresponding gauge theory is depicted in figure .
[ d5 ] with @xmath47 . in the following sections
, we will classify all possible 3 dimensional weighted homogeneous icis , and describe the miniversal deformation .
the coulomb branch is then solved using formulas [ sw ] and [ spectrum ] .
in this section , we recall some definitions and known results about the gorenstein singularity . for a commutative noetherian local ring @xmath48 , the depth of @xmath48 ( the maximum length of a regular sequence in the maximal ideal of @xmath48 ) is at most the krull dimension of @xmath48 .
the ring @xmath48 is called cohen - macaulay if its depth is equal to its dimension . more generally ,
a commutative ring is called cohen - macaulay if it is noetherian and all of its localizations at prime ideals are cohen - macaulay . in geometric terms , a scheme is called cohen - macaulay if it is locally noetherian and its local ring at every point is cohen - macaulay .
let @xmath49 be an isolated singularity of dimension @xmath50 .
@xmath49 is said to be normal or cohen - macaulay if the local ring @xmath51 has such a property .
let @xmath49 be an isolated singularity of dimension @xmath50 .
then we have the following propositions .
( corollary 3.10 , @xcite ) @xmath49 is called cohen - macaulay iff @xmath52 .
@xmath49 is normal iff @xmath53 .
let @xmath49 be an isolated singularity of dimension @xmath50 and @xmath54 be a resolution of @xmath49 .
then @xmath55 following laufer @xcite , we consider the sheaf cohomology with support at infinity .
the following sequence is exact : @xmath56 by serre duality , @xmath57 by grauert - riemenschneider vanishing theorem , we have @xmath58 it follows that @xmath59 on the other hand , we also have the following exact sequence : @xmath60 since @xmath61 .
thus we have @xmath62 take a 1-convex exhaustion function @xmath63 on @xmath64 such that @xmath65 on @xmath64 and @xmath66 if and only if @xmath67 where @xmath68 is the irreducible component of @xmath69 .
put @xmath70 .
then by laufer @xcite , @xmath71 on the other hand , by andreoti and grauert ( thorme 15 of @xcite ) , @xmath72 is isomorphic to @xmath73 for @xmath74 and @xmath75 is injective .
thus we have @xmath76 combining with ( [ 2.1]),([2.2 ] ) and ( [ 2.3 ] ) we have @xmath77 this completes the proof .
\(1 ) @xmath49 is cohen - macaulay @xmath78 . + ( 2 )
@xmath49 is normal and @xmath79 is cohen - macaulay .
a normal variety @xmath80 is called gorenstein if it is cohen - macaulay and the sheaf @xmath81 is locally free .
a gorenstein point @xmath82 of an @xmath50-dimensional variety @xmath80 is rational ( respectively minimally elliptic ) if for a resolution @xmath83 we have @xmath84(respectively @xmath85 , where @xmath86 is the ideal of @xmath87 ) .
( this is equivalent via duality to the cohomological assertion @xmath88 ( respectively , is a 1-dimensional @xmath89-vector space at @xmath87 ) .
it is convenient to make intrinsic ( and generalize slightly ) the notion of `` a general hyperplane section through @xmath87 '' : let @xmath90 be the local ring of a point @xmath82 of a @xmath89- scheme , and let @xmath91 be a finite - dimensional @xmath89-vector space which maps onto @xmath92 ( equivalently , by nakayama s lemma , @xmath93 generates the @xmath51-ideal @xmath86 ) ; by a general hyperplane section through @xmath87 is mean the sub - scheme @xmath94 defined in a suitable neighborhood @xmath95 of @xmath87 by the ideal @xmath96 , where @xmath97 is a sufficiently general element ( that is , @xmath98 is a @xmath89-point of a certain dense zariski open @xmath99 ) .
( @xcite , theorem 2.6)[mth ] if @xmath82 is a rational gorenstein point ( dim @xmath100 .
then for a general hyperplane section @xmath101 through @xmath87 , @xmath102 is minimally elliptic or rational gorenstein .
suppose that @xmath101 runs through any linear system of sections @xmath103 whose equations generate the maximal ideal @xmath86 of @xmath51 .
then a general element @xmath101 of this linear system is normal .
let @xmath104 be any resolution of @xmath80 which dominates the blow - up of the maximal ideal @xmath86 ; by definition of the blow - up , the scheme - theoretic fiber over @xmath87 is an effective divisor @xmath69 such that @xmath105 hence @xmath106 , where @xmath107 runs through a free linear system on @xmath108 . by bertini
s theorem , @xmath109 is a resolution of @xmath101 .
now we use the adjunction formula to compare @xmath110 and @xmath111 . in the diagram ( figure 1 ) , @xmath112^{f } & \supset & t \ar@{->}[d]^{\varphi}+e \\ x&\supset & s } \ ] ] we have @xmath113 and @xmath114 so that @xmath115 and @xmath116 this just means that any @xmath117 has at worst @xmath118 as pole on @xmath107 . on the other hand ,
since the maximal ideal @xmath119 is the restriction to @xmath101 of the maximal ideal @xmath120 , it follows that every element of @xmath121 vanishes along @xmath122 .
hence every element of @xmath123 is regular on @xmath107 , that is @xmath124 thus @xmath125 implies @xmath102 is minimally elliptic and @xmath126 implies @xmath102 is rational .
@xcite [ lth ] let @xmath87 be a minimally elliptic singularity .
let @xmath127 be a resolution of a stein neighborhood @xmath93 of @xmath87 with @xmath87 as its only singular point .
let @xmath128 be the maximal ideal in @xmath129 .
let @xmath130 be the fundamental cycle on @xmath131 .
* if @xmath132 , then @xmath133 on @xmath69 . *
if @xmath134 , and @xmath135 is the minimal resolution or the minimal resolution with non - singular @xmath68 and normal crossings , @xmath136 is the structure sheaf for an embedded point . *
if @xmath134 or @xmath137 , then @xmath87 is a double point .
* if @xmath138 , then for all integers @xmath139 , @xmath140 and @xmath141 .
* if @xmath142 , then @xmath87 is a hypersurface singularity . * if @xmath143 , then @xmath87 is a complete intersection and in fact a tangential complete intersection . * if @xmath144 , then @xmath87 is not a complete intersection .
let @xmath145 be an isolated singularity with dimension @xmath50 , a deformation of @xmath145 will be simply a realization of @xmath145 as the fiber of a map - germ between complex manifolds whose dimensions differ by @xmath50 . to be precise
, it consists of holomorphic map - germ @xmath146 between complex manifold germs with dim@xmath49-dim@xmath147 and an isomorphism @xmath148 of @xmath145 onto the fiber @xmath149 of @xmath20 .
a morphism from a deformation @xmath150 to another @xmath151 is a pair of map - germs @xmath152 such that the diagram @xmath153^{f ' } & \stackrel{\tilde{g}}{\longrightarrow } & ( x , x ) \ar@{->}[d]^{f } \\ ( s ' , o')&\stackrel{g}{\longrightarrow } & ( s , o ) } \ ] ] + is cartesian and @xmath154 .
we say that a deformation @xmath151 of @xmath145 is versal if for any deformation @xmath150 of @xmath145 there exists a morphism @xmath152 from @xmath150 to @xmath151 .
notice that we do not require this morphism to be unique in any sense .
if , however , the derivative of @xmath155 in @xmath156 , @xmath157 is unique , then we say that @xmath151 is miniversal .
( @xcite ( 2.10 ) ) let @xmath158 define an icis at the origin and has dimension @xmath50 .
a miniversal deformation of @xmath159 can be taken in the form @xmath160 where @xmath161 are the representative of a basis of the linear space : @xmath162 here @xmath163 is the tyurina number and is equal to the milnor number @xmath34 if @xmath20 is weighted homogeneous .
in this section we shall prove the following conjecture for homogeneous isolated complete intersection singularity ( icis ) in theorem [ th1 ] and three dimensional isolated complete intersection singularity in theorem [ t1.2 ] .
we shall also give a classification of three - dimensional rational homogeneous isolated complete intersection singularities in theorems [ th2 ] and [ th3 ] .
let @xmath164 be the dimension of rational isolated complete intersection singularity with @xmath22-action .
the the embedding dimension of the singularity is at most @xmath165 .
let @xmath166 be the analytic germ of an @xmath50-dimensional complex homogeneous isolated complete intersection singularity .
let @xmath167 be a resolution of singularity of dimension @xmath50 with exceptional set @xmath168 .
the geometric genus @xmath169 of the singularity @xmath170 is the dimension of @xmath171 and is independent of the resolution @xmath172 .
we have the following proposition .
@xcite[prop1 ] let @xmath173 be a homogeneous isolated complete intersetion singularity of multidegree @xmath174 and dimension @xmath50 , that is deg@xmath175 , then @xmath176 where @xmath177 we prove that the above conjectue is true in homogeneous case .
[ th1 ] let @xmath178 be a homogeneous rational isolated complete intersetion singularity of multidegree @xmath174 and dimension @xmath50 , then @xmath179 ( i.e. @xmath180 ) . since @xmath181 is a homogeneous isolated rational complete intersetion singularity , so by proposition [ prop1 ] , we have @xmath182 thus for any @xmath183 , we have @xmath184 . if we assume the contrary , @xmath185 , without loss of generality , we consider thel @xmath186 with @xmath187 .
then for this choice @xmath188 , we have @xmath189 since @xmath190 .
this contradicts with @xmath184 .
therefore we have @xmath179 .
we have the following two classification theorems for homogeneous case . [ th2 ]
let @xmath178 be a three dimensional homogeneous rational isolated complete intersection singularity of multidegree @xmath174 which is not a hypersurface singularity , then @xmath191 and @xmath192 .
if follows from theorem [ th1 ] that @xmath193 . since @xmath181 is not a hypersurface singularity , so @xmath194
let @xmath195 , we have @xmath196 or @xmath197 . for @xmath198 , by proposition [ prop1 ] @xmath199 we have @xmath200 which implies @xmath201 since @xmath202 and @xmath203 are at least 2 .
similarly , @xmath204 implies @xmath205 .
thus @xmath192 .
[ th3 ] let @xmath206 be a three dimensional homogeneous rational isolated hypersurface singularity of degree @xmath207 , then @xmath208 . by proposition [ prop1 ] ,
we have @xmath209 so we @xmath210 or @xmath211 .
we can also prove the conjecture for @xmath212 .
[ l1.1 ] let @xmath170 be a @xmath50-dimensional isolated singularity in @xmath213 . if @xmath214 , then @xmath170 can not have two components of dimension @xmath50 .
if @xmath93 is union of two components @xmath215 and @xmath216 , each of which is of dimension @xmath50 , then @xmath215 and @xmath216 will intersect with at least dimension @xmath214 .
this is a contradiction since @xmath93 is singular along the intersection .
[ t1.2 ] let @xmath170 be a three dimensional rational isolated complete intersection singularity .
then the embedding dimension of @xmath181 is at most 5 .
take a generic section @xmath217 of @xmath170 .
then by theorem [ mth ] , @xmath217 is either a 2-dimensional rational gorenstein singularity or minimally elliptic singularity .
it is well - known that 2-dimensional rational gorenstein singularity must be rational double points .
so the embedding dimension of @xmath217 is 3 . on the other hand , by theorem [ lth ]
asserts that minimally elliptic complete intersection isolated singularity has embedding dimension at most 4 .
so the embedding dimension of @xmath170 is at most 5 .
an immediate corollary is as follows .
[ c1.1 ] the three dimensional weighted homogeneous rational isolated complete intersection singularity which is not hypersurface singularity is defined by two weighted homogeneous polynomials in 5 variables . both @xmath25 and @xmath26 in corollary [ c1.1 ] are irreducible .
assume that @xmath218 and @xmath219,where @xmath220 and @xmath221 are irreducible .
@xmath222 are @xmath223 irreducible components of dimension 3 . since @xmath224 , so @xmath225 . by lemma [ l1.1 ]
, we have @xmath226 , which implies @xmath227 , thus both @xmath25 and @xmath26 are irreducible .
in this section , we shall give an complete classification of three - dimensional rational weighted homogeneous complete intersection singularities .
we first recall some definitions and then we prove some properties which are used in the proof of classification theorem .
let @xmath228 be an @xmath229-tuple of positive rational numbers .
a polynomial @xmath230 is said to be a weighted homogeneous polynomial with weights @xmath231 if each monomial @xmath232 of @xmath20 satisfies @xmath233 .
and we say a pair of polynomials @xmath234 are weighted homogeneous of type @xmath235 if @xmath25 is weighted homogeneous of type @xmath236 and @xmath26 is weighted homogeneous of type @xmath237 . by corollary [ c1.1 ] , in order to classify three dimensional weighted homogeneous rational isolated complete intersection singularity , we only need to study weighted homogeneous polynomials in 5 variables . by theorem [ ma],we
have @xmath246 .
thus @xmath80 is rational if and only if @xmath247 it is easy to see that @xmath248 implies @xmath249 , it follows that @xmath80 is rational . if @xmath80 is rational , and without lose of generality , we assume that @xmath250 , then @xmath251 is not empty
. let @xmath252 such that @xmath253 for any @xmath254 , where @xmath255 for @xmath256 .
then let @xmath257 so we have @xmath258 for any @xmath259 and @xmath260 .
it is easy to seen that there exist @xmath261 such that @xmath262 , because if not then @xmath263 . without lose of generality , we may assume that @xmath264 .
let @xmath265 , then @xmath266 .
notice that @xmath267 and @xmath268 , thus we have @xmath269 .
it follows that @xmath270 .
it contradicts with @xmath80 is rational , so we conclude that @xmath271 .
similarly we can prove that @xmath272 .
so @xmath273 which implies @xmath248 .
[ l3 ] let @xmath243 be a three dimensional weighted homogeneous icis of type @xmath274
. then we have + ( 1 ) for any @xmath275 , one of the following cases occurs : + ( 1a ) @xmath276 appears in @xmath25 for some n , + ( 1b ) @xmath276 appears in @xmath26 for some n , + ( 1c ) there exist @xmath277 such that @xmath278 appears in @xmath25 for some @xmath50 and @xmath279 appears in @xmath26 for some @xmath128 .
+ + ( 2 ) for any @xmath280 and any @xmath281 , one of the following cases occurs : + ( 2a ) @xmath282 appears in @xmath283 , for some non - negative integers @xmath284 , + ( 2b ) there exists @xmath285 such that @xmath286 appears in @xmath283 , for some non - negative integers @xmath284 .
+ + ( 3 ) for any @xmath281 , one of the following cases occurs : + ( 3a ) @xmath282 appears in @xmath25 for some non - negative integer @xmath284 , + ( 3b ) @xmath282 appears in @xmath26 for some non - negative integer @xmath284 , + ( 3c ) there exist @xmath287 and @xmath288 such that @xmath289 appear in @xmath25 for some non - negative integers @xmath290 and + @xmath291 appear in @xmath26 for some non - negative integers @xmath292 .
+ + ( 4 ) for any @xmath280 and any @xmath293 , let @xmath294 , then one of the following cases occurs : + ( 4a ) @xmath295 appears in @xmath283 , for some non - negative integers @xmath296 , + ( 4b ) @xmath297 and @xmath298 appear in @xmath283 , for some non - negative integers @xmath299 .
+ + ( 5 ) for any @xmath300 , there exists @xmath301 such that @xmath302 appears in @xmath283 for some non - negative integers @xmath296 .
( 1 ) without lose of generality , we may assume that @xmath303 .
assume on the contrary that neither of @xmath304 and @xmath305 occurs .
then @xmath306 does not appear in @xmath283 for any @xmath280 and integer @xmath50 , so we have @xmath307 when @xmath308 . and for any @xmath309 , we have @xmath310 does nt appear in @xmath25 for any non - negative integer @xmath311 or @xmath312 does nt appear in @xmath26 for any non - negative integer @xmath313 .
it follows that @xmath314 or @xmath315 when @xmath308 .
similarly we have @xmath316 or @xmath317 when @xmath308 .
thus we have @xmath318 which implies @xmath319 and @xmath320 are linear dependent .
thus @xmath321 , which has dimension one , is contained in the singular locus of @xmath80 .
this contradicts with @xmath80 has an isolated singularity .
+ + ( 2 ) we may assume that @xmath322 and @xmath323 .
assume on the contrary that neither of @xmath324 and @xmath325 occurs , then @xmath326 does not appear in @xmath25 , for any non - negative integers @xmath284 and @xmath327 does not appear in @xmath25 . and for any @xmath328 and for any non - negative integers @xmath284 , we have @xmath329 and @xmath330 when @xmath331 .
thus @xmath332 , which has dimension at least one , is contained in the singular locus of @xmath80 .
this contradicts with @xmath80 has an isolated singularity .
+ + ( 3 ) we may assume that @xmath333 .
assume on the contrary that neither of @xmath334 and @xmath335 occurs , then one of the following two cases occurs : + * subcase ( a ) * @xmath336 does not appear in @xmath337 for any @xmath338 and any non - negative integer @xmath284 , and there exist @xmath339 such that both @xmath340 and @xmath341 do not appear in @xmath283 for any non - negative integer @xmath290 . +
* subcase ( b ) * @xmath336 does not appear in @xmath337 for any @xmath338 and any non - negative integer @xmath284 , and there exists @xmath328 such that @xmath342 does not appear in @xmath337 for any @xmath338 and any non - negative integer @xmath343 .
+ if * subcase ( a ) * occurs , without lose of generality , we may assume that @xmath322 and @xmath344 , then @xmath345 when @xmath331 .
thus @xmath346 , which has dimension at least one , is contained in the singular locus of @xmath80 .
this contradicts with @xmath80 has an isolated singularity .
+ if * subcase ( b ) * occurs , without lose of generality , we may assume @xmath347 . then when @xmath331 , we have @xmath1 , @xmath348 and @xmath349 .
thus @xmath350 , which has dimension at least one , is contained in the singular locus of @xmath80 .
this contradicts with @xmath80 has an isolated singularity .
+ + ( 4 ) we may assume that @xmath351 and @xmath322 .
assume on the contrary that neither of @xmath352 and @xmath353 occurs , thus @xmath354 does not appear in @xmath25 for any non - negative integers @xmath296 and there exists @xmath355 such that @xmath356 does not appear in @xmath25 for any non - negative integers @xmath296 .
without lose of generality , we may assume that @xmath357 .
then @xmath358 when @xmath359 .
thus @xmath360 , which has dimension at least one , is contained in the singular locus of @xmath80 .
this contradicts with @xmath80 has an isolated singularity .
+ + ( 5 ) we may assume that @xmath351 .
assume on the contrary that @xmath354 does not appear in @xmath283 for any @xmath280 and any non - negative integers @xmath296 .
then when @xmath359 we have @xmath1 and @xmath348 and @xmath349 .
thus @xmath361 , which has dimension at least two , is contained in the singular locus of @xmath80 .
this contradicts with @xmath80 has an isolated singularity .
[ c2 ] let @xmath243 be a weighted homogeneous icis of type + @xmath244 , then we have : + ( 1 ) for any @xmath261 , we have @xmath366 or @xmath367 or there exist @xmath368 such that @xmath369 and @xmath370 . + ( 2 ) for any @xmath281 and any @xmath371 , we have @xmath372 or there exists @xmath373 such that @xmath374 .
+ ( 3 ) for any @xmath281 , if @xmath375 , then there exist @xmath287 and @xmath376 such that @xmath377 , @xmath378 and @xmath379 , @xmath380 .
+ ( 4 ) for any @xmath381 and any @xmath371 , let @xmath294 , if @xmath382 , then we have @xmath383 .
+ ( 5 ) for any @xmath381 , we have @xmath384 or @xmath385 .
let @xmath243 be a weighted homogeneous icis of type @xmath386 , + with @xmath387 .
then @xmath234 has the same type as one of the following weight homogeneous singularities in the list up to permutation of coordinates .
we also list the milnor number @xmath34 and the vector basis of the miniversal deformation of the singularities in the list . in order to save space
, we only list the set of maximum elements ( i.e. * mini * ) of the vector basis of the corresponding singularity .
that is , @xmath388 \mid a\in \textbf{m},\ ; \exists \ ; [ b,0 ] \in
\textbf{mini}~s.t.~b \ge a\ } \cup \{[0,a ] \mid a\in \textbf{m},\ ; \exists \ ; [ 0,b ] \in \textbf{mini}~s.t.~b \ge a\}$ ] form a basis of miniversal deformation where @xmath389 ideal @xmath390 . a monomial @xmath391 if and only if @xmath392 for @xmath393 .
for example , if @xmath394,[z_1z_2 ^ 2,0],[0,z_3z_4]\}$ ] , then @xmath395,[1,0],[z_1,0 ] , [ z_1 ^ 2,0 ] , [ z_1 ^ 3,0 ] , [ z_2,0 ] , [ z_2 ^ 2,0 ] , [ z_1z_2,0 ] , [ z_1z_2 ^ 2,0 ] , [ 0,z_3 ] , [ 0,z_4 ] , [ 0,z_3z_4]\}\ ] ] form a vector basis of miniversal deformation .
mini=\{[z_1,0],[0,z_4z_5 ^ -2 + 2 n],[0,z_3z_5 ^ -3 + 2 n],[0,z_2z_5 ^ -3 + 2 n],[0,z_3z_4z_5 ^ -5 + 2 n],[0,z_3 ^ 2z_5 ^ -5 + 2 n],[0,z_4 ^ 2z_5 ^ -1 +
n ] , + [ 0,z_3 ^ -3 + nz_4z_5 ^ -2 + n],[0,z_2z_4z_5 ^ -2 + n],[0,z_3 ^ -2 + nz_5 ^ -2 + n],[0,z_2 ^ 2z_5 ^ -2 + n],[0,z_3 ^ -2 + nz_4z_5],[0,z_3 ^ -1 + nz_5 ] } mini=\{[z_1,0],[0,z_5 ^ 4],[0,z_4z_5 ^ 3],[0,z_2z_5 ^ 3],[0,z_3 ^ 2z_4 ^ 2z_5 ^ 2],[0,z_3 ^ 3z_4z_5 ^ 2],[0,z_3 ^ 4z_5 ^ 2],[0,z_3 ^ 3z_4 ^ 2z_5],[0,z_3 ^ 7z_4z_5 ] , + [ 0,z_2z_4z_5],[0,z_2z_4 ^ 3],[0,z_3 ^ 4z_4 ^ 2],[0,z_3 ^ 8],[0,z_2 ^ 2 ] } mini=\{[z_1,0],[0,z_3z_4z_5 ^ 3],[0,z_3 ^ 2z_5 ^ 3],[0,z_2z_5 ^ 3],[0,z_3 ^ 4z_4z_5 ^ 2],[0,z_3 ^ 5z_5 ^ 2],[0,z_2 ^ 2z_5 ^ 2],[0,z_3 ^ 4z_4 ^ 2z_5 ] , + [ 0,z_3 ^ 7z_4z_5],[0,z_3 ^ 8z_5],[0,z_4 ^ 3],[0,z_3 ^ 5z_4 ^ 2],[0,z_3 ^ 10z_4],[0,z_2z_4 ] } mini=\{[z_1,0],[0,z_3z_5 ^ 4],[0,z_4z_5 ^ 3],[0,z_3 ^ 3z_5 ^ 3],[0,z_3 ^ 3z_4z_5 ^ 2],[0,z_3 ^ 6z_5 ^ 2],[0,z_2z_5 ^ 2],[0,z_3 ^ 2z_4 ^ 2z_5 ] , + [ 0,z_3 ^ 5z_4z_5],[0,z_3 ^ 9z_5],[0,z_4 ^ 3],[0,z_3 ^ 3z_4 ^ 2],[0,z_3 ^ 7z_4],[0,z_2z_4],[0,z_2 ^ 2 ] } mini=\{[z_1,0],[0,z_5 ^ 5],[0,z_3 ^ 2z_5 ^ 4],[0,z_4z_5 ^ 3],[0,z_3 ^ 4z_5 ^ 3],[0,z_2z_5 ^ 3],[0,z_3 ^ 2z_4z_5 ^ 2],[0,z_3 ^ 6z_5 ^ 2],[0,z_3 ^ 3z_4z_5 ] , + [ 0,z_2z_4z_5],[0,z_3 ^ 7z_5],[0,z_3z_4 ^ 2],[0,z_3 ^ 6z_4],[0,z_3 ^ 10],[0,z_2 ^ 2 ] } mini=\{[z_1,0],[0,z_3 ^
2z_5 ^ 5],[0,z_4z_5 ^ 4],[0,z_3 ^ 5z_5 ^ 4],[0,z_3z_4z_5 ^ 3],[0,z_3 ^ 7z_5 ^ 3],[0,z_2z_5 ^ 3],[0,z_3 ^ 4z_4z_5 ^ 2],[0,z_3 ^ 10z_5 ^ 2 ] , + [ 0,z_3 ^ 6z_4z_5],[0,z_2z_4z_5],[0,z_3 ^ 11z_5],[0,z_3 ^ 2z_4 ^ 2],[0,z_3 ^ 9z_4],[0,z_3 ^ 15],[0,z_2 ^ 2 ] } it is easy to check that each singularities defined by pairs of polynomials in the list above are thee dimensional isolated rational complete intersection singularities . by lemma [ l3 ] ( 1 ) , we know that for any @xmath275 , one of the following cases occurs : + ( 1a ) @xmath1259 appears in @xmath25 for some @xmath1260 , + ( 1b ) @xmath1259 appears in @xmath26 for some @xmath1260 , + ( 1c ) there exist @xmath1261 such that @xmath1262 appears in @xmath25 for some @xmath1260 and @xmath1263 appears in @xmath26 for some @xmath14 .
for each @xmath261 , if one of ( 1a ) , ( 1b ) and ( 1c ) occurs , then there are @xmath1264 cases . if ( 1a ) or ( 1b ) occurs , one monomial which appear in @xmath25 and @xmath26 can be determined . and
if ( 1c ) occurs , then two monomials which appear in @xmath25 and @xmath26 can be determined .
now we consider the following two cases : + ( i ) there exists @xmath261 such that ( 1c ) occurs . therefore more than 6 monomials in @xmath25 and @xmath26 are determined .
thus we get more than 6 equations of @xmath1265 ( for instance , if we have @xmath1266 appears in @xmath26 , then we have @xmath1267 ) .
so @xmath1268 is uniquely determined by solving these 6 linear equations .
and we have checked that each weight type @xmath386 obtained by this way , which satisfies the rational condition @xmath1269 and the conditions listed in corollary [ c2 ] , is the same as one of the weight types of the singularities in the list up to permutation of coordinates .
more explicitly , for example , we treat the case that @xmath1270 appear in @xmath25 and @xmath1271 appears in @xmath26 .
then we can get @xmath1272 by solving the 6 corresponding linear equations . without lose of generality
, we may assume that @xmath1273 .
since we have @xmath1269 and @xmath387 , so we conclude that @xmath1274 can only be one of the following cases : + ( 1 ) @xmath1275 + ( 2 ) @xmath1276 + ( 3 ) @xmath1277 + ( 4 ) @xmath1278 + ( 5 ) @xmath1279 + ( 6 ) @xmath1280 + ( 7 ) @xmath1281 + ( 8) @xmath1282 + ( 9 ) @xmath1283 .
+ then we only consider the infinite cases ( 1 ) , ( 5 ) and ( 7 ) .
the other finite cases can be checked easily .
+ for infinite case ( 1 ) , we have @xmath1284 by corollary [ c2 ] ( 2 ) , we have @xmath1285 or there exists @xmath1286 such that @xmath1287 , it follows that one of following cases occurs : + ( i ) . @xmath1288 + ( ii )
. @xmath1289 + ( iii ) .
@xmath1290 + ( iv ) .
@xmath1291 .
+ if ( iii ) @xmath1290 occurs , since @xmath1300 , we have @xmath1305 .
thus we have @xmath1306 or 1 . if @xmath1307 , notice that @xmath387 and @xmath1300 , thus we have @xmath1308 .
if @xmath1309 , we have @xmath1310 .
+ in conclusion , we have @xmath1311 or @xmath1312 or @xmath1313 .
so @xmath1314 is same as one of the following cases : + case ( 1 ) @xmath1315 + case ( 2 ) @xmath1316 + case ( 3 ) @xmath1317 . +
and we can see that case ( 1 ) ; case ( 2 ) and case ( 3 ) corresponds to the 1st @xmath1318 , 59th ; 1th ; and 55th , 60th singularities in the list respectively .
@xmath1320 it is easy to seen that @xmath1314 of infinite cases ( 5 ) and ( 7 ) is included in the weight types of the 1th singularities in the list up to permutation of coordinates .
+ + ( ii ) for each @xmath261 , ( 1a ) or ( 1b ) occurs .
then there are only 5 monomials in @xmath25 and @xmath26 can be determined . in order to determine @xmath1321 , we need at least one more monomial included in @xmath25 and @xmath26 . since 5 monomials in @xmath25 and @xmath26 are known , so it is easy to seen that one of following will occurs : + ( a ) there exists @xmath1322 such that ( 1a ) occurs when @xmath1323 + ( b ) there exists @xmath1324 such that ( 1b ) occurs when @xmath1325 .
+ if ( a ) occurs , then by lemma [ l3 ] ( 2 ) , we have @xmath1326 appears in @xmath26 for some non - negative integer @xmath284 or there exist @xmath1327 such that @xmath1328 appears in @xmath26 for some non - negative integer @xmath284 .
thus we have 6 monomials in @xmath25 and @xmath26 are determined now .
+ if ( b ) occurs , then by lemma [ l3 ] ( 2 ) , we have @xmath1329 appears in @xmath25 for some non - negative integer @xmath284 or there exist @xmath1330 such that @xmath1331 appears in @xmath25 for some non - negative integer @xmath284 .
thus there are 6 monomials in @xmath25 and @xmath26 are determined now .
+ more explicitly , let us consider the example that @xmath1332 appear in @xmath25 , and @xmath1333 appear in @xmath26 .
then by lemma [ l3 ] ( 2 ) we have @xmath1334 appears in @xmath25 for some non - negative integers @xmath284 or there exist @xmath1335 such that @xmath1336 appears in @xmath25 for some non - negative integer @xmath284 .
thus there are 6 monomials in @xmath25 and @xmath26 are determined .
it follows that @xmath1337 is determined as above .
and we have checked that each weight type @xmath386 gotten by this way , which satisfies the rational condition @xmath1269 and the conditions listed in corollary [ c2 ] , is the same as one of the weight types of the singularities in the above list up to permutation of coordinates . | we classify three dimensional isolated weighted homogeneous rational complete intersection singularities , which define many new four dimensional @xmath0 superconformal field theories .
we also determine the mini - versal deformation of these singularities , and therefore solve the coulomb branch spectrum and seiberg - witten solution . |
square lattice of antiferromagnetically coupled heisenberg spins is the simplest spin model in two dimensions ( 2d).@xcite its properties are nowadays well established by extensive numerical studies.@xcite the case of spin-@xmath15 entails strong quantum effects that reduce the sublattice magnetization@xcite and have an impact on the correlation length@xcite and spin dynamics.@xcite the ideal 2d model lacks long - range order ( lro ) above zero temperature , following the mermin - wagner theorem.@xcite however , any real material features a non - negligible interplane coupling that triggers the lro at a non - zero temperature @xmath16.@xcite when interplane couplings are frustrated and inactive , the lro is driven by anisotropy terms in the spin hamiltonian.@xcite suppression of the lro in square - lattice - based magnets is possible via two mechanisms , frustration of in - plane couplings or dilution of the spin lattice .
the former mechanism is revealed by the model of the @xmath17 frustrated square lattice , where the competition between nearest - neighbor couplings @xmath18 and second - neighbor couplings @xmath19 destroys the magnetic order in the vicinity of the quantum critical point at @xmath20 for .
this is the well - established theoretically but hitherto never observed experimentally case of the spin - liquid ground state in 2d .
the majority of the @xmath17 systems , mostly layered v@xmath2 phosphates , feature columnar antiferromagnetic ( afm ) order@xcite induced by @xmath21 .
materials with @xmath22 developing nel afm order remain low in number and sometimes challenging to investigate.@xcite the second mechanism is the dilution of the spin lattice with non - magnetic impurity atoms .
diluted systems are largely classical even for spin-@xmath15.@xcite the lro vanishes at the classical percolation threshold of @xmath23,@xcite where @xmath24 is the doping level .
the doping leads to a gradual suppression of the nel temperature @xmath16,@xcite but in many spin-@xmath15 materials the @xmath16 drops much faster than expected , because non - magnetic impurity atoms introduce magnetic frustration that contributes to the suppression of the lro.@xcite on the other hand , li@xmath12vosio@xmath25 , which is a frustrated square - lattice antiferromagnet even without dilution,@xcite exhibits a weaker effect on the sublattice magnetization and @xmath16 when diluted with non - magnetic ti@xmath2.@xcite apparently , the dilution of real materials never follows the idealized models and entails a modification of individual exchange couplings . here , we address magnetic properties and dilution behavior of the spin-@xmath15 antiferromagnet zn@xmath12vo(po@xmath13 , where magnetic v@xmath2 ions can be replaced by the non - magnetic ti@xmath2 .
the crystal structure of zn@xmath12vo(po@xmath13 features v@xmath2o@xmath26 pyramids that are linked into layers via po@xmath25 tetrahedra in the @xmath27 plane ( fig .
[ fig : structure]).@xcite given the small size of the interlayer zn@xmath28 cations , the interplane v
v distance ( 4.52a ) is shorter than the distance in the @xmath27 plane ( 6.31a ) .
this led earlier studies@xcite to conclude that zn@xmath12vo(po@xmath13 is a quasi - one - dimensional ( 1d ) magnet with @xmath29 , where @xmath11 and @xmath30 stand for the in - plane and interplane couplings , respectively ( fig .
[ fig : structure ] ) . a careful evaluation of thermodynamic data put forward the opposite , quasi-2d scenario with @xmath31.@xcite magnetic order observed below @xmath32k is afm in the @xmath27 plane and ferromagnetic ( fm ) along the @xmath33 direction.@xcite it is consistent with _ ab initio _ results by kanungo _ et al._,@xcite who also addressed the diluted , ti - doped case and proposed that the 25% ti doping should induce a monoclinic distortion reinstating the 1d physics , but this time in the @xmath27 plane and not along the @xmath33 direction .
in this study , we attempt to verify the prediction@xcite concerning the ti - doped zn@xmath12vo(po@xmath13 experimentally , and show that within the feasible doping levels , neither the monoclinic distortion nor the 1d physics are observed .
instead , thermodynamics of ti - doped zn@xmath12vo(po@xmath13 above @xmath16 is largely consistent with expectations for the diluted square lattice of heisenberg spins .
the nel temperature is systematically suppressed upon the dilution , and the rate of suppression is similar to that in li@xmath12vosio@xmath25 .
we provide accurate estimates of the in - plane exchange coupling in order to assess this effect quantitatively , and discuss our results in the light of available experimental data on diluted afm square lattices .
we also report additional characterization for thermodynamic properties , ground state and spin dynamics of the parent zn@xmath12vo(po@xmath13 compound .
these data will serve as a starting point for detailed studies of the doped material .
left panel : crystal structure of zn@xmath12vo(po@xmath13 .
right panel : magnetic layer in the @xmath27 plane .
dotted lines show hyperfine couplings between the p atom and the neighboring v@xmath2 spins ; each p atom is coupled to two magnetic ions with opposite spin directions .
the spins are along the @xmath33 direction.@xcite ]
polycrystalline samples of zn@xmath12(v@xmath34ti@xmath35)o(po@xmath25)@xmath12 ( @xmath24 = 0% , 5% , 10% , and 15% ) were prepared by a conventional solid - state reaction route . in the first step
, zn@xmath12p@xmath12o@xmath36 was synthesized using zno ( aldrich , 99.999% ) and nh@xmath25h@xmath12po@xmath25 ( aldrich , 99.999% ) as starting materials .
the stoichiometric mixtures were fired at @xmath37@xmath38c in air with one intermediate grinding . in the second step ,
zn@xmath12p@xmath12o@xmath36 was mixed with vo@xmath12 ( aldrich , 99.999% ) and tio@xmath12 ( aldrich , 99.999% ) and then the stoichiometric mixtures were fired in flowing ar - gas atmosphere with several intermediate grindings and palletizations at @xmath39@xmath38c . to check the sample purity , powder x - ray diffraction ( xrd , panalytical powder diffractometer and cuk@xmath40 radiation , @xmath41 ) was performed at room temperature .
the samples with @xmath24 = 0% , 5% , and 10% , were single - phase , but at higher doping concentrations several impurity phases including ti@xmath25o@xmath42(po@xmath25)@xmath42 emerged .
our repeated attempts to achieve higher doping levels by increasing or lowering the firing temperature were unsuccessful .
therefore , we focus on studying the samples with @xmath43% , where a minor amount of non - magnetic ti - containing impurities does not hinder the data analysis
. le bail fit of the powder xrd data was performed using the ` fullprof ` software package based on the tetragonal structure with space group @xmath44 to determine the lattice parameters.@xcite no indications of a symmetry lowering were observed .
all the data sets could be fitted using structural data of the parent compound as the initial parameters .
the refined lattice parameters and the goodness of fits ( @xmath45 ) are listed in table [ tab : latticeparameters ] .
no significant change in lattice constants ( @xmath46 and @xmath33 ) and unit cell volume ( @xmath47 ) was observed with increasing @xmath24 .
given the fact that ti@xmath2 features nearly the same ionic radius ( 0.51 ) as v@xmath2 ( 0.53 ) , we do not expect any substantial changes in the cell volume .
thus our experimental observation is consistent with expectations based on the ionic radii .
.[tab : latticeparameters ] lattice parameters ( @xmath46 , @xmath33 , and @xmath47 ) and the goodness of fit ( @xmath45 ) obtained from the le bail fit of the powder xrd data for zn@xmath12v@xmath34ti@xmath35o(po@xmath13 . [ cols="^,^,^,^,^",options="header " , ] nel temperature of an antiferromagnet depends on its exchange couplings .
therefore , for a proper interpretation of @xmath48 and its slope , one has to evaluate the change in @xmath11 upon doping .
to this end , we fitted magnetic susceptibility and specific heat of ti - doped samples in the same manner as we did in sec .
[ sec : thermo ] for the parent compound .
model curves were obtained by qmc simulations for the diluted square lattice of spins-@xmath15 .
fitted parameters are listed in table [ tab : fits ] and show a good match between the susceptibility and specific heat data .
the error bar for the values of @xmath11 is somewhat difficult to define , because statistical errors largely depend on the temperature range of the fitting .
however , even with a very optimistic error bar of 0.1k for the susceptibility fits above 7k , the change in @xmath11 between @xmath49 and @xmath50 is only marginal .
moreover , @xmath16 depends on @xmath51,@xcite so the 4% change in the @xmath11 value should have negligible effect on the @xmath16 .
its reduction is , therefore , solely due to the dilution , and the slope of @xmath48 reflects the dilution effect on the spin-@xmath15 afm square lattice in zn@xmath12vo(po@xmath13 .
( color online ) specific heat of ti - doped zn@xmath12vo(po@xmath13 measured in zero magnetic field .
the dashed lines are qmc fits , and the arrows mark nel temperatures @xmath16 depicted in the inset as a function of the doping level @xmath24 .
the solid line in the inset is the tentative linear fit @xmath52 with @xmath53
the exchange couplings extracted from @xmath54 , @xmath55 , and @xmath56 data for zn@xmath0vo(po@xmath1)@xmath0 are consistent with the values reported before from the @xmath54 analysis@xcite and neutron scattering experiments within the error bar.@xcite according to the @xmath57 ratio , zn@xmath0vo(po@xmath1)@xmath0 features the nel antiferromagnetic ground state with antiparallel spins on nearest neighbors in the @xmath27 plane ( fig .
[ fig : structure ] , right).@xcite in the crystal structure , squares are formed via v o p o v superexchange interaction paths .
in contrast to pb@xmath0vo(po@xmath1)@xmath0 where each p atom is coupled to four v@xmath2 spins,@xcite in zn@xmath0vo(po@xmath1)@xmath0 each p atom is coupled to two v@xmath2 spins only ( fig .
[ fig : structure ] , right ) .
the total hyperfine coupling constant at the p site is the sum of transferred hyperfine ( @xmath58 ) and dipolar ( @xmath59 ) couplings produced by v@xmath2 spins , i.e. , @xmath60 , where @xmath61 is the number of nearest - neighbor v@xmath2 spins of the p site .
the anisotropic dipolar couplings were calculated using lattice sums to be @xmath62oe/@xmath8 , @xmath63 oe/@xmath8 , and @xmath64 oe/@xmath8 along the @xmath46- , @xmath65- , and @xmath33-directions , respectively . clearly , the value of dipolar coupling is almost negligible compared to the total hyperfine coupling [ @xmath7 oe/@xmath8 ] suggesting that the dominant contribution to the total hyperfine coupling is due to the transferred hyperfine coupling at the p site .
the magnitude of this coupling depends on the relative orientation and the extent of overlap between the v(@xmath67 ) , p(@xmath68 ) , and o(@xmath69 ) orbitals . the internal field at the p site
will be canceled out if the p ion is located at a symmetric position with respect to two nearest neighbor v@xmath2 up and down spins .
however , the observation of a small remnant internal field at the p sites in the afm ordered state indicates that the p sites are not located at the perfect symmetric position and there is a small displacement of the p sites from the perfect symmetric position .
this is also consistent with the crystal structure where the p is sitting slightly above or below the line joining the neighboring up and down spins ( see the right panel of fig . [
fig : structure ] ) .
the @xmath3p line in the magnetically ordered state takes a typical rectangular shape , reflecting that the magnetic ordering is commensurate in nature .
if the magnetic structure were incommensurate with the lattice , the internal field would be distributed and the spectrum would not exhibit the rectangular shape seen in fig .
[ belowtn ] .
our spectra are , therefore , consistent with the collinear magnetic order determined from the neutron diffraction experiments.@xcite the central line does not disappear from the experimental spectra completely even at the lowest measured temperature .
nmr experiments on many other compounds , especially on powder samples , are reported to show similar coexistence of the high-@xmath70 and low-@xmath70 phases , e.g. , in bacup@xmath12o@xmath36 ( ref . ) , ( li , na)vge@xmath12o@xmath71 ( refs . ) , ( ca@xmath25al@xmath12o@xmath71)fe@xmath12(as@xmath34p@xmath35)@xmath12 ( ref . ) , bimn@xmath12po@xmath71 ( ref . ) , and ligacr@xmath25o@xmath72 ( ref . ) .
the origin of this central line remains an open question .
one could argue that the coexistence of two phases is due to a spread of the transition temperatures within the polycrystalline sample , but in such a case it is quite unlikely to observe a distinct peak in the temperature dependence of @xmath4 , as seen in fig .
[ t1 ] . one possible origin of the central line is the impurity phases . in order to check that , we measured @xmath73 below @xmath14 at the positions corresponding to the central peak and the rhs edge of the spectra .
it is to be noted that for any phosphorus containing impurity phase , the corresponding @xmath74 is expected to be different from the intrinsic @xmath74 of the sample .
moreover , if the central peak is a superposition of intrinsic and extrinsic contributions , one would observe a double exponential behaviour of the longitudinal recovery curves .
however our recovery curves at both positions follow single exponential behaviour and the magnitude and the temperature dependence of @xmath73 at both positions are the same , which clearly suggests that the central peak is an intrinsic feature of the sample and completely rules out the contribution of impurity phases . as discussed earlier , the p site in the ordered state experiences a finite internal field due to a slightly asymmetric position with respect to the neighboring up- and down- spins . on the contrary ,
a perfectly symmetric position of p should results in a single narrow spectral line at the zero - shift position .
hence it appears that the central peak may be originating from some p sites which are located close to the perfect symmetric position .
another possible origin of the central line could also be the presence of crystal defects or local dislocations in the polycrystalline sample .
nmr on a high - quality single crystal can probably resolve this issue .
the temperature dependence of @xmath75 in the critical region provides the critical exponent @xmath76 reflecting the universality class of the spin system .
the @xmath76 values expected for different spin- and lattice - dimensionalities are listed in table ii of ref . .
the value of @xmath76 obtained from the experiment is @xmath77 , which would be consistent with any of the 3d spin models ( heisenberg , ising , or xy ) .
given the direction of spins along the @xmath33-axis in the magnetically ordered state,@xcite the 3d ising case looks plausible .
on the other hand , the 3d behavior in the vicinity of @xmath16 should not be confused with the 2d - like behavior above @xmath16 , where the data are well described by the 2d model and 2d spin correlations manifest themselves in neutron scattering.@xcite however , the critical exponent for the 2d ising model@xcite @xmath78 would not be consistent with the experiment . given the fact that below @xmath16 spins are aligned with the @xmath33 direction,@xcite we may expect a weak ising anisotropy , but it is impossible to quantify this putative anisotropy using the data at hand .
interestingly , the critical behavior of zn@xmath12vo(po@xmath13 deviates from that of other square - lattice v@xmath2 antiferromagnets , where @xmath79 ( 2d xy universality class ) was systematically observed in li@xmath12vosio@xmath25 and li@xmath12vogeo@xmath25 ( refs . ) , pb@xmath12vo(po@xmath13 ( refs . ) , srznvo(po@xmath13 ( refs . ) , and other compounds.@xcite the origin of this difference should be addressed in future studies . as shown in fig .
[ t1](b ) , @xmath80 above @xmath81k is @xmath70-independent and increases slowly below 10k where the system begins to show antiferromagnetic short - range order .
the general expression for @xmath82 in terms of the dynamic susceptibility @xmath83 is@xcite @xmath84 where the sum is over wave vectors @xmath85 within the first brillouin zone , @xmath86 is the form factor of the hyperfine interactions as a function of @xmath85 , and @xmath87 is the imaginary part of the dynamic susceptibility at the nuclear larmor frequency @xmath88 . for @xmath89 and @xmath90 ,
the real component of @xmath91 corresponds to the uniform static susceptibility @xmath92 .
thus the temperature - independent @xmath80 above 10k in fig .
[ t1](b ) demonstrates the dominant contribution of @xmath92 to @xmath93 . on the other hand ,
a slight increase in @xmath80 below 10k indicates the growth of antiferromagnetic correlations with decreasing @xmath70 .
the symmetric location of phosphorous between the two v@xmath2 spins implies that nel - type afm spin fluctuations [ @xmath94 from neighboring spins should be largely filtered out ( @xmath95 ) because the p nuclei interact with v@xmath2 spins having opposite directions ( fig .
[ fig : structure ] , right ) .
when the coupling to the two v@xmath2 spins is equivalent , the afm fluctuations do not contribute to @xmath80 .
the residual enhancement of @xmath80 below 10k reflects the asymmetry of the hyperfine couplings .
this asymmetry is consistent with the crystal structure of zn@xmath12vo(po@xmath13 , where the p atoms are located on mirror planes running perpendicular to the ( @xmath96 ) or ( @xmath97 ) crystallographic directions .
the tensor of hyperfine couplings may change its orientation upon the reflection in the mirror plane , thus leading to non - equivalent interactions between p and the up- and down - spins on the neighboring v@xmath2 ions . in the afm
ordered state , @xmath73 is mainly driven by scattering of magnons , leading to a power - law temperature dependence.@xcite for @xmath98 , where @xmath99 is the gap in the spin - wave spectrum , @xmath73 either follows a @xmath100 behavior due to a two - magnon raman process or a @xmath101 behavior due to a three - magnon process , while for @xmath102 , it follows an activated behavior @xmath103 . as seen from fig .
[ t1](a ) , our @xmath3p @xmath73 data in the lowest temperature region ( 1.5k @xmath104 3.25k ) follow the @xmath101 behavior rather than the @xmath100 behavior suggesting that the relaxation is mainly governed by the three - magnon process .
the lack of activated behavior down to 1.5k indicates that the upper limit of @xmath99 is 1.5k . at sufficiently high temperatures , @xmath73 due to local moments is @xmath70-independent and can be expressed within the gaussian approximation of the auto - correlation function of the electronic spin as:@xcite @xmath105 where @xmath106 is the heisenberg exchange frequency , @xmath107 is the number of nearest - neighbor spins of each v@xmath2 ion , and @xmath108 is the number of nearest - neighbor v@xmath2 spins for a given p site .
the @xmath108 in the numerator takes into account the number of nearest - neighbor v@xmath2 spins responsible for producing fluctuations at the p site . using the relevant parameters , @xmath109oe/@xmath8 , @xmath110 s@xmath111t@xmath111 , @xmath112 , @xmath113 , @xmath114 , @xmath115 , and the high - temperature ( 150k ) relaxation rate of @xmath116 s@xmath111 for the p site in eq .
, the magnitude of the exchange coupling is calculated to be @xmath117k in good agreement with @xmath10k determined from the thermodynamic data ( sec .
[ sec : thermo ] ) .
( color online ) @xmath4 vs. @xmath118 ( in log scale ) measured at @xmath119 k for zn@xmath12v@xmath34ti@xmath35o(po@xmath25)@xmath12 ( @xmath24 = 0 and 0.10 ) samples .
the solid line is the fit by @xmath120.,width=288 ] one can see in fig .
[ t1](a ) that for @xmath121k a slight increase in @xmath4 was observed at 9.4mhz compared to the data measured at 75.5mhz . in order to check whether this difference is due to the effect of spin diffusion , we measured @xmath4 at different applied fields at @xmath119k .
long - wavelength ( @xmath122 ) spin fluctuations in a heisenberg magnet show diffusive dynamics . in 1d compounds , such spin diffusion results in a @xmath123 magnetic field dependence of @xmath4 , which has been observed in ( ch@xmath42)@xmath25nmncl@xmath42 , [email protected]@xmath26h@xmath26 , and
sr@xmath12cuo@xmath42.@xcite on the other hand , in 2d materials @xmath4 varies as @xmath124.@xcite in fig [ t1vsh ] , @xmath4 is plotted against @xmath118 ( in log scale ) measured at @xmath119 k for zn@xmath12v@xmath34ti@xmath35o(po@xmath25)@xmath12 ( @xmath24 = 0 and 0.10 ) samples .
both the data sets resemble with each other and show the same field dependency .
they can be fitted by the form @xmath120 where @xmath46 and @xmath65 are constants .
the linearity of the @xmath4 vs. @xmath125 dependence is indicative of the 2d nature of both the parent and 10% ti@xmath2 doped samples above @xmath16 .
zn@xmath12vo(po@xmath13 reveals a very clean case of a diluted antiferromagnet .
we have shown that the change in the nearest - neighbor coupling @xmath11 is marginal ( table [ tab : fits ] ) , and the frustration by second - neighbor couplings @xmath19 is negligible in the parent compound .
however , the nel temperature of zn@xmath12vo(po@xmath13 drops much slower than expected for the diluted afm square lattice of spins-@xmath15 . in a diluted system ,
the @xmath16 can be written as follows:@xcite @xmath126 where @xmath127 reflects the reduction in the interlayer coupling ( the probability to find two coupled spins in the adjacent layers ) , @xmath128 is the in - plane correlation length , and @xmath129 is the staggered magnetization at a given value of @xmath24 .
all these factors taken together should yield the slope @xmath130 ( ref . ) for the linear dependence of @xmath48 and spin-@xmath15 .
experimentally , carretta _ _ et al.__@xcite report @xmath131 and 3.5 for mg- and zn - doped la@xmath12cuo@xmath25 , respectively.@xcite this finding can be rationalized by assuming that zn atoms introduce additional frustration , whereas mg atoms do not.@xcite in ti - doped li@xmath12vosio@xmath25 , the slope of @xmath48 is @xmath132 , only .
papinutto _ _ et al.__@xcite proposed that this slope is solely due to the first term @xmath127 , while @xmath129 is only weakly influenced by doping because the effect of dilution is countered by the change in the frustration ratio .
this explanation looks plausible for li@xmath12vosio@xmath25 indeed , because the physics of this compound is determined by the competing nearest - neighbor and second - neighbor couplings on the square lattice.@xcite ti - doped zn@xmath12vo(po@xmath13 reveals an even lower @xmath133 , and in this compound frustration is clearly inactive . we have shown that the frustration is vanishingly small ( @xmath134 ) in the pristine zn@xmath12vo(po@xmath13 , while its increase ( if any ) will have an opposite effect on the system and increase @xmath135 above 3.2 instead of decreasing it to the experimental @xmath133 value .
the different doping behavior of zn@xmath12vo(po@xmath13 and li@xmath12vosio@xmath25 on one hand and la@xmath12cuo@xmath25 on the other can be ascribed to a different magnitude of their interlayer couplings .
while zn@xmath12vo(po@xmath13 shows signatures of the 2d physics above @xmath16 , the nel temperature of this compound is quite high , @xmath136 , hence @xmath137 . in li@xmath12vosio@xmath25 ,
the lower nel temperature of @xmath138 corresponds to an order - of - magnitude weaker interlayer coupling @xmath139,@xcite which is still much stronger than in la@xmath12cuo@xmath25 with its @xmath140 and @xmath141 .
magnetic anisotropy could be another reason for the different evolution of @xmath16 upon doping , but its effect is difficult to quantify . in la@xmath12cuo@xmath25 , dzyaloshinsky - moriya ( dm ) terms , the leading component of the anisotropy in spin-@xmath15 magnets , are about 1.5% of @xmath11.@xcite crystallographic symmetries of both li@xmath12vosio@xmath25 and zn@xmath12vo(po@xmath13 allow for non - zero dm couplings as well , but their magnitude is presently unknown .
regarding zn@xmath12vo(po@xmath13 , our nmr data provide an upper threshold of about 1.5k for the anisotropy gap .
this value is , however , nearly 20% of @xmath11 and exceeds typical dm anisotropies in v@xmath2 oxides.@xcite the variable interlayer coupling is a plausible reason for the different doping evolution of @xmath16 in square - lattice antiferromagnets . in la@xmath12cuo@xmath25 ,
the long - range order emerges only at low temperatures where the in - plane correlation length is about 100 lattice spacings,@xcite and the magnetic order is vulnerable to the dilution and disorder . in li@xmath12vosio@xmath25 and especially in zn@xmath12vo(po@xmath13 , the in - plane correlation length at @xmath16 is on the order of several lattice spacings , and interlayer couplings have larger influence on the long - range ordering , thus reducing the slope of @xmath48 compared to the ideal 2d case where @xmath142 .
therefore , the doping scenario of zn@xmath12vo(po@xmath13 may be of 3d type and will require one to view this compound as a spatially anisotropic 3d antiferromagnet , even though the physics above @xmath16 is 2d - like.@xcite finally , we note that our data do not support the _ ab initio
_ predictions by kanungo _ _ et al.__@xcite regarding the 1d physics of doped zn@xmath12vo(po@xmath13 .
while probably correct for the ordered monoclinic structure at the 25% doping level , their results do not apply to our case , where magnetic v@xmath2 and non - magnetic ti@xmath2 ions are randomly distributed in the structure , and the overall tetragonal symmetry is retained .
zn@xmath12vo(po@xmath13 is an antiferromagnetic compound with the in - plane coupling of @xmath10k , negligible in - plane frustration , and long - range magnetic order below @xmath143k .
thermodynamic properties above @xmath16 are well described by the heisenberg model on the afm square lattice .
nmr results confirm the commensurate nature of the magnetic order .
the spin - lattice relaxation rate @xmath4 below @xmath14 follows the @xmath101 behavior reflecting that the relaxation is governed by the three - magnon process .
@xmath4 at 15k varies as @xmath124 and supports the presence of strong 2d spatial anisotropy in both the parent and 10% ti@xmath2 doped compounds above @xmath16 . on the other hand , the critical exponent for the sublattice magnetization is consistent with any of the 3d universality classes and may reflect the sizeable interlayer exchange in zn@xmath12vo(po@xmath13 .
ti@xmath2 doping with up to 15% of ti@xmath2 leads to a uniform dilution of the spin lattice and only a marginal change in the in - plane exchange coupling .
@xmath16 goes down in a linear manner , but its slope is well below theoretical expectations for the diluted heisenberg antiferromagnet on the square lattice of spins-@xmath15 and may indiciate the importance of the interlayer exchange .
ay , na , and rcn would like to acknowledge dst india for financial support . at was funded by the mobilitas program of the esf ( grant no .
mtt77 ) and by the iut23 - 3 grant of the estonian research agency .
work at the ames laboratory was supported by the department of energy - basic energy sciences under contact no .
de - ac02 - 07ch11358 .
[ fig : structure ] was prepared using the ` vesta ` software.@xcite 75ifxundefined [ 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 * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop @noop @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) | we report static and dynamic properties of the antiferromagnetic compound zn@xmath0(vo)(po@xmath1)@xmath0 , and the consequences of non - magnetic ti@xmath2 doping at the v@xmath2 site . @xmath3p nuclear magnetic resonance ( nmr ) spectra and spin - lattice relaxation rate ( @xmath4 )
consistently show the formation of the long - range antiferromagnetic order below @xmath5k .
the critical exponent @xmath6 estimated from the temperature dependence of the sublattice magnetization measured by @xmath3p nmr at 9.4mhz is consistent with universality classes of three - dimensional spin models .
the isotropic and axial hyperfine couplings between the @xmath3p nuclei and v@xmath2 spins are @xmath7
oe/@xmath8 and @xmath9 oe/@xmath8 , respectively .
magnetic susceptibility data above 6.5k and heat capacity data above 4.5k are well described by quantum monte - carlo simulations for the heisenberg model on the square lattice with @xmath10k .
this value of @xmath11 is consistent with the values obtained from the nmr shift , @xmath4 and electron spin resonance ( esr ) intensity analysis . doping zn@xmath12vo(po@xmath13 with non - magnetic ti@xmath2 leads to a marginal increase in the @xmath11 value and the overall dilution of the spin lattice .
in contrast to the recent _ ab initio _ results , we find neither evidence for the monoclinic structural distortion nor signatures of the magnetic one - dimensionality for doped samples with up to 15% of ti@xmath2 .
the nel temperature @xmath14 decreases linearly with increasing the amount of the non - magnetic dopant . |
we are going to consider two search tasks asking us to find formulas with certain properties ( and in section [ pairs ] we add one more ) .
both are more complex than the mere task to construct hard tautologies for a given proof system that was discussed in the introduction . to motivate them
we shall describe two thought situations in proof complexity ; the search tasks are then abstract ( and simplified ) versions of those .
first assume that you can prove ( i.e. zfc can ) that @xmath28 and thus , in particular , super - polynomial lower bounds for all proof systems .
for a proof system @xmath0 and a constant @xmath16 denote by @xmath29 the statement @xmath30\ ] ] formalizing a polynomial lower bound for @xmath0 with degree @xmath31 .
it is easy to see that for any decent proof system ( see section [ 5 ] for a formal definition of decency ) , as long as we can prove some specific polynomial lower bound we can also prove its soundness .
the decency assumption allows to extend a proof of a falsifiable formula @xmath32 to a proof of @xmath33 and further to a proof of any @xmath12 , all in polynomial time .
but by a simple application of gdel s theorem zfc is not able to prove the soundness of all proof systems .
this suggests that we should be proving lower bounds conditioned upon the assumption that @xmath0 is indeed a cook - reckhow proof system .
if @xmath0 were not complete we do not need to bother with lower bounds for it , so the interesting clause of the cook - reckhow definition is the soundness and we are lead to implications : @xmath34 where @xmath35 is a universal sentence ( in the language @xmath36 of section [ 2 ] ) formalizing that any formula with a @xmath0-proof must be a tautology .
now we simplify the situation a bit more .
let @xmath37 be a circuit in @xmath3 variables @xmath38 and @xmath39 variables @xmath40 which we interpret as the provability relation of a proof system restricted to formulas of size @xmath3 and proofs of size at most @xmath41 .
this motivates the following * search task * @xmath42 defined for any constant @xmath16 : * input : @xmath1 and a size @xmath43 circuit @xmath37 in @xmath3 variables @xmath38 and @xmath39 variables @xmath40 * required output : either a size @xmath3 falsifiable formula @xmath32 such that @xmath44 is satisfiable or a size @xmath3 tautology @xmath12 such that @xmath45 is unsatisfiable .
the output of @xmath46 thus certifies that @xmath47 is not a non - deterministic circuit ( with input @xmath48 and non - deterministic variables @xmath49 ) that accepts @xmath5 . in other words ,
the output either witnesses that @xmath47 does not correspond to a sound proof system or that the proof system admits a lower bound .
the provability relation of a proof system restricted to size @xmath3 formulas and size @xmath41 proofs can be computed by circuits of size @xmath50 . in the formulation of the problem
we have represented the @xmath51 constant by @xmath31 as well .
in addition * cert * ignores the uniformity of such circuits corresponding to a particular proof system ( they can be constructed in log space from @xmath52 ) .
this is in line with the prevailing approach in complexity theory to reduce uniform problems to non - uniform finite combinatorial problems .
finally note that in our simplification we are taking the reflection principle just for the proof lengths corresponding to the lower bound we should witness ; this make sense due to the non - uniformity .
the second search task we shall consider is motivated by another thought experiment .
assume that you can prove that no optimal proof system exists and , in fact , that you have a uniform construction that from a proof system @xmath0 produces a stronger proof system @xmath53 ( i.e. not simulable by @xmath0 ) .
for definiteness , assume that there is one oracle polynomial time machine that for all @xmath0 defines @xmath53 when having the oracle for @xmath0 .
then we expect to be able to prove @xmath54 and , most importantly , that it is stronger @xmath55\ ] ] where @xmath56 is a size @xmath57 tautology formalizing the soundness of @xmath53 w.r.t .
all proof of size at most @xmath3 ( we assume for simplicity that a proof is always at least as long as the formula it proves so one parameter suffices ) . see a similar formula in ( [ m0 ] ) in the proof of lemma [ 5.3 ] .
any decent proof system can simulate @xmath53 if it can use @xmath56 as extra axioms ( see section [ 5 ] ) . in the following problem
@xmath6 represents a bit more adding @xmath6 is equivalent to adding the reflection principles for @xmath58 .
] generally any extra axiom .
let @xmath0 be a proof system and @xmath59 be constants .
consider the following promise * computational task * @xmath60 : * input : @xmath1 and a tautology @xmath6 such that @xmath61 * required output : any size @xmath3 tautology @xmath8 that has no proof in proof system @xmath62 , @xmath0 augmented by @xmath6 as an extra axiom scheme will be defined in section [ 5 ] . ] , of size less than @xmath9 .
the requirement that the size of @xmath8 is exactly @xmath3 is just for a technical convenience ; we could allow any interval @xmath63 $ ] instead .
let @xmath64 be an injective function such that @xmath65 and let @xmath66 be a boolean function .
the * nisan - wigderson generator * @xmath67 is defined using the notion of a design .
a @xmath68-design on @xmath69 $ ] is a set system @xmath70\}_{i\in [ m]}$ ] on @xmath69 = \{1 , \dots , n\}$ ] such that : * @xmath71 , for all @xmath72 , * @xmath73 , for all @xmath74 . the @xmath72-th bit of @xmath75 is computed by @xmath76 from the @xmath41-bit string @xmath77 , where @xmath78 and @xmath79 is the restriction of @xmath80 to @xmath81 . in the future the parameter @xmath41 will be determined by @xmath82 and we shall denote the restriction @xmath80 as well .
nisan and wigderson@xcite showed that there are such designs for a wide range of parameters @xmath83 and that one can construct them uniformly and feasibly . in particular , we can fix the parameters as follows : @xmath84 where @xmath85 is arbitrary .
we shall thus assume that fixing @xmath82 and @xmath86 fixes the other parameters and also some set system @xmath87 constructed from @xmath88 in time @xmath89 and with parameters meeting the requirements .
in fact , we need that @xmath90 the design from ( * ? ? ?
* l.2.5 ) has this property . in our construction
the function @xmath80 will be @xmath91 . by this
we mean that it is the characteristic function of a language in @xmath91 .
hence @xmath80 is defined by two @xmath27 predicates @xmath92 with @xmath93 and @xmath94 polynomial - time relations and @xmath31 a constant such that @xmath95 for @xmath96 .
any string @xmath49 witnessing the existential quantifier will be called a witness for @xmath97
. we shall use results from @xcite and those do assume that @xmath80 has unique witnesses , meaning that for each @xmath98 there is exactly one witness for @xmath97 .
a natural source of @xmath91 functions with unique witnesses are hard bits of one - way permutations .
that is , for a polynomial time ( and intended to be one - way ) permutation @xmath99 we have @xmath100 where @xmath101 is a hard bit predicate for @xmath102 .
the hardness of one - way permutations is measured as follows .
a polynomial time permutation @xmath102 is defined to be * @xmath103 one way with security parameter @xmath104 * iff for all @xmath41 and any circuit @xmath47 with @xmath41 inputs and of size at most @xmath104 it holds : @xmath105 \ \le \ \epsilon(\ell)\ .\ ] ] using the goldreich - levin theorem we may assume that such a permutation @xmath102 has a hard bit function @xmath101 .
the details can be found in goldreich @xcite .
our construction needs to assume that @xmath80 is hard in the sense of nisan and wigderson @xcite .
they define @xmath80 to be @xmath106-hard if for every @xmath41 and every circuit @xmath107 with @xmath41 inputs and of size at most @xmath108 it holds : @xmath109\ < \ 1/2 + \epsilon/2\ .\ ] ] they define then the * hardness of @xmath80 * , denoted * @xmath110 * , to be the maximal @xmath111 such that the function is @xmath112-hard .
this simplification makes sense when @xmath113 has the rate about @xmath114 as in nisan and wigderson @xcite . in the proof complexity situations studied in @xcite the parameter
@xmath111 plays the main role , with @xmath113 being primarily of the rate @xmath115 .
this corresponds to the fact that in applications of the original nisan - wigderson generators @xmath116 is usually exponentially large but for various purposes of proof complexity ( especially lengths - of - proofs lower bounds ) the best choice would be at the opposite end : @xmath117 .
this lead in @xcite to keeping @xmath113 and @xmath111 separate and using the notion of the approximating hardness ( defined there ) in place of @xmath110 . in this paper , however , we shall use only those results from @xcite where @xmath116 is exponentially large as in ( [ e1 ] ) and thus using the measure @xmath110 suffices here .
a one - way permutation @xmath102 with a hard bit @xmath118 is * exponentially hard * iff it is @xmath119 one - way with security parameter @xmath120 .
the hardness @xmath110 of @xmath80 is then @xmath120 as well .
details of these constructions can be found in goldreich @xcite .
we will use in sections [ 4 ] and [ pairs ] the hypothesis that an exponentially hard one - way permutation exists instead of the presumably weaker assumption that an @xmath91 function @xmath80 with unique witnesses and with exponential hardness @xmath121 exists .
the only reason is that the former hypothesis is more familiar than the latter one .
although the formulation of the search tasks * cert * and * find * may not suggests so explicitly this investigation resulted from a research program in proof complexity about the so called proof complexity generators and we shall use some ideas from this theory
. we shall start with a proof complexity conjecture of razborov ( * ? ? ? * conjecture 2 ) .
take an arbitrary string @xmath122 that is outside of the range @xmath123 of @xmath124 .
the statement @xmath125 is a @xmath126 property of @xmath127 and can be expressed by a propositional formula @xmath128 in the sense that @xmath129 the construction of the propositional translation of the @xmath126 statement is analogous to the usual proof of the np - completeness of sat .
the details can be found in any of @xcite ) .
note that the size of the formulas is polynomial in @xmath116 .
* razborov s conjecture * says that these tautologies are hard for extended frege system ef for @xmath124 defined as above , with @xmath130 and based on an @xmath91 function @xmath80 that is hard on average for @xmath131 .
pich @xcite proved the conjecture for all proof systems admitting feasible interpolation in place of ef . in @xcite
we have considered a generalization of the conjecture .
we shall recall only one part of that generalization dealing with exponentially large @xmath116 ; in the other parts @xmath132 and they use the notion of approximating hardness of a function mentioned in the previous section .
[ 2.1 ] assume @xmath80 is an @xmath133 function with unique witnesses that has an exponential nisan - wigderson hardness @xmath134 .
then there is @xmath135 such that for @xmath136 and for any infinite @xmath27 set @xmath137 that has infinitely many elements whose length equals to @xmath138 for @xmath139 it holds : @xmath140 let us observe that conjecture [ 2.1 ] has a proof complexity corollary including razborov s conjecture .
[ 2.2 ] let @xmath0 be any proof system .
assume that conjecture [ 2.1 ] holds and that the nisan - wigderson hardness @xmath110 of an @xmath91 function @xmath80 with unique witnesses is @xmath141 .
then there exists @xmath135 such that for all @xmath16 , the size of @xmath0-proofs of formulas
@xmath128 for all large enough @xmath82 and all @xmath125 of size @xmath142 is bigger than @xmath143 .
note that the set @xmath137 of all @xmath127 of lengths @xmath138 for @xmath144 for which @xmath128 has a @xmath0-proof of size at most @xmath143 is in @xmath27 .
now we recall ( a part of ) the consistency result from @xcite concerning conjecture [ 2.1 ] .
its technical heart is a lower bound on complexity of functions solving a certain search task associated with @xmath124 and that would , in principle , suffice for our purposes here . using the consistency
result itself , however , seems to decrease the number of technicalities one otherwise needs to discuss .
we first recapitulate a few basic definitions .
cook @xcite has defined a theory pv whose language @xmath36 has a name for every polynomial - time algorithm obtained from a few basic algorithms by the composition and by the limited recursion on notation , following cobham s @xcite characterization of polynomial time .
the details of the definition of @xmath36 can be found in @xcite but are not important here .
in fact , neither is the theory pv itself as we shall work with the true universal first - order theory of @xmath145 in the language @xmath36 .
we shall denote this theory @xmath146 , as in @xcite .
note that @xmath146 contains formulas expressing the soundness of all proof systems .
let @xmath80 be an @xmath147 function defined as in ( [ i0 ] ) .
let us abbreviate by @xmath148 the open @xmath36 formula @xmath149 where @xmath150 is from the set system @xmath87 ( polynomial time definable from @xmath151 and @xmath48 ) and @xmath93 , @xmath94 are from ( [ e3 ] ) .
we do not have a symbol in @xmath36 for the function on @xmath152 computed for @xmath144 on @xmath153 by @xmath124 as it is not a polynomial time function , and the function has to be defined .
one possible formalization of the statement @xmath154 for @xmath155 and @xmath156 is then @xmath157 \exists y ( |y| \le \ell^c)\ g(w , z , x , y)\ ] ] with @xmath31 from ( [ e3 ] ) .
now we are ready to state the result from @xcite we shall need . [
3.1 ] assume @xmath80 is an @xmath133 function with unique witnesses having the nisan - wigderson hardness @xmath110 at least @xmath141 .
then there is @xmath135 such that for any @xmath27 set @xmath137 that has infinitely many elements whose length equals to @xmath138 for @xmath139 and defined by @xmath36 formula @xmath158 , with @xmath159 open , theory @xmath146 does not prove the universal closure of the formula @xmath160 where @xmath161 is the formula with variables @xmath162 @xmath163 and @xmath118 is the formula @xmath164 \forall y ( |y| \le \ell^c ) \
\neg g(w , z , x , y)\ .\ ] ] this statement is in @xcite derived from a bit finer model - theoretic result .
the argument we shall use to derive the hardness of * cert * applies to a more general situation which we describe now . by an @xmath165 * algorithm * we shall mean two polynomial time predicates @xmath166 and @xmath167 and a constant @xmath16 similarly as in ( [ e3 ] ) but now with an extra argument @xmath168 for the non - uniform advice , and a sequence of advice strings @xmath169 such that @xmath170 ( w.l.o.g . we use constant @xmath31 also in the bound to the length of advice strings ) .
we shall assume that @xmath171 \oplus [ \exists y ( |y| \le |x|^c ) f_1(x , y , z)]\ ] ] is valid where @xmath172 is the exclusive disjunction .
thus an @xmath173 algorithm is an @xmath91 set of pairs @xmath174 of appropriate lengths augmented by a sequence of advice strings substituted for @xmath168 . in our situation it is more natural to talk about algorithms than sets as we shall be looking for `` errors they make '' .
we shall denote such an algorithm @xmath175 where @xmath176 is the triple @xmath177 from ( [ h1 ] ) .
for @xmath178 a language let us denote by @xmath179 the truth table of the characteristic function of @xmath178 on @xmath180 .
if @xmath181 then the set @xmath182 of such strings @xmath183 is in @xmath27 and can be defined by an @xmath36 formula as @xmath184 with @xmath185 an open formula .
any @xmath186 witnessing the existential quantifier for @xmath168 will be called a witness for @xmath187 .
note that @xmath188 . for
a language @xmath181 and a triple @xmath189 as in ( [ h1 ] ) define the * search task @xmath190 * as follows : * input : @xmath1 , string @xmath179 and a witness @xmath186 for @xmath191 , and a string @xmath192 such that @xmath193 * required output : a string @xmath194 such that @xmath189 using @xmath192 as an advice string makes an error on @xmath48 : @xmath195\ ] ] [ 4.1 ] assume that an exponentially hard one - way permutation exists .
let @xmath178 be a language such that @xmath181 .
then there exists a triple @xmath176 as in ( [ h1 ] ) such that no deterministic polynomial time algorithm can solve @xmath196 on all inputs for all sufficiently large lengths @xmath3 .
assume that language @xmath178 satisfies the hypothesis of the theorem and let @xmath189 be any triple as in ( [ h1 ] ) .
assume that @xmath197 is a deterministic polynomial time algorithm that attempts to solve @xmath198 on all inputs for all @xmath199 , for some @xmath200 .
we are going to define a universal @xmath36 sentence @xmath201 that is true iff @xmath197 solves @xmath198 for all inputs for all @xmath199 .
the sentence @xmath202 is the universal closure of : @xmath203 where @xmath107 is the formula @xmath204 with @xmath197 represented by an @xmath36 function symbol , and @xmath47 is the formula @xmath205\ .\ ] ] the following should be obvious : * claim 1 : * _ algorithm @xmath197 solves @xmath198 for all inputs for all @xmath199 iff the sentence @xmath202 is true . _ we are going now to define a specific @xmath173 triple @xmath176 as in ( [ h1 ] ) such that @xmath202 will be false for all @xmath181 , all polynomial time algorithms @xmath197 and all @xmath200 .
let @xmath102 be an exponentially hard one - way permutation with a hard bit @xmath118 .
hence the function @xmath80 from ( [ i0 ] ) has the exponential hardness @xmath206 .
the existence of such @xmath102 and @xmath118 is guaranteed by the hypothesis of the theorem .
take @xmath135 provided by theorem [ 3.1 ] and put @xmath207 . using @xmath208 define an @xmath173 triple @xmath209 as n ( [ h1 ] ) as follows : put @xmath210 and for @xmath194 , @xmath211 , @xmath192 of size @xmath212 and @xmath213 define : @xmath214\ .\ ] ] in other words , on input @xmath48
the algorithm computes the @xmath48-th bit of @xmath215 .
* claim 2 : * _ for no @xmath181 , no polynomial time algorithm @xmath197 and no @xmath200 is the sentence @xmath202 true . _ to see this note that by substituting term @xmath216 for @xmath48 in @xmath217 and quantifying it existentially @xmath218)$ ] allows to deduce from @xmath202 the universal closure of @xmath219 from theorem [ 3.1 ] .
hence , by that theorem , @xmath202 can not be true .
claims 1 and 2 imply the theorem .
we remark that the argument can be actually extended to rule out a larger class of algorithms @xmath197 : the so called student - teacher interactive computations of @xcite ( see also @xcite ) .
let @xmath175 be as above .
define circuits @xmath220 to be ( some canonical ) circuits with @xmath3 inputs @xmath48 and @xmath221 inputs @xmath49 that outputs @xmath222 iff @xmath223 holds .
we can choose @xmath16 large enough so that @xmath224 has size at most @xmath43 .
given @xmath1 , a time @xmath10 algorithm can compute the string @xmath225 ( as well as the witness for @xmath226 required in the general formulation of the theorem ) .
such an algorithm is then polynomial in the size of @xmath225 . if @xmath12 is a solution to @xmath42 on input @xmath224 then either @xmath227 and @xmath228 or @xmath229 and @xmath230 is satisfiable , in which case @xmath146 implies that @xmath231 .
in other words , an algorithm solving @xmath42 solves @xmath232 too .
this yields the following statement as a corollary to theorem [ 4.1 ] .
assume that an exponentially hard one - way permutation exists .
then there is @xmath16 such that no deterministic time @xmath10 algorithm can solve @xmath42 on all lengths @xmath233 .
the task * find * was formulated using the provability in a proof system and in this section we develop a technical tool allowing us to move from @xmath173 algorithms to proof systems .
we shall recall first the notion of a * proof system with advice * as introduced by cook - k .
* def.6.1 ) .
it is defined as the ordinary cook - reckhow proof system ( cf.the introduction ) except that the binary relation @xmath11 is decidable in polynomial time using an advice string that depends only on the length of @xmath48 ( the formula ) .
we say that the advice is polynomial iff its length is @xmath234 .
this concept has some interesting properties ; for example , in the classes of these proof systems - with varying bounds on the size of advice strings - there exists an optimal one .
we refer the reader to ( * ? ? ?
* sec.6 ) and to subsequent @xcite for further information .
our aim in this section is to link proof systems with polynomial advice with proof systems with extra axioms , as defined below .
a sequence of formulas @xmath235 will be called * p - bounded * iff @xmath236 for all @xmath3 . [ 5.1 ] let @xmath11 be an ordinary cook - reckhow proof system . 1 . for a tautology
@xmath6 the proof system @xmath58 is defined as follows : + a string @xmath14 is a @xmath237-proof of formula @xmath12 iff @xmath14 is a @xmath0-proof of a disjunction of the form @xmath238 where @xmath239 are arbitrary substitution instances of @xmath6 obtained by substituting constants and variables for variables .
2 . for a p - bounded sequence @xmath235 of tautologies define a string @xmath14 to be an @xmath240-proof of formula @xmath12 iff it is a @xmath241-proof of @xmath12 for @xmath242 .
we allow only substitutions of constants and variables in instances @xmath243 in part 1 as that makes sense for all proof systems ( e.g. we do not have to discuss various limitations on depth for constant depth frege systems ) and it suffices here .
systems @xmath240 are not meant to genuinely formalize the informal notion of proof systems with extra axioms ; such systems should not pose restrictions on which extra axioms can be used in proofs of which formulas .
we use them here only as a technical vehicle allowing us to move from proof systems with advice to ordinary proof systems . note that while @xmath58 is a cook - reckhow proof system , @xmath244 is generally not .
the following lemma is obvious as we can use the sequence @xmath235 as advice strings to recognize @xmath240 - proofs .
[ 5.2 ] let @xmath0 be a cook - reckhow proof system . for every p - bounded sequence @xmath235 of tautologies @xmath244 is a proof system with polynomial advice in the sense of cook - k.@xcite . in section [ tasks ]
we used informally the notion of a decent proof system , meaning a proof system that can perform efficiently a few simple manipulations with proofs .
we shall use the formalization of this notion from ( * ? ? ?
* sec.2 ) . in the following @xmath245
are formulas for @xmath233 and suitable @xmath246 with @xmath247 , @xmath38 and @xmath248 such that for all @xmath249 it holds that : * _ @xmath250 iff @xmath251 is a truth assignment satisfying formula @xmath32 .
_ the extra variables @xmath186 are used to compute the truth value , as in the @xmath27-completeness of sat . a proof system ( ordinary or with advice )
@xmath0 is *
decent * iff the following tasks can be performed by polynomial time algorithms : 1 . from a @xmath0-proof @xmath14 of formula @xmath252 and a truth assignment @xmath251 to variables @xmath48
construct a @xmath0-proof of @xmath253 .
2 . given a true sentence @xmath254 ( i.e. no variables ) construct its @xmath0-proof .
3 . given @xmath0-proofs @xmath255 of @xmath254 and @xmath256 of @xmath257
construct a proof of @xmath258 .
4 . given a formula @xmath259 and a @xmath0-proof of formula @xmath260 with variables @xmath261
construct a @xmath0-proof of @xmath32 .
conditions d1 - 3 are easy to verify for many of the usual proof systems ( e.g. frege systems mentioned in the introduction or resolution ) .
the algorithm for condition d4 is defined by induction on the number of connectives in @xmath32 , cf .
* chpt.9 ) .
[ 5.3 ] let @xmath26 be a proof system with polynomial advice and @xmath0 a decent cook - reckhow proof system . then for every constant @xmath16 there exists a p - bounded sequence @xmath235 of tautologies and @xmath262 such that : any tautology @xmath12 having a @xmath26-proof of size @xmath263 has an @xmath240-proof of size @xmath264 .
assume that @xmath26 uses polynomial size advice string @xmath265 for formulas of size @xmath3 .
for @xmath233 denote by @xmath266 a propositional formula such that : * @xmath267 has @xmath3 atoms @xmath268 for bits of a formula , @xmath221 atoms @xmath269 for bits of a @xmath26-proof , @xmath221 atoms @xmath270 for bits of an advice and @xmath57 atoms @xmath271 for bits of the computation of the truth value of @xmath272 with advice @xmath273 , * for size @xmath3 formula @xmath32 and a @xmath221 size strings @xmath14 and @xmath192 : @xmath274 iff @xmath275 with advice @xmath192 is true .
then take for @xmath276 the formula with variables @xmath277 @xmath278 where @xmath265 is the string used by @xmath26 for size @xmath3 formulas .
the formula expresses the soundness of @xmath26 and hence it is a tautology .
its total size is @xmath279 .
let @xmath32 be a size @xmath3 formula with variables among @xmath280 and having a size @xmath281 @xmath26-proof @xmath14 .
let @xmath282 be bits of the computation of @xmath275 with advice @xmath265 .
take the following substitution instance of @xmath276 : @xmath283 * claim : * _ for some @xmath262 depending only on @xmath16 and @xmath0 the formula @xmath32 has a @xmath241-proof of size @xmath284 .
_ we shall use the decency of @xmath0 . by the choice of @xmath14 and @xmath282 the sentence @xmath285 is true and hence has , by d2 , a size @xmath279 @xmath0-proof .
then , using d3 , use modus ponens to derive in size @xmath279 the formula @xmath286 d4 then allows to derive in @xmath0 the formula @xmath32 , in size @xmath57 .
the total size of the @xmath0-proof is @xmath279 .
note that we have not used the decency condition d1 explicitly ; it s role is replaced here by the definition of the system @xmath287 which takes as axioms all substitution instances of @xmath276 .
formulas @xmath276 depend not only on @xmath3 and @xmath265 but also on the bound @xmath221 to the length of @xmath49 .
this is the reason why we can not simply say that @xmath288 simulates @xmath26 ( it simulates only augmented by all such formulas @xmath276 , constructed for all @xmath3 , @xmath265 and all @xmath16 , does simulate @xmath26 as by the proof of the claim @xmath289 . ]
@xmath26-proofs of size at most @xmath221 ) .
we will need the following notion . for
a complexity class @xmath290 and a language @xmath178 define the property that * @xmath178 is infinitely often in @xmath291 * , denoted @xmath292 , iff there exists @xmath293 such that @xmath294 for infinitely many lengths @xmath3 .
recall the definition of the class @xmath173 at the beginning of section [ 4 ] .
the following consequence of conjecture [ 2.1 ] was noted at the end of ( * ? ? ?
* sec.30.2 ) and it uses an idea linking the output / input ratio of proof complexity generators with the unprovability of circuit lower bounds due to razborov @xcite , quite similar to the reasoning in razborov - rudich @xcite . [ 2.3 ] assume that conjecture [ 2.1 ] holds and that an exponentially hard one - way permutation exists .
then for every @xmath181 : @xmath295 in particular , @xmath296 .
take @xmath135 from conjecture [ 2.1 ] .
put @xmath297 and think of a string @xmath122 as of the truth - table of the characteristic function of the language @xmath178 on inputs of length @xmath3 ; denote it @xmath179 as in section [ 4 ] .
for any language @xmath178 in @xmath298 the set of strings @xmath183 is in @xmath27 : the @xmath27 witness can collect all @xmath299 @xmath300 witnesses for each variable setting - this will have size @xmath10 - and check their validity . in particular , if some @xmath181 would determine the truth - tables @xmath179 for @xmath301 such that all but finitely many lie outside the range of @xmath124 we would get a contradiction with conjecture [ 2.1 ] .
hence we get : * claim : * _ for infinitely many @xmath82 , for @xmath301 and @xmath302 : @xmath303 _ for @xmath304 let @xmath305 be such that @xmath306
. then computing @xmath178 on @xmath307 amounts to computing @xmath80 on @xmath308 .
but by the requirement ( [ e2 ] ) posed on @xmath87 the set @xmath308 can be computed from @xmath72 and @xmath251 ( taken as the advice string ) in time polynomial in @xmath82 and @xmath80 is an @xmath147 function .
this lemma has an immediate consequence for problem * cert*. assume that conjecture [ 2.1 ] holds and that an exponentially hard one - way permutation exists .
then for some @xmath16 the task @xmath42 is only partially defined for infinitely many lengths @xmath233 , i.e. there are inputs corresponding to @xmath3 for which the problem has no solution .
we shall derive the same consequence for the task * find * using the results of section [ 5 ] .
for a triple @xmath189 as in ( [ h1 ] ) define a proof system with polynomial advice @xmath272 by : * either @xmath309 , * or @xmath192 is the empty word and @xmath49 is a frege proof of @xmath48 , thinking of @xmath192 as the advice . note that condition ( [ h1 ] ) does not guarantee that @xmath310 is a proof system for some @xmath192 but the second clause of the definition of @xmath272 allows us to fall back on a frege system by taking for the advice the empty word .
now let @xmath169 such that @xmath170 be a sequence of advice words defining a proof system with advice @xmath26 ( there exists at least one such : the sequence of empty strings ) .
let @xmath0 be any decent cook - reckhow proof system and let formulas @xmath276 and constant @xmath262 be those provided by lemma [ 5.3 ] for @xmath31 from @xmath189 .
assume @xmath311 is such that @xmath312 and assume also w.l.o.g .
that @xmath313 .
consider the task @xmath314 .
a solution for input @xmath1 and @xmath276 is a size @xmath3 tautology @xmath8 and by the choice of @xmath315 it must be that @xmath316 but conjecture [ 2.1 ] implies analogously as above that for the triple @xmath189 coming from @xmath124 there will be infinitely many lengths @xmath317 and strings @xmath265 for which @xmath276 is a tautology but no such @xmath8 exists .
hence we have the following statement .
assume that conjecture [ 2.1 ] holds and that an exponentially hard one - way permutation exists .
then for all decent cook - reckhow proof systems @xmath0 there are constants @xmath59 such that the task @xmath318 has no solution for infinitely many lengths @xmath233 .
let @xmath319 and @xmath320 be two pairs of disjoint subsets of @xmath152 .
a * reduction * of @xmath320 to @xmath319 is a function @xmath66 such that for all @xmath98 : @xmath321 it is ( non - uniform ) * p - reduction * if @xmath80 is ( non - uniform ) p - time .
disjoint @xmath27 pairs appear quite naturally in several places of proof complexity .
most notably , the reflection principle for a proof system just asserts that two @xmath27 sets ( that of formulas with bounded size proofs and of falsifiable formulas ) are disjoint .
a particularly elegant form of this observation was found by razborov @xcite in the notion of the canonical pair of a proof system .
shadowing the relation of the provability of reflection principles to simulations , a similar one exists between the provability of disjointness of such pairs and simulations .
we refer the reader to @xcite for more background . given two pairs of disjoint sets @xmath319 and @xmath320 and a constant @xmath16 consider the * search task * @xmath322 : * input : @xmath1 and a circuit @xmath107 with @xmath3 inputs , several outputs and of size at most @xmath221 * required output : a string @xmath323 such that @xmath324 in other words , the output string @xmath98 certifies that circuit @xmath107 is not a reduction of @xmath325 to @xmath319 on @xmath180 . take a triple @xmath189 as in ( [ h1 ] ) and define @xmath326 and @xmath327 to be the sets of pairs @xmath328 such that @xmath329 and @xmath330 or @xmath331 , respectively . for a disjoint pair @xmath332 of sets such that @xmath333 take for language @xmath178 on @xmath180 simply @xmath161 . for @xmath192 of size @xmath281 consider circuit @xmath334 that takes size @xmath3 input @xmath48 and outputs the pair @xmath335 ; note that @xmath336 for @xmath337 .
then a solution to @xmath338 for input @xmath1 and @xmath334 is also a solution to @xmath196 .
hence theorem [ 4.1 ] implies the following statement .
assume that an exponentially hard one - way permutation exists .
then there are two disjoint @xmath27 sets @xmath339 and @xmath16 such that for any pair @xmath332 of disjoint sets such that @xmath333 the task @xmath322 is not solvable by a deterministic time @xmath10 algorithm .
the reader familiar with the canonical pairs of proof systems mentioned earlier may note that these sets are in @xmath340 and thus the theorem applies to them .
the role of conjecture [ 2.1 ] is rather ambivalent : it implies that @xmath4 ( lemma [ 2.2 ] ) but also that @xmath341 ( lemma [ 2.3 ] ) .
this is caused by the dual role of the nisan - wigderson generator ; it is a source of hard tautologies but also a strong proof system .
the reader should consider , before dismissing conjecture [ 2.1 ] because of lemma [ 2.3 ] , how little contemporary complexity theory understands about the power of non - uniform advice .
it would be interesting to have a variety of candidate combinatorial constructions @xmath53 of a proof system stronger than @xmath0 , as discussed in section [ tasks ] . at present only the construction of @xmath342 ,
the implicit @xmath0 , from @xcite applies to all proof systems and it is consistent with the present knowledge that it indeed yields stronger proof systems .
an indirect plausible construction of @xmath53 may use the relation between proof systems and first - order theories : take theory @xmath23 corresponding to @xmath0 , extend @xmath23 by @xmath343 ( or in some other gdelian fashion ) getting @xmath111 , and then take for @xmath53 the proof system simulating @xmath111 ( cf.@xcite about @xmath23 etc . ) .
but it is hardly combinatorially transparent .
the referee pointed out that problems @xmath344 and @xmath345 seem to be very close to each other and asked whether they could be , in fact , reduced to each other .
one can reduce @xmath346 to @xmath46 for suitable @xmath31 : for @xmath233 and a tautology @xmath6 of size at most @xmath7 take for @xmath47 a circuit ( canonically constructed ) computing the provability relation for @xmath62 on formulas of size @xmath3 and proofs of size at most @xmath9 .
the size of @xmath47 is polynomial in @xmath9 and so @xmath47 is a valid input to @xmath46 for some @xmath347 .
a solution to @xmath46 can not certify that @xmath47 is unsound and so it has to be a size @xmath3 formula certifying the lower bound @xmath348 for @xmath47 and hence the @xmath9 lower bound for @xmath62 too .
i do not know whether @xmath344 can be also reduced to @xmath345 but i suspect it can not be : valid inputs to @xmath344 are allowed either to fail to be a sound proof system or to fail to have short proofs while inputs to @xmath345 can not be unsound . ,
nondeterministic instance complexity and proof systems with advice , in : proc .
3rd international conference on language and automata theory and applications ( lata ) , springer - verlag , lncs 5457 , ( 2009 ) , pp.164 - 175 . , does advice help to prove propositional tautologies ?
, in : proc .
12th international conference on theory and applications of satisfiability testing ( sat ) , springer - verlag , lncs 5584 , ( 2009 ) , pp.65 - 72 . , characterizing the existence of optimal proof systems and complete sets for promise classes . in : a.e.frid , a.morozov ,
a.rybalchenko , k.w.wagner ( eds . ) , _ computer science - theory and applications _ , fourth international computer science symposium in russia , csr 2009 , novosibirsk , russia , ( august 18 - 23 , 2009 ) .
lecture notes in computer science * 5675 * , springer .
( 2009 ) , pp.47 - 58 .
, on methods for proving lower bounds in propositional logic , in : _ logic and scientific methods _ eds . m. l. dalla chiara et al .
, ( vol . 1 of proc .
of the tenth international congress of logic , methodology and philosophy of science , florence ( august 19 - 25 , 1995 ) ) , synthese library , vol.259 , kluwer academic publ . , dordrecht , ( 1997 ) , pp.69 - 83 .
, interactive computations of optimal solutions , in : b. rovan ( ed . ) : _ mathematical foundations of computer science _ ( b. bystrica , august 90 ) , lecture notes in computer science 452 , springer - verlag , ( 1990 ) , pp . | it is well - known ( cf .
k .-
pudlk@xcite ) that a polynomial time algorithm finding tautologies hard for a propositional proof system @xmath0 exists iff @xmath0 is not optimal .
such an algorithm takes @xmath1 and outputs a tautology @xmath2 of size at least @xmath3 such that @xmath0 is not p - bounded on the set of all @xmath2 s .
we consider two more general search problems involving finding a hard formula , * cert * and * find * , motivated by two hypothetical situations : that one can prove that @xmath4 and that no optimal proof system exists . in * cert
* one is asked to find a witness that a given non - deterministic circuit with @xmath3 inputs does not define @xmath5 . in *
find * , given @xmath1 and a tautology @xmath6 of size at most @xmath7 , one should output a size @xmath3 tautology @xmath8 that has no size @xmath9 @xmath0-proof from substitution instances of @xmath6 .
we shall prove , assuming the existence of an exponentially hard one - way permutation , that * cert * can not be solved by a time @xmath10 algorithm . using a stronger hypothesis about the proof complexity of nisan - wigderson generator
we show that both problems * cert * and * find * are actually only partially defined for infinitely many @xmath3 ( i.e. there are inputs corresponding to @xmath3 for which the problem has no solution ) .
the results are based on interpreting the nisan - wigderson generator as a proof system . a * propositional proof system * in the sense of cook and reckhow @xcite is a polynomial time relation @xmath11 such that for a binary string @xmath12 : @xmath13 where taut is the set of propositional tautologies ( in demorgan language for the definiteness ) . any string @xmath14 for which @xmath15 holds
is called a * @xmath0-proof * of @xmath12 . a proof system
( tacitly propositional from now on ) is * p - bounded * iff there exists a constant @xmath16 such that the above holds even with the requirement that @xmath17 .
cook and reckhow @xcite noted that a p - bounded proof system exists iff @xmath18 . hence proving that no p - bounded proof system exists would imply @xmath19 and thus also @xmath20 .
this fact elevated the investigation of lengths of proofs into a fundamental topic of mathematical logic approach to computational complexity .
strong lower bounds were proved for a variety of proof systems and several different methods for this purpose were invented .
examples of proof systems that appear to be outside of the scope of current methods are the so called * frege systems * : the usual text - book propositional calculi based on a finite number of axioms schemes and inference rules ( only quadratic lower bounds are known for them , cf.@xcite ) .
this apparent failure could cause an uninformed reader to dismiss the whole area of proof complexity .
however , although we may not be near proving that @xmath19 , the lower bounds for weaker proof systems proved so far do have consequences interesting in their own right .
for example , a single lower bound for a proof system @xmath0 implies time lower bounds for a class @xmath21 of sat algorithms associated with @xmath0 and all commonly used sat algorithms belong to such a class for some @xmath0 for which we have an exponential lower bound ( cf.@xcite and references given there ) .
another type of consequences can be found in bounded arithmetic : a lengths - of - proofs lower bound for @xmath0 often implies the unprovability of a true @xmath22 sentence in a first - order theory @xmath23 associated with @xmath0 .
these unprovability arguments do not use gdel s theorem and the @xmath22 sentences involved have typically a clear combinatorial meaning . and last but not least , any super - polynomial lower bound for @xmath0 also implies that @xmath20 is consistent with @xmath23 .
we shall not survey these proof complexity topics in detail here and instead refer the reader to expositions in @xcite or in ( * ? ? ?
* chpt.27 ) .
in this paper we are interested in the question how hard it is - to be measured in terms of computational complexity here - to find tautologies hard ( i.e. requiring long proofs ) for a given proof system . proposing plausible candidates for tautologies
hard for frege systems and for stronger proof systems turned out to be a quite delicate issue .
the lack of a variety of good candidates is one of principal obstacles for proving lower bounds for strong systems .
of course , in principle one would be happy to accept a suggestion for such a hard tautology from a friendly oracle .
however , the experience with known lower bound proofs shows that it is essential to have explicit formulas with a transparent combinatorial meaning . in particular , all first super - polynomial lower bounds for proof systems for which we have any such bounds were proved for some sequence of tautologies @xmath24 of size @xmath25 and constructible in polynomial time ( or even log space ) from @xmath1 .
this type of sequences of hard tautologies has been considered in @xcite and ( * ? ? ?
* chpt.14 ) and it exists for a proof system @xmath0 iff @xmath0 is not * optimal * , i.e. there exists a proof system @xmath26 that has a super - polynomial speed - up over @xmath0 ( w.r.t .
lengths of proofs ) on an infinite set of formulas .
it is consistent with the present knowledge , and indeed most researchers seem to conjecture that , that no optimal proof system exist and hence that for each @xmath0 a p - time constructible sequence of hard formulas exist .
however , deriving the existence of hard formulas from the assumption of non - optimality is not very illuminating : it is a basic proof complexity result that @xmath0 can not admit polynomial size proofs of formulas expressing the soundness of a proof system @xmath26 ( these formulas are log space constructible ) if @xmath26 has a super - polynomial speed - up over @xmath0 on an infinite set of formulas ( cf .
( * ? ? ?
* chpt.14 ) ) .
hence , in a sense , deriving the existence of a polynomial time sequence of hard tautologies from the non - optimality assumption amounts just to restating the assumption in a different terminology .
we refer the reader to beyersdorff - sadowski@xcite for further information and up - to - date references .
we shall consider in this paper two more general search problems in which the task includes a requirement to find a hard tautology .
the two problems model in their ways two hypothetical situations : a situation when one can prove @xmath4 ( i.e. super - polynomial lower bounds for all proof systems ) and a situation when one can prove that no optimal proof system exists by having a uniform method how to construct from a given proof system a stronger one .
these two tasks , * cert * and * find * , will be defined in section [ tasks ] ( in section [ pairs ] we add one more search task * pair * involving disjoint pairs of @xmath27 sets )
. we will prove ( using the hypothesis of the existence of a hard one - way permutation ) that * cert * can not be solved by exponential time algorithms and ( using a stronger hypothesis about the proof complexity of the nisan - wigderson generator ) that both * cert * and * find * actually can not be solved at all on infinitely many input lengths .
our primary motivation for this research is to understand what kind of consequences do various - both proven and conjectural - statements about the proof complexity of the nisan - wigderson generator have .
the paper is organized as follows .
after the motivation and the definition of the search tasks * cert * and * find * in section [ tasks ] we review some complexity theory in section [ 1 ] and some proof complexity in section [ 2 ] .
the hardness results are proved in sections [ 4 ] and [ 6 ] , respectively ( after a proof complexity interlude in section [ 5 ] ) .
the paper is concluded by section [ pairs ] considering a related search task for disjoint pairs of sets and a few remarks in section [ rema ] .
we do assume only basic complexity theory and proof complexity ( e.g. the well - known relation between reflection principles and simulations ) .
but the reader may still benefit from understanding a wider proof complexity context .
in particular , ( * ? ? ? * chpt.27 ) overviews some fundamental problems of proof complexity and ( * ? ? ? * chpts.29 and 30 ) survey the theory proof complexity generators ( and list relevant literature ) . |
in teleparallel model of gravity , instead of torsion - less levi - civita , curvature - less weitzenbck connections are used @xcite . in this model
the action reads @xmath1\sqrt{-g}d^4x.\ ] ] @xmath2 is matter lagrangian density .
@xmath3 , and @xmath4 denotes the newtonian gravitational constant .
@xmath5 is the torsion scalar given by @xmath6 where @xmath7 is the torsion tensor .
@xmath8 are vierbeins whose duals are @xmath9 .
the metric tensor is by @xmath10 .
also , @xmath11 and @xmath12 ( for more details bout teleprallel gravity see @xcite ) .
the sum of the ricci scalar , @xmath13 , and the torsion scalar , @xmath5 , is a total divergence @xcite @xmath14 where @xmath15 .
therefore we find that the equations of motion derived from ( [ 1 ] ) are equivalent to those derived from einstein - hilbert action @xmath16\sqrt{-g}d^4x.\ ] ] one way to explain the inflationary regime of the early universe and also the late time acceleration is to introduce exotic scalar field in the einstein - hilbert action . if this scalar field is minimally coupled to gravity , we find that the difference between ( [ 1 ] ) and ( [ 4 ] ) becomes again a total divergence and both theories give the same equations of motion , but for non - minimal coupling this situation changes .
let us consider an action with non - minimal coupling term @xmath17 @xcite , @xmath18\sqrt{-g}d^4x.\ ] ] this non - minimal coupling is required for renormalizability of the theory , and has also root in quantum corrections to scalar field theory on curved space - time @xcite .
this model has been employed to study the inflationary epoch by attributing the higgs scalar to the inflaton @xcite .
inspired by ( [ 5 ] ) , the action @xmath19\sqrt{-g}d^4x\ ] ] has been introduced to describe the positive acceleration of the universe as well as the possible phantom divide line crossing @xcite . due to nonminimal coupling , i.e. @xmath20 , ( [ 5 ] ) is not equivalent to ( [ 6 ] ) . in @xcite
a more general action was proposed @xmath21\sqrt{-g}d^4x,\ ] ] where @xmath22 .
this model includes ( [ 5 ] ) , and ( [ 6 ] ) as its subclasses .
this can be seen by setting @xmath23 , and @xmath24 .
recently , different cosmological and gravitational aspects of this model was discussed in the literature @xcite . in this paper
we consider ( [ 7 ] ) in a spatially flat friedmann - lematre - robertson - walker ( flrw ) space - time filled nearly with pressureless matter and dark energy . based on this action
we try to explain the late time cosmic acceleration after an epoch of matter domination .
to do so , beside the non - minimal couplings , we also need to a conformal coupling between the quintessence and pressureless matter .
this kind of coupling has been previously considered in screening model , such as chameleon @xcite and symmetron @xcite . in the symmetron model ,
cosmic acceleration is related to a @xmath0 symmetry breaking triggered by reduction of dark matter density in the late time .
but due to large mass of quintessence the scalar field overshoots rapidly and oscillates about the minimum of its effective potential and ceases promptly the acceleration .
onset of acceleration in our model is somehow similar to the symmetron model , but the presence of non - minimal couplings in teleparallel model provides a framework such that the quintessence can reside at the minimum of its effective potential giving rise to a de sitter expansion ultimately . the scheme of the paper is as follows : in the second section we introduce our model and , by obtaining vacuum solutions of the system , study @xmath0 symmetry breaking which leads to an accelerated universe after a decelerated matter dominated phase . in the third section , by employing a dynamical phase analysis , we study the stability of late time vacuum solution . in the fourth section
we conclude our results .
we will use units @xmath25 .
we start with the action @xmath26 where @xmath27 and @xmath28 .
@xmath29 is a positive conformal coupling .
we consider a spatially flat flrw space time filled nearly with the scalar field dark energy and pressureless matter .
the scalar field equation of motion derived from ( [ 8 ] ) is @xmath30 the matter continuity equation reads @xmath31 by defining the density @xmath32 we find @xmath33 note that @xmath34 is a mathematical @xmath35 independent variable , which facilitates our computations and interpretations .
in this space - time we have @xmath36 and @xmath37 , where @xmath38 is the hubble parameter which in terms of the scalar factor , @xmath39 , is given by @xmath40 .
variation of ( [ 8 ] ) with respect to the vierbeins yields @xmath41 and @xmath42 two of the three equations ( [ 13 ] ) , ( [ 14 ] ) , and ( [ 15 ] ) are independent .
we consider ( [ 13 ] ) , and ( [ 14 ] ) as independent equations . to solve them
we must identify @xmath43 and @xmath29 . as we want to study the @xmath0 symmetry breaking
, we choose the potential as @xmath44 and take @xmath29 as an even function of @xmath35 , such that @xmath45 in ( ) behaves as a @xmath34 dependent mass term .
based the coincidence problem and structures formation , we expect that in the early universe the contribution of dark energy was negligible .
therefore we assume that , initially the scalar field stays at the vacuum identified by @xmath46 , where @xmath47 and @xmath48 so @xmath49 in this epoch the universe is dominated by matter evolving according to : @xmath50 note that @xmath51 is a nonzero number .
the scalar field stays in the vacuum , @xmath46 , as long as its squared effective mass is still positive @xmath52 from ( [ 13 ] ) , and ( [ 16 ] ) , we find @xmath53 to build our model we need that @xmath54 becomes a nonzero number . in this way , the sign of the effective squared mass depends on the value of the dark matter density .
like @xcite , we take @xmath55 where @xmath56 is a mass scale . hence ( [ 19 ] ) becomes @xmath57 thus provided that @xmath58 as long as @xmath59 @xmath60 holds and the scalar field stays at the stable point @xmath46 .
when @xmath61 , @xmath62 and this point becomes unstable .
the scalar field becomes tachyonic and rolls to the new minimum of its effective potential .
so in the first stage we have a universe with zero energy density dominated by baryonic and cold dark matter . in the second stage
we require that dark energy density increases and gives rise to universe acceleration . to see how this expectation can be realized ,
let us obtain the vacuum after the symmetry breaking . by bearing in mind that @xmath63
, the late time vacuum solution can be found from ( [ 13 ] ) and ( [ 14 ] ) .
this solutions is characterized by @xmath48 , @xmath64 and @xmath65 from ( [ 15 ] ) we have also @xmath66 .
this solution corresponds to a de sitter expansion with a constant @xmath38 . in the first stage
the deceleration parameter @xmath67 , is @xmath68 . in the second stage ,
the late time critical point is given by @xmath69 .
so we expect that universe enters a positive accelerated phase after matter dominated era .
finally the system settles down in a de sitter state .
solutions of ( [ 24 ] ) are @xmath70 these solutions are real provided that @xmath71 also , when @xmath72 only @xmath73 and @xmath74 exist .
( [ 26 ] ) or ( [ 27 ] ) imply that @xmath75 .
whence , if we insist to have late time acceleration , the initial stability condition ( [ 22 ] ) does not hold if we ignore the conformal coupling .
in other words the conformal coupling retains the quintessence in initial stable state for @xmath76 , while the nonminimal couplings ( @xmath77 ) are necessary for late time acceleration . to elucidate this subject
let us see whether the same scenario can be set up in the minimal case , @xmath78 . in the first stage
when @xmath79 , the quintessence settles down in the minimum of its effective potential at @xmath46 . when @xmath80 , the @xmath0 symmetry breaks and the field moves towards the new vacuum @xmath81
. finally as @xmath82 , @xmath35 oscillates about @xmath83 .
note that as @xmath84 , the solution ( [ 24 ] ) , which gives a negative @xmath85 , is not valid in this case and the field can not stay at the minimum of its potential but oscillates about it and acts as a pressureless matter @xcite , giving rise to a decelerated expansion .
moreover , if one considers local gravity tests to screen the new force mediated by the scalar field , he obtains a large mass for the scalar field which expedites this overshoot . besides , after the symmetry breaking the quintessence potential ( [ 16 ] ) becomes negative and as it is known a quintessence with negative potential can not drive the acceleration in the minimal case @xcite . to remedy this ,
some authors add a positive constant to ( [ 16 ] ) playing the role of a cosmological constant rendering the problem to @xmath86 scenario .
phase space analysis may be employed to investigate the stability of the solution ( [ 25 ] ) . to do
so we define @xmath87 these variables satisfy the autonomous equations @xmath88 where @xmath89 and prime denotes derivative with respect to @xmath90 .
@xmath91 is given by @xmath92 to derive ( [ 30 ] ) we have used the friedmann equation @xmath93 the critical point corresponding to the late time evolution of the universe is given by @xmath94 corresponding to ( [ 25 ] ) , where @xmath95 .
the stability of this point may be checked as follows .
setting @xmath96 , @xmath97 , @xmath98 , we find @xmath99 where @xmath100 at the critical point @xmath101 , @xmath102 becomes @xmath103 where @xmath104 from ( [ 29 ] ) , we obtain @xmath105 inserting ( [ 36 ] ) into ( [ 35 ] ) , and by considering the potential ( [ 16 ] ) , we find @xmath106 where @xmath107
so eigenvalues of ( [ 34 ] ) are @xmath108 if we require that all eigenvalues to be negative , which is sufficient condition to have an attractor , we must take @xmath109 ( [ 37 ] ) and ( [ 40 ] ) give @xmath110 which is the condition to have an attractor at late time .
we tried to attribute the late time cosmic acceleration to symmetry breaking in a teleparallel model of cosmology in which quintessence is coupled no - minimally to both the scalar torsion and a boundary term .
we derived equations of motion , from which suitable solutions were derived .
the solutions were classified into two classes : before symmetry breaking and after symmetry breaking . before symmetry breaking , provided that ( [ 22 ] ) holds , the conformal coupling holds the quintessence in the bottom of its effective potential . in this epoch dark energy
has no contribution in total energy and so the coincidence problem may be alleviated in this model . when the density of matter becomes less than a critical value ( given by ( [ 23 ] ) ) , the initial vacuum becomes a false vacuum and the quintessence moves towards its new vacuum which describes a state with de sitter expansion ( provided that the conditions ( [ 26 ] ) or ( [ 27 ] ) be satisfied ) .
finally , if ( [ 41 ] ) holds , the scalar filed resides at this vacuum due to the presence of the nonminimal couplings . in figures ( [ fig1 ] ) and ( [ fig2 ] ) ,
the behaviors of the quintessence and the deceleration parameter are depicted .
figure ( [ fig1 ] ) shows that the scalar field leaves @xmath46 after the symmetry breaking and tends towards the final critical point and finally settles down in it .
figure ( [ fig2 ] ) shows that , during this evolution , the deceleration factor becomes negative in a time of order of the present hubble time .
this figure also shows the possibility to have super - acceleration phase @xmath111 which is forbidden in the minimal case ( see([15 ] ) ) .
99 a. unzicker and t. case , arxiv : physics/0503046 [ physics.hist-ph ] ; k. hayashi and t. nakano , prog .
38 , 491 ( 1967 ) ; c. pellegrini and j. plebanski , k. dan . vidensk .
selsk . mat .
2 ( 1962);e .
v. linder , phys .
d 81 , 127301 ( 2010 ) .
c. q. geng , c. c. lee , e. n. saridakis , and y. p. wu , phys .
b 704 , 384 ( 2011 ) ; c. xu , e. n. saridakis , and g. leon , arxiv:1202.3781 ; g. r. bengochea and r. ferraro , phys . rev .
d 79 , 124019 , ( 2009 ) ; j. t. li , y. p. wu , and c. q. geng , phys .
d 89 , 044040 ( 2014 ) ; l. jarv and a. toporensky , phys .
d 93 , 024051 ( 2016 ) , arxiv:1511.03933 ; k. bamba , r. myrzakulov , s. nojiri , and s. d. odintsov , phys . rev .
d 85 , 104036 ( 2012 ) ; h. m. sadjadi , phys .
b 718 , 270 ( 2012 ) ; h. m. sadjadi , phys . rev .
d 87 , 064028 ( 2013 ) ; d. horvat , s. ilijic , a. kirin , and z. narancic , class .
quantum grav .
32 , 035023 ( 2015 ) ; y. kucukakca , arxiv:1404.7315 ; c. q. geng , c. c. lee , and e. n. saridakis , arxiv:1110.0913 ; g. otalora , phys . rev .
d 88 , 063503 ( 2013 ) .
s. bahamonde and m. wright , phys . rev .
d 92 , 084034 ( 2015 ) , arxiv:1508.06580 .
s. bahamonde , c. g. boehmer , and m. wright , phys . rev .
d 92 , 104042 ( 2015 ) , arxiv:1508.05120 ; m. wright , phys . rev .
d 93 , 103002 ( 2016 ) .
d. freedman and e. weinberg , ann .
( ny ) 87 , 354 ( 1974 ) ; d. freedman , i. muzinich , and e. weinberg , ann
87 , 95 ( 1974 ) ; c. g. callan , s. coleman , and r. jackiw , ann . phys .
( ny ) 59 , 42 ( 1970 ) ; v. sahni and s. habib , phys .
81 , 1766 ( 1998 ) ; v. faraoni , phys . rev .
d 62 , 023504 ( 2000 ) ; e. elizalde , s. nojiri , and s. odintsov , phys . rev .
d 70 , 043539 ( 2004 ) ; p. wang , p. wu , and h. yu , eur . phys .
j. c 72 , 2245 ( 2012 ) , arxiv:1301.5832 .
f. l. bezrukov , a. magnin , and m. shaposhnikov , phys .
b 675 , 88 ( 2009 ) .
m. zubair , and s. bahamonde , arxiv:1604.02996 ; h. m. sadjadi , arxiv:1606.04362 ; h. m. sadjadi , phys . rev .
d 92 , 123538 ( 2015 ) ; s. bahamonde , u. camci , s. capozziello , and m. jamil , arxiv:1608.03918 . j. khoury and a. weltman , phys .
lett.93 , 171104 ( 2004 ) , s. nojiri , and s. d. odintsov , mod .
a 19 , 1273 ( 2004 ) ; s. das , and n. banerjee , phys . rev .
d 78 , 043512 ( 2008 ) .
k. hinterbichler , j. khoury , phys .
104 , 231301 ( 2010 ) ; a. c. davis , b. li , d. f. mota , and h. a. winther , arxiv:1108.3081 ; r. hagala , c. llinares , and d. f. mota , arxiv:1607.02600 .
m. sami and r. myrzakulov , arxiv:1309.4188 ; k. hinterbichler , j. khoury , a. levy , and a. matas , phys .
d 84 , 103521 ( 2011 ) .
m. honardoost , h. mohseni sadjadi , and h. r. sepangi , gen .
48 , 125 ( 2016 ) ; h. mohseni sadjadi , m. honardoost , and h. r. sepangi , arxiv:1504.05678 . | in the teleparallel framework of cosmology , a quintessence with non - minimal couplings to the scalar torsion and a boundary term is considered . a conformal coupling to matter density
is also taken into account .
it is shown that the model can describe onset of cosmic acceleration after an epoch of matter dominated era , where dark energy is negligible , via @xmath0 symmetry breaking .
while the conformal coupling holds the universe in a vacuum with zero dark energy density in the early epoch , the non - minimal couplings lead the universe to a stable state with de sitter expansion at late time . |
robins et al , 2008 , published a theory of higher order influence functions for inference in semi- and non - parametric models .
this paper is a comprehensive manuscript from which robins et al , was drawn .
the current paper includes many results and proofs that were not included in robins et al due to space limitation .
particular results contained in the present paper that were not reported in robins et al include the following . given a set of functionals and their corresponding higher order influence functions ,
we show how to derive the higher order influence function of their product .
we apply this result to obtain higher order influence functions and associated estimators for the mean of a response @xmath0 subject to monotone missingness under missing at random .
these results also apply to estimating the causal effect of a time dependent treatment on an outcome @xmath0 in the presence of time - varying confounding .
finally , we include an appendix that contains proofs for all theorems that were stated without proof in robins et al , 2008 . the initial part of the paper is closely related to robins et al , , the latter parts differ .
we have developed a theory of point and interval estimation for nonlinear functionals @xmath2 in parametric , semi- , and non - parametric models based on higher order likelihood scores and influence functions that applies equally to both @xmath1 and non-@xmath1 problems ( robins 2004 , sec .
9 ; li et al , 2006 , tchetgen et al , 2006 , robins et al , 2007 ) .
the theory reproduces results previously obtained by the modern theory of non - parametric inference , produces many new non-@xmath1 results , and most importantly opens up the ability to perform non-@xmath1 inference in complex high dimensional models , such as models for the estimation of the causal effect of time varying treatments in the presence of time varying confounding and informative censoring .
see tchetgen et al .
( 2007 ) for examples of the latter .
higher order influence functions are higher order u - statistics .
our theory extends the first order semiparametric theory of bickel et al .
( 1993 ) and van der vaart ( 1991 ) by incorporating the theory of higher order scores and bhattacharrya bases considered by pfanzagl ( 1990 ) , small and mcleish ( 1994 ) and lindsay and waterman ( 1996 ) .
the purpose of this paper is to demonstrate the scope and flexibility of our methodology by deriving rate - optimal point and interval estimators for various functionals that are of central importance to biostatistics .
we now describe some of these functionals .
we suppose we observe i.i.d copies of a random vector @xmath3 with unknown distribution @xmath4 on each of @xmath5 study subjects . in this paper
, we largely study non - parametric models that place no restrictions on @xmath6 other than bounds on both the @xmath7 norms and on the smoothness of certain density and conditional expectation functions .
the variable @xmath8 represents a random vector of baseline covariates such as age , height , weight , hematocrit , and laboratory measures of lung , renal , liver , brain , and heart function .
@xmath8 is assumed to have compact support and a density @xmath9 with respect to the lebesgue measure in @xmath10where , in typical applications , @xmath11 is in the range 5 to 100 .
@xmath12 is a binary treatment and @xmath0 is a response , higher values of which are desirable .
then , in the absence of confounding by additional unmeasured factors , the functional @xmath13 \right\ } -e\left\ { e\left [ y|a=0,x\right ] \right\ } $ ] is the mean effect of treatment in the total study population .
our results for @xmath14 \right\ } -e\left\ { e\left [ y|a=0,x\right ] \right\ } $ ] follow from results for the functional @xmath15 \right\ } $ ] based on data @xmath16 rather than @xmath17 if @xmath0 is missing for some study subjects , and @xmath12 is now the indicator that takes the value @xmath18 when @xmath0 is observed and zero otherwise , then the functional @xmath14 \right\ } $ ] is the marginal mean of @xmath0 under the missing at random assumption that the probability @xmath19 = p\left [ a=0|x\right ] \ $ ] that @xmath0 is missing does not depend on the unobserved @xmath20 returning to data @xmath21 the functional @xmath22 \\ & = e\left [ w\left ( x\right ) \left\ { e\left [ y|a=1,x\right ] -e\left [ y|a=0,x\right ] \right\ } \right]\end{aligned}\ ] ] with @xmath23 \ $ ] is the variance weighted average treatment effect . our results for @xmath24 $ ] are derived from results for the functionals @xmath25 and @xmath26 .$ ] we note that robins and van der vaart s ( 2006 ) construction of an adaptive confidence set for a regression function @xmath27 depended on being able to construct a confidence interval for @xmath28 .$ ] they constructed an interval for @xmath29 $ ] when the marginal distribution of @xmath8 was known . in this paper , we construct a confidence interval for @xmath30 $ ] when the marginal of @xmath8 is unknown and , in section [ adaptive_section ] , use it to obtain an adaptive confidence set for @xmath27 . the functional @xmath31 is the functional @xmath32 in the special case in which @xmath33 wp1 .
minimax estimation of @xmath34 has recently been discussed by wang et al .
( 2006 ) and cai et al .
( 2006 ) in the setting of non - random @xmath8 .
the function @xmath35 -e\left [ y|a=0,x = x\right ] $ ] is the effect of treatment on the subgroup with @xmath36 it is important to estimate the function @xmath37 , in addition to the average treatment effect in the total population , because treatment should be given , since beneficial , to those subjects with @xmath38 but withheld , since harmful , from subjects with @xmath39 we show that one can obtain adaptive confidence sets for @xmath37 if one can set confidence intervals for the functional @xmath40 $ ] .
we construct intervals for @xmath41 $ ] under the additional assumption that the data @xmath42 came from a randomized trial . in a randomized trial , in contrast to an observational study ,
the randomization probabilities , @xmath43 are known by design .
we plan to report confidence intervals for @xmath44 $ ] with @xmath45 unknown elsewhere .
all of the above functionals @xmath2 have a positive semiparametric information bound ( sib ) and thus a ( first order ) efficient influence function with a finite variance .
in fact all the functionals @xmath2 have efficient influence function @xmath46 where @xmath47 are monotone functions of certain conditional expectations , and , for any @xmath48@xmath49 = e_{f}\left [ h_{1}\left ( o\right ) \left\ { b^{\ast}\left ( x\right ) -b\left ( x;f\right ) \right\ } \left\ { p^{\ast}\left ( x\right ) -p\left ( x;f\right ) \right\ } \right]\ ] ] where @xmath50 is a known function .
we refer to functionals in our class as doubly - robust to indicate that @xmath51 continues to have mean zero when either ( but not both ) @xmath52 is misspecified as @xmath53 or @xmath54 is misspecified as @xmath55 the functions @xmath56 and @xmath50 differ depending on the functional @xmath57 of interest
. as the functionals @xmath2 are all closely related , we shall use @xmath31 as a prototype in this introduction . for @xmath58 @xmath59 @xmath60@xmath61 and @xmath62 .
whenever a functional @xmath2 has a non - zero sib , given sufficiently stringent bounds on @xmath7 norms and on smoothness , it is possible to use the estimated first order influence function to construct regular estimators and honest asymptotic confidence intervals whose width shrinks at the usual parametric rate of @xmath63 @xmath64 we recall that , by definition , regular estimators are @xmath65-consistent .
when @xmath8 is high dimensional , the apriori smoothness restrictions on @xmath66 and @xmath67 necessary for point or interval estimators of @xmath31 to achieve the parametric rate of @xmath63 are so severe as to be substantively implausible . as a consequence ,
we replace the usual approach based on first order influence functions by one based on higher order influence functions . to provide quantitative results , we require a measure of the maximal possible complexity ( e.g. smoothness ) of @xmath68 and @xmath69 believed substantively plausible.@xmath70 @xmath70we use hlder balls for concreteness , although our methods extend to other measures of complexity .
a function @xmath71 lies in the hlder ball @xmath72 with hlder exponent @xmath73 and radius @xmath74 if and only if @xmath71 is bounded in supremum norm by @xmath75 and all partial derivatives of @xmath76 up to order @xmath77 exist , and all partial derivatives of order @xmath77 are lipschitz with exponent @xmath78and constant @xmath75 .
we make the assumption that @xmath79lie in given hlder balls @xmath80 @xmath81 furthermore , it turns out we must also make assumptions about the complexity of the function @xmath82 f_{x}\left ( x\right ) , $ ] which we take to lie in a given @xmath83 for @xmath84 using higher order influence functions , we construct regular estimators and honest ( i.e uniform over our model ) asymptotic confidence intervals for functionals @xmath2 in our class whose width shrinks at the usual parametric rate of @xmath63 whenever @xmath85 and @xmath86 this result can not be improved on , since even when @xmath87 is known apriori , @xmath88 is necessary for a regular estimator to exist . when @xmath89 and @xmath87 is known apriori@xmath90 we have shown using arguments similar to those of birge and massart ( 1995 ) that the minimax rate of convergence for an estimator and minimax rate of shrinkage of a confidence interval is @xmath91 when @xmath87 is unknown , we construct point and interval estimators with this same rate of @xmath92 whenever @xmath93 where @xmath94 for example if @xmath95 we require @xmath96exceed @xmath97 to achieve the rate @xmath98 when the previous inequality does not hold and @xmath99 we have constructed , in a yet unpublished paper , estimators that converge at rate @xmath100 ^{1/2}-\left ( 1 + 2\beta / d\right ) \right ) \rceil .\nonumber\end{aligned}\ ] ] we conjecture that this rate is minimax , possibly up to log factors . in this paper , however , the estimators we construct are inefficient when the previous inequality fails to hold , converging at rates less than the conjectured minimax rate of eq @xmath101 .
let us return to the case where @xmath33 @xmath102@xmath103@xmath103@xmath103wp1 .
then @xmath104 and @xmath105 @xmath106
so @xmath107 now , for fixed @xmath108 eq @xmath101 converges to @xmath109 as @xmath110 which agrees ( up to a log factor ) with the minimax rate of @xmath111 given by wang et al .
( 2006 ) and cai et al .
( 2006 ) under the semiparametric homoscedastic model @xmath112 with equal - spaced non - random @xmath113 this result might suggest that @xmath114being random rather than equal - spaced can result in faster rates of convergence only when the density of @xmath8 has some smoothness , as quantified here by @xmath86 but this suggestion is not correct . recall that we obtained the rate @xmath115 for @xmath104 as @xmath116 under a non - parametric model . in section [ minimax_section ]
, we construct a simple estimator of @xmath117 under the homoscedastic model with @xmath8 random with unknown density that , for @xmath118 , @xmath119 and without smoothness restrictions on @xmath9 , converges at the rate @xmath120 which is faster than the equal - spaced non - random minimax rate of @xmath121 the paper is organized as follows . in section 2 , we define the higher order ( estimation ) influence functions of a functional @xmath2 for @xmath4 contained in a model @xmath122 and prove two fundamental theorems - the extended information equality theorem and the efficient estimation influence function theorem .
further , in the context of a parametric model whose dimension increases with sample size , we outline why estimators based on higher order influence can outperform those based on first order influence functions in high - dimensional models . in section 3
, we introduce the class of functionals we study in the remainder of the paper and describe their importance in biostatistics .
the theory of section 2 , however , is not directly applicable to these functionals because they have first order but not higher order influence functions .
we show that higher order influence functions fail to exist precisely because the dirac delta function is not an element of the hilbert space @xmath123 of square integrable functions .
we describe two approaches to overcoming this difficulty .
the first approach is based on approximating the dirac delta function by a projection operator onto a subspace of @xmath123 of dimension @xmath124 where @xmath125 can be as large as the square of the sample size @xmath126 the second approach is based on approximating the functional @xmath2 by a truncated functional @xmath127 .
the truncated functional has influence functions of all orders , is equal to @xmath2 if either a @xmath128 dimensional working parametric model ( with @xmath129 for the function @xmath130 or the function @xmath131 in eq .
@xmath132 is correct , and remains close to @xmath2 even if both working models are misspecified .
we then use higher order influence function based estimators of @xmath133 as estimators of @xmath57 .
these estimators @xmath134 are asymptotically normal with variance and bias for @xmath2 depending both on the choice of the dimension @xmath128 of the working models and on the order @xmath135 of the influence function of @xmath136 we show that these same estimators @xmath134 can also be obtained under the approximate dirac delta function approach .
we derive the optimal estimator @xmath137 in the class as a function of the hlder balls in which the functions @xmath138 and @xmath139 are assumed to lie .
finally we conclude section 3 by showing that the estimators @xmath140 have a multi - robustness property that extends the double - robustness property of the first order influence function estimator @xmath141 in section 4 , we consider whether the estimators @xmath142 are rate - minimax .
we show that whenever @xmath143and @xmath144,@xmath145 is not only rate minimax but is semiparametric efficient .
further , by letting the order @xmath146 of the u - statistic depend on sample size , we construct a single estimator @xmath147 that is semiparametric efficient for all @xmath88 even when @xmath148 can not be estimated at an algebraic rate .
we show , however , that when @xmath149 does not in general converge at the minimax rate . in section 4.1 , however , we construct a new estimator @xmath150 that converges at the minimax rate of @xmath151 whenever eq .
@xmath152 holds . in section 5
, we use the results obtained in earlier sections to construct adaptive confidence intervals for a regression function @xmath153 $ ] when the marginal of @xmath8 is unknown and for the treatment effect function and optimal treatment regime in a randomized clinical trial . in section 6.1 , we discuss how to obtain higher order u - statistic point estimators and confidence intervals for functionals @xmath154 that are implicitly defined as the solution to an equation @xmath155 in section 6.2 , we define higher order testing influence functions and efficient scores and describe their relationship to the higher order estimation influence functions and efficient influence functions of section 2 . finally ,
in section 6.3 , we discuss the relationship between the higher order u - statistic point estimators of an implicitly defined functional @xmath154 and higher order testing influence functions . before proceeding ,
several additional comments are in order . in this paper
, we investigate the asymptotic properties of our higher order u - statistic point and interval estimators .
the reader is referred to li et al ( 2006 ) for an investigation of the finite sample properties of our procedures through simulation .
in addition , precise regularity conditions are sometimes omitted from both the statements and the proofs of various theorems .
this reflects the fact that the goal of this paper is to provide a broad overview of our theory as it currently stands .
different subject matter experts will clearly disagree as to the maximum possible complexity of @xmath156,@xmath157 and @xmath158 thus it is important to have methods that adapt to the actual smoothness of these functions .
elsewhere , we plan to provide point estimators that optimally adapt to unknown smoothness .
in contrast to point estimators , however , for honest confidence intervals , the degree of possible adaption to unknown smoothness is small .
therefore we propose that an analyst should report a mapping from apriori smoothness assumptions encoded in the exponents and radii of hlder balls ( or in other measures of complexity ) to the associated optimal @xmath159 honest confidence intervals proposed in this paper .
such a mapping is finally only useful if substantive experts can approximately quantify their informal opinions concerning the shape and wiggliness of @xmath160and @xmath139 using the measure of complexity on offer by the analyst .
it is an open question which , if any , complexity measure is suitable for this purpose .
finally , most of our mathematical results concern rates of convergence .
we offer only a few results on the constants in front of those rates .
this is not because the constant is less important than the rate in predicting how a proposed procedure will perform in the moderate sized samples occurring in practice .
rather , at present , we do not possess the mathematical tools necessary to obtain useful results concerning constants .
a more extended discussion of the issue is found in section 3 of li et al .
( 2006 ) . in the following , we use @xmath161 to mean @xmath162 and @xmath163 @xmath164 to mean @xmath165 @xmath166 and @xmath167 @xmath168 to respectively mean @xmath169 @xmath170 as @xmath171
given @xmath5 i.i.d observations @xmath172 from a model @xmath173 we consider inference on a nonlinear functional @xmath174 in general , @xmath175 can be infinite dimensional but for now we only consider the one dimensional case . in the following all quantities
can depend on the sample size @xmath176 including the support of @xmath177 the parameter space @xmath178 and the functional @xmath175 .
we generally suppress the dependence on @xmath5 in the notation
. we will be particularly interested in models in which the parameter @xmath179 is infinite dimensional and @xmath180 and @xmath181 do not depend on @xmath126 we also briefly discuss models in which subvectors of @xmath179 are finite - dimensional parameters whose dimension @xmath182 increases as power @xmath183 ( often @xmath184 of @xmath5 and thus @xmath185 and @xmath186 depend on @xmath126 our first task is to define higher order influence functions . before proceeding we recall some facts about @xmath187statistics .
consider a function @xmath188 where we often suppress @xmath189 subscript @xmath190 for integers @xmath191lying in @xmath192 we define @xmath193 and@xmath194 \mathbf{\equiv}\frac{(n - m)!}{n!}\sum\limits_{i_{1}\neq i_{2} ... \neq i_{m}}b_{m , i_{1}, ... ,i_{m}}.\ ] ] in an abuse of notation , we will consider the following expressions to be equivalent @xmath195 \mathbf{\equiv}\mathbb{v}_{n}\left [ b_{m,,i_{1}, ... ,i_{m}}\right ] \mathbf{\equiv}\mathbb{v}_{n}\left [ b_{m\ } \right ] .\ ] ] thus @xmath196 $ ] is a @xmath197 order u - statistic with kernel @xmath198 .
we do not assume that @xmath198 is symmetric .
we will write @xmath199 $ ] as @xmath200 so , suppressing the dependence on @xmath176 @xmath201 $ ] .
any @xmath202has a unique ( up to permutation ) decomposition @xmath203 under any @xmath204 as a sum of degenerate u - statistics @xmath205 where degeneracy of @xmath206 means that @xmath207 satisfies @xmath208 = 0,l=1, .... ,s\ ] ] where upper and lower case letters , respectively , denote random variables and their possible realizations .
let @xmath209 be the hilbert space of all @xmath187statistics of order @xmath135 with mean zero and finite variance with inner product defined by covariances with respect to the @xmath5-fold product measure @xmath210 note that any @xmath187statistic @xmath211 of order @xmath212 , @xmath213 is also an mth order @xmath187 statistic with @xmath214 identically zero for @xmath215 . since any two degenerate @xmath187 statistics of different orders
are uncorrelated , the @xmath209-hilbert space projection of @xmath216 on @xmath217 is @xmath218 for @xmath219 thus a @xmath187statistic @xmath216 is degenerate @xmath220 @xmath221 @xmath222 @xmath220 @xmath223 = 0\leftrightarrow\mathbb{b}_{m}\in\mathcal{u}_{m-1}\left ( \theta\right ) ^{\perp_{m,\theta}},\ $ ] where @xmath224 \mathbf{\equiv}\pi_{\theta , m}\left [ \cdot|\cdot\right ] $ ] is the projection operator of the hilbert space @xmath225 ( with the dependence on @xmath135 suppressed when no ambiguity can arise ) and , for any linear subspace @xmath226 of @xmath227 @xmath228 is its orthocomplement in the hilbert space @xmath229 given any @xmath230 , $ ] @xmath231 is explicitly given by @xmath232 $ ] where @xmath233 maps @xmath234 to @xmath235 define for @xmath236 @xmath237 with @xmath238 where the @xmath239 symbol denotes differentiation by the variables occurring to its right and the overbar @xmath240 denotes the vector @xmath241 .
given a sufficiently smooth @xmath242dimensional parametric submodel @xmath243 mapping @xmath244 injectively into @xmath245 , define for @xmath179 in the range of @xmath246 @xmath247 and @xmath248 where @xmath249 is the density of @xmath250wrt a dominating measure .
that is @xmath251 and @xmath252 are higher order derivatives of @xmath253and @xmath254 under a parametric submodel @xmath255 where the model @xmath256 has been suppressed in the notation .
an @xmath257 order score associated with the submodel @xmath243 is defined to be @xmath258 where @xmath259 is a u - statistic of order @xmath260 to understand why @xmath259 is a @xmath187statistic we provide formulae for an arbitrary score @xmath259 in terms of the subject specific scores @xmath261
@xmath262 for @xmath263 suppressing the @xmath264dependence , results in waterman and lindsay ( 1996 ) imply@xmath265 @xmath266 @xmath267 note these equations express each @xmath268 as a sum of degenerate u - statistics .
we now define a @xmath197 order estimation influence function @xmath269 for @xmath270 where we suppress the dependence on @xmath271 when no ambiguity will arise .
a u - statistic @xmath272 of order @xmath135 and finite variance is said to be an @xmath197 order estimation influence function for @xmath175 if ( i ) @xmath273 = 0,$ ] @xmath274 and ( ii ) for @xmath275 and every suitably smooth and regular ( see appendix ) @xmath276 dimensional parametric submodel @xmath277@xmath278 .\ ] ] estimation influence functions need not always exist , but when they do they are useful for deriving point estimators of @xmath271 with small bias and for deriving confidence interval estimators centered on an estimate of @xmath279 we will generally refer to estimation influence functions simply as influence functions .
we remark that @xmath272 is an influence function under the above definition if and only if it is one under the modified version in which the dimension of the parametric submodel @xmath243 is unrestricted .
a key result is the following theorem which is related to results of small and mcleish ( 1994 ) .
[ eiet]*extended information equality theorem : * given a @xmath197 order influence function @xmath280 for any smooth and regular submodels @xmath243 and @xmath281@xmath282 /\partial\zeta_{l_{1}} ... \partial \zeta_{l_{_{s}}}|_{\zeta=\widetilde{\theta}^{-1}\left\ { \theta\right\ } } = -\psi_{\backslash\overline{l}_{s}}\left ( \theta\right)\ ] ] thus , if the functionals @xmath283 $ ] and @xmath284 $ ] have bounded frchet derivatives w.r.t .
@xmath285 to order @xmath286 for a norm
@xmath287 @xmath288 = -\left [ \psi\left ( \theta+\delta\theta\ \right ) -\psi\left ( \theta\right ) \right ] + \text { } o\left ( \left\vert \left\vert \delta \theta\ \right\vert \right\vert ^{m+1}\right)\ ] ] since the functions @xmath289 $ ] and @xmath290 $ ] of @xmath285 have the same taylor expansion around @xmath179 up to order @xmath190 the proof is in the appendix .
define the @xmath135th order tangent space @xmath291 at @xmath179 for the model @xmath292 to be the subspace of @xmath293 formed by taking the closed linear span of all scores of order @xmath135 or less as we vary over all regular parametric submodels @xmath294 ( whose range includes @xmath295 of our model @xmath296 we say a model is ( locally ) nonparametric for @xmath197 order inference if @xmath297 given any @xmath197 order estimation influence function @xmath298 define the mth order efficient estimation influence function to be @xmath299\ ] ] @xmath70where @xmath300 \mathbf{\equiv}\pi _ { \theta , m}\left [ \cdot|\cdot\right ] $ ] is the @xmath301projection operator . in the appendix , we prove the following : [ eift]*efficient estimation influence function theorem :* 1 .
@xmath302 is unique in the sense that for any two mth order influence functions @xmath303
\text { a.s.}\ ] ] * * 2 .
@xmath302 is a mth order estimation influence function and has variance less than or equal to any other @xmath197 order estimation influence function .
@xmath272 is a @xmath197 order estimation influence function if and only if @xmath70@xmath304 where @xmath305 is the ortho - complement of @xmath306 in @xmath229 4 .
if @xmath272 exists then @xmath307exists for @xmath308 and @xmath309 = \mathbb{if}_{s}^{eff}\left ( \theta\right ) .$ ] 5 . if the model @xmath310 is ( locally ) nonparametric , then 1 .
there is at most one @xmath135th order estimation influence function @xmath272 for @xmath311 2 .
@xmath312 where@xmath313\ ] ] and @xmath314 is a degenerate @xmath197 order u - statistic and thus @xmath315 = 0.\ ] ] 3 .
( i ) : suppose , for a given @xmath316 @xmath317 exists and a kernel @xmath318 of @xmath319 has a first order influence function with kernel @xmath320 for all @xmath321 in a set @xmath322 which has probability @xmath18 under @xmath323 then @xmath272 exists and @xmath324 \right ) \label{mm1}\ ] ] where the operator @xmath325 is given in @xmath326 + \(ii ) conversely , if @xmath327 exists then the symmetric kernel @xmath328 of @xmath329 has a first order influence function for all @xmath321 in a set @xmath322 which has probability @xmath18 under @xmath330 further @xmath331 = if_{m , m}^{sym}\left ( o_{i_{1}}, ... ,o_{i_{m}};\theta\right ) .\ ] ] pfanzagl ( 1990 ) previously proved part 5.c(i ) for @xmath332 our theorem offers a generalization of his result .
note , in part ( i ) of 5(c ) , we can always take the kernel to be the symmetric kernel .
[ proj]provided one knows how to calculate first order influence functions , one can obtain @xmath333 recursively using part @xmath334 an example of such a calculation is given in section [ dhoif_subsection ] below .
thus part @xmath335 has the interesting implication that even though higher order influence functions are defined in terms of their inner products with higher order scores @xmath336 nevertheless , in ( locally ) nonparametric models , one can derive all the higher order influence functions of a functional @xmath175 without even knowing how to compute the scores @xmath268 for any @xmath337 in fact , one need not even be aware of the structure of the scores @xmath268 in terms of the subject - specific higher order scores @xmath338 in contrast , in parametric or semiparametric models whose tangent space @xmath339 does not equal the set @xmath225 of all @xmath135th order @xmath340 , one can often ( but not always ) still obtain an inefficient influence @xmath341 by applying part @xmath335 of the theorem .
however , calculation of the efficient influence function @xmath342 $ ] by projection generally requires explicit knowledge of the scores @xmath343to derive @xmath344 for this reason , it can be considerably more difficult to analyze certain parametric models ( with dimension increasing with sample size ) than to analyze ( locally ) nonparametric models .
we will consider derivation of and projections onto @xmath345 in a forthcoming paper . in the current paper , however , we do calculate @xmath346 in one model that is not ( locally ) nonparametric so as to provide some sense of the issues that arise . specifically in section [ minimax_section ] , we calculate @xmath347 for the functional @xmath348 \right\ } ^{2}\right ] $ ] in a model that assumes the marginal distribution of @xmath8 is known .
* [ jj1 ] : * * implications of theorem * @xmath349 * * for the variance of unbiased estimators : * * suppose we have @xmath5 iid draws @xmath350 from @xmath351and a u - statistic @xmath352 of order @xmath353 with @xmath354 <
\infty$ ] for @xmath355satisfying @xmath356 = \psi\left ( \theta\right ) $ ] for all @xmath357 that is , @xmath352 is unbiased for @xmath175 .
we will use theorem @xmath358 to generalize a number of well - known results on minimum variance unbiased estimation to arbitrary models .
by @xmath356 = \psi\left ( \theta\right ) , $ ] we immediately conclude that , viewing @xmath359as a kth order u - statistic , @xmath360 is a @xmath361 order estimation influence function for @xmath175 for @xmath362 by theorem @xmath363
@xmath364 \geq var_{\theta}\left [ \mathbb{if}_{m}^{eff}\left ( \theta\right ) \right ] .$ ] we refer to @xmath365 $ ] as the @xmath197 order bhattacharyya variance bound at @xmath179 for the parameter @xmath270 in model @xmath366 as this definition , in a precise analogy to bickel et al .
( 1993 ) s generalization of the cramer - rao variance bound , generalizes bhattacharyya s ( 1947 ) variance bound to arbitrary semi- and non- parametric models .
indeed our 1st order bhattacharyya bound is precisely bickel et al.s ( 1993 ) generalization of the cramer - rao variance bound .
we shall refer to an mth order u - statistic estimator @xmath352 as mth order ` unbiased locally efficient ' at @xmath285 for @xmath270 in model @xmath310 if it is unbiased for @xmath175 under the model with variance at @xmath285 equal to the @xmath197 order bhattacharyya bound at @xmath367 if @xmath352 is ` unbiased locally efficient ' at @xmath368 for all @xmath369 we say it is ` unbiased globally efficient ' . by theorem @xmath363
@xmath370 \geq var_{\theta}\left [ \mathbb{if}_{m}^{eff}\left ( \theta\right ) \right ] $ ] for @xmath371 as a consequence if an mth order ` unbiased locally efficient ' estimator @xmath372 exists at @xmath285 then , for @xmath373 @xmath374 so the @xmath197 and @xmath375 order bhattacharyya bounds are equal at @xmath285 and @xmath376 is also kth order ` unbiased locally efficient ' at @xmath285 . from the fact that for , an unbiased estimator @xmath377 @xmath360 is an @xmath197 order influence function ,
we conclude that the variance of @xmath352 attains the the bound @xmath378 $ ] at @xmath285 if and only if @xmath379 it follows that @xmath352 is ` unbiased globally efficient ' if and only if @xmath380 for all @xmath357 we thus have proved the following theorem in the @xmath381 direction .
the @xmath382 direction is immediate .
[ global_eff ] * : * in a model @xmath366 there exists an mth order unbiased globally efficient u - statistic estimator of @xmath311 if and only if , for all @xmath383 @xmath384 is a function @xmath372 of the data @xmath385 not depending on @xmath386 in that case , @xmath372 is the unique unbiased globally efficient estimator . in a locally nonparametric model all unbiased mth order estimators are unbiased globally efficient , as there is a unique @xmath197 order influence function .
for example , the usual unbiased estimator @xmath387 of the variance of a random variable @xmath8 is a second order u - statistic and thus is a kth order unbiased globally efficient u - statistic for @xmath388 in the locally nonparametric model consisting of all distributions under which @xmath389 has a finite variance . in section [ minimax_section ]
we use the results from this remark to compare the relative efficiencies of competing rate - optimal unbiased estimators in a model which is not locally nonparametric .
we now describe the main heuristic idea behind using higher order influence functions .
technical details are suppressed .
consider the estimator @xmath390 based on a sample size @xmath176 where @xmath391 is an initial rate optimal estimator of @xmath179 from a separate independent training sample .
that is we assume that our actual sample size is @xmath392 and we randomly split the @xmath392 observations into two samples : an analysis sample of size @xmath5 and a training sample of size @xmath393 where @xmath394 @xmath90 @xmath395we obtain our initial estimate @xmath391 from the training sample data .
sample splitting has no effect on optimal rates of convergence , although in the form described here does affect constants. throughout the paper , we derive the properties of our estimators conditional on the data in the training sample . in a later section ,
we describe how one can sometimes obtain an optimal constant by choosing @xmath396 rather than @xmath397 [ semi ] note that sample splitting is avoided in most statistical applications by using modern empirical process theory to prove that ` plug - in ' estimators such as @xmath398 that estimate @xmath179 from the same sample used to calculate @xmath399 have nice statistical properties .
however empirical process theory is not applicable in our setting because we are interested in function classes whose size ( entropy ) is so large that they fail to be donsker .
for this reason we initially believed that explicit sample splitting would be difficult to avoid in our methodology .
however , in robins et al .
( 2007 ) , we describe a new method , more analogous to the jackknife than to sample splitting , that effectively allows one to use all the data for estimator construction . expanding and evaluating conditionally on the training sample ( or equivalently on @xmath400
we find by theorem @xmath401 that the conditional bias @xmath402\ ] ] is @xmath403which decreases with @xmath404provided @xmath405 in theorem [ 3.19 ] below , we show that if @xmath406 as @xmath407 where @xmath408 is the density of @xmath409 under @xmath179 and @xmath410 has probability one under all @xmath383 then @xmath411 \equiv var_{\theta}\left [ \mathbb{if}_{m}^{eff}\left ( \widehat{\theta}\right ) now , by theorem [ eift ] , @xmath412 $ ] increases with @xmath135 .
further , @xmath413
\asymp1/n$ ] , since , conditional on @xmath414 @xmath415 is the sample average of @xmath416 random variables . to proceed further
we shall need to be more explicit about the model @xmath296 for now , we consider finite - dimensional parametric models whose dimension @xmath128 increases with sample size .
that is @xmath417 depends on @xmath5 and the dimension of @xmath418 is @xmath128 .
suppose @xmath419 let @xmath420 be the maximum likelihood estimator of @xmath386 if @xmath128 increases slower than the sample size ( i.e. , @xmath421 then , a ) under regularity conditions , @xmath422 with @xmath423 the usual euclidean norm in @xmath424 and b ) @xmath412 , \ $ ] although increasing with @xmath425 remains order @xmath426 as a consequence , if @xmath404is chosen greater than the solution @xmath427 to @xmath428 the bias of @xmath352 will be @xmath429 the rate of convergence will be the usual parametric rate of @xmath430 and thus , for @xmath5 sufficiently large , the squared bias of @xmath352 will be less than the variance . as a consequence , as discussed in section [ dr_ci_section ] , we can construct honest ( i.e uniform over @xmath431 asymptotic confidence intervals centered at @xmath432 with width of order @xmath433 here is a concrete example .
* example : * suppose @xmath434 with @xmath0 bernouilli and with @xmath114having a density with respect to the uniform measure @xmath435 on the unit cube @xmath436 ^{d}$ ] in @xmath437 suppose @xmath438 \right ) ^{2}\right ] .$ ] let @xmath439 be a countable , linearly independent , sequence of either spline , polynomial , or compact wavelet basis functions dense in @xmath440 set @xmath441 we assume @xmath442{c}b\left ( x;\overline{\eta}_{k^{\ast}\left ( n\right ) } \right ) \equiv\left [ 1+\exp\left ( -\overline{\eta}_{k^{\ast}\left ( n\right ) } ^{t}\overline { z}_{k^{\ast}\left ( n\right ) } \left ( x\right ) \right ) \right ] ^{-1};\\ \overline{\eta}_{k^{\ast}\left ( n\right ) } \in\mathcal{n}_{k^{\ast}\left ( n\right ) } \end{array } \right\ } , \ ] ] @xmath443{c}f\left ( x;\overline{\omega}_{k^{\ast\ast}\left ( n\right ) } \right ) \equiv c\left ( \overline{\omega}_{k^{\ast\ast}\left ( n\right ) }
\right ) \exp\left [ \overline{\omega}_{k^{\ast\ast}\left ( n\right ) } ^{t}\overline { z}_{k^{\ast\ast}\left ( n\right ) } \left ( x\right ) \right ] ; \\ \overline{\omega}_{k^{\ast\ast}\left ( n\right ) } \in\mathcal{w}_{k^{\ast\ast } \left ( n\right ) } \end{array } \right\}\ ] ] where @xmath444 is a normalizing constant and @xmath445 and @xmath446 are open bounded subsets of @xmath447 and @xmath448 hence , @xmath449 has dimension @xmath450and @xmath451 he ( 2000 ) and portnoy ( 1988 ) prove that , under regularity conditions , @xmath452 when @xmath453 below we shall see that @xmath454 \asymp1/n$ ] for @xmath455 consider next models whose dimension @xmath456 increases faster than @xmath5 ( i.e. , @xmath457 in such models , the mle @xmath420 is generally inconsistent and indeed there may exist no consistent estimator of @xmath458 . in that case , @xmath459fails to be @xmath460 and the conditional bias @xmath461exist@xmath90 it is necessary to place further apriori restrictions on the complexity of @xmath462 typical examples of complexity - reducing assumptions would be an @xmath463sparseness assumption that only @xmath464 of the @xmath128 parameters are non - zero or a smoothness assumption that specifies that the rate of decrease of the @xmath465component of @xmath458 is equal to @xmath466 raised to a given ( positive ) power . even after imposing such complexity - reducing assumptions , @xmath467 may not be estimable at rate @xmath63 .
for instance consider the previous example but now with @xmath468 and @xmath469 exceeding @xmath470 so @xmath471 , @xmath472 and @xmath473with @xmath474 consider the norms @xmath475 = @xmath476@xmath477 and @xmath478 .
suppose , under a particular smoothness assumption , optimal rate estimators @xmath479 and @xmath480 of @xmath481 and @xmath482 satisfy @xmath483 and @xmath484for some @xmath485 and all @xmath486 hence , @xmath487 for @xmath488 based on arguments given later , we expect that @xmath489 \asymp\frac{n^{\left ( \gamma -1\right ) \left ( m-1\right ) } } { n}$ ] and @xmath490 = o_{p}\left ( \left\vert \widehat{\overline{\eta}}_{k^{\ast}\left ( n\right ) } -\overline{\eta}_{k^{\ast}\left ( n\right ) } \right\vert ^{2}\left\vert \widehat{\overline{\omega}}_{k^{\ast\ast}\left ( n\right ) } -\overline{\omega } _ { k^{\ast\ast}\left ( n\right ) } \right\vert _ { m-1}^{m-1}\right ) = o_{p}\left ( n^{-2\gamma_{\eta}-\left ( m-1\right ) \gamma_{\omega}}\right ) $ ] @xmath491 to find the estimator @xmath492 in the class @xmath352with optimal rate of convergence@xmath90 let @xmath493 be the value of @xmath135 that equates the order @xmath494 of the squared bias and the order @xmath495 of the variance .
then @xmath496 if the order @xmath497of the mean squared error@xmath70at @xmath498 is less than or equal to that at @xmath499 otherwise , @xmath500 .
the rate of convergence of @xmath492 will often be slower than @xmath433 note @xmath501 whenever @xmath502 regardless of @xmath503 and @xmath504 by using the estimator @xmath505 rather than @xmath506we can guarantee that the variance asymptotically dominates bias and construct honest ( i.e uniform over @xmath431 asymptotic confidence intervals centered at @xmath507 of course , the sample size @xmath5 at which , for all @xmath508 the finite sample coverage of the intervals discussed above is close to the asymptotic ( i.e. nominal ) coverage is generally unknown and could be very large .
for this reason , a better , but unfortunately as yet technically out of reach , approach to confidence interval construction is discussed in section [ dr_ci_section ] .
in contrast to the case of parametric models of increasing dimension , in the infinite dimensional models which we consider in the following section , the functionals @xmath175 of interest have first order influence functions @xmath509 but do not have higher order influence functions . as a consequence ,
an initial truncation step is needed before we can apply the approach outlined in the preceding paragraph .
finally , even in the case of parametric models with @xmath510 and complexity reducing assumptions imposed , , when the minimax rate for estimation of @xmath175 is slower than @xmath430 the optimal estimator @xmath492 in the class @xmath511 will generally not be rate minimax .
see section 3.2.6 and sections 4.1.1 for additional discussion .
in this section we consider models in which the parameter @xmath179 is infinite dimensional and @xmath180 and @xmath181 do not depend on @xmath126 we make the following 4 assumptions @xmath512 : ai ) the data @xmath513includes a vector @xmath514where , for all @xmath515 the distribution of @xmath114 is supported on the unit cube @xmath436 ^{d}$ ] ( or more generally a compact set ) in @xmath516and has a density @xmath517 with respect to the lebesgue measure .
further @xmath518 where @xmath519 governs the marginal law of @xmath8 and @xmath520 governs the conditional distribution of @xmath521 aii ) the parameter @xmath522 contains components @xmath523 and @xmath524 , @xmath525 ^{d}\mathcal{\rightarrow r}$ ] and @xmath526 ^{d}\mathcal{\rightarrow r},$ ] such that the functional @xmath175 of interest has a first order influence function @xmath527 , $ ] where @xmath528 and the known functions @xmath529do not depend on @xmath386 aiii ) \a ) @xmath530 where @xmath531 and @xmath532 are the parameter spaces for the functions @xmath533 and @xmath534 .
furthermore the sets @xmath535 and @xmath532 are dense in @xmath536 at each @xmath537 or \b ) @xmath538 , @xmath539and @xmath540 is dense in @xmath541 at each @xmath537 * remark : * aiiib ) can be viewed as a special case of aiiia ) as discussed in example 1a below , so we need only prove results under assumption aiiia ) .
assumptions ai)-aiii ) have a number of important implications that we summarize in a theorem and two lemmas .
[ dr]double - robustness : assume ai)-aiii ) hold , and recall @xmath542and @xmath533 are elements of @xmath386 then @xmath543 = e_{\theta}\left [ h\left ( b^{\ast},p\right ) \right ] = e_{\theta}\left [ h\left ( b , p\right ) \right ] = \psi\left ( \theta\right)\ ] ] for all @xmath544 : @xmath545 -e_{\theta}\left [ h\left ( b , p\right ) \right ] = e_{\theta}\left [ \left\ { h_{1}p\left ( x\right ) + h_{2}\right\ } \left\ { b\left ( x\right ) -b^{\ast}\left ( x\right ) \right\ } \right ] $ ] and @xmath546 -e_{\theta}\left [ h\left ( b , p\right ) \right ] = e_{\theta}\left [ \left\ { h_{1}b\left ( x\right ) + h_{3}\right\ } \left\ { p\left ( x\right ) -p^{\ast}\left ( x\right ) \right\ } \right ] .$ ] the theorem then follows from part 1 ) of the following lemma .
theorem [ dr ] states that @xmath547 has mean @xmath175 under @xmath548 even when @xmath549 is misspecified as @xmath53 or @xmath533 is misspecified as @xmath550 we refer to the functional @xmath175 as doubly robust because of this property .
the next lemma shows that @xmath551 is not unbiased if both @xmath533 and @xmath534 are simultaneously misspecified .
that is , @xmath552 \neq\psi\left ( \theta\right ) .$ ] [ conde ] assume ai)-aiii ) hold . then 1
. @xmath553 = e_{\theta } \left [ \left\ { h_{1}p+h_{2}\right\ } |x\right ] = 0 $ ] 2 .
@xmath554 -e_{\theta}\left [ h\left ( b , p\right ) \right ] = e_{\theta}\left [ \left ( b - b^{\ast}\right ) \left ( p - p^{\ast}\right ) h_{1}\right ] $ ] + and @xmath555 = e_{\theta}\left [ -bph_{1}+h_{4}\right ] $ ] part 1 ) : by assumptions @xmath556 and @xmath557 we have paths @xmath558 in our model with @xmath559 and@xmath560 for @xmath561 where the sequence @xmath562 is dense in @xmath563 .$ ] let @xmath564be the score for path @xmath565 at @xmath566 then by @xmath567 $ ] @xmath568 \\ & + e_{\theta}\left [ h\left ( b , p\right ) s_{l}\left ( \theta\right ) \right]\end{aligned}\ ] ] by @xmath569=@xmath570@xmath571\ ] ] thus @xmath572 = 0.$ ] but @xmath573 is dense in @xmath574 \ $ ] so @xmath575 = 0\ ] ] an analogous argument proves latexmath:[$e_{\theta}\left [ \left\ { h_{1}p+h_{2}\right\ } @xmath552 -e_{\theta}\left [ h\left ( b , p\right ) \right ] = $ ] @xmath577 \\ & = e_{\theta}\left [ \left ( b^{\ast}p^{\ast}-bp\right ) h_{1}-\left ( b^{\ast}-b\right ) ph_{1}-\left ( p^{\ast}-p\right ) bh_{1}\right ] \\ & = e_{\theta}\left [ \left ( b - b^{\ast}\right ) \left ( p - p^{\ast}\right ) h_{1}\right]\end{aligned}\ ] ] where the second equality is by part 1 ) .
choosing @xmath578 wp1 completes the proof of the theorem since then @xmath579 = e_{\theta}\left [ h_{4}\right ] $ ] .
below we will need the following partial converse to lemma [ conde ] .
[ if1 ] let @xmath580 and @xmath245 and @xmath581 be as defined in ai)- aiiia ) .
@xmath582 = e_{\theta}\left [ \left\ { h_{1}p+h_{2}\right\ } |x\right ] = 0\ ] ] and @xmath583 .$ ] then @xmath584 \ $ ] is the first order influence function of @xmath174 : the influence function of the functional @xmath579 $ ] for known functions @xmath585 is @xmath586 \right ] .$ ] thus by the linearity of first order influence functions , the lemma is true if and only if for each @xmath587 the functional @xmath588 $ ] with @xmath589 fixed has influence function equal to @xmath590 wp1 at @xmath591 that the influence function is equal to @xmath590 follows from the fact that , under the assumptions of the lemma , for sets @xmath592 and @xmath593 dense in @xmath594 $ ] , @xmath595 /dt_{|t=0}\\ & = e_{\theta}\left [ \left\ { h_{1}b_{0}\left ( x\right ) + h_{3}\right\ } d_{l}\left ( x\right ) \right ] + e_{\theta}\left [ \left\ { h_{1}p_{0}\left ( x\right ) + h_{2}\right\ } c_{l}\left ( x\right ) \right ] = 0\end{aligned}\ ] ] results of ritov and bickel ( 1990 ) and robins and ritov ( 1997 ) imply it is not possible to construct honest asymptotic confidence intervals for @xmath270 whose width shrinks to @xmath590 as @xmath596 if @xmath597 and @xmath598 are too rough@xmath64 therefore we also place apriori bounds on their roughness .
our bounds will be based on the following definition . a function @xmath599 with domain @xmath436 ^{d}$ ]
is said to belong to a hlder ball @xmath72 with hlder exponent @xmath600 and radius @xmath74 if and only if @xmath71 is uniformly bounded by @xmath75 , all partial derivatives of @xmath599 up to order @xmath77 exist and are bounded , and all partial derivatives @xmath601 of order @xmath77 satisfy @xmath602 ^{d}}\left\vert \nabla^{\left\lfloor \beta\right\rfloor } h(x+\delta x)-\nabla^{\left\lfloor \beta\right\rfloor } h(x)\right\vert \leq c||\delta x||^{\beta-\left\lfloor \beta\right\rfloor } .\ ] ] we note that the @xmath603 and @xmath604 rates of convergence for estimation of a marginal density or conditional expectation @xmath605 @xmath606 are @xmath607 and @xmath608 respectively@xmath64 we refer to an estimator attaining these rates as rate optimal .
aiv ) we assume @xmath609 and g@xmath610 lie in given hlder balls @xmath80 @xmath611 @xmath612 where @xmath613 furthermore we assume @xmath614 wp1 . finally we assume
, as can always be arranged by a suitable choice of estimator , that the initial training sample estimators @xmath615 and @xmath616 are rate optimal , have more than max@xmath617 derivatives , and have @xmath604 norm bounded by a constant @xmath618 further inf@xmath619 ^{d}}$ ] @xmath620 the reason for the restrictions on @xmath148 will become clear below . the restrictions @xmath512 are the only restrictions common to all functionals and models in the class . additional model and/or functional specific restrictions will be given below . to motivate our interest in such a class of functionals and models we provide a number of examples .
in each case , one can use lemma @xmath621 to verify that the influence function of @xmath270 is as given .
all but examples 3 and 4 are examples of ( locally ) nonparametric models .
* example 1
: * suppose @xmath409=@xmath622with @xmath12 and @xmath0 univariate random variables .
* example 1a : * * expected product of conditional expectations : * let * * * * @xmath623 $ ] where @xmath624 , p\left ( x\ \right ) = e_{\theta}\left [ a|x\right ] .$ ] in this model @xmath625 so @xmath626 we also consider the special case of this model where @xmath627 with probability one @xmath628 . then , as in assumption @xmath629 @xmath630 @xmath631 @xmath632 wp1
. then @xmath633 .$ ] in section [ adaptive_section ] , we show how our confidence interval for @xmath634 $ ] can be used to obtain an adaptive confidence interval for the regression function @xmath635 .
* example 1b :* * expected conditional covariance * @xmath636 \ -e_{\theta}\left [ p\left ( x\ \right ) b\left ( x\ \right ) \right ] = e_{\theta}\left [ cov_{\theta}\left\ { y , a|x\right\ } \right]\ ] ] has influence function @xmath637 so @xmath638 .
the next example 1c shows that a confidence interval and point estimators for @xmath639 $ ] can be used to obtain confidence intervals and point estimator for the variance weighted average treatment effect in an observational study .
* example 1c : * * variance - weighted average treatment effect : * suppose , in an observational study , @xmath640 , @xmath12 is a binary treatment taking values in @xmath641 , @xmath642 is a univariate response and @xmath8 is a vector of pretreatment covariates . consider the parameter @xmath643 given by:@xmath644 } { e_{\theta}\left [ \ var_{\theta}(a|x)\right ] } = \frac{e_{\theta}\left [ \ cov_{\theta}(y^{\ast},a|x)\right ] } { e_{\theta } \left [ \ \pi\left ( x\right ) \left\ { 1-\pi\left ( x\right ) \right\ } \right ] } , \text { } \ ] ] where @xmath645 is often referred to as the propensity score .
we are interested in @xmath646 for several reasons .
first , in the absence of confounding by unmeasured factors , @xmath643 is the variance - weighted average treatment effect since @xmath643 can be rewritten as @xmath647 $ ] where @xmath648 } $ ] and @xmath649 is the average conditional treatment effect at level @xmath650 of the covariates .
second , under the semiparametric model @xmath651 that assumes the treatment effect does not depend on @xmath652 @xmath653 however since the model @xmath654 may not hold and therefore the parameter @xmath655 may be undefined , we choose to make inference on @xmath646 without imposing @xmath656 . now if for any @xmath657 we define @xmath658 to be @xmath659\ ] ] with @xmath660 it is easy to verify that @xmath643 may also be characterized as the solution @xmath661 to the equation @xmath662 .
thus inference on @xmath646 is easily obtained from inference on @xmath663 in particular a @xmath664confidence set for @xmath643 is the set of @xmath665such that a @xmath664ci interval for @xmath666 contains @xmath667therefore , with no loss of generality , we consider the construction of a @xmath664ci for @xmath668 for a fixed value @xmath669 , and write @xmath670 and @xmath671 thus @xmath672 $ ] and we are in the setting of example 1b .
in section [ testing_section]@xmath90 we show the rates at which the width of the confidence sets for @xmath673 and for @xmath643 shrink with @xmath5 are equal@xmath64 * example 2a : * * missing at random : * suppose @xmath674 where @xmath0 is an outcome that is not always observed , @xmath675is the binary missingness indicator , @xmath8 is a @xmath676dimensional vector of always observed continuous covariates , and let @xmath677 be the propensity score , and @xmath678 .
we suppose @xmath679and define @xmath680 = e_{\theta}\left [ b\left ( x\right ) \right]\ ] ] interest in @xmath175 lies in the fact that @xmath270 is the marginal mean of @xmath0 under the missing ( equivalently , coarsening ) at random ( mar ) assumption that @xmath681 in this model @xmath682 so @xmath683 .
note that if one has assumed apriori that @xmath684 and @xmath685 lay in hlder balls with respective exponents @xmath686 and @xmath687 then @xmath688 would be @xmath689 since @xmath690 .
* example 2b : * * missing not - at random : * consider again the setting of example 2a but we no longer assume mar .
rather we assume @xmath691 \right\ } \right\ } ^{-1}\ ] ] may depend on @xmath0 , where now @xmath692 is an unknown function and @xmath693 is a known constant ( to be later varied in a sensitivity analysis ) .
in this case the marginal mean of @xmath0 is given by @xmath694 \right\ } \right ] \right ) .$ ] robins and rotnitzky @xmath695 proved this model places no restrictions on @xmath696 and derived @xmath697 where , now , @xmath698 /e\left [ \exp\left\ { -\alpha y\right\ } |a=1,x\right]\ ] ] @xmath70 and @xmath699
. thus @xmath700 and @xmath701 when @xmath702 this provides an alternate parametrization of example 2a .
* example 3 : marginal structural models and the average treatment effect : * consider the set - up of example 1c including the non - identifiable assumption of no unmeasured confounders , except now @xmath12 is discrete with possibly many levels and @xmath703 wp1 .
a marginal structural model assumes @xmath704 where @xmath705 is a known function and @xmath706 is an unknown vector parameter of dimension @xmath707 when @xmath12 is dichotomous with @xmath708 and @xmath709 , then @xmath710 is the average treatment effect parameter .
let @xmath711 be any density with the same support as @xmath12 and let @xmath712 be a @xmath713-vector function , both chosen by the analyst .
then @xmath714is identified as the ( assumed ) unique value of @xmath715 satisfying @xmath716 = 0,\ ] ] where @xmath717 .
thus a @xmath664confidence set for @xmath655 is the set of vectors @xmath718such that a @xmath664ci for @xmath719 contains @xmath667therefore , with no loss of generality , we consider the construction of a @xmath664ci for the @xmath676vector functional @xmath720 for a fixed value @xmath721 and define @xmath722 and @xmath723 .$ ] then @xmath724 has influence function @xmath725 next define @xmath726 $ ] . then @xmath727 is the integral @xmath728 it follows that @xmath727 is a integral over @xmath729 of influence functions @xmath730 for parameters @xmath731 in our class with @xmath732 .
thus we can estimate @xmath175 by @xmath733 where @xmath734 is an estimator of @xmath735 if the support of @xmath12 is of greater cardinality than @xmath736 the model is not locally nonparametric .
different choices for @xmath712 and @xmath737 for which @xmath738 $ ] is invertible may result in difference influence functions .
all yield the same rate of convergence , although the constants differ .
see remark [ proj ] above .
extension of our methods to continuous @xmath12 will be treated elsewhere .
* example 4 : confidence intervals for the optimal treatment strategy in a randomized clinical trial : * consider a randomized clinical trial with data @xmath739 , @xmath12 is a binary treatment taking values in @xmath740 , @xmath642 and @xmath0 are univariate responses , @xmath8 is a vector of pretreatment covariates . in a randomized trial , the randomization probabilities @xmath741
are known by design .
let @xmath742 and @xmath743 be the average treatment effects at level @xmath744 on @xmath642 and @xmath20 we assume @xmath745 and @xmath642 have been coded so that positive treatment effects are desirable .
let @xmath746 .$ ] because the model is not locally nonparametric there exists more than a single first order influence function . indeed , for any given function @xmath747 @xmath748 \\ & \times\left\ { a-\pi_{0}\left ( x\right ) \right\ } \sigma_{0}^{-2}\left ( x\right ) + c\left ( x\right ) \left\ { a-\pi_{0}\left ( x\right ) \right\}\end{aligned}\ ] ] with @xmath749 is an influence function in our class [ provided it is square integrable ] with @xmath750 @xmath751 @xmath752 @xmath753 .
as @xmath754 is varied , one obtains all first order influence functions .
we do not discuss the efficient choice of @xmath754 in this paper .
our interest lies in the special case where @xmath755 wp1 ( so there is but one response of interest ) and thus , as in assumption @xmath629 @xmath756and we construct confidence interval for @xmath757 .$ ] in section [ adaptive_section ] we describe how we can use a confidence interval for @xmath758 $ ] to obtain confidence intervals for the treatment effect function @xmath759 and , most importantly , for the optimal treatment strategy @xmath760 $ ] under which a subject with covariate value @xmath650 is treated if and only if the treatment effect @xmath761 is positive ( i.e. , @xmath762 . in all of our examples
the functions @xmath598 and @xmath763 are functions of conditional expectations given the continuous random variable @xmath113 it is well known that the associated point - evaluation functional @xmath764 and @xmath759 do not have first order influence functions .
it then follows from part 5c of theorem [ eift ] and the dependence of @xmath765 $ ] on @xmath130 and @xmath598 evaluated at the point @xmath8 that , in none of our examples , does @xmath175 have a second ( or higher ) order influence function .
as a precise understanding of the reason for the nonexistence of higher order influence functions for @xmath175 is fundamental to our approach , we now use part 5c of theorem [ eift ] to prove that @xmath766 does not exist by showing that the functional @xmath767 does not have a first order influence function @xmath768 .$ ] in this proof , we do not assume that @xmath130 and @xmath598 are functions of conditional expectations .
rather we only assume that our functional satisfies assumptions a(i - iv ) .
let @xmath769 and @xmath770 denote the marginal cdf and density of @xmath113 consider paths ( parametric submodels ) @xmath771 such that @xmath772satisfying @xmath773 where the sequences @xmath562 and @xmath774 are each dense in @xmath775 .$ ] let @xmath776 @xmath777 and @xmath778 denote the overall , conditional , and marginal scores@xmath779 by linearity , @xmath767 has an influence function only if the functionals @xmath759 and @xmath764 have one as well .
now by differentiating the identity @xmath780 = 0\ ] ] wrt to @xmath781 and evaluating at @xmath782 we have @xmath783 = e_{\mathbb{\theta}}\left [ h_{1}|x = x\right ] a_{l}\left ( x\right)\ ] ] however , by definition , @xmath759 has an influence function @xmath784 $ ] at @xmath785only if for @xmath786 both @xmath787 equals @xmath788 $ ] and @xmath789 = 0.$ ] thus if @xmath790 exists , it must satisfy @xmath791 \\ & = e_{\theta}\left [ h_{1}|x = x\right ] e_{\theta}\left [ if_{1,b\left ( x\right ) \ } \left ( o;\theta\right ) s_{l}\left ( o;\theta\right ) \right]\end{aligned}\ ] ] without loss of generality , suppose @xmath792 wp 1@xmath64 now if we could find a kernel @xmath793 such that@xmath794 \nonumber\\ & \equiv\int k_{f_{x},\infty}\left ( x , x^{\ast}\right ) r\left ( x^{\ast } \right ) f_{x}\left ( x^{\ast}\right ) dx^{\ast}\text { for all } r\left ( \cdot\right ) \in l_{2}\left ( f_{x}\right ) \label{dirac2}\ ] ] then @xmath795{c}\left\ { e_{\theta}\left [ h_{1}|x = x\right ] \right\ } ^{-1/2}k_{f_{x},\infty } \left ( x , x\ \right ) \\ \times\left\ { e_{\theta}\left [ h_{1}|x\right ] \right\ } ^{-1/2}\left\ { h_{1}b\left ( x\ \right ) + h_{3}\right\ } \end{array } \right]\ ] ] would be an influence function since @xmath796 e_{\theta}\left [ \begin{array } [ c]{c}-\left\ { e_{\theta}\left [ h_{1}|x = x\right ] \right\ } ^{-1/2}k_{f_{x},\infty}\left ( x , x\right ) \times\\ \left\ { e_{\theta}\left [ h_{1}|x\right ] \right\ } ^{-1/2}\left\ { h_{1}b\left ( x\ \right ) + h_{3}\right\ } s_{l}\left ( o;\theta\right ) \end{array } \right ] \\ & = e\left [ h_{1}|x = x\right ] ^{1/2}\ e_{\theta}\left [ \begin{array } [ c]{c}-k_{f_{x},\infty}\left ( x , x\right ) \left\ { e_{\theta}\left [ h_{1}|x\right ] \right\ } ^{-1/2}\\ \times\left\ { h_{1}b\left ( x\ \right ) + h_{3}\right\ } \left\ { s_{l}\left ( o|x\right ) + s_{l}\left ( x\right ) \right\ } \end{array } \right ] \\ & = e\left [ h_{1}|x = x\right ] ^{1/2}\ e_{f_{x}}\left\ { e_{\theta}\left [ \begin{array } [ c]{c}-k_{f_{x},\infty}\left ( x , x\right ) \left\ { e_{\theta}\left [ h_{1}|x\right ] \right\ } ^{-1/2}\times\\ \left\ { h_{1}b\left ( x\ \right ) + h_{3}\right\ } s_{l}\left ( o|x\right ) |x \end{array } \right ] \right\ } \\ & = -e_{\theta}\left [ \left ( h_{1}b\left ( x\right ) + h_{3}\right ) s_{l}\left ( o|x\right ) |x = x\right]\end{aligned}\ ] ] by an analogous argument @xmath797{c}\left\ { e_{\theta}\left [ h_{1}|x = x\right ] \right\ } ^{-1/2}k_{f_{x},\infty } \left ( x , x\right ) \\ \times\left\ { e_{\theta}\left [ h_{1}|x\right ] \right\ } ^{-1/2}\left\ { h_{1}p\left ( x\ \right ) + h_{2}\right\ } \end{array } \right]\ ] ] would be an influence function . indeed since the sequences @xmath798 and @xmath799 are dense the existence of such a kernel is also a necessary condition for @xmath800 and @xmath801 to exist and thus for @xmath802 to exist .
a kernel satisfying eq.@xmath803 is referred to as the dirac delta function wrt to the measure @xmath804 and would clearly have to satisfy @xmath805 were it to exist . of course a kernel satisfying eq .
( [ dirac2 ] ) is known not to exist in @xmath806 \times l_{2}\left [ f_{x}\right ] .$ ] we conclude that @xmath767 does not have an influence function and therefore @xmath807 does not exist . * a formal approach : * to motivate how one might overcome this difficulty , we note that kernels satisfying eq .
( [ dirac2 ] ) exist as generalized functions or kernels ( also known as schwartz functions or distributions ) .
we shall `` formally derive higher order influence functions that appear to be elements of the space of generalized functions .
however , we use these calculations only as motivation for statistical procedures based on ordinary kernels living in @xmath806 \times l_{2}\left [ f_{x}\right ] .$ ] thus it does not matter whether these formal calculations could be made rigorous with appropriate redefinitions . rather we can simply regard the following as results obtained by applying a ' ' formal calculus " to part 5c of theorem [ eift ] that adds to the usual calculus additional identities licensed by eqs .
( [ dirac2 ] ) and ( [ diracpoint])@xmath64 we will need the fact that@xmath90 for any function @xmath808 eq .
@xmath809implies that @xmath810 we now show that@xmath811 = \pi_{\theta , 2}\left [ \mathbb{v}\left [ if_{1,if_{1,\psi}\left ( o_{i_{1}};\cdot\right ) \ } \left ( o_{i_{2}};\theta\right ) /2\right ] |\mathcal{u}_{1}^{\perp _ { 2,\theta}}\left ( \theta\right ) \right]\ ] ] would formally have u - statistic kernel @xmath812{c}\varepsilon_{b , i_{1}}\left ( \theta\right ) e_{\theta}\left [ h_{1}|x_{i_{1}}\right ] ^{-\frac{1}{2}}k_{f_{x},\infty}\left ( x_{i_{1}},x_{i_{2}}\right ) \\ e_{\theta}\left [ h_{1}|x_{i_{2}}\right ] ^{-\frac{1}{2}}\varepsilon_{p , i_{2}}\left ( \theta\right ) \end{array } \right ] , \label{infin2}\\ \text{with } \varepsilon_{b , i_{1}}\left ( \theta\right ) & = \ \left\ { b_{i_{1}}h_{1,i_{1}}+h_{3,i_{1}}\right\ } , \text { } \varepsilon_{p , i_{2}}\left ( \theta\right ) = \
\left\ { h_{1,i_{2}}p_{i_{2}}+h_{2,i_{2}}\right\ } .\nonumber\end{aligned}\ ] ] to show eq [ infin2 ] note , by @xmath813 @xmath70and@xmath814 we have @xmath815 where @xmath816 ^{-\frac{1}{2}}\\ & \times k_{f_{x},\infty}\left ( x_{i_{1}},x_{i_{2}}\right ) e_{\theta } \left [ h_{1}|x_{i_{2}}\right ] ^{-\frac{1}{2}}\left\ { p_{i_{2}}h_{1,i_{2}}+h_{2,i_{2}}\right\ } \\ & = -\varepsilon_{b , i_{1}}\left ( \theta\right ) e_{\theta}\left [ h_{1}|x_{i_{1}}\right ] ^{-\frac{1}{2}}k_{f_{x},\infty}\left ( x_{i_{1}},x_{i_{2}}\right ) e_{\theta}\left [ h_{1}|x_{i_{2}}\right ] ^{-\frac{1}{2}}\varepsilon_{p , i_{2}}\left ( \theta\right ) \\ q_{2,b,\overline{i}_{2}}\left ( \theta\right ) & \equiv\left\ { p_{i_{1}}h_{1,i_{1}}+h_{2,i_{1}}\right\ } if_{1,b\left ( x_{i_{1}}\right ) \ } \left ( o_{i_{2}};\theta\right ) \\ & = -\varepsilon_{b , i_{2}}\left ( \theta\right ) e_{\theta}\left [ h_{1}|x_{i_{2}}\right ] ^{-\frac{1}{2}}k_{f_{x},\infty}\left ( x_{i_{2}},x_{i_{1}}\right ) e_{\theta}\left [ h_{1}|x_{i_{1}}\right ] ^{-\frac{1}{2}}\varepsilon_{p , i_{1}}\left ( \theta\right)\end{aligned}\ ] ] thus , by part 5c of theorem [ eift ] @xmath70@xmath817 \\ & = \frac{1}{2}\left\ { \mathbb{q}_{2,p,\overline{i}_{2}}\left ( \theta\right ) + \mathbb{q}_{2,b,\overline{i}_{2}}\left ( \theta\right ) \right\ } = \mathbb{q}_{2,p,\overline{i}_{2}}\left ( \theta\right ) \equiv\mathbb{v}\left [ rhs\text { of eq}.(\ref{infin2})\right]\end{aligned}\ ] ] since @xmath818 is a function of only one subject s data and @xmath819 and @xmath820 are the same up to a permutation that exchanges @xmath821 with @xmath822 to obtain @xmath823 one must derive the influence function @xmath824of @xmath825 the formula for @xmath826 is given in eq . @xmath827
a detailed derivation is given in the appendix .
here we simply note that the only essentially new point is that we now require the influence function of @xmath828 , which , as shown next , is given by @xmath829{c}k_{f_{x},\infty}\left ( x_{i_{1}},x_{i_{3}}\right ) k_{f_{x},\infty}\left ( x_{i_{3}},x_{i_{2}}\right ) \\
-k_{f_{x},\infty}\left ( x_{i_{1}\ } , x_{i_{2}}\right ) \end{array } \right\ } \label{kernelif}\ ] ] to see that if eq.@xmath803 held , eq.@xmath830 would hold , note that for any path @xmath831 with @xmath832 @xmath833 .\ $ ] differentiating wrt to @xmath781 and evaluating at @xmath834 , we have @xmath835 + e_{_{\theta}}\left [ \left\ { \frac { \partial}{\partial t}k_{\widetilde{\mathbb{\theta}}\left ( t\right ) , \infty } \left ( x , x_{i_{1}}\right ) _ { |t=0}\right\ } h\left ( x_{i_{1}}\right ) \right]\ ] ] hence it suffices to show that @xmath836 \\ & = e_{_{\theta}}\left [ \left\ { e_{\theta}\left\ { -k_{f_{x},\infty}\left ( x , x_{i_{2}}\right ) k_{f_{x},\infty}\left ( x_{i_{2}},x_{i_{1}}\right ) s_{i_{2}}\left ( \theta\right ) |x_{i_{1}}\right\ } \right\ } h\left ( x_{i_{1}}\right ) \right]\end{aligned}\ ] ] but , by eq.@xmath837 @xmath838 \\ & = e_{_{\theta}}\left [ -k_{f_{x},\infty}\left ( x , x_{i_{1}}\right ) s_{i_{1}}\left ( \theta\right ) h\left ( x_{i_{1}}\right ) \right ] .\end{aligned}\ ] ] * feasible estimators : * these `` formal '' calculations motivate a `` truncated dirac '' approach to estimate @xmath174 let @xmath839 be a countable sequence of known basis functions with dense span in @xmath840 and define @xmath841 define @xmath842 \right\ } ^{-1}\overline{z}_{k}\left ( x_{i_{2}}\right)\ ] ] to be the projection kernel in @xmath840 onto the subspace @xmath843 spanned by the elements of @xmath844 that is , for any @xmath845@xmath846 \\ & = e_{f_{x}}\left [ k_{f_{x},k}\left ( x , x\right ) h\left ( x\right ) \right ] \\ & = \overline{z}_{k}\left ( x\right ) ^{t}\left\ { e_{f_{x}}\left [ \overline{z}_{k}\left ( x\ \right ) \overline{z}_{k}\left ( x\right ) ^{t}\right ] \right\ } ^{-1}e_{f_{x}}\left [ \overline{z}_{k}\left ( x\ \right ) h\left ( x\right ) \right]\end{aligned}\ ] ] then we can view @xmath847 as a truncated at @xmath848 approximation to @xmath849 that is in @xmath806 \times l_{2}\left [ f_{x}\right ] $ ] and satisfies eq.@xmath803 for all @xmath850 then a natural idea would be to substitute @xmath851{c}-\varepsilon_{b , i_{1}}\left ( \widehat{\theta}\right ) e_{\widehat{\theta}}\left [ h_{1}|x_{i_{1}}\right ] ^{-\frac{1}{2}}k_{\widehat{f}_{x},k}\left ( x_{i_{1}},x_{i_{2}}\right ) \\
\times e_{\widehat{\theta}}\left [ h_{1}|x_{i_{2}}\right ] ^{-\frac{1}{2}}\varepsilon_{p , i_{2}}\left ( \widehat{\theta}\right ) \end{array } \right)\ ] ] with , for example , @xmath852 for the generalized function @xmath853based on eqs .
[ infin2 ] resulting in the feasible 2nd u - statistic estimator@xmath854 where@xmath855\ ] ] to avoid having to do a matrix inversion it is convenient to choose @xmath856 where @xmath857 is a complete orthonormal basis wrt to lebesgue measure in @xmath437 then @xmath858 = i_{k\times k}$ ] so @xmath859 this choice corresponds to having taken @xmath860 in our formal calculations where @xmath861 is the dirac delta function wrt to lebesgue measure . in that case with @xmath862 $ ] and @xmath863 , $ ] @xmath864 in the appendix , we show one can proceed by induction to formally obtain that for @xmath865@xmath866{c}\sum_{j=0}^{m-2}c(m , j)\times\\ \prod\limits_{s=1}^{j}\frac{h_{1,i_{s+1}}}{g\left ( x_{i_{s+1}}\right ) } k_{leb,\infty}\left ( x_{i_{s}},x_{i_{s+1}}\right ) \\
\times k_{leb,\infty}\left ( x_{i_{j+1}},x_{i_{m}}\right ) \end{array } \right ] g\left ( x_{i_{m}}\right ) ^{-\frac{1}{2}}\varepsilon_{p , i_{m}}\left ( \theta\right ) \label{imm}\ ] ] where @xmath867 which we then use to obtain @xmath868and @xmath869 * statistical properties : * we shall prove below that the estimator @xmath870 has variance @xmath871 \asymp\left ( \frac{1}{n}\max\left [ 1,\left ( \frac{k}{n}\right ) ^{m-1}\right ] \right)\ ] ] when @xmath872 is a compact wavelet basis@xmath64 ( robins et al .
( 2007 ) proves this result for more general bases ) .
we also prove that the bias @xmath873
-\psi\left ( \theta\right ) = tb_{k}\left ( \theta\right ) + eb_{m}\left ( \theta\right ) , \ ] ] of @xmath870 is the sum of a truncation bias term of order @xmath874 ( for a basis @xmath872 that provides optimal rate approximation for hlder balls ) and an estimation bias term of order @xmath875 the truncation bias is of this order only if @xmath139 has smoothness exceeding @xmath876 .
this restriction on @xmath139 is removed later by using kernels based on eq .
note this estimation bias is @xmath878 it gets its name from the fact that , unlike the truncation bias , it would be exactly zero if the initial estimator @xmath391 happened to equal @xmath386 thus , the u - statistic estimator @xmath870 for our functional @xmath175 ( which does not admit a second order influence function ) differs from the u - statistic estimators @xmath352 of eq .
( [ est ] ) for functionals that admit second order influence functions in that , owing to truncation bias , the total bias of @xmath879 is not @xmath880 the choice of @xmath881determines the trade - off between the variance and truncation bias . as @xmath882 with @xmath5 fixed@xmath90 @xmath883 \rightarrow\infty$ ] and @xmath884
thus , we can heuristically view the non - existent estimator @xmath885 as the choice of @xmath848 that results in no truncation bias [ and therefore a total bias of @xmath886 $ ] at the expense of an infinite variance .
writing @xmath887 the order of the asymptotic mse of @xmath870 is minimized at the value of @xmath888for which order of the variance equals the order of the sum of the truncation and estimation bias . [ 1proj]the models of examples 1 - 4 exhibit a spectrum of different likelihood functions and therefore a spectrum of different first order and higher order scores .
nonetheless , because the first order influence functions of the functionals @xmath175 share a common structure , we were able to use part 5c of theorem [ eift ] to formally derive @xmath889 and , thus , the feasible @xmath890 in examples 1 - 4 in a unified manner without needing to consult the full likelihood function for any of the models
. see remark ( [ proj ] ) above for a closely related discussion . * a critical non - uniqueness : *
we have as yet neglected a critical non - uniqueness in our definition of @xmath891 and thus @xmath870 that poses a significant problem for our `` truncated dirac '' approach .
for instance , when @xmath892 the two generalized @xmath893 kernels @xmath894 of eq [ imm ] and @xmath895{c}\frac{h_{1,i_{2}}}{g\left ( x_{i_{2}}\right ) } k_{leb,\infty}\left ( x_{i_{1}},x_{i_{2}}\right ) \\ -e_{\theta}\left [ \frac{k_{leb,\infty}\left ( x_{i_{1}},x_{i_{2}}\right ) } { f\left ( x_{i_{2}}\right ) } |x_{i_{1}}\right ] \end{array } \right ] \\ & \times k_{leb,\infty}\left ( x_{i_{1}},x_{i_{3}}\right ) \frac { \varepsilon_{p , i_{3}}\left ( \theta\right ) } { g\left ( x_{i_{3}}\right ) ^{\frac{1}{2}}}\ ] ] are precisely equal , by eq .
@xmath896 ; nonetheless , upon truncation , they result in different feasible kernels ; @xmath897{c}\frac{h_{1,i_{2}}}{\widehat{g}\left ( x_{i_{2}}\right ) } k_{leb , k}\left ( x_{i_{1}},x_{i_{2}}\right ) k_{leb , k}\left ( x_{i_{2}},x_{i_{3}}\right ) \\
-k_{leb , k}\left ( x_{i_{1}},x_{i_{3}}\right ) \end{array } \right ] \times\frac{\widehat{\varepsilon}_{p , i_{3}}\left ( \theta\right ) } { \widehat{g}\left ( x_{i_{3}}\right ) ^{\frac{1}{2}}}\ ] ] and @xmath898{c}\frac{h_{1,i_{2}}}{\widehat{g}\left ( x_{i_{2}}\right ) } k_{leb , k}\left ( x_{i_{1}},x_{i_{2}}\right ) \\ -e_{\widehat{\theta}}\left [ \frac{k_{leb , k}\left ( x_{i_{1}},x_{i_{2}}\right ) } { \widehat{f}\left ( x_{i_{2}}\right ) } |x_{i_{1}}\right ] \end{array } \right ] \\ & \times k_{leb , k}\left ( x_{i_{1}},x_{i_{3}}\right ) \frac { \widehat{\varepsilon}_{p , i_{3}}\left ( \theta\right ) } { \widehat{g}\left ( x_{i_{3}}\right ) ^{\frac{1}{2}}}\ ] ] with different orders of bias . for simplicity
, we consider the case where @xmath899 as in examples * 1a*-**1c . *
* let @xmath900 @xmath901 @xmath902 and @xmath903 then,@xmath904 \\ & = e_{\theta}\left [ \begin{array } [ c]{c}\frac{\delta b_{i_{1}}}{\widehat{f}\left ( x_{i_{1}}\right ) ^{\frac{1}{2}}}\times\\ e_{\mu}\left [ \left ( \frac{f\left ( x_{i_{2}}\right ) } { \widehat{f}\left ( x_{i_{2}}\right ) } -1\right ) \overline{\varphi}_{k}\left ( x_{i_{2}}\right ) ^{t}\right ] \overline{\varphi}_{k}\left ( x_{i_{1}}\right ) \\
\times e_{\theta}\left [ \frac{\delta p_{i_{3}}}{\widehat{f}\left ( x_{i_{3}}\right ) ^{\frac{1}{2}}}\overline{\varphi}_{k}\left ( x_{i_{3}}\right ) ^{t}\right ] \overline{\varphi}_{k}\left ( x_{i_{1}}\right ) \end{array } \right ] \\ & = e_{\widehat{\theta}}\left [ \begin{array } [ c]{c}\left ( \frac{f\left ( x_{i_{1}}\right ) } { \widehat{f}\left ( x_{i_{1}}\right ) } -1 + 1\right ) \widehat{f}\left ( x_{i_{1}}\right ) ^{\frac{1}{2}}\delta b_{i_{1}}\times\\ e_{\widehat{\theta}}\left [ \left ( \frac{f\left ( x_{i_{2}}\right ) } { \widehat{f}\left ( x_{i_{2}}\right ) } -1\right ) \widehat{f}\left ( x_{i_{2}}\right ) ^{-\frac{1}{2}}\frac{\overline{\varphi}_{k}\left ( x_{i_{2}}\right ) ^{t}}{\widehat{f}\left ( x_{i_{2}}\right ) ^{\frac{1}{2}}}\right ] \frac{\overline{\varphi}_{k}\left ( x_{i_{1}}\right ) } { \widehat{f}\left ( x_{i_{1}}\right ) ^{\frac{1}{2}}}\\ \times e_{\widehat{\theta}}\left [ \left ( \frac{f\left ( x_{i_{3}}\right ) } { \widehat{f}\left ( x_{i_{3}}\right ) } -1 + 1\right ) \delta p_{i_{3}}\frac{\overline{\varphi}_{k}\left ( x_{i_{3}}\right ) ^{t}}{\widehat{f}\left ( x_{i_{3}}\right ) ^{\frac{1}{2}}}\right ] \frac{\overline{\varphi}_{k}\left ( x_{i_{1}}\right ) } { \widehat{f}\left ( x_{i_{1}}\right ) ^{\frac{1}{2}}}\end{array } \right ] \\ & = e_{\widehat{\theta}}\left [ \begin{array } [ c]{c}\widehat{f}\left ( x_{i_{1}}\right ) ^{\frac{1}{2}}\delta b_{i_{1}}e_{\widehat{\theta}}\left [ \delta f\left ( x_{i_{2}}\right ) \widehat{f}\left ( x_{i_{2}}\right ) ^{-\frac{1}{2}}\overline{z}_{k , i_{2}}^{t}\right ] \overline{z}_{k , i_{1}}\\ \times e_{\widehat{\theta}}\left [ \delta p_{i_{3}}\overline{z}_{k , i_{3}}^{t}\right ] \overline{z}_{k , i_{1}}\end{array } \right ] \\ & + o_{p}\left ( \left\ { b-\widehat{b}\right\ } \left\ { p-\widehat{p}\right\ } \left\ { g-\widehat{g}\right\ } ^{2}\right)\end{aligned}\ ] ] and@xmath905 \\ & = e_{\mu}\left [ \begin{array } [ c]{c}e_{\theta}\left [ \frac{\delta b_{i_{1}}}{\widehat{f}\left ( x_{i_{1}}\right ) ^{\frac{1}{2}}}\overline{\varphi}_{k}\left ( x_{i_{1}}\right ) ^{t}\right ] \overline{\varphi}_{k}\left ( x_{i_{2}}\right ) \left ( \frac{f\left ( x_{i_{2}}\right ) } { \widehat{f}\left ( x_{i_{2}}\right ) } -1\right ) \\ \times\overline{\varphi}_{k}\left ( x_{i_{2}}\right ) ^{t}e_{\theta}\left [ \frac{\delta p_{i_{3}}}{\widehat{f}\left ( x_{i_{3}}\right ) ^{\frac{1}{2}}}\overline{\varphi}_{k}\left ( x_{i_{3}}\right ) \right ] \end{array } \right ] \\ & = e_{\widehat{\theta}}\left [ \begin{array } [ c]{c}e_{\widehat{\theta}}\left [ \left ( \delta f\left ( x_{i_{1}}\right ) + 1\right ) \delta b_{i_{1}}\overline{z}_{k , i_{1}}^{t}\right ] \\
\times\overline{z}_{k , i_{2}}\left ( \frac{f\left ( x_{i_{2}}\right ) } { \widehat{f}\left ( x_{i_{2}}\right ) } -1\right ) \\
\times\overline{z}_{k , i_{2}}^{t}e_{\widehat{\theta}}\left [ \left ( \delta f\left ( x_{i_{3}}\right ) + 1\right ) \delta p_{i_{3}}\overline{z}_{k , i_{3}}\right ] \end{array } \right ] \\ & = e_{\widehat{\theta}}\left [ \begin{array } [ c]{c}e_{\widehat{\theta}}\left [ \delta b_{i_{1}}\overline{z}_{k , i_{1}}^{t}\right ] \overline{z}_{k , i_{2}}\left ( \frac{f\left ( x_{i_{2}}\right ) } { \widehat{f}\left ( x_{i_{2}}\right ) } -1\right ) \\
\times\overline{z}_{k , i_{2}}^{t}e_{\widehat{\theta}}\left [ \delta p_{i_{3}}\overline{z}_{k , i_{3}}\right ] \end{array } \right ] \\ & + o_{p}\left ( \left\ { b-\widehat{b}\right\ } \left\ { p-\widehat{p}\right\ } \left\ { g-\widehat{g}\right\ } ^{2}\right)\end{aligned}\ ] ] that is , @xmath904 -e_{\theta}\left [ if_{3,3,\psi , i_{1},i_{2},i_{3}}^{\left ( k\right ) } \left ( \widehat{\theta}\right ) \right ] \\ & = e_{\widehat{\theta}}\left [ \begin{array } [ c]{c}\widehat{f}\left ( x_{i_{1}}\right ) ^{\frac{1}{2}}\delta b_{i_{1}}e_{\widehat{\theta}}\left [ \delta f\left ( x_{i_{2}}\right ) \widehat{f}\left ( x_{i_{2}}\right ) ^{-\frac{1}{2}}\overline{z}_{k , i_{2}}^{t}\right ] \overline{z}_{k , i_{1}}\\ e_{\widehat{\theta}}\left [ \delta p_{i_{3}}\overline{z}_{k , i_{3}}^{t}\right ] \overline{z}_{k , i_{1}}\end{array } \right ] \\ & -e_{\widehat{\theta}}\left [ \begin{array } [ c]{c}e_{\widehat{\theta}}\left [ \delta b_{i_{1}}\overline{z}_{k , i_{1}}^{t}\right ] \overline{z}_{k , i_{2}}\delta f\left ( x_{i_{2}}\right ) \\ \times\overline{z}_{k , i_{2}}^{t}e_{\widehat{\theta}}\left [ \delta p_{i_{3}}\overline{z}_{k , i_{3}}\right ] \end{array } \right ] \\ & + o_{p}\left ( \left\ { b-\widehat{b}\right\ } \left\ { p-\widehat{p}\right\ } \left\ { g-\widehat{g}\right\ } ^{2}\right)\end{aligned}\]]@xmath906@xmath907 \widehat{f}\left ( x\right ) ^{\frac{1}{2}}\delta b\pi_{\widehat{\theta}}\left [ \delta f\left ( x\right ) \widehat{f}\left ( x\right ) ^{-\frac{1}{2}}|\overline{z}_{k}\right ] \right ] \\ & -e_{\widehat{\theta}}\left [ \pi_{\widehat{\theta}}\left [ \delta p|\overline{z}_{k}\right ] \widehat{f}\left ( x\right ) ^{\frac{1}{2}}\delta f\left ( x\right ) \widehat{f}\left ( x\right ) ^{-\frac{1}{2}}\pi _ { \widehat{\theta}}\left [ \delta b|\overline{z}_{k}\right ] \right ] \\ & + o_{p}\left ( \left\ { b-\widehat{b}\right\ } \left\ { p-\widehat{p}\right\ } \left\ { g-\widehat{g}\right\ } ^{2}\right)\end{aligned}\]]@xmath906@xmath908{c}\pi_{\widehat{\theta}}\left [ \delta p|\overline{z}_{k}\right ] \widehat{f}\left ( x\right ) ^{\frac{1}{2}}\times\\ \left\ { \begin{array } [ c]{c}\pi_{\widehat{\theta}}^{\bot}\left [ \delta b|\overline{z}_{k}\right ] \pi_{\widehat{\theta}}\left [ \delta f\left ( x\right ) \widehat{f}\left ( x\right ) ^{-\frac{1}{2}}|\overline{z}_{k}\right ] \\ -\pi_{\widehat{\theta}}\left [ \delta b|\overline{z}_{k}\right ] \pi_{\widehat{\theta}}^{\bot}\left [ \delta f\left ( x\right ) \widehat{f}\left ( x\right ) ^{-\frac{1}{2}}|\overline{z}_{k}\right ] \end{array } \right\ } \end{array } \right ] \\ & + o_{p}\left ( \left\ { b-\widehat{b}\right\ } \left\ { p-\widehat{p}\right\ } \left\ { g-\widehat{g}\right\ } ^{2}\right)\end{aligned}\ ] ] where @xmath909 $ ] and @xmath910 in @xmath911 of @xmath912 on the @xmath848 dimensional linear subspace @xmath913 spanned by the components of the vector @xmath914 and the projection on the orthocomplement of this subspace .
since the basis @xmath872 provides optimal rate approximation for hlder balls , it is easy to verify that the difference is of order @xmath915{c}n^{-\frac{\beta_{p}/d}{1 + 2\beta_{p}/d}-\frac{\beta g / d}{1 + 2\beta_{g}/d}}k^{-\beta_{b}/d}+n^{-\frac{\beta_{p}/d}{1 + 2\beta_{p}/d}-\frac{\beta_{b}/d}{1 + 2\beta_{b}/d}}k^{-\beta_{g}/d}\\ + n^{-\frac{\beta_{p}/d}{1 +
2\beta_{p}/d}-\frac{\beta_{b}/d}{1 + 2\beta_{b}/d}-\frac{2\beta g / d}{1 + 2\beta_{g}/d}}\end{array } \right)\ ] ] provided @xmath139 has smoothness exceeding @xmath916 for concreteness , we shall look at an example .
suppose @xmath917 and @xmath918 thus , by choosing @xmath919 @xmath920 converges to @xmath270 at rate @xmath921 in contrast , the order , @xmath922 of the optimal root mean squares error of @xmath923that uses @xmath924 is @xmath925 thus @xmath926converges to @xmath175 at a slower rate than @xmath927 which uses @xmath928 nothing in our development up to this point provides any guidance as to which of the many equivalent generalized u - statistic kernels should be selected for truncation . to provide some guidance
, we introduce an alternative approach to the estimation of @xmath175 based on truncated parameters that admit higher order influence functions .
the class of estimators we derive using this alternative approach includes members algebraically identical to the estimators @xmath870 but does not include estimators equivalent to less efficient estimators such as @xmath929 . * an approach based on truncated parameters : * we introduce a class of truncated parameters @xmath930 that ( i ) depend on the sample size through a positive integer index @xmath931which we refer to as the truncation index and will be optimized below ) , ( ii ) have influence functions @xmath932 of all orders @xmath135 , ( iii ) equals @xmath270 on a large subset @xmath933 of @xmath245 and ( iv ) the initial estimator @xmath934 is an element of @xmath935so that the plug - ins @xmath936and @xmath937 are equal . to prepare we introduce a simplified notation . for functions
@xmath938 or @xmath939 of @xmath940 we will often write @xmath941 and @xmath942 as @xmath943 and @xmath944 and @xmath945 $ ] as @xmath946 $ ] . similarly , we often write @xmath947 and @xmath948 as @xmath949and @xmath950 and @xmath951 $ ] as @xmath952 .$ ] further we shall introduce slightly different definitions of truncation and estimation bias .
define the estimator @xmath953or , equivalently , @xmath954 . then the conditional bias @xmath955
-\psi$ ] of @xmath956 is @xmath957 where the truncation bias @xmath958 is zero for @xmath959 and does not depend on @xmath135 and the estimation bias @xmath960 -\widetilde{\psi}_{k}$ ] is @xmath961 does not depend on @xmath962 we will abbreviate @xmath963 as @xmath964 suppressing the dependence on @xmath965 under minimal conditions , the conditional variance of @xmath956 is of the order of @xmath966 $ ] whenever @xmath967 the rate of convergence of @xmath956 to @xmath271 can depend on the choice of @xmath968 nevertheless , many different choices @xmath969 result in estimators @xmath956 that achieve what we conjecture to be the optimal rate for estimators in our class of the form @xmath956 .
we choose , among all such @xmath970 the class that minimizes the computational complexity of @xmath971 specifically for all @xmath969 in our chosen class and all @xmath972 @xmath973 consists of a single term rather than a sum of many terms .
we conjecture this appealing property does not hold for any @xmath969 outside our class .
we now describe this choice .
the parameter @xmath974is defined in terms of @xmath975dimensional working linear parametric submodels for @xmath976 and @xmath130 depending on unknown parameters @xmath977and @xmath978 through the basepoints @xmath979 and @xmath980 where @xmath979 and @xmath981 are initial estimators from the training sample .
specifically let @xmath982 and @xmath983 be arbitrary bounded known functions chosen by the analyst satisfying eqs @xmath984 below .
@xmath985 \geq0\text { } w.p.1\label{dot1}\\ \left\vert \left\vert \frac{\dot{p}\left ( x\right ) } { \dot{b}\left ( x\right ) } \right\vert \right\vert _ { \infty}<c^{\ast},\left\vert \left\vert \frac{\dot{b}\left ( x\right ) } { \dot{p}\left ( x\right ) } \right\vert \right\vert _ { \infty}<c^{\ast}\label{dot2}\\ \frac{\dot{p}\left ( x\right ) } { \dot{b}\left ( x\right ) } \text { has at least } \left\lceil max\left\ { \beta_{b},\ \beta_{p}\right\ } \right\rceil \text { derivatives}\label{dot3}\ ] ] particular choices of @xmath982 and @xmath986 can make the form of @xmath987 more aesthetic .
the choice has no bearing on the rate of convergence of the estimator @xmath956 to @xmath988 often there are fairly natural choices for @xmath989 and @xmath990 see remark [ q ] below for examples@xmath64 let @xmath991 be @xmath992vectors of unknown parameters and consider the working linear models @xmath993 we define the parameters @xmath994 and @xmath995 respectively to be the solution
to@xmath996 = e_{\theta}\left [ \left\ { h_{1}b^{\ast}\left ( x,\overline{\eta}_{k}\right ) + h_{3}\right\ } \dot { p}\overline{z}_{k}\right ] \label{eta}\\ 0 & = e_{\theta}\left [ \partial h\left ( b^{\ast}\left ( x,\overline{\eta } _ { k}\right ) , p^{\ast}\left ( x,\overline{\alpha}_{k}\right ) \right ) /\partial\overline{\eta}_{k}\right ] = e_{\theta}\left [ \left\ { h_{1}p^{\ast } \left ( x,\overline{\alpha}_{k}\right ) + h_{2}\right\ } \dot{b}\overline { z}_{k}\right ] .\label{alpha}\ ] ] the solution to @xmath997 and @xmath998 exist in closed form as @xmath999 ^{-1}e_{\theta}\left [ \overline{z}_{k}\dot{p}\left\ { h_{1}\widehat{b}+h_{3}\right\ } \right ] \label{eta1}\\ \widetilde{\overline{\alpha}}_{k}\left ( \theta\right ) & = -e_{\theta } \left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] ^{-1}e_{\theta}\left [ \overline{z}_{k}\dot{b}\left\ { h_{1}\widehat{p}+h_{2}\right\ } \right ] .\label{alpha1}\ ] ] next define @xmath1000 and @xmath1001and @xmath1002\ ] ] note the models @xmath1003 and @xmath1004 are used only to define the truncated parameter @xmath1005 they are not assumed to be correctly specified . in particular , the training sample estimates @xmath1006 @xmath1007 need not be based on the models @xmath1003,@xmath1008we now compare our truncated parameter @xmath930 with @xmath1009 and calculate the truncation bias .
it is important to keep in mind that @xmath1010 are components of the unknown @xmath785while @xmath1011,@xmath1012are regarded as known functions .
if our model satisfies @xmath1013 @xmath70and @xmath1014 then @xmath1015 further @xmath1016 $ ] immediate from theorem [ dr ] and lemma [ conde ] . we know from the above theorem that @xmath1017 for @xmath1018 however to control the truncation bias in forming confidence intervals for @xmath175 we will need to know how fast sup@xmath1019 decreases as @xmath848 increases@xmath1020the following theorem is a key step towards determining an upper bound . [
tbformula]suppose @xmath983 and @xmath1021 are chosen so that @xmath1022 \geq0 $ ] wp1 .
let @xmath1023 \right\ } ^{1/2}\ ] ] and @xmath1024 $ ] and @xmath1025 $ ] be , respectively , the projection in @xmath1026 of @xmath912 on the @xmath848 dimensional linear subspace @xmath1027 spanned by the components of the vector @xmath1028 and the projection on the orthocomplement of this subspace .
then if @xmath1013 are satisfied @xmath90@xmath1029 \pi^{\perp}\left [ \left ( \frac{b-\widehat{b}}{\dot{b}}\right ) q|q\overline{z}_{k}\right ] \right]\ ] ] [ q ] to simplify various formulae it is often convenient and aesthetically pleasing to have @xmath1030 we can choose @xmath1031 and @xmath1032 to guarantee @xmath1033 wp1@xmath64 for the functional @xmath1034 $ ] of example 1a , @xmath1035 wp1 . thus choosing @xmath1031 and @xmath1036 equal to @xmath18 and @xmath1037 , respectively , @xmath1038 makes @xmath1033 wp1 . in the missing data example 2a , the function @xmath1039
so @xmath1040 = 1/\widehat{p}$ ] and thus the choice @xmath1041makes @xmath1033 wp1 . note since inference on @xmath175 is conditional on the training sample data , we view the initial estimator @xmath979 of @xmath976 from the training sample as known and thus an analyst is free to choose @xmath1032 to be @xmath1042 * examples continued .
* in * example * * 1a , * recall @xmath1043 .$ ] choose @xmath1044 @xmath1038 so @xmath1045 and take @xmath1046 then @xmath1047 , \\ tb_{k } & = e\left\ { \left [ \pi^{\perp}\left [ b|\overline{z}_{k}\right ] \pi^{\perp}\left [ p|\overline{z}_{k}\right ] \right ] \right\ } , \\ \widetilde{\psi}_{k } & = \psi - tb_{k}=e\left\ { \pi\left [ b|\overline{z}_{k}\right ] \pi\left [ p|\overline{z}_{k}\right ] \right\}\end{aligned}\ ] ] thus @xmath969 appears to be the natural choice for a truncated parameter . in the * example 2a * with @xmath1048 , \dot{b}=-1,\dot { p}=\widehat{p}=1/\widehat{\pi},$ ]
@xmath1049 we obtain@xmath1050{c}\pi^{\perp}\left [ \widehat{\pi}\left ( \frac{1}{\pi}-\frac{1}{\widehat{\pi}}\right ) \left\ { \frac{\pi}{\widehat{\pi}}\right\ } ^{1/2}|\left\ { \frac{\pi}{\widehat{\pi}}\right\ } ^{1/2}\overline{z}_{k}\right ] \\ \times\pi^{\perp}\left [ \left\ { \frac{\pi}{\widehat{\pi}}\right\ } ^{1/2}\left ( b-\widehat{b}\right ) |\left\ { \frac{\pi}{\widehat{\pi}}\right\ } ^{1/2}\overline{z}_{k}\right ] \end{array } \right]\ ] ] thus the truncated parameter @xmath1051 does not seem to be a particular natural or obvious choice .
the complexity of @xmath969 is not simply due to the fact that we chose @xmath1052 rather than @xmath1053 as we now demonstrate . in * example 2a * with @xmath1054 @xmath1055@xmath1050{c}\pi^{\perp}\left [ \left ( \frac{1}{\pi}-\frac{1}{\widehat{\pi}}\right ) \pi^{1/2}|\pi^{1/2}\overline{z}_{k}\right ] \times\\ \pi^{\perp}\left [ \left\ { \frac{\pi}{\widehat{\pi}}\right\ } ^{1/2}\left ( b-\widehat{b}\right ) |\pi^{1/2}\overline{z}_{k}\right ] \end{array } \right]\ ] ] nonetheless we will see that , for either choice of @xmath1056 the parameter @xmath969 will result in estimators with good properties .
henceforth , given @xmath1057 @xmath1058 will always denote a complete orthonormal basis wrt to lebesgue measure in @xmath1059 or in the unit cube in @xmath1059 that provides optimal rate approximation for hlder balls @xmath1060 , i.e.@xmath1061 the basis consisting of @xmath1062 tensor products of univariate orthonormal polynomials satisfies @xmath1063 for all @xmath1064 the basis consisting of @xmath1062 tensor products of a univariate daubechies compact wavelet basis with mother wavelet @xmath1065 satisfying @xmath1066 also satisfies @xmath1063 for @xmath1067 [ tbrate]suppose that @xmath512 are satisfied , that @xmath986 and @xmath982 satisfy @xmath1068 and in the remainder of the paper , unless stated otherwise , we take @xmath1069 where recall that @xmath1070 \right\ } .$ ] then @xmath1071 we begin by proving that the first order influence functions of @xmath974and @xmath271 are identical except with @xmath1072 replacing @xmath1073 [ foif]@xmath1074\ ] ] with@xmath1075 since @xmath1076 $ ] , @xmath1077 if_{1,\widetilde{\overline{\eta}}_{k}\left ( \cdot\right ) } \left ( \theta\right ) \\ & + e\left [ h\left ( b^{\ast}\left ( x,\widetilde{\overline{\eta}}_{k}\left ( \theta\right ) \right ) , p^{\ast}\left ( x,\widetilde{\overline{\alpha}}_{k}\left ( \theta\right )
\right ) \right ) /\partial\overline{\alpha}_{k}^{t}\right ] if_{1,\widetilde{\overline{\alpha}}_{k}\left ( \cdot\right ) } \left ( \theta\right)\end{aligned}\ ] ] but , by definition of @xmath1078 and @xmath1079 both expectations are zero .
note that @xmath1078 and @xmath1080 are not maximizers of the expected log - likelihood for @xmath1081 and @xmath1082 this choice was deliberate . had we defined @xmath1083 and @xmath1080 as the maximizers of the expected log - likelihood , then @xmath1084 would have had additional terms since the expectations in the preceding proof would not be zero .
the existence of these extra terms would translate to many extra terms in @xmath932 for large @xmath135 leading to computational difficulties .
similarly had we chosen models @xmath1085 and @xmath1086 with @xmath1087 a non - linear inverse - link function , @xmath932 would also have had many extra terms without an improvement in the rate of convergence .
the following is proved in the appendix .
[ drhoif]@xmath1088 where @xmath973=@xmath1089 $ ] is a jth order degenerate @xmath187statistic given by @xmath1090{c}\left [ \left ( h_{1}\widetilde{p}+h_{2}\right ) \dot{b}\overline{z}_{k}^{t}\right ] _ { i_{1}}\left\ { e\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\\ \times\left [ \overline{z}_{k}\left ( h_{1}\widetilde{b}+h_{3}\right ) \dot { p}\right ] _ { i_{2}}\end{array } \right\ } \\ if_{jj,\widetilde{\psi}_{k},\overline{i}_{j } } & = \left ( -1\right ) ^{j-1}\left [ \left ( h_{1}\widetilde{p}+h_{2}\right ) \dot{b}\overline{z}_{k}^{t}\right ] _ { i_{1}}\\ \times & \left [ \begin{array } [ c]{c}{\displaystyle\prod\limits_{s=3}^{j } } \left\ { e\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\\ \left\ { \left ( \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{s}}-e\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline { z}_{k}^{t}\right ] \right\ } \end{array } \right ] \\ & \times\left\ { e\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\left [ \overline{z}_{k}\left ( h_{1}\widetilde{b}+h_{3}\right ) \dot{p}\right ] _ { i_{2}}\ ] ] we can now calculate the estimator @xmath1092by substitution of @xmath391 for @xmath179 in @xmath1093 to obtain the following .
suppose @xmath877 holds .
then @xmath1094 where @xmath1095 _ { i_{1}}\left [ \overline{z}_{k}\left ( h_{1}\widehat{b}+h_{3}\right ) \dot { p}\right ] _ { i_{2}}\\ \widehat{if}_{jj,\widetilde{\psi}_{k},\overline{i}_{j } } & = \left ( -1\right ) ^{j-1}\left\ { \begin{array } [ c]{c}\left [ \left ( h_{1}\widehat{p}+h_{2}\right ) \dot{b}\overline{z}_{k}^{t}\right ] _ { i_{1}}\left [ { \displaystyle\prod\limits_{s=3}^{j } } \left\ { \begin{array } [ c]{c}\left ( \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{s}}\\ -i_{k\times k}\end{array } \right\ } \right ] \\ \times\left [ \overline{z}_{k}\left ( h_{1}\widehat{b}+h_{3}\right ) \dot { p}\right ] _ { i_{2}}\end{array } \right\}\end{aligned}\ ] ] by lemma [ conde ]
@xmath1096 = e_{\widehat{\theta } } \left [ \left\ { h_{1}\widehat{p}+h_{2}\right\ } \dot{b}\overline{z}_{k}\right ] = 0.$ ] thus by eqs .
@xmath1097and @xmath1098 @xmath1099= @xmath1100 so @xmath1101 and @xmath1102 .
further , by eq.@xmath1103 @xmath1104 = \widehat{e}\left [ \dot { p}\dot{b}\widehat{e}\left [ h_{1}|x\right ] \overline{z}_{k}\overline{z}_{k}^{t}\right ] = \widehat{e}\left [ \widehat{q}^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] = \int\overline{\varphi}_{k}\left ( x\right ) \overline{\varphi}_{k}\left ( x\right ) ^{t}=i_{k\times k}$ ] .
it follows that by our judicious choice of @xmath1105 in eq.@xmath1106 we have avoided the need to invert a @xmath1107 matrix to compute @xmath971 the reader can easily check that when we take @xmath1108 \right\ } ^{1/2}\ ] ] @xmath1109 and @xmath792 wp 1@xmath90 @xmath1110 is precisely the same as @xmath1111 of equation ( [ 22 ] ) in section 3.2.1 . to make our procedures less abstract , we provide explicit expressions for @xmath1112 in examples 1a and 2a .
* example 1a * * continued : * @xmath1043 , $ ] @xmath1113 @xmath1114 @xmath1115 @xmath1116 = i_{k\times k}.$ ] then @xmath1117 and thus @xmath1118 _ { i_{1}}\left [ \overline{z}_{k}\left ( y-\widehat{b}\right ) \right ] _ { i_{2}}\\ \widehat{if}_{jj,\widetilde{\psi}_{k},\overline{i}_{j } } & = \left ( -1\right ) ^{j-1}\left [ \left ( a-\widehat{p}\right ) \overline{z}_{k}^{t}\right ] _ { i_{1}}\left [ { \displaystyle\prod\limits_{s=3}^{j } } \left\ { \overline{z}_{ki_{s}}\overline{z}_{ki_{s}}^{t}-i_{k\times k}\right\ } \right ] \left [ \overline{z}_{k}\left ( y-\widehat{b}\right ) \right ] _ { i_{2}}\ ] ] * example 2a continued : * @xmath1119 @xmath1120 , @xmath1048 , $ ] @xmath1121 and @xmath1122 so @xmath1123 = i_{k\times k}.$ ] then @xmath1124 @xmath1125 so@xmath1126 _ { i_{1}}\left [ \overline{z}_{k}\frac{a}{\widehat{\pi}}\left ( y-\widehat{b}\right ) \right ] _ { i_{2}}\\ \widehat{if}_{jj,\widetilde{\psi}_{k},\overline{i}_{j } } & = \left ( -1\right ) ^{j-1}\left [ \left ( \frac{a}{\widehat{\pi}}-1\right ) \overline{z}_{k}^{t}\right ] _ { i_{1}}\left [ { \displaystyle\prod\limits_{s=3}^{j } } \left\ { \frac{a}{\widehat{\pi}}\overline{z}_{k}\overline{z}_{k}^{t}-i_{k\times k}\right\ } _ { i_{s}}\right ] \left [ \overline{z}_{k}\frac { a}{\widehat{\pi}}\left ( y-\widehat{b}\right ) \right ] _ { i_{2}}\ ] ] consider example 2a with @xmath1127 , @xmath1128 = i_{k\times k},$ ] @xmath1129 , $ ] @xmath1125 so @xmath1130 _ { i_{1}}\left [ \overline{z}_{k}a\left ( y-\widehat{b}\right ) \right ] _ { i_{2}}\\ \widehat{if}_{jj,\widetilde{\psi}_{k},\overline{i}_{j } } & = \left ( -1\right ) ^{j-1}\left [ \left ( \frac{a}{\widehat{\pi}}-1\right ) \overline{z}_{k}^{t}\right ] _ { i_{1}}\left [ { \displaystyle\prod\limits_{s=3}^{j } } \left\ { a\overline{z}_{k}\overline{z}_{k}^{t}-i_{k\times k}\right\ } _ { i_{s}}\right ] \left [ \overline{z}_{k}a\left ( y-\widehat{b}\right ) \right ] _ { i_{2}}\ ] ] our next theorem , proved in the appendix , derives the estimation bias @xmath1131 -\widetilde{\psi}_{k}.$ ] [ ebrate]suppose @xmath1132 and @xmath512 hold then@xmath1133{c}e\left [ q^{2}\left ( \frac{b-\widehat{b}}{\dot{b}}\right ) \overline{z}_{k}^{t}\right ] \left\ { e\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] -i_{k\times k}\right\ } ^{m-1}\\ \times\left\ { e\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}e\left [ \overline{z}_{k}q^{2}\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) \right ] \end{array } \right\ } \label{eb3}\\ \leq\left\ { \begin{array } [ c]{c}\left\vert \left\vert \left\ { \frac{\dot{b}}{\dot{p}}g\right\ } ^{1/2}\right\vert \right\vert _ { \infty}\left\vert \left\vert \left\ { \frac{\dot{p}}{\dot{b}}g\right\ } ^{1/2}\right\vert \right\vert _ { \infty } \left\vert \left\vert \delta g\right\vert \right\vert _ { \infty}^{m-1}\left ( 1+o_{p}\left ( 1\right ) \right ) \times\\ \left\ { \int\left ( p\left ( x\right ) -\widehat{p}\left ( x\right ) \right ) ^{2}dx\right\ } ^{1/2}\left\ { \int\left ( b\left ( x\right ) -\widehat{b}\left ( x\right ) \right ) ^{2}dx\right\ } ^{1/2}\end{array } \right\ } \label{eb4}\\ = o_{p}\left ( \left ( \frac{\log n}{n}\right ) ^{\frac{\left ( m-1\right ) \beta_{g}}{d+2\beta_{g}}}n^{-\left ( \frac{\beta_{b}}{d+2\beta_{b}}+\frac{\beta_{p}}{d+2\beta_{p}}\right ) } \right ) \label{eb6}\ ] ] for @xmath1134 where @xmath1135 [ pnorm]at the cost of a longer proof we could have used hlder s inequality repeatedly to control @xmath1136 in the @xmath7 norm @xmath1137 with @xmath1138 to show that @xmath1139 thus , @xmath1140 is @xmath1141 consistent with the form of the bias given in our fundamental theorem [ eiet ] . * an alternate derivation of * @xmath971 the above derivation of @xmath956 required that one have facility in calculating higher order influence functions @xmath1142as done in the proof of theorem [ drhoif ] in the appendix .
however , there exists an alternate derivation of @xmath956 that does not require one learn how to calculate influence functions . specifically , we know from theorems [ eiet ] and [ eift ] that in a ( locally ) nonparametric model @xmath1143 is the unique @xmath465 order u - statistic that is degenerate under @xmath391 and satisfies @xmath1144 \equiv eb_{j}=o_{p}\left ( \left\vert \left\vert \widehat{\theta}-\theta\right\vert \right\vert ^{j+1}\right)\ ] ] with @xmath1145 -\widetilde{\psi}_{k}.$ ] in fact , we first derived @xmath956 by beginning with @xmath1146 , calculating @xmath1145 -\widetilde{\psi}_{k},$ ] and then , recursively for @xmath1147 finding @xmath1148 satisfying the above equation . for explicit details
see the appendix.in fact if one did not even know how to derive @xmath1149 one could begin the recursion by obtaining @xmath1150 as the unique first order u - statistic with mean zero under @xmath391 satisfying @xmath1151 = o_{p}\left ( \left\vert \left\vert \widehat{\theta}-\theta \right\vert \right\vert ^{2}\right ) $ ] . in this section ,
we derive the order of the variance of @xmath1152 when the orthonormal system @xmath1153 used to construct our @xmath1154-statistics are a compact wavelet basis .
first consider the case where @xmath8 is univariate ; without loss of generality , assume that @xmath1155.$ ] because we are primarily interested in convergence rates , the fact that @xmath8 may not follow the uniform distribution will not affect the rate results given below , but can influence the size of the constants .
we use @xmath1156 in place of @xmath1157 to indicate univariate basis functions .
let @xmath1158 @xmath848 be integer powers of two with @xmath1159 .
denote by @xmath1160 the @xmath992 dimensional basis vector whose first @xmath1161 components @xmath1162 are the @xmath1163vector of level @xmath1164 scaled and translated versions of a compactly supported father wavelet ( mallat , 1998 ) and whose last @xmath1165 components @xmath1166 are the associated compact mother wavelets between levels @xmath1164 and @xmath1167 . in particular , one may use periodic wavelets , folded wavelets or daubechies boundary wavelets with enough vanishing moments to obtain the optimal approximation rate of @xmath1168 for @xmath1169 @xmath64 the multiresolution analysis ( mra ) property of wavelets allows us to decompose the vector space spanned by the @xmath1170-level father wavelets @xmath1171 into the direct sum of the subspace spanned by @xmath1172-level father wavelets @xmath1173 and the span of mother wavelets for each level between @xmath1172 and @xmath1174 which we respectively write as @xmath1175 then for any integer @xmath212 with @xmath1176 we have@xmath1177 as @xmath1178 the resulting basis system is dense in @xmath1179 @xmath1180 since , in fact , @xmath8 is @xmath676dimensional we require a generalization that allows for multivariate tensor wavelet basis functions .
in fact , suppose @xmath1181 is now multivariate , and we again assume @xmath1182 on @xmath436 ^{d}.$ ] given @xmath11 univariate vector spaces @xmath1183 respectively spanned by vectors @xmath1184 so that for @xmath1185@xmath1186 and @xmath1187 one may define @xmath11 dimensional tensor vector spaces @xmath1188 such that@xmath1189 where for @xmath1190 @xmath1191 as @xmath1178 the resulting tensor basis system is dense in @xmath1192 next , suppose that we have a set of multivariate basis functions @xmath1193 such that for each @xmath1194 @xmath1195 spans @xmath1196 where @xmath1197 .
define @xmath1198 as the @xmath123 norm with respect to the lebesgue measure .
the following theorem is key to our derivation of the order of the variance of @xmath956 [ var_multi]for @xmath1199 @xmath1200 the following theorem is an immediate consequence of theorem ( [ var_multi ] ) obtained by taking @xmath1201 ( which implies we use the father wavelets at level log@xmath1202but no mother wavelets . )
[ var_if]for all @xmath383 @xmath1203 \asymp\frac{1}{n}\]]@xmath1204 \asymp\left ( \frac{1}{n}\max\left\ { 1,\left ( \frac{k}{n}\right ) ^{j-1}\right\ } \right)\ ] ] @xmath90 @xmath1205 \approx var_{\widehat{\theta}}\left [ \widehat{\mathbb{if}}_{m,\widetilde{\psi}_{k}}|\widehat{\theta}\right ] \asymp\frac{1}{n}\max\left\ { 1,\left ( \frac{k}{n}\right ) ^{m-1}\right\}\ ] ] we now use theorem @xmath1206 to derive the order of the conditional variance of @xmath956 given @xmath391 .
[ 3.19]if sup@xmath1207as @xmath1208 then for a fixed @xmath425 @xmath1209 & = var_{\theta}\left [ \widehat{\mathbb{if}}_{m,\widetilde{\psi}_{k}}|\widehat{\theta}\right ] \\ & = var_{\widehat{\theta}}\left [ \widehat{\mathbb{if}}_{m,\widetilde{\psi } _ { k}}|\widehat{\theta}\right ] \left ( 1+o_{p}\left ( 1\right ) \right ) \\ & \asymp\left ( \frac{1}{n}\max\left\ { 1,\left ( \frac{k}{n}\right ) ^{m-1}\right\ } \right)\end{aligned}\ ] ] the proof is in the appendix . for a given @xmath425 the estimator @xmath1210 that minimizes the maximum asymptotic mse over the model @xmath292 @xmath70defined by @xmath512 among the candidates @xmath1211 uses the value @xmath1212 of @xmath848 that equates the order @xmath1213 of @xmath1214 \ $ ] to the order @xmath1215 = \\ & \max\left [ \begin{array } [ c]{c}\left ( \frac{\log n}{n}\right ) ^{\frac{2\left ( m-1\right ) \beta_{g}}{d+2\beta_{g}}}n^{-\left ( \frac{2\beta_{b}}{d+2\beta_{b}}+\frac{2\beta_{p}}{d+2\beta_{p}}\right ) } , \\
k^{-\frac{2\left ( \beta_{b}+\beta_{p}\right ) } { d}}\end{array } \right]\end{aligned}\ ] ] of the maximal squared bias .
the estimator @xmath1216 that minimizes the maximum asymptotic mse over the model @xmath292 among all candidates @xmath1217is the estimator @xmath1218 which minimizes @xmath1219 we derive a consistent estimator of the variance and give the asymptotic distribution of @xmath1217 for any model and functional satisfying @xmath1220 let @xmath1221 be the upper @xmath1222quantile of a standard normal , i.e. a @xmath1223 distribution . * [ hoci ] * : \a ) letting @xmath1224 $ ] , @xmath1225 , \ ] ] for @xmath1226 and @xmath1227 where @xmath1228 is the symmetric kernel of @xmath1229 we have@xmath90 @xmath1230 & = \widehat{var}\left [ \widehat{\mathbb{if}}_{1,\widetilde{\psi}_{k}}|\widehat{\theta}\right ] \\ \widehat{e}\left [ \widehat{\mathbb{w}}_{jj,\widetilde{\psi}_{k}}^{2}\right ] & = \widehat{var}\left [ \widehat{\mathbb{if}}_{jj,\widetilde{\psi}_{k}}|\widehat{\theta}\right ] , \\
\widehat{e}\left [ \widehat{\mathbb{w}}_{m,\widetilde{\psi}_{k}}^{2}\right ] & = \widehat{var}\left [ \widehat{\mathbb{if}}_{m,\widetilde{\psi}_{k}}|\widehat{\theta}\right]\end{aligned}\ ] ] where @xmath1231 = var_{\widehat{\theta}}\left [ \cdot\right ] .$ ] \b ) conditional on the training sample , @xmath1232 \right\}\ ] ] converges uniformly for @xmath274 to a normal distribution with finite variance as @xmath596 .
the asymptotic variance is uniformly consistently estimated by @xmath1233 thus @xmath1234 \right\ } /\widehat{\mathbb{w}}_{m,\widetilde{\psi}_{k_{opt}\left ( m , n\right ) } } \ ] ] is converging in distribution to a standard normal distribution .
\c ) define the interval @xmath1235 suppose @xmath1236 then for @xmath1237 @xmath1238 @xmath1239 } { \sqrt{var_{\theta}\left [ \widehat{\psi}_{2,k^{\ast}}|\widehat{\theta}\right ] } } \right ] = o_{p}\left ( 1\right)\ ] ] and @xmath1240 converges uniformly in @xmath274 to a @xmath1241 .
moreover , @xmath1242 is a conservative uniform asymptotic @xmath1243 confidence interval for @xmath175 .
\d ) suppose we could derive a constant @xmath1244 and a constant @xmath1245 such that @xmath1246 \right\vert \\ & = \sup_{\theta}\left\vert \left\ { tb_{k_{opt}\left ( m , n\right ) } \left ( \theta\right ) + eb_{m}\left ( \theta\right ) \right\ } \right\vert \\ & \leq c_{bias}\left\ { \frac{1}{n}\max\left\ { 1,\left ( \frac { n^{\rho_{_{opt}\left ( m , n\right ) } } } { n}\right ) ^{m-1}\right\ } \right\ } ^{1/2}\ ] ] for @xmath1247 .
then @xmath1248 is a conservative uniform asymptotic @xmath1243 confidence interval for @xmath174 part a ) of the theorem is an easy calculation .
the asymptotic normality of @xmath1218 is based on new results on the asymptotic distribution of higher order @xmath340 with kernels depending on @xmath5 to be published elsewhere ( robins et al , 2007 ) .
part c of the theorem implies we obtain a conservative uniform asymptotic @xmath1243 confidence interval for any value of @xmath1249 exceeding @xmath1250 however , for the actual fixed sample size of our study , say @xmath12515000@xmath90 there is no guarantee the interval of part c based on given difference @xmath1252 say @xmath1253 will provide conservative finite sample coverage . because of this difficulty , a better approach , described in part d , would be to determine a constant @xmath1244 that can be used to bound the maximal bias under the model at a sample sizes exceeding @xmath1254 with @xmath1255 no greater than the actual fixed sample size @xmath5 of the study .
then the interval @xmath1256 will be a honest conservative finite sample @xmath159 confidence interval , provided that @xmath1257 has nearly converged to its normal limit at sample size @xmath126 unfortunately as yet we do not know how to determine the constants @xmath1244 and @xmath1258 of part d as a function of our model and of our initial estimator @xmath1259 this is an important open problem . * a model of increasing dimension : * the previous results can also be used for the analysis of models whose dimension increases with sample size . in fact , consider the @xmath1260 @xmath1261known , that differs from model @xmath310 in that , rather than assuming @xmath759 and @xmath764 live in particular hlder balls , we instead assume the working models of eqs .
[ worka ] and [ workb ] are precisely true for @xmath1262 so @xmath1263 and the dimensions of @xmath759 and @xmath1264 increase as @xmath1265 .
valid point and interval estimation for @xmath1266can still be based on the estimators @xmath956 except now ( i ) there is truncation bias only when @xmath1267 , ( ii ) the variance remains of the order of @xmath1268 and ( iii ) the estimation and trunction bias ( when it exists ) orders will be determined by any additional complexity reducing restrictions placed on the fraction of non - zero components or on the rate of decay of the components of the vectors @xmath1269and @xmath1270 and , for estimation bias , by @xmath688 as well . as a consequence , @xmath1271 and @xmath1272 under model
@xmath1273 will differ from their values under model @xmath296 note we need not take @xmath1274 as we did in the heurisitic discussion following remark @xmath1275 indeed @xmath492 in that discussion corresponds to the estimator in the class @xmath1276 with the fastest rate of convergence . in general , @xmath492 will have convergence rate slower than @xmath1277 furthermore , the discussion in section 4.1.1 implies that , when @xmath1278 and the minimax rate for estimation of @xmath175 is slower than @xmath430 even @xmath1279 will typically fail to converge at the minimax rate when complexity reducing restrictions have been imposed on @xmath1280and @xmath1281 .
* multi - robustness and a practical data analysis strategy : * [ multirob copy(1 ) ] conditional on @xmath414 for @xmath316 @xmath1282 is zero and thus estimator @xmath956 is unbiased for @xmath969 if @xmath1283 or @xmath1284 we refer to @xmath956 as triply - robust for @xmath970 generalizing robins and rotnitzky ( 2001 ) and van der laan and robins ( 2003 ) who referred to @xmath1285 as doubly - robust because of its being unbiased for @xmath969 if either @xmath1286 or @xmath1287 in fact , for @xmath1288 we can construct a modified estimator @xmath1289 that is @xmath1290 robust as follows . let @xmath1291 @xmath1292 denote @xmath1293 additional initial estimators of @xmath1294that differ from one another and from @xmath1295 define @xmath1296 where@xmath1297 @xmath1298 _ { i_{1}}\left\ { \left ( \dot{p}\dot { b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{2}}-i_{k\times k}\right\ } \\ \times & \left [ \begin{array } [ c]{c}{\displaystyle\prod\limits_{s=3}^{j-1 } } \left\ { \widehat{e}_{s}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\\ \left\ { \left ( \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{s}}-\widehat{e}_{s}\left [ \dot{p}\dot{b}h_{1}\overline { z}_{k}\overline{z}_{k}^{t}\right ] \right\ } \end{array } \right ] \\ & \times\left\ { \widehat{e}_{j}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\times\left [ \overline{z}_{k}\left ( h_{1}\widehat{b}+h_{3}\right ) \dot{p}\right ] _ { i_{j}}\ ] ] with @xmath1299 defined like @xmath1300 except with @xmath1301 replacing @xmath1302 in the appendix , we prove that @xmath1303 -\widetilde{\psi}_{k}$ ] is @xmath1304{c}e\left [ \dot{b}\dot{p}h_{1}\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) \overline{z}_{k}^{t}\right ] \left\ { e\left [ \dot{b}\dot{p}h_{1}\overline { z}_{k}\overline{z}_{k}^{t}\right ] -i_{k\times k}\right\ } \\ \times{\displaystyle\prod\limits_{s=3}^{m } } \left\ { \widehat{e}_{s}\left [ \dot{b}\dot{p}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\\ \times\left\ { e\left [ \dot{b}\dot{p}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] -\widehat{e}_{s}\left [ \dot{b}\dot{p}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } \\
\times\left\ { e\left [ \dot{b}\dot{p}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}e\left [ \dot{b}\dot{p}h_{1}\left ( \frac{b-\widehat{b}}{\dot{b}}\right ) \right ] \end{array } \right\ } \label{multirob_bias}\ ] ] which is zero if @xmath1305 @xmath1306 or if any of the @xmath1293 @xmath1301 equals @xmath1307 ( we note that if @xmath1308 or @xmath1309 @xmath1310 and thus @xmath1289 and @xmath956 are unbiased for @xmath271 . ) in settings where the dimension @xmath11 of @xmath8 is so large ( say @xmath1311 that the above asymptotic results fail as a guide to the finite sample performance of our procedures at the moderate sample sizes , say @xmath1312 commonly found in practice , one might consider , as a practical data analysis strategy , using the @xmath1290 robust estimator @xmath1313with @xmath1314 and the @xmath1315 selected by cross - validation as in van der laan and dudoit ( 2003 ) .
specifically , the training sample is split into two random subsamples - a candidate estimator subsample of size @xmath1316 and a validation subsample of size @xmath1317 where both @xmath1318 and @xmath1319 are bounded away from @xmath590 as @xmath171 a large number ( e.g. , @xmath1320 candidate parametric models of various dimensions and functional forms for @xmath160 and @xmath139 are fit to the candidate estimator subsample and the validation sample is used to find the candidate estimators @xmath1321and @xmath1322 for @xmath534 and @xmath1323 and the @xmath1324 candidate estimators @xmath616 and @xmath1325 @xmath1292 for @xmath139 with the smallest estimated risks ( with respect to an appropriate risk function such as squared error or kullback - leibler . ) an alternative approach would be to use the triply robust estimator @xmath956 with @xmath616 the candidate for @xmath139 with minimum estimated risk .
we plan to explore through simulation whether @xmath1289 outperforms @xmath956 in the setting of very high dimensional @xmath113
we consider a generic version in which we only assume a model and functional satisfying @xmath1220 to examine efficiency issues , we first consider the estimator @xmath1285 based on the first order influence function and sample splitting . without loss of generality we assume @xmath1326 ( otherwise simply interchange @xmath1327and @xmath1328 in what follows . ) it will be useful to consider the alternative parametrization @xmath1329 the ( conditional ) variance of @xmath1285 is of the order of @xmath1330 and the ( conditional ) bias of @xmath1285 in estimating @xmath271 is @xmath1331 if @xmath1332 and thus @xmath1333 the bias of @xmath1334 is @xmath1335 and @xmath1285 is not @xmath1336 consistent for @xmath271 when @xmath1337 at the other extreme , as @xmath1338 i.e. @xmath1339 the bias of @xmath1285 is @xmath1340which fails to be @xmath1336consistent@xmath70for any finite @xmath1341 * minimaxity with * @xmath139 * * known : * * to further examine efficiency issues , it is instructive to first consider the estimation of @xmath271 with @xmath1342 known .
if @xmath1342 were known , we could set @xmath1343 when calculating @xmath971 then @xmath1344 and @xmath1345 would therefore be an unbiased estimator of @xmath968 letting a superscript @xmath139 denote the model with @xmath139 known , it is easy to see that @xmath1346 would be @xmath1347 where @xmath1348 satisfies @xmath1349solving this , we find that when @xmath1350 is greater than or equal to @xmath1351 we can take @xmath1352 and @xmath1353 regardless of @xmath1354 which is , of course , the minimax rate .
in contrast if @xmath1350 @xmath1355 @xmath1356 and @xmath1357 in an unpublished paper , we have proved that this is the minimax rate when @xmath148 is known .
this raises the question of whether the lower bounds of rate @xmath1358 for @xmath1350 @xmath1359 and/or rate @xmath1360 for @xmath1350 @xmath1361are still achievable when @xmath139 is unknown , without restrictions on the smoothness of @xmath1362 before addressing this question , we take the opportunity to compare the relative efficiencies of competing rate - optimal unbiased estimators in the case of @xmath139 known
. this discussion will provide further insight into the results given in remark [ jj1 ] for models which are not locally nonparametric . * relative efficiency of various unbiased estimator with * @xmath139 * * known * * : for simplicity , we restrict the following discussion to the truncated version of the parameter @xmath1363 , $ ] with @xmath1364,$ ] @xmath1342 known , and @xmath0 bernouilli . for
this choice of @xmath1365 @xmath1366 is the marginal density of @xmath113 in this subsection , we assume @xmath1367is chosen equal to the known @xmath1368 so @xmath1369 = i_{k\times k}.$ ] also we choose @xmath1370 and take @xmath1371 so @xmath1372 = e\left [ b\overline{z}_{k}^{t}\right ] \overline{z}_{k}$ ] and @xmath1373 \right\ } ^{2}\right ] $ ] do not depend on @xmath1374 further we only concern ourselves with efficiency relative to the @xmath5 observations in the estimation sample .
we thus ignore any efficiency loss from using @xmath393 observations to construct @xmath1007 .
let @xmath1375 \right\ } \subset\theta$ ] denote the subset of @xmath245 corresponding to the known @xmath139 , which consists of all functions from the unit cube in @xmath1059 to the unit interval .
the model @xmath1376 is not locally nonparametric .
for example , the @xmath1377 order tangent space @xmath1378 does not include first order scores for @xmath1362 its 2nd order tangent space @xmath1379 does not contain second order scores for @xmath139 or mixed scores for @xmath139 and @xmath1380 rather , @xmath1381 is the closed linear span of the first and second order scores for @xmath533
. thus @xmath1382 \right\ } < \infty ; a\in\mathcal{a},c\in\mathcal{c}\ \}\ ] ] where @xmath1383 , \ ] ] and @xmath1384 and @xmath1385 are the set of one and two dimensional functions of @xmath1386 since , for @xmath1387 @xmath1388{c}\left [ \widehat{b}^{2}+2\widehat{b}\left ( y-\widehat{b}\right ) \right ] _ { i}\\ + \left [ \left ( y-\widehat{b}\right ) \overline{z}_{k}^{t}\right ] _ { i}\left [ \overline{z}_{k}\left ( y-\widehat{b}\right ) \right ] _ { j}\end{array } \right\}\end{aligned}\ ] ] is unbiased for @xmath1389 \right\ } ^{2}\right ] $ ] in model @xmath1376 , we know , by remark [ jj1 ] , that @xmath1390for @xmath1391 is the projection @xmath1392 $ ] of the 2nd order influence function @xmath1393 onto @xmath1394 now if @xmath1395 was an element of @xmath1396 @xmath1393 would equal @xmath1397 and thus be 2nd order ` unbiased locally efficient ' , at @xmath1398 as defined earlier in remark [ jj1 ]
. however we show below that , when @xmath1399 for some @xmath1400 wp1 does not hold , @xmath1401 is not an element of @xmath1381 for any @xmath533 .
rather , a straightforward calculation gives @xmath1402{c}\left [ 2e\left [ b\overline{z}_{k}^{t}\right ] \overline{z}_{k}\left ( y - b\right ) \right ] _ { i}\\ + \left [ \left ( y - b\right ) \overline{z}_{k}^{t}\right ] _ { i}\left [ \overline{z}_{k}\left ( y - b\right ) \right ] _ { j}\end{array } \right\ } .\ ] ] now one can check that @xmath1403 is a function of @xmath1404 so by theorem [ global_eff ] of remark [ jj1 ] , we conclude no unbiased globally efficient estimator exists .
however , we prove below that @xmath1405 and @xmath1345 have identical means .
it follows that @xmath1406 is an unbiased estimator of @xmath1407 \right ) ^{2}\right ] $ ] for any @xmath1408 .
thus , for a given choice of @xmath1409 @xmath1405 is 2nd order unbiased locally efficient at @xmath1410 however , one can show using a proof analogous to that in theorem [ 3.19 ] that for @xmath1411 @xmath1412 /var_{b}\left [ \mathbb{if}_{2,\widetilde{\psi}_{k}}^{eff}\left ( b\right ) \right ] \\ & = 1+o_{p}\left ( \left\vert \left\vert \widehat{b}-b\right\vert \right\vert _
{ \infty}\right ) .\end{aligned}\ ] ] henceforth we assume that @xmath533 lies in a holder ball @xmath1413 that is we consider the submodel @xmath1414and assume @xmath1415 converges to @xmath533 in sup norm at the optimal rate of @xmath1416uniformly over @xmath1417 the submodel and the model @xmath1418 have identical tangent spaces
. for all @xmath1419 @xmath1420 has an asymptotic distribution with mean zero and variance equal to @xmath1421 $ ] for all @xmath1422 _ _ _ _ in a slight abuse of language , we shall refer to @xmath1423 $ ] as the asymptotic variance of @xmath1424thus , as with standard first order theory , even when no unbiased estimator has finite sample variance that attains the bhattacharyya bound for all @xmath1425 , there can exist an unbiased estimator sequence whose asymptotic variance does attain the bound globally .
we next compare the means and variances of @xmath1426 and @xmath1427 now the two estimators are algebraically related by @xmath1428 -e\left [ \widehat{b}^{2}\right ] \right\ } .\ ] ] since @xmath1429 -e\left [ \widehat{b}^{2}\right ] $ ] is unbiased for zero , we conclude that @xmath1345 and @xmath1405 have the same mean but @xmath1430 /var_{b}\left [ \mathbb{if}_{2,\widetilde{\psi}_{k}}^{eff}\left ( b\right ) \right ] > 1 $ ] except when @xmath1431 @xmath1432for some @xmath1433 thus , since @xmath1434has asymptotic variance @xmath1435 $ ] and , except when @xmath1436 @xmath1437 \right ) \asymp\ n^{-1},$ ] we conclude the asymptotic variance of @xmath1345 attains the bound @xmath1438 $ ] when @xmath1439 but exceeds the bound when @xmath1440 , except when @xmath1441 .
finally , for completeness , robins and van der vaart ( 2006 ) considered an alternative particularly simple rate - optimal unbiased estimator of @xmath1442 \right\ } ^{2}\right ] $ ] given by @xmath1443 _ { i}\left [ \overline{z}_{k}y\right ] _ { j}\right\ } $ ] .
the hoeffding decomposition of @xmath1444 is @xmath1445 \overline{z}_{k}y-\widetilde{\psi}_{k}\left ( b\right ) \right ] + \mathbb{v}\left\ { \left [ y\overline{z}_{k}^{t}-e\left [ b\overline{z}_{k}^{t}\right ] \right ] _
{ i}\left [ \overline{z}_{k}y - e\left [ b\overline{z}_{k}\right ] \right ] _ { j}\right\ } \\
& = \mathbb{if}_{2,\widetilde{\psi}_{k}}^{eff}\left ( b\right ) + q+t\end{aligned}\ ] ] with @xmath1446 b-\psi\right\ } \right ] \\
t & = \mathbb{v}\left\ { \begin{array } [ c]{c}2\left ( b_{i}\overline{z}_{k , i}^{t}\overline{z}_{k , j}-\pi\left [ b|\overline{z}_{k}\right ] _ { j}\right ) \left ( y - b\right ) _ { j}\\ + b_{i}\overline{z}_{k , i}^{t}\overline{z}_{k , j}b_{j}-\pi\left [ b|\overline { z}_{k}\right ] _ { i}b_{i}-\pi\left [ b|\overline{z}_{k}\right ] _ { j}b_{j}+\psi \end{array } \right\}\end{aligned}\ ] ] since , except when @xmath1447 wp1 , @xmath1448 and @xmath1449 we conclude that the asymptotic variance of @xmath1450 exceeds the bound @xmath1438 $ ] regardless of whether @xmath1451 does or does not hold except when @xmath1452 wp1 .
* minimaxity with unknown * @xmath139 * * and * * @xmath1350 * * * * @xmath1453 we now show that the bound @xmath1358 for @xmath1350 @xmath1359 is achievable for each @xmath86 consider the estimator @xmath956 with @xmath1454 and @xmath1455 so that @xmath1456 is @xmath1457 then @xmath1458 @xmath1459 and @xmath1460 so @xmath956 will be @xmath1461consistent for @xmath279 if @xmath1462 and @xmath1463 the above expression implies that @xmath1464 for @xmath1461consistency . similarly , if @xmath1465 i.e. @xmath1339 it is necessary that @xmath1466 for @xmath1467consistency .
these results imply that estimators @xmath956 in our class can always achieve @xmath1461consistency whenever @xmath144 , but for fixed @xmath1468 the order @xmath135 of the required @xmath187statistic increases without bound as the smoothness @xmath688 of @xmath139 approaches zero .
* efficiency : * we now show that when @xmath1350 is strictly greater than @xmath1351 we can construct an unconditional asymptotically linear estimator based on all @xmath392 subjects with influence function @xmath1469 by having the number of the @xmath392 subjects allotted to the validation sample and analysis sample be @xmath1470 and @xmath1471 respectively , for @xmath1472 .
it then follows from van der vaart@xmath1473 that the estimator is regular and semiparametric efficient .
specifically , suppose @xmath1474 @xmath1475 consider the estimator @xmath1476 with @xmath1477 so that @xmath1478 is @xmath1479and @xmath1480 so that @xmath1481 and @xmath1482 = o_{p}\left ( 1/n\right ) $ ] for @xmath1483then , by our previous results , @xmath1484 it remains to show that @xmath1485 but the lhs is @xmath1486 * adaptivity when * @xmath1350 * * * * @xmath1487 * * :* * we next prove that if we let @xmath1488with @xmath1489 and @xmath1490 @xmath1491 will be semiparametric efficient for each @xmath1492 , provided @xmath1493 clearly , the truncation bias is @xmath1494 the estimation bias @xmath1495 is @xmath1496 } n^{\ -\left ( 1-\epsilon\right ) \left\ { \frac{\beta_{b}}{d+2\beta_{b}}+\frac{\beta_{p}}{d+2\beta_{p}}\right\ } } \right ) .$ ] thus @xmath1497 if @xmath1498 } = o\left ( n^{-\frac{1}{2}+\left ( 1-\epsilon\right ) \left\ { \frac{\beta_{b}}{d+2\beta_{b}}+\frac{\beta_{p}}{d+2\beta_{p}}\right\ } } \right ) .$ ] so we require @xmath1499 ln\left\ { m\left ( n^{\left ( 1-\epsilon\right ) } \right ) \right\ } /\left [ \frac{1}{2}-\left ( 1-\epsilon\right )
\left\ { \frac{\beta_{b}}{d+2\beta_{b}}+\frac{\beta_{p}}{d+2\beta_{p}}\right\ } \right ] \ln\left ( n\right ) \rightarrow\infty,\ ] ] which is satisfied if @xmath1500 in the appendix we prove that var@xmath1501 = var_{\widehat{\theta}}\left [ \widehat{\psi}_{m , k\ } \right ] \left\ { 1+o_{p}\left ( 1\right ) \right\ } $ ] provided @xmath1502 now @xmath1503 = \frac{1}{n}var_{\widehat{\theta}}\left\ { if_{1,\psi , i}\left ( \widehat{\theta } \right ) \right\ } \left [ o\left ( \sum_{l=0}^{m\left ( n\right ) } \left\ { \ln n\right\ } ^{-l}\right ) \right ] .{\huge \ } $ ] but @xmath1504 @xmath1505 } } { 1-\left\ { \ln n\right\ } ^{-1}}\right ) = o\left ( 1+\left\ { \ln n\right\ } ^{-1}\right ) , $ ] so var@xmath1506 $ ] is @xmath1507 the proof of efficiency now proceeds as above .
* alternative estimators when * @xmath1350 * * * * @xmath1487 * * :* * when @xmath1350 @xmath1508 there actually exist , at least for certain functionals in our class , @xmath1461consistent estimators of @xmath271 that are much simpler than our very high order u - statistic estimators.@xmath70for example consider the expected conditional covariance @xmath1509 $ ] of example 1b with @xmath1510 .
* example 1b ( cont ) * : number the study subjects @xmath1511 ordered by their realized values @xmath1512 where we have not split the sample@xmath64 following wang et al .
( 2006 ) , consider the difference -based estimator estimator @xmath1513 which has conditional mean given @xmath1514 of @xmath1515 hence @xmath1516 \\ & = n^{-1}e\left [ \sum_{i=0}^{n/2 - 1}\left\ { b\left ( x_{i+1}\right ) -b\left ( x_{i}\right ) \right\ } \left\ { p\left ( x_{i+1}\right ) -p\left ( x_{i}\right ) \right\ } \right ] \\
& = o_{p}\left ( n^{-1}{\textstyle\sum\limits_{i=0}^{\frac{n}{2}-1 } } e\left\ { x_{i+1}-x_{i}\right\ } ^{2\beta}\right ) = o\left ( n^{-2\beta } \right)\end{aligned}\ ] ] by the theory of spacings ( pyke , 1965 ) . but
@xmath1517 is @xmath1518 when @xmath1519 @xmath1520 the variance of @xmath1521 is @xmath1522so @xmath1521 is @xmath1523consistent@xmath64 however , @xmath1524 so @xmath1521 is not ( semiparametric ) efficient . as discussed by arellano ( 2003 ) , by using a @xmath135th order rather than a second order difference operator and letting @xmath1525 at an appropriate rate as @xmath1526 the @xmath135th order estimator @xmath1521 can be made efficient .
* minimaxity with unknown * @xmath139 * * and * * @xmath1350 * * * * @xmath1527 consider next whether the lower bound of @xmath1360 for @xmath1350 @xmath1528 is achievable when @xmath139 is unknown@xmath70but @xmath144 .
we will show the next section that the bound @xmath1529 is achievable provided @xmath1530 i.e. , @xmath1531 to attain the bound @xmath1532 whenever eq.@xmath1533 holds , we introduce new more efficient estimators , owing to the fact that an estimator @xmath956 in our class can attain the bound @xmath1532 only in the special case where the second order estimation bias @xmath1534 is less than @xmath1535 for a fixed @xmath1536 the right hand side of eq.@xmath1537 is minimized over @xmath1538 at @xmath107 at @xmath99 eq.@xmath1537 reduces to @xmath1539 the right hand side of eq.@xmath1537 increases with @xmath1540 with asymptote equal to twice the rhs of eq.([bounde ] ) as @xmath1541 hence , in order to attain the optimal rate @xmath1532 when @xmath1542 and @xmath1543 the quantity @xmath1544 must be twice as large as when @xmath1545 in the next section , we construct an estimator with a convergence rate of @xmath1546 at the cut - point @xmath1547 in this paper we do not consider the construction of estimators that are rate optimal below this cutpoint . however , for the special case @xmath99 in an unpublished paper li et .
2007 ) have constructed estimators which converge at a rate given in eq.@xmath101 , whenever inequality @xmath1537 fails to hold .
we conjecture that this rate is minimax , possibly only up to @xmath1548 factors , when inequality @xmath1537 fails to hold and @xmath107 at the cut - point @xmath1549 we obtain @xmath1550 and thus@xmath70eq.@xmath101 becomes@xmath1551 in agreement with the rate of the estimator of section [ case2 ] below . in the extreme case in which @xmath116 with @xmath1519 remaining fixed , @xmath1552 @xmath1553 which agrees ( up to a log factor ) with the rate of @xmath111 given by the simple estimator of wang et al .
( 2006 ) analyzed above under `` example 1b ( cont ) '' .
* improved rates of convergence with * @xmath8 * * random in a semiparametric model : * * we now , as promised in the introduction , construct an estimator of @xmath117 under the homoscedastic model @xmath1554 = b\left ( x\right ) , $ ] @xmath1555 = \sigma^{2}$ ] with @xmath8 random with unknown density that , whenever @xmath1556 and , regardless of the smoothness of @xmath9 , converges at the rate @xmath120 which is faster than equal - spaced non - random minimax rate of @xmath121 specifically we divide the support of @xmath652 i.e. , the unit cube in @xmath1557 into @xmath1558 identical subcubes with edge length @xmath1559 we continue to assume the unknown density @xmath9 is absolutely continuous wrt to lebesgue measure and both it and its inverse are bounded in sup - norm . then it is a standard probability calculation that the number of subcubes containing at least two observations is @xmath1560 we estimate @xmath117 in each such subcube by @xmath1561 where , for any subcube with @xmath1562 or more observations,@xmath1563 and @xmath1564 are chosen randomly , without replacement .
our final estimator of @xmath117 is the average of our subcube - specific estimates @xmath1565 over the @xmath1566 subcubes with at least two observations .
the rate of convergence of the estimator is minimized at @xmath92 by taking @xmath1567 as we now show .
we note that @xmath1568 = \sigma^{2}+\left\ { b\left ( x_{i}\right ) -b\left ( x_{j}\right ) \right\ } ^{2}/2,$ ] @xmath1569 by @xmath119 and @xmath1570when @xmath1571 and@xmath1572 are in the same subcube .
it follows that the estimator has variance @xmath1573 and bias of @xmath1574 to minimize the convergence rate we equate the orders of the variance and the squared bias by solving @xmath1575 which gives @xmath1576 our random design estimator has better bias control and hence converges faster than the optimal equal - spaced fixed @xmath8 estimator , because the random design estimator exploits the @xmath1577 random fluctuations for which @xmath1578 corresponding to two different observations are a distance of @xmath1579 apart .
our estimator will not converge at rate @xmath92 to @xmath1580 $ ] in our nonparametric model , because it then no longer suffices to average estimates of @xmath1581 only over subcubes containing 2 or more observations . in a ( locally ) nonparametric model
@xmath366 the estimator @xmath1582 is essentially the unique @xmath1583 order u - statistic estimator of the truncated parameter @xmath969 for which the leading term in the bias is @xmath1584 however , when the minimax rate of convergence for @xmath271 is slower than @xmath63 , other @xmath1585 order u - statistics estimators will often converge to @xmath969 ( and thus @xmath1586 at a faster rate uniformely over the model than does any estimator @xmath956 ( constructed from an estimated higher order influence function @xmath1587 for @xmath1588 by tolerating bias at orders less than @xmath286 in exchange for a savings in variance .
[ t7]a heuristic understanding as to why this is so can be gained from the following considerations .
the theory of higher order influence functions as developed in theorems 2.2 and 2.3 is a theory of score functions ( derivatives ) .
thus it can directly incorporate the restriction that a function , say @xmath1589 has an expansion @xmath1590 for which @xmath1591 for @xmath1592 as the restriction is equivalent to various scores being equal to zero .
however the theory can not directly incorporate restrictions such as @xmath1593 or @xmath1594 that do not imply any restrictions on score functions .
thus to find an optimal estimator , one must perform additional side calculations to quantify the estimation and truncation bias of various candidate estimators under these restrictions . as the assumption that @xmath759 lies in a holder ball
can be expressed in terms of such restrictions , this remark is relevant to a search for an optimal rate estimator .
we now construct such estimators .
we first consider the case where @xmath1595 and @xmath688 are such that the estimation bias @xmath1596 of the second order estimator is greater than @xmath1597 but the estimation bias @xmath1598 of the third order estimator is less than @xmath1599 that is @xmath1600 then the most efficient estimator @xmath1491 in our class has rate of convergence slower than @xmath1532 because @xmath1601 converges at rate @xmath1602 determined by the 2nd order estimation bias and , for @xmath1603 @xmath1210 converges at a rate no faster than @xmath1604 [ we obtained @xmath1605 as @xmath1606 where @xmath848 solves the equation @xmath1607 that equates the variance @xmath1608 of @xmath327 to the squared truncation bias @xmath1609 to describe our more efficient estimator , define for nonnegative integers @xmath1610 with @xmath1611 and @xmath1612 the @xmath187statistic @xmath1613 with @xmath1614 _ { i_{2}}-i_{\left\ { k\left ( 1\right ) -k\left ( 0\right ) \right\ } \times\left\ { k^{\ast } \left ( 1\right ) -k^{\ast}\left ( 0\right ) \right\ } } \right ) \overline { z}_{k^{\ast}\left ( 0\right ) , i_{3}}^{k^{\ast}\left ( 1\right ) } \widehat{\delta}_{i_{3}}\\ & = \sum_{s_{1}=k\left ( 0\right ) + 1}^{k\left ( 1\right ) } \sum _ { s_{2}=k^{\ast}\left ( 0\right ) + 1}^{k^{\ast}\left ( 1\right ) } \left\ { \begin{array } [ c]{c}\widehat{\epsilon}_{i_{1}}z_{s_{1}}\left ( x_{i_{1}}\right ) \times\\ \left\ { \left [ \dot{b}\dot{p}h_{1}\right ] _ { i_{2}}z_{s_{1}}\left ( x_{i_{2}}\right ) z_{s_{2}}\left ( x_{i_{2}}\right ) -i\left [ s_{1}=s_{2}\right ] \right\ } z_{s_{2}}\left ( x_{i_{3}}\right ) \widehat{\delta } _ { i_{3}}\end{array } \right\ } , \end{aligned}\ ] ] where @xmath1615 @xmath1616 with @xmath1617 as an example @xmath1618 we can identify @xmath1619 with the rectangle in @xmath1620 defined by @xmath1621 with @xmath1622 and @xmath1623 respectively , the vertices closest and furthest from the origin . thus @xmath1624 is identified with the rectangle
@xmath1625 indeed we can write @xmath1626{c}\widehat{\epsilon}_{i_{1}}z_{s_{1}}\left ( x_{i_{1}}\right ) \times\\ \left\ { \left [ \dot{b}\dot{p}h_{1}\right ] _ { i_{2}}z_{s_{1}}\left ( x_{i_{2}}\right ) z_{s_{2}}\left ( x_{i_{2}}\right ) -i\left [ s_{1}=s_{2}\right ] \right\ } z_{s_{2}}\left ( x_{i_{3}}\right ) \widehat{\delta } _ { i_{3}}\end{array } \right\}\end{aligned}\ ] ] where , here and below ,
@xmath1627 and @xmath1628 are restricted to be integers , so @xmath1629 are the lattice points of the rectangle .
we next study the variance of @xmath1630 it follows from theorem [ var_multi ] above that the number of lattice points in @xmath1619 is proportional to the variance of @xmath1631 so if @xmath1632 and @xmath1633 then @xmath1634 $ ] and @xmath1635 $ ] are both of order @xmath1636 hence the order of the variance of @xmath1637 is determined by the vertex of the rectangle @xmath1638 furthest from the origin .
in contrast by a theorem in the appendix , the mean @xmath1639 $ ] is @xmath1640 \delta g\widehat{q}^{2}\widehat{\pi } \left [ \delta p|\overline{z}_{k^{\ast}\left ( 0\right ) } ^{k^{\ast}\left ( 1\right ) } \right ] \right ) \left ( 1+o_{p}\left ( 1\right ) \right)\ ] ] with @xmath1641 @xmath1642 @xmath1643 @xmath1644 and @xmath1645 it follows that if @xmath1632 and @xmath1633 then @xmath1639 $ ] and @xmath1646 $ ] are both of order @xmath1647 .\ ] ] to see this for @xmath1648 $ ] , we @xmath1649sup out @xmath1650 from @xmath1651 \delta g\widehat{q}^{2}\widehat{\pi}\left [ \delta p|\overline{z}_{k^{\ast } \left ( 0\right ) } ^{k^{\ast}\left ( 1\right ) } \right ] \right\vert \right ) $ ] which is @xmath1652 \widehat{e}\left ( \left\vert \widehat{\pi}\left [ \delta b|\overline{z}_{k\left ( 0\right ) } ^{k\left ( 1\right ) } \right ] \widehat{\pi}\left [ \delta p|\overline{z}_{k^{\ast}\left ( 0\right ) } ^{k^{\ast}\left ( 1\right ) } \right ] \right\vert \right ) .\ ] ] we then apply cauchy schwartz to @xmath1653 \widehat{\pi}\left [ \delta p|\overline{z}_{k^{\ast}\left ( 0\right ) } ^{k^{\ast}\left ( 1\right ) } \right ] \right\vert \right ) , $ ] noting that @xmath1654 \right\ } ^{2}\right ) ^{1/2}=o\left ( k\left ( 0\right ) ^{-\beta_{b}}\right ) $ ] .
again a more careful argument using hlder s inequality would show the log factor is unnecessary .
hence the order of the mean of @xmath1637 is determined by the vertex of the rectangle @xmath1638 closest to the origin .
* motivation : * with this background we are ready to motivate our new estimator . recall from section [ dr_ci_section ] , that with @xmath139 known , the choice @xmath1655 gives @xmath1656 because the truncation bias @xmath1657 and variance are of order @xmath1360 and the estimation bias is zero .
any choice of @xmath848 larger than @xmath1658 will result in a slower rate of convergence .
however , when @xmath139 is unknown and thus estimated , @xmath1659 does not attain the optimal rate of convergence because the estimation bias @xmath1660 exceeds @xmath1661 the estimator @xmath1662 also fails to attain the rate @xmath1360 because it has variance of the order of @xmath1663 which exceeds @xmath1664 on the other hand , @xmath1665 has bias of @xmath1666 because the truncation bias is @xmath1666 and the estimation bias @xmath1667 is also @xmath1666 under our assumption @xmath1668 .
our strategy will be to try to replace the term @xmath1669 in the estimator @xmath1670 by @xmath1671{c}\widehat{\epsilon}_{i_{1}}z_{s_{1}}\left ( x_{i_{1}}\right ) z_{s_{2}}\left ( x_{i_{3}}\right ) \widehat{\delta}_{i_{3}}\times\\ \left\ { \left [ \dot{b}\dot{p}h_{1}\right ] _ { i_{2}}z_{s_{1}}\left ( x_{i_{2}}\right ) z_{s_{2}}\left ( x_{i_{2}}\right ) -i\left [ s_{1}=s_{2}\right ] \right\ } \end{array } \right\}\ ] ] where @xmath1672 is a subset of the rectangle @xmath1673 such that @xmath1674 but the additional bias @xmath1675 \\ & = e\left [ \widehat{\mathbb{u}}_{3}\left ( \left ( _ { 0,}^{k_{opt}^{g}\left ( 2\ \right ) , } \genfrac{}{}{0pt}{}{k_{opt}^{g}\left ( 2\ \right ) } { 0}\right ) \backslash\omega\right ) \right ] \\ & \equiv e\left [ \sum_{\left ( s_{1},s_{2}\right ) \in\left ( _ { 0,}^{k_{opt}^{g}\left ( 2\ \right ) , } \genfrac{}{}{0pt}{}{k_{opt}^{g}\left ( 2\ \right ) } { 0}\right ) \backslash\omega}\left\ { \begin{array } [ c]{c}\widehat{\epsilon}_{i_{1}}z_{s_{1}}\left ( x_{i_{1}}\right ) \left\ { \begin{array } [ c]{c}\left [ \dot{b}\dot{p}h_{1}\right ] _ { i_{2}}z_{s_{1}}\left ( x_{i_{2}}\right ) z_{s_{2}}\left ( x_{i_{2}}\right ) \\ -i\left [ s_{1}=s_{2}\right ] \end{array } \right\ } \\
\times z_{s_{2}}\left ( x_{i_{3}}\right ) \widehat{\delta}_{i_{3}}\end{array } \right\ } \right]\end{aligned}\ ] ] is @xmath1676 this approach will succeed if we can chose @xmath1672 and thus @xmath1677 to be sums of rectangles ( whose number does not increase with @xmath1678 such that ( i ) each rectangle in @xmath1679 has its closest vertex to the origin , say @xmath1680 satisfying @xmath1681 \leq n^{-\frac { 4\beta\ } { 4\beta+d}}$ ] and ( ii ) simultaneously each rectangle in @xmath1672 has its furthest vertex from the origin , say @xmath1682 satisfying @xmath1683 we index the vertices of our set of rectangles as follows . consider a natural number @xmath1684 and a set of non - negative integers @xmath1685 satisfying @xmath1686 note the elements with even subscripts increase from @xmath590 to @xmath1687 while elements with odd subscripts decrease from @xmath1037 to @xmath1688 further the smallest element with odd subscript equals the largest element with even subscript .
we will use two such sets @xmath1689 and @xmath1690 with corresponding elements @xmath1691 and @xmath1692with @xmath1693.@xmath70 @xmath70set for @xmath1694 @xmath1695 we leave @xmath1684,@xmath1696 and @xmath1697 unspecified for now but derive optimal values below .
let @xmath1698 be the union of rectangles @xmath1699 the points @xmath1700 for @xmath1701 lie on a hyperbola @xmath1702 in @xmath1620 defined by @xmath1703 shown in figure 1 for @xmath1704 the set @xmath1705 lies below @xmath1706 define@xmath1707 we then have [ uy](i ) : the estimator @xmath1708 has variance of the order of @xmath1709 and bias @xmath1710 of order @xmath1711 proof : each of the @xmath1712 rectangles whose union is @xmath1713 has @xmath1714 or @xmath1715 for some @xmath1694 as the vertex furthest from the origin and thus contributes @xmath1716 to the variance of @xmath1717 the variance of @xmath1718 now @xmath1719 -e\left [ \widehat{\mathbb{u}}_{3}\left\ { \left ( _ { 0,}^{k_{-1},}\genfrac{}{}{0pt}{}{k_{-1}}{0}\right ) \right\ } \right ] \right\ } \\ & = o_{p}\left ( k_{-1}^{-\left ( \beta_{p}+\beta_{b}\right ) /d}\right ) + o_{p}\left ( n^{-\left ( \frac{2\beta_{g}}{2\beta_{g}+d}+\frac{\beta_{b}}{d+2\beta_{b}}+\frac{\beta_{p}}{d+2\beta_{p}}\right ) } \right ) \\ & + e\left [ \widehat{\mathbb{u}}_{3}\left\ { \left ( _ { 0,}^{k_{-1},}\genfrac{}{}{0pt}{}{k_{-1}}{0}\right ) \backslash\omega\left ( \left ( \mathcal{k}_{pj},\mathcal{k}_{bj}\right ) \right ) \right\ } \right]\end{aligned}\ ] ] as is evident from figure 1 , @xmath1720 is the union of rectangles @xmath1721which have @xmath1722 as the set of vertices closest to the origin , leading to the expression for the bias given in the theorem .
[ uz]given @xmath1723with @xmath1724 so @xmath1725 eq.(@xmath1726 ) holds if and only if there exists @xmath1727 such that @xmath1728 if eq.(@xmath1726 ) holds , @xmath1729 = o_{p}\left ( n^{-\frac{4\beta } { 4\beta+d}}\right ) $ ] and thus @xmath1730 when we choose @xmath1684 to be the smallest integer such that @xmath1731 with @xmath1732 @xmath1733 , @xmath1734for @xmath1735 with @xmath1736 note @xmath1684 does not depend on the sample size @xmath126 proof : from theorem [ uy ] , for the variance of @xmath1737 to be @xmath1738 @xmath1684 can not increase with @xmath126 further for the second order truncation bias @xmath1739 and the square root of the variance @xmath1740 of @xmath1741 both to be @xmath1742we must have @xmath1743 it then follows from eqs .
( @xmath1744 ) and ( [ variance3 ] ) that @xmath1745 in order for @xmath1746 = o_{p}\left ( n^{-\frac{4\beta } { 4\beta+d}}\right ) , $ ] we require for @xmath1747 @xmath1748 substituting for @xmath1749 in eq .
( [ bias2 ] ) using eq.([variance1 ] ) and recalling that @xmath1724 so @xmath1538 , we obtain @xmath1750 since @xmath1751 solving the last expression for @xmath1752,we obtain @xmath1753 which is equation eq.@xmath1754 except with a nonstrict inequality .
we have just deduced that the constraint @xmath1755 was due to restriction ( [ bias2 ] ) .
we have not yet considered whether the restriction @xmath1756 implies additional constraints .
we now show that it does not .
specifically if we set @xmath1757 for all @xmath1758 then eq.@xmath1756 is true whenever eq.@xmath1759 holds because of our assumption that @xmath1760 thus we can set @xmath1761 thus we have shown that if @xmath1762 then @xmath1763 @xmath1755 holds , and @xmath1684 must not increase with @xmath126 we next show that when the inequality is strict in @xmath1764and eq.@xmath1668 holds , we can find @xmath1765 for which @xmath1766 we then complete the proof of the theorem by showing that when @xmath1755 holds with an equality , there is no choice of @xmath1767 for which @xmath1768converges at a rate better than @xmath1769 suppose the inequality is strict in @xmath1770 since @xmath1771 eq.@xmath1772 applied recursively suggests we define @xmath1773 for @xmath1774 and take @xmath1775 however , this will not generally give @xmath1776 as required when @xmath1761 instead we use the modified algorithm given in the statement of the theorem which insures that @xmath1777 , as required .
since @xmath1684 is not a function of @xmath176 in order to show @xmath1778 converges at rate @xmath1779 we only need to check the bias
. now @xmath1780 @xmath1781 since @xmath1782so the bias of @xmath1783 is @xmath1784 as required .
suppose now the equality holds in eq.@xmath1755 so @xmath1785 and continue to assume eq.@xmath1668 holds .
we now construct an estimator @xmath1783 that converges at rate @xmath1786and show that no estimator in our class @xmath1783 converges at a faster rate .
we conjecture this rate is minimax when the equality in eq.@xmath1755 holds . again
@xmath1787 and by the previous arguments , @xmath1788 @xmath1789 we can suppose that @xmath1790it remains to determine @xmath1791 and @xmath1792 we know@xmath1793 must satisfy @xmath1794 the variance of @xmath1783 is of order @xmath1795 thus the order of the bias will still equal that of the variance provided we multiply the rhs of eq.@xmath1796 by @xmath1797 . then eq.@xmath1798 becomes @xmath1799 since , @xmath1800and @xmath1801 we substitute @xmath1802 for @xmath1803 in the modified eq.@xmath1798 which gives @xmath1804 hence @xmath1805 which implies that@xmath64 @xmath1806 \right ) \label{log}\ ] ] to minimize the variance , we want the slowest growing function of @xmath5 that satisfies eq.@xmath1807 which is @xmath1808 as claimed .
in this section we no longer assume that the estimation bias @xmath1809of a third order estimator is less than @xmath1535 then even when eq.@xmath1810 holds with a strict inequality , @xmath1778 does not achieve a @xmath1532 rate of convergence because the fourth order bias @xmath1811exceeds @xmath1535 however , we will now construct an estimator @xmath1812 that under our assumptions @xmath512 does converge at rate @xmath1813 whenever @xmath1814 given in assumption @xmath1815 satisfy eq.@xmath1755 with a strict inequality . because the estimator is very complicated ,
we have chosen to only define the estimator and give its properties in the text .
the motivating ideas for and the formal proofs of these properties are provided in the appendix . to define the estimator
, we need some additional notation .
define@xmath1816 where , @xmath1817 @xmath1818 with @xmath1617 then define @xmath1819 as @xmath1820 @xmath1821 is defined as @xmath1822 with @xmath1823 for @xmath1824 and @xmath1825 @xmath1826 next @xmath1827 is defined as @xmath1822 with @xmath1823 for @xmath1828 and @xmath1829 @xmath1830 @xmath1831 @xmath1832 @xmath1833 we will use this notation for @xmath892 even though @xmath1834 does not depend on @xmath1835 and is equal to @xmath1836 of the previous subsection .
finally define@xmath1837 [ beyond4th]given @xmath1814 satisfying eq.[ne2 ] with a strict inequality@xmath90 define @xmath1838 @xmath70to be the smallest integer such that @xmath1839 where @xmath1840 let @xmath1767 , @xmath1841 @xmath1778 be as in theorem [ uz ] and define @xmath1842 then@xmath1843{c}k_{-1}^{2\beta / d},\left ( \frac{\log n}{n}\right ) ^{-\frac{\beta_{g}}{d+2\beta_{g}}}k_{2s}^{-\beta_{b}/d}k_{2s+1}^{-\beta_{p}/d},\left ( \frac{\log n}{n}\right ) ^{-\frac{\beta_{g}}{d+2\beta_{g}}}k_{2s+1}^{-\beta_{b}/d}k_{2s}^{-\beta_{p}/d},\\ \left ( \frac{\log n}{n}\right ) ^{-\frac{2\beta_{g}}{d+2\beta_{g}}}k_{0}^{-2\beta / d},\left ( \frac{\log n}{n}\right ) ^{-\frac{\left ( m-1\right ) \beta_{g}}{d+2\beta_{g}}}n^{-\frac{\beta_{b}}{d+2\beta_{b}}-\frac{\beta_{p}}{d+2\beta_{p}}},\\ \left ( \frac{\log n}{n}\right ) ^{-\frac{2\beta_{g}}{d+2\beta_{g}}}\underset{1\leq s\leq j}{\max}\left ( k_{2s}^{-\beta_{b}/d}k_{0}^{-\beta _ { p}/d},k_{0}^{-\beta_{b}/d}k_{2s}^{-\beta_{p}/d}\right ) \end{array } \right ] \right ) \\ & = o_{p}\left ( \max\left [ \begin{array } [ c]{c}k_{-1}^{2\beta / d},\left ( \frac{\log n}{n}\right ) ^{-\frac{\beta_{g}}{d+2\beta_{g}}}k_{2s}^{-\beta_{b}/d}k_{2s+1}^{-\beta_{p}/d},\left ( \frac{\log n}{n}\right ) ^{-\frac{\beta_{g}}{d+2\beta_{g}}}k_{2s+1}^{-\beta_{b}/d}k_{2s}^{-\beta_{p}/d},\\ \left ( \frac{\log n}{n}\right ) ^{-\frac{2\beta_{g}}{d+2\beta_{g}}}k_{0}^{-2\beta / d},\left ( \frac{\log n}{n}\right ) ^{-\frac{\left ( m-1\right ) \beta_{g}}{d+2\beta_{g}}}n^{-\frac{\beta_{b}}{d+2\beta_{b}}-\frac{\beta_{p}}{d+2\beta_{p}}}\end{array } \right ] \right ) \\ & = o_{p}\left ( n^{-\frac{4\beta}{d+4\beta}}\right)\end{aligned}\ ] ] and @xmath1844 * inference : * elsewhere , we prove that @xmath1845 is asymptotically normal . here , to avoid the problem of unknown constants for confidence interval construction that we discussed in section [ dr_ci_section ] , we will construct nearly optimal rather than optimal confidence intervals .
we suppose that eq .
@xmath1755 holds with strict equality for the @xmath1846associated with the parameter space @xmath245 .
then there exists @xmath1847 such that for all @xmath1848 @xmath1849 satisfies eq @xmath1755 with strict equality@xmath90 @xmath1850 } { var_{\theta}\left [ \widehat{\psi } _ { \mathcal{k}_{j}}^{eff}\left ( \beta_{g},\beta_{b}-\sigma,\beta_{p}-\sigma\right ) |\widehat{\theta}\right ] } \right ] = o_{p}\left ( 1\right)\ ] ] and @xmath1851 \right\ } \asymp n^{-\frac{8\left ( \beta-\sigma\right ) } { d+4\left ( \beta-\sigma\right ) } } .\ ] ] let @xmath1852 $ ] be a uniformly consistent estimator of ( the properly standardized ) @xmath1853 .
then , for all @xmath1854 @xmath1855 \right ) ^{-1}\ ] ] converges uniformly in @xmath274 to a @xmath1241 .
moreover , @xmath1856\ ] ] is a conservative uniform asymptotic @xmath1243 confidence interval for @xmath175 with diameter of the order of @xmath1857 if eq.[ne2 ] holds with an equality and @xmath1767 , @xmath1841 @xmath1778 are as in the final paragraph of the preceding subsection then the proof of theorem [ beyond4th ] in the appendix implies @xmath1858
in this section we describe how to construct adaptive confidence intervals ( i ) for a regression function @xmath1859 $ ] when the marginal of @xmath8 is unknown and ( ii ) for the treatment effect function and optimal treatment regime in a randomized clinical trial .
* example 1a continued : * consider the case @xmath1860 , @xmath1861with @xmath1862 as usual , we assume for all @xmath383 @xmath130 and the density @xmath148 of @xmath8 are contained in known hlder balls @xmath1863 and @xmath1864 redefine @xmath1865 $ ] where @xmath1322 is an adaptive estimate of @xmath763 from the training sample and expectations and probabilities remain conditional on the training sample .
adaptivity of @xmath1866implies that if @xmath1867 is also contained in a smaller hlder ball @xmath1868 @xmath1869 , then @xmath1322 will converge to @xmath130 under @xmath1870 at rate @xmath1871 robins and van der vaart ( 2006 ) showed that , when the marginal density @xmath87 of @xmath8 is known , the key to constructing optimal ( rate ) adaptive confidence balls for @xmath1872 was to find a rate optimal estimator of @xmath1873 .$ ] we shall show that their approach fails when the marginal of @xmath8 is unknown , but that a modification described below succeeds . specifically , if @xmath1874 lies in a smaller hlder ball @xmath1875 @xmath1876 our modification results in honest asymptotic confidence balls under @xmath1877 @xmath1878 whose diameter is ( essentially ) of the same order @xmath1879 as the diameter of robins and van der vaart s optimal adaptive region or ball , provided either @xmath1880 @xmath1881 and @xmath1882 or @xmath1883 @xmath1884 and eq.@xmath1885 holds with @xmath1886 this order is the maximum of the minimax rate @xmath1887 of convergence of @xmath1888 to @xmath1889 were @xmath1890 known to lie in @xmath1891 and the square root of the minimax rate of convergence of an estimator of @xmath1892 \ $ ] in the larger model @xmath310 with @xmath763 and @xmath1342 only known to lie in @xmath1893and @xmath1864 the case where @xmath1884 and eq.@xmath1537 does not hold will be considered elsewhere . now , since @xmath1894 = e_{\theta}\left [ \widehat{b}\left ( x\right ) y\right ] , $ ]
@xmath1895 = e_{\theta}\left [ \left\ { b\left ( x\right ) \right\ } ^{2}\right ] -2e_{\theta}\left [ \widehat{b}\left ( x\right ) b\left ( x\right ) \right ] + e_{\theta}\left [ \left\ { \widehat{b}\left ( x\right ) \right\ } ^{2}\right]\ ] ] has first order influence function @xmath1896 \ $ ] where @xmath1897 -2\widehat{b}\left ( x\right ) y+\widehat{b}^{2}\left ( x\right ) , \ ] ] @xmath70so @xmath1898 thus @xmath1899 for @xmath1900 $ ] differs from @xmath1901 for @xmath1902 $ ] only in @xmath1903 since the truncation bias @xmath1904 , higher order influence functions of @xmath930 and estimation bias do not depend on @xmath1905 , it follows that @xmath1906,@xmath1907 and @xmath1908 are identical for @xmath1909 $ ] and @xmath1910 .$ ] in contrast , @xmath1911 is identically zero for @xmath1865 $ ] but not for @xmath1912 .$ ] thus , by theorem @xmath1913 for @xmath1865 , $ ] var@xmath1914 \asymp\frac{1}{n}\left ( \frac{k}{n}\right ) ^{m-1}$ ] if @xmath1915 and @xmath1916 and var@xmath1914 = 0 $ ] if @xmath1440 and @xmath1917 in the case when @xmath1918 and @xmath1916 by the hoeffding decomposition,@xmath1919 = var_{\theta}\left ( \sum_{s=1}^{m}\left ( \mathbb{d}_{s}^{\left ( \widehat{\mathbb{\psi}}_{m,\widetilde{\psi}_{k}}\right ) } \left ( \theta\right ) \right ) \right)\ ] ] where @xmath1920 is a @xmath212th order degenerate u - statistic .
further by theorem @xmath1913 we have@xmath1919 \asymp\max\left ( var_{\theta}\left ( \mathbb{d}_{1}^{\left ( \widehat{\mathbb{\psi}}_{m,\widetilde{\psi}_{k}}\right ) } \right ) , var_{\theta}\left ( \mathbb{d}_{2}^{\left ( \widehat{\mathbb{\psi}}_{m,\widetilde{\psi}_{k}}\right ) } \right ) \right)\ ] ] as @xmath1921 for any @xmath1922 moreover , @xmath1923 since the kernel of @xmath1924 is of order @xmath1925 in summary @xmath1926 & \asymp\max\left ( \frac{\left\vert \left\vert b\left ( x\right ) -\widehat{b}\left ( x\right ) \right\vert \right\vert _ { 2}^{2}}{n},\frac { k}{n^{2}}\right ) \\ & = \max\left ( n^{-\frac{2\beta_{b}/d}{1 + 2\beta_{b}/d}-1},\frac{k}{n^{2}}\right)\end{aligned}\ ] ] if @xmath1918 and @xmath1927 ( in contrast , for @xmath1928 , $ ] var@xmath1929 \asymp \max\left ( \frac{1}{n},\frac{k}{n^{2}}\right ) = \frac{1}{n}$ ] if @xmath1440 ) .
thus if @xmath1930 @xmath1880 @xmath1931 has @xmath1932 of @xmath1933 where @xmath1934 comes from equating the order @xmath1935 of @xmath1936 to the order @xmath1937 of the variance ( @xmath1938 for @xmath1939 @xmath1940 ) and @xmath1941 is the smallest integer @xmath135 such that the order @xmath1942 of @xmath1943is less than the order @xmath1944 of the standard error@xmath64 it follows that , for @xmath1930 in contrast to @xmath1910 , $ ] we can estimate @xmath1945 $ ] at ( the minimax ) rate @xmath1946 which is faster ( i.e. , less ) than the usual parametric rate of @xmath433 when @xmath1947 , the minimax rates for @xmath1948 $ ] and @xmath1949 $ ] are identical and , when eq.@xmath1537 holds , it follows from theorem [ beyond4th ] that @xmath1950 achieves the minimax rate of @xmath1951 . henceforth assume either @xmath1952 or @xmath1953 and eq.@xmath1537 holds .
pick an @xmath1954 so that eq.@xmath1537 holds for @xmath1955 let @xmath1956 and define @xmath1957 and @xmath1958if @xmath1881 and @xmath1959 and @xmath1960 \ $ ] if @xmath1961 note @xmath1962 is @xmath1963 uniformly over @xmath178 where @xmath245 is the parameter space with smoothness parameters @xmath1964 then , by eq.@xmath1537 and results in section [ case2 ] , as @xmath1965 @xmath1966pr@xmath1967 \geq1-\alpha.$ ] thus , if @xmath175 were a function of @xmath179 only through @xmath597so @xmath1968 , the set @xmath1969 would be an uniform asymptotic @xmath1243 confidence region for @xmath1970 however , for @xmath1971 , $ ] this approach fails because @xmath270 also depends on @xmath179 through the unknown density @xmath87 of @xmath113 this approach succeeded in robins and van der vaart ( 2006 ) because @xmath87 was assumed known .
we consider two solutions .
the first gives ( near ) optimal adaptive honest intervals .
the second would give honest , but non - optimal , intervals .
the first solution is to replace @xmath175 with its empirical mean @xmath1972 \ $ ] in eq.([cia])@xmath64 @xmath1973 n^{-1/2}\right ) = o_{p}\left ( n^{-\left ( \frac{2\beta_{b}}{d+2\beta_{b}}+\frac{1}{2}\right ) } \right)\ ] ] uniformly in @xmath357 it is straightforward to check that for all @xmath1974 @xmath1975 thus , for @xmath1854 @xmath1976 uniformly over @xmath1977 so @xmath1966pr@xmath1978 \geq1-\alpha$ ] and @xmath1979 \leq\widehat{\psi}_{\ } ^{\ast}+z_{\alpha}\widehat{\mathbb{w}}^{\ast}\right\ } \label{citrue}\ ] ] is a uniform asymptotic @xmath1243 confidence region for @xmath1970 moreover , if @xmath1867 lies in a smaller hlder ball @xmath1868 @xmath1876 then , under @xmath1980 the diameter @xmath1981 since @xmath1982 and @xmath1983 and @xmath1962 are @xmath1984 the second , non - optimal , solution would be to replace the functional @xmath1865 $ ] with @xmath1985 the functional @xmath1986 is the first functional we have considered that is not in our doubly robust class of functionals . arguing as above ,
if we can construct an asymptotically normal higher order @xmath187 statistic estimator @xmath1987 that converges to @xmath1986 at rate @xmath1988 on @xmath310 and a consistent estimator @xmath1989of its standard error , then @xmath1990 would be an honest adaptive confidence interval of diameter @xmath1991 we conjecture , based on arguments given elsewhere , that the minimax rate for estimation of @xmath1992 exceeds @xmath1993 $ ] whenever @xmath1994 . since @xmath1995for all
@xmath1996 it follows that , when the marginal of @xmath8 is unknown and @xmath1997 , intervals based on @xmath1998 $ ] will , but intervals based on @xmath1999 will not , have diameter of the same order as the optimal interval with the marginal of @xmath8 known .
* example 4 continued : * consider the case @xmath1860 , @xmath755 wp1 so we have data @xmath2000 , where @xmath12 is a binary treatment , @xmath0 is the response , and @xmath8 is a vector of prerandomization covariates .
the randomization probabilities @xmath741 are known by design and @xmath2001 is the average treatment effects function . for @xmath1977 @xmath130 and the density @xmath1366 of @xmath8
are contained in known hlder balls @xmath2002 and @xmath1864 suppose we have an adaptive estimator @xmath1322 of @xmath2003based on the training sample constructed as described below .
now , since @xmath1894 = e_{\theta}\left [ \widehat{b}\left ( x\right ) y|a=1\right ] -e_{\theta}\left [ \widehat{b}\left ( x\right ) y|a=0\right ] $ ] has influence function @xmath2004 = \left ( a-\pi_{0}\left ( x\right ) \right ) \sigma_{0}^{-2}\left ( x\right ) y\widehat{b}\left ( x\right ) -e_{\theta}\left [ \widehat{b}\left ( x\right ) b\left ( x\right ) \right ] , $ ] where @xmath2005 @xmath1865 $ ] has first order influence functions , indexed by arbitrary functions @xmath2006 @xmath2007 \ $ ] with @xmath2008 thus @xmath1899 for @xmath2009 $ ] differs from @xmath1901 for @xmath2010 $ ] only in @xmath1903 it follows that all the properties of the confidence ball [ citrue ] for @xmath2011 in the setting of the last subsection remain true for @xmath2012 in the setting of this subsection .
now define @xmath2013 .$ ] then it then follows that an honest @xmath159 uniform asymptotic confidence set for the optimal treatment regime @xmath2014 $ ] is given by @xmath2015 \leq\widehat{\psi}_{\ } ^{\ast}+z_{\alpha}\widehat{\mathbb{w}}^{\ast}\right\ } .$ ] * adaptive estimator of the treatment effect function : * one among many approaches to constructing a rate - adaptive estimator of @xmath763 is as follows .
split the training sample into two random subsamples - a candidate estimator subsample of size @xmath1316 and a validation subsample of size @xmath1317 where both @xmath1318 and @xmath1319 are bounded away from @xmath590 as @xmath171 noting that @xmath2016 $ ] for all @xmath2017 we construct candidate estimators of @xmath130 as follows . for @xmath2018 @xmath2019 let @xmath2020 be the solution , if any , to the @xmath212 equations @xmath2021\ ] ] where @xmath857 is a complete basis@xmath70wrt to lebesgue measure in @xmath1059 that provides optimal rate approximation for hlder balls and @xmath2022 is the empirical measure for the candidate estimator subsample .
our candidates for @xmath1890 are the @xmath2023 robins ( 2004 ) proved that @xmath130 is the unique function @xmath2024 * * * * minimizing @xmath2025 b^{\ast}\left ( x\right ) \right\ } ^{2}\right ] .$ ] in fact , the candidate @xmath2026 in our set for which @xmath2027is smallest is also the candidate that minimizes @xmath2028 $ ] since @xmath2029 .
$ ] specifically , @xmath2030{c}\sigma_{0}^{-2}\left ( x\right ) \left\ { y-\left [ a-\pi_{0}\left ( x\right ) \right ] \widehat{b}^{\left ( s\right ) } \left ( x\right ) \right\ } ^{2}\\ -\sigma_{0}^{-2}\left ( x\right ) \left\ { y-\left [ a-\pi_{0}\left ( x\right ) \right ] b\left ( x\right ) \right\ } ^{2}\end{array } \right ] \\ & = e\left [ \begin{array } [ c]{c}\sigma_{0}^{-2}\left ( x\right ) \left ( a-\pi_{0}\left ( x\right ) \right ) \left ( b\left ( x\right ) -\widehat{b}^{\left ( s\right ) } \left ( x\right ) \right ) \times\\ \left ( 2\left ( ab\left ( x\right ) -e\left ( y|a=0,x\right ) \right ) -\left ( a-\pi_{0}\left ( x\right ) \right ) \left ( b\left ( x\right ) + \widehat{b}^{\left ( s\right ) } \left ( x\right ) \right ) \right ) \end{array } \right ] \\ & = e\left ( \sigma_{0}^{-2}\left ( x\right ) \left ( a-\pi_{0}\left ( x\right ) \right ) a\left ( b\left ( x\right ) -\widehat{b}^{\left ( s\right ) } \left ( x\right ) \right ) ^{2}\right ) \\ & = e\left [ \left ( b\left ( x\right ) -\widehat{b}^{\left ( s\right ) } \left ( x\right ) \right ) ^{2}\right]\end{aligned}\ ] ] we use these results to select among our candidates by cross - validation .
let @xmath1322 be the @xmath2031 minimizing @xmath2032 \widehat{b}^{\left ( s\right ) } \left ( x\right ) \right\ } ^{2}\right ] $ ] over @xmath2033 , where @xmath2034 is the validation subsample empirical measure .
if @xmath130 were known to lie in a hlder ball @xmath2035 it is easy to check that the candidate @xmath2036 with @xmath2037 obtains the optimal rate of @xmath2038 for estimating @xmath2039 .$ ] since the number of candidates at sample size @xmath5 is less than @xmath176 it then follows at once from van der laan and dudoit s ( 2003 ) results on model selection by cross validation that @xmath1322 is adaptive over hlder balls .
in example @xmath2040 of section 3.1@xmath90 we considered the following problem .
we were given a functional @xmath658 indexed by a real number @xmath2041 and the parameter @xmath357 the implicitly defined - functional @xmath643 was the assumed unique solution to @xmath2042 we noted that a @xmath664confidence set for @xmath643 is the set of @xmath665such that a @xmath664ci interval for @xmath658 contains @xmath2043 in the following subsection we derive the width of the confidence set for @xmath2044 we then generalize the problem in the second subsection by introducing the notions of the testing tangent space , a testing influence function , and the higher order efficient testing score . in the final subsection , we show how the two earlier subsections are related . to derive the order of the length of the confidence interval for the parameter @xmath643 in example 1c , we can use the next theorem as follows .
assume eq ( [ ne11 ] ) holds and @xmath2045 .
then we can take the estimator @xmath2046 and rate @xmath2047 in the theorem to be the estimator @xmath2048 and rate @xmath2049 for a very small positive @xmath2050 and conclude that the length of the confidence interval for @xmath643 in example 1c to be @xmath2051 [ j1 ] * : * suppose for an estimator @xmath2052 and functional @xmath2053 there is a scale estimator @xmath2054 such that @xmath2055 in @xmath264probability @xmath2056 and @xmath2057 converges in law to @xmath2058uniformly for @xmath274 , @xmath2059 .
then , ( i ) with @xmath1221 the @xmath1222quantile and @xmath2060 the cdf of a @xmath1223 the confidence set @xmath2061 is a uniform asymptotic @xmath159 confidence set for the ( assumed ) unique solution @xmath643 to @xmath2062 @xmath1883 the probability under @xmath179 that a sequence @xmath2063satisfying @xmath2064 is contained in @xmath2065 converges to @xmath18
when @xmath2066 is @xmath2067 when @xmath2068 and converges to @xmath2069when @xmath2070 @xmath2071 if @xmath666 is uniformly twice continuously differentiable in @xmath2041 and _ _ _ _ @xmath2072 _ _ _ _ and _ _ _ _ @xmath2073 _ _ _ _ for constants _ _ _ _
@xmath2074 then @xmath2075 holds for a sequence @xmath2076 satisfying @xmath2077 ( i ) : that @xmath2065 is a uniform asymptotic @xmath159 confidence set is immediate .
( ii ) : now @xmath2078 ( iii ) : since @xmath2079 for some @xmath2080 between @xmath643 and @xmath2081 we have that @xmath2082 satisfies the assumption in ( ii ) .
* remark : * under some further regularity conditions , the solution @xmath2083 to @xmath2084 is asymptotically normal with mean @xmath643 and variance @xmath2085 $ ] uniformly over @xmath274,@xmath2086 in the following , we repeatedly use definitions from sec .
2 , which might usefully be reviewed at this point . *
* m**@xmath2087 * order testing nuisance tangent space , testing tangent space , testing influence functions , efficient score , efficient information , and efficient testing variance : * given a model @xmath310 with parameter space @xmath245 and a functional @xmath2088 define @xmath2089 to be the submodel with parameter space @xmath2090 . thus @xmath2091 is the submodel with @xmath643 equal to @xmath2092 define , for @xmath2093 the @xmath135th order ( i ) testing nuisance tangent space @xmath2094 to be the @xmath1585 order tangent space for the submodel @xmath2095 ( ii ) testing tangent space @xmath2096 to be the closed linear span of @xmath2097 ( iiia ) set @xmath2098 of testing influence functions to be the orthocomplement of @xmath2094 in @xmath209 @xmath90 ( iiib ) set @xmath2099 of standardized testing influence functions to be @xmath2100 = var_{\theta}\left [ \mathbb{if}_{1,\tau\left ( \cdot\right ) } ^{eff}\left ( \theta\right ) \right ] \right\ } , \ ] ] @xmath2101 efficient testing score @xmath2102 to be @xmath2103 \equiv \pi_{\theta}\left [ \mathbb{es}_{1}^{test}\left ( \theta\right ) |\gamma _ { m}^{nuis , test,\perp}\left ( \theta,\tau^{\dagger}\right ) \right]\ ] ] where @xmath2104 @xmath2105 efficient testing information to be @xmath2106 and @xmath2107 the efficient testing variance to be @xmath2108 ^{-1}.$ ] further define , for @xmath383 the @xmath135th order ( i ) estimation nuisance tangent space @xmath2109 to be @xmath2110 @xmath2111 = 0\right\ } , $ ] and ( ii ) efficient estimation variance to be @xmath2112 $ ] . *
remark : * for @xmath2113 the testing and estimation nuisance tangent spaces @xmath2114 and @xmath2115 are identical .
however for @xmath1916 @xmath2114 is generally a strict subset of @xmath2116 for example , if the model can be parametrized as @xmath2117 and @xmath245 is the product of the parameter spaces for @xmath2041 and @xmath2118 the @xmath2114 is the space of mth order scores for @xmath2119 however , @xmath2120 also includes the mixed scores that have @xmath212 derivatives in the direction @xmath2041 and @xmath2121 derivatives in @xmath2122 directions .
it is this strict inclusion that gives rise to higher order phenomena that do not occur in the first order theory .
[ gg ] * : * suppose * * * * @xmath2123 exists in @xmath229 then for @xmath2093 ( i ) the set of estimation nuisance scores @xmath2109 includes the set of testing nuisance scores @xmath2094 with equality of the sets when @xmath2124 , ( ii ) @xmath2125 is standardized if and only if @xmath2126 = 1\ $ ] if and only if @xmath2127 = 1,$ ] ( iii)@xmath2128 ^{-1}\mathbb{if}_{m,\tau\left ( \cdot\right ) } ^{test};\text { } \mathbb{if}_{m,\tau\left ( \cdot\right ) } ^{test}\in\left\ { \mathbb{if}_{m,\tau\left ( \cdot\right ) } ^{test}\right\ } \right\ } , \ ] ] ( iv ) the set @xmath2129 of all mth order estimation influence functions is contained in @xmath2130 with equality of the sets when @xmath2113 ( v)@xmath2131 = \left\ { var\left [ \mathbb{es}_{m}^{test}\left ( \theta\right ) \right ] \right\ } ^{-1}\mathbb{es}_{m}^{test}\left ( \theta\right ) , \ ] ] ( vi ) @xmath2132 \right\ } ^{-1}\mathbb{es}_{m}^{test}\left ( \theta\right ) \in\left\ { \mathbb{if}_{m,\tau\left ( \cdot\right ) } ^{std , test}\right\ } $ ] and has the minimum variance @xmath2133 \right\ } ^{-1}$ ] among members of @xmath2134 . in particular @xmath2135 \right\ } ^{-1}\leq var_{\theta}\left [ \mathbb{if}_{m,\tau\left ( \cdot\right ) } ^{eff}\left ( \theta\right ) \right ] $ ] with equality when @xmath2113 ( vii ) given @xmath2136,any smooth submodel @xmath243 with range containing @xmath179 and contained in @xmath2137and @xmath281 we have @xmath2138 /\partial\zeta_{l_{1}} ...
\partial\zeta_{l_{_{s}}}|_{\zeta=\widetilde{\theta } ^{-1}\left\ { \theta\right\ } } = 0.\ ] ] thus , if @xmath2139 $ ] is frchet differentiable w.r.t .
@xmath285 to order @xmath286 for a norm
@xmath2140 @xmath2141 = o\left ( \left\vert \left\vert \delta\theta\ \right\vert \right\vert ^{m+1}\right ) $ ] for @xmath179 and @xmath2142 in an open neighborhood contained in @xmath2143 , since the taylor expansion of @xmath2144 \ $ ] around @xmath179 through order @xmath135 is identically zero@xmath64 the proof of the theorem will use the following two lemmas : [ a ] : for any @xmath2145@xmath2146 = e_{\theta}\left [ \mathbb{if}_{m,\tau\left ( \cdot\right ) } ^{test}\left ( \theta\right ) \mathbb{es}_{m}^{test}\left ( \theta\right ) \right]\ ] ] @xmath2147 \\ & = e_{\theta}\left [ \mathbb{if}_{m,\tau\left ( \cdot\right ) } ^{test}\pi_{\theta}\left [ \mathbb{es}_{1}^{test}\left ( \theta\right ) |\gamma _ { m}^{nuis , test,\perp}\left ( \theta,\tau^{\dagger}\right ) \right ] \right ] = e_{\theta}\left [ \mathbb{if}_{m,\tau\left ( \cdot\right ) } ^{test}\mathbb{es}_{1}^{test}\left ( \theta\right ) \right ] , \end{aligned}\ ] ] where the last equality holds by @xmath2148 [ b ] for any @xmath2149 @xmath2150 \\ & = e\left [ \mathbb{if}_{m,\tau\left ( \cdot\right ) } ^{test}\left ( \theta\right ) \mathbb{es}_{m}^{test}\left ( \theta\right ) \right ] \left\ { var\left [ \mathbb{es}_{m}^{test}\left ( \theta\right ) \right ] \right\ } ^{-1}\mathbb{es}_{m}^{test}\left ( \theta\right ) \\ & = e\left [ \mathbb{if}_{m,\tau\left ( \cdot\right ) } ^{test}\left ( \theta\right ) \mathbb{es}_{1}^{test}\left ( \theta\right ) \right ] \left\ { var\left [ \mathbb{es}_{m}^{test}\left ( \theta\right ) \right ] \right\ } ^{-1}\mathbb{es}_{m}^{test}\left ( \theta\right)\end{aligned}\ ] ] @xmath2151 thus , by @xmath2152 @xmath2153 & = \pi_{\theta}\left [ \mathbb{if}_{m,\tau\left ( \cdot\right ) } ^{test}\left ( \theta\right ) |\left\ { c\mathbb{es}_{m}^{test}\left ( \theta\right ) ; c\in r^{1}\right\ } \right ] \\ & = e\left [ \mathbb{if}_{m,\tau\left ( \cdot\right ) } ^{test}\left ( \theta\right ) \mathbb{es}_{m}^{test}\left ( \theta\right ) \right ] \left\ { var\left [ \mathbb{es}_{m}^{test}\left ( \theta\right ) \right ] \right\ } ^{-1}\mathbb{es}_{m}^{test}\left ( \theta\right ) .\end{aligned}\ ] ] now apply lemma [ a ] .
( theorem [ gg ] ) ( i ) is immediate from the definitions .
( ii ) and ( iiii ) follow from @xmath2154 & = 1\leftrightarrow e\left [ \mathbb{if}_{m,\tau\left ( \cdot\right ) } ^{test}\left ( \theta\right ) \mathbb{es}_{1}^{test}\left ( \theta\right ) \right ] = 1\\ & \leftrightarrow e_{\theta}\left [ \mathbb{if}_{m,\tau\left ( \cdot\right ) } ^{test}\mathbb{if}_{1,\tau\left ( \cdot\right ) } ^{eff}\left ( \theta\right ) \right ] = var_{\theta}\left [ \mathbb{if}_{1,\tau\left ( \cdot\right ) } ^{eff}\left ( \theta\right ) \right ] , \end{aligned}\ ] ] where we have used lemma [ a ] . for
( iv ) , note @xmath2155 follows from the fact that every smooth submodel through @xmath179 in model @xmath2091 is a smooth submodel through @xmath179 in model @xmath310 .
thus it remains to prove that @xmath2156 is standardized .
but , by part 4 of theorem [ eift ] , @xmath2157 = var_{\theta}\left [ \mathbb{if}_{1,\tau\left ( \cdot\right ) } ^{eff}\left ( \theta\right ) \right ] $ ] .
( v ) follows at once from lemma [ a ] and part ( ii ) . for ( vi ) , note that @xmath2158 \right\ } ^{-1}\mathbb{es}_{m}^{test}\left ( \theta\right ) \in\left\ { \mathbb{if}_{m,\tau\left ( \cdot\right ) } ^{std , test}\right\ } $ ] by definition .
thus @xmath2159 ^{-1}\mathbb{if}_{m,\tau\left ( \cdot\right ) } ^{test}\right\ } \geq\left\ { var_{\theta}\left [ \mathbb{es}_{m}^{test}\left ( \theta\right ) \right ] \right\ } ^{-1}\ ] ] follows from ( v ) .
the result then follows from part ( iii ) .
part ( vii ) is proved analogously to theorem [ eiet ] except now all scores lie in @xmath2109 by range @xmath2160 in @xmath2161 in the case of ( locally ) nonparametric models , we can explicitly characterize @xmath2162 .
let @xmath2163 be the set of all @xmath2164 $ ] with the @xmath2165 indexed by constants @xmath2166 and functions @xmath2167 satisfying @xmath2168 = 0.$ ] we remark that the subset of @xmath2169 comprised of all @xmath1564th order degenerate u - statistics can be written @xmath2170 \right\ } $ ] .
thus @xmath2163 simply restricts one of the functions @xmath2171 to be @xmath2172 [ ff ] if the model @xmath310 is ( locally ) nonparametric , then @xmath2173 . * * * * since the model is locally nonparametric @xmath2174 includes the set of all mean zero first order @xmath187statistics @xmath2175and thus any element of @xmath2162 must be a sum of degenerate @xmath340 of orders @xmath2176 through @xmath190 we continue by induction .
first we prove the theorem for @xmath332 @xmath70now , @xmath2177 where @xmath2178 is the closed linear span of the 2nd order degenerate part @xmath2179 of 2nd order scores @xmath2180 in model @xmath2181 where @xmath2179 is a sum of products @xmath2182 of first order scores in model @xmath2183 for two different subjects . by model @xmath292 being ( locally )
nonparametric , the set of first order scores in model @xmath2091 is precisely the set of random variables @xmath2184 orthogonal to @xmath2185 but the set of degenerate @xmath340 of order @xmath2176 orthogonal to the product of two scores in @xmath2186 is clearly @xmath2187 suppose now the theorem is true for @xmath2188 we show it is true for @xmath2189 by @xmath310 ( locally ) nonparametric and the induction assumption , @xmath2190 where @xmath2191 is the closed linear span of the sum of products of first order scores in model @xmath2192 for @xmath286 different subjects .
but @xmath2193 is the set of set of degenerate @xmath340 of order @xmath286 orthogonal to @xmath2194 in the following theorem we show that estimation influence functions @xmath2195 for the parameter @xmath2196 evaluated at the solution @xmath643 to @xmath2197 is contained in the set @xmath2198 of testing influence functions for @xmath2044 we also derive the estimation influence functions @xmath2199 for @xmath643 in terms of the estimation influence functions @xmath2200 for @xmath2196 and their derivatives with respect to @xmath2201 [ tt ] let @xmath643 be the assumed unique functional defined by @xmath2202 then , for @xmath2203 , whenever @xmath2204 and @xmath2205 exist , ( i ) @xmath2206 ( ii ) @xmath2207 where @xmath2208 ( iii ) @xmath2209where @xmath2210 for @xmath2211 @xmath2212{c}\left ( \begin{array } [ c]{c}\frac{\partial if_{1,\psi\left ( \tau^{\dagger},\cdot\right ) , i_{1}}\left ( \theta\right ) } { \partial\tau}\\ -e_{\theta}\left [ \frac{\partial if_{1,\psi\left ( \tau^{\dagger},\cdot\right ) , i_{1}}\left ( \theta\right ) } { \partial\tau}\right ] \end{array } \right ) if_{1,\tau\left ( \cdot\right ) , i_{2}}\ \left ( \theta\right ) \\ + \left ( \begin{array } [ c]{c}\frac{\partial if_{1,\psi\left ( \tau^{\dagger},\cdot\right ) , i_{2}}\left ( \theta\right ) } { \partial\tau}\\ -e_{\theta}\left [ \frac{\partial if_{1,\psi\left ( \tau^{\dagger},\cdot\right ) , i_{2}}\left ( \theta\right ) } { \partial\tau}\right ] \end{array } \right ) if_{1,\tau , i_{1}}\left ( \theta\right ) \end{array } \right ] \label{q22}\ ] ] where @xmath2213 .
@xmath2214 is given in the appendix as well as the general formula .
\(i ) for @xmath2215 consider any suitably smooth @xmath2216dimensional parametric submodel @xmath243 with range containing @xmath179 and contained in @xmath2161 let @xmath2217 be any associated @xmath2218 score @xmath2219 . by definition of @xmath2220 ,
@xmath2221 hence , @xmath2222 now we expand the rhs using the chain rule and note that the only non - zero term is the term @xmath2223 in which all @xmath2224derivatives are taken with respect to the second @xmath2225 in @xmath2226 all other terms include derivatives of @xmath2227which are zero by range @xmath2228 further @xmath2229 $ ] by the definition of the estimation influence function @xmath2230 .
we conclude that @xmath2204 is in @xmath2231 ( ii ) @xmath2232 @xmath2233 is straightforward .
that @xmath2232 is contained in @xmath2234 follows by part ( iv ) of theorem [ gg ] .
( iii ) see appendix for proof .
we now provide an example to show that , contrary to what one might expect based on part ( vi ) of theorem [ gg ] , inference concerning @xmath2220 may be more efficient when based on an inefficient member of the set @xmath2235 such as @xmath2236 than when based on the efficient score @xmath2237 without loss of generality , it is sufficient to consider the case @xmath2238 . in the following example
it is @xmath2239and @xmath2240 that play the role of @xmath643 and @xmath2241 in the preceding theorem , because @xmath2242 and @xmath2243 have , but @xmath643 and @xmath2244 do not have , higher order estimation and testing influence functions .
* example 1c ( continued ) : * in this example , with @xmath2245 @xmath675and _ _ _ _ @xmath642 _ _ _ _ binary , @xmath659\ ] ] and @xmath643 satisfies @xmath2246 .
let @xmath2247 satisfy @xmath2248 where @xmath2249 -e_{\theta}\left\ { \left [ \pi_{\theta}\left [ b\left ( \tau\right ) |\overline{z}_{k}\right ] \pi_{\theta}\left [ p|\overline{z}_{k}\right ] \right ] \right\ } $ ] is defined in section 3.1 with @xmath2041 a real - valued index and @xmath2250 . note @xmath2251
-e_{\theta}\left [ \left\ { \pi_{\theta } \left [ p|\overline{z}_{k}\right ] \right\ } ^{2}\right ] \right\ } , $ ] @xmath2252 , $ ] @xmath2253 below we freely use results of theorems [ tbrate ] , [ drhoif ] , and [ ebrate ] .
we suppose that @xmath2254 and @xmath2255 < c$ ] for some @xmath2074 @xmath2256 choose @xmath2257 so the truncation bias of @xmath2258 is @xmath2259and @xmath2260 \asymp k / n^{2}=n^{-2\left ( \frac{4\beta}{4\beta+d}+\sigma\right ) } .$ ] we assume the given @xmath1814 are such that the order @xmath2261 $ ] of the estimation bias of @xmath2262 is @xmath2263 then @xmath2264 and @xmath2265 are @xmath2266 which just exceeds the minimax rate @xmath2267 for @xmath2050 very small .
our goal is to compare the coverage and length of confidence intervals for @xmath2242 and @xmath2268 based on@xmath70 _ _ _ _ @xmath2269 where @xmath2270 are appropriate variance estimators , @xmath391 is our usual split sample initial estimator , and @xmath2271 is an initial split sample estimator depending on @xmath2272 that satisfies @xmath2273 if @xmath2274 , i.e. , @xmath2275 = \tau^{\dagger},$ ] we assume that if @xmath2276 then the convergence rate under @xmath179 of our estimator of @xmath2277 for any @xmath2278 remains @xmath2279 . now
the assumption @xmath2280 , $ ] @xmath2281 < c$ ] implies @xmath2282is uniformly bounded away from zero and infinity .
it then follows from earlier results on @xmath2283 the assumption @xmath2284 , e_{\theta}\left [ a^{2}\right ] < c,$ ] and theorem [ tt ] that @xmath2285 is a uniform asymptotic @xmath159 confidence interval for both @xmath643 and @xmath2247 of length @xmath2286 the next theorem gives explicit formulae for @xmath2287 @xmath2288 and @xmath2289 using these formulae we calculate the biases and variances necessary to compare the coverage of the three intervals .
this comparison requires each of our three candidate procedures to be on the same scale .
therefore we used standardized versions of the relevant statistics .
[ aa ] suppose the assumptions described in the preceding example hold . then ( i)@xmath2290 \\ & + \mathbb{v}\left [ \left\ { \left [ y^{\ast}\left ( \tau\right ) -\widehat{b}\left ( x,\tau\right ) \right ] \overline{z}_{k}^{t}\right\ } _ { i_{1}}\left\ { \overline{z}_{k}\left [ a-\widehat{p}\left ( x\right ) \right ] \right\ } _ { i_{2}}\right]\end{aligned}\ ] ] where @xmath2291 @xmath2292 ( ii ) : let @xmath2293 denote @xmath2294 and @xmath2295 denote @xmath2296 thus,@xmath2297 -var\left [ \mathbb{u}_{2,2,\widetilde{\tau}_{k}\left ( \cdot\right ) } ^{\ast , test,\perp } \left ( \widehat{\theta}\left ( \tau^{\dagger}\right ) , \tau^{\dagger } \right ) \right ] \right\ } ^{-1}\\ & \times\left\ { \mathbb{if}_{2,\widetilde{\psi}_{k}\left ( \tau^{\dagger } , \cdot\right ) } \left ( \widehat{\theta}\left ( \tau^{\dagger}\right ) \right ) -\mathbb{u}_{2,2,\widetilde{\tau}_{k}\left ( \cdot\right ) } ^{\ast , test,\perp}\left ( \widehat{\theta}\left ( \tau^{\dagger}\right ) , \tau^{\dagger}\right ) \right\}\end{aligned}\ ] ] where@xmath2298 \right ) ^{-1}\times\widehat{\epsilon}_{i}\widehat{\delta}_{i}\\ & \left\ { \begin{array } [ c]{c}-\left\ { \begin{array } [ c]{c}\left ( e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}^{2}\widehat{\delta } _ { i}^{2}\right ] \right ) ^{-1}e_{\widehat{\theta}}\left [ \widehat{\epsilon } \widehat{\delta}^{2}\overline{z}_{k}^{t}\right ] \\
\times e_{\widehat{\theta}}\left [ \widehat{\epsilon}^{2}\widehat{\delta } \overline{z}_{k}^{t}\right ] \widehat{\epsilon}_{j}\widehat{\delta}_{j}\end{array } \right\ } \\ + e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}^{2}\widehat{\delta}_{i}\overline{z}_{k , i}^{t}\right ] \overline{z}_{k , j}\widehat{\delta}_{j}\\ + e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}\widehat{\delta}_{i}^{2}\overline{z}_{k , i}^{t}\right ] \overline{z}_{k , j}\widehat{\epsilon}_{j}\end{array } \right\}\end{aligned}\ ] ] and @xmath2299\ ] ] also , @xmath2300 ( iii ) @xmath2301 where @xmath2302 \right\}\ ] ] with @xmath2303 @xmath103@xmath103@xmath2304 $ ] @xmath103where @xmath2305{c}\left [ \left\ { a-\widehat{p}\left ( x\right ) \right\ } _ { i_{1}}^{2}-v\left ( \widehat{\theta}\right ) \right ] \left [ \left\ { y-\widehat{b}\left ( x\right ) \right\ } \left\ { a-\widehat{p}\left ( x\right ) \right\ } \right ] _ { i_{2}}+\\ \left [ \left\ { a-\widehat{p}\left ( x\right ) \right\ } _ { i_{2}}^{2}-v\left ( \widehat{\theta}\right ) \right ] \left [ \left\ { y-\widehat{b}\left ( x\right ) \right\ } \left\ { a-\widehat{p}\left ( x\right ) \right\ } \right ] _ { i_{1}}\end{array } \right]\end{aligned}\ ] ] the proof of @xmath1880 was given earlier . the proofs of @xmath2075 and ( iii )
are in the appendix .
[ bb]@xmath64 * * * * suppose @xmath2306 and the assumptions of the preceding theorem hold .
@xmath2307@xmath2308then \(i ) @xmath2309 = o\left ( \frac{1}{n}\right ) , $ ] @xmath2310 \times\left [ var_{\theta}\left\ { var_{\widehat{\theta}\left ( \tau^{\dagger}\right ) } \left\ { \mathbb{es}_{2,\widetilde{\tau}_{k}\left ( \cdot\right ) } ^{test}\left ( \widehat{\theta}\left ( \tau^{\dagger}\right ) \right ) \right\ } ^{-1}\mathbb{es}_{2,\widetilde{\tau}_{k}\left ( \cdot\right ) } ^{test}\left ( \widehat{\theta}\left ( \tau^{\dagger}\right ) \right ) \right\ } \right ] ^{-1}\\ & = 1+o_{p}\left ( 1\right)\end{aligned}\ ] ] \(ii ) @xmath2311 & = o\left ( \frac{1}{n}\right ) , \\ var_{\theta}\left [ v\left ( \widehat{\theta}\right ) ^{-1}\ \psi _ { 2,k\ } \left ( \tau,\widehat{\theta}\right ) \right ] /var_{\theta}\left\
{ \tau_{2,k\ } \left ( \widehat{\theta}\right ) -\tau^{\dagger}\right\ } & = 1+o_{p}\left ( 1\right)\end{aligned}\ ] ] @xmath2071@xmath2312 & = o_{p}\left\ { \left ( p-\widehat{p}\right ) \left ( b\left ( \tau^{\dagger}\right ) -\widehat{b}\left ( \tau^{\dagger}\right ) \right ) \left ( \frac{g\left ( x\right ) } { \widehat{g}\left ( x\right ) } -1\right ) \right\ } \\ & = o_{p}\left ( n^{-\left ( \frac{\beta_{g}}{2\beta_{g}+d}+\frac{\beta_{b}}{d+2\beta_{b}}+\frac{\beta_{p}}{d+2\beta_{p}}\right ) } \right)\end{aligned}\ ] ] \(iv ) @xmath2313 \\ & = o_{p}\left\ { \left ( p-\widehat{p}\right ) \left ( b\left ( \tau ^{\dagger}\right ) -\widehat{b}\left ( \tau^{\dagger}\right ) \right ) \left [ \left ( \frac{g\left ( x\right ) } { \widehat{g}\left ( x\right ) } -1\right ) + \left ( p-\widehat{p}\right ) + \left ( b\left ( \tau^{\dagger}\right ) -\widehat{b}\left ( \tau^{\dagger}\right ) \right ) \right ] \right\ } \\ & = o_{p}\left [ \max\left\ { n^{-\left ( \frac{\beta_{g}}{2\beta_{g}+d}+\frac{\beta_{b}}{d+2\beta_{b}}+\frac{\beta_{p}}{d+2\beta_{p}}\right ) } , n^{-\left ( \frac{\beta_{b}}{d+2\beta_{b}}+\frac{2\beta_{p}}{d+2\beta_{p}}\right ) } , n^{-\left ( \frac{2\beta_{b}}{d+2\beta_{b}}+\frac{\beta_{p}}{d+2\beta_{p}}\right ) } \right\ } \right]\end{aligned}\ ] ] ( v ) @xmath2314 & = o_{p}\left\ { \begin{array } [ c]{c}\left ( p-\widehat{p}\right ) \left ( \frac{g\left ( x\right ) } { \widehat{g}\left ( x\right ) } -1\right ) \left ( b-\widehat{b}\right ) + \\ \left ( p-\widehat{p}\right ) ^{2}\left ( p-\widehat{p}\right ) \left ( b-\widehat{b}\right ) \\ + \left ( \frac{g\left ( x\right ) } { \widehat{g}\left ( x\right ) } -1\right ) ^{2}\left [ \left ( p-\widehat{p}\right ) + \left ( \frac{g\left ( x\right ) } { \widehat{g}\left ( x\right ) } -1\right ) + \left ( b-\widehat{b}\right ) \right ] \end{array } \right\ } \\ & = o_{p}\left\ { \max\left\ { \begin{array } [ c]{c}n^{-\left ( \frac{\beta_{g}}{2\beta_{g}+d}+\frac{\beta_{p}}{d+2\beta_{p}}+\frac{\beta_{b}}{d+2\beta_{b}}\right ) } , \\
n^{-\left ( \frac{2\beta_{p}}{d+2\beta_{p}}\right ) } n^{-\frac{\beta_{p}}{d+2\beta_{p}}}n^{-\frac{\beta_{b}}{d+2\beta_{b}}},\\ n^{-\frac{2\beta_{g}}{2\beta_{g}+d}}\left\ { n^{-\frac{\beta_{b}}{d+2\beta _ { b}}}+n^{-\frac{\beta_{p}}{d+2\beta_{p}}}++n^{-\frac{\beta_{g}}{2\beta_{g}+d}}\right\ } \end{array } \right\ } \right\}\end{aligned}\ ] ] the proof of part ( iii ) was given earlier .
the remaining parts are proved in the appendix .
we conclude from this theorem that the savings in variance that comes with using @xmath2315 rather than @xmath2316 is asymptotically negligible even in regard to constants .
similarly , we conclude that the difference in variance that comes with using @xmath2317 rather than @xmath2318 is asymptotically negligible , again even in regard to constants .
further , because @xmath2319 $ ] and @xmath2320 $ ] are of the order of @xmath2321 as their first order degenerate kernels are both of order @xmath2322 and @xmath2323 \right\ } $ ] is asymptotically normal , we conclude that @xmath2324 \right\ } , \\ & n^{\frac{4\beta}{4\beta+d}-\sigma}\left\ { \mathbb{es}_{2,\widetilde{\tau } _ { k}\left ( \cdot\right ) } ^{test}\left ( \widehat{\theta}\left ( \tau^{\dagger}\right ) \right ) \right\ } ^{-1}\left [ \mathbb{es}_{2,\widetilde{\tau}_{k}\left ( \cdot\right ) } ^{test}\left ( \widehat{\theta } \left ( \tau^{\dagger}\right ) \right ) -e_{\theta}\left [ \mathbb{es}_{2,\widetilde{\tau}_{k}\left ( \cdot\right ) } ^{test}\left ( \widehat{\theta } \left ( \tau^{\dagger}\right ) \right ) \right ] \right]\end{aligned}\ ] ] and @xmath2325 \right\ } $ ] are all asymptotically normal with the same asymptotic variance .
it then follows that a necessary condition for the intervals based on @xmath2326 @xmath2327 and @xmath2328 to cover @xmath2306 at the nominal @xmath159 level as @xmath596 is that @xmath2329 , \\ & var_{\widehat{\theta}\left ( \tau^{\dagger}\right ) } \left\ { \mathbb{es}_{2,\widetilde{\tau}_{k}\left ( \cdot\right ) } ^{test}\left ( \widehat{\theta}\left ( \tau^{\dagger}\right ) \right ) \right\ } ^{-1}e_{\theta}\left [ \ \mathbb{es}_{2,\widetilde{\tau}_{k}\ } ^{test}\left ( \widehat{\theta}\left ( \tau^{\dagger}\right ) \right ) \right]\end{aligned}\ ] ] and @xmath2330 $ ] are @xmath2331 now we know under the assumptions of theorem [ bb ] that this necessary condition holds for @xmath2332 $ ] since @xmath2333 is bounded away from zero and one and , by assumption , @xmath2334 however this necessary condition need not hold for either @xmath2335\ ] ] or @xmath2330 .$ ] for example , consider the following specification consistent with our assumptions : @xmath2336 , @xmath2337 then @xmath2338 so @xmath2339 however , @xmath2340 \ $ ] converges to zero at rate @xmath2341 .
next @xmath2342 \\ & = o_{p}\left ( n^{-\left ( \frac{\beta_{b}}{d+2\beta_{b}}+\frac{2\beta_{p}}{d+2\beta_{p}}\right ) } \right ) = n^{-1/6}>>o_{p}\left ( n^{-\frac{4\beta } { 4\beta+d}+\sigma}\right ) = n^{-1/3+\sigma}\ ] ] for small @xmath2343 we conclude that the intervals based on @xmath2344 and @xmath2345 fail to cover @xmath2306 at the nominal @xmath159 level uniformly over @xmath245 as @xmath596 .
we reach the identical conclusion with regard to the parameter @xmath643 because under our assumptions @xmath2346 furthermore , by the argument used in the proof of theorem [ tt ] , it is easy to see that the length of each interval is @xmath2347 it follows that if we try to improve the coverage of the intervals based on @xmath2344 and @xmath2328 by further increasing @xmath962 the length of the intervals will increase beyond @xmath2331 we conclude that the interval based on @xmath2348 is strictly preferred to the other two intervals when @xmath2349 , @xmath2350 and is never worse in terms of shrinkage rate and coverage than the other two intervals whatever be @xmath2351 , @xmath2352 and @xmath2353 .
we reach the identical conclusion with regard to the coverage of the parameter @xmath643 because , under our assumptions including our choice of @xmath848 , @xmath2354 and @xmath2355 the order of the interval lengths .
these results translate directly into analogous results concerning the associated estimators . under our assumptions
the estimator solving @xmath2356 converges to both @xmath643 and @xmath2357 at rate @xmath2331 in contrast the rate of convergence of @xmath2358 and the estimator solving @xmath2359 converge to @xmath643 and @xmath2360 at the rates given in ( iv ) and ( v ) of theorem [ bb ] .
what is the intuition behind the above findings ?
first note that , as promised by theorem [ eiet ] and part ( vii ) of the theorem in the last subsection , the bias away from zero of @xmath2361 , e_{\theta}\left [ \tau_{2,k}\left ( \widehat{\theta}\right ) -\widetilde{\tau}_{k}\left ( \theta\right ) \right ] , $ ] and @xmath2362 $ ] are all @xmath2363 however the nature and convergence rate of the @xmath2364 term can vary markedly between estimators , attaining a minimum for @xmath2365 .$ ] now it is not surprising that , for the same order of variance , the order of @xmath2366 $ ] often exceeds that of @xmath2365 .$ ] confidence intervals for @xmath2247 based on @xmath2367 are centered at ( i.e are symmetric around ) @xmath2368 which is a quite stringent constraint on the form of the interval . in that sense , intervals based on @xmath2358 are a higher order generalization of the first order asymptotic wald intervals for @xmath2369 it is well known that when @xmath2242 is an implicit parameter that sets a functional such as @xmath2370 to zero , first - order wald confidence intervals are often outperformed in finite samples by confidence sets obtained by inverting a score - like test based on first order estimating functions for the functional that depend on the parameter @xmath2371 and , frequently , on estimated nuisance parameters as well , although this fact is not reflected in the first order asymptotics .
our example is higher order version of this phenomenon , where the benefit of the interval @xmath2372 obtained by inverting tests based on the estimating function @xmath2348 for the functional @xmath2373 is clearly and quantitatively revealed by the asymptotics .
note that , like first order wald intervals , the interval based on @xmath2374 will differ from the interval for @xmath2242 based on applying an inverse nonlinear monotone transform @xmath2375 to the end points of a wald interval for the transformed parameter @xmath2376 that is centered on @xmath2377 in contrast , like first order score - based intervals , the intervals based on @xmath2348 and @xmath2378 are invariant to monotone transformations of the parameter @xmath2242 .
more interesting and perhaps more surprising is that , for the same order of variance , the order of @xmath2379 $ ] exceeds that of @xmath2365 .$ ] the surprise derives from a failure to recognize that the theorem [ gg ] is simply too general to help select among competing procedures .
for example , this theorem implies that under law @xmath2380 ( a ) the variance @xmath2381 of @xmath2382 \right ] $ ] is less ( and generally strictly less ) than the variance of @xmath2383 , $ ] while ( b ) both have bias of @xmath2384 at first blush , this might suggest that the estimator solving @xmath2385 would likely have the same bias but smaller variance than the estimator solving @xmath2386 but we have seen that just the opposite is true .
the reason is that the difference between the variances in ( a ) is negligible in the sense that their ratio is @xmath2387 , while the @xmath2388 biases are often of quite different orders with that of @xmath2383 $ ] always a minimum .
more generally , whenever the functional @xmath658 is in our doubly robust class , eq [ ne11 ] holds so @xmath2389 is rate minimax ( or near minimax if @xmath2050 is chosen positive ) , and the suppositions of theorem [ j1 ] hold for @xmath2390 @xmath2391 theorem [ j1 ] then implies the width of the interval estimator for @xmath643 based on @xmath2392 converges to zero at the convergence rate of @xmath2393 to @xmath663
in this section , we discuss the construction of higher order influence functions for a more general class of functionals than the one thus far considered .
an important application of this construction is in the derivation of higher order influence functions in monotone missing data problems . to begin
, we must learn to construct higher order influence functions for a functional with the product form:@xmath2394 here @xmath2395 @xmath2396 are known to be higher order pathwise differentiable functionals and @xmath2397 is a known constant . the following lemma gives the general form of higher order influence functions of @xmath2398 as a function of influence functions of @xmath2399 before stating our lemma , we need additional notation .
given an ordered set of @xmath135 positive integers @xmath2400 for any @xmath276 non - negative integers @xmath2401 , satisfying @xmath2402 we define @xmath2403 to be an ordered partition of degree @xmath135 and order @xmath276 if and only if for @xmath2404 and @xmath2405 we have@xmath2406 @xmath2407{c}i_{j(s^{\ast})+1},i_{j(s^{\ast})+2}, ... ,i_{j(s^{\ast})+t_{s^{\ast}}}:\\ j(s^{\ast})=\\ \left\ { \begin{array } [ c]{c}\sum\limits_{q=1}^{s^{\ast}-1}t_{q}\text { \ \ \ \ \ for } 1<s^{\ast}\leq r\\ 0\text{\ \ \ \ \ \ \ \ \ for \ \ } s^{\ast}=1 \end{array } \right .
\end{array } \right\}\ ] ] and for all @xmath2408 we have @xmath2409 any such ordered partition satisfies@xmath2410 [ hoiprod]let @xmath2411 for @xmath2412 be the @xmath2413 order influence function of @xmath2395 and define @xmath2414 then the @xmath197 order influence function of @xmath2398 is given by:@xmath2415\end{aligned}\ ] ] where @xmath2416 = \mathbb{v}\left [ { \displaystyle\sum\limits_{\left\ { t_{1}, ... t_{\zeta}\right\ } \in \upsilon_{\zeta;j } } } { \displaystyle\prod\limits_{s=1}^{\zeta } } if_{\psi_{s}\left ( \theta\right ) ; t_{s},t_{s};\overline{i}_{s , t_{s}}}\left ( \theta\right ) \right]\]]@xmath2417 it is easy to generalize this lemma to functionals of the more general form @xmath2418 where both @xmath2419 are known constants , since the higher order influence function of a linear combination of functionals is the linear combination of the influence functions of the functionals .
we next turn to the analysis of missing data models .
suppose that we have derived the @xmath197 order influence function @xmath2420 @xmath2421 for a truncated parameter @xmath969 of a parameter @xmath175 in our doubly robust class of functionals based on i.i.d full data @xmath2422 ; then , according to theorem [ hoiprod ] the estimated @xmath2423 @xmath2424 is of the form @xmath2425 & = \sum_{1\leq l\leq k^{\left ( j-1\right ) } } \left\ { \mathbb{v}\left ( \prod\limits_{s=1}^{j}u_{l , s}^{(j)}\left ( l_{i_{s}}^{full};\widehat{\theta}\right ) \right ) \right\ } \\ & = \sum_{1\leq l\leq k^{\left ( j-1\right ) } } \mathbb{v}\left [ \prod\limits_{s=1}^{j}\widehat{u}_{l , s , i_{s}}^{(j)}\right]\end{aligned}\ ] ] where @xmath2426 @xmath2427 depends of one subject s data and has a functional form which depends on the form of @xmath547 corresponding to the functional of interest .
for example , @xmath2428 _ { i_{1}}$]corresponds to the first component of the vector contributed by person @xmath2429 to @xmath2430 .
note that @xmath2431 = 0.$ ] next , suppose that for a subject some of the data may be missing , so that we observe @xmath2432 where @xmath2433 $ ] instead of the full data @xmath2434 for now , @xmath2435 is a known function of the full data and @xmath2436 is always observed .
below we shall extend our results to monotone missing data .
define @xmath2437 and we suppose @xmath2438if we know @xmath2439 we can use @xmath2440\ ] ] instead of @xmath2441 $ ] , where @xmath2442 is an arbitrary function with finite variance which satisfies @xmath2443 = 0.$ ] it is easy to verify that this statistic is a function of the observed data and an unbiased estimator of@xmath2444 \right]\ ] ] so that in fact , in terms of rates , our observed data statistic has the same bias and variance properties as the original full data higher order influence function . moreover , the most efficient ( in terms of constants ) choice for @xmath2445 in our class of estimators is @xmath2446 while @xmath2447 is an unknown function to be estimated from the observed data .
this last two observations motivate our proposed strategy for mapping higher order influence functions in the full data model to higher order influence functions in the observed data model when the missingness mechanism is unknown .
first , define @xmath2448 \right ] \\ & = e_{\theta}\left [ \mathbb{v}\left [ \prod\limits_{s=1}^{j}\left ( \frac{r_{i_{s}}\left [ \widehat{u}_{l , s , i_{s}}^{(j)}-b_{l , s}^{(j)}\left ( l_{obs , i_{s}};\theta\right ) \right ] } { \pi\left ( l_{obs,,i_{s}};\theta\right ) } + b_{l , s}^{(j)}\left ( l_{obs , i_{s}};\theta\right ) \right ) \right ] \right]\end{aligned}\ ] ] which we notice is of the product form that was discussed earlier in this section .
in order to construct an @xmath197 order influence function for @xmath970 it is now apparent that we must first succeed in constructing @xmath197 order influence functions for the set of parameters @xmath2449
so as to guarantee an estimation bias of order @xmath2450 now , for each @xmath2451 and @xmath212 , @xmath2452 is itself a member of our general doubly robust class so that @xmath2453 $ ] where @xmath2454 @xmath2455and @xmath2456 .
thus , it immediately follows that neither @xmath2457 nor @xmath969 have higher order influence functions .
we proceed by constructing influence functions for the set of truncated parameters @xmath2458 where @xmath2459 are appropriately truncated versions of @xmath2460 as in section 3.2.2 .
we then define the truncated parameter @xmath2461 @xmath2462 note that @xmath2463 differs from the truncated parameters @xmath930 of section 3.2.2 , in that it depends on the order of the influence function we plan to base our inference on .
the next theorem gives the @xmath197 order influence function of @xmath2464 [ map]let @xmath2465the @xmath197 order influence function of @xmath2463 is given by@xmath2466\ ] ] where @xmath2467@xmath2468 the proof follows directly from the lemma [ hoiprod ] applied to each element @xmath2469 the @xmath197 order estimated influence function of @xmath2463 is given by@xmath2470\ ] ] where @xmath2471\]]@xmath2472 this result follows immediately from theorem [ map ] and the fact that @xmath2473 by definition .
we are now ready to use our theorem to derive higher order influence functions for a functional @xmath175 from monotone missing data .
we begin with the simple two - occasion case .
let @xmath2474 @xmath2475 @xmath2476 be @xmath5 @xmath2477 copies constitute the full data .
the outcome of interest , @xmath2478 is univariate . @xmath2479 and @xmath2480 are vectors of continuous covariates with dimensions @xmath2481 and @xmath2482 respectively . the observed data @xmath2483 where both @xmath2484 and @xmath2485 are binary missing indicators .
the outcome @xmath0 is observed if and only if @xmath2486
@xmath2487 = @xmath470 while @xmath2488 is observed if and only if @xmath2486 @xmath2489 .
the data is assumed to be missing at random , that is : @xmath2490 and @xmath2491 under this monotone missing - data pattern , the parameter of interest is given by @xmath2492 we impose the following assumptions : @xmath2493 @xmath2494 \in h\left ( \beta_{b_{0}},c_{b_{0}}\right ) ; $ ] @xmath2495 @xmath2496 \in h\left ( \beta_{b_{1}},c_{b_{1}}\right ) ; $ ] @xmath2497 @xmath2498 and @xmath2499 w.p . 1 @xmath2500 @xmath2501 and @xmath2502 w.p .
1@xmath64 @xmath2503 the marginal density of @xmath2504 @xmath2505 falls in a hlder ball @xmath2506 and @xmath2507 w.p .
1 , @xmath2508 @xmath2509 the marginal density of @xmath2510 @xmath2511 falls in a hlder ball @xmath2512 and @xmath2513 w.p .
1 , @xmath2514 we define @xmath2515 , @xmath2516 with corresponding hlder exponents @xmath2517 and @xmath2518 respectively
. we next show how to apply theorem [ map ] in a nested fashion in order to derive higher order u - statistic estimators @xmath2519 in the observed data model . in the first step of our procedure ,
we derive higher order influence functions for a truncated parameter in the artificial missing data problem in which the observed data is given by @xmath2520 rather than @xmath2521 in the second step , we construct influence functions from a second artificial missing data problem with @xmath2522 now the full data and @xmath2523 the observed data .
this final influence function is , in fact , the influence function for a truncated parameter in the original monotone missing data model . to follow this nested procedure ,
we derive a first stage class of estimators @xmath2524 which are functions of @xmath2525 in this model , @xmath2526 is the full data , @xmath2522 is the observed data , @xmath2484 is the missing indicator , @xmath2527 is a vector of always observed covariates , and @xmath2528 is the outcome which might be missing .
since the parameter of interest @xmath175 is the marginal mean of @xmath2529 this is example 2a * * * * of section 3.1**. * * therefore results from this example may be applied , hence @xmath2530 and @xmath2531 moreover , @xmath2532 so that we can define @xmath2533 ^{-1}e_{\theta}\left [ \frac{r_{0}}{\widehat{\pi}_{0}}\left ( y-\widehat{b}_{0}\right ) \overline{z}_{k_{0}}\right ] \\ \widetilde{\pi}_{0}^{-1 } & = \widehat{\pi}_{0}^{-1}\left ( 1-\overline { z}_{k_{0}}^{t}e_{\theta}\left [ \frac{r_{0}}{\widehat{\pi}_{0}}\overline { z}_{k_{0}}\overline{z}_{k_{0}}^{t}\right ] ^{-1}e_{\theta}\left [ \left ( \frac{r_{0}}{\widehat{\pi}_{0}}-1\right ) \overline{z}_{k_{0}}\right ] \right ) \\ \widetilde{\psi}^{\dag}\left ( \theta\right ) & = e_{\theta}\left [ \frac{r_{0}}{\widetilde{\pi}_{0}}\left ( y-\widetilde{b}_{0}\right ) + \widetilde{b}_{0}\right]\end{aligned}\ ] ] with @xmath2534 and @xmath2535 moreover@xmath2536 @xmath2537 are rate optimal nonparametric estimators of @xmath2538 respectively estimated from the training sample and @xmath2539 from theorem [ tbrate ] and eq.([eb6 ] ) we also know that:@xmath2540 \right)\end{gathered}\ ] ] next , we proceed with the second step of our procedure , where @xmath2522 is now the full data and @xmath2523 becomes the observed data
. then , @xmath2541 if @xmath2542 or @xmath2543 therefore we may define a new missing indicator @xmath2544 with @xmath2545 in contrast with the first phase of our procedure , the full data influence function now has non - zero higher order contributions ; thus we can proceed with the strategy layed out in the first part of this section .
we can put @xmath2546 in the format of eq.([sum ] ) as @xmath2547{c}\left ( \frac{r_{0}}{\widehat{\pi}_{0}}\left ( y-\widehat{b}_{0}\right ) z_{s_{1}}\right ) _ { i_{1}}\left ( \left ( \frac{r_{0}}{\widehat{\pi}_{0}}-1\right ) z_{s_{j-1}}\right ) _ { i_{j}}\times\\{\textstyle\prod\limits_{t=2}^{j-1 } } \left ( \frac{r_{0}}{\widehat{\pi}_{0}}z_{s_{t-1}}z_{s_{t}}-i\left ( s_{t-1}=s_{t}\right ) \right ) _ { i_{t}}\end{array } \right\ } \\ & = \left ( -1\right ) ^{j-1}\sum_{l=1}^{k_{0}^{j-1}}\left\ { \begin{array } [ c]{c}\left ( \frac{r_{0}}{\widehat{\pi}_{0}}\left ( y-\widehat{b}_{0}\right ) z_{n\left ( l,1\right ) } \right ) _ { i_{1}}\left ( \left ( \frac{r_{0}}{\widehat{\pi}_{0}}-1\right ) z_{_{n\left ( l , j-1\right ) } } \right ) _ { i_{j}}\times\\{\textstyle\prod\limits_{t=2}^{j-1 } } \left ( \frac{r_{0}}{\widehat{\pi}_{0}}z_{n\left ( l , t-1\right ) } z_{n\left ( l , t\right ) } -i\left ( n\left ( l , t-1\right ) = n\left ( l , t\right ) \right ) \right ) _ { i_{t}}\end{array } \right\}\end{aligned}\ ] ] where @xmath2548 is a one - to - one mapping that indexes all permutations of @xmath2549 and @xmath2550 is the @xmath781th entry of @xmath2551 we define @xmath1939 @xmath2552@xmath2553{c}$\frac{r_{0}}{\widehat{\pi}_{0}}\left ( y-\widehat{b}_{0}\right ) z_{n\left ( l,1\right ) } $ \ \ \ for{}{}{}$s=1$\\ $ \left\ { \begin{array } [ c]{c}\frac{r_{0}}{\widehat{\pi}_{0}}z_{n\left ( l , s-1\right ) } z_{n\left ( l , s\right ) } \\
-i\left ( n\left ( l , s-1\right ) = n\left (
l , s\right ) \right ) \end{array } \right\ } $ \ for { } { } $ 1<s < j$\\ $ \left ( -1\right ) ^{j-1}\left ( \frac{r_{0}}{\widehat{\pi}_{0}}-1\right ) z_{_{n\left ( l , j-1\right ) } } $ \ { } for { } $ s = j$\end{tabular } \ \right .
\\ \tau_{l , s}^{\left ( j\right ) } \left ( \theta\right ) & \equiv e_{\theta } \left ( \widehat{u}_{l , s}^{\left ( j\right ) } \right)\end{aligned}\ ] ] let @xmath2554 which is always observed
. then @xmath2555{c}\frac{r_{0}}{\widehat{\pi}_{0}}\left ( y-\widehat{b}_{0}\right ) z_{n\left ( l,1\right ) } -\\ e\left ( \frac{r_{0}}{\widehat{\pi}_{0}}\left ( y-\widehat{b}_{0}\right ) z_{n\left ( l,1\right ) } |v\right ) \end{array } \right ] + e_{\theta}\left ( \frac{r_{0}}{\widehat{\pi}_{0}}\left ( y-\widehat{b}_{0}\right ) z_{n\left ( l,1\right ) } |v\right ) \right ) \\ = e_{\theta}\left ( \frac{r_{1}}{\pi_{1}\left ( \theta\right ) } \frac{r_{0}}{\widehat{\pi}_{0}}\left ( yz_{n\left ( l,1\right ) } -b_{1}\left ( \theta\right ) z_{n\left ( l,1\right ) } \right ) + \frac{r_{0}}{\widehat{\pi } _ { 0}}\left ( b_{1}\left ( \theta\right ) z_{n\left ( l,1\right ) } -\widehat{b}_{0}z_{n\left ( l,1\right ) } \right ) \right)\end{gathered}\ ] ] thus , @xmath2556 falls into the doubly - robust @xmath581 class of functionals with @xmath2557 @xmath2558 and corresponding @xmath2559 @xmath2560 @xmath2561 @xmath2562 for any @xmath2563 @xmath2564 is a known function of the observed data , thus @xmath2565 and @xmath2566 for any @xmath337 moreover , choosing @xmath2567 @xmath2568
so that : @xmath2569{c}\widehat{b}_{1}z_{n\left ( l,1\right ) } + \overline{w}_{k_{1}}^{t}e_{\theta } \left ( \frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}\right ) ^{-1}\\ \times e_{\theta}\left ( \frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\left ( y-\widehat{b}_{1}\right ) z_{n\left ( l,1\right ) } \overline{w}_{k_{1}}\right ) \end{array } \right\ } \\ \widetilde{\pi}_{1}^{-1 } & = \left\ { \widehat{\pi}_{1}^{-1}\left ( \begin{array } [ c]{c}1-\overline{w}_{k_{1}}^{t}e_{\theta}\left [ \frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}\right ] ^{-1}\\ \times e_{\theta}\left [ \frac{r_{0}}{\widehat{\pi}_{0}}\left ( \frac{r_{1}}{\widehat{\pi}_{1}}-1\right ) \overline{w}_{k_{1}}\right ] \end{array } \right ) \right\}\end{aligned}\ ] ] where @xmath2570 and @xmath2571 is a @xmath2572dimensional vector of tensor product basis for functions of @xmath2573 from theorem [ map ] , for @xmath2574 @xmath2575{c}r_{1}\widetilde{\pi}_{1}^{-1}\left ( \theta\right ) \frac{r_{0}}{\widehat{\pi } _ { 0}}\left ( yz_{n\left ( l,1\right ) } -\widetilde{b_{1}\left ( \theta\right ) z_{n\left ( l,1\right ) } } \right ) \\ + \frac{r_{0}}{\widehat{\pi}_{0}}\left ( \widetilde{b_{1}\left ( \theta\right ) z_{n\left ( l,1\right ) } } -\widehat{b}_{0}z_{n\left ( l,1\right ) } \right ) \end{array } \right)\end{aligned}\ ] ] have higher order influence functions @xmath2576 .
let @xmath2577 for any @xmath2578 and @xmath2579 be the higher order influence functions of @xmath2580 . for @xmath2581
define @xmath2582 @xmath2583 also belongs to the @xmath2584 class of models with @xmath2585 @xmath2586 @xmath2587 @xmath2588 @xmath2589 @xmath2590 we may choose @xmath2591 @xmath2592so that @xmath2593{c}\widehat{b}_{1}+\overline{w}_{k_{1}}^{t}e_{\theta}\left ( \frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}\right ) ^{-1}\\ \times e_{\theta}\left ( \frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\left ( y-\widehat{b}_{1}\right ) \overline{w}_{k_{1}}\right ) \end{array } \right\ } \\ \widetilde{\pi}_{1}^{-1 } & = \left\ { \widehat{\pi}_{1}^{-1}\left ( \begin{array } [ c]{c}1-\overline{w}_{k_{1}}^{t}e_{\theta}\left [ \frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}\right ] ^{-1}\\ \times e_{\theta}\left [ \frac{r_{0}}{\widehat{\pi}_{0}}\left ( \frac{r_{1}}{\widehat{\pi}_{1}}-1\right ) \overline{w}_{k_{1}}\right ] \end{array } \right ) \right\}\end{aligned}\ ] ] and@xmath2594 has higher order ifs @xmath2595 define @xmath2596 then @xmath1939 @xmath2597@xmath2598 with @xmath2599{c}\frac{r_{1,i}}{\widehat{\pi}_{1,i}}\frac{r_{0,i}}{\widehat{\pi}_{0,i}}\left ( y_{i}-\widehat{b}_{1,i}\right ) + \\
\frac{r_{0,i}}{\widehat{\pi}_{0,i}}\left ( \widehat{b}_{1,i}-\widehat{b}_{0,i}\right ) + \widehat{b}_{0,i}-\psi\left ( \widehat{\theta}\right ) \end{array } \right\}\ ] ] and @xmath1939 @xmath2600@xmath2601{c}\left\ { \begin{array } [ c]{c}\left [ \frac{r_{0}}{\widehat{\pi}_{0}}\frac{r_{1}}{\widehat{\pi}_{1}}\left ( y-\widehat{b}_{1}\right ) \right ] _ { i_{1}}\overline{w}_{k_{1},i_{1}}^{t}\times\\{\displaystyle\prod\limits_{s=2}^{j-1 } } \left ( \frac{r_{0}}{\widehat{\pi}_{0}}\frac{r_{1}}{\widehat{\pi}_{1}}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}-i\right ) _ { i_{s}}\overline { w}_{k_{1},i_{j}}\left [ \frac{r_{0}}{\widehat{\pi}_{0}}\left ( \frac{r_{1}}{\widehat{\pi}_{1}}-1\right ) \right ] _ { i_{j}}\end{array } \right\ } \\{\textstyle\sum\limits_{t=2}^{j-1 } } \left\ { \begin{array } [ c]{c}\left ( \frac{r_{0}}{\widehat{\pi}_{0}}\left ( \frac{r_{1}}{\widehat{\pi}_{1}}-1\right ) \right ) _ { i_{j}}\overline{w}_{k_{1},i_{j}}^{t}{\textstyle\prod\limits_{l = t+1}^{j-1 } } \left ( \frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}-i\right ) _ { i_{l}}\\ \times\left [ \overline{w}_{k_{1}}\frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\left ( y-\widehat{b}_{1}\right ) \overline{z}_{k_{0}}^{t}\right ] _ { i_{1}}\times\\{\displaystyle\prod\limits_{s=2}^{t-1 } } \left ( \frac{r_{0}}{\widehat{\pi}_{0}}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ) _ { i_{s}}\left ( \frac{r_{0}}{\widehat{\pi}_{0}}-1\right ) _ { i_{t}}\overline{z}_{k_{0,}i_{t}}\end{array } \right\ } \\ + \left\ { \begin{array } [ c]{c}\left [ \frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\left ( y-\widehat{b}_{1}\right ) + \frac{r_{0}}{\widehat{\pi}_{0}}\left ( \widehat{b}_{1}-\widehat{b}_{0}\right ) \right ] _ { i_{1}}\overline{z}_{k_{0},i_{1}}^{t}\times\\{\displaystyle\prod\limits_{s=2}^{j-1 } } \left ( \frac{r_{0}}{\widehat{\pi}_{0}}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ) _ { i_{s}}\left ( \frac{r_{0}}{\widehat{\pi}_{0}}-1\right ) _ { i_{j}}\overline{z}_{k_{0,}i_{j}}\end{array } \right\ } \end{array } \right]\end{aligned}\ ] ] this follows directly from theorem [ map ] and the fact that @xmath2602 @xmath2603 for @xmath2604 suppose the class of estimators @xmath2605 are defined as @xmath2606 then@xmath2607{c}e_{\theta}\left ( \psi\left ( \widehat{\theta}\right ) + \mathbb{if}_{m,\widetilde{\widetilde{\psi}}_{k_{0},k_{1},m}\left ( \theta\right ) } \left ( \widehat{\theta}\right ) -\widetilde{\widetilde{\psi}}_{k_{0},k_{1},m}\left ( \theta\right ) \right ) \\ + \left ( \widetilde{\widetilde{\psi}}_{k_{0},k_{1},m}\left ( \theta\right ) -e_{\theta}\left [ \mathbb{if}_{m,\widetilde{\psi}^{\dag}\left ( \theta\right ) } \left ( \widehat{\theta}\right ) + \psi\left ( \widehat{\theta } \right ) \right ] \right )
\\ + \left [ e_{\theta}\left ( \mathbb{if}_{m,\widetilde{\psi}^{\dag}}\left ( \widehat{\theta}\right ) + \psi\left ( \widehat{\theta}\right ) \right ) -\widetilde{\psi}^{\dag}\left ( \theta\right ) \right ] + \left ( \widetilde{\psi}^{\dag}\left ( \theta\right ) -\psi\left ( \theta\right ) \right ) \end{array } \right\ } \label{bsp}\ ] ] the following theorem examines the bias and variance of @xmath2608define @xmath2609 @xmath2610 and @xmath2611 for @xmath2612 let @xmath2613 and @xmath2614 [ btomm]suppose conditions @xmath2615 hold then@xmath2616\ ] ] and @xmath1939 @xmath1927@xmath2617 where @xmath2618{c}\left\ { \begin{array } [ c]{c}e_{\theta}\left\ { \left [ q_{0}\delta b_{0}\right ] \overline{z}_{k_{0}}^{t}\right\ } e_{\theta}\left [ q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}\right ] ^{-1}\times\\ \left [ e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}\right ) -i\right ] ^{m-1}e_{\theta}\left [ \overline{z}_{k_{0}}\delta p_{0}\right ] \end{array } \right\ } _ { -\left ( eb_{1}^{\left ( 1\right ) } \right ) }
\\ + \left\ { \begin{array } [ c]{c}e_{\theta}\left\ { \left [ q_{01}\delta b_{1}\right ] \overline{w}_{k_{1}}^{t}\right\ } e_{\theta}\left [ q_{01}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}\right ] ^{-1}\times\\ \left [ e_{\theta}\left ( q_{01}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}\right ) -i\right ] ^{m-1}e_{\theta}\left [ \overline{w}_{k_{1}}q_{0}\delta p_{1}\right ] \end{array } \right\ } _ { -\left ( eb_{1}^{\left ( 2\right ) } \right ) } \\
+ { \displaystyle\sum\limits_{j=2}^{m-1 } } \left\ { \begin{array } [ c]{c}e_{\theta}\left [ \overline{z}_{k_{0}}\delta p_{0}^{t}\right ] \left [ e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ) \right ] ^{j-2}\times\\ e_{\theta}\left [ q_{01}\delta b_{1}\overline{z}_{k_{0}}\overline{w}_{k_{1}}^{t}\right ] e_{\theta}\left [ q_{01}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}\right ] ^{-1}\times\\ e_{\theta}\left ( q_{01}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}-i\right ) ^{m - j}e_{\theta}\left [ \overline{w}_{k_{1}}q_{0}\delta p_{1}\right ] \end{array } \right\ } _ { -\left ( eb_{jj}^{\left ( 2\right ) } \right ) } \\ + e_{\theta}\left\ { \left [ q_{0}\delta p_{1}\delta b_{1}\right ] \overline { z}_{k_{0}}^{t}\right\ } \left [ e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}\right ) -i\right ] ^{m-2}e_{\theta}\left [ \overline{z}_{k_{0}}\delta p_{0}\right ] _ { -\left ( eb_{mm}^{\left ( 2\right ) } \right ) } \end{array } \right\ } \label{ebm}\]]@xmath2619{c}\left\ { \begin{array } [ c]{c}\left\vert \left\vert \widehat{f}_{0}\right\vert \right\vert _ { \infty } \left\vert \left\vert q_{0}^{-1/2}\right\vert \right\vert _ { \infty}\left\vert \left\vert q_{0}\right\vert \right\vert _ { \infty}\left\vert \left\vert \frac{\widehat{f}_{0}}{f_{0}}\right\vert \right\vert
_ { \infty}\times\\ \left\vert \left\vert \delta g_{0}\right\vert \right\vert
_ { \infty}^{m-1}\left\ { \int\left ( b_{0}-\widehat{b}_{0}\right ) ^{2}dl_{0}\right\ } ^{1/2}\times\\ \left\vert \left\vert \frac{f_{0}}{\pi_{0}\widehat{\pi}_{0}}\right\vert \right\vert _ { \infty}\left\ { \int\left ( \pi_{0}-\widehat{\pi}_{0}\right ) ^{2}dl_{0}\right\ } ^{1/2}\end{array } \right\ } \\ + \left\ { \begin{array } [ c]{c}\left\vert \left\vert \widehat{f}_{1}\right\vert \right\vert _ { \infty } \left\vert \left\vert q_{01}^{-1/2}\right\vert \right\vert _ { \infty } \left\vert \left\vert q_{01}\right\vert \right\vert _ { \infty}\left\vert \left\vert \frac{\widehat{f}_{1}}{f_{1}}\right\vert \right\vert _
{ \infty } \times\\ \left\vert \left\vert \delta g_{1}\right\vert \right\vert _ { \infty}^{m-1}\left\ { \int\left ( b_{1}-\widehat{b}_{1}\right ) ^{2}dl_{0}dl_{1}\right\ } ^{1/2}\\ \times\left\vert \left\vert \frac{\pi_{0}f_{1}}{\widehat{\pi}_{0}\pi _ { 1}\widehat{\pi}_{1}}\right\vert \right\vert _ { \infty}\left\ { \int\left ( \pi_{1}-\widehat{\pi}_{1}\right ) ^{2}dl_{0}dl_{1}\right\ } ^{1/2}\end{array } \right\ } \\ + { \displaystyle\sum\limits_{j=2}^{m-1 } } \left\ { \begin{array } [ c]{c}\left\vert \left\vert \frac{f_{1}}{\widehat{f}_{1}}\right\vert \right\vert _ { \infty}\left\vert \left\vert \frac{f_{0}^{2}}{\widehat{f}_{0}\widehat{\pi } _ { 0}^{2}}\right\vert \right\vert _ { \infty}\left\vert \left\vert \frac{\widehat{f}_{1}}{f_{1}}\right\vert \right\vert _ { \infty}\left\vert \left\vert \frac{\pi_{0}f_{1}}{\widehat{\pi}_{0}\pi_{1}\widehat{\pi}_{1}}\right\vert \right\vert
_ { \infty}\\ \left\vert \left\vert q_{01}\right\vert \right\vert _ { \infty}\left\vert \left\vert \delta b_{1}\right\vert \right\vert
_ { \infty}\left\vert \left\vert \delta g_{0}\right\vert \right\vert
_ { \infty}^{j-2}\left\ { \int\left ( \pi_{0}-\widehat{\pi}_{0}\right ) ^{2}dl_{0}\right\ } ^{1/2}\times\\ \left\vert \left\vert q_{01}^{-1/2}\right\vert \right\vert _ { \infty}\left\vert \left\vert \delta g_{1}\right\vert \right\vert _ { \infty}^{m - j}\left\ { \int\left ( \pi_{1}-\widehat{\pi}_{1}\right ) ^{2}dl_{0}dl_{1}\right\ } ^{1/2}\end{array } \right\ }
\\ + \left\ { \begin{array } [ c]{c}\left\vert \left\vert \frac{f_{1}}{\widehat{f}_{1}\widehat{\pi}_{1}^{2}}q_{0}\right\vert \right\vert _ { \infty}\left\vert \left\vert \delta b_{1}\right\vert \right\vert _ { \infty}\left\ { \int\left ( \pi_{1}-\widehat{\pi}_{1}\right ) ^{2}dl_{0}dl_{1}\right\ } ^{1/2}\\ \times\left\vert \left\vert \frac{f_{0}^{2}}{\widehat{f}_{0}\widehat{\pi}_{0}^{2}}\right\vert \right\vert \left\vert \left\vert \delta g_{0}\right\vert \right\vert
_ { \infty}^{m-2}\left\ { \int\left ( \pi_{0}-\widehat{\pi}_{0}\right ) ^{2}dl_{0}\right\ } ^{1/2}\end{array } \right\ } \end{array } \right ] \label{ebm2}\\ & = o_{p}\left ( \max\left [ \begin{array } [ c]{c}\left ( \frac{\log n}{n}\right ) ^{\frac{\left ( m-1\right ) \beta_{g_{0}}}{d_{0}+2\beta_{g_{0}}}}n^{-\left ( \frac{\beta_{b_{0}}}{d_{0}+2\beta_{b_{0}}}+\frac{\beta_{\pi_{0}}}{d_{0}+2\beta_{\pi_{0}}}\right ) } , \\ \left ( \frac{\log n}{n}\right ) ^{\frac{\left ( m-1\right ) \beta_{g_{1}}}{d_{1}+2\beta_{g_{1}}}}n^{-\left ( \frac{\beta_{b_{1}}}{d_{1}+2\beta_{b_{1}}}+\frac{\beta_{\pi_{1}}}{d_{1}+2\beta_{\pi_{1}}}\right ) } , \\{\displaystyle\sum\limits_{j=2}^{m } } \left ( \frac{\log n}{n}\right ) ^{\frac{\beta_{b_{1}}}{d_{1}+2\beta_{b_{1}}}+\frac{\left ( j-2\right ) \beta_{g_{0}}}{d_{0}+2\beta_{g_{0}}}+\frac{\left ( m - j\right ) \beta_{g_{1}}}{d_{1}+2\beta_{g_{1}}}}n^{-\frac{\beta_{\pi_{0}}}{d_{0}+2\beta_{\pi_{0}}}-\frac{\beta_{\pi_{1}}}{d_{1}+2\beta_{\pi_{1}}}}\end{array } \right ] \right ) \label{ebm3}\ ] ] and@xmath2620{c}e_{\theta}\left ( \pi_{\theta}^{\bot}\left ( q_{0}^{1/2}\delta b_{0}|\left ( q_{0}^{1/2}\overline{z}_{k_{0}}\right ) \right ) \pi_{\theta}^{\bot}\left ( q_{0}^{-1/2}\delta p_{0}|\left ( q_{0}^{1/2}\overline{z}_{k_{0}}\right ) \right ) \right ) \\ + e_{\theta}\left ( \pi_{\theta}^{\bot}\left ( q_{01}^{1/2}\delta b_{1}|\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ) \pi_{\theta}^{\bot}\left ( q_{0}q_{01}^{-1/2}\delta p_{1}|\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ) \right ) \end{array } \right\ } + i\left ( m>2\right ) \times\nonumber\\ & \left\ { \begin{array } [ c]{c}-e_{\theta}\left [ \begin{array } [ c]{c}\pi_{\theta}^{\bot}\left [ \left ( \begin{array } [ c]{c}\frac{\pi_{1}^{1/2}}{\widehat{\pi}_{1}^{1/2}}\delta b_{1}\pi_{\theta}\left ( q_{0}^{1/2}\delta p_{0}|\left ( q_{0}^{1/2}\overline{z}_{k_{0}}\right ) \right ) \end{array } \right ) |\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ] \\ \times\pi_{\theta}^{\bot}\left [ q_{0}q_{01}^{-1/2}\delta p_{1}|\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ] \end{array } \right ] _ { -\left ( tb\left ( m,2\right ) \right ) } \\ + \left ( -1\right ) ^{m}e_{\theta}\left [ \begin{array } [ c]{c}\pi_{\theta}^{\bot}\left [ \left ( \begin{array } [ c]{c}q_{01}^{1/2}\delta b_{1}e_{\theta}\left [ \delta p_{0}\overline{z}_{k_{0}}^{t}\right ] e_{\theta}\left [ q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}\right ] ^{-1}\\ \times\left ( e_{\theta}\left [ q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ] \right ) ^{m-2}\overline{z}_{k_{0}}\end{array } \right ) |\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ] \\ \times\pi_{\theta}^{\bot}\left [ q_{0}q_{01}^{-1/2}\delta p_{1}|\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ] \end{array } \right ] _ { -\left ( tb\left ( m,3\right ) \right ) } \end{array } \right\ } \label{tbm}\]]@xmath2621{c}k_{0}^{-\left ( \beta_{b_{0}}+\beta_{\pi_{0}}\right ) /d_{0}},k_{1}^{-\left ( \beta_{b_{1}}+\beta_{\pi_{1}}\right ) /d_{1}},k_{1}^{-\beta_{\pi_{1}}/d_{1}}k_{0}^{-\beta_{\pi_{0}}/d_{0}}\left ( \frac{\log n}{n}\right ) ^{\frac { \beta_{b_{1}}}{d_{1}+\beta_{b_{1}}}}\\ k_{1}^{-\left ( \min\left ( \beta_{\pi_{0}},\beta_{b_{1}}\right ) + \beta _ { \pi_{1}}\right ) /d_{1}},\left ( \frac{\log n}{n}\right ) ^{-\frac { \beta_{b_{1}}}{d_{1}+\beta_{b_{1}}}-\frac{\left ( m-2\right ) \beta_{g_{0}}}{d_{0}+\beta_{g_{0}}}}n^{-\frac{\beta_{\pi_{0}}}{d_{0}+\beta_{\pi_{0}}}}k_{1}^{-\beta_{\pi_{1}}/d_{1}}\end{array } \right ] \right ) \label{tbm2}\ ] ] moreover , @xmath2622 the optimal choice of @xmath2623 and @xmath2624 in this class of higher order estimators , will depend on the size of the effective smoothness exponents @xmath2625,@xmath2626 and @xmath2627 relative to the size of the exponents @xmath2628,@xmath2629 and @xmath2630 a general prescription for finding an optimal estimator in our class is to choose a pair @xmath2631 that minimizes the maximum asymptotic mse over the model among the candidate @xmath2632 here , the estimator @xmath2633 uses the pair @xmath2634 of @xmath135 that equates the order of the variance to the order of the maximum between the squared truncation and estimation biases ( which are given in the theorem ) theorem [ map ] can be applied in a nested fashion to the estimation of general functionals in nonparametric models with monotone missingness ( that is with an arbitrary number of occasions ) . building on the two - occasion case ,
@xmath2635 where @xmath2636 is the missing indicator for the third occasion .
we also write @xmath2637 and @xmath2638 let @xmath2639 and @xmath2640 is a @xmath2641dimensional vector of tensor product basis for functions of @xmath2642 with finite variance .
the truncated parameter @xmath2643 is given by : @xmath2644 where for all @xmath2645 @xmath2646 @xmath2647{c}1-\overline{v}_{k_{2}}^{t}e_{\theta}\left ( \frac{r_{2}}{\widehat{\pi}_{2}}\frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\overline { v}_{k_{2}}\overline{v}_{k_{2}}^{t}\right ) ^{-1}\\ \times e_{\theta}\left [ \frac{r_{0}}{\widehat{\pi}_{0}}\frac{r_{1}}{\widehat{\pi}_{1}}\left ( \frac{r_{2}}{\widehat{\pi}_{2}}-1\right ) \overline{v}_{k_{2}}\right ] \end{array } \right ) \right\ } \\ \widetilde{b}_{2}^{\left ( 3\right ) } & = \left\ { \begin{array } [ c]{c}\widehat{b}_{2}+\overline{v}_{k_{2}}^{t}e_{\theta}\left ( \frac{r_{2}}{\widehat{\pi}_{2}}\frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi } _ { 0}}\overline{v}_{k_{2}}\overline{v}_{k_{2}}^{t}\right ) ^{-1}\\ \times e_{\theta}\left ( \frac{r_{2}}{\widehat{\pi}_{2}}\frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\left ( y-\widehat{b}_{2}\right ) \overline{v}_{k_{2}}\right ) \end{array } \right\ } \\ \widetilde{w_{s}b}_{2}^{\left ( 3\right ) } & = \left\ { \begin{array } [ c]{c}\widehat{b}_{2}w_{s}+\overline{v}_{k_{2}}^{t}e_{\theta}\left ( \frac{r_{2}}{\widehat{\pi}_{2}}\frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi } _ { 0}}\overline{v}_{k_{2}}\overline{v}_{k_{2}}^{t}\right ) ^{-1}\\ \times e_{\theta}\left ( \frac{r_{2}}{\widehat{\pi}_{2}}\frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\left ( y-\widehat{b}_{2}\right ) w_{s}\overline{v}_{k_{2}}\right ) \end{array } \right\ } \\
\widetilde{z_{t}w_{s}b}_{2}^{\left ( 3\right ) } & = \left\ { \begin{array } [ c]{c}\widehat{b}_{2}z_{t}w_{s}+\overline{v}_{k_{2}}^{t}e_{\theta}\left ( \frac{r_{2}}{\widehat{\pi}_{2}}\frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\overline{v}_{k_{2}}\overline{v}_{k_{2}}^{t}\right ) ^{-1}\\ \times e_{\theta}\left ( \frac{r_{2}}{\widehat{\pi}_{2}}\frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\left ( y-\widehat{b}_{2}\right ) z_{t}w_{s}\overline{v}_{k_{2}}\right ) \end{array } \right\ } \\ \widetilde{z_{t}b}_{2}^{\left ( 3\right ) } & = \left\ { \begin{array } [ c]{c}\widehat{b}_{2}z_{t}+\overline{v}_{k_{2}}^{t}e_{\theta}\left ( \frac{r_{2}}{\widehat{\pi}_{2}}\frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi } _ { 0}}\overline{v}_{k_{2}}\overline{v}_{k_{2}}^{t}\right ) ^{-1}\\ \times e_{\theta}\left ( \frac{r_{2}}{\widehat{\pi}_{2}}\frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\left ( y-\widehat{b}_{2}\right ) z_{t}\overline{v}_{k_{2}}\right ) \end{array } \right\}\end{aligned}\ ] ] and@xmath2648{c}e\left ( \frac{r_{2}}{\widetilde{\pi}_{2}^{\left ( 3\right ) } } \frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\left ( y-\widetilde{b}_{2}^{\left ( 3\right ) } \right ) + \frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\left ( \widetilde{b}_{2}^{\left ( 3\right ) } -\widehat{b}_{1}\right ) \right ) \\ e\left ( \frac{r_{0}}{\widehat{\pi}_{0}}\left ( \widehat{b}_{1}-\widehat{b}_{0}\right ) + \widehat{b}_{0}-\psi\left ( \widehat{\theta}\right ) \right ) \end{array } \right\}\ ] ] @xmath2649{c}\left\ { \left ( a_{s}\right ) _ { 1\times k_{1}}\left [ e\left ( \frac{r_{0}}{\widehat{\pi}_{0}}\frac{r_{1}}{\widehat{\pi}_{1}}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}-i\right ) \right ] ^{j-2}e\left [ \overline { w}_{k_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\left ( \frac{r_{1}}{\widehat{\pi } _ { 1}}-1\right ) \right ] \right\ } \\
+ { \textstyle\sum\limits_{q=2}^{j-1 } } \left\ { \begin{array } [ c]{c}e\left ( \left ( \frac{r_{0}}{\widehat{\pi}_{0}}\left ( \frac{r_{1}}{\widehat{\pi}_{1}}-1\right )
\right ) \overline{w}_{k_{1}}^{t}\right ) \left [ e\left ( \frac{r_{1}}{\widehat{\pi}_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}-i\right ) \right ] ^{j - q-1}\\ \times\left ( b_{s , t}\right ) _ { k_{1}\times k_{0}}\times\\ \left [ e\left ( \frac{r_{0}}{\widehat{\pi}_{0}}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ) \right ] ^{q-2}e\left [ \left ( \frac { r_{0}}{\widehat{\pi}_{0}}-1\right ) \overline{z}_{k_{0}}\right ] \end{array } \right\ } \\ + \left\ { \left ( c_{t}\right ) _ { 1\times k_{0}}\times\left [ e\left ( \frac{r_{0}}{\widehat{\pi}_{0}}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ) \right ] ^{j-2}e\left [ \left ( \frac{r_{0}}{\widehat{\pi}_{0}}-1\right ) \overline{z}_{k_{0}}\right ] \right\ } \end{array } \right]\end{aligned}\ ] ] where @xmath2650 we only give out the final expression for @xmath2651 without any technical details ; bias and variance properties of this estimator will be published elsewhere .
@xmath2652 { \displaystyle\prod\limits_{q=2}^{r-1 } } \left ( \frac{r_{0}r_{1}r_{2}}{\widehat{\pi}_{0}\widehat{\pi}_{1}\widehat{\pi } _ { 2}}\overline{v}_{k_{2}}\overline{v}_{k_{2}}^{t}-i\right ) _ { i_{q}}\left [ \overline{v}_{k_{2}}\frac{r_{0}r_{1}}{\widehat{\pi}_{0}\widehat{\pi}_{1}}\left ( \frac{r_{2}}{\widehat{\pi}_{2}}-1\right ) \right ] _ { i_{r}}\\ & + { \displaystyle\sum\limits_{m=2}^{r-1 } } \left\ { \begin{array } [ c]{c}\begin{array } [ c]{c}\left [ \frac{r_{0}}{\widehat{\pi}_{0}}\left ( \frac{r_{1}}{\widehat{\pi}_{1}}-1\right ) \right ] _ { i_{1}}\overline{w}_{k_{1}}^{t}{\displaystyle\prod\limits_{q=2}^{m-1 } } \left ( \frac{r_{0}r_{1}}{\widehat{\pi}_{0}\widehat{\pi}_{1}}\overline { w}_{k_{1}}\overline{w}_{k_{1}}^{t}-i\right ) _ { i_{q}}\times\\ \left [ \frac{r_{0}r_{1}r_{2}}{\widehat{\pi}_{0}\widehat{\pi}_{1}\widehat{\pi } _ { 2}}\left ( y-\widehat{b}_{2}\right ) \overline{w}_{k_{1}}\overline { v}_{k_{2}}^{t}\right ] _ { i_{m}}{\displaystyle\prod\limits_{s = m+1}^{r-1 } } \left ( \frac{r_{0}r_{1}r_{2}}{\widehat{\pi}_{0}\widehat{\pi}_{1}\widehat{\pi } _ { 2}}\overline{v}_{k_{2}}\overline{v}_{k_{2}}^{t}-i\right ) _ { i_{s}}\times\\ \left [ \overline{v}_{k_{2}}\frac{r_{0}r_{1}}{\widehat{\pi}_{0}\widehat{\pi } _ { 1}}\left ( \frac{r_{2}}{\widehat{\pi}_{2}}-1\right ) \right ] _ { i_{r}}\end{array } \\ + { \displaystyle\sum\limits_{j=2}^{m-1 } } \left [ \begin{array } [ c]{c}\left ( \frac{r_{0}}{\widehat{\pi}_{0}}-1\right ) _ { i_{1}}\overline{z}_{k_{0},i_{1}}^{t}{\displaystyle\prod\limits_{s=2}^{j-1 } } \left ( \frac{r_{0}}{\widehat{\pi}_{0}}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ) _ { i_{s}}\left [ \frac{r_{0}r_{1}r_{2}}{\widehat{\pi } _ { 0}\widehat{\pi}_{1}\widehat{\pi}_{2}}\left ( y-\widehat{b}_{2}\right ) \overline{z}_{k_{0}}\overline{v}_{k_{2}}^{t}\right ] _ { i_{j}}\\ \times{\displaystyle\prod\limits_{q = j+1}^{r+j - m-1 } } \left ( \frac{r_{0}r_{1}r_{2}}{\widehat{\pi}_{0}\widehat{\pi}_{1}\widehat{\pi } _ { 2}}\overline{v}_{k_{2}}\overline{v}_{k_{2}}^{t}-i\right ) _ { i_{q}}\left [ \overline{v}_{k_{2}}\frac{r_{0}r_{1}}{\widehat{\pi}_{0}\widehat{\pi}_{1}}\left ( \frac{r_{2}}{\widehat{\pi}_{2}}-1\right ) \overline{w}_{k_{1}}^{t}\right ] _ { i_{r+j - m}}\times\\{\displaystyle\prod\limits_{q = r+j - m+1}^{r-1 } } \left ( \frac{r_{0}r_{1}}{\widehat{\pi}_{0}\widehat{\pi}_{1}}\overline { w}_{k_{1}}\overline{w}_{k_{1}}^{t}-i\right ) _ { i_{q}}\left [ \overline { w}_{k_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\left ( \frac{r_{1}}{\widehat{\pi } _ { 1}}-1\right ) \right ] _ { i_{r}}\end{array } \right ] \\ + \left ( \frac{r_{0}}{\widehat{\pi}_{0}}-1\right ) _ { i_{1}}\overline{z}_{k_{0},i_{1}}^{t}{\displaystyle\prod\limits_{s=2}^{m-1 } } \left ( \frac{r_{0}}{\widehat{\pi}_{0}}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ) _ { i_{s}}\left [ \overline{z}_{k_{0}}\frac{r_{0}r_{1}r_{2}}{\widehat{\pi}_{0}\widehat{\pi}_{1}\widehat{\pi}_{2}}\left ( y-\widehat{b}_{2}\right ) \overline{v}_{k_{2}}^{t}\right ] _ { i_{m}}\times\\{\displaystyle\prod\limits_{q = m+1}^{r-1 } } \left ( \frac{r_{0}r_{1}r_{2}}{\widehat{\pi}_{0}\widehat{\pi}_{1}\widehat{\pi } _ { 2}}\overline{v}_{k_{2}}\overline{v}_{k_{2}}^{t}-i\right ) \left [ \overline{v}_{k_{2}}\frac{r_{0}r_{1}}{\widehat{\pi}_{0}\widehat{\pi}_{1}}\left ( \frac{r_{2}}{\widehat{\pi}_{2}}-1\right ) \right ] _ { i_{r}}\end{array } \right\}\end{aligned}\]]@xmath2653{c}\left [ \begin{array } [ c]{c}\frac{r_{0}r_{1}r_{2}}{\widehat{\pi}_{0}\widehat{\pi}_{1}\widehat{\pi}_{2}}\left ( y-\widehat{b}_{2}\right ) + \\ \frac{r_{0}r_{1}}{\widehat{\pi}_{0}\widehat{\pi}_{1}}\left ( \widehat{b}_{2}-\widehat{b}_{1}\right ) \end{array } \right ] _ { i_{1}}\overline{w}_{k_{1},i_{1}}^{t}{\displaystyle\prod\limits_{q=2}^{r-1 } } \left ( \frac{r_{0}r_{1}}{\widehat{\pi}_{0}\widehat{\pi}_{1}}\overline { w}_{k_{1}}\overline{w}_{k_{1}}^{t}-i\right ) _ { i_{q}}\left [ \overline { w}_{k_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\left ( \frac{r_{1}}{\widehat{\pi } _ { 1}}-1\right ) \right ] _ { i_{r}}\\ + { \displaystyle\sum\limits_{j=2}^{r-1 } } \left ( \frac{r_{0}}{\widehat{\pi}_{0}}-1\right ) _ { i_{1}}\overline{z}_{k_{0},i_{1}}^{t}{\displaystyle\prod\limits_{s=2}^{j-1 } } \left ( \frac{r_{0}}{\widehat{\pi}_{0}}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ) _ { i_{s}}\left [ \left ( \begin{array } [ c]{c}\frac{r_{0}r_{1}r_{2}}{\widehat{\pi}_{0}\widehat{\pi}_{1}\widehat{\pi}_{2}}\left ( y-\widehat{b}_{2}\right ) \\ + \frac{r_{0}r_{1}}{\widehat{\pi}_{0}\widehat{\pi}_{1}}\left ( \widehat{b}_{2}-\widehat{b}_{1}\right ) \end{array } \right ) \overline{z}_{k_{0}}\overline{w}_{k_{1}}^{t}\right ] _ { i_{j}}\times\\{\displaystyle\prod\limits_{q = j+1}^{r-1 } } \left ( \frac{r_{0}r_{1}}{\widehat{\pi}_{0}\widehat{\pi}_{1}}\overline { w}_{k_{1}}\overline{w}_{k_{1}}^{t}-i\right ) _ { i_{q}}\left [ \overline { w}_{k_{1}}\frac{r_{0}}{\widehat{\pi}_{0}}\left ( \frac{r_{1}}{\widehat{\pi } _ { 1}}-1\right ) \right ] _ { i_{r}}\\ + \left [ \left ( \begin{array } [ c]{c}\frac{r_{0}r_{1}r_{2}}{\widehat{\pi}_{0}\widehat{\pi}_{1}\widehat{\pi}_{2}}\left ( y-\widehat{b}_{2}\right ) \\ + \frac{r_{0}r_{1}}{\widehat{\pi}_{0}\widehat{\pi}_{1}}\left ( \widehat{b}_{2}-\widehat{b}_{1}\right ) \\ + \frac{r_{0}}{\widehat{\pi}_{0}}\left ( \widehat{b}_{1}-\widehat{b}_{0}\right ) \end{array } \right ) \right ] _ { i_{1}}\overline{z}_{k_{0},i_{1}}^{t}{\displaystyle\prod\limits_{s=2}^{r-1 } } \left ( \frac{r_{0}}{\widehat{\pi}_{0}}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ) _ { i_{s}}\overline{z}_{k_{0},i_{r}}\left ( \frac{r_{0}}{\widehat{\pi}_{0}}-1\right ) _ { i_{r}}\end{array } \right\}\ ] ] 99 arellano m. ( 2003 ) " _ panel data econometrics " .
_ oxford university press : advanced texts in econometrics .
bhattacharyya a. ( 1947 ) `` on some analogues of the amount of information and their use in statistical estimation ii - iii '' _ sankhy _ , vol.8 , 3 , 201218 .
bickel p. , klassen c. , ritov y. , wellner j. ( 1993 ) `` _ efficient and adaptive estimation for semiparametric models _ '' .
springer .
bickel p. , and ritov y. ( 2003 ) `` nonparametric estimators which can be ' ' plugged - in " .
_ annals of statist .
_ 31(4 ) , 103353 . birge l. , massart p.(1995 ) `` estimation of integral functionals of a density '' .
_ annals of statistics_. 23(1 ) , 11 - 29 .
he , x. and shao , q.m .
`` on parameters of increasing dimension '' .
_ journal of multivariate analysis _ , 73(1 ) , 120 - 135 .
klassen c. ( 1987 ) `` consistent estimation of the influence function of locally asymptotically linear estimators '' .
_ annals of statistics .
_ 15(4 ) , 1548 - 62 . lee a.j .
( 1990 ) `` _ _ u - statistics : theory and practice _ _ '' .
marcel dekker , new york .
lindsay r , waterman b. ( 1996 ) `` projected score methods for approximating conditional scores '' .
_ biometrika _ 83(1):1 - 13
. li l. , tchetgen e. , van der vaart aw , and robins jm .
( 2006 ) `` robust inference with higher order inference functions : part ii . ''
2005 _ jsm proceedings . _ american statistical associations .
2558 - 2565 .
mallat sg .
( 1998 ) `` _ _ a wavelet tour of signal processing _ _ '' .
acedemic press .
newey w. , hsieh f. , and robins jm .
( 2004 ) `` twicing kernels and a small bias property of semiparametric estimators '' _ econometrica _ 72 , 947 - 962 .
pfanzagl j. ( 1990 ) `` _ _ estimation in semiparametric models : some recent developments _ _ '' .
lecture notes in statistics 31 , springer - verlag , berlin 1985 , 505 pages portnoy s. ( 1988 ) .
`` asymptotic behavior of likelihood methods for exponential families when the number of parameters tends to infinity '' .
_ annals of statistics .
_ 16 , 356 - 366 pyke r. , `` spacings ( with discussion ) '' , _
b _ 27 ( 1965 ) , 395449 .
ritov y. , bickel p. ( 1990 ) `` achieving information bounds in non- and semi - parametric models '' . _ annals of statistics .
_ 18 , 925 - 938 .
robins j. , ritov , y ( 1997 ) .
`` toward a curse - of - dimensionality appropriate ( coda ) asymptotic theory for semiparametric models '' . _ statistics in medicine .
_ 16 285 - 319 .
robins jm .
( 2004 ) `` optimal structural nested models for optimal sequential decisions '' . in dy lin and p. heagerty ( eds . ) _ proceedings of the second seattle symposium in biostatistics_. new york springer .
robins jm , rotnitzky a. ( 2001).comment on the bickel and kwon article , `` inference for semiparametric models : some questions and an answer statistica sinica , 11(4):920 - 936 . [ ' ' on double robustness . " ] robins j.m , li l , tchetgen eric , van der vaart aw ( 2007 ) , `` asymptotic normality of degenerate u - statistics '' . working paper .
robins j.m , van der vaart aw .
( 2006 ) `` adaptive nonparametric confidence sets '' . _
annals of statistics_. 34(1):229 - 253 .
robins jm , li l , tchetgen e , van der vaart a. ( 2008 ) . higher order influence functions and minimax estimation of nonlinear functionals . probability and statistics :
essays in honor of david a. freedman 2:335 - 421 .
small d. , mcleish c.(1994 ) `` _ _ hilbert space methods in probability and statistical inference _ _ '' .
tchetgen e. , li l. , van der vaart aw , and robins jm .
( 2006 ) `` robust inference with higher order inference functions : part i. '' 2005 _ jsm proceedings .
_ american statistical associations .
2644 - 2651 .
tchetgen e. , li l. , van der vaart aw , and robins jm .
( 2007 ) `` higher order u - statistics estimators for longitudinal missing data and causal inference models '' .
working paper van der laan m and dudoit s.(2003 ) .
asymptotics of cross - validated risk estimation in estimator selection and performance assessment . technical report .
van der laan m , and robins jm .
( 2003 ) " _ unified methods for censored longitudinal data and causality " _
springer series in statistics .
wang , l. , brown , l. d. , cai , t. levine , m. ( 2006 ) .
`` effect of mean on variance function estimation in nonparametric regression '' .
technical report .
cai , t. , levine , m. wang , l. ( 2006 ) `` variance function estimation in multivariate nonparametric regression '' . technical report .
van der vaart , aw ( 1991 ) .
`` on differentiable functionals '' .
statist_. 19 178 - 204 .
van der vaart , aw ( 1998 ) .
`` _ _ asymptotic statistics _ _ '' .
cambridge series in statistical and probabilistic mathematics .
in the following , we assume all parametric submodels are sufficiently smooth and regular that expectation and differentiation operators commute as needed
. we also define @xmath2654 to be @xmath2655 ( theorem [ eiet ] ) define the bias function @xmath2656 $ ] of @xmath272 to be @xmath2657 .$ ] @xmath70 define @xmath2658 = \partial^{s}b_{m}\left [ \widetilde{\theta}\left ( \varsigma^{\ast}\right ) , \widetilde{\theta}\left ( \varsigma\right ) \right ] /\partial\varsigma _ { l_{1} ... }^{\ast}\partial\varsigma_{l_{j}}^{\ast}\partial\varsigma _ { l_{j+1} ... }\partial\varsigma_{l_{s}}|_{\varsigma^{\ast}=\widetilde{\theta } ^{-1}\left\ { \theta\right\ } , \varsigma=\widetilde{\theta}^{-1}\left\ { \theta\right\ } } \ ] ] where we reserve @xmath2659 for differentiation with respect to the first argument of @xmath2660 .$ ] thus for @xmath281 @xmath2661\ ] ] to prove the theorem we will first need to show that : @xmath2658 = 0\text { for } m\geq s > j>0\label{interinfo0}\ ] ] to this end note that for @xmath2662 @xmath2663 /\partial\varsigma_{l_{j+1}}\\ & = b_{m , l_{1}^{\ast} ... l_{j}^{\ast}l_{j+1}^{\ast}}\left [ \theta , \theta\right ] + b_{m , l_{1}^{\ast} ... l_{j}^{\ast}l_{j+1}}\left [ \theta , \theta\right ] \\ & = \psi_{\backslash l_{1} ... l_{j+1}}\left ( \theta\right ) + b_{m , l_{1}^{\ast } ... l_{j}^{\ast}l_{j+1}}\left [ \theta,\theta\right]\end{aligned}\ ] ] where the 2nd equality is by the definition of @xmath341 , the third is by the chain rule , and the fourth is again by the definition of @xmath272 .
hence @xmath2664 = 0.$ ] hence for @xmath2665 @xmath2666 /\partial\varsigma_{l_{j+2} .. }=b_{m , l_{1}^{\ast} ...
l_{j}^{\ast}l_{j+2}^{\ast}l_{j+1}}\left [ \theta,\theta\right ] + b_{m , l_{1}^{\ast } ... l_{j}^{\ast}l_{j+1}l_{j+2}}\left [ \theta,\theta\right ] \\ & = 0+b_{m ,
l_{1}^{\ast} ... l_{j}^{\ast}l_{j+1}l_{j+2}}\left [ \theta , \theta\right]\end{aligned}\ ] ] @xmath70where the last equality holds because we just proved @xmath2667 = 0 $ ] for arbitrary indices . iterating this argument proves @xmath2668 .
we complete the proof by induction on @xmath212 for some @xmath2669 given a @xmath2670 dimensional regular parametric submodel @xmath294 , @xmath2671 = 0 $ ] by assumption .
hence , by regularity of the model , @xmath2672 + b_{m , l_{1}.}\left [ \theta,\theta\right ] .$ ] therefore @xmath2673 = -\psi_{\backslash l_{1}}\left ( \theta\right ) .$ ] now suppose the theorem if true for @xmath260 then @xmath2674 /\partial\varsigma_{l_{s+1}.}\\ & = b_{m , l_{s+1}^{\ast}l_{1} ... l_{s}}\left [ \theta,\theta\right ] + b_{m , l_{1} ... l_{s+1}}\left [ \theta,\theta\right ] = 0+b_{m , l_{1} ... l_{s+1}}\left [ \theta,\theta\right ] \\end{aligned}\ ] ] where the second equality is by the induction assumption , the third by the chain rule , and the last by equation @xmath2668 ( theorem [ eift ] ) ( 1 ) : consider two influence functions @xmath2675 and @xmath2676 for @xmath988 then @xmath2677 = \psi_{\backslash\overline{l}_{s}}\left ( \theta\right ) -\psi_{\backslash\overline{l}_{s}}\left ( \theta\right ) = 0\ $ ] for any score @xmath2678 and hence for any linear combination of scores .
but , by definition , linear combinations of scores are dense in @xmath2679 thus @xmath2680 and @xmath2681 have the same projection on @xmath2682 ( 2 - 3 ) : essentially immediate from the definitions . ( 4 ) : for @xmath2683 = e_{\theta}\left [ \pi_{m,\theta } \left [ \mathbb{if}_{m}\left ( \theta\right ) |\mathcal{u}_{t}\left ( \theta\right ) \right ] \widetilde{\mathbb{s}}_{t,\overline{l}_{t}}\left ( \theta\right ) \right ] $ ] for any @xmath2684 .
( 5.a ) : follows from ( 1 ) .
( 5.b ) : follows from ( 4 ) .
degeneracy of @xmath314 follows at once from the fact that @xmath2685 in @xmath225 . proof of part ( 5.c ) requires the following .
[ joel2]suppose , for @xmath1134 @xmath2686 and @xmath2687 exist w.p.1 for a kernel @xmath2688 then , ( i):@xmath2689@xmath2690 and @xmath2691 each have the same mean given @xmath2692 , ( ii ) @xmath2693 = 0,$ ] so @xmath2694 ( i):by @xmath2698 degenerate@xmath90@xmath2699 .$ ] further , by definition , @xmath2700 \\ & = e_{\theta}\left [ if_{m , m,\backslash l_{t},\overline{i}_{m}}\left ( \theta\right ) |o_{i_{1}}, ... ,o_{i_{m}}\right ] .\end{aligned}\ ] ] ( ii ) : by @xmath2698 degenerate @xmath2701 $ ] wp1 and so ( ii ) follows from ( i).(iii ) : ( i ) and ( ii ) imply @xmath2702 \\ & = e_{\theta}\left [ \begin{array } [ c]{c}e_{\theta}\left\ { if_{1,if_{m\ , m\ } \left ( o_{i_{1},} .. o_{i_{m}};\cdot\right ) } \left ( o_{i_{m+1}};\theta\right ) |\left ( o_{i_{m+1}},o_{i_{1}}, ... ,o_{i_{m-2}}\right ) \right\ } \\ \times s_{l_{t}}\left ( o_{i_{m+1}}\right ) |o_{i_{1}}, ... ,o_{i_{m-2}}\end{array } \right]\end{aligned}\ ] ] but , by @xmath2703 an arbitrary mean zero function , @xmath2704 ( iv ) : by definition , @xmath2705 = \mathbb{v}\left [ \left\ { i - d_{m,\theta}\right\ } \left\ { if_{m , m,\backslash l_{t},\overline{i}_{m}}\left ( \theta\right ) \right\ } \right ] .$ ] the result follows by eq .
@xmath2706 and part ( ii ) . * theorem 5c(ii ) : * consider a @xmath135-dimensional parametric submodel @xmath2707 @xmath2708 with @xmath2709 = 0.$ ] since this model is linear in the @xmath2710 @xmath2711 for @xmath1927 .
hence @xmath2712is degenerate of order @xmath135 , i.e. , @xmath2713 since @xmath2714 exists , on setting @xmath2715for @xmath2716 @xmath2717 .\ ] ] differentiating the last display with respect to @xmath2718 and evaluating at @xmath2719 , we obtain @xmath2720 + e_{\theta}\left [ \mathbb{if}_{m-1_{,}\backslash l_{m}}\left ( \theta\right ) \widetilde{\mathbb{s}}_{m-1,\overline{l}_{m-1}}\left ( \theta\right ) \right ] \\ & = e_{\theta}\left [ \mathbb{if}_{m-1_{,}\backslash l_{m}}\left ( \theta\right ) \widetilde{\mathbb{s}}_{m-1,\overline{l}_{m-1}}\left ( \theta\right ) \right]\end{aligned}\ ] ] now @xmath2721 $ ] @xmath2722 + e_{\theta}\left [ \mathbb{if}_{m-1_{,}m-1,\backslash l_{m}}\left ( \theta\right ) \widetilde{\mathbb{s}}_{m-1,\overline{l}_{m-1}}\left ( \theta\right ) \right ] .$ ] setting @xmath2723 @xmath2724 is degenerate of order @xmath1324 so @xmath2725 \\ & = \left ( m-1\right ) ! e_{\theta}\left ( \left [ if_{m-1,m-1}^{sym},_{\backslash l_{m}}\left ( o_{i_{1},} .. o_{i_{m-1}};\theta\right ) \right ] \prod\limits_{r=1}^{m-1}a_{r}\left ( o_{i_{r}},\theta\right ) \right)\end{aligned}\ ] ] and
@xmath2726 = 0.$ ] hence @xmath2727 now , by the assumed existence of @xmath272 , we also have @xmath2728 = m!e_{\theta}\left ( if_{m , m}^{sym}\left ( o_{i_{1},} .. o_{i_{m}};\theta\right ) \prod\limits_{r=1}^{m}a_{r}\left ( o_{i_{r}},\theta\right ) \right ) $ ] .
it follows that , for any choice of @xmath1324 mean zero functions @xmath2729 under @xmath940 @xmath2730{c}if_{m-1,m-1}^{sym},_{\backslash l_{m}}\left ( o_{i_{1},} .. o_{i_{m-1}};\theta\right ) \\ -me_{\theta}\left [ if_{m , m}^{sym}\left ( o_{i_{1},} .. o_{i_{m}};\theta\right ) a_{m}\left ( o_{i_{m}},\theta\right ) |o_{i_{1},} .. o_{i_{m-1}}\right ] \end{array } \right\ } \times\prod\limits_{r=1}^{m-1}a_{r}\left ( o_{i_{r}},\theta\right ) \right ) \\ & = e_{\theta}\left ( r\left ( o_{i_{1},} .. o_{i_{m-1}};\theta\right ) \prod\limits_{r=1}^{m-1}a_{r}\left ( o_{i_{r}},\theta\right ) \right)\end{aligned}\ ] ] where @xmath2731 -me_{\theta}\left [ if_{m , m}^{sym}\left ( o_{i_{1},} .. o_{i_{m}};\theta\right ) a_{m}\left ( o_{i_{m}},\theta\right ) |o_{i_{1},} .. o_{i_{m-1}}\right]\end{aligned}\ ] ] the last equality follows from @xmath2732 $ ] orthogonal to @xmath2733 we conclude @xmath2734 with probability @xmath18 because @xmath2735 is a degenerate u - statistic kernel of order @xmath1324 and all degenerate u - statistics of order @xmath1324 have kernels that are the ( possibly infinite ) sum of products of @xmath1324 mean zero functions .
it follows that , on a set @xmath322 which has probability @xmath18 under @xmath2736 @xmath2737 \\ & + \left\ { i - d_{m-1,\theta}\right\ } \left [ if_{m-1,m-1}^{sym},_{\backslash l_{m}}\left ( o_{i_{1},} ... ,o_{i_{m-1}};\theta\right ) \right ] \\ & = e_{\theta}\left [ \left\ { \begin{array } [ c]{c}m\times if_{m , m}^{sym}\left ( o_{i_{1},} .. o_{i_{m-1}},o;\theta\right ) \\ -\sum_{j=1}^{m-1}if_{m-1,m-1}^{sym}\left ( o_{i_{1},} ... ,o_{i_{j-1}},o , o_{i_{j+1}}, ... ,o_{i_{m-1}};\theta\right ) \end{array } \right\ } a_{m}\left ( o,\theta\right ) \right]\end{aligned}\ ] ] since , by parts ( i ) and ( ii ) of the lemma [ joel2 ] and eq .
@xmath2739 \\ & = -e_{\theta}\left [ \sum_{j=1}^{m-1}if_{m-1,m-1}^{sym}\left ( o_{i_{1},} ... ,o_{i_{j-1}},o , o_{i_{j+1}}, ... ,o_{i_{m-1}};\theta\right ) a_{m}\left ( o,\theta\right ) \right]\end{aligned}\ ] ] here @xmath2740 is the identity operator .
now since the model @xmath2741 with @xmath2742 for @xmath308 has score @xmath2743 and such scores are dense in the subspace of l@xmath2744 with mean zero , it follows that @xmath2745 has influence function @xmath2746on the set @xmath322 . thus @xmath2747
.$ ] & e_{\theta}\left [ \mathbb{if}_{m,\backslash l_{m+1}}\left ( \theta\right ) \widetilde{\mathbb{s}}_{m,\overline{l}_{m}}\left ( \theta\right ) \right ] \nonumber\\ & = m!e_{\theta}\left ( if_{m , m,\backslash l_{m+1}}^{sym}\left ( o_{i_{1},} .. o_{i_{l_{m}}};\theta\right ) \prod\limits_{r=1}^{m}s_{l_{r}}\left ( o_{i_{r}},\theta\right ) \right ) \label{ccc}\ ] ] @xmath2754:by lemma [ joel2 ] and theorem 5c(ii)@xmath90 @xmath2755 \\ & = \mathbb{v}\left [ me_{\theta}\left ( if_{m , m,\overline{i}_{m}}^{sym}\left ( \theta\right ) s_{l_{t}}\left ( o_{i_{m}}\right ) |o_{i_{1}}, ..
,o_{i_{m-1}}\right ) \right ] \\ & = \mathbb{v}\left [ me_{\theta}\left ( m^{-1}d_{m,\theta}\left\ { if_{1,if_{m-1,m-1}^{sym}\left ( o_{i_{1},} .. o_{i_{m-1}};\cdot\right ) } \left ( o_{i_{m}};\theta\right ) \right\ } s_{l_{t}}\left ( o_{i_{m}}\right ) ( iii ) of lemma [ joel2 ] and eq . @xmath2756
the rhs is @xmath2757 \\ & -\mathbb{v}\left\ { e\left [ e\left [ \left ( m-1\right ) e\left [ if_{1,if_{m-1,m-1}^{sym}\left ( o_{i_{1},} .. o_{i_{m-1}};\cdot\right ) } \left ( o_{i_{m}};\theta\right ) |o_{i_{m}},o_{i_{1}}, .. ,o_{i_{m-2}}\right ] \right ] s_{l_{t}}\left ( o_{i_{m}}\right ) |o_{i_{1}}, .. ,o_{i_{m-1}}\right ] \right\ } \\ & = \mathbb{v}\left [ if_{m-1,m-1,\backslash l_{t}}^{sym}\left ( \theta\right ) \right ] -\mathbb{v}\left\ { \left ( m-1\right ) e_{\theta } \left [ if_{m-1,m-1,\backslash l_{t}}^{sym}\left ( \theta\right ) |o_{i_{1}}, ... ,o_{i_{m-2}}\right ] \right\}\end{aligned}\ ] ] on the other hand , by part ( iv ) of the lemma [ joel2 ] , @xmath2758 = \mathbb{v}\left [ if_{m-1,m-1,\backslash l_{t}}\left ( \theta\right ) \right ] -\mathbb{v}\left [ \left ( m-1\right ) e_{\theta}\left [ if_{m-1,m-1,\backslash l_{t},\overline{i}_{m-1}}^{sym}\left ( \theta\right ) |o_{i_{1}}, ... ,o_{i_{m-2}}\right ] \right ] .
$ ] @xmath2759 write @xmath2760 + \mathbb{if}_{1,\backslash l_{t}}\left ( \theta\right ) \right\ } \\ & + \sum_{j=2}^{m-1}\left\ { \pi\left [ \mathbb{if}_{j+1,j+1,\backslash l_{t}}\left ( \theta\right ) |\mathcal{u}_{j}\left ( \theta\right ) \right ] + \pi\left [ \mathbb{if}_{jj,\backslash l_{t}}\left ( \theta\right ) @xmath2763 @xmath2764 = e_{\theta}\left [ \pi\left [ \mathbb{if}_{m , m,\backslash l_{m+1}}\left ( \theta\right ) |\mathcal{u}_{m-1}^{\bot } \left ( \theta\right ) \right ] \widetilde{\mathbb{s}}_{m,\overline{l}_{m}}\left ( \theta\right ) \right ] $ ] by eq . @xmath2765
but the rhs of this equation is the rhs of eq .
@xmath2766 ( * theorem * * 5c(i ) ) : * by assumption @xmath2767 hence @xmath2768\ ] ] by eq . @xmath2769 and
the assumption @xmath2770 has an influence function , we obtain @xmath2771 \\ & = \left ( m-1\right ) !
e_{\theta}\left ( if_{1if_{m-1,m-1}^{sym}\left ( o_{i_{1},} .. o_{i_{m-1}};\cdot\right ) } \left ( o_{i_{m}},\theta\right ) s_{l_{m}}\left ( o_{i_{m}},\theta\right ) \prod\limits_{r=1}^{m-1}s_{l_{r}}\left ( o_{i_{r}},\theta\right ) \right ) .\end{aligned}\ ] ] we conclude that @xmath2772 exists and equals @xmath2773 .$ ] first we show that for any @xmath1564-dimensional parametric submodel @xmath255 @xmath2774 where @xmath2775 from eq ( [ interinfo0 ] ) we know that @xmath2776 for all @xmath2777 since @xmath310 is locally nonparametric , i.e. , @xmath2778 and @xmath2779 therefore we have @xmath2780 \ii ) if @xmath327 exists , we have @xmath2781 with @xmath2782 symmetric , such that@xmath2783 \mathbb{d}_{m}^{(\mathbb{s}_{m,\overline{l}_{m}})}\right ) \\ & + e_{\theta}\left ( \mathbb{v}\left ( me\left [ if_{m}\left ( o_{i_{1}}, ...
,o_{i_{m}};\theta\right ) |o_{i_{1}}, ... o_{i_{m-1}}\right ] ^{c}\right ) \mathbb{d}_{m-1}^{(\mathbb{s}_{m,\overline{l}_{m}})}\right ) \\ & + e\left ( \mathbb{if}_{m-2}\mathbb{s}_{m,\overline{l}_{m}}\right)\end{aligned}\ ] ] therefore , @xmath2784
^{c}$ ] wp1@xmath64 so that the lhs has an influence function since : @xmath2785 ^{c}\right\ } _ { \backslash l_{m}}\\ & = \left\ { e_{\theta}\left [ if_{m}\left ( o_{i_{1}}, ... ,o_{i_{m}};\theta\right ) |o_{i_{1}}, ... o_{i_{m-1}}\right ] \right\ } _ { \backslash l_{m}}\\ & + \sum_{t=0}^{m-2}\left ( -1\right ) ^{m-1-t}\sum_{\substack{i_{r_{1}}\neq
i_{r_{2}} .. \neq i_{r_{t } } \\\overline{i}_{r_{t}}\subset\overline{i}_{m-1 } } } e_{\theta}\left ( if_{m}\left ( o_{i_{1}},o_{i_{2}}, .. o_{i_{m}};\theta\right ) |o_{i_{r_{1}}},o_{i_{r_{2}}}, .. o_{i_{r_{t}}}\right ) _
{ \backslash l_{m}}\\ & = \left\ { e_{\theta}\left [ if_{m}\left ( o_{i_{1}}, ... ,o_{i_{m}};\theta\right ) s_{l_{m}}\left ( o_{i_{m}}\right ) |o_{i_{1}}, ... o_{i_{m-1}}\right ] \right\ } \\ & + \sum_{t=0}^{m-2}\left [ \begin{array } [ c]{c}\left ( -1\right ) ^{m-1-t}\left ( m - t\right ) \\
\sum_{\substack{i_{r_{1}}\neq i_{r_{2}} .. \neq i_{r_{t } } \\\overline{i}_{r_{t}}\subset\overline{i}_{m-1}}}e\left ( \begin{array } [ c]{c}e\left [ if_{m}\left ( o_{i_{1}}, ... ,o_{i_{m}};\theta\right ) |o_{i_{r_{1}}},o_{i_{t}}, .. o_{i_{t+1}}\right ] s_{l_{t+1}}\left ( o_{i_{t+1}}\right ) \\ which means @xmath2794{c}if_{1,if_{m-1,m-1,\theta}\left ( o_{i_{1},} ..
o_{i_{m-1}};\cdot\right ) } ^{c}\left ( o_{i_{m}},\theta\right ) \\ -if_{1,if_{m-1,m-1,\theta}\left ( o_{i_{1},}.o_{i_{m}}.o_{i_{m-1}};\cdot\right ) } ^{c}\left ( o_{i_{j}},\theta\right ) \end{array } \right ] { \textstyle\prod\limits_{r=1}^{m } } s_{l_{r}}\left ( o_{i_{r}},\theta\right ) \right ) = 0\\ \longleftrightarrow if_{1,if_{m-1,m-1,\theta}\left ( o_{i_{1},} .. o_{i_{m-1}};\cdot\right ) } ^{c}\left ( o_{i_{m}},\theta\right ) = if_{1,if_{m-1,m-1,\theta}\left ( o_{i_{1},}.o_{i_{m}}.o_{i_{m-1}};\cdot\right ) } ^{c}\left ( o_{i_{j}},\theta\right ) \text { w.p.1 } \ ] ] ( proof of eq . ( [ imm ] ) ) we have proved in the text the following results that will be used repeatedly throughout the proof:@xmath2797@xmath2798 ^{-\frac{1}{2}}k_{f_{x},\infty } \left ( x_{i_{1}},x_{i_{2}}\right ) e_{\theta}\left [ h_{1}|x_{i_{2}}\right ] ^{-\frac{1}{2}}\varepsilon_{b , i_{2}}\left ( \theta\right ) \\ & = -\frac{k_{leb,\infty}\left ( x_{i_{1}},x_{i_{2}}\right ) } { g\left ( x_{i_{1}}\right ) ^{\frac{1}{2}}g\left ( x_{i_{2}}\right ) ^{\frac{1}{2}}}\varepsilon_{b , i_{2}}\left ( \theta\right ) , \end{aligned}\]]@xmath2799 ^{-\frac{1}{2}}k_{f_{x},\infty } \left ( x_{i_{1}},x_{i_{2}}\right ) e_{\theta}\left [ h_{1}|x_{i_{2}}\right ] ^{-\frac{1}{2}}\varepsilon_{p , i_{2}}\left ( \theta\right ) \\ & = -\frac{k_{leb,\infty}\left ( x_{i_{1}},x_{i_{2}}\right ) } { g\left ( x_{i_{1}}\right ) ^{\frac{1}{2}}g\left ( x_{i_{2}}\right ) ^{\frac{1}{2}}}\varepsilon_{p , i_{2}}\left ( \theta\right ) , \end{aligned}\ ] ] and @xmath2800 in addition , by an analogous argument , we can also show that @xmath2801 } f_{x}\left ( x\right ) + e_{\theta}\left [ h_{1}|x\right ] if_{1,f_{x}\left ( x;\cdot\right ) } \\ & = \left ( h_{1}-e\left [ h_{1}|x\right ] \right ) k_{leb,\infty}\left ( x , x\right ) + e_{\theta}\left [ h_{1}|x\right ] \left ( k_{leb,\infty}\left ( x , x\right ) -f_{x}\left ( x\right ) \right ) \\ & = h_{1}k_{leb,\infty}\left ( x , x\right ) -g\left ( x\right)\end{aligned}\ ] ] \ii ) we now assume eq .
( [ imm ] ) holds for @xmath135 for some @xmath316 and shall prove it is also true for @xmath2450 @xmath70by assumption,@xmath866{c}\sum_{j=0}^{m-2}c(m , j)\times\\ \prod\limits_{s=1}^{j}\frac{h_{1,i_{s+1}}}{g\left ( x_{i_{s+1}}\right ) } k_{leb,\infty}\left ( x_{i_{s}},x_{i_{s+1}}\right ) \\
k_{leb,\infty}\left ( x_{i_{j+1}},x_{i_{m}}\right ) \end{array } \right ] g\left ( x_{i_{m}}\right ) ^{-\frac{1}{2}}\varepsilon_{p , i_{m}}\left ( \theta\right)\end{aligned}\ ] ] following the results from part 5c ) of theorem [ eift ] , @xmath70 @xmath2802 exists if and only if @xmath2803 exists . by the chain rule,@xmath2804{c}\left ( \begin{array } [ c]{c}h_{1,i_{1}}if_{1,\varepsilon_{b , i_{1}}\left ( \cdot\right ) } \left ( o_{i_{m+1}}\right ) \varepsilon_{p , i_{m}}\left ( \theta\right ) + \\ h_{1,i_{m}}\varepsilon_{b , i_{1}}\left ( \theta\right ) if_{1,}\varepsilon _ { p , i_{m}}\left ( \cdot\right ) \left ( o_{i_{m+1}}\right ) \end{array } \right ) g\left ( x_{i_{1}}\right ) ^{-\frac{1}{2}}g\left ( x_{i_{m}}\right ) ^{-\frac{1}{2}}\\ -\frac{1}{2}\varepsilon_{b , i_{1}}\left ( \theta\right ) \varepsilon_{p , i_{m}}\left ( \theta\right ) g\left ( x_{i_{1}}\right ) ^{-\frac{1}{2}}g\left ( x_{i_{m}}\right ) ^{-\frac{1}{2}}\times\\ \left [ \frac{if_{1,g\left ( x_{i_{1}}\right ) } \left ( o_{i_{m+1}}\right ) } { g\left ( x_{i_{1}}\right ) } + \frac{if_{1,g\left ( x_{i_{m}}\right ) } \left ( o_{i_{m+1}}\right ) } { g\left ( x_{i_{m}}\right ) } \right ] \end{array } \right\ } \times\\ & \left [ \begin{array } [ c]{c}\sum_{j=0}^{m-2}c(m , j)\times\\ \prod\limits_{s=1}^{j}\frac{h_{1,i_{s+1}}}{g\left ( x_{i_{s+1}}\right ) } k_{leb,\infty}\left ( x_{i_{s}},x_{i_{s+1}}\right ) \\ \times k_{leb,\infty}\left ( x_{i_{j+1}},x_{i_{m}}\right ) \end{array } \right ] \\ & -\varepsilon_{b , i_{1}}\left ( \theta\right ) \varepsilon_{p , i_{m}}\left ( \theta\right ) g\left ( x_{i_{1}}\right ) ^{-\frac{1}{2}}g\left ( x_{i_{m}}\right ) ^{-\frac{1}{2}}\sum_{j=0}^{m-2}c(m , j)\times\\ & \left [ \begin{array } [ c]{c}\sum_{t=1}^{j}\frac{h_{1,i_{t+1}}k_{leb,\infty}\left ( x_{i_{t}},x_{i_{t+1}}\right ) } { g^{2}\left ( x_{i_{t+1}}\right ) } if_{1,g\left ( x_{i_{t+1}}\right ) } \left ( o_{i_{m+1}}\right ) \times\\ \prod\limits_{s\neq t}\frac{h_{1,i_{s+1}}}{g\left ( x_{i_{s+1}}\right ) } k_{leb,\infty}\left ( x_{i_{s}},x_{i_{s+1}}\right ) k_{leb,\infty}\left ( x_{i_{j+1}},x_{i_{m}}\right ) \end{array } \right]\end{aligned}\]]@xmath906@xmath2805{c}-\frac{h_{1,i_{1}}k_{leb,\infty}\left ( x_{i_{1}},x_{i_{m+1}}\right ) } { g\left ( x_{i_{1}}\right ) g\left ( x_{i_{m+1}}\right ) ^{\frac{1}{2}}g\left ( x_{i_{m}}\right ) ^{\frac{1}{2}}}\varepsilon_{b , i_{m+1}}\left ( \theta\right ) \varepsilon_{p , i_{m}}\left ( \theta\right ) \\ -\varepsilon_{b , i_{1}}\left ( \theta\right ) \frac{h_{1,i_{m}}k_{leb,\infty } \left ( x_{i_{m}},x_{i_{m+1}}\right ) } { g\left ( x_{i_{m}}\right ) g\left ( x_{i_{1}}\right ) ^{\frac{1}{2}}g\left ( x_{i_{m+1}}\right ) ^{\frac{1}{2}}}\varepsilon_{p , i_{m+1}}\\ -\frac{1}{2}\varepsilon_{b , i_{1}}\left ( \theta\right ) \varepsilon_{p , i_{m}}\left ( \theta\right ) g\left ( x_{i_{1}}\right ) ^{-\frac{1}{2}}g\left ( x_{i_{m}}\right ) ^{-\frac{1}{2}}\times\\ \left [ \frac{h_{1,i_{m+1}}k_{leb,\infty}\left ( x_{i_{m+1}},x_{i_{1}}\right ) } { g\left ( x_{i_{1}}\right ) } + \frac{h_{1,i_{m+1}}k_{leb,\infty}\left ( x_{i_{m+1}},x_{i_{m}}\right ) } { g\left ( x_{i_{m}}\right ) } -2\right ] \end{array } \right\ } \\ \times & \left [ \begin{array } [ c]{c}\sum_{j=0}^{m-2}c(m , j)\times\\ \prod\limits_{s=1}^{j}\frac{h_{1,i_{s+1}}}{g\left ( x_{i_{s+1}}\right ) } k_{leb,\infty}\left ( x_{i_{s}},x_{i_{s+1}}\right ) \\ \times k_{leb,\infty}\left ( x_{i_{j+1}},x_{i_{m}}\right ) \end{array } \right ] \\ & -\varepsilon_{b , i_{1}}\left ( \theta\right ) \varepsilon_{p , i_{m}}\left ( \theta\right ) g\left ( x_{i_{1}}\right ) ^{-\frac{1}{2}}g\left ( x_{i_{m}}\right ) ^{-\frac{1}{2}}\sum_{j=0}^{m-2}c(m , j)\times\\ & \left [ \begin{array } [ c]{c}\sum_{t=1}^{j}\frac{h_{1,i_{t+1}}k_{leb,\infty}\left ( x_{i_{t}},x_{i_{t+1}}\right ) } { g^{2}\left ( x_{i_{t+1}}\right ) } \left ( h_{1,i_{m+1}}k_{leb,\infty}\left ( x_{i_{m+1}},x_{i_{t+1}}\right ) -g\left ( x_{i_{t+1}}\right ) \right ) \times\\ \prod\limits_{s\neq t}\frac{h_{1,i_{s+1}}}{g\left ( x_{i_{s+1}}\right ) } k_{leb,\infty}\left ( x_{i_{s}},x_{i_{s+1}}\right ) k_{leb,\infty}\left ( x_{i_{j+1}},x_{i_{m}}\right ) \end{array } \right]\end{aligned}\]]@xmath906 @xmath2806{c}\varepsilon_{b , i_{m+1}}\left ( \theta\right ) g\left ( x_{i_{m+1}}\right ) ^{-\frac{1}{2}}\frac{h_{1,i_{1}}k_{leb,\infty}\left ( x_{i_{m+1}},x_{i_{1}}\right ) } { g\left ( x_{i_{1}}\right ) } \\ \prod\limits_{s=1}^{j}\frac{h_{1,i_{s+1}}}{g\left ( x_{i_{s+1}}\right ) } k_{leb,\infty}\left ( x_{i_{s}},x_{i_{s+1}}\right ) k_{leb,\infty}\left ( x_{i_{j+1}},x_{i_{m}}\right ) \\
\times g\left ( x_{i_{m}}\right ) ^{-\frac{1}{2}}\varepsilon_{p , i_{m}}\left ( \theta\right ) \end{array } \right ) \\ & + \sum_{j=0}^{m-2}\left ( -1\right ) ^{j}\binom{m-2}{j}\left ( \begin{array } [ c]{c}\varepsilon_{b , i_{1}}\left ( \theta\right ) g\left ( x_{i_{1}}\right ) ^{-\frac{1}{2}}\prod\limits_{s=1}^{j}\frac{h_{1,i_{s+1}}}{g\left ( x_{i_{s+1}}\right ) } k_{leb,\infty}\left ( x_{i_{s}},x_{i_{s+1}}\right ) \\
\frac{h_{1,i_{m}}k_{leb,\infty}\left ( x_{i_{j+1}},x_{i_{m}}\right ) } { g\left ( x_{i_{m}}\right ) } k_{leb,\infty}\left ( x_{i_{m}},x_{i_{m+1}}\right ) ^{\frac{1}{2}}\\ \times g\left ( x_{i_{m+1}}\right ) ^{-\frac{1}{2}}\varepsilon_{p , i_{m+1}}\end{array } \right ) \\ & + \frac{1}{2}\sum_{j=0}^{m-2}\left ( -1\right ) ^{j}\binom{m-2}{j}\left\ { \begin{array } [ c]{c}\varepsilon_{b , i_{1}}\left ( \theta\right ) \varepsilon_{p , i_{m}}\left ( \theta\right ) \left ( \begin{array } [ c]{c}\frac{h_{1,i_{m+1}}k_{leb,\infty}\left ( x_{i_{m+1}},x_{i_{1}}\right ) } { g\left ( x_{i_{1}}\right ) }
\\ + \frac{h_{1,i_{m+1}}k_{leb,\infty}\left ( x_{i_{m+1}},x_{i_{j+1}}\right ) } { g\left ( x_{i_{m+1}}\right ) } \end{array } \right ) \\ \times\prod\limits_{s=1}^{j}\frac{h_{1,i_{s+1}}}{g\left ( x_{i_{s+1}}\right ) } k_{leb,\infty}\left ( x_{i_{s}},x_{i_{s+1}}\right ) \times\\ k_{leb,\infty}\left ( x_{i_{j+1}},x_{i_{m}}\right ) g\left ( x_{i_{1}}\right ) ^{-\frac{1}{2}}g\left ( x_{i_{m}}\right ) ^{-\frac{1}{2}}\end{array } \right\ } \\ & + if_{m , m,\psi,\overline{i}_{m}}\left ( \theta\right ) \\ & + \sum_{j=0}^{m-2}\left ( -1\right ) ^{j}\binom{m-2}{j}\left [ \sum _ { t=1}^{j}\left ( \begin{array } [ c]{c}\varepsilon_{b , i_{1}}\left ( \theta\right ) \varepsilon_{p , i_{m}}\left ( \theta\right ) g\left ( x_{i_{1}}\right ) ^{-\frac{1}{2}}g\left ( x_{i_{m}}\right ) ^{-\frac{1}{2}}\times\\ \frac{h_{1,i_{t+1}}k_{leb,\infty}\left ( x_{i_{t}},x_{i_{t+1}}\right ) } { g\left ( x_{i_{t+1}}\right ) } h_{1,i_{m+1}}\frac{k_{leb,\infty}\left ( x_{i_{t+1}},x_{i_{m+1}}\right ) } { g\left ( x_{i_{m+1}}\right ) } \\ \times\prod\limits_{s\neq t}\frac{h_{1,i_{s+1}}}{g\left ( x_{i_{s+1}}\right ) } k_{leb,\infty}\left ( x_{i_{s}},x_{i_{s+1}}\right ) k_{leb,\infty}\left ( x_{i_{j+1}},x_{i_{m}}\right ) \end{array } \right ) \right ] \\ & -\sum_{j=0}^{m-2}\left ( -1\right ) ^{j}\binom{m-2}{j}j\left ( \begin{array } [ c]{c}\varepsilon_{b , i_{1}}\left ( \theta\right ) \varepsilon_{p , i_{m}}\left ( \theta\right ) g\left ( x_{i_{1}}\right ) ^{-\frac{1}{2}}g\left ( x_{i_{m}}\right ) ^{-\frac{1}{2}}\times\\ \prod\limits_{s=1}^{j}\frac{h_{1,i_{s+1}}}{g\left ( x_{i_{s+1}}\right ) } k_{leb,\infty}\left ( x_{i_{s}},x_{i_{s+1}}\right ) k_{leb,\infty}\left ( x_{i_{j+1}},x_{i_{m}}\right ) \end{array } \right)\end{aligned}\ ] ] applying the operator @xmath2807 which is defined in eq@xmath2808 on the statistic above , it is straightforward to show that @xmath2809 \right\ } \right )
\\ = \mathbb{v}\left\ { d_{m+1,\theta}\left [ \begin{array } [ c]{c}\varepsilon_{b , i_{1}}\left ( \theta\right ) g\left ( x_{i_{1}}\right ) ^{-\frac{1}{2}}\prod\limits_{s=1}^{m-1}\frac{h_{1,i_{s+1}}}{g\left ( x_{i_{s+1}}\right ) } k_{leb,\infty}\left ( x_{i_{s}},x_{i_{s+1}}\right ) \\
\times k_{leb,\infty}\left ( x_{i_{m}},x_{i_{m+1}}\right ) g\left ( x_{i_{m+1}}\right ) ^{-\frac{1}{2}}\varepsilon_{p , i_{m+1}}\left ( \theta\right ) \end{array } \right ] \right\ } \\
= \mathbb{v}\left\ { \begin{array } [ c]{c}\varepsilon_{b , i_{1}}\left ( \theta\right ) g\left ( x_{i_{1}}\right ) ^{-\frac{1}{2}}\left [ \begin{array } [ c]{c}\sum_{j=0}^{m-1}c(m+1,j)\times\\ \prod\limits_{s=1}^{j}\frac{h_{1,i_{s+1}}}{g\left ( x_{i_{s+1}}\right ) } k_{leb,\infty}\left ( x_{i_{s}},x_{i_{s+1}}\right ) \\ \times k_{leb,\infty}\left ( x_{i_{j+1}},x_{i_{m+1}}\right ) \end{array } \right ] \\
\times g\left ( x_{i_{m+1}}\right ) ^{-\frac{1}{2}}\varepsilon_{p , i_{m+1}}\left ( \theta\right ) \end{array } \right\}\end{gathered}\ ] ] ( theorem [ tbformula ] ) by eq ( [ eta1 ] ) and eq ( [ alpha1 ] ) and part a of the preceding lemma @xmath2810 ^{-1}e\left [ \overline{z}_{k}\dot { p}\left ( \widehat{b}-b\right ) h_{1}\right ] \\
\widetilde{\overline{\alpha}}_{k } & = -e\ \left [ \dot{b}\dot{p}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] ^{-1}e\ \left [ \overline { z}_{k}\dot{b}\left ( \widehat{p}-p\right ) h_{1}\right]\end{aligned}\ ] ] and hence @xmath2811 ^{-1}e\left [ \dot{p}\dot{b}\overline{z}_{k}\left ( b-\widehat{b}\ \right ) \left\ { \dot{b}\right\ } ^{-1}h_{1}\right ] \\ \widetilde{p}-\widehat{p } & = -\dot{p}\overline{z}_{k}^{t}e_{\ } \left [ \dot{b}\dot{p}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] ^{-1}e\left [ \dot{b}\dot{p}\overline{z}_{k}\left ( p-\widehat{p}\right ) \left\ { \dot { p}\right\ } ^{-1}h_{1}\right]\end{aligned}\ ] ] thus @xmath2812 ^{-1}e\left [ q^{2}\overline{z}_{k}\frac{\left ( p-\widehat{p}\right ) } { \dot{p}}\right ] \\ & = \pi\left [ \frac{p-\widehat{p}}{\dot{p}}q|q\overline{z}_{k}\right ] \\ q\frac{\widetilde{b}-\widehat{b}}{\dot{b } } & = -q\overline{z}_{k}^{t}e_{\theta}\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] ^{-1}e\left [ q^{2}\overline{z}_{k}\frac{\left ( b-\widehat{b}\right ) } { \dot{b}}\right ] \\ & = \pi\left [ \left ( \frac{b-\widehat{b}}{\dot{b}}\right ) q|q\overline { z}_{k}\right]\end{aligned}\ ] ] and hence @xmath2813 \\ q\left ( \frac{b-\widetilde{b}}{\dot{b}}\right ) & = \pi^{\perp}\left [ \left ( \frac{b-\widehat{b}}{\dot{b}}\right ) q|q\overline{z}_{k}\right]\end{aligned}\ ] ] but by the previous theorem , @xmath2814 = e\left [ q\left ( \frac{b-\widetilde{b}}{\dot{b}}\right ) q\left ( \frac{\left ( p-\widetilde{p}\right ) } { \dot{p}}\right ) \right]\ ] ] proving the theorem .
( theorem [ tbrate ] ) under our assumptions , the following holds uniformly for @xmath357@xmath2815 \pi^{\perp } \left [ \left ( \frac{b-\widehat{b}}{\dot{b}}\right ) q|q\overline{z}_{k}\right ] \right ] \right\ } ^{2}\\ & \leq e\left\ { \pi^{\perp}\left [ \left ( \frac{p-\widehat{p}}{\dot{p}}\right ) q|q\overline{z}_{k}\right ] ^{2}\right\ } \times e\left\ { \pi^{\perp}\left [ \left ( \frac{b-\widehat{b}}{\dot{b}}\right ) q|q\overline{z}_{k}\right ] ^{2}\right\}\end{aligned}\ ] ] by cauchy shwartz .
now @xmath2816 ^{2}\right\ } \\ & = inf_{\varsigma_{_{l}}}\int_{r^{d}}q^{2}\left ( \frac{p\left ( x\right ) -\widehat{p}\left ( x\right ) } { \dot{p}\left ( x\right ) } -\sum_{l=1}^{k}\varsigma_{_{l}}z_{l}\left ( x\right ) \right ) ^{2}f\left ( x\right ) dx\\ & = inf_{\varsigma_{_{l}}}\int_{r^{d}}q^{2}\left ( \frac{\left ( p\left ( x\right ) -\widehat{p}\left ( x\right ) \right ) } { \dot{p}\left ( x\right ) } -\sum_{l=1}^{k}\varsigma_{_{l}}^{\ast}\varphi_{l}\left ( x\right ) \right ) ^{2}q^{2}f\left ( x\right ) dx\\ & = inf_{\varsigma_{_{l}}}\int_{r^{d}}\left ( \left ( p\left ( x\right ) -\widehat{p}\left ( x\right ) \right ) -\sum_{l=1}^{k}\varsigma_{_{l}}^{\ast } \varphi_{l}\left ( x\right ) \right ) ^{22}q^{2}\ f\left ( x\right ) dx\\ & \leq\left\vert \left\vert q^{2}f\left ( x\right ) \right\vert \right\vert
_ { \infty}inf_{\varsigma_{_{l}}}\int_{r^{d}}\left ( \left ( p\left ( x\right ) -\widehat{p}\left ( x\right ) \right ) -\sum_{l=1}^{k}\varsigma_{_{l}}^{\ast } \varphi_{l}\left ( x\right ) \right ) ^{2}dx\\ & \leq\left\vert \left\vert q^{2}f\left ( x\right ) \right\vert \right\vert
_ { \infty}o_{p}\left ( k^{-2\beta_{p}/d}\right ) = o_{p}\left ( k^{-2\beta _ { p}/d}\right)\end{aligned}\ ] ] the last equality follows from the fact that under the stated assumptions @xmath2817 .
similarly @xmath2818 ^{2}\right\ } = o_{p}\left ( k^{-2\beta_{b}/d}\right ) .$ ] ( theorem [ drhoif ] ) by theorem [ foif ] @xmath2819 @xmath2820and by part 5.c of theorem [ eift ] @xmath2821 = \frac{1}{2}\left\ { \pi_{\theta}\left [ \mathbb{v}\left [ if_{1,if_{1,\widetilde{\psi}_{k}}\left ( o_{i_{1}},\cdot\right ) , i_{2}}\,\left ( \theta\right ) \right ] |\mathcal{u}_{1}^{\perp_{\theta,2}}\left ( \theta\right ) \right ] \right\ } .\ ] ] now @xmath2822 where @xmath2823@xmath2824 \right\ } ^{-1}\left [ \left\ { h_{1}\widetilde{b}\left ( x,\theta\right ) + h_{3}\right\ } \dot{p}\overline{z}_{k}\right ] _ { i_{2}},\end{aligned}\ ] ] and @xmath2825 \right\ } ^{-1}\left [ \left\ { h_{1}\widetilde{p}\left ( x,\theta\right ) + h_{2}\right\ } \dot{b}\overline{z}_{k}\right ] _ { i_{2}}.\]]@xmath2826 \right\ } ^{-1}\\ & \times\left [ \left\ { h_{1}\widetilde{b}\left ( x,\theta\right ) + h_{3}\right\ } \dot{p}\overline{z}_{k}\right ] _ { i_{2}}\\ & -\left\ { h_{1}\widetilde{b}\left ( x,\theta\right ) + h_{3}\right\ } _ { i_{1}}\dot{p}_{i_{1}}\overline{z}_{ki_{1}}^{t}\left\ { e_{\theta}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\\ & \times\left [ \left\ { h_{1}\widetilde{p}\left ( x,\theta\right ) + h_{2}\right\ } \dot{b}\overline{z}_{k}\right ] _ { i_{2}}\ ] ] and further @xmath2827 @xmath2828 = e_{\theta}\left [ \left\ { h_{1}\widetilde{b}\left ( x,\theta\right ) + h_{3}\right\ } \dot{p}\overline { z}_{k}\right ] = 0\ ] ] and thus @xmath2829 is degenerate .
because @xmath2830 has two terms , it appears that @xmath2831 will consist of two terms .
however by the symmetry upon interchange of @xmath821 and @xmath2832 and the permutation invariance of the operator @xmath2833 @xmath2834 \\ & = \mathbb{v}\left [ \begin{array } [ c]{c}-2\left\ { h_{1}\widetilde{p}\left ( x,\theta\right ) + h_{2}\right\ } _ { i_{1}}\dot{b}_{i_{1}}\overline{z}_{ki_{1}}^{t}\left\ { e_{\theta}\left [ \dot { p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\\ \times\left [ \overline{z}_{k}\left\ { h_{1}\widetilde{b}\left ( x,\theta\right ) + h_{3}\right\ } \dot{p}\right ] _ { i_{2}}\end{array } \right]\end{aligned}\ ] ] thus we can take @xmath2835 \right\ } ^{-1}\\ & \times\left [ \overline{z}_{k}\left\ { h_{1}\widetilde{b}\left ( x,\theta\right ) + h_{3}\right\ } \dot{p}\right ] _ { i_{2}}\ ] ] as was to be proved .
we now complete the proof of the theorem by induction .
we assume it is true for @xmath2836 and prove it is true for @xmath2837
now@xmath2838 = \frac{1}{m}\mathbb{v}\left [ \pi_{\theta}\left [ if_{1,if_{mm,\widetilde{\psi}_{k},}\,\left ( o_{\overline{i}_{m}},\cdot\right ) , i_{m+1}}\,\left ( \theta\right ) |\mathcal{u}_{m}^{\perp_{\theta , m+1}}\left ( \theta\right ) \right ] \right]\ ] ] now by the induction hypothesis , @xmath2839 _ { i_{1}}\\ & \times\left [ { \displaystyle\prod\limits_{s=3}^{m } } \left\ { e_{\theta}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\left\ { \begin{array } [ c]{c}\left ( \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{s}}\\ -e_{\theta}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \end{array } \right\ } \right ] \\ & \times\left\ { e_{\theta}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\left [ \overline{z}_{k}\left ( h_{1}\widetilde{b}\left ( \theta\right ) + h_{3}\right ) \dot { p}\right ] _ { i_{2}}\ ] ] the derivatives with respect to the @xmath2840 in @xmath2841and in the @xmath1324 terms @xmath2842 \right\ } ^{-1}$ ] will each contribute a term to @xmath2843 .$ ] however differentiating wrt to the @xmath179 in the @xmath1293 terms @xmath2844 $ ] will not contribute to @xmath2845 $ ] as the contribution from each of these @xmath1293 terms to @xmath2846 is only a function of @xmath135 units data and is thus an element of @xmath209 which is orthogonal to the space @xmath2847 that is projected on .. now @xmath2848 \right\ } ^{-1},i_{m+1}}\left ( \theta\right ) \\ & = -\left\ { e_{\theta}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\left\ { \begin{array } [ c]{c}\left ( \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{m+1}}\\ -e_{\theta}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \end{array } \right\ } \left\ { e_{\theta}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\ ] ] so upon permuting the unit indices , the contribution of each of these @xmath1324 terms to @xmath2846 is@xmath2849 _ { i_{1}}\label{ifm+1}\\ & \times\left [ { \displaystyle\prod\limits_{s=3}^{m+1 } } \left\ { e_{\theta}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\left\ { \begin{array } [ c]{c}\left ( \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{s}}\\ -e_{\theta}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \end{array } \right\ } \right ] \nonumber\\ & \times\left\ { e_{\theta}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\left [ \overline{z}_{k}\left ( h_{1}\widetilde{b}\left ( \theta\right ) + h_{3}\right ) \dot { p}\right ] _ { i_{2}}\nonumber\end{aligned}\ ] ] which is already degenerate ( i.e. , orthogonal to @xmath2850 differentiating with respect to the @xmath2840 of @xmath2851 in @xmath2846 we obtain @xmath2852 _ { i_{1}}\\ & \left [ { \displaystyle\prod\limits_{s=3}^{m } } \left\ { e_{\theta}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\left\ { \left ( \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{s}}-e_{\theta}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } \right ] \times\\ & \left\ { e_{\theta}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline { z}_{k}^{t}\right ] \right\ } ^{-1}\left [ \overline{z}_{k}\left ( h_{1}\widetilde{b}\left ( \theta\right ) + h_{3}\right ) \dot{p}\right ] _ { i_{2}}\\ & + \left ( -1\right ) ^{m-1}\left [ \left ( h_{1}\widetilde{p}\left ( \theta\right ) + h_{2}\right ) \dot{b}\overline{z}_{k}^{t}\right ] _ { i_{1}}\times\\ & \left [ { \displaystyle\prod\limits_{s=3}^{m } } \left\ { e_{\theta}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\left\ { \left ( \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{s}}-e_{\theta}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } \right ] \times\\ & \left [ \overline{z}_{k}h_{1}\dot{p}\right ] if_{1,\widetilde{p}\left ( x_{i_{2}},\cdot\right ) , i_{m+1}}\,\left ( \theta\right)\end{aligned}\ ] ] substituting in the above expressions for @xmath2853 and @xmath2854 then projecting on @xmath2855 and again permuting unit indices , we obtain two identical terms both equal to eq ( @xmath2856 .
thus we obtain @xmath286 identical terms in all . upon dividing by @xmath286 ,
we conclude that @xmath2857 \ $ ] equals @xmath2833 operating on @xmath2858 proving the theorem .
( theorem [ ebrate ] ) equation ( [ eb6 ] ) follows from eq .
( [ eb4 ] ) by our assumption of rate optimality of the initial estimators .
we next prove eq .
( [ eb3 ] ) by induction . for @xmath1917@xmath2859
-\widetilde{\psi}_{k}\\ & = e\left [ \widehat{b}\widehat{p}h_{1}+\widehat{b}h_{2}+\widehat{p}h_{3}\right ] -e\left [ \widetilde{b}\widetilde{p}h_{1}+\widetilde{b}h_{2}+\widetilde{p}h_{3}\right ] \\ & = e\left [ \dot{b}\dot{p}\overline{z}_{k}^{t}\left ( p-\widehat{p}\right ) \left\ { \dot{p}\right\ } ^{-1}h_{1}\right ] e_{\theta}\left [ \dot{b}\dot { p}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] ^{-1}\\ & \times e\left [ \dot{p}\dot{b}\overline{z}_{k}\left ( b-\widehat{b}\ \right ) \left\ { \dot{b}\right\ } ^{-1}h_{1}\right]\end{aligned}\ ] ] where the last equality follows from @xmath2860 ^{-1}e\left [ \dot{p}\dot{b}\overline{z}_{k}\left ( b-\widehat{b}\ \right ) \left\ { \dot{b}\right\ } ^{-1}h_{1}\right ] \\ \widetilde{p } & = \widehat{p}-\dot{p}\overline{z}_{k}^{t}e_{\theta}\left [ \dot{b}\dot{p}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] ^{-1}e\left [ \dot{b}\dot{p}\overline{z}_{k}\left ( p-\widehat{p}\right ) \left\ { \dot { p}\right\ } ^{-1}h_{1}\right]\end{aligned}\ ] ] next , we proceed by induction .
assume [ eb3 ] holds for @xmath2861 we next show that it holds for @xmath190@xmath2862 \\ & = \left\ { \begin{array } [ c]{c}\left ( -1\right ) ^{m-2}e\left [ q^{2}\left ( \frac{b-\widehat{b}}{\dot{b}}\right ) \overline{z}_{k}^{t}\right ] \\
\left\ { e\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] -i_{k\times k}\right\ } ^{m-2}\left\ { e\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}e\left [ q^{2}\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) \right ] \end{array } \right\ } \\ & + \left\ { \begin{array } [ c]{c}\left ( -1\right ) ^{m-1}e\left [ q^{2}\left ( \frac{b-\widehat{b}}{\dot{b}}\right ) \overline{z}_{k}^{t}\right ] \\
\times\left\ { e\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] -i_{k\times k}\right\ } ^{m-2}e\left [ q^{2}\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) \right ] \end{array } \right\ } \\ & = \left ( -1\right ) ^{m-1}e\left [ q^{2}\left ( \frac{b-\widehat{b}}{\dot{b}}\right ) \overline{z}_{k}^{t}\right ] \left\ { e\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] -i_{k\times k}\right\ } ^{m-1}\\ & \times\left\ { e\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}e\left [ q^{2}\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) \right]\end{aligned}\ ] ] finally we prove that [ eb3 ] implies [ eb4 ] . for any random variable @xmath2863 define @xmath2864 = \widehat{q}\overline{z}_{k}^{t}\widehat{e}\left [ \delta g\widehat{q}\overline{z}_{k}h\right ] \\
\widehat{r}^{t}\left ( h\right ) & = \widehat{r}\circ\widehat{r}^{t-1}\left ( h\right ) \text { \ for } t\geq2\end{aligned}\ ] ] then @xmath2867{c}e\left [ q^{2}\left ( \frac{b-\widehat{b}}{\dot{b}}\right ) \overline{z}_{k}^{t}\right ] \left\ { \widehat{e}\left [ \delta g\text { } \widehat{q}^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{m-2}\times\\ \widehat{e}\left [ \delta g\text { } \widehat{q}\overline{z}_{k}\frac { \widehat{q}}{q}\pi\left ( q\overline{z}_{k}\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) |\left ( q\overline{z}_{k}\right ) \right ) \right ] \end{array } \right\ } \\ & = e\left [ q\left ( \frac{b-\widehat{b}}{\dot{b}}\right ) \frac { q}{\widehat{q}}\widehat{r}^{m-1}\left ( \frac{\widehat{q}}{q}\pi\left ( q\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) |\left ( q\overline{z}_{k}\right ) \right ) \right ) \right]\end{aligned}\]]@xmath2868 \\ & \leq\left\ { \widehat{e}\left [ q\left ( \frac{b-\widehat{b}}{\dot{b}}\right ) \frac{qf}{\widehat{q}\widehat{f}}\right ] ^{2}\right\ } ^{\frac { 1}{2}}\left\ { \widehat{e}\left [ \widehat{r}^{m-1}\left ( \frac{\widehat{q}}{q}\pi\left ( q\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) |\left ( q\overline{z}_{k}\right )
\right ) \right ) \right ] ^{2}\right\ } ^{\frac { 1}{2}}\ ] ] by cauchy shwartz . now @xmath2869 ^{2}\\ & \leq\widehat{e}\left [ \left ( \delta g\right ) ^{2}\left [ \widehat{r}^{m-2}\left ( \frac{\widehat{q}}{q}\pi\left ( q\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) |\left ( q\overline{z}_{k}\right ) \right ) \right ) \right ] ^{2}\right]\end{aligned}\ ] ] iterating this calculation @xmath1324 times we find @xmath2869 ^{2}\\ & \leq\left\vert \left\vert \delta g\right\vert \right\vert _ { \infty } ^{2\left ( m-1\right ) } \widehat{e}\left [ \frac{\widehat{q}}{q}\pi\left ( q\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) |\left ( q\overline{z}_{k}\right ) \right ) \right ] ^{2}\\ & \leq\left\vert \left\vert \delta g\right\vert \right\vert _ { \infty } ^{2\left ( m-1\right ) } \left\vert \left\vert \frac{\widehat{g}}{g}\right\vert \right\vert _ { \infty}e\left ( \pi\left ( q\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) |\left ( q\overline{z}_{k}\right )
\right ) \right ) ^{2}\\ & \leq\left\vert \left\vert \delta g\right\vert \right\vert _ { \infty } ^{2\left ( m-1\right ) } \left\vert \left\vert \frac{\widehat{g}}{g}\right\vert \right\vert _ { \infty}\int q^{2}\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) ^{2}f\left ( x\right ) dx\\ & \leq\left\vert \left\vert \delta g\right\vert \right\vert
_ { \infty } ^{2\left ( m-1\right ) } \left\vert \left\vert \frac{\widehat{g}}{g}\right\vert \right\vert _ { \infty}\left\vert \left\vert \frac{q^{2}f}{\dot{p}^{2}}\right\vert \right\vert
_ { \infty}\int\left ( p\left ( x\right ) -\widehat{p}\left ( x\right ) \right ) ^{2}dx\\ & = \left\vert \left\vert \delta g\right\vert \right\vert _ { \infty}^{2\left ( m-1\right ) } \left\vert \left\vert \frac{q^{2}f}{\dot{p}^{2}}\right\vert \right\vert _ { \infty}\left [ \int\left ( p\left ( x\right ) -\widehat{p}\left ( x\right ) \right ) ^{2}dx\right ] \left ( 1+o_{p}\left ( 1\right ) \right)\end{aligned}\ ] ] by @xmath2871 next @xmath2872 ^{2}\\ & = \int\frac{q^{2}f}{\dot{b}^{2}}\frac{g}{\widehat{g}}\left ( b\left ( x\right ) -\widehat{b}\left ( x\right ) \right ) ^{2}dx\\ & \leq\left\vert \left\vert \frac{q^{2}f}{\dot{b}^{2}}\right\vert \right\vert _ { \infty}\left\vert \left\vert \frac{g}{\widehat{g}}\right\vert \right\vert
_ { \infty}\int\left ( b\left ( x\right ) -\widehat{b}\left ( x\right ) \right ) ^{2}dx\\ & = \left\vert \left\vert \frac{q^{2}f}{\dot{b}^{2}}\right\vert \right\vert _ { \infty}\left [ \int\left ( b\left ( x\right ) -\widehat{b}\left ( x\right ) \right ) ^{2}dx\right ] \left ( 1+o_{p}\left ( 1\right ) \right)\end{aligned}\ ] ]
then we know @xmath2873{c}\left\vert \left\vert \delta g\right\vert \right\vert _ { \infty}^{m-1}\left\vert \left\vert \left ( \frac{\dot{b}}{\dot{p}}g\right ) \right\vert \right\vert _ { \infty}^{\frac{1}{2}}\left\vert \left\vert \frac{\dot{p}}{\dot{b}}g\right\vert \right\vert _ { \infty}^{\frac{1}{2}}\left ( 1+o_{p}\left ( 1\right ) \right ) \times\\ \left\ { \int\left ( p\left ( x\right ) -\widehat{p}\left ( x\right ) \right ) ^{2}dx\right\ } ^{\frac{1}{2}}\left\ { \int\left ( b\left ( x\right ) -\widehat{b}\left ( x\right ) \right ) ^{2}dx\right\ } ^{\frac{1}{2}}\end{array } \right\}\ ] ] to prove theorem [ var_multi ] , we first give the following univariate result .
suppose that we have a set @xmath2874 such that for each @xmath2875 pair , @xmath2876 either spans @xmath2877 or spans @xmath2878 where @xmath2879 and @xmath2880 are both nonnegative integers and @xmath2881
let @xmath2882 we first introduce an important preliminary lemma .
[ var_kernel]for @xmath2883 and @xmath2884 if @xmath2885 then @xmath2886 otherwise if @xmath2887 then @xmath2888 and @xmath2889 are both nonnegative integers and @xmath2890 . thus , @xmath2891 \right ) ^{-1}\left ( \pi_{j=1}^{m+1}k_{2j-1}\right ) = o\left ( 1\right)\ ] ] case 2 : @xmath1917 if @xmath2894 then , @xmath2895 \\ & \geq e\left [ \left ( k_{\left ( 1,k^{\ast}\right ) } \left ( x , x\right ) \right ) ^{2}\right ] \geq\left ( e\left [ k_{\left ( 1,k^{\ast}\right ) } \left ( x , x\right ) \right ] \right ) ^{2}=\left ( k^{\ast}\right ) ^{2}\asymp k_{1}k_{3}.\end{aligned}\ ] ] similarly , if @xmath2896 and we further assume @xmath2897 wlog , then@xmath2895 \\ & \geq e\left [ \left ( k_{\left ( \frac{k_{1}}{2}+1,k_{1}\right ) } \left ( x , x\right ) \right ) ^{2}\right ] \geq\left ( e\left [ k_{\left ( \frac{k_{1}}{2}+1,k_{1}\right ) } \left ( x , x\right ) \right ] \right ) ^{2}\\ & = \left ( \frac{k_{1}}{2}\right ) ^{2}\asymp k_{1}k_{3};\end{aligned}\ ] ] otherwise , wlog , we assume that @xmath2898 then@xmath2895 \\ & = e\left [ \sum_{r = k_{0}}^{k_{1}}\phi_{r}^{2}\left ( x\right ) \sum _ { s = k_{2}}^{k_{3}}\phi_{s}^{2}\left ( x\right ) \right ] \\ & \geq e\left [ \sum_{r = k_{0}}^{k_{1}}\phi_{r}^{2}\left ( x\right ) \sum_{s=\frac{k_{3}}{2}+1}^{k_{3}}\phi_{s}^{2}\left ( x\right ) \right]\end{aligned}\ ] ] let @xmath2901 indicate the correponding compactly supported father / mother wavelet on @xmath2902 , $ ] whose absolute value is bounded above by @xmath2903 by the continuity of @xmath2904 there exists a set @xmath2905 which is a union of finite number of disjoint open intervals , such that @xmath2906 is greater than @xmath2907 on @xmath12 . moreover , the lebesgue measure ( i.e. , the length ) of @xmath2908 is greater than @xmath2909 because @xmath2901 is bounded and has unit length .
specifically,@xmath2910 by definition @xmath2911 is a sequence of level @xmath1164 scaled and translated father wavelets on the unit interval @xmath436 $ ] , therefore it is obvious that the lebesgue measure of the set @xmath2912 is greater than @xmath2913 furthermore , for @xmath2914 , the set @xmath2915 consists of multiple disjoint open intervals whose lengths are all of order @xmath2916 in contrast , the support for any level @xmath2917 scaled and translated mother wavelet @xmath2918 @xmath2919 , is of order @xmath2920 as @xmath2921 therefore , at least @xmath2922 proportion of the level @xmath2917 scaled and translated mother wavelets @xmath2923 have their support inside @xmath2924 hence , @xmath2895 \\ & > e\left [ 1_{\widetilde{a}}\sum_{r=1}^{k^{\ast}}\phi_{r}^{2}\left ( x\right ) \sum_{s=\frac{k_{3}}{2}+1}^{k_{3}}\phi_{s}^{2}\left ( x\right ) \right ] \\ & \geq\frac{k^{\ast}}{2\left ( u_{w}-l_{w}\right ) } e\left [ 1_{\widetilde{a}}\sum_{s=\frac{k_{3}}{2}+1}^{k_{3}}\phi_{s}^{2}\left ( x\right ) \right ] \\ & > \frac{k^{\ast}}{2\left ( u_{w}-l_{w}\right ) } \frac{1}{4m_{w}^{2}\left ( u_{w}-l_{w}\right ) } \frac{k_{3}}{2}\\ & \asymp k_{1}k_{3}.\end{aligned}\ ] ] case 3 : @xmath337 wlog , we assume that @xmath2925 and @xmath2926 for @xmath2927 @xmath2928 then @xmath2929 \\ & \geq e\left [ \left ( k_{\left ( k_{0},k_{1}\right ) } \right ) ^{l_{1}}{\displaystyle\prod\limits_{r=2}^{t } } \left ( k_{\left ( \frac{k_{2l_{r-1}+1}}{2}+1,k_{2l_{r-1}+1}\right ) } \right ) ^{l_{r}-l_{r-1}}\right ] \\ & > e\left [ \left ( \sum_{r = k_{0}}^{k_{1}}\phi_{r}^{2}\left ( x\right ) \right ) ^{l_{1}}{\displaystyle\prod\limits_{r=2}^{t } } \left ( \sum_{r=\frac{k_{2l_{r-1}+1}}{2}+1}^{k_{2l_{r-1}+1}}\phi_{r}^{2}\left ( x\right ) \right ) ^{l_{r}-l_{r-1}}\right]\end{aligned}\ ] ] by a similar argument as that in case 2 , we can show that there exists a set @xmath2930 which consists of multiple disjoint open intervals , such that @xmath2931 @xmath1939 @xmath2932 and @xmath2933 for some positive constants @xmath2934 moreover , by the multiresolution analysis ( mra ) property of compactly supported wavelets and the fact that @xmath2935 , it is obvious that at least @xmath2936 proportion of the level @xmath2937 scaled and translated mother wavelets @xmath2938 have support inside @xmath2939 hence we can find a set @xmath2940 , which is also a union of disjoint open intervals , such that @xmath2941 @xmath1939 @xmath2942 and @xmath2943 for some positive constants @xmath2944 furthermore , by applying this algorithm in a nested fashion @xmath2945 times , we can find a decreasing sequence of sets @xmath2946 such that for any @xmath2947 @xmath2948 @xmath1939 @xmath2949 and @xmath2950 for some positive constants @xmath2951 and @xmath2952 in addition , @xmath2953 proportion of the level @xmath2954 scaled and translated mother wavelets @xmath2955 have support inside @xmath2956 hence , @xmath2929 \\ & > \left ( { \displaystyle\prod\limits_{r=1}^{t-1 } } k_{2l_{r-1}+1}^{l_{r}-l_{r-1}}\right ) e \left [ \left ( 1_{a_{t-1}}\sum_{r=\frac{k_{2l_{t-1}+1}}{2}+1}^{k_{2l_{t-1}+1}}\phi_{r}^{2}\left ( x\right ) \right ) ^{l_{t}-l_{t-1}}\right ] \\ & \geq\left ( { \displaystyle\prod\limits_{r=1}^{t-1 } } k_{2l_{r-1}+1}^{l_{r}-l_{r-1}}\right ) e\left [ \left ( 1_{a_{t-1}}\sum_{r=\frac{k_{2l_{t-1}+1}}{2}+1}^{k_{2l_{t-1}+1}}\phi_{r}^{2}\left ( x\right ) \right ) \right ] ^{l_{t}-l_{t-1}}\\ & > \left ( { \displaystyle\prod\limits_{r=1}^{t-1 } } k_{2l_{r-1}+1}^{l_{r}-l_{r-1}}\right ) \left ( \frac{k_{2l_{t-1}+1}}{2}{\displaystyle\prod\limits_{r=1}^{t-1 } } \frac{\delta_{r}^{\phi}}{2}\right ) ^{l_{t}-l_{t-1}}\times{\displaystyle\prod\limits_{j=1}^{m+1 } } k_{2j-1}\ ] ] ( lemma : [ var_uni ] ) case 1:@xmath2958@xmath2959 ^{2}\\ & = e\left [ \left ( k_{\left ( k_{0},k_{1}\right ) } \left ( x_{i_{1}},x_{i_{2}}\right ) \right ) ^{2}\right ] \\ & = e\left [ \left ( k_{\left ( k_{0},k_{1}\right ) } \left ( x_{i_{1}},x_{i_{1}}\right ) \right ) \right ] \\ & = k_{1}-k_{0}\ ] ] which follows from the orthonormality of wavelets .
consider the case @xmath2124 @xmath2962 \\ & = e\left [ \left ( k_{\left ( k_{0},k_{1}\right ) } \left ( x_{i_{2}},x_{i_{2}}\right ) k_{\left ( k_{2},k_{3}\right ) } \left ( x_{i_{2}},x_{i_{2}}\right ) \right ) \right ] \\ & \leq||k_{\left ( k_{0},k_{1}\right ) } \left ( x , x\right ) ||_{\infty } ||k_{\left ( k_{2},k_{3}\right ) } \left ( x , x\right ) ||_{\infty}\ ] ] if @xmath2963 then @xmath2964 .
similarly , if @xmath2965 and @xmath2966 then @xmath2967 otherwise , if @xmath2965 and @xmath2968 then @xmath2969 finally , for @xmath1927@xmath2970 \\ & = e\left [ k_{\left ( k_{0},k_{1}\right ) } \left ( x_{i_{1}},x_{i_{2}}\right ) ^{2}\left ( { \displaystyle\prod\limits_{j=2}^{m } } k_{\left ( k_{2j-2},k_{2j-1}\right ) } \left ( x_{i_{j}},x_{i_{j+1}}\right ) \right ) ^{2}k_{\left ( k_{2m},k_{2m+1}\right ) } \left ( x_{i_{m+1}},x_{i_{m+2}}\right ) ^{2}\right ] \\ & = e\left [ k_{\left ( k_{0},k_{1}\right ) } \left ( x_{i_{2}},x_{i_{2}}\right ) \left ( { \displaystyle\prod\limits_{j=2}^{m } } k_{\left ( k_{2j-2},k_{2j-1}\right ) } \left ( x_{i_{j}},x_{i_{j+1}}\right ) \right ) ^{2}k_{\left ( j_{2m},k_{2m+1}\right ) } \left ( x_{i_{m+2}},x_{i_{m+2}}\right ) \right ] \\ & \leq||k_{\left ( k_{0},k_{1}\right ) } \left ( x , x\right ) ||_{\infty } e\left [ \left ( { \displaystyle\prod\limits_{j=2}^{m } } k_{\left ( k_{2j-2},k_{2j-1}\right ) } \left ( x_{i_{j}},x_{i_{j+1}}\right ) \right ) ^{2}\right ] ||k_{\left ( k_{2m},k_{2m+1}\right ) } \left ( x , x\right ) ||_{\infty}\\ & = ||k_{\left ( k_{0},k_{1}\right ) } \left ( x , x\right ) ||_{\infty } ||k_{\left ( k_{2m},k_{2m+1}\right ) } \left ( x , x\right ) ||_{\infty}e\left [ \left ( { \displaystyle\prod\limits_{j=2}^{m } } k_{\left ( k_{2j-2},k_{2j-1}\right ) } \left ( x_{i_{j}},x_{i_{j+1}}\right ) \right ) ^{2}\right ] \\ & \leq{\displaystyle\prod\limits_{j=1}^{m+1 } } we shall first show that @xmath2972 and then complete the proof by showing that @xmath2973 specifically , @xmath2974 \\ & = e\left [ k_{\left ( 1,k_{1}\right ) } \left ( x_{i_{3}},x_{i_{3}}\right ) k_{\left ( 1,k_{3}\right ) } \left ( x_{i_{3}},x_{i_{3}}\right ) { \displaystyle\prod\limits_{j=3}^{m+1 } } \left ( k_{\left ( 1,k_{2j-1}\right ) } \left ( x_{i_{j}},x_{i_{j+1}}\right ) \right ) ^{2}\right ] \\ & -e\left [ k_{1}k_{3}\left\ { h_{x_{i_{3}}}\left ( x_{i_{3}}\right ) -\left ( \mathcal{k}_{\left ( 1,k_{3}\right ) } \mathcal{\circ}h_{x_{i_{3}}}\right ) \left ( x_{i_{3}}\right ) \right\ } { \displaystyle\prod\limits_{j=3}^{m+1 } } \left ( k_{\left ( 1,k_{2j-1}\right ) } \left ( x_{i_{j}},x_{i_{j+1}}\right ) \right ) ^{2}\right]\end{aligned}\ ] ] where @xmath2975 and @xmath2976 .$ ] as previously shown , there exists a positive constant @xmath2977 such that @xmath2978 by the regularity property of compactly supported wavelets and the approximation property of the kernel @xmath2979 it is obvious that @xmath2980 thus , @xmath2030{c}k_{1}k_{3}\left\ { h_{x_{i_{3}}}\left ( x_{i_{2}}\right ) -\left ( \mathcal{k}_{\left ( 1,k_{3}\right ) } \mathcal{\circ}h_{x_{i_{3}}}\right ) \left ( x_{i_{3}}\right ) \right\ } \\ \times{\displaystyle\prod\limits_{j=3}^{m+1 } } \left ( k_{\left ( 1,k_{2j-1}\right ) } \left ( x_{i_{j}},x_{i_{j+1}}\right ) \right ) ^{2}\end{array } \right ] \\ & = o\left ( \pi_{j=1}^{m+1}k_{2j-1}\right)\end{aligned}\ ] ] in fact , by arguing similarly as above , @xmath2981{c}k_{\left ( 1,k_{1}\right ) } \left ( x_{i_{3}},x_{i_{3}}\right ) k_{\left ( 1,k_{3}\right ) } \left ( x_{i_{3}},x_{i_{3}}\right ) \\ \times{\displaystyle\prod\limits_{j=3}^{m+1 } } \left ( k_{\left ( 1,k_{2j-1}\right ) } \left ( x_{i_{j}},x_{i_{j+1}}\right ) \right ) ^{2}\end{array } \right\ } \right ] + o\left ( \pi_{j=1}^{m+1}k_{2j-1}\right ) \\ & = e\left [ \left\ { \begin{array } [ c]{c}k_{\left ( 1,k_{1}\right ) } \left ( x_{i_{4}},x_{i_{4}}\right ) k_{\left ( 1,k_{3}\right ) } \left ( x_{i_{4}},x_{i_{4}}\right ) k_{\left ( 1,k_{5}\right ) } \left ( x_{i_{4}},x_{i_{4}}\right ) \\ \times{\displaystyle\prod\limits_{j=4}^{m+1 } } \left ( k_{\left ( 1,k_{2j-1}\right ) } \left ( x_{i_{j}},x_{i_{j+1}}\right ) \right ) ^{2}\end{array } \right\ } \right ] \\ & + o\left ( \pi_{j=1}^{m+1}k_{2j-1}\right ) \\ & = e\left [ { \displaystyle\prod\limits_{j=1}^{m+1 } } k_{\left ( 1,k_{2j-1}\right ) } \left ( x , x\right ) \right ] + o\left ( \pi_{j=1}^{m+1}k_{2j-1}\right ) \\ & \asymp{\displaystyle\prod\limits_{j=1}^{m+1 } } k_{2j-1}\ ] ] now , we prove eq .
( [ eq1 ] ) .
@xmath2982@xmath906@xmath2983 where @xmath2984 may be @xmath2985 or @xmath2986 but @xmath2987 therefore @xmath2988 as a direct consequence of the previously proved result that @xmath2989 the proof is complete .
( theorem [ var_multi ] ) we note that since the linear span of @xmath2990 equals @xmath2991we have @xmath2992{c}t_{1}, ... ,t_{l}:\\ 1\leq t_{u}\leq k_{j , u}\\ u=1, ...
,l \end{array } \right\ } } \prod\limits_{u=1}^{l}\overline{\varphi}_{t_{u}}\left ( x_{i_{j}}^{u}\right ) \prod\limits_{u^{\prime}=1}^{l}\overline{\varphi}_{t_{u^{\prime } } } \left ( x_{i_{j+1}}^{u^{\prime}}\right)\end{aligned}\ ] ] so that @xmath2993 the remainder of the proof then follows from lemma [ var_uni ] .
( theorem [ 3.19 ] ) in the proof of theorem [ 3.19 ] , the following lemma plays a central role .
note that expectations and probabilities remain conditional on @xmath391 even when it is suppressed in the notation .
[ 3.19_lemma]let @xmath2994 be the symmetric kernel of @xmath2995 then for any @xmath2421 and @xmath2996 @xmath2997 -var_{\widehat{\theta}}\left [ \left ( \widehat{\mathbb{if}}_{m , m,\widetilde{\psi}_{k}\left ( \cdot\right ) } \right ) ^{2}\right ] \nonumber\\ & = o\left ( n^{-m}\left\ { e_{\theta}\left [ \left ( \widehat{if}_{m , m,\overline{i}_{m}}^{\left ( s\right ) } \right ) ^{2}\right ] -e_{\widehat{\theta}}\left [ \left ( \widehat{if}_{m , m,\overline{i}_{m}}^{\left ( s\right ) } \right ) ^{2}\right ] \right\ } \right ) \nonumber\\ & + o\left ( \frac{1}{n}\left\ { e_{\theta}\left [ \widehat{if}_{m , m,\overline { i}_{m}}^{\left ( s\right ) } \right ] \right\ } ^{2}\right ) \nonumber\\ & + o\left ( \max\left ( \frac{1}{n},\frac{k^{m-2}}{n^{m-1}}\right ) \right ) \label{lemma_1}\ ] ] and @xmath2998 e_{\theta}\left [ \widehat{if}_{m_{2},m_{2},\overline{i}_{m}}^{\left ( s\right ) } \right ] \right ) + o\left ( \max\left ( \frac{1}{n},\frac{k^{m_{1}-2}}{n^{m_{1}-1}}\right ) \right ) \label{lemma_2}\ ] ] ( theorem [ 3.19 ] ) . by the degeneracy of @xmath2999 for any @xmath3000 under @xmath3001
@xmath3002 -var_{\widehat{\theta}}\left [ \widehat{\mathbb{if}}_{m,\widetilde{\psi}_{k}\left ( \cdot\right ) } |\widehat{\theta}\right ] } { var_{\widehat{\theta}}\left [ \widehat{\mathbb{if}}_{m,\widetilde{\psi}_{k}\left ( \cdot\right ) } |\widehat{\theta}\right ] } \\ & = \frac{\left\ { \begin{array } [ c]{c}{\displaystyle\sum\limits_{t=1}^{m } } \left ( var_{\theta}\left [ \widehat{\mathbb{if}}_{t , t,\widetilde{\psi}_{k}\left ( \cdot\right ) } \right ] -var_{\widehat{\theta}}\left [ \widehat{\mathbb{if}}_{t , t,\widetilde{\psi}_{k}\left ( \cdot\right ) } \right ] \right ) \\ + 2{\displaystyle\sum\limits_{1\leq t_{1}<t_{2}\leq m } } cov_{\theta}\left [ \widehat{\mathbb{if}}_{t_{1},t_{1},\widetilde{\psi}_{k}\left ( \cdot\right ) } , \widehat{\mathbb{if}}_{t_{2},t_{2},\widetilde{\psi } _ { k}\left ( \cdot\right ) } \right ] \end{array } \right\ } } { var_{\widehat{\theta}}\left [ \widehat{\mathbb{if}}_{m,\widetilde{\psi}_{k}\left ( \cdot\right ) } |\widehat{\theta}\right ] } , \end{aligned}\ ] ] which equals @xmath3003 -e_{\widehat{\theta}}\left [ \left ( \widehat{if}_{t , t,\overline{i}_{m}}^{\left ( s\right ) } \right ) ^{2}\right ] \right ) } { { \displaystyle\sum\limits_{t=1}^{m } } n^{-t}e_{\widehat{\theta}}\left [ \left ( \widehat{if}_{t , t,\overline{i}_{m}}^{\left ( s\right ) } \right ) ^{2}\right ] } \left ( 1+o\left ( 1\right ) \right ) \\ & + \frac{o\left ( \max\left ( \frac{1}{n},\frac{k^{m-2}}{n^{m-1}}\right ) \right ) } { var_{\widehat{\theta}}\left [ \widehat{\mathbb{if}}_{m,\widetilde{\psi}_{k}\left ( \cdot\right ) } |\widehat{\theta}\right ] } \ ] ] by assumption , @xmath3004 as @xmath3005 , hence @xmath3006 -e_{\widehat{\theta}}\left [ \left ( \widehat{if}_{t , t,\overline{i}_{m}}^{\left ( s\right ) } \right ) ^{2}\right ] } { e_{\widehat{\theta}}\left [ \left ( \widehat{if}_{t , t,\overline { i}_{m}}^{\left ( s\right ) } \right ) ^{2}\right ] } \\ & = \frac{e_{\widehat{\theta}}\left [ \left ( \widehat{if}_{t , t,\overline { i}_{m}}^{\left ( s\right ) } \right ) ^{2}\left ( { \displaystyle\prod\limits_{s=1}^{t } } \frac{f\left ( o_{i_{s}};\theta\right ) } { f\left ( o_{i_{s}};\widehat{\theta } \right ) } -1\right ) \right ] } { e_{\widehat{\theta}}\left [ \left ( \widehat{if}_{t , t,\overline{i}_{m}}^{\left ( s\right ) } \right ) ^{2}\right ] } \\ & \leq\sup_{\mathbf{o}_{\overline{i}_{t}}}\left\vert { \displaystyle\prod\limits_{s=1}^{t } } \frac{f\left ( o_{i_{s}};\theta\right ) } { f\left ( o_{i_{s}};\widehat{\theta } \right ) } -1\right\vert \\ & = o\left ( \underset{o\in\mathcal{o}}{sup}\left\vert f\left ( o;\widehat{\theta}\right ) -f\left ( o;\theta\right ) \right\vert \right ) = o\left ( 1\right ) .\end{aligned}\ ] ] furthermore , @xmath3007 } = o\left ( 1\right ) $ ] as well since @xmath3008 \times\max\left ( \frac{1}{n},\frac{k^{m-1}}{n^{m}}\right ) $ ] from theorem [ var_multi ] .
thus the proof of thoerem [ 3.19 ] is complete . as proved in theorem [ on ] , for any @xmath316 @xmath3011{c}\left [ \left ( h_{1}\widehat{p}+h_{2}\right ) \dot{b}\overline{z}_{k}^{t}\right ] _ { i_{1}}\left [ { \displaystyle\prod\limits_{s=3}^{m } } \left\ { \left ( \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{s}}-i_{k\times k}\right\ } \right ] \\ \times\left [ \overline{z}_{k}\left ( h_{1}\widehat{b}+h_{3}\right ) \dot { p}\right ] _ { i_{2}}\end{array } \right\}\end{aligned}\ ] ] we first consider the case when @xmath332 @xmath3012 \\ & = e_{\theta}\left [ q^{2}\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) \overline{z}_{k}^{t}\right ] \left [ \overline{z}_{k}\left ( h_{1}\widehat{b}+h_{3}\right ) \dot{p}\right ] _ { i_{2}}\\ & = \widehat{e}\left [ \frac{f\left ( x\right ) } { \widehat{f}\left ( x\right ) } \frac{q^{2}}{\widehat{q}}\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) \widehat{q}\overline{z}_{k}^{t}\right ] \left [ \frac{\widehat{q}}{\widehat{q}}\overline{z}_{k}\left ( h_{1}\widehat{b}+h_{3}\right ) \dot { p}\right ] _ { i_{2}}\\ & = \widehat{\pi}\left [ \left ( \frac{f\left ( x\right ) } { \widehat{f}\left ( x\right ) } \frac{q^{2}}{\widehat{q}}\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) \right ) |\left ( \widehat{q}\overline{z}_{k}\right ) _ { i_{2}}\right ] \left ( \frac{\left ( h_{1}\widehat{b}+h_{3}\right ) \dot{p}}{\widehat{q}}\right ) _ { i_{2}}\\ & = \left\vert \left\vert p-\widehat{p}\right\vert \right\vert _ { 2}\frac { t_{c}\left ( o_{i_{2}}\right ) } { \left\vert \left\vert p-\widehat{p}\right\vert \right\vert _ { 2}}\left ( \frac{\left ( h_{1}\widehat{b}+h_{3}\right ) \dot{p}}{\widehat{q}}\right ) _ { i_{2}}\ ] ] \i ) if @xmath3018 then @xmath3019 \left [ { \displaystyle\prod\limits_{s=3}^{m } } \left\ { \begin{array } [ c]{c}\left ( \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{s}}\\ -i_{k\times k}\end{array } \right\ } \right ] \\ & \times\left [ \overline{z}_{k}\left ( h_{1}\widehat{b}+h_{3}\right ) \dot{p}\right ] _ { i_{2}}\\ & = \left\ { \begin{array } [ c]{c}\widehat{e}\left [ \frac{f\left ( x\right ) } { \widehat{f}\left ( x\right ) } q^{2}\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) \overline{z}_{k}^{t}\right ] \left ( \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{3}}\\ -e_{\theta}\left [ q^{2}\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) \overline{z}_{k}^{t}\right ] \end{array } \right\ } \times\\ & { \displaystyle\prod\limits_{s=4}^{m } } \left\ { \begin{array } [ c]{c}\left ( \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{s}}\\ -i_{k\times k}\end{array } \right\ } \left [ \overline{z}_{k}\left ( h_{1}\widehat{b}+h_{3}\right ) \dot{p}\right ] _ { i_{2}}\]]@xmath906@xmath3020{c}\left ( \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{s}}\\ -i_{k\times k}\end{array } \right\ } \left [ \overline{z}_{k}\left ( h_{1}\widehat{b}+h_{3}\right ) \dot{p}\right ] _ { i_{2}}\\ & -e_{\theta}\left ( \widehat{if}_{m-1,m-1,i_{1}i_{2}i_{4} ... i_{m}}|\mathbf{o}_{-i_{1}}\right ) \\ & = \left ( \frac{\dot{p}\dot{b}h_{1}}{\widehat{q}}\right ) _ { i_{3}}t_{c}\left ( o_{i_{3}}\right ) \overline{z}_{k , i_{3}}^{t}{\displaystyle\prod\limits_{s=4}^{m } } \left\ { \begin{array } [ c]{c}\left ( \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{s}}\\ -i_{k\times k}\end{array } \right\ } \\ & \times\left [ \overline{z}_{k}\left ( h_{1}\widehat{b}+h_{3}\right ) \dot{p}\right ] _ { i_{2}}-e_{\theta}\left ( \widehat{if}_{m-1,m-1,i_{1}i_{2}i_{4} ... i_{m}}|\mathbf{o}_{-i_{1}}\right)\end{aligned}\ ] ] from the fact that @xmath3021 we have @xmath3022 |\mathbf{o}_{-i_{1}}\right ) ^{2}\right ] \\ & \leq2e\left [ \left ( \begin{array } [ c]{c}\left ( \frac{\dot{p}\dot{b}h_{1}}{\widehat{q}}\right ) _ { i_{3}}t_{c}\left ( o_{i_{3}}\right ) \overline{z}_{k , i_{3}}^{t}\times\\{\displaystyle\prod\limits_{s=4}^{m } } \left\ { \begin{array } [ c]{c}\left ( \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _
{ i_{s}}\\ -i_{k\times k}\end{array } \right\ } \left [ \overline{z}_{k}\left ( h_{1}\widehat{b}+h_{3}\right ) \dot{p}\right ] _ { i_{2}}\end{array } \right ) ^{2}\right ] \\ & + 2e\left [ \left ( e_{\theta}\left [ \widehat{if}_{m-1,m-1,i_{1}i_{2}i_{4} ... i_{m}}|\mathbf{o}_{-i_{1}}\right ] \right ) ^{2}\right]\end{aligned}\ ] ] from theorem [ var_multi ] , it can be shown that @xmath3023{c}\left ( \frac{\dot{p}\dot{b}h_{1}}{\widehat{q}}\right ) _ { i_{3}}t_{c}\left ( o_{i_{3}}\right ) \overline{z}_{k , i_{3}}^{t}\times\\{\displaystyle\prod\limits_{s=4}^{m } } \left\ { \begin{array } [ c]{c}\left ( \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _
{ i_{s}}\\ -i_{k\times k}\end{array } \right\ } \left [ \overline{z}_{k}\left ( h_{1}\widehat{b}+h_{3}\right ) \dot{p}\right ] _ { i_{2}}\end{array } \right ) ^{2}\right ] \\ & = \left\vert \left\vert p-\widehat{p}\right\vert \right\vert _ { \infty } o\left ( k^{m-2}\right ) \\ & = o\left ( k^{m-2}\right)\end{aligned}\ ] ] \iii ) if @xmath3027 wlog , assume @xmath3028 then@xmath3029 \\ & = \left [ \left ( h_{1}\widehat{p}+h_{2}\right ) \dot{b}\overline{z}_{k}^{t}\right ] _ { i_{1}}\left ( e_{\theta}\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] -i\right ) \times\\ & { \displaystyle\prod\limits_{s=4}^{m } } \left\ { \begin{array } [ c]{c}\left ( \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{s}}\\ -i_{k\times k}\end{array } \right\ } \left [ \overline{z}_{k}\left ( h_{1}\widehat{b}+h_{3}\right ) \dot{p}\right ] _ { i_{2}}\\ & = \left [ \left ( h_{1}\widehat{p}+h_{2}\right ) \dot{b}\overline{z}_{k}^{t}\right ] _ { i_{1}}\widehat{e}\left [ \delta g\text { } \widehat{q}^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \left ( \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{4}}\\ & \times{\displaystyle\prod\limits_{s=5}^{m } } \left\ { \left ( \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{s}}-i_{k\times k}\right\ } \left [ \overline{z}_{k}\left ( h_{1}\widehat{b}+h_{3}\right ) \dot{p}\right ] _ { i_{2}}\left ( \equiv \widehat{t}\right ) \\ & -e_{\theta}\left [ \widehat{if}_{m-1,m-1,i_{1}i_{2}i_{3}i_{5} .. i_{m}}|\mathbf{o}_{-i_{3}}\right]\end{aligned}\ ] ] moreover , it can be shown that @xmath3030 = o\left ( \left\vert \left\vert \delta g\right\vert \right\vert _ { \infty}^{2}k^{m-2}\right ) = o\left ( k^{m-2}\right ) $ ] following the proof of theorem [ var_multi ] but replacing @xmath3031 with @xmath3032 \overline{\phi } _ { 0}^{k}\left ( x_{i_{2}}\right ) .$ ] specifically , @xmath3033 \overline{\phi}_{0}^{k}\left ( x_{i_{2}}\right ) \overline{\phi}_{0}^{k , t}\left ( x_{i_{2}}\right ) \widehat{e}\left [ \delta g\text { } \widehat{q}^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right ) \\ & = \overline{\phi}_{0}^{k , t}\left ( x_{i_{2}}\right ) \left ( \widehat{e}\left [ \delta g\text { } \widehat{q}^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right ) ^{2}\overline{\phi}_{0}^{k}\left ( x_{i_{2}}\right ) \\ & \leq\left\vert \left\vert \delta g\right\vert \right\vert _ { \infty}^{2}\overline{\phi}_{0}^{k , t}\left ( x_{i_{2}}\right ) \overline{\phi}_{0}^{k}\left ( x_{i_{2}}\right)\end{aligned}\ ] ] the last inequality holds because @xmath3034 $ ] is a semi - positive definite symmetric matrix . @xmath3035
\right ) ^{2}\right ] $ ] is of order @xmath3036 by induction assumption .
now , the proof of this proposition is complete .
moreover , by arguing similarly , the result above can be generalized in a straightforward manner to the following proposition .
( lemma [ 3.19_lemma ] ) throughout the proof , we repeatedly use the result that @xmath3039 for any @xmath2421 is degenerate under @xmath3040 we first prove eq .
( [ lemma_1 ] ) . @xmath2820@xmath2997 -var_{\widehat{\theta}}\left [ \left ( \widehat{\mathbb{if}}_{m , m,\widetilde{\psi}_{k}\left ( \cdot\right ) } \right ) ^{2}\right ] \\ & = e_{\theta}\left [ \left ( \widehat{\mathbb{if}}_{m , m,\widetilde{\psi}_{k}\left ( \cdot\right ) } \right ) ^{2}\right ] -e_{\widehat{\theta}}\left [ \left ( \widehat{\mathbb{if}}_{m , m,\widetilde{\psi}_{k}\left ( \cdot\right ) } \right ) ^{2}\right ] -\left ( e_{\theta}\left [ \widehat{\mathbb{if}}_{m , m,\widetilde{\psi}_{k}\left ( \cdot\right ) } \right ] \right ) ^{2}\\ & = e_{\widehat{\theta}}\left [ \left ( \widehat{\mathbb{if}}_{m , m,\widetilde{\psi}_{k}\left ( \cdot\right ) } \right ) ^{2}\left ( { \textstyle\prod\limits_{i=1}^{n } } \frac{f\left ( o_{i}\right ) } { \widehat{f}\left ( o_{i}\right ) } -1\right ) \right ] -\left ( e_{\theta}\left [ \widehat{\mathbb{if}}_{m , m,\widetilde{\psi } _ { k}\left ( \cdot\right ) } \right ] \right ) ^{2}\ ] ] can be written as a sum of four terms as below:@xmath3041 -\left ( e_{\theta}\left [ \widehat{\mathbb{if}}_{m , m,\widetilde{\psi } _ { k}\left ( \cdot\right ) } \right ] \right ) ^{2}\\ & = e_{\widehat{\theta}}\left\ { \begin{array } [ c]{c}\left [ \binom{n}{m}\right ] ^{-2}\left ( { \displaystyle\sum\limits_{i_{1}<i_{2} .. <i_{m } } } \widehat{if}_{m , m,\overline{i}_{m}}^{\left ( s\right ) } \right )
\\ \times\left ( { \displaystyle\sum\limits_{r_{1}<r_{2} .. <r_{m } } } \widehat{if}_{m , m,\overline{r}_{m}}^{\left ( s\right ) } \right ) \left ( { \textstyle\prod\limits_{i=1}^{n } } \frac{f\left ( o_{i}\right ) } { \widehat{f}\left ( o_{i}\right ) } -1\right ) \end{array } \right\ } \\ & -\left ( e_{\theta}\left [ \widehat{\mathbb{if}}_{m , m,\widetilde{\psi}_{k}\left ( \cdot\right ) } \right ] \right ) ^{2}\\ & = e_{\widehat{\theta}}\left\ { \begin{array } [ c]{c}\left [ \binom{n}{m}\right ] ^{-2}\left [ \begin{array } [ c]{c}{\displaystyle\sum\limits_{i_{1}<i_{2} .. <i_{m } } } \left ( \widehat{if}_{m , m,\overline{i}_{m}}^{\left ( s\right ) } \right ) ^{2}\\ + { \displaystyle\sum\limits_{\overline{i}_{m}\cap\overline{r}_{m}=\emptyset } } \widehat{if}_{m , m,\overline{i}_{m}}^{\left ( s\right ) } \widehat{if}_{m , m,\overline{r}_{m}}^{\left ( s\right ) } \\ + { \displaystyle\sum\limits_{1\leq\#\left ( \overline{i}_{m}\cap\overline{r}_{m}\right ) < m } } \widehat{if}_{m , m,\overline{i}_{m}}^{\left ( s\right ) } \widehat{if}_{m , m,\overline{r}_{m}}^{\left ( s\right ) } \end{array } \right ]
\\ \times\left ( { \textstyle\prod\limits_{i=1}^{n } } \frac{f\left ( o_{i}\right ) } { \widehat{f}\left ( o_{i}\right ) } -1\right ) \end{array } \right\ } \\ & -\left ( e_{\theta}\left [ \widehat{\mathbb{if}}_{m , m,\widetilde{\psi}_{k}\left ( \cdot\right ) } \right ] \right ) ^{2}\ ] ] where @xmath3042 is the number of elements in the intersection set @xmath3043 the first term : @xmath3044 ^{-2}{\displaystyle\sum\limits_{i_{1}<i_{2} .. <i_{m } } } \widehat{if}_{m , m,\overline{i}_{m}}^{\left ( s\right ) 2}\left ( { \textstyle\prod\limits_{i=1}^{n } } \frac{f\left ( o_{i}\right ) } { \widehat{f}\left ( o_{i}\right ) } -1\right ) \right ] \\ & = \left [ \binom{n}{m}\right ] ^{-1}e_{\widehat{\theta}}\left [ \widehat{if}_{m , m,\overline{i}_{m}}^{\left ( s\right ) 2}\left ( { \textstyle\prod\limits_{i=1}^{n } } \frac{f\left ( o_{i}\right ) } { \widehat{f}\left ( o_{i}\right ) } -1\right ) \right ] \\ & = \left [ \binom{n}{m}\right ] ^{-1}\left ( e_{\theta}\left [ \widehat{if}_{m , m,\overline{i}_{m}}^{\left ( s\right ) 2}\right ] -e_{\widehat{\theta}}\left [ \widehat{if}_{m , m,\overline{i}_{m}}^{\left ( s\right ) 2}\right ] \right)\end{aligned}\ ] ] the second term : @xmath3044 ^{-2}\left ( { \displaystyle\sum\limits_{\overline{i}_{m}\cap\overline{r}_{m}=\emptyset } } \widehat{if}_{m , m,\overline{i}_{m}}^{\left ( s\right ) } \widehat{if}_{m , m,\overline{r}_{m}}^{\left ( s\right ) } \right ) \left ( { \textstyle\prod\limits_{i=1}^{n } } \frac{f\left ( o_{i}\right ) } { \widehat{f}\left ( o_{i}\right ) } -1\right ) \right ] \\ & = \left [ \binom{n}{m}\right ] ^{-2}e_{\theta}\left [ { \displaystyle\sum\limits_{\overline{i}_{m}\cap\overline{r}_{m}=\emptyset } } \widehat{if}_{m , m,\overline{i}_{m}}^{\left ( s\right ) } \widehat{if}_{m , m,\overline{r}_{m}}^{\left ( s\right ) } \right ] \\ & = \left [ \binom{n}{m}\right ] ^{-2}\binom{n}{m}\binom{n - m}{m}e_{\theta } \left [ \widehat{if}_{m , m,\overline{i}_{m}}^{\left ( s\right ) } \widehat{if}_{m , m,\overline{r}_{m}}^{\left ( s\right ) } \right ] \\ & = \frac{\left ( n - m\right ) ... \left ( n-2m+1\right ) } { n\left ( n-1\right ) ... \left ( n - m+1\right ) } \left ( e_{\theta}\left [ \widehat{if}_{m , m,\overline{i}_{m}}^{\left ( s\right ) } \right ] \right ) ^{2}\ ] ] substracting the fourth term from the second term , we have @xmath3045 \left ( e_{\theta}\left [ \widehat{if}_{m , m,\overline{i}_{m}}^{\left ( s\right ) } \right ] \right ) ^{2}\\ & = o\left ( \frac{1}{n}\left ( e_{\theta}\left [ \widehat{if}_{m , m,\overline { i}_{m}}^{\left ( s\right ) } \right ] \right ) ^{2}\right)\end{aligned}\ ] ] the third term:@xmath3044 ^{-2}\left ( { \displaystyle\sum\limits_{1\leq\#\left ( \overline{i}_{m}\cap\overline{r}_{m}\right ) < m } } \widehat{if}_{m , m,\overline{i}_{m}}^{\left ( s\right ) } \widehat{if}_{m , m,\overline{r}_{m}}^{\left ( s\right ) } \right ) \left ( { \textstyle\prod\limits_{i=1}^{n } } \frac{f\left ( o_{i}\right ) } { \widehat{f}\left ( o_{i}\right ) } -1\right ) \right ] \\ & = \sum_{t=1}^{m-1}e_{\theta}\left [ \left [ \binom{n}{m}\right ] ^{-2}\left ( { \displaystyle\sum\limits_{\#\left ( \overline{i}_{m}\cap\overline{r}_{m}\right ) = t } } \widehat{if}_{m , m,\overline{i}_{m}}^{\left ( s\right ) } \widehat{if}_{m , m,\overline{r}_{m}}^{\left ( s\right ) } \right ) \right ] \\ & = \sum_{t=1}^{m-1}\left [ \binom{n}{m}\right ] ^{-2}\binom{n}{2m - t}\left [ \binom{m}{t}\right ] ^{2}e_{\theta}\left [ \widehat{if}_{m , m , i_{1}i_{2} .. i_{t}i_{t+1} .. i_{m}}^{\left ( s\right ) } \widehat{if}_{m , m , i_{1}i_{2} .. i_{t}i_{m+1} .. i_{2m - t}}^{\left ( s\right ) } \right ] \\ & = o\left ( \sum_{t=1}^{m-1}n^{-t}e_{\theta}\left [ \left ( e_{\theta}\left [ \widehat{if}_{m , m,\overline{i}_{m}}^{\left ( s\right ) } |\mathbf{o}_{\overline{i}_{t}}\right ] \right ) ^{2}\right ] \right ) \\ & = o\left ( \sum_{t=1}^{m-1}n^{-t}\left\ { e_{\theta}\left [ \left ( \widehat{if}_{m , m,\overline{i}_{m}}|o_{s_{1}}, ... o_{s_{t}}\right ) ^{2}\right ] e_{\theta}\left [ \left ( \widehat{if}_{m , m,\overline{i}_{m}}|o_{v_{1}}, ...
o_{v_{t}}\right ) ^{2}\right ] \right\ } ^{1/2}\right ) \\ & = o\left ( \max\left ( \frac{1}{n},\frac{k^{m-2}}{n^{m-1}}\right ) \right)\end{aligned}\ ] ] the last two equalities follow from cauchy - shwartz inequality and lemma [ 3.19_lemma ] .
specifically , @xmath3046 \right ) ^{2}\\ & = e_{\theta}\left [ e_{\theta}\left ( \widehat{if}_{m , m,\overline{i}_{m}^{\ast}}|o_{i_{1}}, ...
o_{i_{t}}\right ) e_{\theta}\left ( \widehat{if}_{m , m,\overline{r}_{m}^{\ast}}|o_{i_{1}}, ...
o_{i_{t}}\right ) \right ] \\ & \leq\left\ { e_{\theta}\left [ \left ( e_{\theta}\left ( \widehat{if}_{m , m,\overline{i}_{m}^{\ast}}|o_{i_{1}}, ...
o_{i_{t}}\right ) \right ) ^{2}\right ] \right\ } ^{1/2}\\ & \times\left\ { e_{\theta}\left [ \left ( e_{\theta}\left ( \widehat{if}_{m , m,\overline{r}_{m}^{\ast}}|o_{i_{1}}, ...
o_{i_{t}}\right ) \right ) ^{2}\right ] \right\ } ^{1/2}.\end{aligned}\ ] ] next , we prove eq .
[ lemma_2 ] for any @xmath3050 here we also rewrite @xmath3051 as a sum of four terms . @xmath3052 -e_{\theta}\left [ \widehat{if}_{m_{1},m_{1},\overline{i}_{m}}^{\left ( s\right ) } \right ] e_{\theta}\left [ \widehat{if}_{m_{2},m_{2},\overline{i}_{m}}^{\left ( s\right ) } \right ] \\ & = e_{\theta}\left [ \frac{1}{\binom{n}{m_{1}}\binom{n}{m_{2}}}\sum _ { i_{1}<i_{2} .. <i_{m_{1}}}\widehat{if}_{m_{1},m_{1},\overline{i}_{m_{1}}}^{\left ( s\right ) } \sum_{r_{1}<r_{2} .. <r_{m_{2}}}\widehat{if}_{m_{2},m_{2},\overline{r}_{m_{2}}}^{\left ( s\right ) } \right ] \\ & -e_{\theta}\left [ \widehat{if}_{m_{1},m_{1},\overline{i}_{m}}^{\left ( s\right ) } \right ] e_{\theta}\left [ \widehat{if}_{m_{2},m_{2},\overline { i}_{m}}^{\left ( s\right ) } \right ] \\ & = e_{\theta}\left [ \frac{1}{\binom{n}{m_{1}}\binom{n}{m_{2}}}\left\ { \begin{array } [ c]{c}\left ( { \displaystyle\sum\limits_{\overline{i}_{m_{1}}\subset\overline{r}_{m_{2 } } } } + { \displaystyle\sum\limits_{\overline{i}_{m_{1}}\cap\overline{r}_{m_{2}}=\emptyset } } + { \displaystyle\sum\limits_{1\leq\#\left ( \overline{i}_{m_{1}}\cap\overline { r}_{m_{2}}\right ) < m_{1 } } } \right ) \\ \times\widehat{if}_{m_{1},m_{1},\overline{i}_{m_{1}}}^{\left ( s\right ) } \widehat{if}_{m_{2},m_{2},\overline{r}_{m_{2}}}^{\left ( s\right ) } \end{array } \right\ } \right ] \\ & -e_{\theta}\left [ \widehat{if}_{m_{1},m_{1},\overline{i}_{m}}^{\left ( s\right ) } \right ] e_{\theta}\left [ \widehat{if}_{m_{2},m_{2},\overline { i}_{m}}^{\left ( s\right ) } \right]\end{aligned}\ ] ] the first term : @xmath3053 \\ & = \frac{1}{\binom{n}{m_{1}}\binom{n}{m_{2}}}\binom{n}{m_{2}}\binom{m_{2}}{m_{1}}e_{\theta}\left [ \widehat{if}_{m_{1},m_{1},\overline{i}_{m_{1}}}^{\left ( s\right ) } \widehat{if}_{m_{2},m_{2},\overline{r}_{m_{2}}}^{\left ( s\right ) } \right ] \\ & = \frac{\binom{m_{2}}{m_{1}}}{\binom{n}{m_{1}}}e_{\theta}\left [ \widehat{if}_{m_{1},m_{1},\overline{i}_{m_{1}}}^{\left ( s\right ) } e_{\theta } \left [ \widehat{if}_{m_{2},m_{2},\overline{r}_{m_{2}}}^{\left ( s\right ) } |\mathbf{o}_{\overline{i}_{m_{1}}}\right ] \right ] \\ & \leq\frac{\binom{m_{2}}{m_{1}}}{\binom{n}{m_{1}}}\left\ { e_{\theta}\left [ \left ( \widehat{if}_{m_{1},m_{1},\overline{i}_{m_{1}}}^{\left ( s\right ) } \right ) ^{2}\right ] e_{\theta}\left [ \left ( e_{\theta}\left [ \widehat{if}_{m_{2},m_{2},\overline{r}_{m_{2}}}^{\left ( s\right ) } |\mathbf{o}_{\overline{i}_{m_{1}}}\right ] \right ) ^{2}\right ] \right\ } ^{1/2}\\ & = o\left ( \frac{k^{m_{1}-1}}{n^{m_{1}}}\right ) , \end{aligned}\ ] ] which follows from cauchy - shwartz inequality , theorem ( [ var_multi ] ) , and lemma [ 3.19_lemma ] . the difference between the second and the fourth terms equals@xmath3054 \\ & -e_{\theta}\left [ \widehat{if}_{m_{1},m_{1},\overline{i}_{m}}^{\left ( s\right ) } \right ] e_{\theta}\left [ \widehat{if}_{m_{2},m_{2},\overline { i}_{m}}^{\left ( s\right ) } \right ] \\ & = \left ( \frac{1}{\binom{n}{m_{1}}\binom{n}{m_{2}}}\binom{n}{m_{1}}\binom{n - m_{1}}{m_{2}}-1\right ) e_{\theta}\left [ \widehat{if}_{m_{1},m_{1},\overline{i}_{m}}^{\left ( s\right ) } \right ] e_{\theta}\left [ \widehat{if}_{m_{2},m_{2},\overline{i}_{m}}^{\left ( s\right ) } \right ] \\ & = o\left ( \frac{1}{n}e_{\theta}\left [ \widehat{if}_{m_{1},m_{1},\overline{i}_{m}}^{\left ( s\right ) } \right ] e_{\theta}\left [ \widehat{if}_{m_{2},m_{2},\overline{i}_{m}}^{\left ( s\right ) } \right ] \right)\end{aligned}\ ] ] the third term:@xmath3055 \\ & = { \displaystyle\sum\limits_{t=1}^{m_{1}-1 } } \frac{1}{\binom{n}{m_{1}}\binom{n}{m_{2}}}e_{\theta}\left [ { \displaystyle\sum\limits_{\#\left ( \overline{i}_{m_{1}}\cap\overline { r}_{m_{2}}\right ) = t } } \widehat{if}_{m_{1},m_{1},\overline{i}_{m_{1}}}^{\left ( s\right ) } \widehat{if}_{m_{2},m_{2},\overline{r}_{m_{2}}}^{\left ( s\right ) } \right ] \\ & = { \displaystyle\sum\limits_{t=1}^{m_{1}-1 } } \left\ { \begin{array } [ c]{c}\frac{\binom{n}{m_{2}+m_{1}-t}\binom{m_{2}+m_{1}-t}{m_{2}}\binom{m_{2}}{t}}{\binom{n}{m_{1}}\binom{n}{m_{2}}}\times\\ e_{\theta}\left [ \widehat{if}_{m_{1},m_{1},i_{1}i_{2} .. i_{t}i_{t+1} .. i_{m_{1}}}^{\left ( s\right ) } \widehat{if}_{m_{2},m_{2},i_{1} .. i_{t}i_{m_{1}+1} .. i_{m_{1}+m_{2}-t}}^{\left ( s\right ) } \right ] \end{array } \right\ } \\ & = o\left ( \max\left ( \frac{1}{n},\frac{k^{m_{1}-2}}{n^{m_{1}-1}}\right ) \right ) , \end{aligned}\ ] ] which also follows from cauchy - shwartz inequality and lemma [ 3.19_lemma ] . for @xmath2211the estimation bias is given by@xmath3057
e\left [ q^{2}\left ( \frac{b-\widehat{b}}{\dot{b}}\right ) \overline{z}_{k}\right ] \right\ } \\ & + \left\ { \begin{array } [ c]{c}e\left [ q^{2}\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) \overline{z}_{k}^{t}\right ] \left\ { e\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\\ \times e\left [ \overline{z}_{k}q^{2}\left ( \frac{b-\widehat{b}}{\dot{b}}\right ) \right ] \end{array } \right\}\end{aligned}\]]@xmath906@xmath2805{c}e\left [ \left ( h_{1}\widehat{p}+h_{2}\right ) \dot{b}\overline{z}_{k}^{t}\right ] \left [ \left\ { e\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}-i\right ] \\
\times e\left [ \overline{z}_{k}\left ( h_{1}\widehat{b}+h_{3}\right ) \dot { p}\right ] \end{array } \right\ } \\ & = -\left\ { \begin{array } [ c]{c}e\left [ \left ( h_{1}\widehat{p}+h_{2}\right ) \dot{b}\overline{z}_{k}^{t}\right ] \left\ { e\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline { z}_{k}^{t}\right ] -i\right\ } \\
\times e\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] ^{-1}e\left [ \overline{z}_{k}\left ( h_{1}\widehat{b}+h_{3}\right ) \dot{p}\right ] \end{array } \right\}\end{aligned}\ ] ] suppose the bias formula holds for @xmath135 , then the bias at @xmath286 is@xmath3058{c}e\left [ q^{2}\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) \overline{z}_{k}^{t}\right ] \left\ { e\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] -i\right\ } \\ \times{\displaystyle\prod\limits_{s=3}^{m } } \left\ { \widehat{e}_{s}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\left\ { e\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] -\widehat{e}_{s}\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ }
\\ \times\left\ { e\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}e\left [ q^{2}\overline{z}_{k}\left ( \frac{b-\widehat{b}}{\dot{b}}\right ) \right ] \end{array } \right\ } \\ & + \left ( -1\right ) ^{m}e\left [ q^{2}\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) \overline{z}_{k}^{t}\right ] \left ( e\left [ \left ( \dot{p}\dot { b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{2}}\right ] -i\right ) \times\\ & \left [ { \displaystyle\prod\limits_{s=3}^{m } } \left\ { \widehat{e}_{s}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\left\ { e\left [ \left ( \dot { p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ) _ { i_{s}}\right ] -\widehat{e}_{s}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } \right ] \\ & \times\left\ { \widehat{e}_{m+1}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}e\left [ \overline{z}_{k}q^{2}\left ( \frac{b-\widehat{b}}{\dot{b}}\right ) \right]\end{aligned}\]]@xmath906@xmath3059{c}e\left [ q^{2}\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) \overline{z}_{k}^{t}\right ] \left\ { e\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] -i\right\ } \\
\times{\displaystyle\prod\limits_{s=3}^{m } } \left\ { \widehat{e}_{s}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\left\ { \begin{array } [ c]{c}e\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \\ -\widehat{e}_{s}\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \end{array } \right\ } \\ \times\left [ \left\ { \widehat{e}_{m+1}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}-\left\ { e\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\right ] \\ \times
e\left [ \overline{z}_{k}q^{2}\left ( \frac{b-\widehat{b}}{\dot{b}}\right ) \right ] \end{array } \right\ } \\ & = \left ( -1\right ) ^{m}\left\ { \begin{array } [ c]{c}e\left [ q^{2}\left ( \frac{p-\widehat{p}}{\dot{p}}\right ) \overline{z}_{k}^{t}\right ] \left\ { e\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] -i\right\ } \\
\times{\displaystyle\prod\limits_{s=3}^{m+1 } } \left\ { \widehat{e}_{s}\left [ \dot{p}\dot{b}h_{1}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}\left\ { \begin{array } [ c]{c}e\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \\ -\widehat{e}_{s}\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \end{array } \right\ } \\
\times\left\ { e\left [ q^{2}\overline{z}_{k}\overline{z}_{k}^{t}\right ] \right\ } ^{-1}e\left [ \overline{z}_{k}q^{2}\left ( \frac{b-\widehat{b}}{\dot{b}}\right ) \right ] \end{array } \right\}\end{aligned}\ ] ] which completes the proof .
since @xmath3060 we can show that @xmath3061 e_{\theta } \left ( \overline{z}_{k\left ( m-1,0\right ) } ^{k\left ( m-1,1\right ) } \widehat{\delta}\right ) \right ) \\ & = \left\ { \widehat{e}\left ( \left ( \delta g+1\right ) \delta b\overline{z}_{k\left ( 1,0\right ) } ^{k\left ( 1,1\right ) t}\right ) { \textstyle\prod\limits_{u=2}^{m-1 } } \widehat{e}\left ( \delta g\widehat{q}^{2}\overline{z}_{k\left ( u-1,0\right ) } ^{k\left ( u-1,1\right ) } \overline{z}_{k\left ( u,0\right ) } ^{k\left ( u,1\right ) t}\right ) \widehat{e}\left ( \left ( \delta g+1\right ) \delta p\overline{z}_{k\left ( m-1,0\right ) } ^{k\left ( m-1,1\right ) } \right ) \right\}\end{aligned}\ ] ] with @xmath1641 @xmath1642 @xmath1643 @xmath3062 and @xmath1645 below we give a useful representation of @xmath3063 .\mathit{\ \ } $ ] let @xmath3064 \widehat{e}\left ( \delta p\overline { z}_{k\left ( m-1,0\right ) } ^{k\left ( m-1,1\right ) } \right ) \right\ } \\ & b_{m+1}^{bg}\left ( \widehat{\mathbb{u}}_{m}\left ( \left ( l\right ) _
{ k\left ( l,0\right ) } ^{k\left ( l,1\right ) } , 1\leq l\leq m-1\right ) \right ) \\ & \equiv\left\ { \widehat{e}\left ( \delta g\delta b\overline{z}_{k\left ( 1,0\right ) } ^{k\left ( 1,1\right ) t}\right ) { \textstyle\prod\limits_{u=2}^{m-1 } } \widehat{e}\left ( \delta g\widehat{q}^{2}\overline{z}_{k\left ( u-1,0\right ) } ^{k\left ( u-1,1\right ) } \overline{z}_{k\left ( u,0\right ) } ^{k\left ( u,1\right ) t}\right ) \widehat{e}\left ( \delta p\overline{z}_{k\left ( m-1,0\right ) } ^{k\left ( m-1,1\right ) } \right ) \right\}\end{aligned}\]]@xmath3065 [ cp]assume @xmath3067 is a @xmath3068 dimensional matrix with @xmath3069 @xmath3070 and @xmath3071 @xmath103for @xmath1939 @xmath3072 if @xmath3073{c}\left ( \begin{array } [ c]{c}b_{t}\left ( \widehat{\mathbb{u}}_{t-2}\left ( \left ( l\right ) _ { k_{1}\left ( l,0\right ) } ^{k_{1}\left ( l,1\right ) } , 1\leq l\leq t-3\right ) \right ) \\ -b_{t}^{bg}\left ( \widehat{\mathbb{u}}_{t-1}\left ( \left ( l\right ) _ { k_{1}\left ( l,0\right ) } ^{k_{1}\left ( l,1\right ) } , \left ( t-2\right ) _ { k_{1}\left ( t-2,0\right ) } ^{k_{1}\left ( t-2,0\right ) } , 1\leq l\leq t-3\right ) \right ) \end{array } \right ) \\ -\left ( \begin{array } [ c]{c}b_{t}^{pg}\left ( \widehat{\mathbb{u}}_{t-1}\left ( \left ( 1\right ) _ { k_{1}\left ( 0,0\right ) } ^{k_{1}\left ( 0,1\right ) } , \left ( l+1\right ) _ { k_{1}\left ( l,0\right ) } ^{k_{1}\left ( l,1\right ) } , 1\leq l\leq t-3\right ) \right ) \\ -b_{t}\left ( \widehat{\mathbb{u}}_{t}\left ( \left ( 1\right ) _ { k_{1}\left ( 0,0\right ) } ^{k_{1}\left ( 0,1\right ) } , \left ( l+1\right ) _ { k_{1}\left ( l,0\right ) } ^{k_{1}\left ( l,1\right ) } , \left ( t-1\right ) _ { k_{1}\left ( t-2,0\right ) } ^{k_{1}\left ( t-2,0\right ) } , 1\leq l\leq t-3\right ) \right ) \end{array } \right ) \end{array } \right\}\end{gathered}\ ] ] this lemma explains how to use higher order u - statistics to estimate the @xmath781th order contribution of @xmath3075 with a residual bias not exceeding @xmath3076
our estimator uses this idea to reduce fourth and higher order estimation bias to the optimal rate .
( lemma [ cp])@xmath3077{c}\widehat{e}\left ( \delta b\delta g\overline{z}_{k_{1}\left ( 1,0\right ) } ^{k_{1}\left ( 1,1\right ) t}\right ) { \textstyle\prod\limits_{u=2}^{t-3 } } \widehat{e}\left ( \delta g\widehat{q}^{2}\overline{z}_{k_{1}\left ( u-1,0\right ) } ^{k_{1}\left ( u-1,1\right ) } \overline{z}_{k_{1}\left ( u,0\right ) } ^{k_{1}\left ( u,1\right ) t}\right ) \times\\ \widehat{e}\left ( \delta g\widehat{q}\overline{z}_{k_{1}\left ( t-3,0\right ) } ^{k_{1}\left ( t-3,1\right ) } \left ( \widehat{q}^{-1}\delta p-\widehat{\pi } \left ( \widehat{q}^{-1}\delta p|\widehat{q}\overline{z}_{k_{1}\left ( t-2,0\right ) } ^{k_{1}\left ( t-2,1\right ) } \right ) \right ) \right ) \end{array } \right\ } \\ & -\left\ { \begin{array } [ c]{c}\widehat{e}\left ( \widehat{q}\delta g\overline{z}_{k_{1}\left ( 1,0\right ) } ^{k_{1}\left ( 1,1\right ) t}\widehat{\pi}\left ( \widehat{q}^{-1}\delta b|\left ( \widehat{q}\overline{z}_{k_{1}\left ( 0,0\right ) } ^{k_{1}\left ( 0,1\right ) } \right ) \right ) \right ) { \textstyle\prod\limits_{u=2}^{t-3 } } \widehat{e}\left ( \delta g\widehat{q}^{2}\overline{z}_{k_{1}\left ( u-1,0\right ) } ^{k_{1}\left ( u-1,1\right ) } \overline{z}_{k_{1}\left ( u,0\right ) } ^{k_{1}\left ( u,1\right ) t}\right ) \\ \times\widehat{e}\left ( \delta g\widehat{q}\overline{z}_{k_{1}\left ( t-3,0\right ) } ^{k_{1}\left ( t-3,1\right ) } \left ( \widehat{q}^{-1}\delta p-\widehat{\pi}\left ( \widehat{q}^{-1}\delta p|\left ( \widehat{q}\overline{z}_{k_{1}\left ( t-2,0\right ) } ^{k_{1}\left ( t-2,1\right ) } \right ) \right ) \right ) \right ) \end{array } \right\ } \\ & = \left\ { \begin{array } [ c]{c}\widehat{e}\left ( \widehat{q}\delta g\overline{z}_{k_{1}\left ( 1,0\right ) } ^{k_{1}\left ( 1,1\right ) t}\widehat{\pi}^{\bot}\left ( \widehat{q}^{-1}\delta b|\left ( \widehat{q}\overline{z}_{k_{1}\left ( 0,0\right ) } ^{k_{1}\left ( 0,1\right ) } \right ) \right ) \right ) { \textstyle\prod\limits_{u=2}^{t-3 } } \widehat{e}\left ( \delta g\widehat{q}^{2}\overline{z}_{k_{1}\left ( u-1,0\right ) } ^{k_{1}\left ( u-1,1\right ) } \overline{z}_{k_{1}\left ( u,0\right ) } ^{k_{1}\left ( u,1\right ) t}\right ) \\ \times\widehat{e}\left ( \delta g\widehat{q}\overline{z}_{k_{1}\left ( t-3,0\right ) } ^{k_{1}\left ( t-3,1\right ) } \widehat{\pi}^{\bot}\left ( \widehat{q}^{-1}\delta p|\left ( \widehat{q}\overline{z}_{k_{1}\left ( t-2,0\right ) } ^{k_{1}\left ( t-2,1\right ) } \right ) \right ) \right ) \end{array } \right\ } \\ & = \left\ { \widehat{e}\left ( \delta g\left [ \left ( { \textstyle\prod\limits_{u=1}^{t-3 } } r_{u}\right ) \left ( \widehat{\pi}^{\bot}\left ( \widehat{q}^{-1}\delta b|\left ( \widehat{q}\overline{z}_{k_{1}\left ( 0,0\right ) } ^{k_{1}\left ( 0,1\right ) } \right ) \right ) \right ) \right ] \widehat{\pi}^{\bot}\left ( \widehat{q}^{-1}\delta p|\left ( \widehat{q}\overline{z}_{k_{1}\left ( t-2,0\right ) } ^{k_{1}\left ( t-2,1\right ) } \right ) \right ) \right ) \right\}\end{aligned}\ ] ] where@xmath3078 @xmath3079 , $ ]
@xmath3080 @xmath3081 since projection operator has operator norm of 1 , we have @xmath3082{c}\left\vert \left\vert \delta g\right\vert \right\vert _ { \infty}^{t-2}\left\ { \widehat{e}\left ( \widehat{\pi}^{\bot}\left ( \widehat{q}^{-1}\delta b|\left ( \widehat{q}\overline{z}_{0}^{k_{1}\left ( 0,1\right ) } \right )
\right ) \right ) ^{2}\right\ } ^{1/2}\\ \times\left\ { \widehat{e}\left ( \widehat{\pi}^{\bot}\left ( \widehat{q}^{-1}\delta p|\left ( \widehat{q}\overline{z}_{0}^{k_{1}\left ( t-2,1\right ) } \right ) \right ) \right ) ^{2}\right\ } ^{1/2}\end{array } \right ] \\ & = o_{p}\left ( \left ( \frac{\log n}{n}\right ) ^{-\frac{\left ( t-2\right ) \beta_{g}}{d+2\beta_{g}}}\left ( k_{1}\left ( 0,1\right ) \right ) ^{-\frac{\beta_{b}}{d}}\left ( k_{1}\left ( t-2,1\right ) \right ) ^{-\frac{\beta p}{d}}\right)\end{aligned}\ ] ] moreover@xmath3089 \\ & = \widehat{e}\left ( \delta g\widehat{\pi}\left ( \widehat{q}^{-1}\delta b|\left ( \widehat{q}\overline{z}_{k\left ( 1,0\right ) } ^{k\left ( 1,1\right ) } \right ) \right ) \widehat{\pi}\left ( \widehat{q}^{-1}\delta p|\left ( \widehat{q}\overline{z}_{k\left ( 2,0\right ) } ^{k\left ( 2,1\right ) } \right ) \right ) \right)\end{aligned}\ ] ] ( theorem [ beyond4th ] ) let @xmath3098 for @xmath3099 and @xmath3100 . for @xmath1939
@xmath3101 define @xmath3102{c}b_{t}\left ( \mathbb{g}\left ( s , t-2\right ) \right ) -b_{t}^{bg}\left ( \mathbb{g}\left ( s , t-1\right ) \right ) \\
-b_{t}^{pg}\left ( \mathbb{g}\left ( s , t-1\right ) \right ) + b_{t}\left ( \mathbb{g}\left ( s , t\right ) \right ) \end{array } \right ) \\ b_{t}^{\left ( q\right ) } & = \left ( -1\right ) ^{t-1}\left ( b_{t}\left ( \mathbb{q}_{t-2}\right ) -b_{t}^{bg}\left ( \mathbb{q}_{t-1}\right ) -b_{t}^{pg}\left ( \mathbb{q}_{t-1}\right ) + b_{t}\left ( \mathbb{q}_{t}\right ) \right)\end{aligned}\ ] ] then @xmath3103 _ { -\left ( l4.1\right ) } \\ & -b_{2}\left ( \mathbb{h}_{2}^{\ast}\right ) + b_{3}\left ( \mathbb{h}_{3}^{\ast}-\mathbb{h}_{2}^{\ast\ } \right ) + { \displaystyle\sum\limits_{s=1}^{j } } b_{3}\left ( \mathbb{g}\left ( s,3\right ) \right ) + b_{3}\left ( \mathbb{q}_{3}\right ) \\ & + \sum_{t=4}^{m\left ( \beta_{g},\beta_{b},\beta_{p}\right ) + 2}\left ( b_{t}^{\left ( h^{\ast}\right ) } + b_{t}^{\left ( g\right ) } + b_{t}^{\left ( q\right ) } \right)\end{aligned}\]]@xmath3104 \\ + \widehat{e}\left ( \widehat{e}\left ( h_{1}|x\right ) \left ( b-\widehat{b}\right ) \left ( p-\widehat{p}\right ) \right ) \\ = \widehat{e}\left ( \delta g\widehat{q}^{-2}\delta b\delta p\right ) _ { -\left ( l4.2\right ) } + \widehat{e}\left ( \widehat{q}^{-2}\delta b\delta p\right ) _ { -\left ( l4.3\right ) } \ ] ] @xmath3105 \\ & = \widehat{e}\left [ \widehat{\pi}^{\bot}\left ( \widehat{q}^{-1}\delta b|\left ( \widehat{q}\overline{z}_{0}^{k_{-1}}\right ) \right ) \widehat{\pi } ^{\bot}\left ( \widehat{q}^{-1}\delta p|\left ( \widehat{q}\overline{z}_{0}^{k_{-1}}\right ) \right ) \right]\end{aligned}\ ] ] then @xmath3106@xmath3107{c}\widehat{e}\left ( \widehat{q}^{-2}\delta b\delta p\delta
g\right ) -\widehat{e}\left ( \delta b\delta g\overline{z}_{0}^{k_{-1}t}\right ) \widehat{e}\left ( \delta p\overline{z}_{0}^{k_{-1}}\right ) \\ -\widehat{e}\left ( \delta b\overline{z}_{0}^{k_{-1}t}\right ) \widehat{e}\left ( \delta p\delta g\overline{z}_{0}^{k_{-1}}\right ) \end{array } \right\ } _ { -\left ( l5.1\right ) } \\ & + \widehat{e}\left ( \delta b\overline{z}_{0}^{k_{0}t}\right ) \widehat{e}\left ( \delta g\widehat{q}^{2}\overline{z}_{0}^{k_{0}}\overline { z}_{0}^{k_{0}t}\right ) \widehat{e}\left [ \delta p\overline{z}_{0}^{k_{0}}\right ] \\ & + \widehat{e}\left ( \delta b\overline{z}_{k_{0}}^{k_{-1}t}\right ) \widehat{e}\left ( \delta g\widehat{q}^{2}\overline{z}_{k_{0}}^{k_{-1}}\overline{z}_{0}^{k_{0}t}\right ) \widehat{e}\left [ \delta p\overline{z}_{0}^{k_{0}}\right ] \\ & + \widehat{e}\left ( \delta b\overline{z}_{0}^{k_{0}t}\right ) \widehat{e}\left ( \delta g\widehat{q}^{2}\overline{z}_{0}^{k_{0}}\overline { z}_{k_{0}}^{k_{-1}t}\right ) \widehat{e}\left [ \delta p\overline{z}_{k_{0}}^{k_{-1}}\right ] \\ & + \sum_{s=1}^{j}\left\ { \begin{array } [ c]{c}\widehat{e}\left ( \delta b\overline{z}_{k_{2s-2}}^{k_{2s-1}t}\right ) \widehat{e}\left ( \delta g\widehat{q}^{2}\overline{z}_{k_{2s-2}}^{k_{2s-1}}\overline{z}_{k_{2s-2}}^{k_{2s}t}\right ) \widehat{e}\left [ \delta p\overline{z}_{k_{2s-2}}^{k_{2s}}\right ] \\ + \widehat{e}\left ( \delta b\overline{z}_{k_{2s-2}}^{k_{2s}t}\right ) \widehat{e}\left ( \delta g\widehat{q}^{2}\overline{z}_{k_{2s-2}}^{k_{2s}}\overline{z}_{k_{2s}}^{k_{2s-1}t}\right ) \widehat{e}\left [ \delta p\overline{z}_{k_{2s}}^{k_{2s-1}}\right ] \end{array } \right\ } \\ & + \widehat{e}\left ( \delta b\overline{z}_{k_{2j}}^{k_{2j+1}t}\right ) \widehat{e}\left ( \delta g\widehat{q}^{2}\overline{z}_{k_{2j}}^{k_{2j+1}}\overline{z}_{k_{2j}}^{k_{2j+1}t}\right ) \widehat{e}\left [ \delta p\overline{z}_{k_{2j}}^{k_{2j+1}}\right]\end{aligned}\ ] ] because @xmath3108{c}\delta g\left\ { \widehat{\pi}\left ( \widehat{q}^{-1}\delta b|\left ( \widehat{q}\overline{z}_{0}^{k_{-1}}\right ) \right ) + \widehat{\pi}^{\bot } \left ( \widehat{q}^{-1}\delta b|\left ( \widehat{q}\overline{z}_{0}^{k_{-1}}\right ) \right ) \right\ } \\
\times\left\ { \widehat{\pi}\left ( \widehat{q}^{-1}\delta p|\left ( \widehat{q}\overline{z}_{0}^{k_{-1}}\right ) \right ) + \widehat{\pi}^{\bot } \left ( \widehat{q}^{-1}\delta p|\left ( \widehat{q}\overline{z}_{0}^{k_{-1}}\right ) \right ) \right\ } \end{array } \right ) \\ & -\widehat{e}\left ( \delta g\delta p\widehat{q}^{-1}\widehat{\pi}\left ( \widehat{q}^{-1}\delta b|\left ( \widehat{q}\overline{z}_{0}^{k_{-1}}\right ) \right ) \right ) \\ & -\widehat{e}\left ( \delta g\delta b\widehat{q}^{-1}\widehat{\pi}\left ( \widehat{q}^{-1}\delta p|\left ( \widehat{q}\overline{z}_{0}^{k_{-1}}\right ) \right ) \right ) \\ & = \widehat{e}\left ( \delta g\widehat{\pi}^{\bot}\left ( \widehat{q}^{-1}\delta b|\left ( \widehat{q}\overline{z}_{0}^{k_{-1}}\right ) \right ) \widehat{\pi}^{\bot}\left ( \widehat{q}^{-1}\delta p|\left ( \widehat{q}\overline{z}_{0}^{k_{-1}}\right ) \right ) \right ) \\ & -\widehat{e}\left ( \delta g\widehat{\pi}\left ( \widehat{q}^{-1}\delta b|\left ( \widehat{q}\overline{z}_{0}^{k_{-1}}\right ) \right ) \widehat{\pi } \left ( \widehat{q}^{-1}\delta p|\left ( \widehat{q}\overline{z}_{0}^{k_{-1}}\right ) \right ) \right)\end{aligned}\ ] ] and by lemma [ on2 ] , we now have @xmath3109{c}\widehat{e}\left [ \delta g\widehat{\pi}\left ( \widehat{q}^{-1}\delta b|\left ( \widehat{q}\overline{z}_{k_{2s}}^{k_{2s-1}}\right ) \right ) \widehat{\pi}\left ( \widehat{q}^{-1}\delta p|\left ( \widehat{q}\overline { z}_{k_{2s+1}}^{k_{2s-1}}\right ) \right ) \right ] _ { -\left ( l5.3\right ) } \\ + \widehat{e}\left [ \delta g\widehat{\pi}\left ( \widehat{q}^{-1}\delta b|\left ( \widehat{q}\overline{z}_{k_{2s+1}}^{k_{2s-1}}\right ) \right ) \widehat{\pi}\left ( \widehat{q}^{-1}\delta p|\left ( \widehat{q}\overline { z}_{k_{2s}}^{k_{2s+1}}\right ) \right ) \right ] _ { -\left ( l5.4\right ) } \end{array}\end{aligned}\ ] ] therefore it is easy to show from theorem [ tbrate ] that @xmath3110 @xmath1939 @xmath3111@xmath3112{c}b_{t}\left ( \widehat{\mathbb{u}}_{t-2}\left ( _ { 0}^{k_{0}}\right ) \right ) -b_{t}^{bg}\left ( \widehat{\mathbb{u}}_{t-1}\left ( _ { 0}^{k_{0}}\right ) \right ) \\ -\left [ b_{t}^{pg}\left ( \widehat{\mathbb{u}}_{t-1}\left ( _ { 0}^{k_{0}}\right ) \right ) -b_{t}\left ( \widehat{\mathbb{u}}_{t}\left ( _ { 0}^{k_{0}}\right ) \right ) \right ] \end{array } \right\ } _ { -\left ( l6.1\right ) } \\ & + { \displaystyle\sum\limits_{u=1}^{t-3 } } \left\ { \begin{array } [ c]{c}\left [ b_{t}\left ( \widehat{\mathbb{u}}_{t-2}^{\left ( u\right ) } \left ( _ { k_{0}}^{k_{-1}},_{0}^{k_{0}}\right ) \right ) -b_{t}^{bg}\left ( \widehat{\mathbb{u}}_{t-1}^{\left ( u\right ) } \left ( _ { k_{0}}^{k_{-1}},_{0}^{k_{0}}\right ) \right ) \right ] \\ -\left [ b_{t}^{pg}\left ( \widehat{\mathbb{u}}_{t-1}^{\left ( u+1\right ) } \left ( _ { k_{0}}^{k_{-1}},_{0}^{k_{0}}\right ) \right ) -b_{t}\left ( \widehat{\mathbb{u}}_{t}^{\left ( u+1\right ) } \left ( _ { k_{0}}^{k_{-1}},_{0}^{k_{0}}\right ) \right ) \right ] \end{array } \right\ } _ { -\left ( l6.2.u\right ) } \\ & -b_{t}^{pg}\left ( \widehat{\mathbb{u}}_{t-1}^{\left ( 1\right ) } \left ( _ { k_{0}}^{k_{-1}},_{0}^{k_{0}}\right ) \right ) -b_{t}\left ( \widehat{\mathbb{u}}_{t}^{\left ( 1\right ) } \left ( _ { k_{0}}^{k_{-1}},_{0}^{k_{0}}\right ) \right ) _ { -\left ( l6.3\right ) } \\ & -b_{t}^{bg}\left ( \widehat{\mathbb{u}}_{t-1}^{\left ( t-2\right ) } \left ( _ { k_{0}}^{k_{-1}},_{0}^{k_{0}}\right ) \right ) -b_{t}\left ( \widehat{\mathbb{u}}_{t}^{\left ( t-1\right ) } \left ( _ { k_{0}}^{k_{-1}},_{0}^{k_{0}}\right ) \right ) _ { -\left ( l6.4\right ) } \ ] ] it can be shown that @xmath3113 with @xmath3114 @xmath3115 @xmath3116 @xmath3117 with @xmath3118 @xmath3119 for @xmath1828 and @xmath3120 @xmath3121 and @xmath3122{c}\widehat{e}\left ( \delta g\widehat{q}\widehat{\pi}\left ( \widehat{q}^{-1}\delta b|\left ( \widehat{q}\overline{z}_{k_{0}}^{k_{-1}}\right ) \right ) \overline{z}_{0}^{k_{0}t}\right ) \left ( \widehat{e}\left ( \delta g\widehat{q}^{2}\overline{z}_{0}^{k_{0}}\overline{z}_{0}^{k_{0}t}\right ) \right ) ^{t-4}\\ \times e\left ( \delta g\widehat{q}\overline{z}_{0}^{k_{0}}\widehat{\pi}^{\bot}\left ( \widehat{q}^{-1}\delta b|\left ( \widehat{q}\overline{z}_{0}^{k_{0}}\right ) \right ) \right ) \end{array } \right\ } \right\vert \\ & = o_{p}\left ( \left ( \frac{\log n}{n}\right ) ^{-\frac{\left ( t-2\right ) \beta_{g}}{d+2\beta_{g}}}k_{0}^{-2\beta / d}\right)\end{aligned}\ ] ] similarly @xmath3123 from lemma [ cp ] , we now have@xmath3124 therefore @xmath3125 @xmath1939 @xmath3126 following the same argument as above , we know as well @xmath3127 @xmath1939 @xmath3128@xmath3129 from lemma [ on2 ] @xmath3130\end{aligned}\]]@xmath3131 \\ & = \widehat{e}\left ( \widehat{\pi}^{\bot}\left ( \widehat{q}^{-1}\delta b|\left ( \widehat{q}\overline{z}_{0}^{k_{0}}\right ) \right ) \delta g\overline{z}_{k_{2s-2}}^{k_{2s-1}t}\right ) \widehat{e}\left ( \delta g\widehat{q}\overline{z}_{k_{2s-2}}^{k_{2s-1}}\widehat{\pi}\left ( \widehat{q}^{-1}\delta p|\left ( \widehat{q}\overline{z}_{k_{2s-2}}^{k_{2s}}\right ) \right ) \right ) \\ & + \widehat{e}\left ( \widehat{\pi}\left ( \widehat{q}^{-1}\delta b|\left ( \widehat{q}\overline{z}_{k_{2s-2}}^{k_{2s-1}}\right ) \right ) \delta g\overline{z}_{k_{2s-2}}^{k_{2s}t}\right ) \widehat{e}\left ( \delta g\widehat{q}\overline{z}_{k_{2s-2}}^{k_{2s}}\widehat{\pi}^{\bot}\left ( \widehat{q}^{-1}\delta p|\left ( \widehat{q}\overline{z}_{0}^{k_{0}}\right ) \right ) \right)\end{aligned}\ ] ] similarly,@xmath3132 therefore@xmath3133 applying the above argument to @xmath3134 we can show that : @xmath3135 in addition , @xmath1939 @xmath3136 @xmath3137 @xmath3138 which completes the proof of bias
. the order of variance follows directly from theorem [ var_multi ] .
( theorem [ tt ] ( iii ) ) as proved in ( ii ) , @xmath70@xmath3139 by part 5c ) of theorem [ eift ] , consider any suitably smooth one dimensional parametric submodel @xmath243 with range containing @xmath179 and contained in @xmath3140 , and differentiate both sides of eq .
( [ if1_tau ] ) wrt .
@xmath2752 then , @xmath3141 \\ & -e_{\theta}\left [ if_{1,if_{1,\psi}\left ( o;\cdot\right ) } \left ( o_{2};\theta\right ) s_{\zeta}\left ( \theta\right ) \right ] .\end{aligned}\ ] ] further , @xmath3142 $ ] and @xmath3143 } { \partial\tau}|_{\tau=\tau^{\ast}}\\ & = e_{\theta}\left [ \frac{\partial if_{1,\psi\left ( \tau^{\dagger},\cdot\right ) } \left ( o_{2};\theta\right ) } { \partial\tau}s_{\zeta}\left ( \theta\right ) \right]\end{aligned}\ ] ] thus , @xmath3144 \\ & -e_{\theta}\left [ if_{1,if_{1,\psi}\left ( o;\cdot\right ) } \left ( o_{2},\theta\right ) s_{\zeta}\left ( \theta\right ) \right ] \\ & -if_{1,\tau\left ( \cdot\right ) } \left ( o;\theta\right ) \left ( \begin{array } [ c]{c}\psi_{\tau\tau}e_{\theta}\left [ if_{1,\tau\left ( \cdot\right ) } \left ( o_{2},\theta\right ) s_{\zeta}\left ( \theta\right ) \right ]
\\ + e_{\theta}\left [ \frac{\partial if_{1,\psi\left ( \tau^{\dagger},\cdot\right ) } \left ( o_{2};\theta\right ) } { \partial\tau}s_{\zeta}\left ( \theta\right ) \right ] \end{array } \right)\end{aligned}\ ] ] for any @xmath3145 wp
that is , there exists a first order influence function for @xmath3146 , and @xmath3147 \right\ } \\ & = -\psi_{\tau}^{-1}\left\ { \begin{array } [ c]{c}\mathbb{if}_{2,2,\tau\left ( \cdot\right ) } \left ( \theta\right ) + \frac { 1}{2}\psi_{\tau\tau}\mathbb{v}\left ( if_{1,\tau\left ( \cdot\right ) } \left ( o_{1};\theta\right ) if_{1,\tau\left ( \cdot\right ) } \left ( o_{2},\theta\right ) \right ) \\
+ \frac{1}{2}\mathbb{v}\left ( \begin{array } [ c]{c}if_{1,\tau\left ( \cdot\right ) } \left ( o_{1};\theta\right ) d_{1,\theta } \left ( \frac{\partial if_{1,\psi\left ( \tau^{\dagger},\cdot\right ) } \left ( o_{2};\theta\right ) } { \partial\tau}\right ) \\ + d_{1,\theta}\left ( \frac{\partial if_{1,\psi\left ( \tau^{\dagger},\cdot\right ) } \left ( o_{1};\theta\right ) } { \partial\tau}\right ) if_{1,\tau\left ( \cdot\right ) } \left ( o_{2};\theta\right ) \end{array } \right ) \end{array } \right\}\end{aligned}\ ] ] which completes the proof .
note that @xmath3148 is defined in eq .
( [ deg ] ) .
( ii ) : to obtain eq ( [ joel ] ) , note by theorem [ gg ] , we have @xmath3155 ^{-1}\times\\ & \pi_{\widehat{\theta}\left ( \tau^{\dagger}\right ) } \left [ \mathbb{if}_{2,\widetilde{\psi}_{k}\left ( \tau^{\dagger},\cdot\right ) } \left ( \widehat{\theta}\left ( \tau^{\dagger}\right ) \right ) |\gamma_{2}^{test}\left ( \widehat{\theta}\left ( \tau^{\dagger}\right ) , \tau^{\dagger } \right ) \right]\end{aligned}\ ] ] but by theorem [ tt ] and the definition of @xmath3156 we have @xmath3157 ^{-1}\\ \times\left\ { y^{\ast}\left ( \tau^{\dagger}\right ) -\widehat{b}\left ( x,\tau^{\dagger}\right ) \right\ } \left\ { a-\widehat{p}\left ( x\right ) \right\}\end{gathered}\ ] ] thus , we obtain @xmath3158 ^{-1}=v\left ( \widehat{\theta}\left ( \tau^{\dagger}\right ) \right ) ^{-1}$ ] now @xmath3159 \\ & = \mathbb{if}_{2,\widetilde{\psi}_{k}\left ( \tau^{\dagger},\cdot\right ) } \left ( \widehat{\theta}\left ( \tau^{\dagger}\right ) \right ) -\pi_{\widehat{\theta}}\left [ \mathbb{if}_{2,\widetilde{\psi}_{k}\left ( \tau^{\dagger},\cdot\right ) } \left ( \widehat{\theta}\left ( \tau^{\dagger } \right ) \right ) |\left\ { \mathbb{u}_{2,2,\widetilde{\tau}_{k}\left ( \cdot\right ) } ^{test,\perp}\left ( \widehat{\theta}\left ( \tau^{\dagger } \right ) , \tau^{\dagger}\right ) \right\ } \right]\end{aligned}\ ] ] let @xmath2293 denote @xmath2294 and @xmath2295 denote @xmath2296 next , we show that @xmath3160 = \mathbb{u}_{2,2,\widetilde{\tau}_{k}\left ( \cdot\right ) } ^{\ast , test,\perp}\left ( \widehat{\theta}\left ( \tau^{\dagger}\right ) , \tau^{\dagger}\right)\ ] ] where@xmath2298 \right ) ^{-1}\times\widehat{\epsilon}_{i}\widehat{\delta}_{i}\\ & \left\ { \begin{array } [ c]{c}-\left\ { \begin{array } [ c]{c}\left ( e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}^{2}\widehat{\delta } _ { i}^{2}\right ] \right ) ^{-1}e_{\widehat{\theta}}\left [ \widehat{\epsilon } \widehat{\delta}^{2}\overline{z}_{k}^{t}\right ] \\
\times e_{\widehat{\theta}}\left [ \widehat{\epsilon}^{2}\widehat{\delta } \overline{z}_{k}^{t}\right ] \widehat{\epsilon}_{j}\widehat{\delta}_{j}\end{array } \right\ } \\ + e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}^{2}\widehat{\delta}_{i}\overline{z}_{k , i}^{t}\right ] \overline{z}_{k , j}\widehat{\delta}_{j}\\ + e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}\widehat{\delta}_{i}^{2}\overline{z}_{k , i}^{t}\right ] \overline{z}_{k , j}\widehat{\epsilon}_{j}\end{array } \right\}\end{aligned}\ ] ] as proved in theorem [ ff],@xmath3161 = 0\right\}\ ] ] we assume that @xmath3162 \\ & = \mathbb{v}\left\ { if_{1,\tau\left ( \cdot\right ) , i}^{eff}\left ( \theta\right ) h^{\ast}\left ( o_{j};\theta\right ) \right\ } , \end{aligned}\ ] ] then by the definition of the projection , for any @xmath3163 such that @xmath3164 = 0,$ ] we have @xmath3165 \\ & = e_{\widehat{\theta}}\left [ \mathbb{v}\left\ { if_{1,\tau\left ( \cdot\right ) , i}^{eff}\left ( \theta\right ) h^{\ast}\left ( o_{j};\theta\right ) \right\ } \mathbb{v}\left\ { if_{1,\tau\left ( \cdot\right ) , i}^{eff}\left ( \theta\right ) h\left ( o_{j};\theta\right ) \right\ } \right ] , \end{aligned}\ ] ] which is equivalent to @xmath3166 + e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}\widehat{\delta}_{i}^{2}\overline{z}_{k , i}^{t}\overline{z}_{k , j}\widehat{\epsilon}_{j}h\left ( o_{j}\right ) \right ] \right\ } \\ & = v\left ( \widehat{\theta}\right ) ^{-2}\left\ { e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}^{2}\widehat{\delta}_{i}^{2}h^{\ast}\left ( o_{j}\right ) h\left ( o_{j}\right ) \right ] + e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}\widehat{\delta}_{i}h\left ( o_{i}\right ) \widehat{\epsilon}_{j}\widehat{\delta}_{j}h^{\ast}\left ( o_{j}\right ) \right ] \right\ } .\end{aligned}\ ] ] as the equation above holds for any mean zero function @xmath3167 therefore@xmath3168 h^{\ast}\left ( o\right ) + \widehat{\epsilon } \widehat{\delta}e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{j}\widehat{\delta}_{j}h^{\ast}\left ( o_{j}\right ) \right ] \right\ } \\ & = v\left ( \widehat{\theta}\right ) \left\ { e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}^{2}\widehat{\delta}_{i}\overline{z}_{k , i}^{t}\right ] \overline{z}_{k}\widehat{\delta}+e_{\widehat{\theta}}\left [ \widehat{\epsilon } _ { i}\widehat{\delta}_{i}^{2}\overline{z}_{k , i}^{t}\right ] \overline{z}_{k}\widehat{\epsilon}\right\}\end{aligned}\]]@xmath3169@xmath3170 \right ) ^{-1}\left\ { \begin{array } [ c]{c}c_{h}\left ( \widehat{\theta}\right ) \widehat{\epsilon}\widehat{\delta}\\ + v\left ( \widehat{\theta}\right ) e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}^{2}\widehat{\delta}_{i}\overline{z}_{k , i}^{t}\right ] \overline{z}_{k}\widehat{\delta}\\ + v\left ( \widehat{\theta}\right ) e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}\widehat{\delta}_{i}^{2}\overline{z}_{k , i}^{t}\right ] \overline{z}_{k}\widehat{\epsilon}\end{array } \right\}\end{aligned}\ ] ] and @xmath3171 is determined by the following equation @xmath3172 \overline{z}_{k}\widehat{\delta}\\ & + v\left ( \widehat{\theta}\right ) e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}\widehat{\delta}_{i}^{2}\overline{z}_{k , i}^{t}\right ] \overline{z}_{k}\widehat{\epsilon}\\ & + \widehat{\epsilon}\widehat{\delta}e_{\widehat{\theta}}\left [ \widehat{\epsilon}\widehat{\delta}\left ( e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}^{2}\widehat{\delta}_{i}^{2}\right ] \right ) ^{-1}\left\ { \begin{array } [ c]{c}c_{h}\left ( \widehat{\theta}\right ) \widehat{\epsilon}\widehat{\delta}\\ + v\left ( \widehat{\theta}\right ) e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}^{2}\widehat{\delta}_{i}\overline{z}_{k , i}^{t}\right ] \overline{z}_{k}\widehat{\delta}\\ + v\left ( \widehat{\theta}\right ) e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}\widehat{\delta}_{i}^{2}\overline{z}_{k , i}^{t}\right ] \overline{z}_{k}\widehat{\epsilon}\end{array } \right\ } \right ] \\ & = v\left ( \widehat{\theta}\right ) \left\ { e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}^{2}\widehat{\delta}_{i}\overline{z}_{k , i}^{t}\right ] \overline{z}_{k}\widehat{\delta}+e_{\widehat{\theta}}\left [ \widehat{\epsilon } _ { i}\widehat{\delta}_{i}^{2}\overline{z}_{k , i}^{t}\right ] \overline{z}_{k}\widehat{\epsilon}\right\}\end{aligned}\]]@xmath3169@xmath3173{c}c_{h}\left ( \widehat{\theta}\right ) + \left ( e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}^{2}\widehat{\delta}_{i}^{2}\right ] \right ) ^{-1}\times\\ \left\ { \begin{array } [ c]{c}e_{\widehat{\theta}}\left [ \widehat{\epsilon}^{2}\widehat{\delta}^{2}\right ] c_{h}\left ( \widehat{\theta}\right ) \\ + 2v\left ( \widehat{\theta}\right )
e_{\widehat{\theta}}\left [ \widehat{\epsilon}\widehat{\delta}^{2}\overline{z}_{k}^{t}\right ] e_{\widehat{\theta}}\left [ \widehat{\epsilon}^{2}\widehat{\delta}\overline { z}_{k}^{t}\right ] \end{array } \right\ } \end{array } \right ] = 0\]]@xmath3169@xmath3174 \right ) ^{-1}e_{\widehat{\theta}}\left [ \widehat{\epsilon}\widehat{\delta}^{2}\overline{z}_{k}^{t}\right ] e_{\widehat{\theta}}\left [ \widehat{\epsilon } ^{2}\widehat{\delta}\overline{z}_{k}^{t}\right]\end{aligned}\ ] ] in summary,@xmath3175 \right ) ^{-1}\times\\ & \mathbb{v}\left\ { \begin{array } [ c]{c}\widehat{\epsilon}_{i}\widehat{\delta}_{i}\times\\ \left\ { \begin{array } [ c]{c}-\left\ { \begin{array } [ c]{c}\left ( e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}^{2}\widehat{\delta } _ { i}^{2}\right ] \right ) ^{-1}e_{\widehat{\theta}}\left [ \widehat{\epsilon } \widehat{\delta}^{2}\overline{z}_{k}^{t}\right ] \\
\times e_{\widehat{\theta}}\left [ \widehat{\epsilon}^{2}\widehat{\delta } \overline{z}_{k}^{t}\right ] \widehat{\epsilon}_{j}\widehat{\delta}_{j}\end{array } \right\ } \\ + e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}^{2}\widehat{\delta}_{i}\overline{z}_{k , i}^{t}\right ] \overline{z}_{k , j}\widehat{\delta}_{j}\\ + e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}\widehat{\delta}_{i}^{2}\overline{z}_{k , i}^{t}\right ] \overline{z}_{k , j}\widehat{\epsilon}_{j}\end{array } \right\ } \end{array } \right\}\end{aligned}\ ] ] to obtain @xmath3176 we divide eq . (
[ joel ] ) by @xmath3177 . to obtain @xmath3178
, we take the variance of both sides of eq .
( [ joel ] ) under law @xmath2271 giving @xmath3179 \right ] \\ & = v\left ( \widehat{\theta}\left ( \tau^{\dagger}\right ) \right ) ^{-2}\left\ { var_{\widehat{\theta}\left ( \tau^{\dagger}\right ) } \left [ \mathbb{if}_{2,\widetilde{\psi}_{k}\left ( \tau^{\dagger},\cdot\right ) } \left ( \widehat{\theta}\left ( \tau^{\dagger}\right ) \right ) \right ] -var\left [ \mathbb{u}_{2,2,\widetilde{\tau}_{k}\left ( \cdot\right ) } ^{\ast , test,\perp}\left ( \widehat{\theta}\left ( \tau^{\dagger}\right ) , \tau^{\dagger}\right ) \right ] \right\}\end{aligned}\ ] ] parts ( i ) and ( ii ) : that @xmath3180 = o_{p}\left ( 1/n\right ) \ \ $ ] and @xmath2320 = o_{p}\left ( 1/n\right ) \ $ ] is a straightforward calculation .
the remainder of @xmath1880 and @xmath1883 follows from the fact that @xmath3181 \asymp\max\left ( \frac{1}{n},\frac{k}{n^{2}}\right ) .$ ] part ( iv ) : by part ( ii ) of theorem [ aa ] , it is sufficient to show that @xmath3182 \\ & = o_{p}\left\ { \left ( p-\widehat{p}\right ) \left ( b\left ( \tau ^{\dagger}\right ) -\widehat{b}\left ( \tau^{\dagger}\right ) \right ) \left [ \left ( p-\widehat{p}\right ) + \left ( b\left ( \tau^{\dagger}\right ) -\widehat{b}\left ( \tau^{\dagger}\right )
\right ) \right ] \right\}\end{aligned}\ ] ] below we show @xmath3183 \\ & = \left ( e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}^{2}\widehat{\delta}_{i}^{2}\right ] \right ) ^{-1}\times\\ & \left\ { \begin{array } [ c]{c}e_{\theta}\left [ \left ( b\left ( \tau^{\dagger}\right ) -\widehat{b}\left ( \tau^{\dagger}\right ) \right ) \left ( p-\widehat{p}\right ) \right ] \times\\ \left\ { \begin{array } [ c]{c}e_{\theta}\left [ \left ( p-\widehat{p}\right ) \overline{z}_{k}^{t}\right ] e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}^{2}\widehat{\delta}_{i}\overline{z}_{k , i}^{t}\right ] \\ + e_{\theta}\left [ \left ( b\left ( \tau^{\dagger}\right ) -\widehat{b}\left ( \tau^{\dagger}\right ) \right ) \overline{z}_{k}^{t}\right ] e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}\widehat{\delta}_{i}^{2}\overline{z}_{k , i}^{t}\right ]
\\ -\left\ { \begin{array } [ c]{c}\left ( e_{\widehat{\theta}}\left [ \widehat{\epsilon}_{i}^{2}\widehat{\delta } _ { i}^{2}\right ] \right ) ^{-1}e_{\widehat{\theta}}\left [ \widehat{\epsilon } \widehat{\delta}^{2}\overline{z}_{k}^{t}\right ] \\
\times e_{\widehat{\theta}}\left [ \widehat{\epsilon}^{2}\widehat{\delta } \overline{z}_{k}^{t}\right ] \times\\ e_{\theta}\left [ \left ( b\left ( \tau^{\dagger}\right ) -\widehat{b}\left ( \tau^{\dagger}\right ) \right ) \left ( p-\widehat{p}\right ) \right ] \end{array } \right\ } \end{array } \right\ } \end{array } \right\}\end{aligned}\ ] ] which is @xmath3184 \right\}\ ] ]
when , as is the case under our assumptions @xmath3185 $ ] and @xmath3185 $ ] are both order @xmath3186 but would be @xmath3187 \right\}\ ] ] in the ( unlikely ) special case in which the semiparametric regression model was precisely true , since then @xmath3188 /\left\ { e_{\widehat{\theta}}\left [ y\left ( \tau^{\dagger}\right ) 1\right)\ ] ] so @xmath3189 = o_{p}\left ( p-\widehat{p}\right)\ ] ] and @xmath3190 = o_{p}\left ( b\left ( \tau^{\dagger}\right ) -\widehat{b}\right ) .\ ] ] the expression for @xmath3191 $ ] is obtained from the formula for @xmath3192 in theorem [ aa ] by noting that , because @xmath3193 = \psi_{1,k}\left ( \tau^{\dagger},\widehat{\theta}\left ( \tau^{\dagger } \right ) \right ) $ ] and @xmath3194 \\ = e_{\theta}\left [ \psi_{1,k}\left ( \tau^{\dagger},\widehat{\theta}\left ( \tau^{\dagger}\right ) \right ) -\widetilde{\psi}_{k}\left ( \tau^{\dagger } , \theta\right ) \right ] \\ = e_{\theta}\left [ \left ( p-\widehat{p}\right ) \overline{z}_{k}^{t}\right ] e_{\theta}\left [ \overline{z}_{k}\overline{z}_{k}^{t}\right ] e_{\theta } \left [ \left ( b\left ( \tau^{\dagger}\right ) -\widehat{b}\left ( \tau^{\dagger}\right ) \right ) \overline{z}_{k}\right]\end{gathered}\ ] ] by theorem [ ebrate ] . part ( v ) : we first note that by theorem [ eiet ] , @xmath3195 } { e_{\theta}\left [ \mathbb{if}_{3,3,\widetilde{\tau } _ { k}\left ( \cdot\right ) } \left ( \widehat{\theta}\right ) \right ] } = -\left ( 1+o_{p}\left ( 1\right ) \right ) .$ ] it can be shown that @xmath3196{c}if_{3,3,\psi\left ( \tau^{\dag},\cdot\right ) , i_{1}i_{2}i_{3}}\left ( \theta\right ) + \frac{1}{6}\psi_{\backslash\tau^{3}}if_{1,\tau\left ( \cdot\right ) , i_{1}}\left ( \theta\right ) if_{1,\tau\left ( \cdot\right ) , i_{2}}\left ( \theta\right ) if_{1,\tau\left ( \cdot\right ) , i_{3}}\left ( \theta\right ) \\ + \frac{1}{3}\psi_{\backslash\tau^{2}}\left ( \begin{array } [ c]{c}if_{1,\tau\left ( \cdot\right ) , i_{1}}\left ( \theta\right ) if_{2,2,\tau \left ( \cdot\right ) , i_{2}i_{3}}\left ( \theta\right ) \\ + if_{1,\tau\left ( \cdot\right ) , i_{2}}\left ( \theta\right ) if_{2,2,\tau \left ( \cdot\right ) , i_{1}i_{3}}\left ( \theta\right ) \\ + if_{1,\tau\left ( \cdot\right ) , i_{3}}\left ( \theta\right ) if_{2,2,\tau \left ( \cdot\right ) , i_{1}i_{2}}\left ( \theta\right ) \end{array } \right ) \\ + \frac{1}{3}\left ( \begin{array } [ c]{c}d_{1,\theta}\left ( \frac{\partial if_{1,\psi\left ( \tau^{\dag},\cdot\right ) , i_{1}}\left ( \theta\right ) } { \partial\tau}\right ) if_{2,2,\tau\left ( \cdot\right ) , i_{2}i_{3}}\left ( \theta\right ) \\ + d_{1,\theta}\left ( \frac{\partial if_{1,\psi\left ( \tau^{\dag},\cdot\right ) , i_{2}}\left ( \theta\right ) } { \partial\tau}\right ) if_{2,2,\tau\left ( \cdot\right ) , i_{1}i_{3}}\left ( \theta\right ) \\ + d_{1,\theta}\left ( \frac{\partial if_{1,\psi\left ( \tau^{\dag},\cdot\right ) , i_{3}}\left ( \theta\right ) } { \partial\tau}\right ) if_{2,2,\tau\left ( \cdot\right ) , i_{1}i_{2}}\left ( \theta\right ) \end{array } \right ) \\ + \frac{1}{3}\left ( \begin{array } [ c]{c}if_{1,\tau\left ( \cdot\right ) , i_{1}}\left ( \theta\right ) d_{2,\theta } \left ( \frac{\partial if_{2,2,\psi\left ( \tau^{\dag},\cdot\right ) i_{2}i_{3}}\left ( \theta\right ) } { \partial\tau}\right ) \\ + if_{1,\tau\left ( \cdot\right ) , i_{2}}\left ( \theta\right ) d_{2,\theta } \left ( \frac{\partial if_{2,2,\psi\left ( \tau^{\dag},\cdot\right ) i_{1}i_{3}}\left ( \theta\right ) } { \partial\tau}\right ) \\ + if_{1,\tau\left ( \cdot\right ) , i_{3}}\left ( \theta\right ) d_{2,\theta } \left ( \frac{\partial if_{2,2,\psi\left ( \tau^{\dag},\cdot\right ) i_{1}i_{2}}\left ( \theta\right ) } { \partial\tau}\right ) \end{array } \right ) \\ + \frac{1}{6}\left ( \begin{array } [ c]{c}d_{1,\theta}\left ( \frac{\partial^{2}if_{1,\psi\left ( \tau^{\dag},\cdot\right ) , i_{1}}\left ( \theta\right ) } { \partial\tau^{2}}\right ) if_{1,\tau\left ( \cdot\right ) , i_{2}}\left ( \theta\right ) if_{1,\tau \left ( \cdot\right ) , i_{3}}\left ( \theta\right ) \\ + d_{1,\theta}\left ( \frac{\partial^{2}if_{1,\psi\left ( \tau^{\dag},\cdot\right ) , i_{2}}\left ( \theta\right ) } { \partial\tau^{2}}\right ) if_{1,\tau\left ( \cdot\right ) , i_{1}}\left ( \theta\right ) if_{1,\tau \left ( \cdot\right ) , i_{3}}\left ( \theta\right ) \\ + d_{1,\theta}\left ( \frac{\partial^{2}if_{1,\psi\left ( \tau^{\dag},\cdot\right ) , i_{3}}\left ( \theta\right ) } { \partial\tau^{2}}\right ) if_{1,\tau\left ( \cdot\right ) , i_{1}}\left ( \theta\right ) if_{1,\tau \left ( \cdot\right ) , i_{2}}\left ( \theta\right ) \end{array } \right ) \end{array } \right\}\end{gathered}\ ] ] from the fact that @xmath3197 we conclude that the order of @xmath3198 $ ] is equal to the order of@xmath3199 + e_{\theta}\left [ d_{1,\theta } \left ( \partial if_{1,\widetilde{\psi}_{k}\left ( \tau,\cdot\right ) } \left ( \widehat{\theta}\right ) /\partial\tau\right ) \right ] e_{\theta } \left [ if_{2,2,\widetilde{\tau}_{k}\left ( \cdot\right ) } \left ( \widehat{\theta}\right ) \right ] \\ & + e_{\theta}\left [ d_{2,\theta}\left ( \partial if_{2,2,\widetilde{\psi } _ { k}\left ( \tau,\cdot\right ) } \left ( \widehat{\theta}\right ) /\partial\tau\right ) \right ] e_{\theta}\left [ if_{1,\widetilde{\tau}_{k}\left ( \cdot\right ) } \left ( \widehat{\theta}\right ) \right]\end{aligned}\ ] ] now @xmath3200 = o_{p}\left [ \left ( p-\widehat{p}\right ) \left ( b\ -\widehat{b}\ \right ) \left ( \frac{g\left ( x\right ) } { \widehat{g}\left ( x\right ) } -1\right ) \right ] , \\ e_{\theta}\left [ d_{1,\theta}\left ( \partial if_{1,\widetilde{\psi}_{k}\left ( \tau,\cdot\right ) } \left ( \widehat{\theta}\right ) /\partial \tau\right ) \right ] = e_{\theta}\left [ \left ( a-\widehat{p}\right ) ^{2}\right ] -e_{\widehat{\theta}}\left [ \left ( a-\widehat{p}\right ) ^{2}\right ]
\\ = e_{\widehat{\theta}}\left [ \left ( \frac{f\left ( a|x\right ) } { \widehat{f}\left ( a|x\right ) } \frac{g\left ( x\right ) } { \widehat{g}\left ( x\right ) } -1\right ) \left ( a-\widehat{p}\right ) ^{2}\right ]
\\ = e_{\widehat{\theta}}\left [ \left ( \left [ \left ( \frac{f\left ( a|x\right ) } { \widehat{f}\left ( a|x\right ) } -1\right ) \frac{g\left ( x\right ) } { \widehat{g}\left ( x\right ) } \right ] \right ) \left ( a-\widehat{p}\right ) ^{2}\right ]
\\ + e_{\widehat{\theta}}\left [ \left ( \frac{g\left ( x\right ) } { \widehat{g}\left ( x\right ) } -1\right ) \left ( a-\widehat{p}\right ) ^{2}\right ] = o_{p}\left [ \left ( p-\widehat{p}\right ) + \left ( \frac{g\left ( x\right ) } { \widehat{g}\left ( x\right ) } -1\right ) \right]\end{gathered}\ ] ] by @xmath12 binary , @xmath3201 = -\left\ { e_{\theta}\left [ \left ( a-\widehat{p}\right ) \right ] \right\ } ^{2}=o_{p}\left [ \left ( p-\widehat{p}\right ) ^{2}\right ] \\ e_{\theta}\left [ if_{1,\widetilde{\tau}_{k}\left ( \cdot\right ) } \left ( \widehat{\theta}\right ) \right ] = e_{\theta}\left [ \left ( a-\widehat{p}\right ) \left ( y\ -\widehat{b}\ \right ) \right ]
\\ = e_{\widehat{\theta}}\left [ \left ( p-\widehat{p}\right ) \left ( b\ -\widehat{b}\ \right ) \right]\end{gathered}\ ] ] and using @xmath3202 and the explicit expression for @xmath3203 in eq ( [ q22 ] ) @xmath3204 \\ & = o_{p}\left [ e_{\theta}\left [ if_{2,2,\widetilde{\psi}_{k}\left ( \tau^{\dagger},\cdot\right ) } \left ( \widehat{\theta}\right ) \right ] \right ] + e_{\theta}\left [ \partial if_{1,\widetilde{\psi}_{k}\left ( \tau,\cdot\right ) } \left ( \widehat{\theta}\right ) /\partial\tau\right ] \\ & \times e_{\theta}\left [ if_{1,\widetilde{\tau}_{k}\left ( \cdot\right ) } \left ( \widehat{\theta}\right ) \right ] \\ & = o_{p}\left [ \left ( p-\widehat{p}\right ) \left ( b\ -\widehat{b}\ \right ) \right ] \\ & + o_{p}\left [ \left ( p-\widehat{p}\right ) + \left ( \frac{g\left ( x\right ) } { \widehat{g}\left ( x\right ) } -1\right ) \right ] \times\\ & o_{p}\left [ \left ( p-\widehat{p}\right ) + \left ( \frac{g\left ( x\right ) } { \widehat{g}\left ( x\right ) } -1\right ) + \left ( b\ -\widehat{b}\ \right ) \right]\end{aligned}\ ] ] combining terms completes the proof .
next , we motivate and derive the formula of @xmath3205 for an assumed unique functional @xmath643 implicitly defined by @xmath3206 @xmath357 to motivate the general formula of @xmath3207 for arbitrary @xmath425 we first consider the following formula for @xmath3208 which was derived from @xmath2232 following part 5c ) of theorem [ eift ] .
@xmath3209{c}\psi_{\backslash\tau^{4}}if_{1,\tau}\left ( i\right ) if_{1,\tau}\left ( j\right ) if_{1,\tau}\left ( s\right ) if_{1,\tau}\left ( t\right ) \\ + d_{1,\theta}\left ( \frac{\partial if_{1,\tau}\left ( i\right ) } { \partial\tau^{3}}\right ) if_{1,\tau}\left ( j\right ) if_{1,\tau}\left ( s\right ) if_{1,\psi\left ( \tau,\cdot\right ) } \left ( t\right ) \\ + if_{1,\tau}\left ( i\right ) d_{1,\theta}\left ( \frac{\partial if_{1,\tau } \left ( j\right ) } { \partial\tau^{3}}\right ) if_{1,\tau}\left ( s\right ) if_{1,\psi\left ( \tau,\cdot\right ) } \left ( t\right ) \\ + if_{1,\tau}\left ( i\right ) if_{1,\tau}\left ( j\right ) d_{1,\theta}\left ( \frac{\partial if_{1,\tau}\left ( s\right ) } { \partial\tau^{3}}\right ) if_{1,\psi\left ( \tau,\cdot\right ) } \left ( t\right ) \\ + if_{1,\tau}\left ( i\right ) if_{1,\tau}\left ( j\right ) if_{1,\tau}\left ( s\right ) d_{1,\theta}\left ( \frac{\partial if_{1,\psi\left ( \tau , \cdot\right ) } \left ( t\right ) } { \partial\tau^{3}}\right ) \end{array } \right\ } \\ & + \frac{1}{2}\mathbb{v}\left\ { \begin{array } [ c]{c}\psi_{\backslash\tau^{3}}if_{22,\tau}\left ( ij\right ) if_{1,\tau}\left ( s\right ) if_{1,\tau}\left ( t\right ) \\ + d_{1,\theta}\left ( \frac{\partial if_{1,\psi}\left ( t\right ) } { \partial\tau^{2}}\right ) if_{22,\tau}\left ( ij\right ) if_{1,\tau}\left ( s\right ) \\ + d_{1,\theta}\left ( \frac{\partial if_{1,\psi}\left ( s\right ) } { \partial\tau^{2}}\right ) if_{22,\tau}\left ( ij\right ) if_{1,\tau}\left ( t\right ) \\ + d_{1,\theta}\left ( \frac{\partial if_{22,\psi}\left ( ij\right ) } { \partial\tau^{2}}\right ) if_{1,\tau}\left ( t\right ) if_{1,\tau}\left ( s\right ) \end{array } \right\ } \\ & + \frac{1}{2}\mathbb{v}\left\ { \begin{array } [ c]{c}\psi_{\backslash\tau^{2}}if_{22,\tau}\left ( ij\right ) if_{22,\tau}\left ( st\right ) + \\ d_{2,\theta}\left ( \frac{\partial if_{22,\psi}\left ( ij\right ) } { \partial\tau}\right ) if_{22,\tau}\left ( st\right ) \\ + if_{22,\tau}\left ( ij\right ) d_{2,\theta}\left ( \frac{\partial if_{22,\psi}\left ( st\right ) } { \partial\tau}\right ) \end{array } \right\ } \\ & + \mathbb{v}\left\ { \begin{array } [ c]{c}\psi_{\backslash\tau^{2}}if_{1,\tau}\left ( i\right ) if_{33,\tau}\left ( jst\right ) + \\ d_{1,\theta}\left ( \frac{\partial if_{1,\psi}\left ( i\right ) } { \partial \tau}\right ) if_{33,\tau}\left ( jst\right ) + d_{3,\theta}\left ( \frac{\partial if_{33,\psi}\left ( jst\right ) } { \partial\tau}\right ) if_{1,\tau}\left ( i\right ) \end{array } \right\}\end{aligned}\ ] ] this formula for @xmath3210 reveals a very nice pattern .
note that in addition to @xmath3211 the rhs consists of four pieces with leading terms @xmath3212 @xmath3213 @xmath3214 and @xmath3215 respectively . within each piece , the remaining terms can be constructed simply by applying the algorithm * te * described below . * te * i )
remove the partial derivative of @xmath271 wrt .
@xmath3216 ii ) for each factor of the leading term , replace the @xmath3217 in the subscript with @xmath3218 and iii ) partially differentiate the newly replaced factor wrt .
@xmath2041 , and make the partial derivative degenerate . here
the order of the partial derivative equals the total number of the factors in the leading term minus 1 . moreover , each piece corresponds to one of the following ways of representing the number 4 as a sum : @xmath3219 @xmath3220 , @xmath3221 , and @xmath3222 furthermore , assume @xmath135 can be written as @xmath3223 @xmath2820with @xmath3224 e.g. , @xmath3225 then the number in front of the piece corresponding to the sum representation @xmath3226 equals @xmath3227 note that @xmath3228 [ testing ] let @xmath643 be the assumed unique functional defined by @xmath3229 then , for @xmath3230 , whenever @xmath3231 and @xmath3232 exist , @xmath3233{c}\psi_{\backslash\tau^{sum\left ( \overline{\varkappa}_{m}\right ) } } { \displaystyle\prod\limits_{r=1}^{m-1 } } \left ( { \displaystyle\prod\limits_{s=1}^{\varkappa_{m , r } } } if_{r , r,\tau\left ( \cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \left ( \theta\right ) \right ) \\ + { \displaystyle\sum\limits_{r , s}^{m-1,\varkappa_{m , r } } } \left ( \begin{array } [ c]{c}d_{r,\theta}\left ( \frac{\partial^{\left [ sum\left ( \overline{\varkappa } _ { m}\right ) -1\right ] } \left ( if_{r , r,\psi\left ( \tau,\cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \right ) } { \partial\tau^{\left [ sum\left ( \overline{\varkappa}_{m}\right ) -1\right ] } } \right ) \times\\{\displaystyle\prod\limits_{\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } if_{r_{1},r_{1},\tau\left ( \cdot\right ) , \overline{i}_{r_{1},s_{1}}}^{\left ( s\right ) } \left ( \theta\right ) \end{array } \right ) \end{array } \right\ } \label{tsm}\ ] ] note that @xmath3235{c}d_{r,\theta}\left ( \frac{\partial^{\left [ sum\left ( \overline{\varkappa } _ { m}\right ) -1\right ] } \left ( if_{r , r,\psi\left ( \tau,\cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \right ) } { \partial\tau^{\left [ sum\left ( \overline{\varkappa}_{m}\right ) -1\right ] } } \right ) \times\\{\displaystyle\prod\limits_{\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } if_{r_{1},r_{1},\tau\left ( \cdot\right ) , \overline{i}_{r_{1},s_{1}}}^{\left ( s\right ) } \left ( \theta\right ) \end{array } \right ) \right\}\ ] ] can be constructed by applying algorithm * te * to the leading term @xmath3236 eq . ( [ ts1 ] )
has been proved in theorem [ tt ] .
next , we shall prove eq .
( [ tsm ] ) by induction .
the case where @xmath2238 was proved in theorem [ tt ] as well .
now , we assume eq .
( [ tsm ] ) holds for @xmath3237and prove it is also true for @xmath286 by part 5c ) of theorem [ eift ] . specifically , by induction assumption , @xmath3238{c}\psi_{\backslash\tau^{sum\left ( \overline{\varkappa}_{m}\right ) } } { \displaystyle\prod\limits_{r=1}^{m-1 } } \left ( { \displaystyle\prod\limits_{s=1}^{\varkappa_{m , r } } } if_{r , r,\tau\left ( \cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \left ( \theta\right ) \right ) \\ + { \displaystyle\sum\limits_{r , s}^{m-1,\varkappa_{m , r } } } \left ( \begin{array } [ c]{c}d_{r,\theta}\left ( \frac{\partial^{\left [ sum\left ( \overline{\varkappa } _ { m}\right ) -1\right ] } \left ( if_{r , r,\psi\left ( \tau,\cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \right ) } { \partial\tau^{\left [ sum\left ( \overline{\varkappa}_{m}\right ) -1\right ] } } \right ) \times\\{\displaystyle\prod\limits_{\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } if_{r_{1},r_{1},\tau\left ( \cdot\right ) , \overline{i}_{r_{1},s_{1}}}^{\left ( s\right ) } \left ( \theta\right ) \end{array } \right ) \end{array } \right\}\end{aligned}\ ] ] consider any sufficiently smooth @xmath3239dimensional parametric submodel @xmath3240 mapping @xmath3241 to @xmath245 . for any @xmath179 in the range of @xmath3242 differentiate both sides of the above equation w.r.t @xmath781 and evaluate at @xmath3243 @xmath3244 then @xmath3245{c}\left [ \begin{array } [ c]{c}\psi_{\backslash\tau^{sum\left ( \overline{\varkappa}_{m}\right ) + 1}}\tau_{\backslash t}\left ( \theta\right ) \\ + \frac{\partial\psi_{\backslash\tau^{sum\left ( \overline{\varkappa}_{m}\right ) } } \left ( \tau,\theta_{t}\right ) } { \partial t}|_{t = t^{\ast}}\end{array } \right ] { \displaystyle\prod\limits_{r=1}^{m-1 } } \left ( { \displaystyle\prod\limits_{s=1}^{\varkappa_{m , r } } } if_{r , r,\tau\left ( \cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \left ( \theta\right ) \right ) \\ + \psi_{\backslash\tau^{sum\left ( \overline{\varkappa}_{m}\right ) } } { \displaystyle\sum\limits_{\left ( r , s\right ) } ^{sum\left ( \overline { \varkappa}_{m}\right ) } } if_{r , r,\tau\left ( \cdot\right ) , \overline{i}_{r , s},\backslash t}^{\left ( s\right ) } \left ( \theta\right ) { \displaystyle\prod\limits_{\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } if_{r_{1},r_{1},\tau\left ( \cdot\right ) , \overline{i}_{r_{1},s_{1}}}^{\left ( s\right ) } \left ( \theta\right ) \\ + { \displaystyle\sum\limits_{\left ( r , s\right ) } ^{sum\left ( \overline { \varkappa}_{m}\right ) } } \left\ { \begin{array } [ c]{c}\left [ \begin{array } [ c]{c}\frac{\partial^{sum\left ( \overline{\varkappa}_{m}\right ) } \left ( if_{r , r,\psi\left ( \tau,\cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \right ) } { \partial\tau^{sum\left ( \overline{\varkappa}_{m}\right ) } } \tau_{\backslash t}\left ( \theta\right ) \\ + \frac{\partial^{sum\left ( \overline{\varkappa}_{m}\right ) } \left ( if_{r , r,\psi\left ( \tau,\cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \left ( \tau,\theta_{t}\right ) \right ) } { \partial\tau^{\left [ sum\left ( \overline{\varkappa}_{m}\right ) -1\right ] } \partial t}|_{t = t^{\ast}}\end{array } \right ] \\ \times{\displaystyle\prod\limits_{\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } if_{r_{1},r_{1},\tau\left ( \cdot\right ) , \overline{i}_{r_{1},s_{1}}}^{\left ( s\right ) } \left ( \theta\right ) \\ + \frac{\partial^{\left [ sum\left ( \overline{\varkappa}_{m}\right ) -1\right ] } \left ( if_{r , r,\psi\left ( \tau,\cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \right ) } { \partial\tau^{\left [ sum\left ( \overline{\varkappa}_{m}\right ) -1\right ] } } \\ \times{\displaystyle\sum\limits_{\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } if_{r_{1},r_{1},\tau\left ( \cdot\right ) , \overline{i}_{r_{1},s_{1}},\backslash t}^{\left ( s\right ) } \left ( \theta\right ) \\ \times{\displaystyle\prod\limits_{\left ( r_{2},s_{2}\right ) \neq\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } if_{r_{2},r_{2},\tau\left ( \cdot\right ) , \overline{i}_{r_{2},s_{2}}}^{\left ( s\right ) } \left ( \theta\right ) \end{array } \right\ } \end{array } \right\ } \label{m2m+1}\ ] ] note that @xmath3246 is the derivative of @xmath3247 w.r.t .
@xmath781 while fixing @xmath2041 at @xmath2044 therefore , @xmath3248 \\ & = e_{\theta}\left [ \frac{\partial^{p-1}if_{1,if_{r , r,\psi\left ( \tau , \cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \left ( \cdot\right ) } \left ( o_{m+1};\theta\right ) } { \partial\tau^{p-1}}s_{1,t}\left ( o_{m+1}\right ) \right ] , \end{aligned}\ ] ] and@xmath3249{c}\ \psi_{\backslash\tau^{2}}if_{1,\tau\left ( \cdot\right ) , i_{m+1}}\left ( \theta\right ) if_{m , m,\tau\left ( \cdot\right ) } ^{\left ( s\right ) } \left ( \mathbf{o}_{i_{m}};\theta\right )
\\ + \frac{\partial}{\partial\tau}if_{1,\psi\left ( \tau,\cdot\right ) , i_{m+1}}if_{m , m,\tau\left ( \cdot\right ) } ^{\left ( s\right ) } \left ( \mathbf{o}_{i_{m}};\theta\right ) \\ + if_{m , m,\psi\left ( \tau,\cdot\right ) , \backslash\tau}^{\left ( s\right ) } if_{1,\tau\left ( \cdot\right ) , i_{m+1}}\end{array } \right\ } \\ & + \sum_{\left ( \varkappa_{m,1},\varkappa_{m,2}, ... ,\varkappa_{m , m-1}\right ) } \left ( { \displaystyle\prod\limits_{r=1}^{m-1 } } \varkappa_{m , r}!\right ) ^{-1}\times\\ & \left\ { \begin{array } [ c]{c}\left\ { \begin{array } [ c]{c}\left\ { \begin{array } [ c]{c}\left [ \begin{array } [ c]{c}\psi_{\backslash\tau^{sum\left ( \overline{\varkappa}_{m}\right ) + 1}}if_{1,\tau\left ( \cdot\right ) , i_{m+1}}\left ( \theta\right ) \\ + \frac{\partial^{sum\left ( \overline{\varkappa}_{m}\right ) } } { \partial \tau^{sum\left ( \overline{\varkappa}_{m}\right ) } } \left ( if_{1,\psi\left ( \tau,\cdot\right ) , i_{m+1}}\right ) \end{array } \right ] \\ \times{\displaystyle\prod\limits_{r=1}^{m-1 } } \left ( { \displaystyle\prod\limits_{s=1}^{\varkappa_{m , r } } } if_{r , r,\tau\left ( \cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \left ( \theta\right ) \right ) \end{array } \right\ } \\ + { \displaystyle\sum\limits_{\left ( r , s\right ) } ^{sum\left ( \overline { \varkappa}_{m}\right ) } } \left [ \begin{array } [ c]{c}\frac{\partial^{sum\left ( \overline{\varkappa}_{m}\right ) } \left ( if_{r , r,\psi\left ( \tau,\cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \right ) } { \partial\tau^{sum\left ( \overline{\varkappa}_{m}\right ) } } if_{1,\tau\left ( \cdot\right ) , i_{m+1}}\left ( \theta\right ) \\ \times{\displaystyle\prod\limits_{\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } if_{r_{1},r_{1},\tau\left ( \cdot\right ) , \overline{i}_{r_{1},s_{1}}}^{\left ( s\right ) } \left ( \theta\right ) \end{array } \right ] \end{array } \right\ } \\ + { \displaystyle\sum\limits_{\left ( r , s\right ) } ^{sum\left ( \overline { \varkappa}_{m}\right ) } } \left ( \begin{array } [ c]{c}\left [ \begin{array } [ c]{c}\psi_{\backslash\tau^{sum\left ( \overline{\varkappa}_{m}\right ) } } if_{1,if_{r , r,\tau\left ( \cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \left ( \cdot\right ) } \left ( o_{i_{m+1}};\theta\right ) \\ \times{\displaystyle\prod\limits_{\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } if_{r_{1},r_{1},\tau\left ( \cdot\right ) , \overline{i}_{r_{1},s_{1}}}^{\left ( s\right ) } \left ( \theta\right ) \end{array } \right ] \\ + \left [ \begin{array } [ c]{c}\frac{\partial^{sum\left ( \overline{\varkappa}_{m}\right ) } \left ( if_{1,if_{r , r,\psi\left ( \tau,\cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \left ( \cdot\right ) } \left ( o_{i_{m+1}};\theta\right ) \right ) } { \partial\tau^{\left [ sum\left ( \overline{\varkappa}_{m}\right ) -1\right ] } } \\ \times{\displaystyle\prod\limits_{\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } if_{r_{1},r_{1},\tau\left ( \cdot\right ) , \overline{i}_{r_{1},s_{1}}}^{\left ( s\right ) } \left ( \theta\right ) \end{array } \right ] \end{array } \right ) \\ + { \displaystyle\sum\limits_{\left ( r , s\right ) } ^{sum\left ( \overline { \varkappa}_{m}\right ) } } \left\ { \begin{array } [ c]{c}\frac{\partial^{\left [ sum\left ( \overline{\varkappa}_{m}\right ) -1\right ] } \left ( if_{r , r,\psi\left ( \tau,\cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \right ) } { \partial\tau^{\left [ sum\left ( \overline{\varkappa } _ { m}\right ) -1\right ] } } \times\\{\displaystyle\sum\limits_{\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } \left ( \begin{array } [ c]{c}if_{1,if_{r_{1},r_{1},\tau\left ( \cdot\right ) , \overline{i}_{r_{1},s_{1}}}^{\left ( s\right ) } \left ( \cdot\right ) } \left ( o_{i_{m+1}};\theta\right ) \times\\{\displaystyle\prod\limits_{\left ( r_{2},s_{2}\right ) \neq\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } if_{r_{2},r_{2},\tau\left ( \cdot\right ) , \overline{i}_{r_{2},s_{2}}}^{\left ( s\right ) } \left ( \theta\right ) \end{array } \right ) \end{array } \right\ } \end{array } \right\}\end{aligned}\ ] ] consider the last term @xmath3250{c}\frac{\partial^{\left [ sum\left ( \overline{\varkappa}_{m}\right ) -1\right ] } \left ( if_{r , r,\psi\left ( \tau,\cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \right ) } { \partial\tau^{\left [ sum\left ( \overline{\varkappa } _ { m}\right ) -1\right ] } } { \displaystyle\sum\limits_{\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } if_{1,if_{r_{1},r_{1},\tau\left ( \cdot\right ) , \overline{i}_{r_{1},s_{1}}}^{\left ( s\right ) } \left ( \cdot\right ) } \left ( o_{i_{m+1}};\theta\right ) \\ \times{\displaystyle\prod\limits_{\left ( r_{2},s_{2}\right ) \neq\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } if_{r_{2},r_{2},\tau\left ( \cdot\right ) , \overline{i}_{r_{2},s_{2}}}^{\left ( s\right ) } \left ( \theta\right ) \end{array } \right\}\ ] ] wlog , we exchange @xmath3251 with @xmath3252 then we have @xmath3253{c}\frac{\partial^{\left [ sum\left ( \overline{\varkappa}_{m}\right ) -1\right ] } \left ( if_{r_{1},r_{1},\psi\left ( \tau,\cdot\right ) , \overline{i}_{r_{1},s_{1}}}^{\left ( s\right ) } \right ) } { \partial\tau^{\left [ sum\left ( \overline{\varkappa}_{m}\right ) -1\right ] } }
\times\\ if_{1,if_{r , r,\tau\left ( \cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \left ( \cdot\right ) } \left ( o_{i_{m+1}};\theta\right ) { \displaystyle\prod\limits_{\left ( r_{2},s_{2}\right ) \neq\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } if_{r_{2},r_{2},\tau\left ( \cdot\right ) , \overline{i}_{r_{2},s_{2}}}^{\left ( s\right ) } \left ( \theta\right ) \end{array } \right]\ ] ] and the sum of the last two terms equals @xmath3254{c}\left [ \begin{array } [ c]{c}\psi_{\backslash\tau^{sum\left ( \overline{\varkappa}_{m}\right ) } } if_{1,if_{r , r,\tau\left ( \cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \left ( \cdot\right ) } \left ( o_{i_{m+1}};\theta\right ) \\
\times{\displaystyle\prod\limits_{\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } if_{r_{1},r_{1},\tau\left ( \cdot\right ) , \overline{i}_{r_{1},s_{1}}}^{\left ( s\right ) } \left ( \theta\right ) \end{array } \right ]
\\ + \left [ \begin{array } [ c]{c}\frac{\partial^{sum\left ( \overline{\varkappa}_{m}\right ) } \left ( if_{1,if_{r , r,\psi\left ( \tau,\cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \left ( \cdot\right ) } \left ( o_{i_{m+1}};\theta\right ) \right ) } { \partial\tau^{\left [ sum\left ( \overline{\varkappa}_{m}\right ) -1\right ] } } \\ \times{\displaystyle\prod\limits_{\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } if_{r_{1},r_{1},\tau\left ( \cdot\right ) , \overline{i}_{r_{1},s_{1}}}^{\left ( s\right ) } \left ( \theta\right ) \end{array } \right ] \\ + \left ( \begin{array } [ c]{c}if_{1,if_{r , r,\tau\left ( \cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \left ( \cdot\right ) } \left ( o_{i_{m+1}};\theta\right ) \times\\{\displaystyle\sum\limits_{\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } \left [ \begin{array } [ c]{c}\frac{\partial^{\left [ sum\left ( \overline{\varkappa}_{m}\right ) -1\right ] } \left ( if_{r_{1},r_{1},\psi\left ( \tau,\cdot\right ) , \overline{i}_{r_{1},s_{1}}}^{\left ( s\right ) } \right ) } { \partial\tau^{\left [ sum\left ( \overline{\varkappa}_{m}\right ) -1\right ] } } \times\\{\displaystyle\prod\limits_{\left ( r_{2},s_{2}\right ) \neq\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } if_{r_{2},r_{2},\tau\left ( \cdot\right ) , \overline{i}_{r_{2},s_{2}}}^{\left ( s\right ) } \left ( \theta\right ) \end{array } \right ] \end{array } \right ) \end{array } \right)\ ] ] now , we have shown that , in addition to @xmath3255 the rhs of eq.([m2m+1 ] ) can be written as the sum of three pieces with the following leading terms@xmath3256{c}\left [ \psi_{\backslash\tau^{sum\left ( \overline{\varkappa}_{m}\right ) + 1}}if_{1,\tau\left ( \cdot\right ) , i_{m+1}}\left ( \theta\right ) \right ] { \displaystyle\prod\limits_{r=1}^{m-1 } } \left ( { \displaystyle\prod\limits_{s=1}^{\varkappa_{m , r } } } if_{r , r,\tau\left ( \cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \left ( \theta\right ) \right ) \\
+ { \displaystyle\sum\limits_{\left ( r , s\right ) } ^{sum\left ( \overline { \varkappa}_{m}\right ) } } \left ( \begin{array } [ c]{c}\psi_{\backslash\tau^{sum\left ( \overline{\varkappa}_{m}\right ) } } if_{1,if_{r , r,\tau\left ( \cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \left ( \cdot\right ) } \left ( o_{i_{m+1}};\theta\right ) \\ \times{\displaystyle\prod\limits_{\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } if_{r_{1},r_{1},\tau\left ( \cdot\right ) , \overline{i}_{r_{1},s_{1}}}^{\left ( s\right ) } \left ( \theta\right ) \end{array } \right ) \end{array } \right\ } , \label{lead_m+1}\ ] ] while the remaining terms can be constructed by applying the algorithm * te * to the above leading terms .
this can be proved following a simple but important fact that , for any sum representation of@xmath3260 either i ) @xmath3261 and @xmath3262 for @xmath1939 @xmath3263 or ii ) @xmath3264 and there exists a sum representation of @xmath3265 @xmath3266 such that , define@xmath3275 \\ & \times{\displaystyle\prod\limits_{r=1}^{m-1 } } \left ( { \displaystyle\prod\limits_{s=1}^{\varkappa_{m , r } } } if_{r , r,\tau\left ( \cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \left ( \theta\right ) \right ) \\ & lt2\left ( \varkappa_{m,1}, ... ,\varkappa_{m , m-1};r\right ) \\ & = \left ( { \displaystyle\prod\limits_{r=1}^{m-1 } } \varkappa_{m , r}!\right ) ^{-1}\sum_{s=1}^{\varkappa_{m , r}}\left ( \begin{array } [ c]{c}\psi_{\backslash\tau^{sum\left ( \overline{\varkappa}_{m}\right ) } } if_{1,if_{r , r,\tau\left ( \cdot\right ) , \overline{i}_{r , s}}^{\left ( s\right ) } \left ( \cdot\right ) } \left ( o_{i_{m+1}};\theta\right ) \\ \times{\displaystyle\prod\limits_{\left ( r_{1},s_{1}\right ) \neq\left ( r , s\right ) } } if_{r_{1},r_{1},\tau\left ( \cdot\right ) , \overline{i}_{r_{1},s_{1}}}^{\left ( s\right ) } \left ( \theta\right ) \end{array } \right ) .\end{aligned}\ ] ] note that for any @xmath3276 if @xmath3277 then @xmath3278 and for any @xmath3279 such that @xmath3280 @xmath3281 as @xmath3282 now , it is obvious that the term with @xmath3283 in eq .
( [ lead2 ] ) comes from the following terms in eq .
( [ lead_m+1 ] ) @xmath3284 while the term with @xmath3285 in eq .
( [ lead2 ] ) comes from the following terms in eq .
( [ lead_m+1 ] ) @xmath3286 ( theorem [ hoiprod ] ) we proceed by induction . for @xmath2581 @xmath3287 \right\ } \\ & = \sum_{r=1}^{\zeta}\left\ { e_{\theta}\left [ \mathbb{if}_{\psi_{r},1}\left ( \theta\right ) \mathbb{\zeta}_{1;\backslash l_{1}}\left ( \theta\right ) \right ] \times\left [ \prod\limits_{s\leq\zeta , s\neq r}\psi_{s}\left ( \theta\right ) \right ] \right\}\end{aligned}\ ] ] therefore @xmath3288 \right\ } \\ & = \mathbb{v}\left [ { \displaystyle\sum\limits_{\left\ { t_{1}, ... t_{\zeta}\right\ } \in \upsilon_{\zeta;1 } } } { \displaystyle\prod\limits_{s=1}^{\zeta } } if_{\psi_{s}\left ( \theta\right ) ; t_{s},t_{s};\overline{i}_{s
, t_{s}}}\left ( \theta\right ) \right]\end{aligned}\ ] ] assume that the lemma holds for @xmath972i.e . :
@xmath3289\ ] ] we now show that it holds for @xmath3290 now , @xmath3291 \\ & -\pi_{\theta , m}\left [ \mathbb{v}\left [ if_{1,if_{\psi\left ( \theta ; \zeta\right ) ; j , j}\left ( o_{i_{1}}, ... ,o_{i_{j}};\cdot\right ) \ } \left ( o_{i_{j+1}};\theta\right ) \right ] |\mathcal{u}_{j}\left ( \theta\right ) \right]\end{aligned}\ ] ] so that @xmath3292 @xmath3293{c}\left\ { \begin{array } [ c]{c}if_{if_{\psi_{r}\left ( \theta\right ) ; t_{r},t_{r};\overline{i}_{r , t_{r}}}\left ( \theta\right ) ; 1,1\ } \left ( o_{i_{r , t_{r}+1}};\theta\right ) \\ -\pi_{\theta , t_{r}}\left [ \mathbb{v}\left [ if_{if_{\psi_{r}\left ( \theta\right ) ; t_{r},t_{r};\overline{i}_{r , t_{r}}}\left ( \theta\right ) ; 1,1\ } \left ( o_{i_{r , t_{r}+1}};\theta\right ) \right ] |\mathcal{u}_{t_{r}}\left ( \theta\right ) \right ] \end{array } \right\ } \\
\times\left ( { \displaystyle\prod\limits_{s\leq\zeta , s\neq r } } if_{\psi_{s}\left ( \theta\right ) ; t_{s},t_{s};\overline{i}_{s , t_{s}}}\left ( \theta\right ) \right ) \end{array } \right\ } \right ] \\ & = \mathbb{v}\left [ { \displaystyle\sum\limits_{\left\ { t_{1}, ... t_{\zeta}\right\ } \in \upsilon_{\zeta;j } } } { \displaystyle\sum\limits_{r=1}^{\zeta } } \left\ { \begin{array } [ c]{c}\left ( t_{r}+1\right ) if_{\psi_{r}\left ( \theta\right ) ; t_{r}+1,t_{r}+1;\overline{i}_{r , t_{r}+1}}\left ( \theta\right ) \\
\times\left ( { \displaystyle\prod\limits_{s\leq\zeta , s\neq r } } if_{\psi_{s}\left ( \theta\right ) ; t_{s},t_{s};\overline{i}_{s , t_{s}}}\left ( \theta\right ) \right ) \end{array } \right\ } \right]\end{aligned}\ ] ] now , consider an arbitrary term in the double sum corresponding to the index vector @xmath3294 where the star indicates the index of the second summation .
then this term and terms corresponding to @xmath3295 all share the common factor @xmath3296 but with respective multiplicative constants @xmath3297 . so that the total contribution of terms in the double summation that share this common factor is given by:@xmath3298{c}if_{\psi_{r^{\ast}}\left ( \theta\right ) ; t_{r^{\ast}}^{\prime}+1,t_{r^{\ast } } ^{\prime}+1;\overline{i}_{r^{\ast},t_{r^{\ast}}^{\prime}+1}}\left ( \theta\right ) \\
\times\left ( { \displaystyle\prod\limits_{s\leq\zeta , s\neq r^{\ast } } } if_{\psi_{s}\left ( \theta\right ) ; t_{s}^{\prime},t_{s}^{\prime};\overline { i}_{s , t_{s}^{^{\prime}}}}\left ( \theta\right ) \right ) \end{array } \right ) \\ & = ( j+1)\left ( \begin{array } [ c]{c}if_{\psi_{r^{\ast}}\left ( \theta\right ) ; t_{r^{\ast}}^{\prime}+1,t_{r^{\ast } } ^{\prime}+1;\overline{i}_{r^{\ast},t_{r^{\ast}}^{\prime}+1}}\left ( \theta\right ) \\ \times\left ( { \displaystyle\prod\limits_{s\leq\zeta , s\neq r^{\ast } } } if_{\psi_{s}\left ( \theta\right ) ; t_{s}^{\prime},t_{s}^{\prime};\overline { i}_{s , t_{s}^{^{\prime}}}}\left ( \theta\right ) \right ) \end{array } \right)\end{aligned}\ ] ] repeating this argument over all common factors in the set @xmath3299{c}if_{\psi_{r^{\ast}}\left ( \theta\right ) ; t_{r^{\ast}}+1,t_{r^{\ast}}+1;\overline{i}_{r^{\ast},t_{r^{\ast}}+1}}\left ( \theta\right ) \\
\times\left ( { \displaystyle\prod\limits_{s\leq\zeta , s\neq r^{\ast } } } if_{\psi_{s}\left ( \theta\right ) ; t_{s}^{\prime},t_{s}^{\prime};\overline { i}_{s , t_{s}^{^{\prime}}}}\left ( \theta\right ) \right ) \end{array } \right ) : \left\ { t_{1}, ... t_{\zeta}\right\ } \in\upsilon_{\zeta;j}\right\}\ ] ] that appear in the double sum , we recover the desired sum @xmath3300\ ] ] this is because the set of all common factors in @xmath1384 is precisely : @xmath3301 this concludes the proof .
( theorem [ btomm ] ) for @xmath2113@xmath3302 + e_{\theta } \left [ \frac{r_{0}}{\widehat{\pi}_{0}}\left ( \widehat{b}_{1}-\widehat{b}_{0}\right ) \right ] + e_{\theta}\left ( \widehat{b}_{0}-b_{0}\right ) \\ & = e_{\theta}\left [ \frac{r_{0}}{\widehat{\pi}_{0}}\delta p_{1}\delta b_{1}+\delta p_{0}\delta b_{0}\right]\end{aligned}\ ] ] next we proceed to prove eq.([ebm ] ) and eq.([tbm ] ) by induction , first of all , @xmath3303 \\ & = e_{\theta}\left ( \frac{r_{0}}{\widehat{\pi}_{0}}\left ( b_{1}-\widehat{b}_{1}\right ) \left ( \frac{r_{1}}{\widehat{\pi}_{1}}-1\right ) + \frac{r_{0}}{\widehat{\pi}_{0}}\left ( b_{1}-\widehat{b}_{1}+\widehat{b}_{1}-\widehat{b}_{0}\right ) |l_{0}\right ) \\ & = e_{\theta}\left ( \frac{r_{0}}{\widehat{\pi}_{0}}\delta b_{1}\delta p_{1}+\frac{r_{0}}{\widehat{\pi}_{0}}\delta b_{0}|l_{0}\right)\end{aligned}\ ] ] and @xmath3304 \\ = e_{\theta}\left ( \pi_{\theta}\left ( q_{01}^{1/2}\delta b_{1}|\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ) \pi_{\theta}\left ( q_{0}q_{01}^{-1/2}\delta p_{1}|\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ) \right ) + \\ e_{\theta}\left ( \pi_{\theta}^{\bot}\left ( q_{01}^{1/2}\delta b_{1}|\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ) \pi_{\theta}^{\bot}\left ( q_{0}q_{01}^{-1/2}\delta p_{1}|\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ) \right)\end{gathered}\ ] ] with @xmath3305 for @xmath2211@xmath3306 + e_{\theta}\left ( \widehat{\psi}_{2,2}\right ) \\ & = e_{\theta}\left [ \frac{r_{0}}{\widehat{\pi}_{0}}\delta p_{1}\delta b_{1}+\delta p_{0}\delta b_{0}\right ] -e_{\theta}\left [ \frac{r_{0}\pi_{1}}{\widehat{\pi}_{0}\widehat{\pi}_{1}}\delta b_{1}\overline{w}_{k_{1}}^{t}\right ] e_{\theta}\left [ \overline{w}_{k_{1}}\frac{r_{0}}{\widehat{\pi } _ { 0}}\delta p_{1}\right ] \\ & -e_{\theta}\left\ { \left ( \frac{r_{0}}{\widehat{\pi}_{0}}\delta p_{1}\delta b_{1}+\frac{r_{0}}{\widehat{\pi}_{0}}\delta b_{0}\right ) \overline{z}_{k_{0}}^{t}\right\ } e_{\theta}\left [ \overline{z}_{k_{0}}\delta p_{0}\right ] \\ & = -e_{\theta}\left ( q_{0}\delta b_{0}\overline{z}_{k_{0}}^{t}\right ) e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}\right ) ^{-1}\left [ e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}\right ) -i\right ] e_{\theta}\left ( \overline{z}_{k_{0}}\delta p_{0}\right ) \\ & -e_{\theta}\left ( q_{01}\delta b_{1}\overline{w}_{k_{1}}^{t}\right ) e_{\theta}\left ( q_{01}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}\right ) ^{-1}\left ( e_{\theta}\left ( q_{01}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}\right ) -i\right ) e_{\theta}\left ( \overline{w}_{k_{1}}q_{0}\delta p_{1}\right ) \\ & -e_{\theta}\left ( q_{0}\delta p_{1}\delta b_{1}\overline{z}_{k_{0}}^{t}\right ) e_{\theta}\left [ \overline{z}_{k_{0}}\delta p_{0}\right ] \\ & + e_{\theta}\left ( \pi_{\theta}^{\bot}\left ( q_{0}^{1/2}\delta b_{0}|\left ( q_{0}^{1/2}\overline{z}_{k_{0}}\right ) \right ) \pi_{\theta } ^{\bot}\left ( q_{0}^{-1/2}\delta p_{0}|\left ( q_{0}^{1/2}\overline { z}_{k_{0}}\right ) \right ) \right ) \\ & + e_{\theta}\left ( \pi_{\theta}^{\bot}\left ( q_{01}^{1/2}\delta b_{1}|\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ) \pi_{\theta } ^{\bot}\left ( q_{0}q_{01}^{-1/2}\delta p_{1}|\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ) \right)\end{aligned}\ ] ] if e.q([ebm ] ) holds for @xmath3307 we next show it also holds for @xmath425@xmath3308{c}e_{\theta}\left [ q_{0}\delta p_{1}\delta b_{1}\overline{z}_{k_{0}}^{t}\right ] \left [ e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline { z}_{k_{0}}^{t}-i\right ) \right ] ^{m-2}e_{\theta}\left ( \overline{z}_{k_{0}}\delta p_{0}\right ) \\ + e_{\theta}\left [ q_{0}\delta b_{0}\overline{z}_{k_{0}}^{t}\right ] \left [ e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ) \right ] ^{m-2}e_{\theta}\left ( \overline{z}_{k_{0}}\delta p_{0}\right ) _ { \_\left ( l1.1\right ) }
\\ + e_{\theta}\left ( q_{01}\delta b_{1}\overline{w}_{k_{1}}^{t}\right ) \left [ e_{\theta}\left ( q_{01}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}-i\right ) \right ] ^{m-2}e_{\theta}\left ( \overline{w}_{k_{1}}q_{0}\delta p_{1}\right ) \__{\left ( l1.2\right ) } \\ + { \textstyle\sum\limits_{j=2}^{m-1 } } \begin{array } [ c]{c}e_{\theta}\left [ \overline{z}_{k_{0}}\delta p_{0}^{t}\right ] \left [ e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ) \right ] ^{j-2}e_{\theta}\left [ q_{01}\delta b_{1}\overline{z}_{k_{0}}\overline{w}_{k_{1}}^{t}\right ] \\
\times e_{\theta}\left ( q_{01}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}-i\right ) ^{m-1-j}e_{\theta}\left [ \overline{w}_{k_{1}}q_{0}\delta p_{1}\right ] \__{\left ( l1.3.j\right ) } \end{array } \end{array } \right\}\end{aligned}\]]@xmath3309@xmath3310{c}\left\ { \begin{array } [ c]{c}e_{\theta}\left\ { \left [ q_{0}\delta b_{0}\right ] \overline{z}_{k_{0}}^{t}\right\ } e_{\theta}\left [ q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}\right ] ^{-1}\times\\ \left [ e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}\right ) -i\right ] ^{m-2}e_{\theta}\left [ \overline{z}_{k_{0}}\delta p_{0}\right ] \end{array } \right\ } _ { -\left ( l2.1\right ) }
\\ + \left\ { \begin{array } [ c]{c}e_{\theta}\left\ { \left [ q_{01}\delta b_{1}\right ] \overline{w}_{k_{1}}^{t}\right\ } e_{\theta}\left [ q_{01}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}\right ] ^{-1}\times\\ \left [ e_{\theta}\left ( q_{01}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}\right ) -i\right ] ^{m-2}e_{\theta}\left [ \overline{w}_{k_{1}}q_{0}\delta p_{1}\right ] \end{array } \right\ } _ { -\left ( l2.2\right ) } \\
+ { \displaystyle\sum\limits_{j=2}^{m-2 } } \left\ { \begin{array } [ c]{c}e_{\theta}\left [ \overline{z}_{k_{0}}\delta p_{0}^{t}\right ] \left [ e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ) \right ] ^{j-2}\times\\ e_{\theta}\left [ q_{01}\delta b_{1}\overline{z}_{k_{0}}\overline{w}_{k_{1}}^{t}\right ] e_{\theta}\left [ q_{01}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}\right ] ^{-1}\times\\ e_{\theta}\left ( q_{01}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}-i\right ) ^{m-1-j}e_{\theta}\left [ \overline{w}_{k_{1}}q_{0}\delta p_{1}\right ] \end{array } \right\ } _ { \_\left ( l2.3.j\right ) } \\ + e_{\theta}\left\ { \left [ q_{0}\delta p_{1}\delta b_{1}\right ] \overline { z}_{k_{0}}^{t}\right\ } \left [ e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}\right ) -i\right ] ^{m-3}e_{\theta}\left [ \overline{z}_{k_{0}}\delta p_{0}\right ] _ { \_\left ( l2.4\right ) } \end{array } \right\}\ ] ] it can be shown @xmath3311{c}e_{\theta}\left\ { \left [ q_{0}\delta b_{0}\right ] \overline{z}_{k_{0}}^{t}\right\ } e_{\theta}\left [ q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}\right ] ^{-1}\times\\ \left [ e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}\right ) -i\right ] ^{m-1}e_{\theta}\left [ \overline{z}_{k_{0}}\delta p_{0}\right ] \end{array } \right\}\end{aligned}\]]@xmath3312{c}e_{\theta}\left\ { \left [ q_{01}\delta b_{1}\right ] \overline{w}_{k_{1}}^{t}\right\ } e_{\theta}\left [ q_{01}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}\right ] ^{-1}\times\\ \left [ e_{\theta}\left ( q_{01}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}\right ) -i\right ] ^{m-1}e_{\theta}\left [ \overline{w}_{k_{1}}q_{0}\delta p_{1}\right ] \end{array } \right\}\end{aligned}\ ] ] @xmath1939 @xmath3313@xmath3314{c}e_{\theta}\left [ \overline{z}_{k_{0}}\delta p_{0}^{t}\right ] \left [ e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ) \right ] ^{j-2}\times\\ e_{\theta}\left [ q_{01}\delta b_{1}\overline{z}_{k_{0}}\overline{w}_{k_{1}}^{t}\right ] e_{\theta}\left [ q_{01}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}\right ] ^{-1}\times\\ e_{\theta}\left ( q_{01}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}-i\right ) ^{m - j}e_{\theta}\left [ \overline{w}_{k_{1}}q_{0}\delta p_{1}\right ] \end{array } \right\}\end{aligned}\ ] ] if @xmath3315 \left [ e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ) \right ] ^{m-3}\overline{z}_{k_{0 } } , $ ] then @xmath3316 e_{\theta}\left [ \overline{w}_{k_{1}}q_{0}\delta p_{1}\right ] -e_{\theta}\left\ { \zeta_{m}\left ( l_{0},\theta\right ) q_{0}\delta p_{1}\delta b_{1}\right\ } \\ & = \left ( -1\right ) ^{m-1}\left\ { \begin{array } [ c]{c}e_{\theta}\left ( q_{01}\delta b_{1}\zeta_{m}\left ( l_{0},\theta\right ) \overline{w}_{k_{1}}^{t}\right ) e_{\theta}\left ( q_{01}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}\right ) ^{-1}\times\\ \left ( e_{\theta}\left ( q_{01}\overline{w}_{k_{1}}\overline{w}_{k_{1}}^{t}-i\right ) \right ) e_{\theta}\left ( \overline{w}_{k_{1}}q_{0}\delta p_{1}\right ) \end{array } \right\ } \\ & + \left ( -1\right ) ^{m}e_{\theta}\left ( \begin{array } [ c]{c}\pi_{\theta}^{\bot}\left [ \left ( \begin{array } [ c]{c}q_{01}^{1/2}\delta b_{1}\zeta_{m}\left ( l_{0},\theta\right ) \end{array } \right ) |\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ] \\ \times\pi_{\theta}^{\bot}\left [ q_{0}q_{01}^{-1/2}\delta p_{1}|\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ] \end{array }
\right ) _ { -\left ( tb_{m-1,m-1}^{\left ( 2\right ) } \right ) } \ ] ] in summary@xmath3317 next we show @xmath3318@xmath3319 \left [ e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ) \right ] ^{m-3}\overline{z}_{k_{0}}\\ & = \left\ { \begin{array } [ c]{c}e_{\theta}\left [ \delta p_{0}\overline{z}_{k_{0}}^{t}\right ] \left ( \left ( \begin{array } [ c]{c}e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}\right ) ^{-1}+\\ i - e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}\right ) ^{-1}\end{array } \right ) \right ) \\ \times\left [ e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ) \right ] ^{m-3}\overline{z}_{k_{0}}\end{array } \right\ } \\ & = \left\ { e_{\theta}\left [ \delta p_{0}\overline{z}_{k_{0}}^{t}\right ] e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}\right ) ^{-1}\left [ e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ) \right ] ^{m-3}\overline{z}_{k_{0}}\right\ } \\ & + \left\ { \begin{array } [ c]{c}e_{\theta}\left [ \delta p_{0}\overline{z}_{k_{0}}^{t}\right ] e_{\theta } \left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}\right ) ^{-1}e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ) \\ \times\left [ e_{\theta}\left ( q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ) \right ] ^{m-3}\overline{z}_{k_{0}}\end{array } \right\}\end{aligned}\ ] ] applying the above expression of @xmath3320 to e.q@xmath3321 we find that @xmath3322 we want to mention that @xmath3323 @xmath3324 equals @xmath3325 , @xmath3326 @xmath3327 @xmath3328 @xmath3329 @xmath3330 and @xmath3331 therefore the results from this theorem is consistent with eq.([bsp ] ) .
technical details are not presented here .
eq.([ebm2 ] ) follows from eq.([ebm ] ) similarly as in the proof of theorem [ btomm ] and eq.([ebm3 ] ) can be derived from eq.([ebm2 ] ) by our assumption of rate optimality of the initial estimator .
next we prove eq.([tbm2]),@xmath3332 \right)\end{aligned}\ ] ] ( cauchy - shwartz inequality . the last equality can easily be shown as in the proof of theorem [ on])@xmath3333{c}\pi_{\theta}^{\bot}\left [ \left ( \begin{array } [ c]{c}\frac{\pi_{1}^{1/2}}{\widehat{\pi}_{1}^{1/2}}\delta b_{1}\pi_{\theta}\left ( q_{0}^{1/2}\delta p_{0}|\left ( q_{0}^{1/2}\overline{z}_{k_{0}}\right ) \right ) \end{array } \right ) |\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ] \\ \times\pi_{\theta}^{\bot}\left [ q_{0}q_{01}^{-1/2}\delta p_{1}|\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ] \end{array } \right ] \right\vert \\ & \leq\left\vert e\left [ \begin{array } [ c]{c}\pi_{\theta}^{\bot}\left [ \left ( \begin{array } [ c]{c}\delta b_{1}q_{01}^{1/2}\delta p_{0}\end{array } \right ) |\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ] \\ \times\pi_{\theta}^{\bot}\left [ q_{0}q_{01}^{-1/2}\delta p_{1}|\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ] \end{array } \right ] \right\vert \\ & + \left\vert e\left [ \begin{array } [ c]{c}\pi_{\theta}^{\bot}\left [ \left ( \begin{array } [ c]{c}\frac{\pi_{1}^{1/2}}{\widehat{\pi}_{1}^{1/2}}\delta b_{1}\pi_{\theta}^{\bot } \left ( q_{0}^{1/2}\delta p_{0}|\left ( q_{0}^{1/2}\overline{z}_{k_{0}}\right ) \right ) \end{array } \right ) |\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ] \\ \times\pi_{\theta}^{\bot}\left [ q_{0}q_{01}^{-1/2}\delta p_{1}|\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ] \end{array } \right ] \right\vert \\ & \leq o_{p}\left ( \max\left ( k_{1}^{-\left ( \min\left ( \beta_{\pi_{0}},\beta_{b_{1}}\right ) + \beta_{\pi_{1}}\right ) /d_{1}},k_{1}^{-\beta _ { \pi_{1}}/d_{1}}k_{0}^{-\beta_{\pi_{0}}/d_{0}}\left ( \frac{\log n}{n}\right ) ^{\frac{\beta_{b_{1}}}{d_{1}+\beta_{b_{1}}}}\right ) \right)\end{aligned}\ ] ] the last inequality holds because @xmath3334{c}\frac{\pi_{1}^{1/2}}{\widehat{\pi}_{1}^{1/2}}\delta b_{1}\pi_{\theta}^{\bot } \left ( q_{0}^{1/2}\delta p_{0}|\left ( q_{0}^{1/2}\overline{z}_{k_{0}}\right ) \right ) \end{array } \right ) |\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ] \right ) ^{2}\\ & \leq e\left ( \frac{\pi_{1}^{1/2}}{\widehat{\pi}_{1}^{1/2}}\delta b_{1}\pi_{\theta}^{\bot}\left ( q_{0}^{1/2}\delta p_{0}|\left ( q_{0}^{1/2}\overline{z}_{k_{0}}\right ) \right ) \right ) ^{2}\\ & \leq\left\vert \left\vert \frac{\pi_{1}^{1/2}}{\widehat{\pi}_{1}^{1/2}}\delta b_{1}\right\vert \right\vert _ { \infty}^{2}e\left ( \pi_{\theta}^{\bot } \left ( q_{0}^{1/2}\delta p_{0}|\left ( q_{0}^{1/2}\overline{z}_{k_{0}}\right ) \right ) \right ) ^{2}\ ] ] ( the first inequality holds because the orthogonal projection operator has a norm no greater than 1)@xmath3335{c}\pi_{\theta}^{\bot}\left [ \left ( \begin{array } [ c]{c}q_{01}^{1/2}\delta b_{1}e\left [ \delta p_{0}\overline{z}_{k_{0}}^{t}\right ] e\left [ q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}\right ] ^{-1}\\ \times\left ( e\left [ q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ] \right ) ^{m-2}\overline{z}_{k_{0}}\end{array } \right ) |\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ] \\ \times\pi_{\theta}^{\bot}\left [ q_{0}q_{01}^{-1/2}\delta p_{1}|\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ] \end{array } \right ] \right\vert \\ & \leq\left\ { \begin{array } [ c]{c}\left\ { e\left ( \begin{array } [ c]{c}q_{01}^{1/2}\delta b_{1}e\left [ \delta p_{0}\overline{z}_{k_{0}}^{t}\right ] e\left [ q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}\right ] ^{-1}\\ \times\left ( e\left [ q_{0}\overline{z}_{k_{0}}\overline{z}_{k_{0}}^{t}-i\right ] \right ) ^{m-2}\overline{z}_{k_{0}}\end{array } \right ) ^{2}\right\ } ^{1/2}\\ \times\left\ { e\left ( \pi_{\theta}^{\bot}\left [ q_{0}q_{01}^{-1/2}\delta p_{1}|\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ] \right ) ^{2}\right\ } ^{1/2}\end{array } \right\}\end{aligned}\ ] ] ( cauchy - shwartz inequality and projection operator has operator norm of 1)@xmath3336{c}\left\vert \left\vert q_{01}^{1/2}\frac{f_{1}}{\widehat{f}_{1}}\right\vert \right\vert _ { \infty}\left\vert \left\vert \delta b_{1}\right\vert \right\vert _ { \infty}\left\vert \left\vert q_{0}^{-1/2}\frac{\widehat{f}_{0}}{f_{0}}\right\vert \right\vert _ { \infty}\left\vert \left\vert \frac{f_{0}}{\pi _ { 0}\widehat{\pi}_{0}}\right\vert \right\vert _ { \infty}\\ \times\left\vert \left\vert \delta g_{0}\right\vert \right\vert _ { \infty } ^{m-2}\left\ { \int\left ( \pi_{0}-\widehat{\pi}_{0}\right ) ^{2}dl_{0}\right\ } ^{1/2}\times\\ \left\ { e\left ( \pi_{\theta}^{\bot}\left [ q_{0}q_{01}^{-1/2}\delta p_{1}|\left ( q_{01}^{1/2}\overline{w}_{k_{1}}\right ) \right ] \right ) ^{2}\right\ } ^{1/2}\end{array } \right\ } \\ & = o_{p}\left ( \left ( \frac{\log n}{n}\right ) ^{-\frac{\beta_{b_{1}}}{d_{1}+\beta_{b_{1}}}-\frac{\left ( m-2\right ) \beta_{g_{0}}}{d_{0}+\beta_{g_{0}}}}n^{-\frac{\beta_{\pi_{0}}}{d_{0}+\beta_{\pi_{0}}}}k_{1}^{-\beta_{\pi_{1}}/d_{1}}\right)\end{aligned}\ ] ] | robins et al , 2008 , published a theory of higher order influence functions for inference in semi- and non - parametric models .
this paper is a comprehensive manuscript from which robins et al , was drawn .
the current paper includes many results and proofs that were not included in robins et al due to space limitation .
particular results contained in the present paper that were not reported in robins et al include the following . given a set of functionals and their corresponding higher order influence functions ,
we show how to derive the higher order influence function of their product .
we apply this result to obtain higher order influence functions and associated estimators for the mean of a response @xmath0 subject to monotone missingness under missing at random .
these results also apply to estimating the causal effect of a time dependent treatment on an outcome @xmath0 in the presence of time - varying confounding .
finally , we include an appendix that contains proofs for all theorems that were stated without proof in robins et al , 2008 . the initial part of the paper is closely related to robins et al , , the latter parts differ .
specifically , we present a theory of point and interval estimation for nonlinear functionals in parametric , semi- , and non - parametric models based on higher order influence functions ( robins @xcite , sec . 9
; li et al .
@xcite , tchetgen et al , @xcite , robins et al , @xcite ) .
higher order influence functions are higher order u - statistics .
our theory extends the first order semiparametric theory of bickel et al . @xcite and van der vaart @xcite by incorporating the theory of higher order scores considered by pfanzagl @xcite , small and mcleish @xcite and lindsay and waterman @xcite .
the theory reproduces many previous results , produces new non-@xmath1 results , and opens up the ability to perform optimal non-@xmath1 inference in complex high dimensional models .
we present novel rate - optimal point and interval estimators for various functionals of central importance to biostatistics in settings in which estimation at the expected @xmath1 rate is not possible , owing to the curse of dimensionality .
we also show that our higher order influence functions have a multi - robustness property that extends the double robustness property of first order influence functions described by robins and rotnitzky @xcite and van der laan and robins @xcite . |
this work started with the discovery of new techniques for solving the friedmann equations analytically @xcite , and finding the complete set of solutions for the gravitational system that includes a scalar field @xmath0 with a special , but still typical , potential @xmath1 in a cosmological model @xcite@xcite .
the special techniques were direct outcomes from 2t - physics @xcite and 2t - gravity @xcite@xcite .
this led to an emergent formulation of gravity in 1t - physics with an additional local conformal ( weyl ) symmetry .
the cosmological model that emerged in this approach had a strong overlap with the cosmological colliding branes scenario @xcite@xcite developed from worldbrane notions @xcite in m - theory @xcite .
this connection led to a fruitful collaboration between our group , that includes c.h .
chen , and the group that includes paul steinhardt and neil turok .
our collaborative work has partially appeared in @xcite . here
i will summarize results that include some that are about to be published in @xcite .
our approach begins with the standard action typically used in cosmological models that describe a scalar field @xmath2 minimally coupled to gravity@xmath3{c}\frac{1}{2\kappa^{2}}r\left ( g\right ) -\frac{1}{2}g^{\mu\nu}\partial_{\mu } \sigma\partial_{\nu}\sigma - v\left ( \sigma\right ) \\
+ radiation+matter \end{array } \right\ } .\label{action}\ ] ] although not necessary , the scalar field @xmath4 could be identified with the dilaton in string theory , written in the einstein frame . for the purpose of studying cosmology ,
the relevant degrees of freedom are homogeneous fields that are only time dependent .
the space dependence is ignored in the leading approximation .
it can be argued @xcite that the space gradients are responsible for non - leading effects as compared to the dominant effects parameterized by only time dependent fields that we study here ( including anisotropy ) .
our cosmological analysis of this theory involves geometries with or without spacial curvature , as well as anisotropy , and includes radiation and other relativistic matter approximated as relativistic dust described by an energy - momentum tensor consistent with conformal symmetry .
the geometries we study have the form@xmath5 where @xmath6 is the scale factor that tracks the size of the universe .
we will use conformal time @xmath7 which is related to cosmic time @xmath8 by @xmath9 as in eq.([metric ] ) .
the 3-dimensional part @xmath10 of the geometries we study include * the isotropic frw universe with or without curvature @xmath11 @xmath12 * the anisotropic kasner metric with zero curvature @xmath13 where @xmath14 and @xmath15 are two dynamical degrees of freedom in the metric along with @xmath16 and * the anisotropic bianchi ix metric with curvature @xmath11@xmath17 where @xmath18 which are given in @xcite , depend on space coordinates ( and @xmath11 ) but not on @xmath19 the bianchi ix metric reduces to the frw metric with curvature @xmath11 in the isotropic limit @xmath20 and it reduces to the anisotropic kasner metric when the curvature vanishes @xmath21 the relevant degrees of freedom are @xmath22 here we have attached the label @xmath23 on the scale factor @xmath24 to emphasize that it is defined in the einstein frame .
we do this because we will also define other frames to solve the equations .
the equations of motion ( namely the friedmann - type equations ) are obtained from the action above .
however , it is convenient to provide an effective worldline mechanics - type action in the einstein frame @xmath25 that reproduces the same equations@xmath26{c}\frac{1}{e}\left [ -\frac{6}{2\kappa^{2}}\dot{a}_{e}^{2}+\frac{1}{2}a_{e}^{2}\dot{\sigma}^{2}+\frac{1}{2}a_{e}^{2}\dot{\alpha}_{1}^{2}+\frac{1}{2}a_{e}^{2}\dot{\alpha}_{2}^{2}\right ] \\ -e\left [ a_{e}^{4}v\left ( \sigma\right ) + \rho_{0}-\frac{6k}{2\kappa^{2}}a_{e}^{2}v\left ( \alpha_{1},\alpha_{2}\right ) \right ] \end{array } \right\ } .\label{seffe}\ ] ] here the constant @xmath27 represents the energy density of radiation and relativistic matter .
it comes from approximating these degrees of freedom as `` dust '' modeled as a perfect fluid , with an energy - momentum tensor of the form @xmath28 that is taken to be traceless ( @xmath29 ) and behave as @xmath30 to be consistent with the conformal symmetry for massless particles ( including radiation and matter ) .
the potential energy term for the anisotropy involving @xmath31 @xmath32 is obtained by computing @xmath33 for the bianchi ix metric @xcite .
the additional degree of freedom @xmath34 plays the role of the metric on the worldline and ensures @xmath7-reparameterization invariance .
this effective action can be derived directly from the 4-dimensional action by inserting the metric including the lapse function @xmath35 where the lapse function @xmath36 is the einbein on the worldline .
the equation of motion for @xmath37 imposes the zero energy constraint which is equivalent to the @xmath38 einstein equation .
eventually we work in the gauge @xmath39 which corresponds to the conformal time in eq.([metric ] ) .
similarly , it can be checked that this effective action reproduces all friedmann - type equations that are derived from the original action in eq.([action ] ) .
for our model we chose the following potential energy @xmath40 after insuring stability by requiring @xmath41 the plot of this potential , which looks like a well or double well for various values and signs of @xmath42 shows that it has features that are similar to other potentials used in cosmological applications .
the reason for choosing this specific potential is that we can use some special tricks to solve the friedmann equations exactly in the frw universe and find all of the solutions analytically @xcite@xcite@xcite .
we can similarly solve the equations analytically for a few other potentials , as presented in a future publication .
the tricks for solving the equations are actually deeply related to the gauge symmetries and shadow phenomena in 2t - physics @xcite@xcite .
but , in the current setting with the potential @xmath43 in ( [ v ] ) , it is possible to introduce them directly as the bars - chen - steinhardt - turok ( bcst ) transformation of the fields @xmath44 to a new basis @xmath45 as follows @xmath46 then the action @xmath25 in eq.([seffe ] ) is rewritten in the new basis as follows @xmath47{c}\frac{1}{e}\left [ -\frac{1}{2}\dot{\phi}_{\gamma}^{2}+\frac{1}{2}\dot { s}_{\gamma}^{2}+\frac{\kappa^{2}}{12}\left ( \phi_{\gamma}^{2}-s_{\gamma}^{2}\right ) \left ( \dot{\alpha}_{1}^{2}+\dot{\alpha}_{2}^{2}\right ) \right ] \\ -e\left [ \phi_{\gamma}^{4}f\left ( s_{\gamma}/\phi_{\gamma}\right ) + \rho _ { 0}-\frac{k}{2}\left ( \phi_{\gamma}^{2}-s_{\gamma}^{2}\right ) v\left ( \alpha_{1},\alpha_{2}\right ) \right ] \end{array } \right\ } , \label{seffg}\ ] ] where @xmath48 when @xmath1 is given as in eq.([v ] ) . in the absence of anisotropy , @xmath20 @xmath49 this form of the action , with the special potential , shows that the fields @xmath50 are decoupled from each other , except for the zero energy condition that results from the @xmath51-equation of motion .
the decoupled second order field equations for @xmath52 and @xmath53 are easily solved because they are analogous to a particle in a potential .
the solutions are given in terms of jacobi elliptic functions .
the zero energy condition merely relates the integration variables @xmath54 ( energy levels for @xmath55 ) in the form @xmath56 and @xmath57 so that there is only one integration variable , @xmath58 and two initial values @xmath59 due to time translation symmetry one of the initial values has no physical significance ; therefore @xmath60 are the only two integration parameters .
together with the 4 parameters of the model @xmath61 these 6 parameters determine the properties of all the solutions in the absence of anisotropy .
this shows that the bcst transformation makes the friedmann equations solvable analytically and leads to the complete set of solutions given in @xcite@xcite@xcite as outlined below .
cclecturefig1_fs_plane.eps + fig.1- gravity @xmath62 in left / right quadrants .
antigravity @xmath63 in top / bottom quadrants .
[ ccrfig1 ] eq.([bcst1 ] ) is only half of the bcst transformation between the @xmath44 and @xmath64 bases .
the form in eq.([bcst1 ] ) is initially defined only for the region of the @xmath65 plane that satisfies @xmath66 which corresponds to the left and right quadrants as depicted in fig.1 .
the cosmological singularity at @xmath67 corresponds to the 45@xmath68 lines , which may be considered a lightcone in the @xmath64 space .
the study of geodesic completeness at the @xmath67 singularity amounts to the question of how geodesics cross the lightcone in @xmath64 field space . through our explicit analytic solutions
@xcite@xcite@xcite we learned that _
generic trajectories in the absence of anisotropy _
cross this lightcone and continue to develop in the entire @xmath65 plane . therefore
, geodesic completeness leads us to expand the domain of the einstein frame beyond the @xmath69 region by including the full @xmath65 plane .
this extension of the field domain , including the top and bottom quadrants , reminds us of a similar extension of spacetime ( rather than field space ) via the kruskal - szekeres coordinates of the schwarzschild blackhole . to complete the bcst transformation
, we also need the inverse transformation which involves the four quadrants in the @xmath50 space .
this is given by @xmath70{c}\frac{\sqrt{6}}{\kappa}\sqrt{\left\vert z\right\vert } \cosh\left ( \frac{\kappa\sigma}{\sqrt{6}}\right ) , \text { if } z>0\\ \frac{\sqrt{6}}{\kappa}\sqrt{\left\vert z\right\vert } \sinh\left ( \frac{\kappa\sigma}{\sqrt{6}}\right ) , \text { if } z<0 \end{array } \right .
, \;s_{\gamma}=\pm\left\ { \begin{array } [ c]{c}\frac{\sqrt{6}}{\kappa}\sqrt{\left\vert z\right\vert } \sinh\left ( \frac{\kappa\sigma}{\sqrt{6}}\right ) , \text { if } z>0\\ \frac{\sqrt{6}}{\kappa}\sqrt{\left\vert z\right\vert } \cosh\left ( \frac{\kappa\sigma}{\sqrt{6}}\right ) , \text { if } z<0 \end{array } \right . \label{bcst2}\ ] ] actually the @xmath71 and @xmath72 field domains are separated from each other by the spacetime singularity at @xmath73 so @xmath71 is satisfied in a patch of spacetime , while @xmath72 is satisfied in a different patch of spacetime .
these patches communicate with each other through the cosmological singularity , or equivalently through the whole lightcone in the @xmath74 plane .
allowing @xmath65 to be extended to the full @xmath65 plane , including the top and bottom quadrants ( as guided by the solutions ) , defines a continuation of the einstein frame to a larger domain of field space .
we found that the full @xmath64 provides a geodesically complete geometry , while the domain @xmath71 by itself is geodesically incomplete .
we find that in the extended domain the physics described by @xmath75 includes patches of spacetime in which there is gravity when @xmath76 or antigravity when @xmath77 the actions @xmath78 are equivalent to each other via the bcst transformation only in the sector in which @xmath79 is positive@xmath80 in the antigravity patches the inverse bcst transformation produces again an einstein frame , and an action similar to @xmath25 , but with a negative newton constant . to better understand the meaning of the different patches we need the complete theory , not only the effective theory @xmath75 for the cosmological degrees of freedom .
this is given from first principles by a slight extension of the einstein action by including only gauge degrees of freedom associated with a local scaling symmetry , called weyl symmetry .
the local scaling transformation , given by @xmath81 does not allow the usual einstein - hilbert term , but allows conformally coupled scalars .
the following action contains two conformally coupled scalars @xmath82
interacting with the curvature term with the coefficient @xmath83 dictated by the gauge symmetry @xmath84 here @xmath85 is an arbitrary function of the scale invariant ratio @xmath86 fermionic and gauge fields , as well as more conformally coupled scalars , can be added , as part of a complete model , without breaking the conformal gauge symmetry @xcite . here
@xmath87 appears as a gravitational parameter that replaces the newton constant @xmath88 . to have the possibility for it to be positive
, we must have one scalar , namely @xmath89 to have the wrong sign kinetic term .
if @xmath90 , like @xmath91 has the correct sign , then we would end up with a purely negative gravitational parameter .
hence the relative minus sign in @xmath92 is required .
the wrong sign kinetic term makes @xmath90 potentially a ghost .
however , the local gauge symmetry compensates for the ghost , thus insuring unitarity of the theory beyond @xmath89 consistently with the weyl symmetry @xcite .
only the form given here parallels the structure @xmath93 of the complete set of solutions .
other schemes of introducing scalars with a weyl symmetry are not useful to understand the solutions and their extension of the einstein domain .
actually , the path of discovery was to first obtain this action from 2t - physics @xcite@xcite , which then enabled us to find the solutions @xcite . ] . the usual einstein gravity in eq.([action ] )
can be obtained by choosing a gauge for the local scale symmetry . in particular ,
in the einstein gauge we denote the fields by @xmath94 with a subscript @xmath58 and we choose @xmath95 for all spacetime @xmath96 thus generating the gravitational constant .
then writing @xmath97 we find that the conformally gauge invariant action ( [ c - action ] ) yields the familiar action in eq.([action ] ) with an einstein - hilbert term @xmath98 and a minimally coupled scalar field @xmath4 , and any potential energy @xmath99 another convenient gauge choice is the @xmath100-gauge in which we denote the fields by @xmath101 with a subscript @xmath100 . in this gauge
we choose the conformal factor of the metric @xmath102 for all spacetime @xmath103 in particular , in the frw , kasner or bianchi ix spacetimes of eqs.([metric]-[bianchi9 ] ) taken in the gamma gauge @xmath104 , we obtain simple expressions for the curvature , such as @xmath105 , etc .. inserting these expressions in the full action in eq.([c - action ] ) we derive the gauge fixed action @xmath75 for cosmological applications , precisely as given in eq.([seffg ] ) .
it is useful to point our that the following quantity is gauge invariant in all spacetime @xmath106 in the einstein gauge in the gravity sector given in eq.([e - gauge ] ) it is equal to @xmath107 , while in the einstein gauge in the antigravity sector ( interchanging @xmath108 and @xmath109 in eq.([e - gauge ] ) ) it is equal to @xmath110 furthermore , in the @xmath100-gauge it is equal to @xmath111 . equating these quantities to each other by gauge invariance immediately gives part of the bcst transformation in eq.([bcst1 ] ) .
this shows that the bcst transformation in eqs.([bcst1],[bcst2 ] ) is actually the gauge transformation from the einstein gauge ( both gravity and antigravity regions ) to the @xmath100-gauge and vice - versa .
now it is clear that the @xmath100-gauge of the action ( [ c - action ] ) is valid in the entire @xmath112 plane in all spacetime , for all values of the gauge invariant @xmath113 .
but the usual einstein theory as given in eq.([action ] ) is possible only in the left and right quadrants of fig.1 where @xmath114 or equivalently when the gauge invariant @xmath115 is positive . to give an einstein frame description of the top and bottom quadrants of fig.1 , where @xmath116 and the gauge invariant @xmath117 is negative , we need an einstein gauge just like eq.([e - gauge ] ) by interchanging @xmath108 with @xmath118 the result of gauge fixing in the region @xmath116 is an einstein action just like ( [ action ] ) , but with the gravitational constant @xmath88 replaced by @xmath119 and the wrong sign kinetic term for the sigma field .
evidently , in this system there is antigravity rather than gravity since the newton constant has the opposite sign . comparing the gravity / antigravity regions further in their respective einstein frames , note that the potential energy @xmath120 does not change sign , but is replaced by a new one @xmath121 which is related to the old one in eq.([v ] ) by interchanging @xmath108 with @xmath118 any additional terms in a complete action do not change sign either , but any terms that initially contain @xmath122 would look different in the gravity / antigravity sectors by the interchange of @xmath108 with @xmath109 .
the @xmath100-gauge interpolates smoothly between the gravity and antigravity regions and this is why , among other things , it is helpful with the geodesic completeness of the geometry . after discussing the classical solutions of the cosmological equations
, we will return to the antigravity region to discuss possible instabilities due to the sign changes in the full action ( [ c - action ] ) , and what they may mean .
first , recall the geodesic incompleteness of the geometry in the usual einstein frame , which is an old problem that has been set aside .
this is typical for any frw cosmological solution with a big bang singularity , because the geometry is incomplete at the singularity as @xmath123 .
a massive particle in the flat frw universe satisfies the geodesic equation @xmath124 where @xmath125 is the conserved momentum of the particle . from this
it is straightforward to compute that it takes a finite amount of conformal time , cosmic time , or proper time , to reach the big bang singularity from anywhere in the universe .
therefore the particle trajectory is artificially stopped at the singularity in a finite amount of proper time .
this is geodesic incompleteness .
it begs the question : what was there just before the big bang ? what would we find if we could allow our watch to run beyond the singularity where it was artificially stopped ?
a geodesically complete geometry would provide at least a partial answer to this question .
of course , modifications of the geometry or even a completely different description may be expected from quantum gravity or string theory .
unfortunately , the question remains even more obscure in these formalisms because of the lack of proper understanding of how they should be applied to the cosmological setting .
it is therefore not unreasonable to at least , at first , try to answer such a question in the setting of classical gravity .
having the complete set of solutions analytically allowed us to answer such questions , and understand how to complete the geometry in the classical physics setting , thus leading to some surprising behavior , as follows .
the complete set of solutions of the cosmological equations have been analyzed in detail in @xcite@xcite@xcite .
these solutions are obtained with _ no restrictions on either the parameters of the model or the initial conditions on the fields_. before the important role of anisotropy is taken into account near the singularity at @xmath126 , there are 6 unrestricted parameters in our solutions as indicated following eq.([seffg ] ) .
all 6 parameters are available to try to fit cosmological data far away from the singularity . in the absence of anisotropy
, the _ generic solution _ for @xmath127 behaves in various detailed ways in various regions of the 6 parameter space .
this is given in the appendix of @xcite where we identified 25 different regions of the 6 parameter space in which the analytic expression is different for each case separately .
the trajectory of the generic solution in the absence of anisotropy can be plotted parameterically in the @xmath64 plane of fig.1 .
this shows that the generic trajectory crosses the lightcone in fig.1 at any point ( as determined by the 6 parameters ) so that @xmath128 keeps changing sign as the universe moves back and forth from the gravity patch to the antigravity patch .
of course , to do so , the scale factor @xmath129 vanishes at each crossing , so that the gravity / antigravity patches are connected to each other only through the cosmological singularity .
the behavior of the generic solution near the singularity changes dramatically in the presence of anisotropy . as will be outlined below , an important effect of anisotropy is to focus the trajectory of the generic solution to pass through the origin of the @xmath64 plane , such that the generic trajectory can cross the lightcone in fig.1 only at the origin where @xmath65 vanish simultaneously . by restricting the parameter space or initial conditions
, we find that there is a subset of special solutions that are geodesically complete in the einstein frame purely in the gravity patch .
these are the zero - size and finite - size bounces shown in figs.(2,3,4,5 ) . as seen in figs.(4,5 )
, their trajectories never reach into the antigravity sector and they never cross the lightcone in fig.1 except at the origin .
@xmath130{cc}{\parbox[b]{2.9006in}{\begin{center } \includegraphics [ height=1.804 in , width=2.9006 in ] { cclecturefig2-a+sigma_b-.eps}\\ fig.2 - zero - size bounces in a cyclic universe .
$ a_e\left ( \tau\right ) $ and $ \sigma\left ( \tau\right ) $ for $ b<0 $ is depicted . for $ b>0
$ turnaround is at $ a_e\rightarrow\infty.$ \end{center } } } & { \parbox[b]{2.8928in}{\begin{center } \includegraphics [ height=1.7997 in , width=2.8928 in ] { cclecturefig3-f+s_b-.eps}\\ fig.3 - $ \phi\left ( \tau\right ) , s\left ( \tau\right ) $ for $ b<0,$ with synchronized initial values and relatively quantized periods . for
$ b>0 $ turnaround is at $ \phi\rightarrow\infty.$ \end{center}}}\end{array } $ ] the zero - size - bounce ( see figs.(2,3 ) ) describes a cyclic universe that _ contracts to zero size and then bounces back smoothly from zero size_. it expands up to either a finite size ( when @xmath131 ) or infinite size ( when @xmath132 ) and then turns around to repeat the cycle . in these solutions , as described in detail in @xcite@xcite , the behavior of @xmath129 near the singularity at @xmath133 is smooth ( if @xmath27 or @xmath11 take generic values , then @xmath134 ; if both @xmath135 vanish or take some special values , then @xmath136 also , the behavior of the potential and kinetic energy terms for the scalar field @xmath137 near the singularity are surprising .
namely , contrary to the generic solution , the potential energy dominates over the kinetic energy so that the equation of state @xmath138 is negative near the singularity . according to common lore
, it was thought that such zero - size - bounce solutions would not exist because they would violate the null energy condition ( nec ) .
however , this is not the case .
the nec is satisfied because there is a singularity at zero size , and this is the exception allowed according to the nec theorems .
we found all such solutions and classified them in @xcite@xcite , thus providing a rich class of examples of zero - size - bounce cyclic universes , characterized by arbitrary values of the parameters @xmath139 plus one additional quantized parameter .
to make the zero - size - bounce happen , a synchronization of initial conditions and a quantization condition among the 6 available parameters must be imposed .
these properties are illustrated in fig.3 , where it is seen @xmath140 is imposed as an initial condition , and the periods of oscillations of @xmath141 and @xmath142 are quantized relative to each other . hence such solutions are characterized by 4 continuous and 1 quantized parameter , rather than the 6 continuous parameters of the generic solutions . in that sense
the zero - size - bounce solutions are a set of measure zero .
so , statistically , it does not seem likely that the universe would choose such a solution over the generic solution .
@xmath130{cc}{\parbox[b]{2.9144in}{\begin{center } \includegraphics [ height=0.5076 in , width=2.9144 in ] { cclecturefig4-parametric - b-.eps}\\ fig.4 - geodesically complete trajectory only in the gravity sector , for $ b<0.$ \end{center } } } & { \parbox[b]{3.2059in}{\begin{center } \includegraphics [ height=0.2741 in , width=3.2059 in ] { cclecturefig5parametric - b+.eps}\\ fig.5 - geodesically complete trajectory only in the gravity sector , for $ b>0.$ \end{center}}}\end{array } $ ] however , anisotropy plays a very important role in providing an attractor mechanism such that , for typical initial conditions away from the singularity , all trajectories are attracted to the origin of the @xmath143 plane and can cross the lightcone in fig.1 only at the origin . in that sense
the type of solution depicted in figs.(2 - 5 ) is not too far from being generic once anisotropy is taken into account .
in particular the behavior away from the origin is a good approximation .
nevertheless , what goes on in the singularity region is quite different as seen in figs.(8 , 9 ) and discussed below . @xmath130{cc}{\parbox[b]{2.8651in}{\begin{center } \includegraphics [ height=1.7824 in , width=2.8651 in ] { cclecturefig6-finitebounce.eps}\\ fig.6 - finite - size - bounce , $ \left\vert \phi_\text{min}\left ( \tau\right ) \right\vert > k/4b>\left\vert s_\text{max}\left ( \tau\right ) \right\vert .$ \end{center } } } & { \parbox[b]{2.9006in}{\begin{center } \includegraphics [ height=1.804 in , width=2.9006 in ] { cclecturefig7-finitebounce2.eps}\\ fig.7 - finite - size - bounce , \\[0pt]$a_e^2\left ( \tau\right ) $ never vanishes .
\end{center}}}\end{array } $ ] the finite - size - bounce ( see figs.(6,7 ) ) describes a universe that contracts up to a minimum non - zero size and then bounces back into an expansion phase up to infinite size . as the universe turns around to repeat such cycles the minimum size is not necessarily the same , as this depends on the parameters .
such solutions occur when the parameters satisfy , @xmath144 and @xmath145 .
note that there are still 6 parameters , so this is not a set of measure zero , but it is a restricted region of parameter space or initial values .
the analytic solution is given explicitly in @xcite@xcite .
it is useful to write the equations of motion for the fields @xmath146 in terms of the canonical momenta instead of their velocities .
these are given by @xmath147 , @xmath148 and @xmath149 then the energy constraint that follows from the action @xmath75 or @xmath25 takes the following form when written in terms of @xmath150@xmath151 -\frac{6kv\left ( \alpha_{1},\alpha_{2}\right ) } { \kappa^{2}z},\label{adot}\ ] ] where @xmath152 for the generic solution ( i.e. unlike the zero - size - bounce solution , assuming typically @xmath153 are non - vanishing at the singularity ) , the kinetic term @xmath154 is the dominant term in the energy equation as we approach the singularity @xmath155 the next to the leading term is @xmath156 while the potential terms @xmath157 are subdominant . keeping the leading and next to the leading terms , it is possible to integrate all the friedmann equations to obtain the behavior of the general solution near the singularity . in the absence of the potentials ,
the momenta @xmath158 are all conserved ; then the solution is uniquely given as follows ( where we use units with @xmath159 @xmath160 as well as @xmath161 @xmath162 where @xmath163 is an integration constant that amounts to an initial value for @xmath4 .
similarly , the solutions for @xmath164 near the singularity are @xmath165 where @xmath166 are integration constants that amount to initial values for the @xmath167 .
the solutions for @xmath168 are plotted in fig.8 , where we see that @xmath169 changes sign .
this shows that the trajectory of the universe inevitably passes through antigravity .
furthermore , the parametric plot in fig.9 shows that there is an attractor mechanism that forces the trajectory to pass through the origin of the @xmath74 plane .
@xmath130{cc}{\parbox[b]{3.0251in}{\begin{center } \includegraphics [ height=1.881 in , width=3.0251 in ] { cclecturefig8-z+sigma.eps}\\ fig.8 - plots of $ z\left ( \tau\right ) $ and $ \sigma\left ( \tau\right ) , $ with $ all$ parameters $ \rho_0,p_\sigma , p_\alpha , t,$ set to 1 numerically .
\end{center } } } & \raisebox{-0pt}{\parbox[b]{3.2128in}{\begin{center } \includegraphics [ height=1.8922 in , width=3.2128 in ] { cclecturefig9-gravloop.eps}\\ fig.9 - antigravity loop in the parametric plot of $ \phi\left ( \tau\right ) , s\left ( \tau\right ) .$ \end{center}}}\end{array } $ ] from these plots it is seen that the universe contracts to zero size with a big crunch , passing through the singularity at @xmath170 , it then expands in a region of antigravity while @xmath171 then contracts again to zero size at @xmath172 , and then emerges with a big bang in a region of gravity where @xmath173 the duration of the antigravity phase is @xmath174 the essential parameters that control the properties of this unique solution in the vicinity of the singularity are @xmath175 they are related to the kinetic energy of the sigma field , anisotropy and radiation(i.e .
all relativistic particles in the early universe ) respectively .
they are all typically non - zero and not negligible in the vicinity of the singularity .
note that _ even if anisotropy _
@xmath176 _ is infinitesimal , the attraction to the origin _
@xmath177 _ _ is inevitable__. the parameter space for which the potentials can be neglected is estimated by inserting the solution back into the the equations and comparing the magnitude of the potential energy terms to the kinetic energy and radiation terms . from this
we determined that the radiation term @xmath178 in eq.([adot ] ) beats the potential @xmath1 by more than two powers of @xmath7 as @xmath179 and similarly for @xmath180 so @xmath181 is negligible near the singularity .
similarly @xmath182 is negligible compared to radiation as a function of time _ near _ the singularity ( even when @xmath1830 ) provided @xmath184 .
this condition is also compatible with a weaker condition ( @xmath185 , see @xcite ) that avoids the mixmaster behavior discussed in @xcite and is consistent with the discussions in @xcite@xcite .
the attractor mechanism is an almost model independent universal behavior since the potential energy @xmath1 is negligible close to the singularity .
anisotropy in even tiny amounts forces the universe to behave as described above .
the issue of _ initial conditions for all degrees of freedom in our model becomes partially resolved _ because at the singularity @xmath186 we must have the following automatically enforced initial values , independent of any other arbitrarily chosen initial conditions at some other time , @xmath187 rather than only @xmath188 .
hence , we must take @xmath189 as the dynamically enforced boundary values at all cosmological singularities .
this reduces the number of arbitrary initial values .
now we see that the zero - size - bounce solutions described in the previous section ( figs.(2 - 5 ) ) are not really of measure zero .
they do take into account effectively the dynamically enforced initial values of eq.([initial ] ) .
although these solutions are not capable of describing the detailed ( antigravity loop ) behavior near the singularity , they do describe the correct analytic behavior away from the singularity where anisotropy becomes negligibly small .
a more accurate geodesically complete solution is the combination of the near - singularity and far - from - singularity solutions which can be constructed by matching them at some time sufficiently away from the singularity .
the physics during the antigravity period remains cloudy in our investigation so far .
we discussed the trajectory of the universe on the average , including the antigravity period .
however , once in this regime , we may expect some violent behavior and backreaction because of antigravity , and these may affect the trajectory during the antigravity period .
first , because of the opposite sign of the gravitational constant we expect that chunks of similar matter will repel each other rather than attract .
this is an invitation to investigate the nature of interactions , including fluctuations on top of our background solution , which so far we have not included in the analysis .
furthermore , there are additional phenomena we should expect due to interactions as seen by investigating the form of the action in the antigravity regime , as follows . because of the sign change of the effective newton constant one may argue that fluctuations of the metric , namely the spin-2 gravitons , will now come with the wrong sign kinetic terms .
there are no gauge symmetries to remove them , hence , they will behave like negative energy fluctuations , which could destabilize the vacuum by the emission of negative energy gravitons together with lots of positive energy matter particles .
there is however a counter effect .
once particle production begins , it automatically implies a back reaction that changes the energy density @xmath190 for all relativistic matter that enters in the cosmological equations .
this effectively amounts to a huge increase of the parameter @xmath27 once we enter the antigravity regime .
so , rather than solving the cosmological equations with a constant @xmath27 in the antigravity region , we should allow @xmath27 to increase dramatically by large amounts . then according to our rough formula in eq.([duration ] ) for the duration of the antigravity regime ,
the duration gets shortened as @xmath27 increases . in other words , particle production by antigravity
shortens its duration .
the particle production and the corresponding jump of @xmath27 need to be better understood .
these are effects that oppose each other .
it is difficult to estimate the balance of these effects with our current understanding of the corresponding physics .
of course , quantum gravity effects need to be also included . therefore , it would be very interesting to study similar circumstances in the framework of string theory , although the requisite technical tools may not be available at this time . to formulate the antigravity aspects in string theory
we could use the bcst field transformations given in eqs.([bcst1],[bcst2 ] ) , but even better would be the inclusion of the analog of the weyl symmetry in the framework of string theory .
we should also mention that the phenomenon of the sign change of the newton constant transcends the specific simple model in eq.([c - action ] ) .
the phenomena we have found should also be expected generically in supergravity theories coupled to matter whose formulation include a similar factor that multiplies @xmath191 in supergravity , that factor is related to what is called the kahler potential , and this factor , combined with the usual einstein - hilbert term , is not generally positive definite @xcite . in the past , typically it was assumed that the overall factor is positive and investigations of supergravity proceeded only in the positive regime .
a discussion of the field space in the positive sector for general @xmath192=2 supergravity can be found in @xcite .
however , our results suggest that generically the overall factor can and will change sign dynamically , and therefore antigravity sectors similar to our discussion in this paper should be expected in typical supergravity theories .
furthermore , we can also draw an analogy to similar factors that occur in front of the kinetic terms for gauge bosons and scalar fields @xcite .
these too can potentially change sign . in the case of gauge bosons , since the kinetic term takes the form @xmath193 where @xmath194 represent electric and magnetic fields , a sign change could be reinterpreted by an electric - magnetic duality transformation .
indeed this type of duality was explored in depth at the quantum level for @xmath192=2 super yang - mills theory in @xcite .
it is therefore interesting to point out that something similar may be part of the phenomena in the case of the gravity / antigravity transitions we have uncovered .
having mentioned a few of the possible physical effects in the antigravity regime , we will proceed by assuming that the destabilizing effects of antigravity are not a disaster , but rather they are the signals of interesting physics that remains to be understood better .
we expect that the general nature of our solution survives and still describes the interesting effects we discovered , namely the attractor mechanism and the presence of antigravity for some period of time between a crunch and the following bang .
in this talk i emphasized the role of analytic solutions of cosmological equations to discuss geodesic completeness through the big bang singularity and the discovery of some new phenomena by focussing on questions that could not be answered with only approximate solutions . until better understood in the context of quantum gravity , or string theory , our results
should be considered to be a first pass for the types of new questions they raise and the answers they provide .
here is a brief summary of the main points we have learned .
* we found new techniques to solve cosmological equations analytically .
this led to all the solutions for several special potentials @xmath195 ( only one of those was discussed in this talk ) .
several model independent general results that followed from this include geodesic completeness , and an attractor mechanism to the origin of the @xmath45 plane , which resolves partially questions on initial values .
* antigravity is very hard to avoid .
anisotropy + radiation + kinetic energy of @xmath196 require the antigravity epoch . for close to a year we tried to find models and mechanisms to avoid antigravity ( i.e. when _ all _ solutions of a geodesically complete model are included ) .
the failure to find such mechanisms finally led us to take antigravity seriously .
* we studied the wheeler - dewitt equation to take into account some quantum effects for the same system that we analyzed classically .
we could solve some cases exactly , others semi - classically .
this will be published at a later time . but let me mention that we arrived substantially to the same conclusions that we learned by studying the purely classical system .
* will the new insights we obtained survive the effects of a full quantum theory ?
my expectation is affirmative .
perhaps the most satisfactory approach is to try to study the same issues in the context of string theory .
however , this is actually quite challenging since string theory has not been developed sufficiently in the cosmological context .
* i emphasize that the structure of the theory in eq.([c - action ] ) , the methods of solutions , and the phenomena that ensued , are historically direct predictions of 2t - physics in 4 + 2 dimensions @xcite@xcite .
the model corresponds to the conformal shadow of 2t - physics which takes the form of a familiar 1t - field theory setting .
other shadows @xcite of the same 2 t theory which would be in the form of dual 1t - field theories , could also play an important role to understand the system and its physical interpretation . *
open questions include : are there observational effects today of a past antigravity period ?
this is an important ongoing project for our group .
this involves the study of small fluctuations and the fitting to current observations of the cmb .
much remains to be understood , including quantum gravity and string theory effects , but it is clear that previously unsuspected phenomena , including an attractor mechanism and antigravity , are at work close to the cosmological singularity . the technical tools to study such issues in the context of a full quantum theory of gravity are yet to be developed .
so this is an important challenge to the theory community , including string theorists , since here are some deep physics issues where string theory , if it is the correct approach at those scales , should play its most important role as a quantum theory of gravity .
perhaps duality transformations of the type that occur in electric - magnetic duality @xcite could play a role as mentioned earlier . our understanding of cosmology and of the beginning of the universe could not be clarified until an understanding of the antigravity effects are obtained in a full quantum theory of gravity , and the associated phenomena are taken into account in comparing to cosmological observations .
this research was partially supported by the u.s .
department of energy under grant number de - fg03 - 84er40168 .
it was also partially supported by princeton university and the perimeter institute during my sabbatical at those institutions .
i would like to thank my collaborators c.h .
chen , paul steinhardt and neil turok with whom this research was conducted .
i also thank niayesh afshordi , tom banks , latham boyle , bernard dewit , and martin rocek for helpful conversations . | this talk is about solving cosmological equations analytically without approximations , and discovering new phenomena that could not be noticed with approximate solutions .
we found all the solutions of the friedmann equations for a specific model , including all the zero - size - bounce solutions that do not violate the null energy condition , as well as all the finite - size - bounce solutions , and then discovered model independent phenomena . among them
is the notion of geodesic completeness for the geometry of the universe . from this
we learned a few new general lessons for cosmology . among them
is that anisotropy provides a model independent attractor mechanism to some specific initial values for cosmological fields , and that there is a period of antigravity in the history of the universe .
the results are obtained only at the classical gravity level .
effects of quantum gravity or string theory are unknown , they are not even formulated , so there are new theoretical challenges . |
for a given sequence @xmath8 of nonnegative random variables on a probability space @xmath9 with @xmath10 a.s .
, consider the stochastic fixed - point equation ( sfpe ) @xmath11 where @xmath12 are i.i.d . ,
nonnegative and independent of @xmath5 and where @xmath13 is stipulated on @xmath14 . a distribution @xmath15 on @xmath16
is called a solution to ( [ minfp1 ] ) if this equation holds true with @xmath17 , and it is called positive if @xmath18 . note that @xmath19 , @xmath20 the dirac measure at 0 , always provides a trivial solution .
the set of all solutions @xmath21 will be denoted as @xmath22 hereafter , or as @xmath23 if we want to emphasize its dependence on @xmath5 .
we will make no notational distinction between a distribution @xmath15 and its left continuous distribution function , and we denote by @xmath24 the associated survival function , _
i.e. _ , @xmath25 . for @xmath26 ,
( [ minfp1 ] ) may be rewritten in terms of @xmath24 as @xmath27 for all @xmath28 .
denote by @xmath29 the spaces of probability measures on @xmath16 and @xmath30 $ ] , respectively . defining the map @xmath31 by @xmath32 we see that , formally speaking , @xmath22 is nothing but the set of fixed points @xmath33 of @xmath34 , that is @xmath35 .
sfpes of type or similar with a min- or max - operation involved turn up in various fields of applied probability like probabilistic combinatorial optimization , the run - time analysis of divide - and - conquer algorithms or branching particle systems , where they typically characterize the asymptotic distribution of some random variable of interest ( see _ e.g. _ @xcite , @xcite and @xcite ) .
in particular , the probabilistic worst - case analysis of hoare s find algorithm leads to the following fixed - point equation for the distributional limit @xmath36 of the linearly scaled maximal number of key comparisons of .@xmath37 where @xmath38 is a uniform @xmath39 $ ] random variable and @xmath40 are independent copies of the random variable @xmath36 ( see ( * ? ? ?
* theorem 1 ) ) .
more general , in the analysis of divide - and - conquer algorithms , equations of the form @xmath41 appear ( _ cf .
_ @xcite and @xcite ) .
further , from a species competition model , the following fixed - point equation @xmath42 arises , where @xmath43 are the points of a poisson process at rate @xmath44 and @xmath45 is an @xmath46 variable independent of the process @xmath43 ( see ( * ? ? ?
* example 38 ) ) . by a theorem of rschendorf ( see ( * ? ? ?
* theorem 4.2 ) ) , under appropriate conditions on the random coefficients of the sfpes above , there is a one - to - one relationship between the solutions of the max - type equation and the corresponding homogeneous equation .
therefore , it is convenient to study the homogeneous equation @xmath47 which is equivalent to eq . by an application of the involution @xmath48 .
( note that if in eq .
@xmath36 has an atom in @xmath49 , then this atom becomes an atom in @xmath50 in the equivalent equation .
this situation is not explicitly covered by the subsequent analysis but the results of this article remain true ( after some minor changes ) if an atom in @xmath50 is permitted . ) a first systematic approach to eq .
( [ minfp1 ] ) was given by jagers and rsler @xcite who pointed out the following connection of with its additive counterpart @xmath51 and the corresponding map @xmath52 , defined by @xmath53 which is usually called the _ smoothing transformation _
due to durrett and liggett @xcite .
namely , rewriting in terms of the laplace transform @xmath54 , say , of @xmath36 , we obtain @xmath55 for all @xmath28 , which is the direct analog of ( [ minfp2 ] ) .
but since any laplace transform vanishing at @xmath50 and thus pertaining to a distribution on @xmath56 can also be viewed as the survival function of a ( continuous ) probability distribution on @xmath16 , one has the implication @xmath57 where @xmath58 denotes the set of all positive solutions to . defining @xmath59 ,
one further has @xmath60 owing to the fact that @xmath61 .
however , jagers and rsler also give an example which shows that this implication can not be reversed .
we take up their example ( the water cascades example ) in section [ cascade ] in a generalized form .
eq . for the situation where the @xmath62 , are deterministic but not necessarily nonnegative is discussed in detail by alsmeyer and rsler @xcite .
their results concerning the case of nonnegative weights can be summarized as follows : except for simple cases nontrivial fixed - points exist iff @xmath5 possesses a _
characteristic exponent _ , defined as the unique positive number @xmath63 such that @xmath64 . in this case
the set of solutions @xmath22 can be described as follows : for @xmath65 and @xmath66 , let @xmath67 be the set of left continuous , multiplicatively @xmath68-periodic functions @xmath69 such that @xmath70 is nondecreasing .
the distribution @xmath15 on @xmath56 with survival function @xmath71 is then called @xmath68-periodic weibull distribution with parameters @xmath72 and @xmath73 , in short @xmath68-@xmath74 .
put @xmath75 for @xmath76 and @xmath77 , and let @xmath78 denote the set of ordinary weibull distributions with parameter @xmath73 , _
i.e. _ , the set of distributions @xmath15 having a survival function @xmath79 ( @xmath28 ) for some positive constant @xmath80 . then @xmath81 or @xmath82 , respectively , depending on whether the closed multiplicative subgroup @xmath83 generated by the positive @xmath84 , which we denote by @xmath85 , equals @xmath86 or @xmath87 for some @xmath66 . in view of this result ,
the result of alsmeyer and rsler extends classical results on extreme value distributions and the problem adressed in this paper , the analysis of eq .
is a further generalization of the analysis of extreme values , namely , of the distributional equation of homogeneous min - stability for in our situation the scaling factor is replaced by random coefficients .
one purpose of this paper is to investigate under which conditions similar results hold true in the situation of random weights @xmath84 , @xmath88 , _
i.e. _ , in which cases weibull distributions or suitable mixtures of them are solutions to .
this calls for extended definitions of @xmath85 and of the characteristic exponent : we define @xmath85 as the minimal closed multiplicative subgroup @xmath89 such that @xmath90 .
we further define @xmath91 $ ] by @xmath92 and then the characteristic exponent as the minimal positive @xmath93 such that @xmath94 if such an @xmath6 exists . with these generalizations ,
we obtain a connection between certain weibull mixtures and _ @xmath6-regular solutions _ , defined as solutions @xmath15 to such that the ratio @xmath95 stays bounded away from 0 and @xmath50 as @xmath96 approaches @xmath49 .
indeed , theorem [ alpha - elementar - darstellung ] will show that any @xmath6-regular fixed point is a mixture of weibull distributions with parameter @xmath6 and particularly _
@xmath6-elementary _ which means that @xmath95 converges to a positive constant as @xmath97 through a residue class relative to @xmath85 .
furthermore , theorem [ alpha - elementar - charakterisierung ] will provide an exact characterization of when @xmath6-elementary solutions to eq . exist .
both , the existence of weibull mixtures as fixed points and the existence of regular fixed points , are related to the existence of the characteristic exponent , which also plays a fundamental role in the analysis of eq . .
the further organization of this article is as follows .
section [ simple ] provides a discussion of trivial and simple cases of , which will be excluded thereafter .
an introduction of the weighted branching model closely related to our sfpe is given in section [ wbp ] , followed by the presentation and discussion of the main results in section [ mainresults ] .
section [ disintreq ] contains the derivation of a certain pathwise renewal equation related to via disintegration , while section [ charakteristischerexponent ] is devoted to a study of the characteristic exponent .
it contains most of the necessary prerequisites to prove theorem [ alpha - elementar - charakterisierung ] and theorem [ alpha - elementar - darstellung ] which is done in section [ fre ] . here
the afore - mentionded pathwise renewal equation will form a key ingredient .
as already mentioned , section [ cascade ] provides a discussion of eq . for a class of examples where the characteristic exponent does not generally exist .
finally , section [ conrem ] contains some results for eq . , which are closely related to ours and can be derived by the same methods .
theorem [ alpha - elementar - charakterisierung2 ] and theorem [ alpha - elementar - darstellung2 ] are extensions of theorem 2 and proposition 3 in @xcite .
this section is devoted to a brief discussion of simple cases and a justification of the following two basic assumptions on @xmath5 ( or , to be more precise , on the distribution of @xmath5 ) : put @xmath98 and consider @xmath99 @xmath100 by our standing assumption , @xmath101 .
hence , if fails to be true , then @xmath102 a.s .
and eq . reduces to @xmath103 , where @xmath5 is independent of @xmath36 and a.s .
but this sfpe can easily be solved , namely @xmath104 iff @xmath105 a.s . , see _
e.g. _ liu @xcite , lemma 1.1 .
validity of condition will therefore always be assumed hereafter . as a consequence , the branching process with offspring distribution @xmath106 ( of simple galton - watson type if @xmath107 a.s . )
survives with probability @xmath44 , a fact that will be used later .
the justification of assumption is slightly more involved and based upon the following two propositions : [ beschrsatz1 ] suppose
. then @xmath108 implies @xmath109 .
let @xmath15 be a solution to . then eq
. gives @xmath110 and thus @xmath111 for all @xmath28 .
let @xmath112 be a sequence of i.i.d .
copies of @xmath113 with associated multiplicative random walk @xmath114 , _
i.e. _ , @xmath115 and @xmath116 for @xmath117 .
then @xmath118 for each @xmath119 .
our assumptions on @xmath120 ensure @xmath121 a.s .
, whence @xmath122 for all @xmath123 , that is @xmath19 . before proceeding with our second proposition , let us note in passing that any @xmath124-measurable finite or infinite rearrangement @xmath125 of @xmath5 leaves the set of solutions to our sfpe unchanged because the @xmath126 are i.i.d . and independent of @xmath5 , thus also of @xmath127 .
so @xmath128 . as a consequence
it is no loss of generality to assume @xmath129 whenever the supremum is a.s.attained .
[ beschrsatz2 ] if holds the following assertions are equivalent : * @xmath130 a.s .
* @xmath131)=1\text { for some } c>0\}$ ] .
* @xmath132 for all @xmath133 .
* @xmath132 for some @xmath133 .
the implications `` ( b ) @xmath134 ( c ) @xmath134 ( d ) @xmath134 ( a ) '' being obvious , we must only prove `` ( a ) @xmath134 ( b ) '' .
define @xmath135 , if @xmath136 , and @xmath137 , if @xmath138 and w.l.o.g .
note for the latter situation that @xmath140 by .
the two cases @xmath141 ( case 1 ) and @xmath142 ( case 2 ) will now be discussed separately .
* pick any @xmath143 and let @xmath144 denote a sequence of i.i.d .
random variables with common distribution @xmath15 . as @xmath145 a.s .
the sfpe reads @xmath146 and clearly entails @xmath147 a.s .
now use the independence of @xmath148 and @xmath149 to infer that this can only hold if @xmath150 for some @xmath80 , w.l.o.g .
@xmath151 which is positive , for @xmath152 .
it remains to show that @xmath153 a.s .
assuming the contrary , we also have @xmath154 for some @xmath155 . since @xmath156 in the present case , the stopping time @xmath157 is finite with positive probability , and we have that @xmath158 because the @xmath126 and @xmath5 are independent . but this leads to the following contradiction : @xmath159 so we have proved @xmath160)=1 $ ] .
conversely , if we pick any @xmath161 with this property for some @xmath133 , then @xmath143 is immediate from because we have there @xmath162 a.s .
* case 2 . *
if @xmath142 we can not assume @xmath129 and then resort to the above argument because with positive probability the supremum may not be attained . on the other hand
, the claim reduces here to @xmath163 and it is easily verified that any @xmath164 is indeed a solution . for the reverse inclusion , pick any solution @xmath15 and suppose it is not concentrated at a single point , thus @xmath165 for some @xmath166 . for @xmath167 , define a r.v . @xmath168 as follows : @xmath169 where @xmath170 .
observe that in the second case @xmath171 and @xmath172 for any choice @xmath173 , so that @xmath174 by left continuity .
for any @xmath28 , we now infer @xmath175 and therefore @xmath176 for any @xmath96 such that @xmath177 . by left continuity , @xmath178 for any such @xmath96 and some @xmath179 .
however , @xmath180 then leads to the contradiction @xmath181 we hence conclude that @xmath15 must be concentrated in a single point .
[ compactsupport ] as a particular consequence of proposition [ beschrsatz2 ] , all solutions to have compact support if and @xmath130 a.s .
hold true .
as to a reverse conclusion , let us point out the following : if holds true and @xmath182 , then the assertions * @xmath130 a.s .
* there exists @xmath183 with compact support .
are equivalent . with only `` ( b ) @xmath134 ( a ) '' to be proved ,
let @xmath15 be an element of @xmath22 with compact support and @xmath36 a random variable with distribution @xmath15 , so @xmath184 .
suppose now there exists @xmath179 such that @xmath185 .
we can pick @xmath68 and @xmath186 in such a way that @xmath187 and thus @xmath188 .
but then @xmath189 which is a contradiction
. consequently , @xmath190 which in combination with @xmath191 and proposition [ beschrsatz1 ] proves ( a ) .
we close this section with a lemma that shows that any @xmath26 is continuous at 0 and that a search for solutions putting mass on @xmath16 is actually no restriction .
[ wlog ] suppose and let @xmath152 be any distribution on @xmath192 solving eq . .
then @xmath15 is continuous at @xmath49 and concentrated either on @xmath16 or @xmath193 $ ] . in the latter case , if @xmath17 and @xmath194 denotes the distribution of @xmath195 ( @xmath196 a.s . ) , then @xmath197 , where @xmath198 .
let @xmath12 be i.i.d . with distribution @xmath15 and independent of @xmath5 .
in view of what has been mentioned before proposition [ beschrsatz2 ] we may assume w.l.o.g . that @xmath199 iff @xmath200
. then @xmath201 whence @xmath202 must be a fixed point of the generating function of @xmath203 in @xmath39 $ ] . now use to infer @xmath204 .
next consider @xmath205 and suppose it to be @xmath206 .
then we infer with the help of @xmath207 for each @xmath117 and thereupon @xmath208 .
on the other hand , by another appeal to , @xmath209 for each @xmath210 with equality holding iff @xmath211 .
consequently , @xmath212 , which is clearly impossible as @xmath152 .
we thus conclude @xmath213 and thereby @xmath214 which proves the continuity of @xmath15 at 0 . as for the final assertion
, it suffices to note that @xmath215 is clearly equivalent to @xmath216 . unless stated otherwise
, we will always assume and hereafter .
as a consequence of , we infer that the closed multiplicative subgroup @xmath85 generated by @xmath5 can not be @xmath217 , the trivial subgroup .
so we have either @xmath218 for some @xmath66 ( @xmath68-geometric case ) or @xmath219 ( continuous case ) .
let @xmath220 be the infinite tree with vertex set @xmath221 , where @xmath222 contains only the empty tuple @xmath223 .
we abbreviate @xmath224 by @xmath225 and write @xmath226 for the vertex @xmath227 , where @xmath228 .
furthermore , @xmath229 and @xmath230 will serve as shorthand notation for @xmath231 and @xmath232 for some @xmath233 , respectively .
@xmath234 , @xmath235 and @xmath236 are defined similarly .
let @xmath237 denote a family of i.i.d .
copies of @xmath5 . for sake of brevity , suppose @xmath238 .
interpret @xmath239 as a weight attached to the edge @xmath240 in the infinite tree @xmath220 . then put @xmath241 and @xmath242 for @xmath243 .
so @xmath244 gives the total multiplicative weight along the unique path from @xmath223 to @xmath245 .
for @xmath246 , let @xmath247 denote the @xmath248-algebra generated by the sequences @xmath249 , @xmath230 , _
i.e. _ , @xmath250 put @xmath251 and @xmath252 . let us further introduce the following bracket operator @xmath253_u$ ] for any @xmath254 . given any function @xmath255 of the weight ensemble
@xmath237 pertaining to @xmath220 , define @xmath256_u\ \gl\ \psi((t(uv))_{v \in \operatorname{\mathbb{v}}})\ ] ] to be the very same function , but for the weight ensemble pertaining to the subtree rooted at @xmath257 .
any branch weight @xmath244 can be viewed as such a function , and we then obtain @xmath258_u = t_{v_1}(u ) \cdot \ldots \cdot t_{v_n}(uv_1 \ldots v_{n-1})$ ] if @xmath259 , and thus @xmath258_u = l(uv)/l(u)$ ] whenever @xmath260 . the weighted branching process ( wbp ) associated with @xmath237 is now defined as @xmath261 for any @xmath262 , we can replace the @xmath249 with @xmath263 which leads to the branch weights @xmath264 and the associated wbp @xmath265 note that @xmath266 , so that @xmath267 counts the positive branch weights in generation @xmath119 . if @xmath107 a.s .
, then @xmath268 forms a galton - watson process with offspring distribution @xmath269 , for @xmath270 .
suppose there exists an @xmath63 such that @xmath271 with @xmath272 as defined in .
then the sequence @xmath273 constitutes a nonnegative supermartingale with respect to @xmath274 and hence converges a.s . to @xmath275 . by fatou s lemma , @xmath276 which gives @xmath277 a.s . if @xmath278 . in the case @xmath279 , we have the dichotomy @xmath280 or @xmath281 ( _ cf .
_ theorem [ biggins ] in the appendix , or biggins @xcite , lyons @xcite and alsmeyer and iksanov @xcite for details ) .
henceforth , let @xmath282 and @xmath283 denote the distribution and laplace transform , respectively , of @xmath284 .
[ remcharexp ] as @xmath279 and @xmath285 ( or , equivalently , @xmath286 ) for some @xmath287 will be a frequent assumption hereafter , it is noteworthy that this forces @xmath6 to be the characteristic exponent of @xmath5 , that is , the _ minimal _ @xmath76 with @xmath288 . for a proof using theorem [ biggins ]
we refer to corollary [ charcharexp ] in the appendix . due to our standing assumption , theorem [ biggins ]
further implies that @xmath286 is actually equivalent to the a.s.positivity of @xmath284 .
as explained before proposition [ beschrsatz2 ] , the mapping @xmath34 defined in is invariant under @xmath124-measurable rearrangements of the @xmath84 , @xmath88 . it is therefore stipulated hereafter that @xmath289 if , and only if , @xmath290 . in order to provide the connection of the previously introduced weighted branching model with the sfpe ,
let @xmath291 be a family of independent copies of @xmath36 which is also independent of @xmath292 .
if @xmath17 , then @xmath119-fold iteration of yields @xmath293 for all @xmath294 which in terms of the survival function @xmath24 becomes @xmath295
we continue with the statement of the two main results that will be derived in this article .
theorem [ alpha - elementar - charakterisierung ] provides the connection between the existence of certain regular solutions to and the existence of the characteristic exponent of @xmath5 , while theorem [ alpha - elementar - darstellung ] is a representation result which states that any regular solution is a certain weibull mixture ( cf . definition [ wrnualpha ] below ) .
the definition of an @xmath6-regular fixed point is part of the following definition .
[ fixptype ] let @xmath287 and @xmath26 .
put @xmath296 for @xmath297
. then @xmath15 is called * _ @xmath6-bounded _ , if @xmath298 . *
_ @xmath6-regular _ , if @xmath299 .
* _ @xmath6-elementary _ , if * * in the case @xmath300 there exists a constant @xmath301 such that @xmath302 . *
* in the case @xmath218 for some @xmath66 for each @xmath303 there exists a positive constant @xmath304 such that @xmath305 for each @xmath306 .
the sets of @xmath6-bounded , @xmath6-regular and @xmath6-elementary fixed point are denoted by @xmath307 and @xmath308 , respectively .
\(a ) the notion of an
_ @xmath6-elementary _ fixed point has been introduced by iksanov @xcite in his study of the smoothing transformation @xmath309 given in and the associated sfpe .
his definition is the same as ours for the continuous case when replacing @xmath310 with @xmath311 where @xmath54 denotes the laplace transform of a solution to .
\(b ) given the existence of the characteristic exponent @xmath6 , guivarch @xcite and later liu @xcite called a ( nonnegative ) solution to with laplace transform @xmath54 _
canonical _ if it can be obtained as the stable transformation of a solution to the very same equation for the weight vector @xmath312 .
if the latter solution has laplace transform @xmath313 this means that @xmath314 for all @xmath315 .
this definition appears to be more restrictive than that of an @xmath6-elementary fixed point because the latter definition is valid for any @xmath287 . on the other hand ,
once shown that an @xmath6-elementary fixed point actually exists only if @xmath6 is the characteristic exponent of @xmath5 ( see theorem [ alpha - elementar - charakterisierung ] ) , `` @xmath6-elementary '' ( at least in the more restrictive sense of iksanov ) and `` canonical '' turn out to be just different names for the same objects ( see theorem 2 in @xcite and also theorem [ alpha - elementar - darstellung ] below ) .
\(c ) note for the @xmath68-geometric case that @xmath316}h(s)$ ] must be positive , for @xmath317 for all @xmath318 $ ] and @xmath319 .
after this observation , we see that any @xmath6-elementary fixed point is also @xmath6-regular , and since @xmath6-regularity trivially implies @xmath6-boundedness , we have that @xmath320 [ alpha - elementar - charakterisierung ] suppose and .
then the following assertions are equivalent for any @xmath287 : * eq . has an @xmath6-elementary solution @xmath321 .
* eq . has an @xmath6-regular solution @xmath322 .
* @xmath279 and @xmath286 . *
@xmath279 , the random walk @xmath323 with increment distribution @xmath324 , @xmath325 , converges to @xmath50 a.s . and @xmath326 \mathds{p } ( w_1^{(\alpha ) } \in \ , du ) < \infty.\ ] ] the proof of this theorem will be given in section [ fre ] .
we note that the equivalence statement of ( c ) and ( d ) is part of theorem [ biggins ] and only included here for completeness .
let us further point out the connection of our result with a similar one on @xmath6-elementary fixed points of the smoothing transformation obtained by iksanov @xcite .
by lemma a.3 in @xcite , each _ continuous _
@xmath6-elementary solution @xmath24 to is the laplace transform of a probability measure on @xmath16 solving .
therefore , under the continuity restriction , a part of our theorem could be deduced from theorem 2 in @xcite . on the other hand ,
the latter result strongly hinges on proposition 1 in the same reference the proof of which contains a serious flaw ( occuring in eq .
( 14 ) on p.36 where it is mistakenly assumed that @xmath327 does not depend on @xmath245 . without this assumption
the subsequent argument breaks down completely and there seems to be no obvious way to fix it under the stated assumptions ) . in order to state our second theorem ,
the following definition of certain classes of weibull mixtures is given , where the definitions of @xmath328 , @xmath329 and @xmath330 should be recalled from the introduction .
[ wrnualpha ] let @xmath63 and @xmath331 be a probability measure on @xmath86
. then * @xmath332 denotes the class of @xmath331-mixtures of @xmath329 distributions @xmath15 of the form @xmath333 where @xmath301 . *
@xmath334 for @xmath77 denotes the class of @xmath331-mixtures of @xmath328 distributions @xmath15 of the form @xmath335 where @xmath336 .
the reader should notice that @xmath332 is always a subclass of @xmath334 for any @xmath77 .
[ alpha - elementar - darstellung ] suppose and that @xmath279 and @xmath337 for some @xmath287 .
recall that @xmath338 . then @xmath339 where @xmath340 in the @xmath68-geometric case @xmath341 and @xmath342 in the continuous case @xmath343 . as to the proof of theorem [ alpha - elementar - darstellung ] ,
let us note that , once @xmath279 and @xmath337 have been verified , the inclusion @xmath344 follows upon direct inspection relying on the well known fact that @xmath345 , see lemma [ wrnualphainfinf ] .
so the nontrivial part is the reverse conclusion which will be shown in section [ fre ] .
the reader should further notice that does not need to be assumed in theorem [ alpha - elementar - darstellung ] because it already follows from and @xmath279 .
[ commentsthm](a ) the two previous theorems can be summarized as follows : the existence of at least one @xmath6-regular fixed point is equivalent to @xmath6 being the characteristic exponent of @xmath5 with @xmath337 , and in this case _ all _ regular solutions are in fact weibull mixtures with mixing distribution @xmath282 and particularly @xmath6-elementary . moreover , there are no further solutions in @xmath346 .
\(b ) let us briefly address two natural questions that arise in connection with our results .
first , do further nontrivial solutions to exist if @xmath279 and @xmath337 for some @xmath287 ? clearly , any further @xmath26 must satisfy either @xmath347 or @xmath348 lemma [ liminflimsup ] will show that only the second alternative ( @xmath310 oscillating at 0 ) might be possible .
however , whether solutions of that kind really exist in certain instances remains an open question .
second , one may wonder about the existence of solutions to if @xmath5 does not possess a characteristic exponent . although we can not provide a general answer to this question , it will emerge from our discussion in section [ cascade ] that there are situations in which there is no characteristic exponent and yet @xmath104 .
this was already observed by jagers and rsler @xcite , and the class of examples studied here forms a natural extension of theirs .
\(c ) there is yet another situation , called the _ boundary case _ by biggins and kyprianou @xcite , that we have deliberately excluded here from our analysis in order to not overburden the subsequent analysis .
it occurs when @xmath279 and @xmath349 contains an element @xmath350 , say , with infinite mean for some @xmath287 .
then @xmath351 a.s .
and @xmath352 provided that @xmath353 exists in a neighborhood of @xmath6 .
this case is quite different from the one in focus here , where @xmath354 , except that @xmath355 with @xmath356 as in theorem [ alpha - elementar - darstellung ] is easily verified by copying the proof of lemma [ wrnualphainfinf ] . if @xmath357 denotes the laplace transform of @xmath350 , then , under mild conditions ( cf .
theorem 5 of @xcite ) , @xmath358 behaves like a constant times @xmath359 as @xmath97 , and thus @xmath360 for any @xmath361 .
we quote this different behavior as opposed to that in the situation of the results above to argue that the boundary case requires separate treatment .
we refrain from going into further details and refer to a future publication .
our further analysis is based on a disintegration of eq .
by which we mean the derivation of a pathwise counterpart ( eq . below ) which reproduces upon integration on both sides .
we embark on the following known result on the sequence @xmath362 appearing under the expected value in .
[ disintegration ] let @xmath143 .
then @xmath363 forms a bounded nonnegative martingale with respect to @xmath364 and thus converges a.s . and in mean to a random variable @xmath365 satisfying @xmath366 the proof of this lemma can be found in biggins and kyprianou ( @xcite , theorem 3.1 ) . in the situation of lemma [ disintegration ] , we put @xmath367 and call the stochastic process @xmath368 a _ disintegration of @xmath15 _ and also a _ disintegrated fixed point_. the announced pathwise fixed - point equation for an arbitrary disintegrated fixed point is next .
[ disintsfpe ] let @xmath143 and @xmath369 a disintegration of @xmath15 .
then @xmath370_v(t l(v))\quad\text { a.s.}\ ] ] for each @xmath371 and @xmath372
. we have @xmath373_v ) \\ & \leq & \liminf_{k \to \infty } \prod_{v \in \{1,\ldots , m\}^n } [ \overline{\mathcal{f}}_k]_v(t l(v))\\ & = & \prod_{v \in \{1,\ldots , m\}^n } [ \overline{\mathcal{f}}]_v(t l(v ) ) \\ & \underset{m \to \infty}{\longrightarrow } & \prod_{|v|=n } [ \overline{\mathcal{f}}]_v(t l(v ) ) \qquad ( t \geq 0).\end{aligned}\ ] ] taking expectations on both sides , this inequality becomes an equality since the @xmath374_v(t l(v))$ ] , @xmath229 , are conditionally independent given @xmath375 and have conditional expectation @xmath376 by lemma [ disintegration ] .
this gives the asserted result .
is of essential importance for our purposes . it can be transformed into an additive one by taking logarithms and a change of the variables @xmath377 . to this
end , fix any @xmath287 and define @xmath378 for @xmath379 . put also @xmath380 for @xmath381 with the usual convention @xmath382 on @xmath383 .
then , by , @xmath384_v(e^t l(v ) ) \right ) \\ & = & \sum_{|v|=n } e^{-\alpha t } \left(-\log [ \overline{\mathcal{f}}]_v(e^t l(v ) ) \right ) \\
& = & \sum_{|v|=n } l(v)^{\alpha } e^{-\alpha ( t - s(v ) ) } \left(-\log [ \overline{\mathcal{f}}]_v(e^{t - s(v ) } ) \right ) \\ & = & \sum_{|v|=n } l(v)^{\alpha } [ \psi]_v(t - s(v))\quad\text { a.s.,}\end{aligned}\ ] ] that is @xmath385 satisfies the following _ pathwise renewal equation _ : @xmath386_v(t - s(v))\quad\text { a.s.}\ ] ] for each @xmath379 . to explain the notion `` pathwise renewal equation '' , we introduce a family of measures related to , namely @xmath387 if @xmath279
, then @xmath388 is a probability distribution on @xmath192 and @xmath389 ( the @xmath119-fold convolution of @xmath390 ) for each @xmath391 , see _ e.g. _ @xcite . in the following ,
we denote by @xmath392 a random walk with increment distribution @xmath390 if @xmath94 .
the renewal measure of @xmath392 shall be denoted by @xmath38 , _
i.e. _ , @xmath393 .
then @xmath394 where @xmath395 denotes the _ random weighted renewal measure _ of the branching random walk @xmath396 , for details see again @xcite .
the connection to the associated random walk has been used by various authors in the analysis of eq . or the branching random walk , see _ e.g. _
@xcite , @xcite , @xcite , or @xcite .
now suppose @xmath271 and define @xmath397 ( @xmath379 ) .
@xmath313 is well defined due to the fact that @xmath398 a.s . for all @xmath379 . by taking expectations on both sides of eq .
, we obtain @xmath399_v(t - s(v ) ) \\ & = & \mathds{e } \left ( \mathds{e } \left [ \sum_{|v|=n } l(v)^{\alpha } [ \psi]_v(t - s(v ) ) \big| \mathcal{a}_n \right ] \right ) \\ & = & \mathds{e } \sum_{|v|=n } l(v)^{\alpha } \psi(t - s(v ) ) \\ & = & \int \psi(t - s ) \ , \sigma_{\alpha , n}(ds),\end{aligned}\ ] ] having utilized that @xmath400_v$ ] is independent of @xmath247 for @xmath229 .
consequently , @xmath313 satisfies the renewal equation @xmath401 of which is a disintegrated version .
this provides the justification for the notion _ `` pathwise renewal equation''_. while uniqueness results for renewal equations of the form are commonly known , uniqueness results for processes solving a pathwise renewal equation are systematically studied in @xcite .
the following result is cited from there : [ pre ] suppose that @xmath402 , @xmath403 and @xmath279 for some @xmath287 . let @xmath404 $ ] denote a @xmath405-measurable stochastic process which solves eq . for @xmath406 . then the following assertions hold true : * suppose @xmath390 is nonarithmetic
. + if , at each @xmath407 , @xmath385 is a.s . left continuous with right hand limit and locally uniformly integrable and if @xmath408 , then @xmath385 is a version of @xmath409 for some @xmath410 . *
suppose @xmath390 is @xmath356-arithmetic ( @xmath411 ) .
+ if @xmath412 for all @xmath413 , then there exists a @xmath356-periodic function @xmath414 such that @xmath385 is a version of @xmath415 . in order to utilize this theorem in the context of fixed - point equations , we need to check whether the additive transformation @xmath385 of the disintegrated fixed - point @xmath369 satisfies the assumptions of the theorem .
clearly , @xmath385 is product measurable and standard arguments also show that it is a.s
. left continuous with right hand limit at any @xmath416 .
applicability of theorem [ pre ] therefore reduces to a verification of the integrability conditions for @xmath385 .
this is not always possible but works for the subclass of @xmath6-bounded fixed points and forms the key ingredient to our proof of theorem [ alpha - elementar - darstellung ] in section [ fre ] .
the purpose of this section is to provide some results related to the existence of the characteristic exponent @xmath6 of @xmath5 .
recall from remark [ remcharexp ] that a sufficient condition for this to be true is that @xmath279 and @xmath417 .
section [ cascade ] is devoted to a class of examples showing that solutions to our sfpe may exist even if @xmath5 does not have a characteristic exponent .
this will subsequently be used to point out some phenomena which may occur in the situation of eq . but not in the situation of its additive counterpart , _
i.e. _ , eq . .
[ transienz ] if @xmath271 for some @xmath287 , then @xmath418 since @xmath419 a.s . for each @xmath391
, there is nothing to show if @xmath420 a.s .
hence suppose that @xmath284 is nondegenerate . in this case ,
@xmath279 and the associated random walk @xmath392 with increment distribution @xmath390 converges a.s .
to @xmath50 by theorem [ biggins ] . a well known result from renewal theory ( cf . _
@xcite ) further tells us that the transience of @xmath392 implies @xmath421 for all compact intervals @xmath422 , where @xmath38 denotes the renewal measure of @xmath392 ( cf .
section [ wbp ] ) . since @xmath423 by eq .
, we infer @xmath424 a.s . for all compact intervals @xmath425 and therefore @xmath426 finally , use @xmath427 a.s . to conclude @xmath428 a.s .
[ fquer<1 ] if the characteristic exponent @xmath6 exists , then @xmath429 for all @xmath123 and @xmath183 .
suppose there exists a @xmath430 with @xmath431 .
pick any @xmath432 and let @xmath433 .
then @xmath434 is an a.s . finite stopping time by lemma [ transienz ] .
hence a combination of the optional sampling theorem applied to the bounded martingale @xmath435 and lemma [ disintegration ] yields @xmath436 since @xmath437 was chosen arbitrarily , we have @xmath438 for all @xmath371 , which is clearly impossible for any proper distribution on @xmath86 .
[ wrnualphainfinf ] suppose that @xmath94 and @xmath337 for some @xmath287 .
then @xmath439 in the @xmath68-geometric case ( @xmath66 ) and @xmath440 in the continuous case . in what follows
we restrict ourselves to the @xmath68-geometric case as the continuous case is similar but easier .
so let @xmath336 and then @xmath15 as in definition [ wrnualpha](a ) with @xmath441 .
recall that @xmath283 denotes the laplace transform of @xmath284 . then @xmath442 ( @xmath371 ) . since @xmath72 is multiplicatively @xmath68-periodic and
all positive @xmath84 take values in @xmath87 a.s . , we infer @xmath443 having used that @xmath444 , which in turn follows from the equation @xmath445_i$ ] a.s .
thus , @xmath24 solves eq . , _
i.e. _ , @xmath183 .
one can easily check that @xmath15 is always @xmath6-elementary .
[ otalphaleqwalpha ] let @xmath446 , @xmath447 be its disintegration and @xmath63 be such that @xmath271 .
then there exists a positive constant @xmath448 such that @xmath449 by assumption , @xmath450 for all sufficiently small @xmath371 and some @xmath448 . using this and @xmath451 for all @xmath452 in a suitable left neighbourhood of @xmath44
, we infer @xmath453 for all @xmath454 with @xmath455 sufficiently small . by lemma [ transienz ] in the following section
, @xmath456 ensures @xmath457 for @xmath458 .
fixing any @xmath123 , we have @xmath459 on @xmath460 for all @xmath235 and some sufficiently large @xmath119 .
hence @xmath461 on @xmath460 for all @xmath235 which in turn implies @xmath462 on the almost certain event @xmath460 .
[ otalphaleqwalpharem ] ( a ) lemma [ otalphaleqwalpha ] allows the following obvious modification : suppose @xmath271 for some @xmath287 and @xmath463 for some @xmath77 .
let @xmath143 be such that @xmath464 for some @xmath465 and all sufficiently large @xmath466 . then @xmath467 a.s . for all @xmath468 , where @xmath369 denotes the disintegration of @xmath15 .
\(b ) if @xmath351 a.s . , which is always true if @xmath469 and may be true if @xmath279 , then the assertion of lemma [ otalphaleqwalpha ] becomes @xmath470 a.s . for all @xmath315 which implies @xmath471 for all @xmath315 which is clearly impossible .
hence lemma [ otalphaleqwalpha ] is really about the case when @xmath279 and @xmath337 .
[ liminflimsup ] suppose that @xmath279 and @xmath337 for some @xmath287 .
then the following assertions hold true for any @xmath26 : * @xmath472 . * @xmath473 .
\(a ) suppose that @xmath474 . then @xmath15 clearly satisfies the crucial assumption of lemma [ otalphaleqwalpha ] for _ every _
@xmath448 , and we conclude for its disintegration @xmath369 that @xmath475 for every @xmath448 . as a consequence , @xmath476 a.s . which in turn implies @xmath477 .
but this contradicts lemma [ fquer<1 ] and we conclude that @xmath472 as claimed .
\(b ) suppose that @xmath478 .
this time we will produce a contradiction by comparison of @xmath479 with the class @xmath480 , @xmath166 , of solutions to ( see lemma [ wrnualphainfinf ] ) . since @xmath481 for any @xmath166 , we infer @xmath482 for any @xmath166 and @xmath483 with @xmath484 sufficiently small .
now fix any @xmath297 and consider the bounded martingales @xmath435 and @xmath485 defined by for @xmath15 and @xmath486 , respectively . by lemma [ transienz ] ,
the stopping time @xmath487 is a.s .
finite and @xmath488 for any @xmath166 .
therefore , by an appeal to the optional sampling theorem , @xmath489 for any @xmath166 . finally , use @xmath490 as @xmath491 to infer @xmath492 and thereupon the contradiction @xmath19 since @xmath297 was chosen arbitrarily .
[ iksanowtrick ] let @xmath287 and @xmath493
. then @xmath271 .
let @xmath494 and recall that @xmath495 for @xmath123 , thus @xmath496 and @xmath497 .
it follows from eq .
that ( see equation ( 10 ) in @xcite ) @xmath498 now use fatou s lemma to infer @xmath499 but the same argument applies to eq .
( that is after @xmath119 iterations of the sfpe ) and gives @xmath500 for each @xmath117 and thereupon the desired conclusion
. given two sequences @xmath501 , @xmath502 of real numbers , we write @xmath503 hereafter if @xmath504 and @xmath505 if @xmath506 . [ suffcharacexp ] let @xmath287 and @xmath493 . then @xmath6 is the characteristic exponent of @xmath5 and @xmath284 a.s .
positive .
note first that @xmath456 follows by lemma [ iksanowtrick ] and then @xmath428 a.s . by lemma [ transienz ] . since @xmath494 is not a dirac measure ( cf .
proposition [ beschrsatz2 ] ) , we have @xmath507 for some @xmath297 . by combining these facts with @xmath508 as @xmath509 and @xmath6-regularity
, we infer @xmath510 which in combination with @xmath511 a.s .
( lemma [ disintegration ] ) shows @xmath512 on the event @xmath513 .
hence @xmath279 , @xmath514 a.s . and @xmath515 ( by theorem [ biggins ] and ) .
`` ( a ) @xmath134 ( b ) '' is trivial as @xmath516 . ``
( b ) @xmath134 ( c ) '' is proposition [ suffcharacexp ] . ``
( c ) @xmath134 ( a ) '' . assuming ( c ) we have @xmath281 ( see theorem [ biggins ] in the appendix ) whence the laplace transform @xmath283 of @xmath284 satisfies @xmath517 . by lemma
[ wrnualphainfinf ] , @xmath518 , @xmath371 , defines a solution to eq . .
furthermore , @xmath519 and thus @xmath15 constitutes an @xmath6-elementary solution to eq . .
`` ( c ) @xmath520 ( d ) '' .
as already pointed out , this is a consequence of biggins martingale limit theorem stated as theorem [ biggins ] in the appendix.@xmath521 in view of lemma [ wrnualphainfinf ] it remains to verify that @xmath522 , where @xmath342 , if @xmath300 , and @xmath523 , if @xmath463 . to this end , let @xmath15 be @xmath6-bounded with disintegration @xmath524 . by lemma [ otalphaleqwalpha ] , we have @xmath525 for a suitable @xmath526 . recall that @xmath524 satisfies the multiplicative eq .
, which , upon logarithmic transformation and setting @xmath527 becomes the following pathwise renewal equation ( see ): @xmath528_v(t - s(v))\quad\text { a.s.}\ ] ] for all @xmath407 .
we want to make use of theorem [ pre ] and must therefore check its conditions as for the random function @xmath385 .
we already mentioned right after theorem [ pre ] that @xmath385 is product measurable and a.s . left - continuous with right hand limits at any @xmath416 .
since , by , @xmath529 for all @xmath379 on a set of probability one , it follows that @xmath408 and that @xmath385 is locally uniformly integrable . hence theorem [ pre ] applies and we infer @xmath530 a.s . , where @xmath531 denotes a measurable @xmath532-periodic function in the @xmath68-geometric case and a positive constant in the continuous case , respectively .
in both cases , @xmath533 for all @xmath123 , where @xmath534 . taking expectations on both sides of this equation provides us with @xmath535 therefore , the proof is complete in the continuous case where @xmath72 is necessarily constant . for
the rest of the proof , suppose we are in the @xmath68-geometric case .
then @xmath531 is @xmath532-periodic as mentioned above and thus @xmath72 multiplicatively @xmath68-periodic .
furthermore , the left continuity of @xmath24 implies the left continuity of the function @xmath536 and , therefore , the left continuity of @xmath72 .
similarly , we conclude that @xmath536 is nondecreasing .
these facts together give @xmath336 , which finally shows @xmath537.@xmath521 a combination of theorem [ alpha - elementar - darstellung ] and lemma [ liminflimsup ] provides us with a very short proof of the following result about the distribution @xmath282 of @xmath284 as a solution to with @xmath538 instead of @xmath5 : [ nualphaunique ] suppose and that @xmath94 and @xmath539 for some @xmath63 .
then @xmath540 , _
i.e. _ , @xmath282 is the unique solution up to scaling to for @xmath538 .
this result ( with @xmath541 instead of condition ) has been obtained under slightly stronger conditions on @xmath353 by biggins and kyprianou as theorem 1.5 in @xcite and as theorem 3 in @xcite , where the latter result also covers the boundary case briefly discussed in remark [ commentsthm](c ) .
suppose @xmath542 is another solution to for @xmath538 with laplace transform @xmath313 .
regard @xmath543 as the survival function of a probability measure @xmath194
. then @xmath544 and @xmath545 , the finiteness following from lemma [ liminflimsup ] .
consequently , @xmath546 and thus @xmath547 for some @xmath548 in the @xmath68 geometric case or a constant @xmath72 in the continuous case . in both cases , @xmath549 implies @xmath550 , _
i.e. _ , @xmath551 with @xmath552 .
this proves @xmath553 .
in the following we shall discuss a class of examples which demonstrates that nontrivial solutions to the sfpe may exist even if @xmath5 does not have a characteristic exponent .
this is in contrast to the additive case , _
i.e. _ , eq . , for which the existence of the characteristic exponent and the existence of nontrivial solutions are equivalent , at least under appropriate conditions on @xmath5 and @xmath203 , see durrett and liggett @xcite and liu @xcite .
we fix @xmath554 , @xmath555 , and denote by @xmath556 independent bernoulli variables with parameter @xmath557 , that is @xmath558 , @xmath559 . put @xmath560 for @xmath561 and @xmath562 for @xmath563 .
notice that @xmath564 ( @xmath565-geometric case ) .
. takes the form @xmath566 for @xmath567 and @xmath568 , this example was studied by jagers and rsler @xcite .
[ kritischlemma ] in the situation of eq .
the following assertions are equivalent : * the characteristic exponent @xmath63 exists .
* @xmath569 . *
if @xmath570 , then the galton - watson process with offspring distribution @xmath571 is subcritical . under stated assumptions , we have @xmath572 therefore @xmath573 is necessary and sufficient for the existence of a positive @xmath6 such that @xmath94 .
this proves that ( a ) and ( b ) are equivalent .
the equivalence of ( b ) and ( c ) is obvious as @xmath574 . according to lemma [ kritischlemma ]
, we distinguish three cases : * _ subcritical case _ : @xmath569 . * _ critical case _ : @xmath575 . * _ supercritical case _ : @xmath576 .
consider the associated wbp as introduced in section [ wbp ] .
now define @xmath577 for @xmath578 . by our model assumptions
, @xmath579 ( @xmath294 ) defines a galton - watson process with offspring distribution @xmath571 . in the supercritical case , @xmath580 survives with positive probability , thus @xmath581 the main outcome of the subsequent discussion will be that eq .
has nontrivial solutions in all three possible cases . in view of in the supercritical case
, this shows that the existence of a nontrivial solution to does not necessarily entail @xmath582 a.s . as intuition might suggest .
we proceed with the construction of nontrivial solutions to equation and start by taking a look at the associated functional equation . given a solution @xmath15
, the latter takes the form @xmath583 which , upon solving for @xmath584 , leads to @xmath585 where @xmath586 for @xmath587 .
the following lemma collects some properties of @xmath588 .
[ gnthetadiskussion ] the following assertions hold true for @xmath589 : * @xmath590:g _ { n,\vartheta}(u ) = u\}=\{0,1\}$ ] . * @xmath591 for all @xmath155 . * if @xmath592 ( subcritical or critical case ) , then @xmath589 is strictly increasing on @xmath593 .
in particular , @xmath594 for all @xmath155 . *
if @xmath595 ( supercritical case ) , there exists a unique @xmath596 satisfying @xmath597 .
@xmath588 is strictly increasing on @xmath598 $ ] and @xmath599 on @xmath600 .
obviously , @xmath601 holds iff @xmath602 and thus iff @xmath603 , for @xmath604 .
this shows ( a ) .
next , @xmath605 for all sufficiently small @xmath606 because @xmath588 is continuously differentiable on @xmath607 $ ] with @xmath608 .
but this gives ( b ) , by the continuity of @xmath588 and ( a ) , and we also infer that @xmath609 is positive in a right neighborhood of @xmath49 . furthermore , @xmath610 , @xmath611 $ ] , implies that @xmath612 for @xmath613 iff @xmath614 . hence @xmath615 on @xmath593 in the subcritical and critical case ( @xmath616 ) , and ( c ) is true . in the supercritical case ( @xmath617 ) , we have @xmath618 for a unique @xmath619 .
consequently , @xmath588 is strictly increasing on @xmath620 and strictly decreasing on @xmath621 .
since @xmath622 , ( d ) must be true ( _ cf .
_ also figure 1 ) . for @xmath567 and @xmath623 and the identity function ( from top to bottom),width=306 ] [ fig : ggraphik ]
* a. critical and subcritical case*. assuming @xmath592 , we have , by lemma [ gnthetadiskussion ] , that @xmath588 is strictly increasing with unique fixed points @xmath49 and @xmath44 in the unit interval .
therefore , its inverse function , denoted by @xmath624 , exists on @xmath39 $ ] . we can rewrite eq . in terms of @xmath624 as @xmath625 equations of this type have been completely solved in theorem 2.1 of @xcite , and its application allows us here to provide a full description of @xmath22 . to this end ,
let @xmath626 denote the @xmath627-fold composition of @xmath588 ( @xmath246 ) or its inverse @xmath624 ( @xmath628 ) , and let @xmath629 be the identity function .
although the situation in @xcite differs slightly from ours , we adopt the notation from there and write @xmath630 for the set of nonincreasing , left continuous functions @xmath631 \rightarrow ( 0,1)$ ] , satisfying @xmath632 for all @xmath633 $ ] .
then the following theorem is proved along the same lines as theorem 2.1 of @xcite with only minor changes in obvious places .
[ maximin ] if @xmath592 ( subcritical or critical case ) , there is a one - to - one correspondence @xmath634 between @xmath22 and @xmath630 , established by @xmath635 where @xmath468 denotes the unique integer satisfying @xmath636 . in the subcritical ,
where the characteristic exponent exists , we can state the following interesting corollary .
[ finf = wralpha ] suppose @xmath637 ( subcritical case ) and let @xmath6 be the characteristic exponent , _
i.e. _ , @xmath279 . then @xmath638 . in particular , any @xmath143 can be written as @xmath639 , @xmath123 for a unique @xmath640
. by proposition [ wrnualphainfinf ] , @xmath641
conversely , fix any @xmath143 and put @xmath642 ( @xmath643 ) , where @xmath644 denotes the inverse function of @xmath283 , the laplace transform of @xmath284 ( note that @xmath645 for all @xmath297 by lemma [ fquer<1 ] and remark [ compactsupport ] ) . extend @xmath72 to a multiplicatively @xmath565-periodic function on @xmath56 .
then it can easily be seen that @xmath640 .
thus , @xmath646 ( @xmath123 ) defines an element of @xmath647 and thus an element of @xmath22 as well .
moreover , @xmath15 and @xmath194 coincide on @xmath648 $ ] and , as a consequence of theorem [ maximin ] , @xmath15 and @xmath194 are uniquely determined by their values on @xmath648 $ ] .
hence @xmath649 .
* b. supercritical case*. assuming now @xmath650 , we follow an idea of rsler and jagers ( @xcite , 2004 ) and construct a particular nontrivial ( and discrete ) solution @xmath15 to eq . which is then shown to be unique up to scaling ( theorem [ superkritischsatz ] ) .
the first step is to define a sequence @xmath651 such that @xmath652 forms the range of @xmath24 .
let @xmath653 be the unique ( by lemma [ gnthetadiskussion](d ) ) value in @xmath593 such that @xmath654 .
lemma [ gnthetadiskussion ] also ensures that @xmath588 is strictly increasing on @xmath598 $ ] and @xmath605 for @xmath655 $ ] .
hence @xmath656\to [ 0,a_0]\subset [ 0,1]$ ] as well as its iterations are well defined , and we we can choose @xmath657 for @xmath246 .
evidently , @xmath651 constitutes a decreasing sequence of positive numbers and thus @xmath658 exists . since @xmath588 is continuous ,
we have @xmath659 so that @xmath660 by lemma [ gnthetadiskussion](a ) .
we now define our candidate @xmath15 as @xmath661 and note that @xmath15 is clearly left continuous and increasing with @xmath662 and therefore a distribution function . moreover , for @xmath372 and @xmath663 , @xmath664 where @xmath665 . as @xmath438 for @xmath666
, this shows that @xmath24 solves and so @xmath26 .
[ superkritischsatz ] if @xmath595 ( supercritical case ) , then @xmath667 where @xmath15 is given by .
we have already proved that @xmath143 .
also , @xmath22 is obviously closed under scaling , _
i.e. _ , @xmath668 .
conversely , let @xmath669 and notice that @xmath194 can not be concentrated in a single point by proposition [ beschrsatz2 ] .
hence we can choose @xmath123 such that @xmath670 .
now suppose @xmath671 .
then there exists a unique @xmath672 satisfying @xmath673 ( where @xmath674 ) . then from the definition of the @xmath675 we obtain @xmath676 which is obviously impossible .
thus , @xmath677 for all @xmath371 .
finally , put @xmath678 , thus @xmath679 , and use once again eq . and the recursive definition of the @xmath675 to obtain @xmath680 .
further details are omitted .
we have already pointed out in the introduction that equations and , and thus also the associated maps @xmath34 in and @xmath309 in , are naturally connected via the functional equation ( see ) @xmath681 valid for all survival functions @xmath682 of solutions @xmath15 to and for laplace transforms @xmath683 of solutions to .
the connection is even closer owing to the fact that any laplace transform vanishing at @xmath50 is also the survival function of a distribution on @xmath16 .
it is therefore not surprising that our results stated in section 4 have corresponding versions for the additive case dealing with the smoothing transform @xmath309 and its fixed points .
the latter has been studied in a large number of articles , see _ e.g. _
@xcite , @xcite , @xcite , @xcite , @xcite and @xcite .
in order to formulate the counterparts of theorem [ alpha - elementar - charakterisierung ] and theorem [ alpha - elementar - darstellung ] for @xmath309 , we first recall from @xcite the definition of @xmath6-stable laws and their @xmath68-periodic extensions which take here the role of the weibull distributions in the min - case .
[ periodicstable ] for @xmath63 and @xmath77 , let @xmath684 be the set of multiplicativley @xmath68-periodic functions @xmath685 such that @xmath686 has a completely monotone derivative . then , for @xmath687 , the _ @xmath68-periodic @xmath6-stable law _
@xmath68-@xmath688 is defined as the distribution on @xmath16 with laplace transform @xmath689 ( @xmath123 ) .
note that @xmath54 really defines a laplace transform by criterion 2 on page 441 of @xcite , for @xmath690 is positive with completely monotone derivative .
furthermore , if @xmath691 , then @xmath54 is the laplace transform of a positive stable law with scale parameter @xmath692 , shift @xmath49 and index of stability @xmath6 ( _ cf . _
@xcite , definition 1.1.6 and proposition 1.2.12 ) .
this distribution , denoted as @xmath693 , does not depend on @xmath68 ( which is thus dropped in the notation ) . for our convenience
, we also define @xmath68-@xmath694 .
we continue with a short proof of the known fact that periodic stable laws as defined above only exist for @xmath695 .
[ stable ] in the situation of definition [ periodicstable ] , any element of @xmath696 is constant , while @xmath697 for any @xmath698 . if @xmath699 , we have for each @xmath700 $ ] @xmath701 since @xmath702 is convex and @xmath703 nonincreasing if @xmath704 , we see that @xmath531 must be nonincreasing in @xmath705 .
in fact , @xmath531 is even strictly decreasing if @xmath706 .
but in view of the periodicity of @xmath531 , the latter is impossible , thus @xmath697 for @xmath707 , while @xmath531 must be constant in the case @xmath708 .
our next definition is just the canonical modification of definition [ wrnualpha ] for the previously defined generalized stable laws .
[ srnualpha ] let @xmath709 $ ] and @xmath331 be a probability measure on @xmath16
. then * @xmath710 denotes the class of @xmath331-mixtures of positive @xmath6-stable laws @xmath15 of the form @xmath711 where @xmath301 . *
@xmath712 for @xmath77 denotes the class of @xmath331-mixtures of @xmath68-@xmath688 distributions @xmath15 of the form @xmath713 where @xmath687 .
finally , the notions `` @xmath6-bounded '' , `` @xmath6-regular '' and `` @xmath6-elementary '' for fixed points of @xmath309 are defined exactly as in definition [ fixptype ] for fixed points of @xmath34 , when substituting @xmath310 with @xmath714 , @xmath54 the laplace transform of @xmath15 .
the respective classes are denoted as @xmath715 and @xmath716 , respectively . we are now ready to formulate the results corresponding to our theorems in section [ mainresults ] for the smoothing transform
. the standing assumptions and for the min - case can be replaced here with the weaker ones @xmath717 @xmath718 [ alpha - elementar - charakterisierung2 ] suppose and .
then the following assertions are equivalent for any @xmath287 : * equation has an @xmath6-elementary solution @xmath719 .
* equation has an @xmath6-regular solution @xmath720 . *
@xmath709 $ ] , @xmath279 and @xmath286 . *
@xmath709 $ ] , @xmath279 , the random walk @xmath721 converges to @xmath50 a.s . and
@xmath326 \mathds{p } ( w_1^{(\alpha ) } \in \ , du ) < \infty.\ ] ] [ alpha - elementar - darstellung2 ] suppose , and that @xmath279 and @xmath337 for some @xmath709 $ ] .
then @xmath722 where @xmath340 in the @xmath68-geometric case @xmath341 and @xmath342 in the continuous case @xmath343 .
the proofs of the previous two theorems are essentially the same as for theorems [ alpha - elementar - charakterisierung ] and [ alpha - elementar - darstellung ] except for the additional assertion @xmath723 .
but the arguments in the proof of theorem [ alpha - elementar - charakterisierung ] show that the laplace transform @xmath54 of any element of @xmath724 can be written as @xmath725 ( @xmath123 ) , where @xmath531 is a multiplicatively @xmath68-periodic function in the @xmath68-geometric case and a constant in the continuous case .
owing to lemma [ stable ] we get @xmath723 if we can prove that @xmath687 in the @xmath68-geometric case . by writing @xmath726 ( @xmath123 ) ( where @xmath727 denotes the inverse function of @xmath54 ) ,
we see that @xmath531 is infinitely often differentiable .
it remains to verify that @xmath728 has a completely monotone derivative . to this end
, we observe that @xmath729 for @xmath730 having utilized that @xmath731 .
since @xmath732 has a completely monotone derivative for each @xmath730 and since the convergence is uniform on compact sets , @xmath690 has a completely monotone derivative as well .
the assumptions and are in force throughout .
let @xmath733 denote the extinction probability of @xmath734 ( an ordinary supercritical galton - watson process if @xmath735 a.s . ) .
the subsequent characterization theorem for martingale limits in branching random walks ( which are nothing but limits of wbp s having the martingale property ) is a crucial ingredient to our analysis of @xmath6-elementary fixed points . in the stated most general form , which imposes no conditions on @xmath5 beyond @xmath279 ,
it was recently obtained by alsmeyer and iksanov @xcite , but the first version of the result under additional assumptions on @xmath5 was obtained more than three decades ago by biggins @xcite and later reproved ( under slightly relaxed conditions ) by lyons @xcite via a measure - change argument ( size - biasing ) .
alsmeyer and iksanov combined lyons argument with goldie and maller s results on perpetuities @xcite .
we further state the following corollary which says that @xmath279 and @xmath286 always imply that @xmath6 equals the characteristic exponent of @xmath5 , that is , the _ minimal _ value at which @xmath272 equals 1 .
this is relevant to be pointed out because @xmath272 , as a strictly convex function on @xmath739 , may equal 1 for two values @xmath740 .
first note that @xmath743 a.s
. follows from lemma [ transienz ] , which is easily seen to remain valid under and .
now , if @xmath279 and @xmath744 for @xmath745 , we see that @xmath746 for any @xmath747 $ ] which is clearly impossible if @xmath748 , where @xmath749 would be a nonnegative supermartingale . 999 ( 2005 )
a survey of max - type recursive distributional equations .
probab . _ * 15 * , 1047 - 1110 .
the objective method : probabilistic combinatorial optimization and local weak convergence . in _ probability on discrete structures _ ( encyclopaedia of mathematical sciences * 110 * ) , ed .
kesten , 1 - 72 .
springer , new york .
a log - type moment result for perpetuities and its application to martingales in the supercritical branching random walk . _
electronic j. probab . _
* 14 * , 289 - 313 .
( 2007 ) . a stochastic fixed - point equation related to game - tree evaluation . _
* 44 * , 586 - 606 .
( 2006 ) . a stochastic fixed - point equation related to weighted branching with determinstic weights . _
electronic j. probab . _ * 11 * , 27 - 56 .
( 2008 ) . a stochastic fixed - point equation related to weighted minima and maxima .
h. poincar _ * 44 * , 89 - 103 .
martingale convergence in the branching random walk .
_ j. appl .
_ * 14 * , 25 - 37 .
seneta - heyde norming in the branching random walk .
. probab . _ * 25 * , 337 - 360 ( 2005 ) .
fixed points of the smoothing transform : the boundary case . _ electronic j. probab . _ * 10 * , 609 - 631 .
representation of fixed points of a smoothing transformation . _ mathematics and computer science iii _ , eds . m. drmota et al .
, trends in math . ,
birkhuser , basel , 311 - 324 .
( 2001 ) . on the probabilistic worst - case time of `` find '' .
_ algorithmica _ * 31 * , 291 - 303 .
. fixed points of the smoothing transformation . _ z. wahrsch .
gebiete _ * 64 * , 275 - 301 .
( 1971 ) . _
an introduction to probability theory and its applications _ , vol .
ii , 2nd edition , wiley , new york .
stability of perpetuities .
_ * 28 * ( 2000 ) , 1195 - 1218 .
sur une extension de la notion semi - stable .
h. poincar _ * 26 * , 261 - 285 . ( 2004 ) .
elementary fixed points of the brw smoothing transforms with infinite number of summands .
appl . _ * 114 * , 27 - 50 .
( 2004 ) stochastic fixed points for the maximum . _
mathematics and computer science iii _ , 325 - 338 .
trends math .
birkhuser , basel .
fixed points of a generalized smoothing transformation and applications to the branching random walk .
probab . _ * 30 * , 85 - 112 .
( 1997 ) . a simple path to biggins martingale convergence for branching random walk . in _ classical and modern branching processes
( i m a vol .
appl . * 84 * ) , eds .
k.b . athreya and p. jagers , 217 - 221 .
springer , new york .
( 2009 ) . weighted branching and a pathwise renewal equation . to appear in _ stoch .
analysis of algorithms by the contraction method : additive and max - recursive sequences .
( 2006 ) . on stochastic recursive equations of sum and max type .
_ j. appl .
_ * 43 * , 687 - 703 . (
_ stable non - gaussian random processes : stochastic models with infinite variance_. chapman & hall , new york . | this paper is devoted to the study of the stochastic fixed - point equation @xmath0 and the connection with its additive counterpart @xmath1 associated with the smoothing transformation . here
@xmath2 means equality in distribution , @xmath3 is a given sequence of nonnegative random variables and @xmath4 is a sequence of nonnegative i.i.d . random variables independent of @xmath5 .
we draw attention to the question of the existence of nontrivial solutions and , in particular , of special solutions named @xmath6-regular solutions @xmath7 .
we give a complete answer to the question of when @xmath6-regular solutions exist and prove that they are always mixtures of weibull distributions or certain periodic variants .
we also give a complete characterization of all fixed points of this kind . a disintegration method which leads to the study of certain multiplicative martingales and a pathwise renewal equation after a suitable transform
are the key tools for our analysis .
finally , we provide corresponding results for the fixed points of the related additive equation mentioned above . to some extent ,
these results have been obtained earlier by iksanov @xcite . _
keywords : _ branching random walk ; elementary fixed points ; multiplicative martingales ; smoothing transformation ; stochastic fixed - point equation ; 2000 mathematics subject classification : 60e05 + 39b22 |
radio galaxies exhibiting jets which terminate in radio lobes on tens of kpc to mpc scales are classified based on their radio morphology and power by @xcite .
they are divided into fanaroff - riley ( fr ) type i and type ii , where type i sources have the highest surface brightness feature at the center , while in type ii sources it is farther from the core .
furthermore , the transition radio luminosity between fri and frii increases with the optical luminosity of the host galaxy @xcite . in the agn unification scheme
, blazars are thought to be radio galaxies with the jet aligned along our line of sight , and are subdivided into flat spectrum radio quasars ( fsrqs ) and bl lacertae objects based on the strength of emission lines in their spectrum , where fsrqs generally have strong emission lines , while bl lacs have weak or none @xcite .
fri galaxies are thought to correspond to misaligned bl lacs , while friis correspond to misaligned fsrqs ( e.g. , * ? ? ?
* and references therein ) , although there is evidence that this unification scheme is too simple ( e.g. , * ? ? ?
apparent superluminal motion observed on milli - arcsecond size scales indicates that their jets must be moving at high relativistic speeds , with bulk lorentz factor @xmath3 1020 for fsrqs and bl lacs @xcite , although some tev bl lacs have @xmath4 @xcite .
the existence of high energy and very high energy ( vhe ) @xmath0-rays observed from these sources provides further evidence for highly relativistic flows , as they are necessary to avoid @xmath0-ray attenuation by electron - positron pair production @xcite .
indeed , this sometimes gives values of @xmath5 greater than that found from very - long baseline interferometry ( vlbi ) superluminal observations ; e.g. , @xmath6 is required for a recent outburst from pks 2155 - 304 ( e.g. , * ? ? ?
* ; * ? ? ?
since blazars are strong sources of beamed @xmath0-rays , it is natural to think that radio galaxies may be also .
several radio galaxies were detected by egret : @xcite , @xcite , and centaurus ( cen ) a @xcite .
the identifications were rather uncertain , due to the large egret error circles .
only two radio galaxies have been detected so far with the latest generation of tev atmospheric cherenkov telescopes , @xcite and cen a @xcite .
the radio galaxy 3c 66b seems to have been seen by magic @xcite , although the detection is questionable due to its proximity to the bl lac 3c 66a and its lack of detection by veritas @xcite .
fermi_-lat collaboration has reported the detections of ( per a ; * ? ? ?
* ) , m87 @xcite , and cen a @xcite .
several more @xmath0-ray detections of radio galaxies have been reported in the first _
fermi_-lat catalog ( 1fgl ; * ? ? ?
* ; * ? ? ?
* ) and a future publication will examine them in more detail ( fermi collaboration 2010 , in preparation ) .
the _ fermi _ gamma ray space telescope was launched on 2008 june 11 and contains the large area telescope ( lat ) , a pair conversion telescope which has a field of view of about 20% of the sky at 20 mev to over 300 gev @xcite .
for the first year of operation , _ fermi _ was operated in a sky - survey observing mode , wherein the lat sees every point on the sky every @xmath7 hours . during the first 3-months of science operation , the _ fermi_-lat confirmed @xcite the egret detection of cen a. here with additional monitoring , we present accumulated data after 10 months of operation .
the new lat observations bridge the gap between egret and hess , providing a detailed look at the @xmath8-ray spectrum essential for addressing emission models .
in addition to the lat @xmath0-ray source in the central few kpc ( hereafter the @xmath0-ray `` core '' ) , @xmath0-rays from the giant lobes of have also been seen with _ fermi _
, with the origin likely to be compton scattering of the cosmic microwave background ( cmb ) and extragalactic background light ( ebl ) , confirming the predictions of @xcite and @xcite
. detailed work on separating the core and lobe emission is presented elsewhere ( * ? ? ?
* hereafter referred to as the lobe paper ) , although we provide a summary of lat observations below . for the purposes of this paper , which is a study of @xmath0-ray emission of the core , the lobes are essentially background sources
we present a summary of cen a and observations of this object in
[ cena - section ] .
the observations of the core of cen a with the lat over the first 10 months of _ fermi _ operation are presented in [ latobs ] .
we also present simultaneous cen a core observations from _ suzaku _ and _ swift _ , and radio data from the tanami program in [ obs ] . in
[ model - section ] we combine these with archival data and model its sed of the cen a core . in [ discussion ]
we discuss the implications in detail , and we conclude with a brief summary ( [ summary ] ) .
the fri is the nearest radio lound active galaxy to earth , making it an excellent source for studying the physics of relativistic outflows and radio lobes . indeed
, it is near enough that its peculiar velocity dominates over the hubble flow , and its redshift ( @xmath9 ) can not be used to accurately calculate its distance . @xcite
have found that the average of several distance indicators gives @xmath10 , which we adopt . at this distance
, an arcsecond corresponds to about 18 pc .
due to its proximity to earth , it has been well studied throughout the electromagnetic spectrum , from radio to @xmath0-rays .
recently , the auger collaboration reported that the arrival directions of the highest energy cosmic rays ( @xmath11 ev ) observed by the auger observatory are correlated with nearby agn , including cen a @xcite , while @xcite found that , if the giant lobes are taken into account , as many as four ultra - high energy cosmic rays ( uhecrs ) may be associated with this source .
although the overall significance of this correlation is reduced in the expanded auger data set , the significance remains high in the direction of cen a @xcite .
this suggests that and other radio galaxies may be sources of uhecrs .
has interesting radio structure on several size scales .
the most prominent features are its giant radio lobes , which subtend @xmath12 on the sky , oriented primarily in the north - south direction .
they have been imaged at 4.8 ghz by the parkes telescope @xcite and studied at up to @xmath1360 ghz by @xcite utilizing wilkinson microwave anisotropy probe ( wmap ; * ? ? ?
* ) observations .
the north lobe contains a bright region a few tens of arcminutes in size often referred to as the northern middle lobe @xcite .
mis - aligned by approximately @xmath14 relative to the outer lobes are inner radio lobes on an arcminute scale @xcite . a strong
, well - collimated jet can be seen on the arcsecond size scale in the radio , and _
chandra _ can resolve x - ray emission from it , which is likely caused by synchrotron emission @xcite .
the innermost region of cen a has been resolved with vlbi , and shown to have a size of @xmath15 cm @xcite .
observations at shorter wavelengths also reveal a small core , namely vlt infrared interferometry which resolves the core size to @xmath16 cm @xcite .
vlbi images reveal a weak counter jet on the milli - arcsecond scale @xcite .
based on the motion of the vlbi blobs , and assuming the brightness differences of the different jets are due to doppler effects , @xcite estimate the angle of the sub - parsec jet to our line of sight to be @xmath17 . applying a similar technique to the 100 pc scale jet which has a larger jet - counterjet ratio , @xcite
estimate a jet angle of @xmath18 .
@xcite speculate that the conflicting angle estimates may be due to the assumption that the jet
counter jet brightness differences are caused by doppler beaming rather than intrinsic differences .
, the giant elliptical host galaxy of , contains a kiloparsec - scale dust lane .
this feature appears to be an edge - on disk obscuring the central region and nucleus , and is probably the remnant of a previous merger @xcite .
it also has a dusty torus within 100 pc of the black hole , with a high column density ( @xmath19 ) @xcite .
x - ray spectra taken at various times over decade timescales indicate a time - varying absorbing column density , which could be due to variations in a warped disk viewed edge - on @xcite .
estimates for the mass of the supermassive black hole at the center of range from @xmath20 @xcite based on the kinematics of stars , as well as h@xmath21 and ionized gas . with the @xmath22 gamma - ray observatory ,
emission was detected by osse @xcite and comptel @xcite at 100s of kev to mev energies .
@xcite suggested the hard x - ray emission from detected with osse was the result of compton - scattered disk radiation by a thermal plasma ( i.e. , a hot corona ) , due to a turnover in the spectrum at a few hundred kev .
however , @xcite noted that the high - energy portion of the osse spectra smoothly connected with the higher energy comptel spectra , and the osse and comptel variability seem to be correlated .
they used this to argue for a nonthermal jet origin for the x - rays . @xcite
have resolved the arcsecond - scale core of cen a with _ chandra _ and _ xmm - newton_. the 27 kev x - ray continuum , when corrected for absorption , is consistent with what is predicted from a correlation between unresolved x - ray emission and 5 ghz core emission for jets of radio galaxies @xcite .
they thus consider it likely that nonthermal emission from the sub - pc ( sub - mas ) scale jet is the origin of the continuum x - rays from the core of cen a. however , hard x - rays observed by _
suzaku _ do not seem to fit on the @xcite correlation , possibly indicating a non - jet origin @xcite .
the nature of the continuum x - ray emission from the core of cen a remains an open question .
cen a has been a target of @xmath8-ray observations dating back to the 1970s ( e.g. , * ? ? ?
* ; * ? ? ?
cen a was seen by egret up to gev energies @xcite .
the @xmath0-rays are thought to originate from a relativistic jet near the central elliptical galaxy ( the radio `` core '' ) analogous to blazars , although it has been suggested that compton - scattering of the cmb and the infrared - optical ebl in the giant radio lobes could be a source of @xmath0-rays from @xcite and other radio galaxies such as @xcite . at the highest , tev energies ,
a detection was recently reported from cen a by the air cherenkov detector hess @xcite .
the egret detection of cen a @xcite was confirmed early on by the _ fermi_-lat . based on 3-months of all - sky survey data ,
the initial lat detection was reported in the lat bright source list ( bsl ) paper @xcite as 0fgl j1325.44303 with a 95@xmath23 confidence localization , 95=0.304= 18.3 . in the companion
lat bright agn sample paper ( lbas ; * ? ? ?
* ) to the bsl , a single power - law fit was reported , which gave @xmath24(@xmath25100 mev ) = 2.15 ( @xmath26 0.45 ) with photon index , @xmath27 = 2.91 @xmath26 0.18 , and a peak flux on a @xmath131 week timescale of ( 3.23 @xmath26 0.80 ) .
note that this only considered the @xmath0-ray emission from cen a as a single point source , i.e. , it did not account for any lobe emission .
to these initial observations , 7 additional months of all - sky survey data are added to the current analysis .
specifically , the observations span the time period from 2008 august 4 to 2009 may 31 , corresponding to met ( mission elapsed time ) 239557420 265507200 .
diffuse event class ( ctbclasslevel=3 ) events were selected with a zenith angle cut of @xmath28105 , and a rocking angle cut of 39 .
the former are well calibrated and have minimal background while the latter greatly reduce earth albedo @xmath8-rays . for the analysis ,
lat science tools version v9r11 was utilized with the p6_v3_diffuse instrument response function ( irf ) .
the standard lat galactic emission model , gll_iem_v02.fit was used and the uniform background was represented by the isotropic diffuse @xmath0-ray background and the instrumental residual background ( isotropic_iem_v02.txt , * ? ? ?
we consider 11 point sources in the 1fgl catalog ( * ? ? ?
* see also figure [ figure - roi ] ) .
figure [ figure - roi ] shows the the 0.2@xmath2930 gev lat image centered on cen a , which is clearly detected .
also prominent is the galactic emission toward the south , and several faint sources in the field .
we obtained a localization of the source at cen a with gtfindsrc , which finds point source locations based on an unbinned likelihood analysis .
the resulting localization was reduced to 95= 0.087 = 5.2 ( 5.7 kpc ) , centered at ra = 201.399 , dec = @xmath30 ( j2000.0 epoch ) which is offset by 0.029 = 1.7 ( 1.9 kpc ) from the vlbi radio position of @xcite .
figure [ rgb ] shows the localization error circle of the lat emission overlaid on the combined radio , optical , x - ray images .
the new lat position is consistent with that of 3eg j1324 - 4314 @xcite , but both are notably offset from egr j1328 - 4337 , the closest egret source in the @xcite catalog .
the latter derived position shifted in such a way that cen a was outside of the 95 localization circle , so that there was some ambiguity as to whether egret was actually detecting cen a , but the new lat position confirms the earlier 3eg result .
the lat significantly improves upon the previous egret @xmath8-ray localization ( 95=0.53= 32 ) .
the binned likelihood fitting was performed with the gtlike tool , first assuming cen a is a point source , i.e. , that there is no @xmath0-ray lobe emission ( model a ) .
the field point source positions were fixed , and their spectra were assumed to be power - laws , with the photon indices allowed to vary .
the location of cen a was fixed at its vlbi radio position @xcite .
in addition to the 11 1fgl point sources used in the lobe paper , in order to treat the lobe emission as a background source , we include two 1fgl sources , 1fgl j@xmath31 and 1fgl j@xmath32 , which are thought to be the local maxima of the lobe emission .
a likelihood analysis with the energy information binned logarithmically in 20 bins in the 0.230 gev band , and the @xmath0-ray directions binned into a @xmath33 grid with a bin size of @xmath34 .
for both the galactic and isotropic emission models , one free parameter was introduced to adjust the normalization .
because the effective area of the lat is rapidly changing below @xmath35 , we use events with energy above this value . above 30 gev the significance of detection is @xmath36 , so we make a cut as this energy as well . as a result ,
the test statistic ( ts ; * ? ? ?
* ) is found to be 378 for cen a , which is smaller than the ts=628 in the 1fgl catalog @xcite , since the lower energy limit is 200 mev in our analysis , instead of 100 mev in the catalog .
the relative normalizations of the galactic and isotropic models become [email protected] and [email protected] , respectively , and the fit is reasonable within the current background model uncertainty .
this fit gives a power - law photon index of cen a between 200 mev and 30 gev of @xmath27=@xmath37 and the flux extrapolated down to @xmath25100 mev is ( [email protected] ) ( where errors are statistical only ) .
as noted in @xcite , the spectrum is very steep in comparison to the typical blazars of @xmath38 the power - law photon index is consistent with the 3eg result of @xmath39 @xcite .
the 3eg flux was reported to be ( @xmath40 ) , and have a peak value of ( @xmath41 ) @xcite , consistent with with the average flux .
we next modeled the region with a radio image of the giant lobe ( model b ) .
this analysis is identical to that described in the lobe paper , and the reader is referred to it for details .
we present a brief description below .
we use the wmap image at 20 ghz from @xcite , and eliminate the cen a core region with a cut radius of @xmath42 . in this analysis , we exclude two point sources ( 1fgl j@xmath31 and 1fgl j@xmath32 ) , which are assumed to be emission from the lobes .
the binned likelihood analysis was performed to extract the flux and spectral indices for the core and lobes .
the relative normalizations of the galactic and isotropic models become [email protected] and [email protected] , respectively .
the @xmath0-ray detection in each energy range is significant at a 4@xmath43 level up to the 5.610 gev energy bin for the core region and the spectrum is consistent with the power - law model .
this fit gives a photon index of the core between 200 mev and 30 gev of @xmath27=@xmath44 and a flux extrapolated down to @xmath25100 mev of ( 1.50@xmath45 ) , with statistical and systematic errors reported .
here , we consider the systematic errors from the effective area , the diffuse model , and wmap inner cut radius , as described in the lobe paper . the photon index is almost identical to that of model a , but the flux is somewhat lower due to some of the core photons from model a being considered as being emitted by the lobes in model b. the results for model b can be seen in figure [ graysed ] .
to quantify variability within the @xmath1310 month lat observation , we generated light curves in 30 and 15 day bins using the unbinned likelihood analysis with gtlike .
we performed the analysis taking into account the lobe emission ( i.e. , model b in [ latspectralsection ] ) .
the power - law normalizations of the core and background point sources are treated as free parameters , but the photon indices of all sources and the normalizations of the lobes and the diffuse background models are fixed to the values obtained in 200 mev
30.0 gev for the whole time region .
figure [ figure - lightcurve]a shows the light curve of the flux ( extrapolated down to @xmath46 mev ) in 30 day bins .
the @xmath47 test results in @xmath47/d.o.f .
= 0.98 , and the light curve with 15 day bins gives @xmath47/d.o.f .
these are consistent with no variability .
the time behavior of cen a is in contrast to large variability of typical blazars in the mev / gev range , and similar to that of perseus a @xcite and m87 @xcite .
observations with several different instruments , both on the earth and in space , were made during the 10 months of lat observations presented here .
cen a was observed in the radio as part of the tracking active galactic nuclei with austral milliarcsecond interferometry ( tanami ) program ( mueller et al .
2009 ; ojha et al .
data were taken with two instruments on the _ swift _ spacecraft @xcite and two instruments on the _ suzaku _ spacecraft @xcite .
a summary of these observations can be found in table [ tab - obs ] , and descriptions are given below .
cen a was observed with vlbi on 2009
november 27/29 , as part of the tanami program using the five antennas of the australian long baseline array ( lba ) , the 70 m dss-43 antenna at nasa s deep space network at tidbinbilla , australia , and two trans - oceanic telescopes tigo ( chile ) and ohiggins ( antarctica ) of the international vlbi service ( ivs ) for geodesy and astrometry ( the latter two participating at 8.4ghz , only ) .
the beam size achieved was ( 0.92 mas @xmath48 0.56 mas ) at 8.4ghz and ( 1.68 mas @xmath48 1.25 mas ) at 22.3ghz using natural weighting . these observations were part of the tanami monitoring of a radio and @xmath8-ray selected sample of 65 blazars at 8.4ghz and 22.3ghz with observations approximately every two months .
tanami data are correlated on the difx software correlator @xcite at curtin university in perth , western australia .
data inspection and fringe fitting was done with aips ( national radio astronomy observatory s astronomical image processing system software ) .
the images were produced by applying the program difmap @xcite , using the clean algorithm .
more details about the data reduction can be found in @xcite .
data from the first epoch ( november 2009 ) of tanami observations are presented in @xcite .
[ figure - sed ] includes the fluxs at 22.3ghz and 8.4ghz measured in 2009
november 27/29 , respectively . the total flux density
, corresponding to the emission distributed over the inner @xmath13 120mas at 8.4ghz , is @xmath49jy . at 22.3ghz ,
a total vlbi flux density of 3.2jy is distributed over the inner @xmath50mas of the jet , with very little emission on the counterjet side . via model fitting , we found a component with an inverted spectrum , which is the brightest at both frequencies and which we identify with the jet core .
the core flux density is 0.92jy at 8.4ghz and 1.54jy at 22.3ghz .
the core size is consistently modeled at both frequencies to be ( 0.91.0)mas @xmath48 ( 0.290.31)mas at the same position angle of 5355 degrees ( see ojha et al .
2009 ) .
cen a was observed with _ suzaku _ on 2009
july 2021 , aug 56 , and aug 1416 with a total exposure of 150 ks , during which time the flux approximately doubled .
we utilized data processed with version 2.4 of the pipeline _ suzaku _ software , and performed the standard data reduction : a pointing difference of @xmath51 , an elevation angle of @xmath52 from the earth rim , a geomagnetic cut - off rigidity ( cor ) of @xmath256 gv .
we did not use events from the time the spacecraft entered the south atlantic anomaly ( saa ) to @xmath53 s after it left the saa .
further selection was applied : earth elevation angle of @xmath54 for the x - ray imaging spectrometer ( xis ) , cor@xmath258 gv and the time elapsed from the saa ( t_saa_hxd ) of @xmath25500 s for the hard x - ray detector ( hxd ) .
the xis response matrices are created with xisrmfgen and xissimarfgen @xcite .
the hxd responses used here are ae_hxd_pinhxnome5_20070914.rsp for the pin and ae_hxd_gsohxnom_20060321.rsp and ae_hxd_gsohxnom_20070424.arf for the gadolinium silicate ( gso ) crystal .
the `` tuned '' ( lcfit ) hxd background files @xcite are utilized .
the detailed _ suzaku _ analysis , including gso data and time variability , will be reported elsewhere ( y. fukazawa et al .
2010 , in preparation ) .
the _ suzaku _
data were fit with a single absorbed power - law , which was found to have a spectral index @xmath55 with dust absorbing column density @xmath56 . the flux in the 12 76 kev band on 2009 july was @xmath57 , about twice the flux measured by _ suzaku _ in 2005 @xcite .
was observed on six days between 2009 jan .
15 28 for a total exposure of 22 ksec ( see table [ tab - obs ] ) .
the xrt @xcite data were processed with the xrtdas software package ( v. 2.5.1 ) developed at the asi science data center ( asdc ) and distributed by the nasa high energy astrophysics archive research center ( heasarc ) within the heasoft package ( v. 6.6 ) .
event files were calibrated and cleaned with standard filtering criteria with the _ xrtpipeline _ task using the latest calibration files available in the caldb .
the xrt dataset was taken entirely in windowed timing mode .
for the spectral analysis we selected events in the energy range 210 kev with grades 02 .
the source events were extracted within a box of 40x40 pixels ( @xmath1394 arcsec ) , centered on the source position and merged to obtain the average spectrum of cen a during the xrt campaign .
the background was estimated by selecting events in a region free of sources .
ancillary response files were generated with the _ xrtmkarf _ task applying corrections for the psf losses and ccd defects .
the combined january x - ray spectrum is highly absorbed .
hence it was fitted with an absorbed power - law model with a photon spectral index of @xmath58 , an intrinsic absorption column of @xmath59 , in excess of the galactic value of @xmath60 in that direction @xcite .
the average absorbed flux over the @xmath61 energy range is @xmath62 , which corresponds to an unabsorbed flux of @xmath63 .
the xrt spectrum included in the broadband sed was binned to ensure a minimum of 2500 counts per bin and was de - absorbed by forcing the absorption column density to zero in xspec , and applying a correction factor to the original spectrum equal to the ratio of the de - absorbed spectral model over the absorbed model .
we used data from the burst alert telescope ( bat ) on board the swift mission to derive a 14195kev spectrum of cen - a contemporary to the lat observations .
the spectrum has been extracted following the recipes presented in @xcite .
this spectrum is constructed by calculating weighted averages of the source spectra extracted over short exposures ( e.g. 300s ) .
these spectra are accurate to the mcrab level and the reader is referred to @xcite for more details .
the lat spectrum of the core of cen a is shown in fig.[graysed ] , extrapolated into the tev regime , along with the hess spectrum observed between 2004 and 2008 @xcite .
also shown is the hess spectrum scaled down by its source flux normalization uncertainty .
it seems that the lat spectrum , with its statistical and systematic errors , extrapolated to higher energies , is just barely consistent with the hess spectrum .
however , one should keep in mind that the hess and lat spectra presented in this figure are not simultaneous , although the hess data did not show any signs of variability .
additionally , @xmath64 absorption makes it unlikely that the hess and lat emission originate from the same region , which is explored below ( [ sscmodel ] ) . since
the cores of many blazars have been shown to be @xmath0-ray loud it is plausible to assume that the radio core is the source of the central @xmath0-rays from cen a. however , one should keep in mind that the error circles of the _ fermi _ and hess @xcite observations are consistent with emission from the inner lobes , jet and radio core , so that these other regions could be sources of @xmath0-rays as well .
we construct the sed for the resolved sub - arcsec and arcsec - scale core as compiled in @xcite , including their mm / ir / optical observations from 20032005 .
they have compiled additional points from the 1990s and have applied an extinction correction of @xmath65 mag to the optical and ir data .
we plot historical data in the x - ray @xcite , hard x - rays @xcite , comptel @xcite , and the hess tev @xmath0-rays @xcite .
the _ swift _ xrt and bat , as well as _ suzaku _ data , corrected for galactic dust as well as dust in ngc 5128 , discussed in
[ obs ] , were collected during time intervals which overlap with much of the _ fermi_-lat data .
furthermore , we add the simultaneous radio data of the tanami vlbi jet components .
all these are shown in fig .
[ figure - sed ] .
the lat data points in fig .
[ figure - sed ] are from model b and include statistical errors only .
single - zone synchrotron / synchrotron self - compton ( ssc ) models have been very successful in explaining the multiwavelength ( including @xmath0-ray ) emission from bl lac objects ( e.g. , * ? ? ?
* ; * ? ? ?
if fris are the misaligned counterpart to bl lacs , one would expect this model to apply to them as well . in this scenario the low energy , radio through optical emission originates from nonthermal synchrotron radiation from a relativistically moving spherical homogeneous plasma blob , and the x - ray through vhe @xmath0-rays from the compton scattering of that synchrotron radiation by electrons in the same blob .
the one - zone ssc model has successfully fit the emission from the other _
detected fris , ( ngc 1275 ; * ? ? ?
* ) and @xcite , and has been successfully applied to previous observations of @xcite . here
we apply the single - zone ssc model to fit the recent multiwavelength observations of , particularly the _ fermi_-lat and hess emission .
one can show ( see appendix [ analytic_appendix ] ) that , on the assumption that all of the emission in the multiwavelength sed of the cen a core originates from the same region in a single zone ssc model , @xmath64 absorption gives the constraint on the doppler factor @xmath66 where the doppler factor is @xmath67^{-1}$ ] , the bulk lorentz factor of the jet is @xmath68 , @xmath69 is the speed of the jet , and @xmath70 is the angle of the jet with respect to our line of sight . solving for @xmath5 in terms of @xmath71 , @xmath72 in order for @xmath5 to be real , the quantity under the radical must be positive , which implies @xmath73 ( e.g. , * ? ? ?
, estimates of @xmath74 vary from @xmath75 to @xmath76 ( see section [ cena - section ] ) .
for the least constraining value , @xmath77 , @xmath78 clearly , the constraints ( [ cenaconstrain1 ] ) and ( [ cenaconstrain2 ] ) are not compatible . thus , if the radio through _ fermi _ @xmath8-ray data presented in fig .
[ figure - sed ] are synchrotron and ssc emission originating from the same region of the jet , then _ the hess emission can not originate from the same part of the jet .
_ note also that the hess emission can not originate from the same region of the jet , yet be emitted from a different mechanism than ssc ( say , compton scattered accretion disk or dust torus radiation ) because even this radiation would be subject to the same @xmath79 attenuation by synchrotron photons .
if the vlbi jet core is assumed to be the origin of the high - energy emission , the tanami core - size measurement can be used to calculate an upper limit on the size of the @xmath0-ray emitting region of @xmath80pc = @xmath81 cm ( 3.1 ) .
this is consistent with the vlbi observations of @xcite and @xcite , and with a variability timescale of @xmath82 day , given that the emitting region radius @xmath83 is constrained by the variability time by @xmath84 .
this variability timescale is consistent with the _ suzaku _ observations , although it is not clear that the _ suzaku _ x - rays come from the same region as the @xmath0-rays .
using this variability timescale and eqns ( [ sedconstrain ] ) and ( [ bconstrain ] ) , one gets @xmath85 and @xmath86 g. more precise modeling @xcite gives the green curve in fig .
[ figure - sed ] with the model parameters in table [ modelparams ] .
this curve demonstrates the emission can be fit with a doppler factor of unity .
this is consistent with a lorentz factor of unity or 7 , a degeneracy which can be seen in eqn.([gammaquad ] ) .
a stationary , nonrelativistic jet can explain the entire sed , except the vhe emission .
this fit is similar to the synchrotron / ssc fit by @xcite who fit similar data .
we further note that a small change in @xmath71 leads to a large change in the lorentz factor .
this , combined with the uncertainty in the inclination angle , leads to the fact that the lorentz factor is not well - constrained by modeling .
we also note that vlbi observations show _ apparent _ motion with @xmath87 @xcite , implying @xmath88 , which is also not a particularly strong constraint .
what if the hard x - ray emission originates from thermal comptonization near the disk , and not from jet emission ?
if we assume the rest of the high - energy sed is from the jet , then @xmath89 and @xmath90 erg s@xmath91 @xmath92 , so that eqns ( [ sedconstrain ] ) and ( [ bconstrain ] ) give @xmath93 and @xmath94 g for a variability timescale of 1 day .
more detailed modeling gives the violet curve seen in fig .
[ figure - sed ] with the parameters in table [ modelparams ] .
the larger doppler factor needed for this model requires a smaller angle to the line of sight .
the lorentz factor is again not strongly constrained , and could plausibly be as high as @xmath95 and still provide a good fit , although this would push the parameters to their extremes .
this model still under - predicts the hess data .
jet powers for these models are given in table [ modelparams ] .
the proton and pair content of the jet are not well known , so the total jet power presented in table [ modelparams ] is for a pure pair jet , and can be considerd a lower limit .
even with 10100 times more energy in ions than leptons , the absolute jet power is far below the eddington luminosity for a @xmath96 black hole ( @xmath97 ) . for the green curve ,
the parameters assume @xmath98 .
the jet power needed to inflate the giant lobes of in their lifetime , as inferred from the radio spectral break , is @xmath99 erg s@xmath91 @xcite .
this value is approximately consistent with the the green curve model presented in fig .
[ figure - sed ] .
a possible explanation for the hess observations is that the tev emission is produced by another blob .
we show in fig.[figure - sed ] ( brown curve ) that another synchrotron / ssc - emitting blob can produce the hess emission without over - producing any of the other multiwavelength data .
the parameters for this blob are in table [ modelparams ] , although this fit is not unique and many parameter sets would fit the hess data and not contribute at other wavelengths . other possible origins for the vhe emission are discussed in [ vheorigin ] .
unification models for blazars suggest that frii galaxies are fsrqs with the jet viewed away from our line of sight , and similarly fris are the parent population of bl lacs . in this case
, one would expect non - thermal emission from the cores of radio galaxies , de - beamed compared to blazars .
however , the cores of fris seem brighter than what is expected from simply de - beamed emission from bl lacs , which implies the radio galaxy core emission is from a slower region than that of bl lacs , since the beaming angle is related to the bulk lorentz factor by @xmath100 . there are ( at least ) two possible explanations for this : ( 1 ) the jet consists of a faster `` spine '' , which is responsible for the on - axis blazar emission , inside a slower outer `` sheath '' , which would be responsible for the off - axis emission seen in the cores of radio galaxies ( e.g. , * ? ? ? * ) ; and ( 2 ) a decelerating jet model where the on - axis blazar emission is produced by a faster flow closer to the black hole , and the off - axis @xmath0-rays seen in radio galaxies are produced by the slower flow farther out along the jet @xcite . as an example
, we provide a fit to the cen a sed using this decelerating flow , as the blue curve in fig .
[ figure - sed ] . in this model ,
the high energy emission is due to upstream compton scattering of synchrotron photons produced in the slower part of the flow being scattered by energetic electrons in the faster , upstream part of the flow .
the jet starts with a bulk lorentz factor @xmath101 and decelerates down to @xmath102 in a length of @xmath103 cm .
the injected power law electron distribution , @xmath104 has an index @xmath105 , and extends from @xmath106 , to @xmath107 , and the magnetic field at the inlet is @xmath108 g. jet powers for this model are similar to the one - zone ssc model fits presented in [ sscmodel ] , although this decelerating model fit is particle- rather than magnetic field - dominated .
we also note that the parameters used in this fit are not unique .
since the single blob model does not seem to be able to reproduce the broadband sed of , could something else be the origin of the vhe @xmath8-rays ?
we have already shown that another blob emitting synchrotron and ssc radiation could explain the hess emission without over - producing any of the other data ( fig .
[ figure - sed ] brown curve ) . @xcite
have presented a model with multiple blobs , moving at different angles to the line of sight from a large opening angle , to m87 and cen a ( among other objects ) .
this model does seem to be able to explain this sed @xcite .
it has also been suggested that absorbed @xmath8-rays which create @xmath109 pairs , creating an isotropic halo of electrons in the ism which compton - scatter the host galaxy s starlight , leading to isotropically - produced @xmath0-rays @xcite .
the hess data do seem to match the @xcite predictions cen a with a galactic magnetic field of 10 @xmath110 g .
compton - scattering off of leptons accelerated by the supermassive black hole magnetosphere , similar to particle acceleration in pulsars , has been proposed to explain the vhe @xmath8-ray radiation from m87 @xcite .
this could also explain the hess data from separate from the other multiwavelength emission .
as we have noted earlier , what we designate in this paper as the @xmath0-ray `` core '' actually encompasses the radio core , jet , and inner lobes of cen a. this is also true for the hess emission . @xcite
have noted that a shock front observed in x - rays in the southwest inner lobe could be a source of tev @xmath0-rays , which seems consistent with these observations .
finally , we note that the sed presented here is constructed from non - simultaneous data .
although _ fermi _ and hess @xmath8-rays do not show appreciable variability , they could still be variable on longer timescales .
perhaps for a good , simultaneous multiwavelength sed , a one - zone synchrotron / ssc model could provide a good fit to all of the data . probably the best way to discriminate between the above models
simple ssc , compton - scattering emission from a pair halo , multiple blobs , etc.is correlated variability between lat @xmath8-rays and other bandpasses .
this emphasizes the importance of simultaneous multiwavelength data .
the auger observatory results indicate some uhecrs could be originating from ( see [ cena - section ] )
. the uhecrs could interact with photons at the source and in the extragalactic background light leading to an observable signature in the hess band .
if the vhe @xmath8-rays originate from cosmic rays this could account for the discrepancy between hess and _ fermi _ @xmath8-rays .
based on the green curve fit presented in fig.[figure - sed ] we can analyze whether it is plausible for cosmic rays to originate from , keeping in mind that the parameters of that model are not well constrained ( 3 ) .
the maximum energy to which cosmic rays can be accelerated is limited by the size scale of the emitting region and the highest energy they can reach before they are cooled .
the former constraint implies that the highest energy a cosmic ray can reach is @xmath111 and the latter implies @xmath112 ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , where @xmath113 is the acceleration efficiency factor , and @xmath114 is the elementary charge , @xmath115 is the atomic number , and @xmath116 the atomic mass of the ion .
note that these timescales , and all quantities expressed above , are in the frame comoving with the blob , although for the particular model considered here , @xmath71=1 so this is not important .
we assume all parameters have values from the green curve model .
thus , it seems for this model that it is unlikely that protons will be accelerated to energies above @xmath117 ev , although it is possible for heavier ions to be accelerated this high before they are disintegrated by interacting with infrared photons from the core .
the threshold energy for photomeson interaction with peak synchrotron photons is similar to @xmath118 .
this process could create observational signatures from secondary emission ( e.g. , * ? ? ? * ) , as well as convert protons to neutrons , which can escape as cosmic rays @xcite .
again , we note that this result is strongly model - dependent , and the parameters of this model are not strongly constrained , so this limit should not be taken too seriously .
for example , a small change in the doppler factor would have little effect on the model fit , but would require a large change in the bulk lorentz factor , @xmath5 .
a large change in @xmath5 would significantly affect the highest energy to which particles could be accelerated , as seen in eqns ( [ esat1 ] ) and ( [ esat2 ] ) .
furthermore , if we are viewing a slower sheath , uhe cosmic rays could be accelerated in the faster spine beamed away from our line of sight , which could have significantly different parameters .
acceleration of protons up to @xmath119 requires jet powers of @xmath120 , which may take place in occasional flaring activities in @xcite .
we have reported on observations of with the lat instrument on board the _ fermi _ gamma - ray space telescope .
this instrument s excellent angular resolution compared to other @xmath0-ray detectors at mev
gev energies makes it possible for the first time to separate the lobe and core emission .
the lat observations have been supplemented with simultaneous observations from _ suzaku _ , _ swift _ , the australia telescope long baseline array , and a variety of non - simultaneous data , including those from hess .
our results are as follows : 1 .
the lat - detected core position is consistent with cen a s vlbi core @xcite and previous egret observations @xcite .
2 . with 10 months of lat exposure
, we find the core flux @xmath46 mev to be ( 1.50@xmath45 ) and the spectral index in the @xmath121 gev range to be @xmath27=@xmath44 , consistent with the egret @xcite and the previously - reported 3-month lat detection @xcite .
3 . extrapolated to higher energies ,
the lat spectrum is barely consistent with the hess spectrum @xcite only if the hess spectrum is lowered in flux by its normalization error .
4 . a single zone ssc model can explain all of the multiwavelength emission from the core except for the non - simultaneous hess emission .
it is not possible to fit the entire sed , including the hess emission , with a single zone compton - scattering model due to internal @xmath79 absorption effects .
modeling results are consistent with suggestions by chiaberge et al .
that we are seeing @xmath0-rays from a different origin that we would if were were looking down the jet .
this could be explained by a spine in sheath @xcite or decelerating jet scenario @xcite . the _ fermi _ lat collaboration acknowledges generous ongoing support from a number of agencies and institutes that have supported both the development and the operation of the lat as well as scientific data analysis .
these include the national aeronautics and space administration and the department of energy in the united states , the commissariat lenergie atomique and the centre national de la recherche scientifique / institut national de physique nuclaire et de physique des particules in france , the agenzia spaziale italiana and the istituto nazionale di fisica nucleare in italy , the ministry of education , culture , sports , science and technology ( mext ) , high energy accelerator research organization ( kek ) and japan aerospace exploration agency ( jaxa ) in japan , and the k. a. wallenberg foundation , the swedish research council and the swedish national space board in sweden . additional support for science analysis during the operations phase is gratefully acknowledged from the istituto nazionale di astrofisica in italy and the centre national dtudes spatiales in france .
we are grateful to the anonymous referee for useful comments which have improved the manuscript .
we thank n. odegard for providing us with the wmap image and the suzaku team for their calibration and satellite operation .
we also acknowledge the swift team and the swift / xrt monitoring program efforts , as well as swift analysis supported by nasa grants nnx08av77 g and nnx09au07 g .
in the ssc model , the doppler factor , @xmath71 , and comoving , tangled , isotropic magnetic field strength , @xmath122 , may be estimated from the dimensionless peak energy and @xmath123 flux , @xmath124 and @xmath125 of the synchrotron and ssc components , respectively , observed in the sed . assuming the comoving blob size can be constrained by @xmath126 , this gives @xmath127 @xmath128 @xcite where @xmath129 is the variability timescale and @xmath130 is the distance to the source .
the doppler factor , @xmath131^{-1}$ ] where the bulk lorentz factor is @xmath132 , @xmath133 is the speed of the jet , and @xmath70 is the angle of the jet with respect to our line of sight . in order for @xmath8-rays to escape an emission region , the @xmath134 absorption optical depth , @xmath135 , can not be too large . assuming the @xmath136 synchrotron flux , @xmath137 , is given by a broken power law , then for @xmath138 for a photon with dimensionless energy @xmath139 , this implies @xmath140^{\frac{1}{6 - 2a}}\end{aligned}\ ] ] @xcite where @xmath141 and @xmath116 is the index of the synchrotron spectrum below the break for @xmath142 and above the break for @xmath143 solving eqn .
( [ sedconstrain ] ) for @xmath129 and inserting this into eqn .
( [ ggconstrain ] ) , one gets the constraint @xmath144^{1/4}\ ; .\end{aligned}\ ] ] for , @xmath145 mpc @xmath146 cm , and @xmath147 .
the spectral parameters can be obtained from the sed of the core of ( see fig .
[ figure - sed ] ) : @xmath148 , @xmath149 , @xmath150 erg s@xmath91 @xmath92 , and @xmath151 erg s@xmath91 @xmath92 .
note that here we assume that the x - ray data is from the jet ; see above .
below the break in the synchrotron spectrum , @xmath152 , and above @xmath153 .
the highest energy photon bin in the hess spectrum is @xmath154 , so that @xmath155 erg s@xmath91@xmath92 .
these values give the constraint @xmath156 which is equation ( [ cenaconstrain1 ] ) .
lccc australian lba and ivs & 2009 nov .
27 & 3.6 ks & 22.3 ghz + & 2009 nov .
29 & 3.6 ks & 8.4 ghz + _ suzaku _ xis & 2009 jul .
16 & 150 ks & 0.4 10 kev + _ suzaku _ hxd - pin & 2009 july 20 aug .
16 & 150 ks & 10 70 kev + _ swift _ xrt & 2009 jan .
15 28 & 22 ks & 0.2 10 kev + _ swift _ bat & 2008 aug .
2009 may & 1.9 ms & 14 200 kev + _ fermi _ lat & 2008 aug .
4 2009 may 31 & 10 months & 0.230 gev + [ tab - obs ] lccccc bulk lorentz factor & @xmath157 & 7.0 & @xmath158 & 3.7 & 2.0 + doppler factor & @xmath71 & 1.0 & @xmath159 & 3.9 & 3.1 + jet angle & @xmath74 & @xmath160 & @xmath161 & @xmath75 & @xmath75 + magnetic field [ g ] & @xmath122 & 6.2 & 0.45 & 0.2 & 0.02 + variability timescale [ sec ] & @xmath129 & @xmath162 & & @xmath163 & @xmath163 + comoving blob size scale [ cm ] & @xmath83 & @xmath164 & @xmath165 & @xmath166 & @xmath167 + low - energy electron spectral index & @xmath168 & 1.8 & 3.2 & 1.8 & 1.8 + high - energy electron spectral index & @xmath169 & 4.3 & & 4.0 & 3.5 + minimum electron lorentz factor & @xmath170 & @xmath171 & @xmath172 & @xmath173 & @xmath173 + maximum electron lorentz factor & @xmath174 & @xmath175 & @xmath176 & @xmath175 & @xmath175 + break electron lorentz factor & @xmath177 & @xmath178 & & @xmath179 & @xmath180 + jet power in magnetic field [ @xmath181 & @xmath182 & @xmath183 & @xmath184 & @xmath185 & @xmath186 + jet power in electrons [ @xmath181 & @xmath187 & @xmath188 & @xmath189 & @xmath190 & @xmath191 + [ modelparams ] | we present @xmath0-ray observations with the large area telescope on board the _ fermi _ gamma - ray telescope of the nearby radio galaxy centaurus a. the previous egret detection is confirmed , and the localization is improved using data from the first 10 months of _ fermi _ science operation . in previous work , we presented the detection of the lobes by the lat ; in this work , we concentrate on the @xmath0-ray core of cen a. flux levels as seen by the lat are not significantly different from that found by egret , nor is the extremely soft lat spectrum ( @xmath1 where the photon flux is @xmath2 ) .
the lat core spectrum , extrapolated to higher energies , is marginally consistent with the non - simultaneous hess spectrum of the source .
the lat observations are complemented by simultaneous observations from _ suzaku _ , the _ swift _ burst alert telescope and x - ray telescope , and radio observations with the tracking active galactic nuclei with austral milliarcsecond interferometry ( tanami ) program , along with a variety of non - simultaneous archival data from a variety of instruments and wavelengths to produce a spectral energy distribution ( sed ) .
we fit this broadband data set with a single - zone synchrotron / synchrotron self - compton model , which describes the radio through gev emission well , but fails to account for the non - simultaneous higher energy tev emission observed by hess from 2004 - 2008 .
the fit requires a low doppler factor , in contrast to bl lacs which generally require larger values to fit their broadband seds .
this indicates the @xmath0-ray emission originates from a slower region than that from bl lacs , consistent with previous modeling results from cen a. this slower region could be a slower moving layer around a fast spine , or a slower region farther out from the black hole in a decelerating flow .
the fit parameters are also consistent with cen a being able to accelerate ultra - high energy cosmic - rays , as hinted at by results from the auger observatory . |
[ cols="^,^,^,^,^,^,^,^,^,^,^ " , ] ( au ) , and inclination , @xmath6 ( ) , for the cn 54 j=9/27/2 and j=11/29/2 lines ( top and bottom panel , respectively ) .
the peak separation versus the line flux is indicated by filled points and solid lines connecting the model predictions for the same @xmath7 .
the fwhm versus the line flux is overplotted with empty circles and dotted lines.,width=332 ] ( @xmath8 ) , the dotted black and blue curves the gas temperature and total hydrogen number density @xmath9.,width=332 ] in order to test if the cn lines originate in the disk of t tau n we use a parametrized disk model calculated with the thermo - chemical disk modeling code prodimo @xcite .
we adopt stellar and disk parameters as inferred from previous studies , which well reproduce the source spectral energy distribution ( sed ) ( e.g. , * ? ? ?
we use stellar spectral type k0 ( @xmath10 k ) , stellar luminosity @xmath11 @xmath12 and stellar mass @xmath13 @xmath14 as determined by @xcite .
the uv spectrum @xcite is reproduced by an uv excess @xmath15 and a power law slope @xmath16 .
the effect of x - ray radiation from the stellar corona of t tau n ( @xmath17 ergs@xmath1 , @xcite ) is taken into account following @xcite and @xcite .
following @xcite we assume a disk inner radius @xmath18 au , a dust mixture of astronomical silicates @xcite and amorphous carbon @xcite with relative abundances of 62.5% and 37.5% , and a grain size distribution @xmath19 with @xmath20 , where @xmath21 is the number of dust particles with radius @xmath22 , and the minimum / maximum grain size are @xmath23 @xmath24 m and @xmath25 mm .
this implies a dust mass of @xmath26 @xmath14 to reproduce the observed mm emission @xcite . using the average ism dust - to - gas ratio of @xmath27 ,
the gas mass is 0.013 @xmath14 .
the dust - to - gas ratio can be different from the ism value in evolved disks ( e.g. , * ? ? ?
* ; * ? ? ?
however , as the cn lines are optically thick in the disk , their flux is independent on the gas mass , hence on the assumed dust - to - gas ratio .
the sed is well reproduced by assuming a disk surface density @xmath28 and scale height @xmath29 @xcite . to obtain a dust distribution similar to the two - layers model adopted by @xcite
, we assume that grains larger than @xmath30 @xmath24 m have settled to a smaller scale height than the gas , @xmath31 , with @xmath32 . as pah emission is generally not detected in ttss ( e.g. , * ? ? ?
* ) , the pah fraction , @xmath33 , is set to @xmath27 with respect to the ism abundance of @xmath34 pah particles / h - nucleus .
lower values , @xmath35 , do not affect the cn emission .
the adopted stellar and disk parameters are summarized in table [ tab : model_par ] .
as the cn line fluxes and profiles are very sensitive to the assumed disk outer radius , @xmath7 , and inclination , @xmath6 , we proceed through two steps .
first , we run a grid of models at low resolution ( 50@xmath3650 grid points ) adopting @xmath7 and @xmath6 values which cover the range of estimates from previous studies ( @xmath37 au , @xmath38 au , @xcite ) .
then , we run a high resolution model ( 100@xmath36100 grid points ) for @xmath7 and @xmath6 which best fit the observations , to better resolve the chemical and temperature gradients and the vertical extent of the line forming region in the disk .
the predicted fwhm and peak separation do not depend on the model resolution while the line fluxes are lower by up to @xmath39% when increasing the resolution .
the line profiles and fluxes are obtained by first solving the statistical equilibrium with 2d escape probability to obtain the level populations , and then using 2d ray - tracing . as the three brightest lines of each cn 54 j
j transition are blended and the three undetected components are more than one order of magnitude fainter , the model computes the sum of the cn 54 j=9/27/2 and j=11/29/2 lines .
the results obtained for the grid of models ( figure [ fig : grid ] ) show that the cn line fluxes depend mainly on the disk size and increase for increasing outer radii .
this is due to the fact that , as suggested by their profiles , the cn lines originate from the outer region of the disk .
the best fit of the cn line fluxes is obtained for @xmath40 au . however , as model - predicted fluxes are affected by @xmath39% uncertainty , values of @xmath7 between @xmath41 au produce cn line fluxes which are still in agreement with the observations .
the ratio between the 566.9 and the 566.7 ghz lines predicted by the models , @xmath42 , is slightly lower but still in agreement with the observed one ( @xmath43 ) .
the peak separation and fwhm predicted by the models depend on both @xmath7 and @xmath6 . for @xmath44 au the observed @xmath45 and fwhm are reproduced if the disk inclination is low ( @xmath46 ) .
this is in agreement with the conclusions by @xcite and @xcite based on the modeling of the sed and near- , mid - infrared visibilities .
larger inclination angles of @xmath47 @xcite are excluded based on the observed cn profiles . following the results obtained from the grid of models we adopt @xmath48 au , @xmath3 and compute the fluxes and profiles of all the lines covered by our observations using the high resolution model .
figure [ fig : cn_model ] shows the region in the disk where 50% of the cn n=54 j=9/27/2 and j=11/29/2 lines originate according to our model .
this is obtained using vertical escape probability and without accounting for disk inclination .
the model indicates that the cn lines are excited in a disk upper layer located at 40110 au distance from the star , which is irradiated by the stellar uv field .
this heats the gas up to temperatures of @xmath49 k , while the gas density is @xmath50 @xmath8 , thus the cn lines are almost thermalized and optically thick ( @xmath51 ) in the line emitting region .
in this disk layer , cn is produced by h + cn@xmath52 , and c + no reactions , and photo - dissociation of hcn , while the main destruction route is photo - dissociation of cn . in figure [ fig : hifi_lines ] and table [ tab : lines ] the model - predicted line profiles and fluxes are compared with the observed ones . for cn lines , the model predicts a fwhm of 3.9 kms@xmath1 and peak separation of 2.6 kms@xmath1 , in agreement with observations ( fwhm@xmath53 kms@xmath1 , @xmath54 kms@xmath1 ) .
also the line fluxes agree with observations within a factor @xmath55 .
the sum of the fluxes of the n=21 components , instead , is about an order of magnitude lower than what observed by @xcite , who suggest that the line is dominated by envelope emission .
the co , @xmath56co , h@xmath0o , and so line fluxes predicted by the disk model are from a factor four to two orders of magnitude smaller than observed , as well as for the high - j co , h@xmath0o , oh , and atomic [ ] , [ ] lines previously observed with pacs ( see table 4 by @xcite ) .
this is a further evidence that these lines are dominated by envelope and outflow emission , as already suggested by the fact that [ ] , [ ] emission is spatially extended with pacs ( @xmath57 ) and by the broad profiles of co , h@xmath0o , so , and [ ] obtained with hifi .
the _ herschel_/hifi observations of the t tau system show emission in a number of molecular and atomic lines .
the origin of the emission lines in embedded , accreting / ejecting sources is highly debated , being crucial to identify a tracer to dig out the faint disk emission ( e.g. , * ? ? ?
* ; * ? ? ?
in the case of t tau , the co , h@xmath0o , and [ ] lines clearly trace high - velocity outflowing gas .
by contrast with those lines and with previously observed cn 21 lines @xcite , the cn 54 lines show narrow double - peaked profiles centered at the systemic velocity , suggesting an origin in the outer disk of t tau n. disk modeling predicts cn line fluxes and profiles in agreement with observed ones and constrains the size of the gaseous disk of t tau n ( @xmath2 au ) and its inclination ( @xmath3@xmath4 ) .
this study demonstrates that high - j cn lines are a unique tool to probe the gaseous disk of strongly accreting / ejecting sources and paves the way for future observations of embedded disks with alma . | we present spectrally resolved _ herschel_/hifi observations of the young multiple system t tau in atomic and molecular lines . while co , h@xmath0o , [ ] , and so lines trace the envelope and the outflowing gas up to velocities of 33 kms@xmath1 with respect to systemic , the cn 54 hyperfine structure lines at 566.7 , 566.9 ghz show a narrow double - peaked profile centered at systemic velocity , consistent with an origin in the outer region of the compact disk of t tau n. disk modeling of the t tau n disk with the thermo - chemical code prodimo produces cn line fluxes and profiles consistent with the observed ones and constrain the size of the gaseous disk ( @xmath2 au ) and its inclination ( @xmath3@xmath4 ) .
the model indicates that the cn lines originate in a disk upper layer at 40110 au from the star , which is irradiated by the stellar uv field and heated up to temperatures of @xmath5 k. with respect to previously observed cn 21 millimeter lines , the cn 54 lines appear to be less affected by envelope emission , due to their larger critical density and excitation temperature .
hence , high - j cn lines are a unique confusion - free tracer of embedded disks , such as the disk of t tau n. |
[ pachner 1986 , @xcite ] [ thm : pachner ] two closed combinatorial n - manifolds are pl - homeomorphic if and only if it is possible to move between their triangulations using a sequence of bistellar moves ( pachner moves ) and simplicial isomorphisms .
this theorem suggests that we can look for quantities invariant under the bistellar moves so as to obtain invariants of closed combinatorial manifolds .
the most famous one is the euler characteristic defined as the alternating sum of numbers of simplices in a combinatorial manifold .. 2 cm [ roberts 2002 , @xcite ] [ thm : justin ] any linear combination of the numbers of simplices which is an invariant of closed combinatorial manifolds must be proportional to the euler characteristic .
a real value invariant @xmath1 of a closed combinatorial manifold @xmath2 is called * _ localizable _ * if there exists a real value function @xmath3 on the set of isomorphism classes of pl @xmath4-spheres such that @xmath5 where @xmath6 is the link of a vertex @xmath7 in @xmath2 .
we call @xmath3 a * _ local formula _ * for @xmath1 .
.2 cm euler characteristic @xmath8 is localizable .
a local formula for @xmath9 is : @xmath10 where @xmath11 is the number of @xmath12-simplices in @xmath6 .
rational pontryagin numbers of closed combinatorial manifolds are localizable ( see @xcite and @xcite ) . all known local formulae for rational pontryagin numbers depend on the geometric patterns of the link of each vertex . but could there exist any local formulae for rational pontryagin numbers that only depend on the quantity of simplices in each link ? more generally , we may ask the following .. 2 cm * question 1 : * can we find new localizable invariants of combinatorial manifolds other than euler characteristic whose local formulae only depend on the quantity of simplices in the link of each vertex ?
.2 cm in this paper , we will give negative answer to the question 1 .
indeed , we will show the following .
[ thm : main ] suppose @xmath1 is a localizable invariant of closed combinatorial manifolds @xmath2 that can be written as @xmath13 where @xmath14 is the number of @xmath15-simplices in the link @xmath6 of @xmath7 in @xmath2 , and @xmath3 is an @xmath16-variable function on @xmath17 . then up to a constant , @xmath1 is proportional to the euler characteristic .. 2 cm there are no local formulae for rational pontryagin numbers of closed combinatorial @xmath18-manifolds that depend only on the quantity of simplices in a link .
.2 cm in the category of all compact combinatorial manifolds ( with or without boundary ) , it is shown in @xcite that : a localizable invariant of compact combinatorial manifolds must be proportional to euler characteristic .
however , in the category of closed combinatorial manifolds , this statement is wrong because the rational pontryagin numbers are localizable too .
.2 cm the paper is organized as following . in section [ sec2 ] ,
we discuss some combinatorial properties of bistellar moves which are useful in this paper . in section [ sec3 ]
, we define a type of @xmath19 @xmath16-disks in each dimension @xmath20 ( called gadget cells ) , which help us to understand how a local formula changes its values under all bistellar moves . in section [ sec4 ] ,
we give a proof of theorem [ thm : main ] . in section [ sec5 ] , we do some simple calculations to verify theorem in dimension @xmath21 .
.2 cm in addition , theorem [ thm : main ] is trivial in dimension @xmath22 .
so we assume @xmath20 in the rest of the paper .
and we use @xmath23 to denote the binomial coefficient @xmath24 through out the paper .
we first recall some definitions in combinatorial topology ( see @xcite ) .
.2 cm suppose @xmath25 is a simplicial complex , the * _ star _ * @xmath26 of a simplex @xmath27 in @xmath25 is the subcomplex consisting of all the simplices of @xmath25 that contain @xmath27 .
the * _ link _ * @xmath28 of @xmath27 is the subcomplex consisting of all the simplices @xmath29 of @xmath25 with @xmath30 and @xmath31 being a simplex in @xmath25 .
a simplicial complex is called an @xmath16-dimensional * _ closed combinatorial manifold _ * if the link of each vertex of the complex is an @xmath32-dimensional @xmath19 sphere .
let @xmath2 be a combinatorial @xmath16-manifold and @xmath33 a @xmath34-simplex ( @xmath35 ) such that its link in @xmath2 is the boundary @xmath36 of an @xmath15-simplex @xmath37 that is not a face of @xmath2 .
then the operation @xmath38 is called a @xmath16-dimensional * _ bistellar _ @xmath15-_move _ * ( or pachner move ) .
bistellar @xmath15-moves with @xmath39 $ ] is also called _
reverse bistellar _
@xmath34-_move_ .. 2 cm for example : @xmath40-move adds a new vertex to a triangulation ( we assume that @xmath41 ) , and a reverse @xmath40-move deletes a vertex , see figure [ p:2dim_move ] and figure [ p:3dim_move ] .
note that except @xmath40-move and reverse @xmath40-move , all other bistellar moves do not change the number of vertices in @xmath2 .. 2 cm , title="fig:",scaledwidth=90.0% ] + , title="fig : " ] + when we apply a bistellar @xmath15-move to @xmath2 , the links of some vertices involved in the move will be changed .
we have the following simple observation .
.2 cm [ lem : move ] for any @xmath42 , doing an @xmath16-dimensional bistellar @xmath15-move @xmath43 in @xmath2 will induce an @xmath4-dimensional bistellar @xmath15-move on the link of each vertex of @xmath27 and induce an @xmath4-dimensional bistellar @xmath44-move on the link of each vertex of @xmath37 .
doing an @xmath16-dimensional bistellar @xmath40-move will induce an @xmath4-dimensional bistellar @xmath40-move on the link of each vertex of @xmath27 .
for each vertex @xmath45 of a @xmath34-simplex @xmath27 , let @xmath46 represent the codimension @xmath22 face of @xmath27 that does not contain @xmath45 .
then the change of @xmath47 under the bistellar @xmath15-move @xmath43 is : @xmath48 which by our notation is an @xmath4-dimensional bistellar @xmath15-move @xmath49 .
similarly , for any @xmath50 , the change of @xmath51 under @xmath43 is @xmath52 which is an @xmath4-dimensional bistellar @xmath44-move @xmath53 .
.2 cm let @xmath54 be the set of all isomorphism classes of @xmath4-dimensional pl - spheres ( without orientations ) .
for any @xmath55 , let @xmath56 be the number of @xmath15-simplices in @xmath57 and call @xmath58 the @xmath0__*-vector * _ _ of @xmath57 .
in addition , we define @xmath59 .
.2 cm let @xmath60 for any @xmath4-dimensional bistellar @xmath15-move @xmath61 on @xmath57 , let @xmath62 be the @xmath0-vector of @xmath57 after the move .
it is easy to see that : @xmath63 where @xmath64 .
it is easy to check @xmath65
so the reverse bistellar @xmath15-move on @xmath57 gives @xmath67 .2 cm suppose @xmath1 is a localizable invariant of closed combinatorial manifolds and @xmath68 for some function @xmath3 that only depends on @xmath69 .
then @xmath3 is a function @xmath70 . by theorem [ thm : pachner ] , @xmath1 is invariant under all bistellar moves .
so for a bistellar @xmath15-move @xmath43 , according to lemma [ lem : move ] , the function @xmath3 must satisfy the following equations : * when @xmath71 or @xmath16 , we have @xmath72 + @xmath73 * when @xmath74 , we have @xmath75 when @xmath76 , we have @xmath77 + where @xmath78 is the @xmath79-vector of the boundary of an @xmath16-simplex . .2 cm * warning : * 1 .
we can not conclude that @xmath80 is a constant directly from equation . because in general , we may not be able to guarantee all the vertices of @xmath27 have isomorphic links .
only when the @xmath19 @xmath16-ball is very symmetric , may we have this kind of configuration in the combinatorial structure .
in general , we can not fit arbitrarily given @xmath19 @xmath16-balls together such that their intersection is a simplex .
.2 cm 2 .
it is not true that an arbitrary link @xmath81 could be involved in the equation for all @xmath15 . because when @xmath82 , there might be no bistellar @xmath15-move involving @xmath83 .
this means that for any @xmath34-simplex or @xmath15-simplex @xmath27 in @xmath84 , either @xmath28 is not isomorphic to the boundary of a simplex of complementary dimension , or @xmath28 is the boundary of some existing simplex in the combinatorial manifold .
.2 cm [ rem : universal_0_1 ] for any vertex @xmath83 in a closed combinatorial manifold , @xmath83 can always be involved in bistellar @xmath40 and @xmath22-moves .
.2 cm +
in this section , we introduce some special type of @xmath19 @xmath16-disks in each dimension @xmath20 and use them to figure out how a local formula @xmath3 changes its value under bistellar moves if @xmath3 depends only on the @xmath0-vector of a link .
.2 cm [ lem : gadgetcell ] for each @xmath20 , there exists a @xmath19 @xmath16-disk @xmath85 and a vertex @xmath86 such that : 1 .
@xmath87 is isomorphic to the boundary of an @xmath16-simplex .
2 . for any @xmath88
, there exists a bistellar @xmath15-move @xmath43 in @xmath85 with @xmath89 and @xmath90 , and @xmath43 does not cause any changes to the star of any vertex on @xmath91 other than @xmath45 . for each @xmath88 ,
let @xmath92 be a simplex of dimension @xmath15 .
let @xmath93 and choose a vertex @xmath94 of @xmath95 in @xmath96 .
let @xmath97 be the one - point union of @xmath98 got by gluing @xmath94 to a base point @xmath99 .
on the other hand , let @xmath100 be two @xmath16-simplices such that @xmath101 and @xmath102 is a vertex @xmath45 of both .
next , we glue @xmath99 to @xmath45 and put @xmath97 inside @xmath103 such that @xmath104 . by making up some new simplices and adding some new vertices inside @xmath103 if necessary , we get a pl @xmath16-disk @xmath85 ( see figure [ p : magic_cell ] for a construction of @xmath105 ) .
it is easy to see that doing the obvious bistellar @xmath15-move @xmath106 ( replaceing @xmath96 by @xmath107 ) associated to @xmath96 inside @xmath103 will not change the star of any vertex on @xmath87 except @xmath45 .
so such a @xmath85 satisfies all our requirements .
.2 cm , title="fig:",scaledwidth=92.0% ] + note that our construction of @xmath85 is far from unique . here
, we only need to choose one such @xmath85 in each dimension @xmath20 .
we call @xmath85 the * _ gadget @xmath16-cell _ * and @xmath45 is called the * _ base point _ * of @xmath85 .
let @xmath108 be the number of @xmath15-simplices of the link of @xmath45 in @xmath85 that lie in the interior of @xmath109 , and define @xmath110 .. 2 cm let @xmath111 .
.2 cm @xmath112 . for any @xmath55 ,
let @xmath113 be an pl @xmath16-ball . obviously , the link of @xmath45 is isomorphic to @xmath57 .
next , we choose an arbitrary @xmath16-simplex in @xmath114 and turn it into the gadget @xmath16-cell @xmath85 by subdivisions such that @xmath45 is the base point
. then the link of @xmath45 in @xmath114 becomes a new @xmath19 @xmath4-sphere whose @xmath0-vector is @xmath115 .
.2 cm [ lem : const_1 ] suppose @xmath3 is local formula of a localizable invariant that only depends on the @xmath0-vector of a link .
for any @xmath88 , @xmath116 is independent on @xmath117 .
suppose @xmath55 such that @xmath118 .
suppose @xmath83 is a vertex in a closed combinatorial manifold @xmath2 such that @xmath119 .
we choose an @xmath16-simplex in @xmath84 and turn it into the gadget @xmath16-cell @xmath85 by subdivisions such that @xmath83 is the base point .
now , @xmath120 . for any @xmath88 , we do the obvious @xmath16-dimensional bistellar @xmath15-move @xmath106 associated to @xmath96 in the gadget cell .
let @xmath121 be all the vertices involved in @xmath106 other than @xmath83 . by the construction of @xmath85 in lemma [ lem
: gadgetcell ] , for each @xmath122 , the star of @xmath123 is completely inside @xmath85 and the change of @xmath124 under @xmath106 is also inside @xmath85 .
so in the equations , all terms are canonically decided by @xmath85 except @xmath125 .
so @xmath116 does not depend on @xmath126 .
.2 cm for any @xmath117 and any @xmath88 , let @xmath127 .2 cm [ equ : d_i ] for any @xmath88 , we have 1 .
@xmath128 , and @xmath129 , @xmath130 .
2 . @xmath131 .
3 . @xmath132 if @xmath133 .
so each @xmath134 is a rational multiple of @xmath135 .
take a bistellar @xmath15-move @xmath43 in a closed combinatorial @xmath16-manifold .
for each vertex @xmath7 of @xmath27 and @xmath37 , we choose an @xmath16-simplex in @xmath136 and turn it into the gadget cell @xmath85 by subdivisions ( of course , sharing gadget cells between different stars are allowed ) .
then each @xmath69 becomes an element in @xmath137 after these subdivisions . by the equations and lemma [ lem : const_1 ] , we have the first equality .
the second and third equalities follow easily from the first one .
[ lem : const ] suppose @xmath3 is a local formula of a localizable invariant that only depends on the @xmath0-vectors of links .
for any @xmath88 and any @xmath4-dimensional bistellar @xmath15-moves on a @xmath19 @xmath4-sphere @xmath57 , @xmath138 .
suppose @xmath61 is an @xmath4-dimensional bistellar @xmath15-move on @xmath139 .
then we can find a vertex @xmath83 in a closed combinatorial manifold @xmath2 such that @xmath119 and an @xmath16-dimensional bistellar @xmath15-move @xmath43 in @xmath2 such that @xmath140 , and the move @xmath61 on @xmath81 is induced from @xmath43 .
for any vertex @xmath141 of @xmath27 and @xmath37 , we can choose an @xmath16-simplex @xmath142 with @xmath143 , and then turn @xmath144 into the gadget cell @xmath85 by subdivisions .
after these subdivisions , @xmath6 is changed and @xmath69 becomes an element in @xmath137 while @xmath84 stays the same .
so by lemma [ lem : const_1 ] and the equation , we have the following .
@xmath145 by lemma [ equ : d_i ] ( 1 ) , we have @xmath146 .2 cm the main idea in the above lemmas is : the change of the @xmath0-vector of a @xmath19-sphere caused by different bistellar @xmath15-moves are the same , i.e. the change does not depend on where the @xmath15-moves take place .
so the variations of @xmath147 under different bistellar @xmath15-moves on a @xmath19-sphere @xmath57 are the same if the value of @xmath147 only depends on the @xmath0-vector of @xmath57 .
for any @xmath19 @xmath4-sphere @xmath57 , by theorem , we can apply a finite sequence of @xmath4-dimensional bistellar moves to the boundary of an @xmath16-simplex @xmath148 to get @xmath57 . for each @xmath88 , suppose we have @xmath149 bistellar @xmath15-moves in the sequence . then by equation , @xmath150 satisfy @xmath151 the solution to equation is not unique because adding a bistellar @xmath15-move and its reverse simultaneously to a solution will give a new solution .
but the only thing important here is @xmath152 which is , in fact , uniquely determined by @xmath153 .
.2 cm for any @xmath154 - 1 $ ] , @xmath55 , @xmath155 is uniquely determined by @xmath156 in equation .
first of all , by equation and , the equation becomes : @xmath157 - 1}{\underset{i=0}{\sum } } \left ( m_i(l ) - m_{n-1-i}(l ) \right ) \cdot r_{k , i } = f_k(l ) - c^{k+1}_{n+1 } , \
\forall\ , 0 \leq k \leq n-1\ ] ] in addition , the dehn - sommerville equations of @xmath19-spheres imply that the @xmath153 is completely decided by @xmath158 - 1}(l)$ ] .
so the solution to above system of @xmath16 linear equations equal the solution to the @xmath159 $ ] linear equations : @xmath160 - 1}{\underset{i=0}{\sum } } \left ( m_i(l ) - m_{n-1-i}(l ) \right ) \cdot r_{k , i } = f_k(l ) - c^{k+1}_{n+1 } , \
\forall\ , 0 \leq k \leq [ \frac{n}{2}]-1.\ ] ] .2 cm notice when @xmath161 - 1 $ ] , @xmath162 - 1 $ ] , @xmath163 .
so * if @xmath164 , @xmath165 .
* if @xmath166 , @xmath167
. the square integral matrix @xmath168 - 1 } $ ] is invertible over @xmath169 .
so the above system of @xmath159 $ ] linear equations has a unique solution . when @xmath170 is odd , by equation , @xmath171 for any @xmath172 .
so in equation , the term @xmath173 is omitted .
indeed , we will see that when @xmath170 , the value of @xmath174 does not affect our calculation of @xmath3 since @xmath175 and @xmath176 ( see lemma [ equ : d_i ] ) .
.2 cm for any @xmath161 - 1 $ ] and @xmath177 , we can assume @xmath178 - 1}_{k=0 } c_{ik } \left ( f_k(l ) - c^{k+1}_{n+1 } \right ) , \
c_{ik } \in { \mathbb z}\]].4 cm * proof of theorem [ thm : main ] : * for a local formula @xmath3 of a real value localizable invariant @xmath1 of closed combinatorial @xmath16-manifolds , by lemma [ equ : d_i ] and lemma [ lem : const ] , for any @xmath179 , we have @xmath180 - 1}_{i=0 } ( m_i(l)-m_{n-1-i}(l ) ) \cdot h^n_i \notag \\ & = \psi(\mathbf{f_{\delta^n } } ) + \sum^{[\frac{n}{2}]-1}_{i=0 } \sum^{[\frac{n}{2}]-1}_{k=0 } h^n_i\cdot c_{ik } \left ( f_k(l ) - c^{k+1}_{n+1 } \right ) .
\end{aligned}\ ] ] so @xmath181 is a linear function of @xmath182 . moreover , since each @xmath134 is a rational multiple of @xmath135 , so up to a constant factor @xmath135 , we can write @xmath183 - 1}_{k=-1 } b_k\cdot f_k(l ) , \;\ b_k \in { \mathbb q}\quad ( \text{note}\ f_{-1}(l ) = 1)\ ] ] then for a closed combinatorial @xmath16-manifold @xmath2 , @xmath184 - 1}_{k=-1 } b_k \cdot f_k(lk(v ) ) \end{aligned}\ ] ] for a closed combinatorial manifold @xmath2 , let @xmath185 be the number of @xmath12-simplices in @xmath2 .
then obviously @xmath186 by equation , @xmath187 - 1}_{k=-1 } b_k ( k+2)\cdot f_{k+1}(m^n ) \end{aligned}\ ] ] so @xmath188 is a linear function of @xmath189 .
then by theorem [ thm : justin ] , @xmath1 must be proportional to the euler characteristic .
@xmath190 + +
when @xmath191 , by the dehn - sommerville equations for pl - spheres , we find that the @xmath0-vector of a pl @xmath192-sphere @xmath57 depends only on the number of vertices and edges in @xmath57 .
so we can write @xmath193 the system of equations are : @xmath194 so @xmath195 , @xmath196 .
in addition , by lemma , we have @xmath197 then by equation , we have @xmath198 this is proportional to the local formula for euler characteristic .
in fact , a local formula for euler characteristic of @xmath21-dimensional closed combinatorial manifolds @xmath199 is : for any vertex @xmath200 , @xmath201 the dehn - sommerville equations for the pl @xmath192-spheres imply : @xmath202 @xmath203 so @xmath204 .
then we have @xmath205 + a. a. gaifullin , _ computation of characteristic classes of a manifold from a triangulization of it _ , uspekhi matem .
nauk * 60 * : 4(2005 ) , 3766 ; english transl . , russian math .
surveys * 60*:4 ( 2005 ) , 615644 | if a real value invariant of closed combinatorial manifolds admits a local formula that depends only on the @xmath0-vector of the link of each vertex , then the invariant must be proportional to the euler characteristic . |
this work is a part of the research plan msm 0021620859 supported by the ministry of education of the czech republic and by the project 202/06/0363 of czech grant agency .
it is also partly supported by grant no .
fis2005 - 02796 ( mec , spain ) .
r. g. n. gratefully acknowledges support from the ramn y cajal programme ( spain ) .
99 a. bohr and b. r. mottelson , _ nuclear structure _ vol .
ii ( benjamin , new york , 1975 ) . see , for example , s. zhu _ et al .
_ , phys .
b 618 ( 2005 ) 51 ; p. mason _
et al . _ ,
c 72 ( 2005 ) 064315 ; m. e. debray _
et al . _ ,
c 73 ( 2006 ) 024314 ; y. j. chen _ et al .
_ , phys .
c 73 ( 2006 ) 054316 .
see for a review p. butler and w. nazarewicz , rev .
phys . 68 ( 1996 ) 349 .
f. garote , j. l. egido , and l. m. robledo , phys .
lett 80 ( 1998 ) 4398 .
m. yamagami and k. matsuyanagi , nucl .
phys . a 672 ( 2000 ) 123 .
t. tanaka , r. g. nazmitdinov , and k. iwasawa , phys .
c 63 ( 2001 ) 034309 .
a. tsvetkov , j. kvasil , and r. g. nazmitdinov , j. phys .
g : nucl . part .
phys . 28 ( 2002 ) 2187 .
j. dobaczewski and j. dudek , comp .
comm . 131 ( 2000 ) 163 .
j. kvasil and r. g. nazmitdinov , phys .
c * 69 * ( 2004 ) 031304(r ) ; pisma v zhetf 83 ( 2006 ) 227 [ jetp lett .
83 ( 2006 ) 187 ] ; phys .
c 73 ( 2006 ) 014312 .
j. kvasil , n. lo iudice , v. o. nesterenko , and m. kopl , phys .
c 58 ( 1998 ) 209 .
r. g. nazmitdinov , d. almehed , and f. dnau , phys .
c 65 ( 2002 ) 041307(r ) .
i. hamamoto , nucl .
phys . a 271 ( 1976 ) 15 .
s. cwiok , j. kvasil , and b. choriev , j. phys .
g : nucl .
phys . 10 ( 1984 ) 903 .
h. sakamoto and t. kishimoto , nucl .
phys . a 501 ( 1989 ) 205 .
r. g. nazmitdinov , yad .
fiz . 46 ( 1987 ) 732 [ sov .
j. nucl .
phys . 46 ( 1987 )
s. frauendorf and f. r. may , phys .
b 125 ( 1983 ) 245 .
j. l. egido , h. j. mang , and p. ring , nucl .
phys . a 339 ( 1980 ) 390
; r. bengtsson , yong - shou chen , jing - ye zhang , and s. berg , nucl . phys . a 405 ( 1983 ) 221
; y. r. shimizu and k. matsuyanagi , prog .
phys . 71 ( 1984 ) 960 . | a shape evolution of @xmath0yb in yrast states is traced using the self - consistent skyrme hartree - fock calculations .
we found that nonaxial octupole deformations ( in particular , @xmath1 term ) become favorable at @xmath2mev in @xmath3yb , while in @xmath4yb a nonaxial quadrupole shape is dominant at fast rotation .
the cranked nilsson model and random phase approximation are used to understand the dynamics of octupole correlations in both nuclei .
we demonstrate that the disappearance of one of the octupole vibrational modes in the rotating frame gives rise to the nonaxial octupole deformations in @xmath3yb , while the octupole modes are nonzero in @xmath4yb .
shell structure plays a prominent role in the formation of ground and excited nuclear states .
it is crucially important at fast rotation , when nuclear stability against fission is determined by a delicate balance of different constituents of a nuclear potential . in a geometrical approach @xcite , where the concept of shape deformation is one of the basic cornerstones ,
effects produced by quadrupole degrees of freedom are well understood for various effective potentials .
whether octupole degrees of freedom are of any importance and how they are manifested, these questions stimulate a noticeable fraction of experimental efforts in high spin physics @xcite .
numerous calculations exploiting the intrinsic reflection symmetry breaking of a nuclear mean field predict the existence of stable axial octupole deformation in the ground state for ra - th @xmath5 , @xmath6 ) and ba - sm @xmath7 , @xmath8 ) nuclei , where strong octupole instability is due to the coupling of @xmath9 , @xmath10 , @xmath11 orbitals @xcite .
the potential energy surface produced by different calculations is , however , quite shallow .
experimental nuclear spectra also show a pronounced parity splitting at low angular momenta .
rather dynamical octupole effects ( vibrations ) are dominant over static effects ( octupole deformations ) at low spins .
the increase of the angular momentum decreases pairing correlations which reduce the octupole interaction , since they couple the orbitals with the same parity . with rotation , the density of different parity states is enhancing markedly near the fermi level .
indeed , one observes a smooth decrease of the parity splitting with the increase of the rotational frequency fairly in a few nuclei .
these features are nicely reproduced within the cranking+hartree - fock - bogoliubov ( hfb ) approach with gogny forces in ba - sm region @xcite .
similar calculations with skyrme @xcite and gogny @xcite forces predict a nonaxial @xmath1 octupole deformation in light nuclei at high spins .
the results based on the skyrme interaction demonstrate the importance of a nonaxial @xmath12 octupole deformation in actinide nuclei at fast rotation @xcite .
notice that rotation induces the contribution of nonaxial quadrupole and octupole components of effective nuclear potentials . a wealth of experimental information accumulated in the past few years on high spin states in the transitional nuclei @xmath13
, @xmath14 prompted us to investigate the question to what extent dynamical octupole effects observed at low spins in this mass region might develop static octupole deformations at sufficiently high angular momenta .
this is a main subject of this letter . to analyse the rotational evolution of equilibrium deformations we have employed the hfodd code @xcite . in the present calculations
the pairing correlations are neglected , since they are not expected to play an important role at large rotational frequencies .
our hf calculations based on the cranked skyrme skp interaction ( see details in ref . )
demonstrate that a spontaneous symmetry breaking phenomenon occurs in @xmath3yb at large rotational frequencies ( see fig.1 ) . and
octupole @xmath15 momenta for a broken ( solid line ) and an unbroken ( dashed line ) reflection symmetry skp skyrme mean field solutions for @xmath3yb . ] the mean field solution with a broken reflection symmetry ( with nonzero @xmath16 , @xmath17 momenta and a small admixture of @xmath18 one , see fig.2 ) becomes favorable at @xmath2mev , in contrast to the solution with a reflection symmetry . in @xmath4yb
both the minimal solutions are very close to each other , while the pure quadrupole one determines properties of yrast states at fast rotation .
hereafter , @xmath19 means the averaging over the mean field vacuum ( yrast ) state at a given rotational frequency @xmath20 .
we obtain a similar pattern with the skiii parametrisation , which conforms our main finding , however , with slightly higher values of the energy minima . at slow rotation @xmath21mev ( when pairing is important )
the results should be taken cautiously , while the calculations provide a most credible tendency at @xmath22mev . a transparent physical idea
that an instability of a nuclear potential with respect to a given deformation implies a softening of the corresponding vibrational mode @xcite enables us to shed light on the mean field results .
the cranked shell model that incorporates a random phase approximation ( rpa ) represents a powerful tool to illuminate dynamics of various correlations , evolving from shape fluctuations ( vibrations ) to a nuclear shape instability .
we use a crpa approach @xcite that consists in a self - consistent solution of the cranked nilsson potential for the yrast line and analysis of the low - lying excitations near the yrast line in the rpa .
our consideration is based on the hamiltonian @xmath23 the term @xmath24 contains the nilsson hamiltonian @xmath25 with quadrupole deformations and the additional term that restores the local galilean invariance of the nilsson potential , broken in the rotating frame . to describe different parity states ,
the interaction @xmath26 includes separable monopole pairing , monopole - monopole , quadrupole - quadrupole and spin - spin terms for the positive parity @xmath27 and dipole - dipole , octupole - octupole terms for the negative parity @xmath28 .
the chemical potentials @xmath29 @xmath30n or p ) are determined so as to give correct average particle numbers @xmath31 .
all multipole and spin - multipole operators have a good isospin @xmath32 and signature @xmath33 ( see the properties of the matrix elements in ref . ) .
they are expressed in terms of doubly stretched coordinates @xmath34 , which ensure the self - consistent conditions at the equilibrium deformation .
all the details of this approach are thoroughly discussed in ref .. although the crpa approach is based on the nilsson potential , it contains static and dynamic effects induced by rotation and serves a useful purpose of independent view of the skyrme results .
the mode associated with the rotation about the @xmath35-axis determines the thouless - valatin moment of inertia @xmath36 @xmath37= \frac{\hat j_x}{\cal j_{tv } } , \quad [ { \hat \phi},{\hat j}_x]=i.\ ] ] here , @xmath38 is the positive signature term of the full hamiltonian ; an angle operator @xmath39 is the canonical partner of the angular momentum operator @xmath40 .
the solution of these equations leads to the definition @xmath41 where the matrix a(b ) has a dimension @xmath42 due to a coupling of different operators involved in @xmath38 @xcite .
we recall that a comparison of the dynamical moment of inertia @xmath43 @xmath44 ) with the thouless - valatin @xmath45 moment of inertia calculated in the rpa provides a faithful test of the self - consistency of a cranking microscopic theory .
the equivalence between these momenta of inertia certainly holds , if one found a self - consistent mean field minimum and spurious solutions are separated from the physical ones ( see exact results for a solvable model in ref . ) .
the use of the phenomenological rotational dependence of the gap parameter near the backbending region ( see ref . ) , the @xmath46 and @xmath47 terms of the nilsson potential affect this equivalence .
however the result achieved in our calculations yields a reasonable self - consistency ( see fig.3 ) .
we obtain a good agreement between our results and experimental kinematic @xmath48 and the dynamic @xmath49 momenta of inertia for the yrast states . here , @xmath50 is an experimental energy for the rotational band @xmath51 ( @xmath52 ) @xcite .
the drastic change of the kinematic momenta of inertia starts slightly earlier and ends slightly later than in experiments .
it is known that in the backbending region the cranking approach is less reliable @xcite .
nevertheless , one observes a similar rotational dependence of the magnitude of the calculated and experimental kinematic momenta of inertia till @xmath53mev in @xmath4yb , while a good agreement holds till @xmath54mev in @xmath3yb ( see below ) .
the results for the dynamic momenta of inertia reproduce with a reasonable accuracy the fluctuations in the backbending region , in spite of the approximations made in our approach .
yb ( left ) and @xmath4yb ( right ) .
top panels : equilibrium deformations in @xmath55-@xmath56 plane as a function of the angular momentum @xmath57 ( in units of @xmath58 ) .
middle panels : the kinematic @xmath59 momenta of inertia . filled square ( circle ) is used for experimental positive ( negative ) parity lowest states .
different symbols display the experimental data associated with b1,b2 ... bands ( the band labels are taken in accordance with the definitions given in ref . ) .
bottom panels : the dynamic @xmath60 momenta of inertia are compared with the corresponding experimental values ( filled squares ) .
experimental values are connected by thin line to guide eyes .
the calculated values of the kinematic ( middle ) and dynamic ( bottom ) momenta of inertia for the yrast line are connected by a solid line .
the dashed line displays the behavior of the thouless - valatin moment of inertia ( bottom ) .
all results are obtained within the cranked nilsson plus rpa ( crpa ) approach . ] the consistency between the mean field and the rpa results was achieved by varying the strength constants of the pairing and multipole interactions in the rpa ( see ref . ) to fulfill the conservation laws for both signatures and both parities @xmath61 & = & 0 , \,\,\,\,\,\,\,\ , \left[\hat h_\omega(^{\pi=-}_{r=+ } )
\,,\hat p_x \right ] = 0 , \nonumber\\ \left[\hat h_\omega(^{\pi=+}_{r=+})\,,\,\hat j_x \right ] & = & 0 , \,\,\,\,\,\,\,\ , \left[\hat h_\omega(^{\pi=+}_{r=-}),\hat \gamma^{\dagger}\right ] = \hbar\omega \hat \gamma^{\dagger } , \label{7}\end{aligned}\ ] ] where @xmath62 .
it results in the separation of physical modes ( excitations ) from those that are related to the symmetries broken by the mean field .
two goldstone modes are associated with the violation of the particle number ( for protons and for neutrons ) .
the other two goldstone modes are related to the translational and spherical symmetries of the mean field .
the last equation yields a negative signature solution @xmath63 , a collective rotational mode arising from the symmetries broken by the external rotational field .
we recall that the decoupling of @xmath64 and @xmath65 components of the center of mass operator from physical modes is uniquely defined in the cranking model with a signature and a parity symmetry @xcite .
the constraint satisfaction of the translational symmetry related to the operator @xmath66 determines , therefore , the isoscalar dipole strength solely and also guarantees the decoupling of the spurious center of mass motion from physical modes .
this is especially important for the calculations of @xmath67-transitions . for the isovector dipole coupling strength
we adopt the standard value @xmath68 with @xmath69mev @xcite .
the self - consistent octupole strength constants obtained for the anisotropic harmonic oscillator @xcite are used in our calculations for octupole excitations , taking into account a rotational dependence of different mean field values in the original expression . by virtue of these constants we nicely reproduce the experimental octupole routhians ( see below ) .
. the rotational evolution of crpa solutions : the positive ( negative ) signature ones with even ( odd ) spins are displayed on left ( right ) panels .
number in a circle denotes the rpa solution number : @xmath70 is the first @xmath71 rpa solution _
etc_. different symbols display the experimental data associated with b1,b2 ... bands .
the yrast line position denoted as `` 0 '' ( top , left ) is displayed by dashed line on all panels .
top panels : positive parity @xmath72 solutions .
the redundant mode ( right ) @xmath73 is denoted as `` 0 '' and is displayed by dotted line .
bottom panels : negative parity @xmath74 solutions . ]
the different response on the rotation can be understood from the analysis of excited states . to this aim
we define the experimental excitation energy in the rotating frame @xmath75 as a function of the rotational frequency @xmath20 @xcite . here , the routhian function @xmath76 .
the experimental states are classified by the parity @xmath77 and a quantum number @xmath78 which is equivalent to our signature @xmath79 ( see figs.4,6 ) .
the positive signature states ( @xmath80 ) correspond to @xmath81 which characterizes rotational bands with even spin in even - even nuclei .
the negative signature states ( @xmath82 ) correspond to @xmath83 and are associated with odd spin states in even - even nuclei .
the energy @xmath84 can be compared with the rpa results , @xmath85 , calculated at a given rotational frequency .
the agreement between our results and experimental data indicates that quadrupole deformation ( axial and nonaxial ) is a major factor that determines the evolution of the mean field till @xmath86mev in @xmath4yb ( @xmath3yb ) .
. reduced @xmath67- and @xmath87- transition probabilities from the lowest negative parity one - phonon states to the yrast states .
we used the effective proton ( neutron ) charge @xmath88 for the calculation of @xmath67-transitions , while only the proton contribution is taken into account for the @xmath87- transitions .
left(right ) panels correspond to the transition with @xmath89 even ( @xmath89 odd ) .
all results are obtained within the crpa approach . ] in @xmath3yb the band b4 crosses the band b1 at the rotational frequency @xmath90mev ( a transition point ) and becomes the yrast one for @xmath91mev . the change of the yrast line structure produces the backbending ( see fig.3 ) .
it should be emphasized that the potential - energy surfaces are very shallow .
the deformations shown in fig.3 should be not taken as precise predictions , since even before the transition point the yrast states may possess small @xmath92 @xmath56-deformation ( see also refs . ) . to elucidate the structure near the transition point we trace the rotational evolution of the crpa solutions and the lowest two - quasiparticle poles . at @xmath93
each quasiparticle orbital ( routhian ) is characterized by asymptotic nilsson quantum numbers .
these numbers are irrelevant , however , in the rotating case and are used only for convenience .
the first positive parity and signature rpa solution ( see fig.4 , top left panel ) may be identified with @xmath55-excitations at @xmath94mev .
the increase of the rotational frequency leads to a strong mixing between @xmath56- ( the second rpa solution at low spins ) and @xmath55-excitations . at the transition point from an axial to nonaxial rotations ,
@xmath95mev , a two - quasiparticle neutron component @xmath963/2[651]$ ] dominates in the quadrupole phonon structure ( @xmath97 ) .
this fact suggests that the alignment of a pair @xmath98 is the main mechanism that drives the nucleus to triaxial shapes .
the first negative signature rpa solution ( fig.4 , top right panel ) carries the properties of a gamma - vibrational mode ( with odd spins ) till @xmath99mev and does not lead to any instability .
yb .
similar to fig.4 . ] from fig.4 ( bottom panel , right ) one observes that the first negative signature and parity rpa ( octupole ) solution tends to zero .
almost a zero energy gap between the negative parity band ( b5 ) and the yrast line is also observed at @xmath100mev .
the collectivity of the lowest negative parity rpa solutions of both signatures increases noticeably with the increase of the rotational frequency .
results for reduced @xmath67- and @xmath87-transition probabilities evidently demonstrate the onset of octupole correlations at @xmath101mev ( see fig.5 ) .
for example , @xmath102-transition probability increases from @xmath103 w.u . at @xmath104 mev to @xmath105 w.u . at @xmath106 mev
( here w.u is the weisskopf unit ) .
several two - quasiparticle components originated from @xmath107 and @xmath108 subshells for protons and @xmath98 and @xmath109 subshells for neutrons contribute to the collectivity of the lowest negative parity one - phonon states .
the maximal weight of two - quasiparticle components is @xmath110 .
all these features reveal the nature of a shape transition at @xmath111mev .
the onset of the static octupole deformations becomes feasible , since the instability point found in the crpa approach coincides with the result of the skyrme mean field calculations ( see fig.1 ) .
these results suggest that the octupole deformations are due to the octupole phonon condensation at fast rotation .
yb .
similar to fig.5 . ] in @xmath4yb ( see fig.6 ) , there are observed 15 rotational bands so far @xcite .
the positive signature family ( with even spin ) is comprised of b1 , b2 , b8 , b9 , b10 , b11 bands of positive and b3 , b4 , b5 bands of negative parities .
the negative signature family ( with odd spins ) includes b2 , b12 , b13 , b14 , and b6 , b7 , b15 of positive and negative parity bands , respectively . at @xmath112mev
the band b11 crosses the ground band b1 .
this crossing gives rise to the backbending ( see fig.3 ) . according to our analysis ,
the first positive parity and signature rpa solution ( see fig.6 , top left panel ) may be identified with @xmath56-excitations at small rotation @xmath94mev .
the rotation induces a strong mixing between @xmath56-and @xmath55-excitations ( the second rpa solution at low spins ) . although the pattern is similar to the one of @xmath3yb , @xmath55- and @xmath113 vibrations are interchanged at small rotation . at @xmath114mev the transition from an axial to nonaxial rotations is determined by two - quasiparticle neutron component from @xmath98 subshell .
a good agreement between data and our results implies that quadrupole degrees of freedom dominate in this domain of rotational frequency values .
the negative parity rpa solutions of both signatures ( see fig.6 ) decrease with the rotational frequency .
the rpa results reproduce with a good accuracy the rotational behaviour of the negative parity states , especially , the lowest one . while the octupole phonons are of collective nature ( the maximal component is @xmath115 ) , the quadrupole deformed field remains stable , even at very high spins .
the growing octupole collectivity is exhibited in the increasing dipole and octupole transitions from the negative - parity one phonon states at @xmath116mev ( see fig.7 ) .
in particular , @xmath102-transition probability increases from @xmath117 w.u . at @xmath104 mev to @xmath118 w.u . at @xmath106 mev .
in contrast with @xmath3yb , all negative parity solutions are nonzero at @xmath119mev .
octupole correlations are unable to break the reflection symmetry in @xmath4yb , which elucidates the result of skyrme calculations ( see fig.1 ) .
summarizing , we found that the internal structure of the low - lying excited ( @xmath72 , @xmath80 ) band , which crosses the ground band , has almost a pure two - quasiparticle character originating from @xmath98 neutron subshell , in agreement with the prediction of the phenomenological analysis @xcite . in both nuclei
the rotation produces a profound affect on shell structure inducing noticeable octupole correlations at high spins . at @xmath116mev
we predict the formation of octupole bands with strong @xmath67- and @xmath87- transitions to the yrast line ( see figs.5,7 ) .
we found that at @xmath120mev the octupole phonon solution vanishes in the rotating frame in @xmath3yb .
it results in the onset of the degeneracy between the lowest negative parity and negative signature band and the positive parity and positive signature yrast line .
this mechanism explains the spontaneous breaking of reflection symmetry of the rotating mean field with skyrme interaction ( see fig.1 ) .
in contrast , in @xmath4yb the octupole correlations manifest themselves as a low - lying octupole vibrations of the quadrupole deformed rotating nucleus , in agreement with the skyrme results . |
the understanding that entanglement serves as a physical resource for various quantum communication or information processing protocols has motivated both experimental and theoretical studies in the preparation and quantification of entangled states @xcite .
entanglement between two subsystems of a quantum system , i.e. , bipartite entanglement , is well understood and can be directly quantified by a number of readily computable measures .
multipartite quantum systems , however , offer a richer variety of entangled states , the characterization of which is , naturally , more complicated and indeed still under active investigation ( see , e.g. , @xcite ) . of particular interest in this context
are the @xmath0-qubit symmetric dicke states @xcite ; in particular , the states @xmath8 ( with @xmath9 ) satisfying @xmath10 with @xmath11 , where @xmath12 are collective angular momentum operators and @xmath13 gives the total number of excitations in the system .
these states are , with regards to entanglement , particularly robust against particle loss @xcite , and they are significant for the study and application of genuine multipartite entanglement @xcite .
photonic systems have proven fruitful for the experimental preparation and detection of three - photon polarization - entangled @xmath14 states ( dicke states with one `` excitation '' ) @xcite and , very recently , four - photon polarization - entangled dicke states with two excitations @xcite . meanwhile ,
single collective atomic excitations
atomic @xmath14 states have been produced in many - atom ensembles @xcite and with small collections of trapped ions @xcite .
the potentially long coherence times of ground electronic states of atoms ( two of which constitute the qubit ) make atomic systems particularly attractive for the storage and manipulation of entangled states , and a variety of schemes have been proposed for the preparation of more general dicke states of trapped atoms or ions @xcite . here
we describe a system that adds to these proposed schemes , in that it may produce collective atomic states with multiple excitations , but also yields very interesting `` entangled - state dynamics '' and conditional quantum evolution , based upon photon detection in the output field of an optical cavity containing the atoms .
the cavity qed setup and excitation scheme that we employ have been considered previously in the context of generating , in steady state , two - qubit ( i.e. , two - atom ) entangled mixed states of any allowed combination of purity and entanglement @xcite , although here we consider a specific limiting case of the operating conditions ( not considered in @xcite ) that does not in fact admit a unique steady state . in examining this specific case
, we gain inspiration from a recent investigation of conditional quantum dynamics in a related cavity qed system involving two atoms in separate , cascaded cavities @xcite , which also focussed on a specific limiting case of the operating conditions for an atom - entanglement scheme proposed earlier @xcite . in particular
, we use the method of quantum trajectories based upon continuous monitoring of the cavity output field by photodetection to study the conditional quantum evolution of the collective state of @xmath0 atoms inside the ( single ) cavity . as in @xcite
, we observe manifestly distinct behaviours from one trajectory to the next ; the system either evolves to a ( steady ) dicke state with @xmath2 excitations without emitting any photons ( for even @xmath0 ) , or it executes a sustained cycle of `` switches '' between particular , well - defined superpositions of dicke states , with a concomitant continuous output stream of photons .
furthermore , for @xmath15 more than one distinct switching cycle is possible , with the different cycles distinguished by the rate of photon emissions .
we start in section [ quanttrajs ] with a brief review of the cavity qed configuration proposed in @xcite , after which we formulate our quantum trajectory model , making use of a convenient change of basis , which is motivated by the particular form of the effective hamiltonian and jump operators . in section [ steadystate ]
we focus briefly on the particular case in which the system evolves into a stable dicke state with @xmath2 excitations .
individual trajectory results for @xmath16 atoms , highlighting cyclic behavior of the state , are presented and discussed in section [ esc ] .
we also make some more general observations concerning cycles for larger @xmath0 , before concluding in section [ conclusion ] .
we consider a system of @xmath0 four - level atoms trapped inside a high finesse optical cavity . the atomic excitation scheme and physical implementation
are depicted in fig .
[ physsys ] .
a single quantized cavity mode couples to the @xmath17 and @xmath18 transitions in each atom with coupling strengths @xmath19 and @xmath20 , respectively .
auxiliary laser fields couple to the @xmath21 and @xmath22 transitions with rabi frequencies @xmath23 and @xmath24 , respectively .
together , the cavity and laser fields drive ( resonant ) raman transitions between the stable atomic ground states @xmath25 and @xmath26 .
such an excitation scheme could be achieved , for example , by using orthogonal , linearly - polarized laser and cavity fields on an @xmath27 atomic transition ( as in @xmath28 ) , or on an @xmath29 transition ( as in @xmath30 ) with an applied magnetic field , as described in @xcite .
finally , we note that the spacings between the atoms are assumed to be sufficiently large that direct dipole - dipole interactions can be neglected .
we also assume that each atom is held tightly in position by a confining trap , so that the atomic motion can be omitted from the analysis . ) and two laser fields ( rabi frequencies @xmath31 ) drive resonant raman transitions between the atomic ground states @xmath25 and @xmath26 .
note that an additional laser or magnetic field may be required in the excitation scheme to provide an independent relative energy shift between the ground states @xmath25 and @xmath26 ( see @xcite ) .
note also that @xmath32 and @xmath33 may be a single level , provided the two raman channels remain distinct from each other ( which would require that the two ground states are nondegenerate in energy ) . for large detunings @xmath34 and @xmath35 , and for @xmath36 , the cavity field decay rate ,
much larger than the raman transition rates , the model may be reduced to that of an effective two - level atom with characteristic transition rates @xmath37 and @xmath38 . , width=321 ] with the assumption of large detunings @xmath34 and @xmath35 , the atomic excited states can be adiabatically eliminated and atomic spontaneous emission neglected .
the coherent dynamics is then characterized by raman transition rates and energy level shifts , which we subsequently assume to be much smaller than the cavity field decay rate , @xmath36 .
this enables us to adiabatically eliminate the cavity field to derive a master equation that describes the atomic dynamics alone . under `` resonant '' conditions ( i.e.
, light shifts of the atomic ground states are balanced , such that raman resonance is maintained ; see @xcite for details ) , this master equation can be written in the form @xmath39 where @xmath40 is the atomic density matrix , @xmath41 with @xmath42 , @xmath43 , and @xmath44 for @xmath45 , the dynamics of ( [ mastereq1 ] ) produces unique steady states that are non - maximally - entangled ( pure or mixed ) collective spin states @xcite . however , for the remainder of this paper we consider the case @xmath46 for which ( [ mastereq1 ] ) does not admit a unique steady state ( mathematically , the system of linear equations for the density matrix elements possesses a second zero eigenvalue when @xmath47 ) . scaling time by @xmath48 , the master equation we thus focus on takes the form @xmath49 where @xmath50 is the @xmath51-component of the total pseudo - spin @xmath52 of all @xmath0 qubits . note
that since @xmath53 commutes with @xmath54 , the dynamics described by ( [ mastereq2a ] ) conserves the magnitude of the total spin angular momentum ( which we will take to be the maximal value in this work ) @xcite .
we are interested in the time - dependent behavior of the atomic state subject to dynamics as described by ( [ mastereq2a ] ) and , as depicted in fig . [ physsys](b ) , in photon counting measurements on the cavity output field .
the method of quantum trajectories ( see , e.g. , @xcite ) , summarized briefly in the appendix , lends itself nicely to this task . in this approach , and with photon counting measurements in mind , we rewrite ( [ mastereq2a ] ) in the form @xmath55 and identify a non - hermitian effective hamiltonian , @xmath56 , and collapse operator , @xmath57 , as @xmath58 in the trajectory model , free evolution of the atomic wave function , subject to @xmath56 , is interspersed , at random times , with collapses ( or quantum jumps ) produced by the action of @xmath57 on the wave function .
these collapses can be identified with photon detections in the cavity output field and the resulting wave function trajectory describes the state of the atomic spin system conditioned upon the sequence of photon detections . given the form of both @xmath56 and @xmath57 , a particularly convenient basis for our study will be the basis of eigenstates of the @xmath51-component of the total spin @xmath52 ; more precisely , simultaneous eigenstates of @xmath59 and @xmath54 . in the more familiar representation of simultaneous eigenstates of @xmath59 and @xmath60 , which we denote by @xmath61 ,
the allowed values of the @xmath62-component are @xmath63 , where , for a system of @xmath0 effective spin-1/2 particles , as we consider here , @xmath64 may take one of the values @xmath65 .
so , for example , for @xmath66 atoms and @xmath67 , @xmath68 while for @xmath69 ( both spins `` up '' or `` down '' ) @xmath70 where @xmath71 @xmath72 denotes the state with atom 1 in state @xmath73 and atom 2 in state @xmath74
. to obtain the eigenstates of @xmath75 , we perform the appropriate set of euler rotations on the coordinate system .
in particular , the @xmath54-eigenstates are related to the @xmath62-eigenstates by a unitary rotation operator with the property @xmath76 : @xmath77 here , @xmath78 are the matrix elements of the rotation operator given by ( sometimes referred to as wigner s formula ) @xcite @xmath79 note that the sum in ( [ wignersformula ] ) is only over terms with non - negative arguments of the factorials . finally , we will confine our study to evolution from the initial state @xmath80 i.e. , @xmath9 ( the maximal value ) and all spins `` down '' . since @xmath64 is conserved by the evolution , only @xmath81 linearly independent basis states are required to describe the state of the @xmath0-atom system . in the basis of @xmath54-eigenstates it is very straightforward and instructive to calculate the general behavior of the spin wave function either undergoing free evolution subject to @xmath56 , or being `` collapsed '' under the action of the jump operator @xmath82 . in particular ,
consider first free evolution from the initial state @xmath83 up to some time @xmath84 , without any collapses having occurred .
for even @xmath0 , the ( unnormalized ) state @xmath85 can be written as @xmath86 & = & d^{(n/2)}_{0,-n/2}\left(-\frac{\pi}{2}\right ) { \left| n/2,0 \right\rangle}_{x } \nonumber \\ & & + \sum^{n/2}_{\substack{m =- n/2 \\ m\ne0 } } d^{(n/2)}_{m ,- n/2}\left(-\frac{\pi}{2}\right ) \:e^{\!-m^{2}t } { \left| n/2,m \right\rangle}_x . \nonumber\\\end{aligned}\ ] ] noting that @xmath87 , it is useful to define @xmath88 and rewrite ( [ timeevo ] ) as @xmath89 where @xmath90 .
this result shows explicitly that the @xmath67 component of the spin wave function in the @xmath54-eigenstate basis is unaffected by the free evolution , while the amplitudes of the @xmath91 components decay exponentially with time .
conversely , the @xmath67 component is completely removed by the action of the jump operator , i.e. , @xmath92 , while @xmath93 for odd @xmath0 , there is no @xmath67 component and the ( unnormalized ) wave function takes the form @xmath94 after a time @xmath84 of free evolution ( from the initial state @xmath83 ) .
hence , all components of the wave function have amplitudes that decay exponentially .
now , in the trajectory formalism , the square of the norm of the wave function , @xmath95 , can be identified as the probability for a collapse ( i.e. , a photon detection ) _ not to have happened _ . from this identification and our observations
above , it follows that the system under investigation can exhibit very distinct behaviors , as we detail in the following two sections .
from ( [ evo ] ) , it follows that for an even number @xmath0 of atoms there is a finite probability , given by @xmath97 ^ 2 } \
, , \label{p_nm0}\end{aligned}\ ] ] that no collapse ever occurs and , hence , that the system evolves to a stable steady state given by @xmath96 .
this is a dicke state , in the basis of @xmath54-eigenstates , with @xmath2 excitations , which , as mentioned in the introduction , is of particular interest in the context of multipartite quantum entanglement .
in fact , for our system a necessary and sufficient condition for genuine @xmath0-partite entanglement is @xcite @xmath98 which is `` optimally '' satisfied for the state @xmath96 , for which @xmath99 .
note that with a standard unitary rotation , implemented for example using coherent laser raman pulses , this state can be transformed into the corresponding state @xmath100 in the ( conventional ) @xmath62-eigenstate basis .
the probability @xmath101 is plotted in fig .
[ pjmx0 ] as a function of @xmath0 and maintains a significant value even for quite large @xmath0 .
in fact , using stirling s approximation ( @xmath102 for large @xmath103 ) , one can show that for @xmath104 , @xmath105 which describes the slow decrease with @xmath0 , as illustrated in fig .
[ pjmx0 ] .
it must be emphasized that in the case that the state @xmath96 is prepared , no photons are emitted from the cavity . with a probability @xmath106 there _ is _ a collapse , or photon detection .
this in fact initiates a sustained sequence of collapses ( and photon detections ) , since the amplitudes of all of the remaining ( @xmath91 ) components of the wave function decay exponentially under free evolution , i.e. , the norm of the wave function is now guaranteed to decay towards zero , meaning that a collapse is also guaranteed to occur .
hence , the system exhibits a distinctly bimodal behavior : either no photons are detected , `` signalling '' preparation of the state @xmath96 , or a continuous stream of photons is detected .
while this stream of photons indicates failure in the preparation of a steady state , it also signals the initiation of some remarkable dynamics of the system , which we label _ entangled state cycles _ and examine in the next section . for preparation of the dicke state @xmath96 versus @xmath2 ( @xmath0 even ) .
the crosses are calculated from eq .
( [ p_nm0 ] ) .
the solid line is the curve @xmath107^{1/2}$ ] , which approximates @xmath101 for @xmath108.,width=302 ]
as described above , a first photon detection initiates a persistent sequence of photon detections and the relevant atomic spin states involved in the dynamics from then on are the states @xmath109 .
note that for odd @xmath0 , a first photon detection and subsequent persistent sequence is guaranteed to occur .
the rate of photon detections is proportional to the expectation value of the operator @xmath110 , and hence to the value @xmath111 .
the photon detection configuration we consider can therefore be interpreted as providing a measurement of @xmath111 , so we can expect each trajectory to `` converge '' towards a particular pair of states , @xmath109 ( or possibly to the state @xmath96 for even @xmath0 ) , conditioned upon the measurement record .
for example , from ( [ jumps ] ) we see that collapses increase the amplitude @xmath112 relative to @xmath113 , while from ( [ evo ] ) , we see that free evolution decreases @xmath112 exponentially relative to @xmath113 . hence , within a given time interval , the more photon detections that occur , the more probable it is for the state to converge towards states @xmath109 with larger @xmath5 ( for which , of course , a higher photon count rate is also expected ) . in contrast , the longer @xmath114 evolves without a collapse the more likely it is for the state to converge towards states @xmath109 with lower values of @xmath5 .
which state or pair of states ( @xmath96 or @xmath109 ) is `` selected '' in a particular trajectory is random . in the case of @xmath109
, the sequence of photon detections corresponds to a sequence of jumps , described by ( [ jumps ] ) , between the states @xmath115 and @xmath116 .
these states are also very interesting entangled states , and we label the system s behavior in this instance as an _ entangled - state cycle_. we now look at specific examples of entangled state cycles for @xmath16 .
consider free evolution of the initial state @xmath118 from @xmath119 up to time @xmath84 , without any collapses having occurred .
the normalized wave function is given by @xmath120 this illustrates very clearly that as time progresses and no collapses occur ( i.e. , no photons are detected ) , the more likely it is for the system to be in the state @xmath121 .
such a no - collapse trajectory ( an example of which is shown in fig .
[ 2qpic1 ] ) occurs with probability @xmath122 i.e. , in 50% of trials the system is prepared in the two - qubit maximally - entangled state @xmath123 .
time is measured in units of @xmath124 .
no collapses occur and the system is eventually projected into the @xmath121 eigenstate.,width=302 ] when a collapse does occur and a photon is detected , the system state is projected onto @xmath125 with the @xmath67 component removed , subsequent time evolution is guaranteed to lead to further collapses and a persistent cycle is established between the two states @xmath126 .
such a cycle is illustrated in fig .
[ 2qpic2 ] , where the abrupt changes in the sign of the amplitude @xmath127 indicate jumps between the states @xmath126 .
it is interesting to consider the compositions of the states @xmath126 in the original @xmath62-eigenstate basis ; in particular , @xmath128 and @xmath129 so , when a cycle is established , the states involved are also maximally - entangled states of the two qubits . .
the system settles into an entangled - state cycle after the first collapse.,width=302 ] considering the initial state @xmath130 we see that there is no @xmath67 component and hence an entangled state cycle always results .
in particular , a first collapse is guaranteed , as exemplified by the wave function after time @xmath84 of free evolution ( ignoring normalization for simplicity , i.e. , eq .
( [ evo_odd ] ) for @xmath131 ) , @xmath132 with probability @xmath133 the system evolves to an entangled - state cycle between the states @xmath134 . a sample trajectory of this cycle is shown in fig .
[ 3qpic1 ] . in this particular example , by the time of the first collapse , or photon detection , the components of @xmath135 have become negligibly small . a trajectory in which the three - qubit state evolves to a cycle between the states @xmath136 ( occurring with probability @xmath137 ) is shown in fig . [ 3qpic2 ] .
in this case , collapses are clearly much more frequent than in the previous example , since , mathematically , the norm of the wave function decays more rapidly between each collapse . .
the @xmath138 components decay to a negligible magnitude by the time of the first collapse and the system evolves into the cycle between the states @xmath134.,width=302 ] , where sufficient jumps occur over a time interval to promote the three - qubit state into a cycle involving @xmath136.,width=302 ] in fact , the ratio of the frequencies of collapses for the two cycles is given by @xmath139 so which cycle the system evolves into for any individual realization is clearly distinguishable from the photon counting record .
it is interesting to consider the states involved in the two cycles . in the @xmath54-eigenstate and @xmath62-eigenstate bases ,
respectively , we have @xmath140 \\ & = & \frac{1}{2\sqrt{3}}\left ( { \left| 011 \right\rangle}_z + { \left| 101 \right\rangle}_z + { \left| 110 \right\rangle}_z\right ) \nonumber \\ & & ~~ + \frac{\sqrt{3}}{2}{\left| 000 \right\rangle}_z \ , , \\[0.2 cm ] \label{3qchi - bare1 } { \left| \chi^{-}_{3}\!\left(\tfrac{1}{2}\right ) \right\rangle } & = & \frac{1}{\sqrt{6}}\left[\left({\left| 110 \right\rangle}_x + { \left| 101 \right\rangle}_x + { \left| 011 \right\rangle}_x\right ) \right . \nonumber \\ & & ~~ - \left .
\left({\left| 001 \right\rangle}_x + { \left| 010 \right\rangle}_x + { \left| 100 \right\rangle}_x \right)\right ] \\ & = & \frac{1}{2\sqrt{3}}\left({\left| 100 \right\rangle}_z + { \left| 010 \right\rangle}_z + { \left| 001 \right\rangle}_z\right ) \nonumber \\ & & ~~ - \frac{\sqrt{3}}{2}{\left| 111 \right\rangle}_z \,,\end{aligned}\ ] ] and @xmath141 in the @xmath54-eigenstate basis , the states @xmath136 are simply ghz states , while @xmath134 are each a linear combination of two @xmath14 states , @xmath142 and @xmath143 .
such combinations are sometimes referred to as @xmath144 states , which fall under the ghz class of entangled states @xcite , but are clearly less entangled than a single @xmath14 state . alternatively , in the @xmath62-eigenstate basis , the states @xmath134 and @xmath136 are all superpositions of a product state and a @xmath14 state @xcite . note again that unitary rotations of states between bases could be implemented using coherent raman transitions between the atomic ground states @xmath25 and @xmath26 . in the case of four qubits with initial state @xmath145 , ( [ evo ] )
reads @xmath146 so again two distinct entangled - state cycles are possible , corresponding to @xmath147 or @xmath148 and occurring with probabilities @xmath149 or @xmath150 , respectively . in the @xmath54-eigenstate basis , the cycles occur again either between two ghz states , @xmath151 , or between two @xmath144 states , @xmath152
. however , when represented in the @xmath62-eigenstate basis , the cycles now occur between quite different forms of entangled states .
in particular , the @xmath147 cycle is between a ghz state and a @xmath144 state , i.e. , @xmath153 \label{4qchi - bare1 } { \left| \chi^{-}_{4}(1 ) \right\rangle } & = & \frac{1}{\sqrt{8}}\left [ \left ( { \left| 0001 \right\rangle}_z + { \left| 0010 \right\rangle}_z + { \left| 0100 \right\rangle}_z + { \left| 1000 \right\rangle}_z\right ) \right .
\nonumber \\ & & - \left .
\left({\left| 1110 \right\rangle}_z + { \left| 1101 \right\rangle}_z + { \left| 1011 \right\rangle}_z + { \left| 0111 \right\rangle}_z \right ) \right ] \ , , \nonumber \\\end{aligned}\ ] ] while the @xmath148 cycle is between a @xmath144 state ( @xmath154 ) and a state that is an equal superposition of a ghz state and a dicke state with two excitations @xcite , i.e. , @xmath155 \label{4qchi - bare2 } { \left| \chi^{-}_{4}(2 ) \right\rangle } & = & \frac{1}{\sqrt{8 } } \left ( { \left| 0001 \right\rangle}_z + { \left| 0010 \right\rangle}_z + { \left| 0100 \right\rangle}_z + { \left| 1000 \right\rangle}_z \right . \nonumber \\ & & \left .
+ { \left| 1110 \right\rangle}_z + { \left| 1101 \right\rangle}_z + { \left| 1011 \right\rangle}_z + { \left| 0111 \right\rangle}_z \right ) \ , .
\nonumber \\
\end{aligned}\ ] ] ( note that the state @xmath156 is of the same form as ( [ 4qchi+bare2 ] ) , but with different weightings for the ghz and dicke states . )
examples of entangled - state cycles for @xmath157 are shown in figs .
[ 4qpic1 ] and [ 4qpic2 ] . with initial state @xmath145 .
in this case , the system settles into the cycle between the states @xmath152.,width=302 ] with initial state @xmath145 . in this case
, the system settles into the cycle between the states @xmath151.,width=302 ] for larger @xmath0 the number of possible cycles obviously increases , as does the variety of entangled states involved . from the decomposition of the initial state @xmath158 in terms of the states
@xmath159 and @xmath160 , one deduces that the probability , @xmath161 , for the system to be projected into the cycle involving the states @xmath162 is given by @xmath163 in the large-@xmath0 limit , and for @xmath164 , this is approximated well by the simple gaussian form ( derived using stirling s approximation ) @xmath165 as demonstrated in fig .
[ p_chi_m ] for @xmath166 and @xmath167 .
this result makes it very clear that cycles with @xmath168 become increasingly unlikely for large @xmath0 .
favored cycles are those with small @xmath5 , corresponding , in fact , to states satisfying the condition for genuine @xmath0-partite entanglement , @xmath98 i.e. , states with @xmath169 .
for the system to evolve into the entangled - state cycle between @xmath162 .
the crosses are calculated from eq .
( [ pchi ] ) the solid lines are the gaussian approximation given in eq .
( [ pchi_gauss]).,width=302 ]
in this work we have described a means of realizing , through suitable laser and cavity qed interactions , a collective atomic spin system with dynamics described by the simple master equation @xmath170 despite its apparent simplicity , this model reveals a remarkable diversity of behavior when viewed in a quantum trajectory picture , given an initial state @xmath171 and assuming photon counting measurements on the cavity output field .
the different behaviors can be readily understood by employing a basis of @xmath54-eigenstates and interpreting the configuration in terms of a projective measurement of the operator @xmath172 . on top of the interesting dynamical behavior
, the actual states produced , either in steady state ( @xmath67 ) or participating in an entangled - state cycle ( @xmath173 ) , typically possess genuine multipartite entanglement .
our work is actually quite closely related , but complementary , to that of stockton _
et al_. @xcite , who consider continuous quantum nondemolition measurement of the atomic spin operator @xmath62 , via a dispersive atom - cavity interaction and homodyne detection of the cavity field , starting from an initial atomic spin state polarized along the @xmath51-axis . there ,
dicke states in the @xmath62-eigenstate basis are prepared , either nondeterministically , with a probability determined by the initial state decomposition in this basis , or deterministically , via the use of state - based feedback control .
such a feedback approach could perhaps also be applied usefully to the present configuration .
this work was supported in part by the marsden fund of the royal society of new zealand .
asp thanks g. tth and e. solano for helpful discussions .
we will not discuss quantum trajectories in any profound way , but we simply outline how a realization of the reduced density operator for the system , called a trajectory , can be built operationally using a monte carlo algorithm .
in essence , the trajectory formalism reduces the problem of mixed - state evolution to that of a pure state , as we describe below .
references @xcite provide an excellent account of the formalism . for our quantum trajectory approach ,
the output of the optical cavity is considered to be monitored continuously by photon counting , as shown in fig .
[ physsys](b ) . in this scenario
, each photodetection sequence corresponds to a trajectory of the @xmath0-qubit collective atomic state .
consider a normalized pure state @xmath174 at time @xmath84 .
the state is evolved `` freely '' over a small ( ideally infinitesimal ) time increment @xmath175 according to @xmath176 where @xmath56 is a non - hermitian effective hamiltonian and the state is renormalized after the non - unitary evolution @xcite .
such free evolution is interspersed , at random times , with `` quantum jumps '' or `` collapses '' , the probability of which must be calculated at each timestep . in particular , the probability that a collapse of the state occurs over the ( infinitesimal ) time step @xmath175 is given by @xmath177 where @xmath57 is the collapse operator . to determine whether a collapse occurs or not , a uniformly distributed random number @xmath178 is drawn from the unit interval @xmath179 $ ] and compared to @xmath180 . if @xmath181 then the state becomes @xmath182 otherwise , for @xmath183 , the state is not collapsed and it continues to evolve , according to ( [ nonhermite ] ) , over another subsequent interval of @xmath175 .
this process is iterated over sufficiently many time steps to build up a trajectory .
note that for our system the action of @xmath57 represents the emission and subsequent detection of a photon in the cavity output field .
m. a. nielsen and i. l. chuang , _ quantum computation and quantum information _
( cambridge university press , cambridge , 2000 ) .
d. bouwmeester , a. ekert , and a. zeilinger ( eds . ) , _ the physics of quantum information _ ( springer , berlin , 2000 ) .
w. dr , g. vidal , and j. i. cirac , phys .
a * 62 * , 062314 ( 2000 ) .
a. wong and n. christensen , phys .
a * 63 * , 044301 ( 2001 ) .
d. collins _ et al_. , phys .
rev . lett . * 88 * , 170405 ( 2002 ) .
f. verstraete , j. dehaene , b. de moor , and h. verschelde , phys .
rev . a * 65 * , 052112 ( 2002 ) .
j. k. stockton , j. m. geremia , a. c. doherty , and h. mabuchi , phys .
a * 67 * , 022112 ( 2003 ) .
g. tth and o. ghne , phys .
lett . * 94 * , 060501 ( 2005 ) .
g. tth , j. opt .
b * 24 * , 275 ( 2007 ) . as defined in @xcite , an @xmath0-qubit state is said to possess genuine multipartite entanglement if it is not biseparable , i.e. , if it can not be expressed in the form @xmath184 , which is separable with respect to the partition @xmath185 .
n. kiesel _
et al_. , j. mod . opt . *
50 * , 1131 ( 2003 ) .
m. bourennane _
et al_. , phys .
lett . * 92 * , 087902 ( 2004 ) .
et al_. , phys .
lett . * 92 * , 077901 ( 2004 ) .
h. mikami , y. li , k. fukuoka , and t. kobayashi , phys .
* 95 * , 150404 ( 2005 )
. m. bourennane _
et al_. , phys .
lett . * 96 * , 100502 ( 2006 ) .
a. kuzmich _
et al_. , nature * 423 * , 731 ( 2003 ) .
m. d. eisaman _
et al_. , phys .
lett . * 93 * , 233602 ( 2004 ) . t. chanelire _
et al_. , nature * 438 * , 833 ( 2005 ) .
m. d. eisaman _
et al_. , nature * 438 * , 837 ( 2005 ) .
d. felinto _
et al_. , phys .
a * 72 * , 053809 ( 2005 ) .
m. d. lukin _
et al_. , phys .
. lett . * 87 * , 037901 ( 2001 ) .
r. g. unanyan and m. fleischhauer , phys .
. lett . * 90 * , 133601 ( 2003 ) .
duan and h. j. kimble , phys .
lett . * 90 * , 253601 ( 2003 ) .
x. zou , k. pahlke , and w. mathis , phys .
a * 68 * , 034306 ( 2003 ) .
j. k. stockton , r. van handel , and h. mabuchi , phys .
a * 70 * , 022106 ( 2004 ) .
xiao , z .- f .
han , j. gao , and g .- c .
guo , j. phys .
b : at . mol .
opt . phys . * 39 * , 485 ( 2006 ) .
a. retzker , e. solano , and b. reznik , phys .
a * 75 * , 022312 ( 2007 ) . in practice
, some ( small ) amount of atomic spontaneous emission , due to finite excitation of the atomic levels @xmath186 , can be expected to occur and this need not conserve the magnitude of @xmath187 .
so , our idealized model ( [ mastereq2a ] ) will strictly only be valid on timescales much shorter than the timescale for spontaneous emission , but we assume this timescale to be sufficiently long that we can observe the effects described in the following sections .
h. j. carmichael , s. singh , r. vyas , and p. r. rice , phys .
a * 39 * , 1200 ( 1989 ) .
h. j. carmichael , _ an open systems approach to quantum optics _
, lecture notes in physics vol . 18 ( springer - verlag , berlin , 1993 )
. j. dalibard , y. castin , and k. mlmer , phys .
* 68 * , 580 ( 1992 ) .
r. dum , a. s. parkins , p. zoller , and c. w. gardiner , phys .
a * 46 * , 4382 ( 1992 ) .
j. j. sakurai , _ modern quantum mechanics , revised edition _ , ( addison - wesley , 1994 ) , chap .
p. meystre and m. sargent iii , _ elements of quantum optics , 3rd edition _ , springer international edition ( springer - verlag , india , 2003 ) , chap . 15 .
it is not necessary to renormalize the state after each period of the non - unitary evolution . in an alternative
( but equivalent ) monte carlo routine , a uniform random number is drawn from the unit interval and the state is only made to collapse when its norm decays to that random number @xcite .
the state is then renormalized and the process is repeated sufficiently many times to obtain the required trajectory . | we study quantum trajectories of collective atomic spin states of @xmath0 effective two - level atoms driven with laser and cavity fields .
we show that interesting `` entangled - state cycles '' arise probabilistically when the ( raman ) transition rates between the two atomic levels are set equal . for odd ( even ) @xmath0 , there are @xmath1 ( @xmath2 ) possible cycles . during each cycle the @xmath0-qubit state switches , with each cavity photon emission , between the states @xmath3 , where @xmath4 is a dicke state in a rotated collective basis .
the quantum number @xmath5 ( @xmath6 ) , which distinguishes the particular cycle , is determined by the photon counting record and varies randomly from one trajectory to the next .
for even @xmath0 it is also possible , under the same conditions , to prepare probabilistically ( but in steady state ) the dicke state @xmath7 , i.e. , an @xmath0-qubit state with @xmath2 excitations , which is of particular interest in the context of multipartite entanglement . |
the @xmath3cdm ( or _ concordance _ ) model @xcite is one of the greatest successes of general relativity .
it reproduces beautifully all the main observational results e.g. , the dimming of type ia supernovae @xcite , cosmic microwave background radiation ( cmbr ) anisotropies @xcite , large scale structure formation @xcite , baryon oscillations @xcite and weak lensing @xcite ) .
unfortunately , this model is also affected by significant fine - tuning problems related to the vacuum energy scale and this has led to a considerable amount of effort throughly exploring other viable theoretical schemes . currently , one of the most popular alternatives to the @xmath3cdm model is based on gravitational actions which are non - linear in the ricci curvature @xmath4 and/or contain terms involving combinations of derivatives of @xmath4 : the so called @xmath0 theories of gravity @xcite .
such models first became popular in the 1980 s because it was shown that they can be derived from fundamental physical theories ( like m - theory ) and naturally admit a phase of accelerated expansion , which could be associated with an early universe inflationary phase @xcite .
the fact that the phenomenology of dark energy requires the presence of a similar phase ( although only a late time - low energy one ) has recently revived interest in these models .
in particular , the idea that dark energy may have a geometrical origin , i.e. , that there is a connection between dark energy and a non - standard behavior of gravitation on cosmological scales is now a very active area of research ( see for example @xcite ) . unfortunately efforts to obtain an understanding of the physics of these theories are hampered by the complexity of the fourth - order field equations , making it difficult to obtain both exact and numerical solutions , which can be compared with observations .
recently however , progress has been made in resolving these issues using a number of useful techniques .
one such method , based on the theory of dynamical systems @xcite has proven to be very successful in providing a simple way of obtaining exact solutions and a ( qualitative ) description of the global dynamics of these models @xcite . the dynamical systems analysis has up to now approached this problem in the conventional way , by first exploring the finite equilibrium points and then computing the asymptotic behavior of the phase - space using the classical method of poincare projections . in this paper
we develop an alternative scheme which involves compactifying the phase space for general @xmath0 theories of gravity , subject to certain conditions on the function @xmath1 .
we then use this framework to study the behaviour of the phase space of universes with a non - negative ricci scalar in @xmath2 gravity .
we find a number of interesting cosmological evolutions , which include the possibility , at least in principle , where the universe begins close to at an unstable power - law inflationary equilibrium point , then evolves towards a curvature fluid dominated phase where the effective equation of state mimics standard radiation with @xmath5 ( we will refer to such phases as radiation - like ) , then passes through a standard matter ( cdm ) era and ultimately evolves asymptotically towards a de - sitter - like late - time accelerated phase .
we also show that as @xmath6 , all the fixed points that approach the @xmath3cdm subspace of the complete state space of @xmath2 gravity , are unstable .
this implies that the behavior of the solutions of a fourth order theory which is close to @xmath3cdm may be completely different from those of @xmath3cdm and one needs to do a careful analysis of the solutions rather than a priori assuming any global behavior of the trajectories .
the natural extension of standard general relativity is to consider a lagrangian that contains curvature invariants of higher than linear order .
in fact , renormalization of quantum field theory suggests that adding such terms to the standard gravitational action appears to be necessary @xcite to give a first approximation to some quantized theory of gravity .
the quadratic lagrangians are at the first level of such modifications and have been studied extensively over true past two decades .
the four possible second - order curvature invariants are @xmath7 where @xmath8 is completely antisymmetric 4-volume element and @xmath4 , @xmath9 , @xmath10 are the ricci scalar , ricci tensor and riemann tensor respectively . however , for homogeneous and isotropic spacetimes , because of the following identities @xcite @xmath11 it follows that the general fourth - order lagrangian for a these highly symmetric spacetimes contain only powers of @xmath4 and we can write the action as = 12 d^4x , [ action ] where @xmath12 represents the matter contribution .
varying the action with respect to the metric gives the following field equations : f g_ab = t^m_ab+ 12 ( f - rf ) g_ab + _
b_af- g_ab_c^cf , [ field1 ] where @xmath13 denotes the derivative of the function @xmath1 w.r.t .
the ricci scalar and @xmath14 is the matter stress energy tensor defined by t^m_ab = ^mu_au_b + p^mh_ab+ q^m_au_b+ q^m_bu_a+^m_ab . here @xmath15 is the direction of a timelike observer , @xmath16 is the projected metric on the 3-space perpendicular to @xmath15 . also @xmath17 , @xmath18 , @xmath19 and @xmath20 denotes the standard matter density , pressure , heat flux and anisotropic stress respectively .
equations ( [ field1 ] ) reduce to the standard einstein field equations when @xmath21 . for the homogeneous and isotropic spacetimes with vanishing 3-curvature and barotropic perfect fluid as the standard matter source with equation of state @xmath22 , the independent field equations for general @xmath0 gravity are as follows .
* _ the raychaudhuri equation _ : @xmath23\;,\end{aligned}\ ] ] where @xmath24 is the volume expansion of the matter flow lines @xmath15 and @xmath25 is the standard matter density . * _ the friedmann equation _ : @xmath26 and * _ conservation of standard matter _ : @xmath27 combining the raychaudhuri and friedman equations ,
we obtain : @xmath28
in this paper we will study the dynamics of friedmannlematrerobertsonwalker ( flrw ) models only in the sector @xmath29 .
this is because the sector @xmath30 is not of much physical interest and also , as we shall see later , the sectors @xmath31 and @xmath30 are connected by the invariant sub - manifold @xmath32 , making the physically interesting dynamics completely confined to the sector @xmath31 .
also , we consider the 3-curvature to be vanishing , which is an invariant sub - manifold by itself . as required by the _ noghost condition _ we also assume @xmath33 . to compactify the phase space we rewrite the friedmann equation ( [ f ] ) in the following form : @xmath34 where @xmath35 we can now define the following set of normalized variables : @xmath36 to guarantee that the propagation equations for these compact variables will result in a dimensionless dynamical system , we need to define a new time variable @xmath37 , such that @xmath38 for @xmath37 to be a monotonously increasing time variable ,
a normalization d is chosen such that it is strictly positive at all times .
it is clear by construction that when @xmath39 the normalized dynamical variables as well as the time variable are well defined , thus this normalization allows the study of general static , re - collapsing and bouncing solutions . from the friedmann equation
we obtain the following constraints , @xmath40 the first constraint comes directly from friedmann equation , while the second one arises from the definition of the normalization parameter d. according to these constraints and considering @xmath31 , @xmath41 and @xmath33 , we see that the above dynamical variables have to be defined in the following ranges , @xmath42 making the complete phase space compact . also since the variable @xmath43 is a normalized hubble parameter , the cosmological solutions will naturally include both expanding and collapsing as well as static solutions and these two sets of solutions are connected via the non - invariant subset @xmath44 .
an autonomous system , which is equivalent to cosmological equations ( [ raychaudhuri]-[combination ] ) can be derived by differentiating the compact variables ( [ cv ] ) , with respect to @xmath37 and using ( [ raychaudhuri]-[conservation ] ) .
the dimensionality of the resultant system can then be reduced by using the two constraints ( [ const ] ) . by eliminating the dynamical variables
@xmath45 and @xmath46 , we obtain the following 3-dimensional effective autonomous system : @xmath47 + x^2\big[4 + 6\omega+z(-3(1+\omega)-2\gamma ) \big ] \big ) \nonumber\\ z'&=&\dfrac{z}{3}\big(-4q^2x - q((5 + 3\omega)x^2 + 3(1+\omega)(-1+z))\nonumber\\ & + & x(5 + 3\omega-(1 + 3\omega)x^2-\gamma\big[-2 - 3n(1+\omega)+2z+3n(1+\omega)(z+(q+x)^2)\big]\big)\\ q'&=&\dfrac{1}{6}\big(-4q^2x - q\big[(5 + 3\omega)x^2 + 3(1+\omega)(-1+z)\big]\nonumber\\ & + & x(5 + 3\omega-(1 + 3\omega)x^2-\gamma\big[-2 - 3n(1+\omega)+2z+3n(1+\omega)((q+x)^2+z)\big]\big)\;\nonumber,\end{aligned}\ ] ] where @xmath48 . in general , the system is not closed unless @xmath49 is expressed in terms of the dynamical variables ( [ cv ] ) . for example , in the case of @xmath2
, we have @xmath50}\ , . \label{gamma}\end{aligned}\ ] ] thus , the above system defines the dynamics of all well defined @xmath0 theories for which @xmath51 is invertible in terms of the dynamical variables . from equations ( [ ds ] ) we can see that @xmath52 is an invariant sub - manifold . and
in the @xmath52 2-surface the line @xmath44 is an invariant subset .
since @xmath52 corresponds to @xmath32 , we obtain an important result : _ for all well defined functions @xmath0 , with @xmath33 and @xmath51 invertible in terms of the dynamical variables defined by ( [ cv ] ) , a flrw universes with nonnegative ricci scalar , continues to be so , both in the future and in the past .
also an @xmath32 universe can never undergo a bounce in the future or the past .
_ in the next section we will fix the function @xmath1 to be the class of theories @xmath53 and study the dynamics of the flat flrw universes and their stability for those theories . in order to study the stability of the fixed points of the dynamical systems ( [ ds ] ) , we will use the very well known techniques , which involve linearizing the dynamical equations around the equilibrium points and then finding the eigenvalues of the linearization matrix ( the jacobian ) at the equilibrium points .
if the jacobian is well defined , then they can be classified according to the sign of the real part of eigenvalues as attractors , repellers and saddle points .
as we have seen from the equation ( [ gamma ] ) , @xmath51 is invertible in terms of the dynamical variables for @xmath53 . it is interesting to note that the constant ` @xmath54 ' couples to the dynamical equations ( [ ds ] ) only via the quantity @xmath49 and the constant @xmath55 does not couple to the equations at all . hence all the fixed point of the system
are necessarily independent of @xmath55 .
the coordinates of the fixed points are shown in table 1 .
note that each fixed point has an expanding ( @xmath56 ) and a collapsing ( @xmath57 ) version as indicated by the subscripts ( @xmath58 ; @xmath59 ) respectively .
also some points only occur in the compact state space defined by ( [ range ] ) for certain ranges of @xmath54 .
the occurrence of the fixed points outside the compact region for specific @xmath54 and @xmath60 means that the constraints ( [ const ] ) are not satisfied and consequently these fixed points are not physical for these values of @xmath54 and @xmath60 ) .
fixed points that are not physical for these values of @xmath54 and @xmath60 have been excluded from the analysis . * table 1*. coordinates of the equilibrium points for @xmath61-gravity .
we will not explicitly state the expressions for s , g1, ... ,g4 and f1, ... ,f4 , which are rational functions of n and @xmath60 , however we give them at the following link @xcite .
5.8pt [ cols="<,<,<",options="header " , ] the exact solutions at the fixed points are also summarized in table 1 and the stability analysis for the dust and radiation cases are summarized in table 2 .
first we discuss the static solutions . from definition ( [ ds ] ) ,
@xmath62 , so any fixed point that lies on the surface @xmath44 represents a static universe . by looking at the coordinates of the fixed points in table 1
, we can see that the point @xmath63 is static for all values of n and @xmath60 .
the point @xmath64 is static only for @xmath65 and @xmath66 and we find that for these values of @xmath54 the point @xmath64 represent an unstable saddle point .
we now proceed to find the exact solutions for the scale factor at the nonstatic fixed points .
the expansion rate and the deceleration parameter @xmath67 are related by the raychaudhuri equation , @xmath68 if we know the value of the deceleration parameter @xmath69 at some fixed point i , we can use the above equation to obtain the behavior of the scale factor at that point .
when @xmath70 we have desitter solutions @xmath71 or static solutions @xmath72 . for @xmath73
we have a milne evolution and when @xmath74 ; @xmath75 we have accelerated and decelerated power law behaviours respectively . to obtain the exact solutions for the scale factor @xmath76 associated with the nonstatic @xmath77 equilibrium points we need to have an expression for @xmath78 in term of the compact variables . from the definition @xmath78
we obtain : @xmath79 the non - invariant surface @xmath80 is the transition surface between accelerated and decelerated expansions phases ( see fig 1 ) . by substituting [ q in term of compact v ] in [ deceleraion ] we obtain : @xmath81 where @xmath82 . the evolution of the scale factor can now be given directly by integrating equation [ scale factor ] : @xmath83 where @xmath84 the constants of integration can be obtained by substituting the solutions into the original equations .
as explained in @xcite , these solutions must satisfy all the cosmological equations in order to be considered physical . by looking at table 1
we can distinguish two classes of nonstatic solutions .
the first class @xmath85 and @xmath86 } contain solutions that are independent of n and @xmath60 .
the fixed point b represents a static phase as mentioned earlier , the points a , d and f are radiationlike phases .
the expanding version of the point a is a saddle for @xmath87 and repeller for @xmath52 and the other two points d an f are saddles .
the fixed point n represent a matter phase and the expanding version of this point is a spiral@xmath88 . the evolution of the scale factor a(t ) for the fixed point @xmath89 is a function of @xmath90 and for fixed point @xmath91 is function of @xmath92 . the dependence of these solutions on n and/or @xmath60 provide us with additional degrees of freedom that can lead to interesting cosmological scenarios . when @xmath93 , the fixed point @xmath89 merges with @xmath94 for @xmath66 , and it merges with the point @xmath95 for @xmath96 .
when @xmath97 it merges with @xmath95 for @xmath66 , and for @xmath98 it corresponds to the matter point @xmath94 . in the case @xmath93 or @xmath97 , we find that for all values n for which this point is physical , the expansion ( contraction ) is never accelerating . as mentioned earlier , the evolution of the scale factor for the point @xmath91 is independent of the equation of state parameter @xmath60 . for @xmath99 the fixed point @xmath91
merges with @xmath100 , for @xmath101 it merges with point @xmath102 and for @xmath103 it is a matter point .
we also find that for this fixed point the expansion ( contraction ) is accelerating for @xmath104 .
the existence of this accelerated phase together with the fact that for @xmath104 the point @xmath89 is a matter - like point , leads to the possibility of an extremely interesting cosmological scenario , where it is possible in principle to find an orbit that starts close to the unstable accelerating phase @xmath105 , evolves past the unstable radiationlike point @xmath106 , followed by the unstable matter point @xmath107 and finally ends up at the desitter attractor @xmath108 . in figs . 3 and 4 we have plotted two interesting orbits .
the orbit in fig 3 is for @xmath93 and @xmath99 .
it begins near the radiationlike points @xmath109 , passes nearby the radiationlike point @xmath106 , followed by the standard matter point @xmath107 and ends up at the de - sitter attractor @xmath108 .
the orbit in fig 4 is for @xmath110 and @xmath103 .
it begins near the radiationlike point @xmath111 , passes nearby the matter point @xmath105 , then close to the matter point @xmath107 and ends up at the de sitter attractor @xmath108 .
it is interesting to see that for all fixed points whose xcoordinate goes to zero as @xmath6 , are unstable . as the @xmath3cdm subspace lies on the @xmath112 surface and this surface
is not an invariant submanifold , this implies that the behavior of the solutions of a fourth order theory which is close to @xmath3cdm , may be completely different from those of @xmath3cdm .
furthermore , this suggests that the best fit model to the current observational data within the complete state space of @xmath2 , may be given by a non - infinitesimal value of @xmath54 . for @xmath113 ; @xmath114 .
the left half of the state space corresponds to collapsing models , while the right half contain expanding models .
this is indicated by the subscripts of the various equilibrium points.,width=302 ] and @xmath99.,width=340 ] and @xmath103.,width=340 ]
in this work , we have presented a careful analysis of the state space of the class of @xmath2 theorys of gravity , focusing on the @xmath31 sector with @xmath115 , together with the no - ghost condition @xmath116 .
due to the complexity of this class of gravity theories , the standard hubble normalization does not lead to a compact dynamical variables . in order to construct variables defining a compact dynamical system
one has to use an appropriate normalization . in this paper
we used the same formalism used in @xcite , where we absorbed all the negative contributions of the friedmann equation into the normalization .
first of all we obtained the following important result : for all well defined function @xmath0 , with @xmath33 and @xmath51 invertible in terms of the dynamical variables defined by @xmath117 , the flrw universes with non - negative ricci scalar , continues to be so both in the future and in the past .
also an r = 0 universe can never undergo a bounce in the future or past .
our compact analysis shows that there are more equilibrium points than in the corresponding non - compact analysis in @xcite .
in particular we find a new finite fixed point @xmath86 .
because of it s very special location in the phase space , it is quite difficult to obtain this point using the standard techniques .
this point is found to represent a matter phase and the expanding version of this point is @xmath118 .
furthermore , we find that for @xmath119 the phase space of @xmath2 , contains two accelerated fixed points @xmath120 , together with two other saddle points ( one represent a radiation phase @xmath106 and the other represent a matter - like phase @xmath107 ) .
although we have obtained all the desired fixed points and desired stability , this does not necessarily imply that there is an orbit connecting them . due to the fact that for @xmath119 , the two accelerated points and the matter - like point are quite close to each other in the phase space , which makes it difficult to prove the existence of an orbit connecting these points together with the radiationlike point .
but the presence of all these phases in the state space of @xmath2 makes a more detailed investigation worth pursuing .
s. perlmutter _
_ , astrophys
. j. * 517 * , 565 ( 1999 ) ; a. g. riess _
et al . _ ,
j. * 116 * , 1009 ( 1998 ) ; j. l. tonry _
_ , _ astrophys . j. _ * 594 * , 1 ( 2003 ) ; r. a. knop _
_ , _ astrophys
. j. _ * 598 * , 102 ( 2003 ) ; a. g. riess _
_ astrophys .
j. _ * 607 * , 665 ( 2004 ) ; s. perlmutter _
_ astrophys .
j. _ * 517 * , 565 ( 1999 ) ; _ astron .
astrophys . _ * 447 * , 31 ( 2006 ) .
s. m. carroll , v. duvvuri , m. trodden and m. s. turner 043528 ( 2004 ) ; s. nojiri and s. d. odintsov , 123512 ( 2003 ) ; s. capozziello , _ int .
_ d * 11 * , 483 ( 2002 ) ; v. faraoni , 124005 ( 2005 ) ; m. l. ruggiero and l. iorio jcap * 0701 * 010 ( 2007 ) ; a.
de la cruz - dombriz and a. dobado a , , 087501 ( 2006 ) ; n. j. poplawski , , 084032 ( 2006 ) ; n. j. poplawski , , 3013 ( 2007 ) ; a. w. brookfield , c. van de bruck and l. m. h. hall , , 064028 ( 2006 ) ; y. song , w. hu and i. sawicki , 044004 ( 2007 ) ; b. li , k. chan and m. chu , , 024002 ( 2007 ) ; x. jin , d. liu and x. li .
[ arxiv : astro - ph/0610854 ] ; t. p. sotiriou and s. liberati s , 935 ( 2007 ) ; t. p. sotiriou , , 5117 ( 2006 ) ; r. bean , d. bernat , l. pogosian , a. silvestri and m. trodden , 064020 ( 2007 ) ; i. navarro and k. van acoleyen , jcap * 0702 * , 022 ( 2007 ) ; a. j. bustelo and d. e. barraco , 2333 ( 2007 ) ; g. j. olmo , 023511 ( 2007 ) ; j. ford , s. giusto and a. saxena , _ nucl .
_ b * 790 * , 258 ( 2008 ) ; f. briscese , e. elizalde , s. nojiri and s. d. odintsov , , 105 ( 2007 ) ; s. baghram , m. farhang and s. rahvar , , 044024 ( 2007 ) ; d. bazeia , b. carneiro da cunha , r. menezes and a. petrov , 445 ( 2007 ) ; p. zhang , 024007 ( 2007 ) ; b. li and j. d. barrow , 084010 ( 2007 ) ; t. rador , _ phys . lett .
_ b * 652 * , 228 ( 2007 ) ; t. rador , , 064033 ( 2007 ) ; l. m. sokolowski , _ class .
* 24 * , 3391 ( 2007 ) ; v. faraoni , , 067302 ( 2007 ) ; o. bertolami , c. g. boehmer , t. harko and f. s. n. lobo f s , , 104016 ( 2007 ) ; s. k. srivastava , _ int
. j. theor .
* 47 * , 1966 ( 2008 ) ; s. capozziello , v. f. cardone and a. troisi , jcap * 08 * , 001 ( 2006 ) ; a. a. starobinsky , _ jetp lett . _
_ 86 _ , 157 ( 2007 ) ; a. de felice , s. tsujikawa , f(r ) theories , arxiv:1002.4928v1 ; e. elizalde , s. d. odintsov , l. sebastiani , s. zerbini , arxiv:1108.6184v1 [ gr - qc ] ; e. elizalde , s. nojiri , s.d .
odintsov , l. sebastiani , s. zerbini , _ phys.rev .
_ d * 83 * 086006 ( 2011);s .
nojiri , s.d .
odintsov _ phys.rept .
_ * 505 * 59 - 144 , ( 2011 ) ; s. nojiri , s.d .
odintsov _ int.j.geom.meth.mod.phys .
_ * 4 * 115 - 146 , ( 2007 ) .
s. capozziello , v. f. cardone , a. troisi , _ jcap _ * 0608 * , 001 ( 2006 ) .
amare abebe , mohamed abdelwahab , alvaro de la cruz - dombriz , peter k.s . dunsby , [ arxiv:1110.1191 ] .
k. i. maeda and n. ohta , _ phys . lett .
_ b * 597 * , 400 ( 2004 ) ; k. i. maeda and n. ohta , _ phys . rev .
_ d * 71 * , 063520 ( 2005 ) ; n. ohta , _ int .
_ a * 20 * , 1 ( 2005 ) ; k. akune , k. i. maeda and n. ohta , _ phys . rev . _
d * 73 * , 103506 ( 2006 ) .
s. carloni , p. k. s. dunsby , a. troisi , _ phys .
_ d * 77 * 024024 ( 2008 ) ; k. n. ananda , s. carloni , p. k. s. dunsby , _ phys .
_ d * 77 * , 024033 ( 2008 ) ; k. n. ananda , s. carloni , p. k. s. dunsby , _ class .
_ * 26 * 235018 ( 2009 ) ; k. n. ananda , s. carloni , p. k. s. dunsby , a characteristic signature of fourth order gravity , [ arxiv:0812.2028 ] .
s. carloni , p. k. s. dunsby , s. capozziello , a. troisi , _ class .
* 22 * , 4839 ( 2005 )
; s. carloni and p. k. s. dunsby , _ j. phys . _ a * 40 * , 6919 ( 2007 ) ; s. carloni , a. troisi , p. k. s. dunsby , _ gen . rel .
grav . _ * 41 * , 1757 ( 2009 ) .
b. de witt , dynamical theory of groups and fields ( new york : gordon and breach , 1965 ) birrell n , davies p 1982 quantum fields in curved space ( cambridge : cambridge univ . press ) pechlaner e and sex1 r 1966 commun .
phys . 2 165 s. carloni , k. n. ananda , p. k. s. dunsby and m. e. s. abdelwahab , arxiv:0812.2211 [ astro - ph ] . n. goheer , j. a. leach and p. k. s. dunsby , _ class .
_ * 25 * , 035013 ( 2008 )
. n. goheer , r. goswami , and p. k. s. dunsby , _ class .
* 26 * , 105003 ( 2009 ) . | we construct a compact phase space for flat flrw spacetimes with standard matter described by a perfect fluid with a barotropic equation of state for general @xmath0 theories of gravity , subject to certain conditions on the function @xmath1 .
we then use this framework to study the behaviour of the phase space of universes with a non - negative ricci scalar in @xmath2 gravity .
we find a number of interesting cosmological evolutions which include the possibility of an initial unstable power - law inflationary point , followed by a curvature fluid dominated phase mimicking standard radiation , then passing through a standard matter ( cdm ) era and ultimately evolving asymptotically towards a de - sitter - like late - time accelerated phase . |
kagome lattice spin systems have been a topic of intense research for quite some time .
it is believed that the high level of frustration in these systems can give rise to exotic spin liquid phases .
a good model to look for these phases is the nearest - neighbor heisenberg antiferromagnet on the kagome lattice . in the past , many theoretical and numerical methods have been used to study such frustrated spin systems .
more recently , with the discovery of materials like volborthite and herbertsmithite there is the possibility of realizing some of these phases in actual experiments .
the ground state of the quantum heisenberg antiferromagnet on the kagome lattice is considered to be a promising candidate for realizing spin liquid phases .
however , so far the many theoretical and numerical studies still remain inconclusive .
recent numerical and theoretical studies strongly agree in favor of a gapped @xmath7 spin liquid state.@xcite .
other studies argue instead in favor of a gapless @xmath8-dirac spin liquid state.@xcite a few other studies also indicate the possibility of a valence - bond ( vbc ) type of crystalline state,@xcite while some other studies find symmetry breaking states @xcite or even a chiral spin liquid state.@xcite further , recent experiments on herbertsmithite do indicate that its ground state may indeed be a quantum spin liquid.@xcite in the presence of an external magnetic field , frustrated quantum antiferromagnets are expected to give rise to magnetization plateaus of which the plateau at @xmath9 should be most prominent and , for this reason , there has been a lot of work analyzing the properties of these plateaus . numerical works focusing on the isotropic and ising regimes of the quantum heisenberg antiferromagnet in an external magnetic field do identify magnetization plateaus at many different values .
however , in the ising regime , the simulations favor a valence bond crystal ( vbc ) type state with an enlarged unit cell based on a @xmath10 structure at these plateaus.@xcite experimentally it is difficult to observe these magnetization plateaus in herbertsmithite ( the most structurally perfect kagome compound ) since its exchange coupling is quite high , @xmath11 , which implies that it would require fields close to 200 t to be able to observe the @xmath1 plateau . as a result okamoto _
looked at volborthite ( with @xmath12 ) and vesignieite ( with @xmath13 ) and found a plateau at @xmath14 which is a little off from the expected value at @xmath9.@xcite another study did observe the @xmath9 plateau in another kagome compound ( cu - titmb ) .
however , the plateau is unstable in cu - titmb and the nearest neighbor interactions and next nearest neighbor interactions are comparable in this material,@xcite complicating the physics of the plateaus .
common approaches used to study frustrated spin systems involve either representing the spin operators in terms of slave fermions@xcite or in terms of slave bosons.@xcite these methods have been used extensively in theoretical and numerical works .
both these approaches work well at the mean - field level but suffer from the limitation that there is no small parameter about which the fluctuations can be treated in a consistent manner .
controlled calculations have been performed by generalizing the @xmath15 quantum heisenberg antiferromagnetic model to an @xmath16 or @xmath17 spin model on the kagome lattice@xcite but it is not clear if the results obtained in the large-@xmath18 limit remain valid for the experimentally relevant case of @xmath19 . here , we present an alternative approach based on a lattice chern - simons theory on a kagome lattice which implements a two - dimensional jordan - wigner transformation that maps hard - core bosons ( flipped spins of the antiferromagnet ) onto spinless fermions coupled to the chern - simons gauge fields.@xcite lattice chern - simons theories were defined for a system on a square lattice,@xcite and within a consistent lattice chern - simons approach , they were used to study the nearest - neighbor heisenberg antiferromagnet on the square lattice.@xcite chern - simons theories have been very successful in studying and explaining fractional quantum hall ( fqh ) type states .
these theories yield reliable results in gapped systems and thus may present some new insight into the problem of frustrated quantum systems . for these reasons
they have been used for quite some time to study quantum antiferromagnets on frustrated lattices ( triangular , kagome , shastry - sutherland , and others ) yielding of intriguing results of possible spin liquid phases.@xcite in these works these systems were treated only at the level of the average field approximation , and the role of the quantum fluctuations of the chern - simons gauge field were ignored .
these fluctuations are crucial to the physics of this systems .
this is a well known issue from the analogous theories of the fqh fluids where these fluctuations they play a key role in the physics of the excitations , and in particular their fractional statistics ( for a detailed recent discussion of this problem in the fqh fluids see ref .
however the existing lattice chern - simons theory@xcite ( and its more refined and consistent version by eliezer and semenoff@xcite ) can only be used for systems on a square lattice and , in particular , it can not be used for frustrated systems on non - bipartite lattices .
the chern - simons action encodes two key features : 1 ) the local constraint requiring the states to be locally gauge invariant ( in the form of a gauss - type law ) and 2 ) a definition of the canonical pairs of fields . the first condition , which for a chern - simons theory is a relation between the charge on a site and the gauge flux in an adjacent `` plaquette '' ,
must be obeyed at all sites of the lattice and not just on average .
this condition requires that the gauge fluxes on different `` plaquettes '' must commute with each other since otherwise the constraints do not commute with each other ( even through while they may still commute locally with the hamiltonian ) .
this consistency condition sets a restriction on the commutation relations of the gauge fields .
eliezer and semenoff@xcite showed how to impose these conditions consistently for the case of square lattice , at the expense of making the chern - simons action less local than would have naively expected.@xcite it is easy to see that these constraints , even in a non - bipartite lattice , can only be imposed consistently provided that there is a one - to - one correspondence between sites and `` plaquettes '' of the lattice . in this paper
we will show that this can be done for the ( non - bipartite ) kagome lattice . in a separate publication
we discuss the generalizations of this construction to more complex lattices .
however , this approach does not work for the triangular lattice , for which there are two different adjacent triangles ( `` plaquettes '' ) for each site of the lattice .
similarly , this procedure does not work for the ( bipartite ) hexagonal lattice for which two sites are associated with each hexagon .
in this paper we study the nearest - neighbor @xmath20 frustrated quantum heisenberg antiferromagnet on the kagome lattice using a generalization of the construction of the chern - simons gauge theory of eliezer and semenoff to the non - bipartite kagome lattice which we present here .
this construction is one of the main results of this paper .
the generalization of this construction for a class of frustrated 2d lattices is presented elsewhere.@xcite in this picture the flipped - spins are represented by hard - core bosons which in turn are described as a problem of fermions coupled to a chern - simons gauge field on the same lattice .
further , the effect of an external magnetic field can also be easily mimicked by adjusting the density of fermions in the equivalent problem .
more importantly , this approach will also allow us to go beyond mean - field theory and analyze the effect of fluctuations in such systems .
it is here that the effects of the consistent constriction are crucial . here
we will focus on the simpler case of the 1/3 magnetization plateau . within our approximations
we find that , in the @xmath2 limit and for a wide range of the anisotropy parameter @xmath21 , the ground state of the 1/3 magnetization plateau of the @xmath20 model on the kagome lattice is equivalent to a laughlin fractional quantum hall state of hard - core bosons with filling fraction @xmath4 .
this fully gapped state is a topological fluid with a broken time reversal invariance .
this is a state with spin currents in the ground state and with a fractional hall spin conductance , has a two - fold ground state degeneracy on the torus , has a single chiral gapless edge state on a disk geometry , and that the excitations of this state are semions .
we also found two other plateaus states , one at magnetization @xmath6 ( which is equivalent to the @xmath3 plateau ) and another one at @xmath5 .
the plateau at @xmath5 magnetization is also a topological fluid and is equivalent to the first jain daughter state of the laughlin fqh state for bosons at filling fraction @xmath6 . in this case
the state has two chiral edge states , has a three - fold ground state degeneracy on the torus and the excitations are anyons with statistical angle @xmath22 .
we also showed that , as expected , the spin - spin correlation functions decay exponentially as a function of distance while exhibiting an oscillatory behavior which reflects the breaking of time - reversal symmetry . for large enough anisotropy we find a quantum phase transition to time - reversal invariant states which extend all the way to the ising limit , where it has been shown@xcite that the ground state is governed by a quantum order - by - disorder mechanism and that it equivalent to a complex valence - bond solid . a possible time - reversal invariant @xmath7 spin liquid has been conjectured to exist at intermediate values of the anisotropy . the paper is organized as follows . in section [
sec : jordan - wigner_transformation ] , the spin-1/2 quantum heisenberg model on the kagome lattice is introduced and the jordan - wigner transformation is summarized along with the difficulties related to defining a chern - simons term on a lattice . the procedure to obtain a consistent chern - simons theory on the kagome lattice is described in subsection [ sec : cs_kagome ] .
a more detailed discussion on the lattice chern - simons theory and when such a procedure works will be presented elsewhere.@xcite t in section [ sec : mft ] , we analyze the consequences of this theory at the mean - field level and set up the saddle - point equations for the nearest - neighbor @xmath20 model after performing the jordan - wigner transformation .
the mean field theory of the magnetization plateaus is presented in section [ sec : mft ] .
subsection [ sec : xy - mft ] deals with the simpler case of the @xmath2 regime and the possible magnetization plateaus that can arise are discussed at the mean - field level where this problem reduces to a problem of interacting fermions hopping on a kagome lattice in the presence of a background statistical gauge field .
this mean field state closely resembles integer quantum hall type states , typical of the composite fermion approach to the fqh states . in subsection
[ sec : xxz - mft ] we discuss the @xmath1 , @xmath23 and @xmath24 plateaus , and analyze the full @xmath20 heisenberg antiferromagnetic model at these fillings .
the effect of fluctuations on the @xmath1 and @xmath24 plateau states are considered in section [ sec : effects_of_fluctuations ] .
we discuss the implications of the fluctuations and how they alter the mean - field physics .
the states now correspond to a _
fractional _ quantum hall type state for bosons with fractional ( spin ) hall conductivity of @xmath25 and @xmath23 respectively and are thus identified with the laughlin fqh state for bosons at @xmath26 and the first jain state ( for bosons ) at @xmath27 .
lastly , we also present asymptotic calculations of the spin correlations in the magnetization plateaus based on the effective continuum theory ion section [ sec : spin - correlations ] .
section [ sec : conclusions ] is devoted to our conclusions and some open problems .
details of the calculations are relegated to the appendix .
we discuss our results and open questions in section [ sec : conclusions ] .
in this section we briefly review the jordan - wigner transformation that maps heisenberg spins to fermions coupled to a chern - simons gauge field .
this transformation was first discussed in ref .
an important property of the chern - simons gauge field is that it imposes a constraint that relates the local density of the fermions to the flux through an adjacent plaquette of the lattice .
this flux attachment allows us to identify the spins ( which are hard - core bosons ) with fermions carrying half a flux quantum . as a result of this constraint , it is crucial to be able to define the chern - simons term in a consistent manner on the lattice so that the flux attachment may be performed at each and every site on the lattice .
the model that we will study is the nearest - neighbor @xmath20 heisenberg model on the kagome lattice in the presence of an external magnetic field @xmath28 @xmath29 - h \sum_i s^z_i \label{eq : heisenberg}\ ] ] where @xmath30 for antiferromagnetic interactions , @xmath31 ( with @xmath32 ) are the three spin-1/2 operators at lattice site @xmath33 , @xmath34 stands for nearest neighboring sites , and @xmath21 is the magnetic anisotropy parameter along the @xmath35-direction .
after the jordan - wigner transformation , the resultant action with fermions and the chern - simons gauge field becomes @xmath36 where the fermionic and interacting parts are @xmath37 \label{eq : lagrangian - free}\\ s_{\rm int}(\psi , \psi^ * ) = & \lambda j \int_t \sum_{<x , x'>}\left(\frac{1}{2 } - n(x , t ) \right ) \left ( \frac{1}{2 } - n(x',t ) \right ) \label{eq : lagrangian_fermions } \end{aligned}\ ] ] where @xmath38 is the covariant time derivative , @xmath39 is the fermion density operator ( i.e. the site occupancy ) , and @xmath40 stands for nearest neighboring sites @xmath41 and @xmath42 on the kagome lattice . under the transformation
the @xmath35 component of the spin operator , @xmath43 , is mapped to the local fermion occupation number @xmath44 we can then absorb the external magnetic field term in the hamiltonian , @xmath45 , in the definition of the chemical potential @xmath46 in eq .. hence , the effect of the external magnetic field can easily be mimicked by adjusting the fermionic filling . in the jordan - wigner transformation
the parameter @xmath47 is selected so that the statistics of the spins ( which are hard - core bosons ) are changed in to fermions
. this can be done by choosing @xmath48 for any @xmath49 .
although the hard - core boson to fermion mapping holds for all integer ( positive and negative ) values of @xmath50 , we will see below that for two special values , @xmath51 ( or , equivalently , @xmath52 ) , there is a mean field approximation with a fully gapped spectrum .
the resulting states for these two choices of @xmath47 are related to each other by time reversal and hence by a reversal of the sign of the magnetization . in order to complete the jordan - wigner transformation ,
the chern - simons term in eq needs to be specified .
if one naively , extends the continuum version of the chern - simons term to a lattice , the flux attachment constraints can not be imposed consistently as @xmath53 \neq 0 $ ] for any two sites @xmath41 and @xmath54 on the lattice.@xcite eliezer and semenoff @xcite developed a form of the chern - simons theory for a square lattice that can be consistently defined on a square lattice .
this lattice chern - simons theory was subsequently used to successfully study the ( unfrustrated ) spin-1/2 quantum heisenberg antiferromagnet on the square lattice.@xcite by generalizing the procedure outlined by eliezer and semenoff , we were able to develop a chern - simons theory that can be consistently defined on the non - bipartite kagome lattice .
this will now allow us to use the jordan - wigner mapping to study the nearest - neighbor spin-1/2 quantum heisenberg antiferromagnet on the kagome lattice .
the next section will briefly outline this procedure .
a construction of the lattice chern - simons term for more general non - bipartite lattices is presented elsewhere.@xcite we begin by writing down the below generic form for the lattice chern - simons term @xmath55 where the @xmath56 fields are defined on the sites of the lattice and the @xmath57 fields are defined on the links of the lattice .
see fig.[fig : kagome_unit_cell ] for our definitions of these gauge fields on the unit cell of the kagome lattice .
note that in eq . we have omitted the factor of @xmath47 . the first term in eq .
is the gauss law term that imposes the constraint between local density and flux through the plaquettes of the kagome lattice .
the vector kernel @xmath58 enforces the condition that relates the charge ( i.e. the site occupancy by a fermions ) to the gauge flux in the adjacent plaquette . once a gauss law has been fixed , the first term in eq .
is completely determined .
a key feature of the kagome lattice ( shared with the square lattice ) is that there is a one - to - one correspondence between sites of the lattice and plaquettes of the same lattice .
this condition is not satisfied in other planar lattices , e.g. honeycomb and triangular , which leads to flux attachment prescriptions which are ambiguous and break the symmetries of the lattice .
we elaborate more on this issue in ref . [ ] .
the structure of the matrix kernel , denoted by @xmath59 in the second term in eq . , determines the commutation relations between the different ( spatial ) gauge fields as follows @xmath60 = - i k_{ij}^{-1}(x - y ) \label{eq : acommutation}\ ] ] it is the structure of the matrix kernel @xmath59 in the second term of eq . that will allow us to consistently impose the gauss law constraints on the lattice .
this @xmath61 matrix also satisfies the condition @xmath62 .
since there are six spatial links in the unit cell , this is a @xmath63 matrix .
_ the key point is that a lattice chern - simons term can be uniquely determined by fixing a gauss law , imposing gauge invariance and demanding that the commutation relations between the @xmath64 fields are local " . _
the last condition is primarily included to obtain the simplest form of the lattice chern - simons term .
now , we will proceed by establishing these conditions on the kagome lattice and obtaining the chern - simons term .
we should note that the matrix @xmath65 of eq .
is unrelated to the so - called @xmath61-matrix that appears in the classification of abelian fqh states.@xcite the first step in writing down the chern - simons term is to determine how to impose the gauss law ( flux attachment ) on the kagome lattice .
the unit cell of the kagome lattice has three inequivalent sites ( denoted by @xmath66 , @xmath67 and @xmath68 in fig.[fig : kagome_unit_cell ] ) and three plaquettes ( two triangles , denoted by @xmath67 and @xmath68 in fig.[fig : kagome_unit_cell ] , and one hexagon , denoted by @xmath66 in fig.[fig : kagome_unit_cell ] ) .
thus , there is a natural correspondence between sites and plaquettes ( just as in the case of a square lattice ) .
it turns out that this is the crucial condition that needs to be satisfied in order to write down a chern - simons lattice term on a generic lattice.@xcite on the kagome lattice , whose unit cell is shown in fig.[fig : kagome_unit_cell ] , we define the flux through each plaquette ( defined by the associated sites @xmath66 , @xmath67 and @xmath68 ) to be @xmath69 where @xmath70 and @xmath71 are vectors along the two directions of the lattice as shown in fig.[fig : kagome_unit_cell ] .
these equations attach the flux at sub - lattice @xmath66 to that of the hexagon ( @xmath72 ) and the fluxes of sub - lattices @xmath67 and @xmath68 to each of the two corresponding triangles ( @xmath73 and @xmath74 respectively ) .
the second step in writing down the chern - simons term is to demand that eq .
is gauge invariant under the gauge transformations @xmath75 for time component of the gauge fields and @xmath76 for the space components of the gauge fields ( where @xmath77 is the difference operator on the kagome lattice ) . as an example , the field @xmath78 gets transformed as @xmath79 under a gauge transformation where the labels @xmath67 and @xmath66 again refer to the sub - lattices and the field @xmath80 lives on the sites of the lattice .
the gauge invariance condition is imposed on each of the sub - lattices and is written as @xmath81 where @xmath82 for each of the sub - lattices . the vectors @xmath83 ( introduced in eq . ) and @xmath84 are most easily written in fourier space ( for the kagome lattice in fig.[fig : kagome_unit_cell ] ) as @xmath85 and @xmath86 so far we have established the gauss law and imposed gauge invariance .
the final step is to look for a form of the matrix @xmath61 that is local " i.e. only links ( spatial gauge fields @xmath64 ) that touch one another have non - trivial commutation relations as given by eq .. this is most easily seen in fig.[fig : a1_commutation ] where the link @xmath87 is shown as an example .
the link @xmath78 has non - trivial commutation relations with just the six links that it touches and it commutes with all the other links on the lattice .
has non - trivial commutation relations with only these six links .
this is the locality " condition that is imposed.,scaledwidth=40.0% ] using the above conditions one can obtain the below unique form of the matrix @xmath61 @xmath88 where @xmath89 are _ lattice _ shift operators along the two different directions ( @xmath70 and @xmath71 ) on the lattice i.e. @xmath90 . also since @xmath91 ,
the matrix @xmath61 is invertible . the above form of the matrix @xmath61 in eq .
ensures that the fluxes commute with each other for any pair of sites @xmath41 and @xmath54 on the kagome lattice ( i.e. @xmath92 = 0 $ ] for any @xmath41 and @xmath54 ) .
this will allow us to impose the gauss law constraint consistently on each and every site of the kagome lattice .
more precisely , any two wilson loops on the lattice commute , and wilson lines on the lattice obey a signed intersection rule analogous to what happens in the continuum case of a chern - simons theory .
this completes the mapping of eq . for the nearest - neighbor @xmath20 heisenberg antiferromagnet on the kagome lattice into a system of fermions coupled to chern - simons gauge fields .
a more detailed discussion of the construction of the chern - simons theory on other non - bipartite lattices will be presented in another publication.@xcite here , we continue by looking at the consequences of this theory on the kagome lattice . in the preceding subsections we showed that a system of hard - core bosons on a kagome lattice , representing the flipped spins of the spin-1/2 anisotropic heisenberg antiferromagnet on the kagome lattice ( relative to the uniformly magnetized reference state ) , is equivalent to a system of fermions on the same kagome lattice minimally coupled to a chern - simons gauge field ( defined on the links of the kagome lattice ) .
the same result was obtained earlier on for the case of the square lattice.@xcite a consequence of this mapping ( actually , an identity ) is the operator identification @xmath93 which follows from the local ( gauss law ) constraint of the chern - simons theory , discussed in subsection [ sec : gauss ] , which relates the _ local _ fermion occupation number of lattice site @xmath94 to the flux @xmath95 on the adjoining plaquette .
a second consequence is the following formal identification of the spin raising and lowering operators @xmath96 in terms of fermion operators coupled to wilson lines of the gauge field @xmath97 where the sum in the exponent has to be interpreted as the oriented sum of gauge fields defined on the links of the lattice on an ( arbitrary ) open path @xmath98 ending at the site @xmath94 .
the operators defined on the right hand sides of eq .
are manifestly gauge invariant and square to zero , and using the commutation relations of the chern - simons gauge fields , they are found to obey bosonic commutation relations ( provided @xmath99 ) . for a system with the geometry of a disk , the exponential factors
can equivalently be rewritten in the form @xmath100 where the operator @xmath101 is a _ disorder operator _ that creates a fluxoid of strength @xmath102 at the plaquette adjoining site @xmath94 of the kagome lattice .
this is the generalized jordan - wigner transformation of ref . [ ] .
in this section we discuss the physics of the magnetization plateaus of the spin-1/2 @xmath20 quantum heisenberg antiferromagnet on the kagome lattice at the mean - field level . as already mentioned above the chern - simons term
imposes the flux attachment condition @xmath103 .
this allows us to re - write the interaction term of the action for the fermions , _ c.f . _ eq . , ( originally the @xmath104 term of the heisenberg hamiltonian ) purely in terms of the chern - simons gauge field as follows @xmath105 as a result of this substitution , the action is now quadratic in fermionic fields and the fermionic degrees of freedom can be integrated out to yield the effective action just in terms of gauge fields .
the effective action has the form @xmath106+s_{\rm
int}(a_{\mu})+\theta s_{cs}(a_{\mu } ) \label{eq : action_gauge_fields}\ ] ] where the hopping hamiltonian @xmath107 is ( in matrix notation ) @xmath108
\label{eq : hamiltonian_matrix}\ ] ] and the label @xmath109 refers to the nearest neighbors sites @xmath41 and @xmath42 on the kagome lattice .
the saddle - point equations are obtained by extremizing the action in eq .
w.r.t the gauge fields @xmath110 before writing down the expressions for the saddle point equations , we first focus on the fermionic part of the action @xmath111 .
its derivative w.r.t the temporal @xmath56 component of the gauge field gives @xmath112 similarly , for the spatial @xmath113 component one gets @xmath114
\end{aligned}\ ] ] here , @xmath115 is the ( gauge - invariant ) fermionic current .
@xmath116 is the propagator for the fermions in an average background field @xmath117 , and it is the solution of the lattice differential equation @xmath118 where @xmath119 is given in eq .. hence , the saddle point equation w.r.t @xmath120 field yields the expectation value of the local fermion density @xmath121 which amounts to imposing the flux attachment constraint on average . similarly for the field @xmath122 , one gets the mean - field equation for the expectation value of the local fermion current @xmath123 the expectation values in eq .
are explicitly given by @xmath124 where @xmath125 is the sub - lattice index and @xmath126 where @xmath127 and @xmath128 are again the shift operators as defined earlier in section [ sec : cs_kagome ] .
the full form of the saddle point - equation for the @xmath113 fields is quite cumbersome and will not be written down explicitly here . instead ,
as we are looking for time - independent / static and uniform solutions , we take the fluxes on any particular sub - lattice to be the same ( i.e. @xmath129 for any @xmath41 and @xmath54 ) , and the resulting simplified mean - field expression for the mean - field currents is @xmath130 \end{aligned } \label{eq : mf_current}\ ] ] with @xmath131 when @xmath132 and @xmath133 when @xmath134 .
the expressions for the mean - field state were derived by assuming translation invariance and allowing for each sublattice to be inequivalent from the others .
if one is looking for other types of states ( that break translational symmetry for example ) , then the mean - field expressions would have to be modified accordingly .
let us now analyze the @xmath2 heisenberg antiferromagnet @xmath135 and its magnetization plateaus at the mean - field level .
setting @xmath136 gets rid of the @xmath137 components and makes the fermions non - interacting .
however , the fermions are still coupled to the chern - simons gauge field .
further , the mean - field equation eq . is satisfied by @xmath138 in the absence of any currents .
this implies that we are just left with the flux attachment condition in eq .. at the mean - field level , we look for uniform flux states i.e. @xmath139 with @xmath140 .
this makes the total flux through the unit cell ( which has three plaquettes ) @xmath141 . by imposing the chern - simons constraint on average
, we deduce that , for uniform states and taking into account that @xmath99 , the average site occupancy ( density ) of each sublattice of the unit cell is @xmath142 .
such a state can be realized with the below choice of gauge fields @xmath143 with @xmath144 where @xmath145 and @xmath146 are the co - ordinates along the @xmath70 and @xmath71 directions respectively in fig.[fig : kagome_unit_cell ] .
the @xmath2 model has now been reduced to a problem of non - interacting fermions hopping in a lattice in the presence of a ( statistical ) magnetic field .
this is very similar to the problem of the integer quantum hall ( iqh ) effect where the one - particle states possess non - trivial chern numbers . for a square lattice
one can obtain these chern numbers by solving the resulting harper equation either numerically or by performing a perturbation theory in the hopping parameters.@xcite the final structure is most easily seen in the hofstadter spectrum as was pointed out by g. misguich et .
@xcite in their studies on the triangular and shastry - sutherland lattices using a similar analysis . by extrapolating the chern numbers from the case of the square lattice ,
one can obtain the chern numbers for the case of the kagome @xmath2 heisenberg model .
the results are shown in fig.[fig : hofstadter ] where the @xmath41-axis is the average filling / density @xmath147 on each sub - lattice and the @xmath54-axis are the single - particle energies of the associated free fermion model of the @xmath2 model . the bottom solid line indicates the fermi level for the occupied bands .
the top solid line is the next excited energy single - particle state available .
model as a function of sublattice density @xmath147 .
the numbers shown are the chern numbers of the respective filled bands . the magnetization plateaus at @xmath148 correspond to the three vertical jumps shown in this figure , respectively at densities @xmath149.,scaledwidth=50.0% ] the numbers shown in fig.[fig : hofstadter ] are the chern numbers of a state with all the below bands completely filled up to that number . in section [ sec : full_continuum_action ] we will see that we have a quantum hall type incompressible state if the total chern number of the occupied bands satisfies @xmath150 .
the discontinuous jumps in the fermi level in the figure indicate the fillings at which the chern number @xmath151 .
these situations are expected to correspond to magnetization plateaus.@xcite these jumps occur at site fillings of @xmath152 , which correspond , respectively , to @xmath153 flux through each unit cell , @xmath154 flux through each unit cell ( and hence @xmath153 flux for two unit cells ) , and @xmath155 flux per unit cell ( or @xmath156 flux in three unit cells ) . since @xmath157 we could find possible magnetization plateaus corresponding to @xmath158 and @xmath24 , with @xmath159 .
hence , at the mean - field level all these magnetization states have integer chern numbers and behave like integer quantum hall ( iqh ) states , much in the same way as with the behavior of composite fermions in the theory of the fractional quantum hall effect.@xcite in what follows we will focus primarily on the simplest case of the @xmath3 magnetization plateau and comment briefly on the other cases .
following our discussion in the previous section , we now extend the results to the case of the @xmath20 model by introducing the @xmath21 term @xmath160 in the hamiltonian .
we begin first by performing the mean - field analysis at the fillings associated with the magnetization plateaus . at the mean - field level
each unit cell satisfies the condition @xmath161 .
this density condition translates to the flux condition @xmath162 as a result of the flux attachment constraint in eq .. ( fluxes on the lattice are defined modulo @xmath153 . ) the @xmath163 fields are gauge invariant quantities and the above ansatz can be satisfied by the below choice of gauge fields @xmath164 where @xmath165 and @xmath166 are some constant parameters that will be determined below and the definitions of the links on the kagome lattice in fig.[fig : kagome_unit_cell ] are used .
this makes @xmath167 further , assuming that these ground states states have no currents i.e. @xmath168 , the second saddle point equation eq .
can be satisfied by the below choice of temporal gauge fields on each of the sublattices @xmath169 the parameters @xmath165 and @xmath166 can be computed ( numerically ) by demanding that the saddle - point equation eq .
is satisfied on each sub - lattice for a given value of the magnetic anisotropy parameter @xmath21 , i.e. @xmath170 where @xmath125 is the sub - lattice index .
the expression for the propagator in momentum space is given by @xmath171 where @xmath172 the values of @xmath165 and @xmath166 that satisfy eq .
are listed in table [ mf_consistency ] for a few different values of @xmath21 .
for the @xmath2 model , @xmath136 , the densities on all the sites are the same ( and equal to 1/3 ) . as @xmath21 is increased ( and
the interactions are turned on ) , the density @xmath173 steadily increases while the density @xmath174 decreases .
for very large anisotropy @xmath21 , @xmath175 while @xmath176 in this model . intuitively , in the ising limit ( @xmath177 ) , this corresponds to the spins either pointing strictly up or down as expected . at one - third
filling , this translates to two up spins and one down spin on average .
the fluxes on each of the plaquettes would then either be @xmath178 or @xmath153 which are equivalent on the lattice .
hence , in the ising limit , this maps to a problem of fermions hopping on the kagome lattice with no flux at the mean - field level .
+ the energy spectrum of the mean - field phases has three bands as shown in fig.[fig : mf_spec_int ] ( for two values of anisotropy parameter @xmath21 ) . at @xmath1 filling only the bottom band is filled .
this state is gapped for all values of @xmath21 except at @xmath179 when the bottom two bands cross and the low - energy fermionic states become gapless dirac fermions .
.sublattice occupation numbers @xmath173 , @xmath180 and @xmath181 , and values of the parameters @xmath182 for different values of the magnetic anisotropy @xmath21 , where @xmath183 is the @xmath2 model and @xmath177 is the ising model . [ cols="^,^,^,^",options="header " , ]
in this appendix we describe the procedure to obtain the spin - spin correlations shown in sec [ sec : spin - correlations ] . the first subsection below focuses specifically on the @xmath184 correlations .
the next subsection then deals with the @xmath2 correlations .
the fluctuating components of the @xmath185 correlation can be expressed in terms of the magnetic field from the jordan - wigner transformation as shown below .
this expectation can then be computed by introducing the usual source term @xmath186 coupled to a chern - simons term as follows @xmath187 where @xmath188 is the continuum green s function of the lagrangian in eq . and
@xmath189 is the partition function .
the @xmath190 fields can now be integrated out using the standard procedure of shifting the fields @xmath191 and then requiring that the terms linear in @xmath192 cancel .
this gives the condition @xmath193 substituting the above expression back into eq .
gives @xmath194j_{\delta } ( x ' ) \right\ } } \\
= & i \theta^2 \epsilon^{0\nu\lambda } \epsilon^{0\rho\epsilon } \partial _ { \nu } \partial^y_{\rho } g_{\lambda \epsilon}(x , y ) \end{aligned}\ ] ] the green s function can be computed in the momentum space .
hence , the fourier transform of the @xmath184 spin correlation @xmath195 can now be expressed as @xmath196 the propagator can be computed by introducing a gauge fixing term @xmath197 @xmath198 where @xmath199 .
the limit @xmath200 corresponds to the lorentz gauge ( @xmath201 ) . in this gauge
the above correlation yields @xmath202 where @xmath203 .
note that in the @xmath2 regime @xmath204 for the case of @xmath1 filling on the kagome lattice making the collective modes massive .
@xmath205 if the same calculation were performed on the square lattice ( which is unfrustrated ) , the two terms would cancel ( since @xmath206 ) as observed in the main text and the collective modes are massless corresponding to goldstone modes in the @xmath2 regime . using the jordan - wigner transformation ,
the spin - spin correlation function @xmath207 can be written as @xmath208 where @xmath209 is the fermion green function in the presence of the statistical gauge field @xmath190 and is obtained by integrating out the fermionic degrees of freedom .
the average in the last step is over the statistical gauge fields @xmath190 .
computing this expectation is involving due to the presence of the non - local string along the path @xmath210 that starts from @xmath41 and ends at @xmath54 . to simplify the above expression
, the fermionic propagator can be expressed in terms of a feynman path integral as a sum over histories of the particles as follows @xmath211}\ ] ] where the action @xmath212 is the action for non - relativistic particles coupled to the statistical gauge field @xmath213 subject to the boundary conditions for a particle traveling from @xmath54 to @xmath41 @xmath214 note that the second term in the action of eq .
corresponds to another wilson line but now traveling from point @xmath54 to point @xmath41 .
this combined with the wilson line in eq .
now creates a closed loop which we will call @xmath215 . for a problem with an energy gap and in the long - distance limit , the dominant trajectories are close to the classical trajectories . in this approximation
the average over the different trajectories @xmath216 can be pulled outside of the integral for averaging over the statistical gauge fields .
hence , eq . can now be written as @xmath217 where @xmath218 is the amplitude associated with a path @xmath215 and the set of closed curves @xmath219 are the oriented sum of paths @xmath220 and the histories of the particle .
now , the computation of the wilson loop expectation value can be performed by introducing a source term @xmath221 as follows @xmath222 where @xmath223 where @xmath224 is a unit vector tangent to the path @xmath215 at @xmath35 . in this form
the above expectation can be written as @xmath225 where @xmath188 is the green s function in the continuum for the statistical gauge fields shown above in eq .. the exponent in the above integral has two contributions , one from the maxwell like terms ( @xmath226 ) and the other from the chern - simons like terms @xmath227 .
first , the @xmath226 term can be simplified to give @xmath228 where the propagator @xmath229 can be approximated in the long - distance limit as follows @xmath230 note here that we are assuming that the above propagator is massive ( i.e @xmath231 ) as in the case of @xmath1 filling in the @xmath2 limit of the kagome lattice .
this argument would clearly fail at other fillings or for instance in the case of the square lattice when @xmath232 changing the physics all together .
hence , the integral @xmath226 can be approximated as @xmath233 where @xmath234 is the length of the path @xmath215 .
the second integral due to the chern - simons term approximates to give ( again in the long - distance limit ) @xmath235 the current @xmath236 can be regarded as an electric current . with this interpretation
the current can be related to a magnetic field @xmath237 as follows @xmath238 now the second integral @xmath227 can be written in terms of a magnetic field as @xmath239 where again at the semi - classical level we have approximated the field by the average flux @xmath240 over the path @xmath215 .
this makes the above integral take the form of an aharonov - bohm phase over the path @xmath215 . in the last step
we have used the fact that @xmath241 and @xmath242 ( in units of @xmath243 ) .
hence , the correlation can be approximated as @xmath244 where @xmath226 decays exponentially as the length of the path @xmath234 increases and @xmath227 is an aharonov - bohm phase associated with the path @xmath215 .
the aharonov - bohm term in the above expression would depend on the two points @xmath41 and @xmath54 and the area enclosed by the path @xmath215 . in the main text we have considered two possible situations , one where the area of the path @xmath215 is zero and does nt lead to any aharonov - bohm phase and the other where there are several different aharonov - bohm phases that can arise . | frustrated spin systems on kagome lattices have long been considered to be a promising candidate for realizing exotic spin liquid phases . recently , there has been a lot of renewed interest in these systems with the discovery of materials such as volborthite and herbertsmithite that have kagome like structures . in the presence of an external magnetic field
, these frustrated systems can give rise to magnetization plateaus of which the plateau at @xmath0 is considered to be the most prominent . here we study the problem of the antiferromagnetic spin-1/2 quantum xxz heisenberg model on a kagome lattice by using a jordan - wigner transformation that maps the spins onto a problem of fermions coupled to a chern - simons gauge field .
this mapping relies on being able to define a consistent chern - simons term on the lattice .
such a lattice chern - simons term had previously only been written down for the square lattice and was used to successfully study the unfrustrated heisenberg antiferromagnet on the square lattice . at a mean - field level , these ideas have also been applied to frustrated systems by ignoring the details of the chern - simons term .
however , fluctuations are generally strong in these models and are expected to affect the mean - field physics . using a recently developed method to rigorously extend the chern - simons term to the frustrated kagome lattice we can now formalize the jordan - wigner transformation on the kagome lattice .
we then discuss the possible phases that can arise at the mean - field level from this mapping and focus specifically on the case of @xmath1-filling ( @xmath0 plateau ) and analyze the effects of fluctuations in our theory .
we show that in the regime of @xmath2 anisotropy the ground state at the @xmath3 plateau is equivalent to a bosonic fractional quantum hall laughlin state with filling fraction @xmath4 and that at the @xmath5 plateau it is equivalent to the first bosonic jain daughter state at filling fraction @xmath6 . |
our wigner islands are constituted by millimetric stainless steel balls ( of diameter @xmath13 mm and weight @xmath14 mg ) located on the bottom electrode of a horizontal plane capacitor ( a doped silicon wafer whereas the top electrode is a transparent conducting glass ) .
an isolated metallic circular frame of diameter @xmath15 mm and height @xmath16 mm intercalated between the two electrodes confines the balls @xcite . when a potential @xmath17 is applied to the top electrode ( the bottom one and the frame being linked to the ground )
, the balls become monodispersely charged , repel each other and spread throughout the whole available space .
for the currently used potential @xmath17 ( around a few hundred of volts ) , the charge of each ball has been evaluated to about @xmath18 electrons .
the whole cell is fixed on a plate linked with two independent loudspeakers supplied by a white noise voltage . as we shall show later , this horizontal shaking results in an erratic movement of the balls which simulates an effective temperature that can be modified by tuning the shaking amplitude . in the experiment presented here ,
the system is always prepared in its ground configuration at low effective temperature and the temperature is progressively increased . throughout the experiment ,
images of the arrays of balls are recorded in real - time using a ccd camera , and their center of mass is detected .
for each temperature , the positions of each ball are followed during 400 s , the minimum time between two snapshots being 100 ms .
the characteristic time of the oscillation of a single ball in its local potential is , at low temperature , about half a second .
then , the choice of the total experiment time and the time interval allow us to track the balls ( at least in the considered temperature range ) and get statistically relevant data .
the different configurations reached are then determined by counting the number of balls in each shell . in this procedure
the radial limits of a shell are defined as the mean values between its radius and its neighbor s one , that have been measured in the ground state .
they are independent from the temperature . before studying the `` melting '' of such systems ,
strong attention has been paid to show that the cell shaking effectively results in a brownian motion of the balls allowing the identification of this shaking with an effective temperature .
we present here the experiments performed in order to calibrate this effective temperature and the in situ " thermometer we have developed . as in @xcite
, we used a system for which the energy is well - known : the calibration was obtained by the use of a single ball rolling on the silicon wafer which has been winded with an angle @xmath19 from the horizontal plane .
the ball can elastically bounce on a bottom wall .
no electrostatic force is at stake and the only energy is the gravitational potential .
for each voltage @xmath20 applied on the loudspeakers , the horizontal distances @xmath21 of the ball center from the bottom wall have been measured through the capture of a few thousand snapshots of the ball .
the recorded random @xmath21 positions are distributed following the density @xmath22 . in order to compare this density with bolzmann theory ,
it was fitted with the function @xmath23 where @xmath24 is the potential energy of the ball and @xmath25 the fit parameter which corresponds to the expected effective temperature , @xmath26 being the boltzmann constant .
figure [ temperature](a ) presents the experimental data obtained for the voltage amplitude @xmath27 a.u . and the corresponding fit .
at evidence boltzmann law is obeyed .
this very good agreement being observed whatever the voltages @xmath20 , we may conclude that the mechanical shaking corresponds to an effective temperature .
this effective temperature is obtained thanks to the fitting analysis , or more simply through the relationship @xmath28 and a calibration curve @xmath29 has then been be obtained . as shown in figure [ temperature](b ) , the relation between @xmath20 and @xmath25 is affine within the range required to study the melting " .
let us indicate that the temperature range is about @xmath30k , which has no other signification than the energy range .
finally , we have used this calibration to develop anin - situ " thermometer which is constituted by a single ball trapped in a second circular frame located near the main one and submitted to the same voltage .
the effective temperature is determined by measuring the radial mean square displacement @xmath31 of this unique ball for each given shaking amplitude @xmath20 and by identifying this displacement with the temperature t(a ) previously determined by the calibration procedure .
whatever the various thermometer diameters tested in order to optimize the sensibility of this thermometer ( @xmath32 and 10 mm ) and the applied potential @xmath17 ( @xmath33 v ) the mean square displacement @xmath31 varies linearly with the temperature .
same linearity is observed for the mean square speed .
figure [ excursionbilleseule ] presents the variation @xmath31 ( t ) corresponding to the potential @xmath34 v and the retained diameter @xmath35 mm
. this thermometer will give a precise determination of the temperature for any experiment to come , independently from the variations associated to the total weight of the support or to the loudspeakers ageing .
the bond - orientational correlation function @xmath36 is relevant in order to characterize phase transitions in two - dimensional systems .
it is defined by @xmath37 where @xmath38 is the angle of a fixed bond between two particles and @xmath39 denotes an averaging over all bonds . in infinite two - dimensional systems
, @xmath40 tends to a constant roughly equal to 1 at low temperature .
the system is then like an ordered crystal .
if the temperature is higher than a temperature @xmath41 , a strong decay with time is observed , which denotes a liquid phase .
such a dynamical criterion offers experimental facilities since it is often easier to have long time acquisition rather than to observe a large sample of particles .
in particular , we can readily use this criterion to characterize order - disorder transitions in small systems .
note that the factor 6 in the exponential is adapted for hexagonal lattices , therefore we can expect the limit value at low temperature to be lower than 1 for our small systems in which the three - fold symmetry is broken . on the other hand , the averaging being made only on @xmath42 links , we ca nt expect @xmath40 to go until zero value in the liquid phase , even at large time .
these points apart , this parameter still measures the correlations and we can expect to measure the same global behaviors according to the temperature . two parameters easy to get experimentally have been identified in order to characterize the configuration transitions .
the first of them is the jump rate @xmath43 : @xmath44 where @xmath45 is the total number of configuration switches during the period @xmath46 .
this parameter is an indicator of the transition activity " of the system .
qualitatively , this parameter is small at low temperature when only a few transitions occurs and increases strongly with temperature when energetical barriers can be overcome . a second way to characterize more quantitatively
the transition rate is to measure the mean time required to escape from each state or the mean residence time in each state .
let us consider for instance a wigner island at low temperature ; this system can be understood as a two level system characterized by a ground state @xmath47 and one metastable state @xmath48 ( the only one that is reachable if the temperature is sufficiently low ) .
the thermal fluctuations induce transitions between those two configurations .
these transitions in the real space can be mapped in the phase space by the jump of a fictive particle from a well to another , characterized by their depth @xmath47 and @xmath48 and a saddle point @xmath49 . under the assumption that all escape attempts are independent and of weak probability ,
the probability for the particle to stay in a state of energy @xmath50 during a time @xmath51 is : @xmath52 where @xmath53 is the mean residence or kramers time .
it depends on the energy barrier and on the temperature and is given by @xcite : @xmath54 where @xmath55 is the barrier energy , and @xmath56 is the relaxation time within the well .
the energetical barriers can be measured by the slope of the curve describing the variations of the mean residence times in log scale as a function of 1/t .
similarly , the spectrum of the excitation energies is determined through the ratio of mean residence times in the two wells .
qualitatively , this analysis can be extended to the case of high temperature for which the system explores more than the two first levels and reaches higher excited levels .
but the possibility for a particle to escape a well in different ways whose relative weight should depends on the geometry leads to a more acute problem .
we have tested the validity of this analysis in the case of the two - level systems @xmath57 and @xmath58 .
these two cases are interesting since they involve in their respective configuration transitions the two kinds of individual jump observed in the wigner excitations . for @xmath57 ,
the ground configuration consists in a unique shell ( 5 ) whereas the metastable configuration ( 1 - 4 ) requires a centripetal displacement of a ball .
by contrast , the @xmath58 ground state is a centered configuration ( 1 - 5 ) and the metastable configuration ( 6 ) is reached after a centrifugal displacement of ball . in figure [ temps56 ]
we present the variation with the temperature of the ratio @xmath59 , where @xmath60 stands for the residence time in the centered configuration , respectively ( 1 - 4 ) or ( 1 - 5 ) , and @xmath61 denotes the residence time in the one - shell configuration , respectively ( 5 ) or ( 6 ) . for the two selected systems
these variations obey kramers relation . according to this boltzmanian description
, we will characterize in the following the configurations transitions by @xmath43 and by the mean residence times in each configuration .
these parameters depend on the configuration energy spectra and thus , are extremely sensitive to the local symmetry of the configurations . in order to explore the collective excitations of wigner islands through a lindemann criterion ,
the mean square deviations of the balls from their equilibrium locations have to be calculated . for three - shell configurations and circular symmetry @xcite
, we have to calculate , for each configuration and for shells 1 and 2 , the radial displacements @xmath62 the relative angular intrashell displacements @xmath63 and , for each configuration and for shell 1 , the relative angular intershell displacements @xmath64 where @xmath65 are the polar coordinates of a ball with respect to the center of the confining frame , @xmath66 indicates the right neighbor of the ball @xmath67 , and @xmath68 indicates its nearest neighbor in the surrounding shell , which is determined every snapshot .
@xmath69 , where @xmath70 is the balls density , is the mean radial free space for a ball and @xmath71 is the mean interball angular distance in the shell consisting in @xmath72 balls .
@xmath39 indicates an average over time .
these deviations are relative displacements ( with respect to another ball or the center of the system ) , therefore they are relevant in order to exhibit a lindemann - like criterion comparable to the one used for larger two - dimensional systems @xcite
. however , we will not be able to discuss as in the latter article about the values of the dimensionless parameter @xmath73 , where @xmath50 is a typical interaction energy between two particles , since we do not have numerical values for the potential energies in our system at the present time .
note that we have discriminated between the different configurations .
this is allowed since the radial displacements , as we shall show , are around a sixth of the distance between the two shells whatever the effective temperature , then the shells are always well identified .
this choice is motivated by our requirements of precise information about the influence of the local order on the `` melting '' .
the temperature dependencies of the displacements averaged over all the configurations have also been calculated .
this procedure is not the same as the one used in @xcite , where the distinction between shells is only made at the beginning of the numerical simulation , so we shall expect lower values for our radial displacements that do not include intershell jumps .
on figure [ correlationsauts](a - c ) we present typical time correlation functions @xmath36 for five different temperatures and for the three systems , for which the same behaviors are observed : at low temperature , a constant value is reached , indicating an ordered state . at higher temperature , the correlations decay strongly . as in large systems , we can define a `` melting '' temperature @xmath74 .
we notice that the @xmath9 system keeps an ordered structure denoted by a quasi - constant correlation function until a much higher temperature than the two other systems .
indeed , the transition temperatures @xmath74 are sensitively different according to the number of balls @xmath1 : @xmath75 is less than @xmath76 , whereas @xmath77 is more than @xmath78 and @xmath79 between both .
note our goal is not to measure precise transition temperatures , but to put in evidence the mechanisms involved in the disordering .
as we shall see , the different transition temperatures are sufficiently separated to do such an analysis . in the following
, we will then investigate the two expected mechanisms for this disordering : configuration transitions and collective excitations through lindemann criterion .
the first indication of the configuration transition activity can be evidenced by observing the evolution of the jump rate @xmath43 while the effective temperature increases .
these variations are shown on figure [ correlationsauts](d ) for the different studied systems .
whatever the number of balls , @xmath43 presents the same qualitative behavior : it is close to zero at low temperature and increases strongly at higher temperature .
these variations correspond to a progressive augmentation of the number of transitions activated at the effective temperature .
whatever the temperature , the smallest jump rate is associated to @xmath80 system and the one associated to @xmath81 is always smaller than the one corresponding to @xmath82 . in order to describe more quantitatively these behaviors and their differences ,
let us introduce transition temperatures which characterize the beginning " of the @xmath43 increases .
we have chosen to name transition temperature the temperature at which the ground configuration begin to switch .
let us nevertheless indicate that this transition temperatures does not correspond to an actual sudden transition since the jump rate @xmath43 rises progressively .
note that experimentally , the infinite time limit in the @xmath43 definition corresponds to the maximum measurement time and its finite value could alter the @xmath43 value .
however a complete analysis has shown that the variation with temperature of @xmath43 is independent from this experimental limit provided that this limit was largely higher than the mean residence time of the system in each configuration . in our experiments we chose @xmath83s which satisfied this condition .
the analysis of the evolution of the residence times distribution shows that , for @xmath9 , the transition temperature @xmath84 is @xmath85k , much higher than the temperatures @xmath86 and @xmath87 which are respectively @xmath88k and @xmath89k . at larger temperature ,
the second excited states are reached .
the corresponding temperatures are respectively @xmath90k , @xmath91k and @xmath90k for @xmath9 , 18 and 20 and are indicated by arrows on figure [ correlationsauts](d ) .
those temperatures might correspond to the critical temperatures given in @xcite where the rate of radial jumps is calculated .
this is exactly the same parameter as our jump rate since each configuration switch involves only one ball jump from one shell to another .
these behavior differences between the systems are still more obvious when we study their mean residence times in the ground state . on figure [ tempsfond ]
we have plotted the logarithm of these times @xmath92 versus the inverse of the effective temperature .
we obtain a linear variation , which is in agreement with the boltzmann law description presented above , the curves slopes being equal to the barrier heights for escaping the configuration .
we can observe that these slopes are different according to the number of balls .
the highest slope is associated to the ground state @xmath9 and is equal to @xmath93 k , those corresponding to @xmath82 and 20 being respectively equal to @xmath94k and @xmath95k .
this indicates that the ground state for @xmath96 is much more stable than the ground states associated to @xmath82 and 20 , whose barrier heights are very similar since their corresponding slopes are roughly identical .
this analysis can be completed by studying the excited states ( figure [ tempsexc ] ) .
the slopes for @xmath82 is @xmath97k and is equal to @xmath98k for @xmath9 ; in both systems , this suggests that the barrier is much lower than the ground state s one . on the contrary ,
the slope associated to the first excited state of @xmath81 is equal to @xmath99k , almost the same as the barrier for the ground state .
all these results are in good agreement with the energy levels calculated in @xcite .
let us indicate that even when they are reachable , the other excited states are not easy to study since their higher energy involves too small statistical occurrences .
finally , these measurements allow us to determine the first excitation energy associated to each system .
these energies are @xmath100k , @xmath101k , @xmath102k respectively for @xmath103 , 19 , 20 .
these measurements show that the magic number " system corresponds to the deeper ground state and confirms its strong stability in comparison with the two other systems .
as expected , the commensurability influences strongly the depth of the well , and consequently the transition temperature in @xmath43 .
let us conclude this section by indicating that obviously the jump rate @xmath43 is related to the mean residence times .
for instance , for a two - level system , @xmath43 is simply equal to @xmath104 , where the subscripts 1 and 2 stand for the two levels .
this relation is satisfied and confirms the self consistency of our results , at least up to a temperature at which a third level is reached .
we now turn to the study of the `` melting '' through lindemann - like criterion . we will successively present the radial , intrashell and intershell mean square displacements averaged over all the configurations . in order to explore more precisely the relation between the local order and these collective parameters ,
we have compared them to the corresponding parameters before averaging .
we shall show in particular that the procedure of averaging over the configurations , commonly used in literature , mask actually subtle effects resulting from the configurations transitions , even though they do nt infer that much on transition temperatures .
the temperature dependencies of the radial mean displacements averaged over all the configurations are presented for the three different numbers of balls on figure [ exc](a ) .
the displacements corresponding to the inner shell ( shell 1 ) present similar behaviors for the three systems .
they vary regularly from 0.005 at low temperature and go on increasing until 0.05 at the highest experimental temperature , the highest and the lowest radial displacements being respectively associated to the systems @xmath81 and @xmath9 , whatever the temperature . in this experimental temperature range ,
the low values of @xmath105 , smaller than the lindemann criterion , indicate that each shell remains very well identified and that we can always discuss qualitatively the results in terms of shell entities .
notice that the highest obtained temperature corresponds to the ball - tracking limit , but we could forecast that the value would increase strongly beyond 0.05 after this temperature limit , since this value is far from being the maximal possible value , even though jumps from one shell to another are not taken into account .
from this point of view , we can consider the highest experimental temperature is very close to the radial transition temperature transition @xmath7 of these systems .
the radial displacements of the outer shell ( shell 2 ) are qualitatively similar to those observed for the shell 1 although their values are smaller and vary from 0.005 until 0.03 .
this can be explained by the fact that this shell is submitted to a regular and constant potential resulting from the confinement frame whereas shell 1 is submitted to fluctuant potential from both sides @xcite .
we have seen that geometrical considerations are responsible for the different configuration transition rates observed .
it is therefore natural to examine if it is the same for the displacements .
the radial mean displacements for the different configurations of the different systems are presented on figure [ excrad ] .
notice that data for the excited states with low residence time present higher statistical error due to the smaller number of their occurrences ; this is the case for instance for the second excited state for 18 and 20 balls and for both excited states for 19 balls . as for the averaged curves , the radial displacements in the two shells of the systems in their ground configurations present regular increases with temperature .
more precisely , for shell 1 , we can observe that the @xmath9 ( 1/6/12 ) and @xmath81 ( 1/6/13 ) curves are identical whereas the displacement corresponding to @xmath82 ( 1 - 6 - 11 ) is higher whatever the temperature . for shell 2 the displacements
are identical for @xmath82 ( 1/6/11 ) and @xmath9 ( 1/6/12 ) whereas those associated to @xmath81 ( 1 - 6 - 13 ) presents a higher value .
the temperature dependencies observed for the excited states look like those associated to the ground states , with a constant switch .
the highest @xmath105 values are now observed for @xmath81 ( 1 - 7 - 12 ) in the case of the shell 1 and for @xmath82 ( 1 - 5 - 12 ) in the case of the shell 2 .
we can not define a precise rule to explain the different behaviors .
it seems however that , for a given temperature , the more balls in the shell , the larger the radial displacements . in the case of equality
( that s to say , for the first shell , the couples ( 1 - 6 - 11),(0 - 6 - 12 ) and ( 1 - 6 - 13),(2 - 6 - 12 ) and for the second shell the couples ( 1 - 5 - 12),(0 - 6 - 12 ) and ( 1 - 7 - 12),(2 - 6 - 12 ) ) , we observe that the commensurable states have lower radial displacements . those differences are higher than a possible distortion due to a renormalization by a different @xmath72 . the first statement can be simply understood by the fact that orthoradialy squeezed balls have more kinetic energy to involve in their radial movement . in the case of a commensurable state , all the balls ( for the first shell ) or half of them ( for the second shell ) find themselves in front of a repelling ball of the other shell , which restrains their radial fluctuations . note this result is opposite to what was reported for quantum dots in @xcite : the authors exhibit higher radial displacements for magic number @xmath9 . beyond these specific results ,
the comparison between the averaged and non averaged radial displacements show that even if some differences can be exhibited concerning their precise and relative values , global behavior remains the same : each shell will radialy melt at the same temperature whatever the configuration . in the case of
the angular mean displacements , the averaging does not introduce any distortion in the analysis , as it can be seen on figure [ excort ] .
so we will only discuss the mean displacements averaged over the configurations ( figures [ exc](c ) and ( d ) ) . like the radial displacements , the intrashell orthoradial displacements associated with the inner shell of ground configurations increase with the temperature .
however , whereas the former keep on rising slowly , the angular ones begin to vary linearly and change very rapidly at almost the same temperature whatever the number of balls .
by contrast , this similarity of behaviors is not observed for the outer shell : whereas the intra - shell orthoradial displacements are identical for @xmath82 and @xmath81 , in the case @xmath9 , it remains smaller than 0.005 and without rapid rise .
the changes are observed when the displacements reach the critical value @xmath106 in accordance with lindemann criterion .
thus , we can define from these data a transition temperature @xmath6 or , rather , a temperature interval centered temperature on @xmath6 ( between @xmath107k and @xmath108k ) after which intra shell orientational order is lost in a given configuration for @xmath82 and @xmath81 .
we can expect that the corresponding temperature is not far from the experimental limit in the case of @xmath9 since the balls in the shell 1 have begun to be non - correlated whereas the shell 2 is still rigid .
let us now consider the intershell relative displacements which measures the ability of the two shells to find a stable position one with respect to the other .
their variations with temperature averaged over configurations are presented in figure [ exc](b ) for the three systems .
we can notice that even at low temperature these displacements are high and until @xmath109k , the intershell angular displacements are larger than the corresponding intrashell and radial displacements .
this indicates that at low temperature these intershell movements are the main effects resulting from the thermal fluctuations .
moreover , this temperature being smaller than @xmath110 and @xmath7 , we are allowed to discuss these results in terms of well defined rigid shells in which the balls are regularly spaced . at large temperature , @xmath111k
, the inter - shell angular displacements reach the same finite value whatever the system , this value being in fact its theoretical maximum . in this temperature range ,
the thermal energy is sufficient to overcome the barrier energies , which correlate the inner and outer shells .
since disorder is characterized by a deviation of the whole island from the symmetrical situation , note that the intershell displacements are calculated considering the nearest neighbor at each time step , in order to take into account the invariance towards some rotations .
consequently , we will always have a small maximum value and will never observe a steep rise as in @xcite .
below this temperature , a temperature range in which the effect of the local order has an essential play is well exhibited .
in this temperature domain , the systems are mainly in their ground state and the intershell angular displacement is lower for 19 balls than for 18 and 20 balls , the latter remaining of the same order .
this can be simply explained by commensurability arguments : let us consider the shells as rigid rings ; in the 1 - 6 - 12 configuration , shell 1 is submitted to a @xmath112 periodic potential due to shell 2 , then each of its balls can find itself in a potential well .
on the contrary , in any other configuration uncommensurability implies that , if rigid , the two shells ca nt find a position for which every ball will be located in a minimum of energy , hence a higher instability .
the study of the transition temperatures @xmath113 through correlation functions show that they depend strongly on the local order .
indeed , those temperatures are very different for the three systems , the @xmath9 `` magic system '' being more stable . as for temperatures around @xmath113 ,
the systems can still be seen as sets of shells , comparison can be made between the temperatures @xmath41 and the temperatures @xmath114 of first configuration transition .
since they are very similar , we can infer that configuration transitions play an important role in the disordering , at least for intermediate temperatures .
it can be simply understood by the fact that the islands are invariant by many rotations whatever the configuration , then a cycle of configuration transitions can induce disorder ( from the correlation point of view ) , since the initial particles positions are not necessary recovered at the end of the transition cycle . moreover a configuration switch implies a complete reorganization in the shells .
this disorder mechanism involving only one particle jump can not be evidenced by the global parameters as the mean displacements used in the lindemann approach . note that even at low temperature , the orientational order between adjacent shells is already lost while their internal order is conserved ; this suggests that the trajectory of the jumping particle might be not only radial but also orthoradial , taking advantage of the relative rotation of the two implied shells .
this is also in accordance with the fact that commensurability considerations play a role for the intershell rotations as well as for the height of the energetical barriers . on the other hand ,
mere intershell rotation is not sufficient to induce disorder as defined by the correlation function , since the system will periodically find itself in the same position as the initial one .
if we focus now on collective displacements , a first transition occurs at the temperature @xmath115 , which corresponds to the emergence of the angular intra - shell diffusion .
this stage in which the systems can be considered as independent shells remains until a second transition temperature @xmath7 involving diffusion of particles between the shells . at higher temperatures ,
the shell structure disappears .
figure [ cm ] shows the mean crystals ( superposition of the different positions of the particles ) obtained for instance for @xmath82 at three key temperatures .
figure [ cm](b ) illustrates the role of the transition activity in the loss of the orientational order : according to the values of the angular displacements , particles positions should be distinguishable .
this kind of two - stage `` melting '' had been suggested by previously cited works , especially in @xcite , where the authors focus on a global lindemann criterion that includes different mecanisms that we have exhibited there . in particular , we have distinguished two contributions in the radial displacements , namely the individual jumps and the mean behavior . in colloidal systems
@xcite@xcite the system is arranged at low temperature in a shell - like structure .
it also exhibits a very similar behavior when the temperature increases , excepted a re - entrant ordered phase which was not observed in our case , this phase being specific to the hard wall confinement @xcite . in @xcite
, the authors studied the melting of @xmath9 and @xmath81 quantum particles interacting with coulombic interaction .
they described the melting as a two stages process : first an orientational inter and intrashell disordering and then a radial melting at higher temperatures .
the relative positions of the transition temperatures found here are in good qualitative agreement with their results .
in addition , we can clearly distinguish two phases in the angular disordering , namely an intershell rotation and then an intrashell melting .
this last point is also presented in numerical works on classical coulombic particles @xcite . whereas local geometry have an influence on the configuration transitions through commensurability considerations
, we have shown that its effects are neglectable for the intrashell displacements as well as for the radial ones .
on the other hand , correlation functions define very different temperatures for our three systems ( all the transition temperatures are summarized in table [ table ] ) . moreover , and contrary to large systems , those temperatures are lower than the transition temperatures given by the lindemann criterion .
we have then to consider that beyond the well - known lindemann scenario , there are other sources of disorder , namely the configuration switches . .[table]summary of the different transition temperatures ( in @xmath116 ) .
[ cols="^,^,^,^,^",options="header " , ]
in this paper we show that for a system constituted of a small number of interacting particles like wigner islands , an increase of temperature results in a disordering of the system more than a real melting .
this disordering process is very sensitive to the local order of the explored configurations .
this disordering results from both individual excitations that induce configuration transitions and collective excitations .
this process is marked by three different transitions .
the temporal correlation functions which describe the correlation loss of the system exhibits an exponential decrease at the `` liquid transition temperature '' @xmath113 ( named after the usual convention ) .
this first transition is identified without ambiguity as corresponding to the increase of configuration transitions between the stable and metastable states of the system .
so the liquid transition temperature depends strongly on the local order as the transition rate . at larger temperatures , two other transitions
appear , @xmath6 and @xmath7 characterizing respectively intrashell and intershell diffusion .
these transitions are evidenced by the change of the mean square displacements with temperature and correspond to collective excitations .
the local order play a less significant role in the transition temperature values .
this disordering process and the importance of the local order on these temperatures are due the small number of configuration states explored by the system at a fixed temperature . from this point of view , it is a specific characteristic of small systems .
indeed , these effects can not be observed for large systems since the number of explored configurations is large enough to mask the individual excitations in the collective ones , hence the coincidence between the transition temperatures described by the correlation functions and those identified by the lindemann criterion . whereas the @xmath113 transition was never discussed before ,
the two following successive transitions have previously been mentioned in the literature .
the corresponding analyses are in agreement with our results but their approach is very different . we show that the description of the transition from well organized arrays towards liquid state resulting from successive excitations requires more detailed analyses than the single use of lindemann criterion . finally , let us conclude by suggesting that small wigner islands that we proposed could be good candidates in order to explore experimentally the thermodynamic laws dedicated to small systems of interacting particles@xcite . | the influence of local order on the disordering scenario of small wigner islands is discussed .
a first disordering step is put in evidence by the time correlation functions and is linked to individual excitations resulting in configuration transitions , which are very sensitive to the local symmetries .
this is followed by two other transitions , corresponding to orthoradial and radial diffusion , for which both individual and collective excitations play a significant role .
finally , we show that , contrary to large systems , the focus that is commonly made on collective excitations for such small systems through the lindemann criterion has to be made carefully in order to clearly identify the relative contributions in the whole disordering process .
many efforts have been intended in past years to understand properties of mesoscopic devices in which interacting particles are confined .
for instance , these particles can be vortices in mesoscopic shaped superconductors @xcite@xcite , electrons in quantum dots @xcite , strongly coupled rf dusty plasma @xcite , trapped cooled ions @xcite , vortices in superfluid @xmath0 @xcite , electron dimples on a liquid helium surface @xcite , vortices in a bose - einstein condensate @xcite or colloidal particles @xcite .
more recently , in order to take advantage of the macroscopic scale to explore the properties of such systems , we have proposed a new macroscopic system consisting of @xmath1 interacting charged balls of millimetric size free to move on a plane conductor and confined electrostatically , the temperature being simulated by a mechanical shaking @xcite . using this system
we observed the equilibrium configurations of the wigner islands obtained for circular @xcite and elliptic confinements @xcite .
our previous study , essentially focused on static properties , conclude that at low temperature and for a circular confining potential , the observed small islands @xmath2 present self - organized patterns constituted by concentric shells in which the balls are located . as it was widely discussed in the literature
@xcite@xcite@xcite , this peculiar structure is due to the competition between the ordering into a triangular lattice symmetry , which appears for infinite two - dimensional electrostatic systems , and the circular symmetry imposed by the confining potential . in the following
, the wigner islands will be described by the configuration @xmath3 where @xmath4 is the number of balls in the i^th^ shell from the center .
surprisingly , the influence of the thermal fluctuations on the phase behavior which was an important question largely studied for the two - dimensional extended systems has been little discussed in such small systems .
the studies devoted to this question are essentially numerical and focused on the temporal stabilities of the different configurations @xcite@xcite , their spectral properties @xcite@xcite or the mean displacements of the particles @xcite .
let us however indicate an interesting experimental work concerning themelting " process of colloidal interacting particles islands @[email protected] " in two - dimensional quantum electron clusters has also been recently studied in @xcite .
phase transitions in two - dimensional large crystals have mainly been described by the changes in the asymptotic behavior of spatial correlation functions .
for instance , the kosterlitz - thouless - halperin - nelson - young ( kthny ) theory predicts a two - step melting scenario according to which the liquid phase is reached when bond - orientational correlations become short - range @xcite@xcite .
parallely , it has been suggested that temporal correlations should have the same behavior as the spatial ones @xcite .
this has recently been put in evidence experimentally for colloidal systems @xcite@xcite . on the other hand ,
the previous studies on small interacting systems always refer to a generalization of the well - known lindemann s method employed to describe the order - disorder transition for large systems @xcite , which considers the liquid phase is reached when the mean square displacements relatively to the lattice parameter of the particles go beyond the value @xmath5 ( which seems to be independent of the interaction @xcite@xcite ) , or 0.05 for each coordinate .
note that it has been shown that , for infinite systems , the transition temperature exhibited there is the same as the one given by the correlation functions @xcite .
let us underline that no such result is known for small systems . for the latter , the studies lying on lindemann s criterion predict or observe that the shell - structured islands become less ordered while the temperature increases .
they describe a two - step process corresponding to two different transition temperatures . at very low temperature ,
each particle is thermally excited in its local potential ; a first transition appears at the temperature @xmath6 when the orientational order between the shells is lost .
this first transition is followed by a second one , at the temperature @xmath7 , which corresponds to the emergence of the radial diffusion of particles between the shells , as well as an angular diffusion in the shells .
for higher temperature , the initial order is completely destroyed .
the transition temperatures @xmath6 and @xmath7 are respectively identified as the temperatures at which the intershell angular and radial mean square displacements have a rapid and strong increase @xcite . in this paper , we discuss the influence of the local order on the `` melting '' . indeed , for small islands , singular events such as a unique jump of only one particle from one shell to another one result to an important modification of the system since such an intershell jump corresponds to a transition between well - separated stable and metastable equilibrium configurations of different geometry .
consequently , such jumps will modify the bond - orientational correlation functions .
we shall show that , for a given temperature , those configurations switches and thus the disordering process are controlled by the local geometry . in order to distinguish between those events and the collective excitations of the particles , we calculate the lindemann parameters on each shell and for each configuration separately , without taking into account jumps between shells .
we determine the temperatures at which the collective excitations appear and destroy the ordered configurations .
this unusual lindemann procedure is in agreement with our definition of an ordered system as a system always close to its most symmetrical configuration .
for instance , an intershell jump which does not destroy the local symmetry does not have to be considered as relevant in our description .
this will explain why the obtained critical temperatures will be higher than the one determined by the usual lindemann method which includes all kind of displacements in the calculation of the parameters , as in @xcite .
notice that this question is specific to small systems since in large systems there is a continuum of states and individual excitations are hidden by collective ones . in order to evaluate the relative importance of these different contributions to the `` melting ''
, we have experimentally observed the evolution of macroscopic wigner crystals while the effective temperature is increased . to emphasize the contribution of the individual excitations with respect to the collective ones
, we have selected a set of systems which have very different local symmetry for a similar number of balls in order to present different configuration transition behaviors for almost the same kind of collective excitations .
we chose wigner islands consisting in @xmath8 and 20 interacting particles confined in a circular frame .
these systems , in spite of a very close number of balls , are very different from the local symmetry point of view and the resulting excitation energy spectra .
indeed , the magic number " system @xmath9 exhibits a three fold symmetry , as its ground configuration is ( 1 - 6 - 12 ) .
in fact , the latter shell can be divided into two subshells of 6 balls , as it was numerically shown in @xcite , however , the difference between the two radii being rapidly of the same order as the thermally induced radial fluctuations , we will still refer to this configuration as the ( 1 - 6 - 12 ) one .
by contrast , the ground configurations of @xmath10 and @xmath11 systems are constituted by incommensurable shells ( in the sense that the ratio between the number of balls on each shell is not an integer ) , their ground states being respectively ( 1 - 6 - 11 ) and ( 1 - 6 - 13 ) .
when the temperature increases , the two lower excited states of each system can be reached . in spite of the radial displacements due to the temperature , the ring - like structure remains and the configurations can be very well identified : * for 18 balls : ( 1 - 6 - 11 ) , ( 1 - 5 - 12 ) , ( 0 - 6 - 12 ) * for 19 balls : ( 1 - 6 - 12 ) , ( 1 - 7 - 11 ) , ( 1 - 5 - 13 ) * for 20 balls : ( 1 - 6 - 13 ) , ( 1 - 7 - 12 ) , ( 2 - 6 - 12 ) we can notice that only one ball jump from a shell to another is necessary to induce a configuration switch .
these configurations correspond exactly to those computed in @xcite for logarithmic interparticle interaction potential , strongly suggesting this kind of interaction between the balls , at least within the range of our experimental inter - particle distance @xcite .
this conclusion has been confirmed later by comparing the ground configuration obtained for elliptic confinement with the configuration of vortices calculated in similar shaped mesoscopic superconductors for which the inter - vortices interaction is logarithmic in the considered range of inter - particle distance @xcite @xcite . from the energetic point of view , the local symmetry differences between the various systems induce strong differences in their excitation energy spectra @xcite .
the configuration energies for @xmath9 and @xmath10 are well separated , however the gap between the ground and the first excited state is larger for the magic number system .
by contrast , the latter is very small for @xmath11 . in section [ procedure ]
, we present the experiments which validate the mechanical shaking as an effective thermodynamic temperature and we describe the parameters used to characterize the disorder , the configuration transitions and the collective excitations .
the evolutions with temperature of these parameters for @xmath12 and 20 islands will be described and discussed in details in section [ transition ] . in section
[ order ] , the respective influence on the disordering of the individual and collective excitations will be discussed .
we will show that , according to the considered parameter , different transition temperatures can be identified .
in particular , the systems present an important configuration transition activity inducing disorder at a temperature smaller than those characterized by the lindemann criterion .
the `` melting '' will have to be considered rather as a disordering than a real melting . |
tunable continuous - wave ( cw ) lasers are the workhorse for quantum technologies based on ultracold atoms and ions . while the visible and near - infrared spectral regions are readily accessible by solid - state laser systems , the extension of these techniques to the near - ultraviolet region remains an active area of research .
many ionic and atomic species have their principal cooling transitions in the near - uv , including species such as be@xmath12 @xcite and hg @xcite that are relevant for optical frequency standards .
in addition to laser cooling , narrowband cw excitation enables precision spectroscopy of highly excited rydberg states .
rydberg atoms have high principal quantum number @xmath10 and exhibit large atomic polarizabilities , which scale as @xmath15 , thus providing tunable long - range interactions and high sensitivity to electric fields . with sufficiently narrowband excitation the long - range interactions between pairs of rydberg atoms can be observed .
in particular , if the interaction shift exceeds the linewidth of the excitation , then a blockade effect occurs @xcite and a single excitation is shared between all the atoms within a blockade radius @xmath16 , where @xmath17 is the van der waals interaction strength and @xmath18 is the excitation rabi frequency .
the rydberg blockade effect has been used to demonstrate quantum effects in neutral atoms , including cooperativity @xcite , entanglement @xcite , a controlled - not gate @xcite and coherent many - body rabi oscillations @xcite .
other applications of precision rydberg spectroscopy include electrometry @xcite , production of single photons @xcite and the creation of ultra - long range molecules @xcite . by optically `` dressing '' the ground or a metastable state with a rydberg state ,
further potential applications include demonstration of a super - solid phase of matter @xcite and spin - squeezing in optical lattice clocks for increased signal - to - noise beyond the standard quantum limit @xcite .
_ @xcite have recently demonstrated rydberg dressing in cs atoms for studying low - decoherence implementation of a controlled - z gate .
so far , most of these experiments have relied on a two - step excitation scheme via a short - lived , low - lying intermediate state .
while this has the advantage that both lasers may be in the visible or near - ir spectral regions , the short lifetime of the intermediate state introduces additional decoherence .
electromagnetically induced transparency @xcite or off - resonant two - photon excitation @xcite can reduce this effect , but at the expense of a lower rydberg excitation efficiency .
more recently , experiments in the alkali atoms have used direct excitation from the ground state , which requires tunable cw light at 297 nm ( rb ) @xcite and 319 nm ( cs ) @xcite . an alternative approach to reducing the decoherence of the two - photon excitation scheme is provided by two - electron atoms such as sr . here , intercombination transitions give access to long - lived intermediate states .
these provide a coherent two - photon excitation route to the triplet rydberg states , including the 5s@xmath10s@xmath19s@xmath20 series which have near - isotropic wavefunctions and interaction strengths @xcite .
isotropic interactions relax the geometric constraints of the atomic sample for many of the proposed rydberg and dressing experiments , and could be key for demonstrating a supersolid phase of matter @xcite . moving to two - electron atoms
also enables applications , such as spin - squeezing @xcite and blackbody thermometry @xcite , in optical lattice clocks . in sr
this requires tunable radiation in the 316 nm 322 nm region [ fig .
[ fig : fig1](a ) ] .
recently , the first experiments using this intercombination excitation route in sr have led to the observation of sr rydberg molecules @xcite and the study of autler - townes spectra in a dense sr rydberg gas @xcite .
combined with the ability to optically manipulate the second valence electron for trapping @xcite , state detection @xcite and imaging @xcite , intercombination line rydberg excitation provides a powerful toolbox for exploring many of these ideas .
the conventional approach to generating tunable cw light in the near - uv region of the spectrum has been to frequency - double a dye laser @xcite .
solid - state systems using sum - frequency generation ( sfg ) have been developed for particular cooling transitions , offering high power and narrow linewidth @xcite , but a tuning range of only a few ghz .
recently these techniques have been extended to lasers for rydberg spectroscopy @xcite , where the restricted tuning range means that only a narrow range of principal quantum numbers are accessible . in this paper
we describe a solid - state , narrow - linewidth uv laser system tunable from 316.3 nm 317.7 nm and 318.0 nm
319.3 nm , which enables the two - step excitation of sr rydberg states with principal quantum number in the range @xmath21@xmath22300 from two different intermediate states that are separated by 5.6thz .
we achieve a stable output power of @xmath23mw . a dual stabilization scheme based on a high - finesse tunable optical transfer cavity and an atomic reference
provides a long - term absolute frequency instability of @xmath24khz , measured relative to a gps - stabilized optical frequency comb .
the utility of the laser for rydberg physics is demonstrated by the spectroscopy of high - lying triplet rydberg states in sr .
the goal was to develop a laser system suitable for exciting to a large range of rydberg states in the triplet series of sr , via two different intermediate states [ fig .
[ fig : fig1](a ) ] .
a narrow laser linewidth is required in order to achieve a large rydberg blockade radius , and high power is necessary for off - resonant rydberg dressing since the coupling matrix elements are weak . with these criteria in mind
, we have built a high power , narrow - linewidth , widely - tunable laser system for excitation to the triplet rydberg series from both the 5s5p@xmath19p@xmath25 and 5s5p@xmath19p@xmath20 intermediate states .
the full range of wavelengths achievable , and the corresponding rydberg states that can be addressed , are given in table [ tab : crystal ] .
@xmath26k in a mot .
two - photon excitation with 689 nm and 319 nm , or 698 nm and 316 nm drives the atoms up to the triplet rydberg series .
( b ) schematic of the laser system .
a ppslt crystal is used to sum the frequencies of two infra - red lasers at wavelengths of @xmath27 and @xmath28 .
the resulting light at @xmath29 is frequency doubled to produce @xmath30mw in the uv ( @xmath31 ) .
the laser frequency is locked to an optical transfer cavity stabilized to the 5s@xmath32@xmath33s@xmath25 @xmath34 5s5p@xmath19p@xmath20 intercombination line in sr at 689 nm . a wideband electro - optic modulator ( eom )
is used to bridge the frequency gap between the cavity mode and the rydberg transition .
the laser frequency is measured on a gps - referenced optical frequency comb .
( edfa = er - doped fiber amplifier , ydfa = yb - doped fiber amplifier .
@xmath35 = dichroic mirrors ) . ] our approach , shown in fig .
[ fig : fig1](b ) , is based on that described in @xcite , and is similar to that used in @xcite .
two infra - red seed lasers are amplified and combined using sum frequency generation ( sfg ) in a periodically - poled stoichiometric lithium tantalate ( ppslt ) crystal to produce red light at @xmath36633 nm
subsequently , uv light is produced using resonantly - enhanced second harmonic generation ( shg ) .
a small amount of the red light is separated off for locking to an atomic - referenced optical cavity , and for independent absolute frequency measurements using a gps - referenced optical frequency comb .
the key challenge compared to previous work was to extend the sfg technique to meet our tuning requirements .
firstly , we exploit the wide gain bandwidth of er- and yb - doped fiber amplifiers to achieve high fundamental power over a wide range of wavelengths .
we employ two commercial fiber amplifier systems ; an er - doped amplifier ( manlight ) that produces up to 5w over the wavelength range @xmath27=1549 nm to 1565 nm , and a yb - doped amplifier ( nufern ) that produces up to 10w over @xmath28=1064 nm to 1083 nm .
table [ tab : crystal ] summarizes how we exploit these wavelength ranges to address the rydberg series of interest .
cccccc .wavelengths achievable with our laser system and the corresponding principal quantum numbers @xmath10 of the sr rydberg states we can access .
wavelengths @xmath37 and @xmath38 are combined using sfg to produce @xmath29 , which is frequency doubled to give the desired wavelength , @xmath31 . [
cols="^ " , ] + 1549.0 1565.0 & 1069.5 & 632.7 635.3 & 316.3 317.7 & @xmath19p@xmath25 & 32 320 + 1549.0
1565.0 & 1079.0 & 636.0 638.7 & 318.0 319.3 & @xmath19p@xmath20 & 35
@xmath39 + [ tab : crystal ] the 5.6thz frequency shift required to change between the two intermediate states is achieved by switching between two different seed lasers centered at @xmath401069.5 nm , or @xmath411079.0 nm .
both seed lasers are narrow - linewidth ( @xmath3610khz ) , high stability fiber lasers ( np photonics ) with polarization - maintaining single - mode optical fiber outputs .
the advantage of switching the seed laser in this way is that it requires no significant optical realignment .
tunability within each rydberg series is achieved by using a highly tunable external cavity diode laser ( ecdl ) ( toptica photonics ) as the other seed laser ( @xmath27 ) .
the laser linewidth is somewhat broader than available fiber laser technology ( @xmath36100khz ) , but the frequency can be tuned over the whole usable gain bandwidth of the fiber amplifier . the final step is to ensure that the sfg process is quasi - phase - matched across these wavelength ranges .
the phase mismatch @xmath42 can be calculated @xcite by considering the extraordinary refractive index @xmath43 of the crystal @xcite for a particular temperature @xmath44 and vacuum wavelength @xmath45 ( where @xmath46 corresponds to the two input , and one output , beams respectively ) , and the poling period of the crystal @xmath47 : @xmath48 we considered both ppslt and periodically - poled lithium niobate ( ppln ) crystals .
the latter offers a higher nonlinear coefficient ( @xmath49 ) @xcite , at the expense of higher light - induced infra - red absorption and a lower photo - refractive damage threshold @xcite , and therefore we chose to work with ppslt . inserting the relevant material parameters @xcite in to eq .
( [ eqn : temp1 ] ) we find that five equally spaced poling periods between 13.05@xmath50 m and 13.45@xmath50 m will provide quasi - phase - matching across the full rydberg series for crystal temperatures in the range 100@xmath51c
140@xmath51c .
the ppslt crystal we use ( laser 2000 ) has a length of @xmath52=50 mm and a thickness of 1 mm .
distributed across its 15 mm width are the five independent periodically - poled channels .
the crystal is mounted in a temperature stabilized oven ( laser 2000 ) , which allows for fine tuning of the refractive index and poling period .
the ir beams are independently mode - matched using lenses of focal length @xmath53 mm and @xmath54 mm in the @xmath27 beam , and @xmath55 mm and @xmath56 mm in the @xmath28 beam ( as shown in fig .
[ fig : fig1](b ) ) .
the beams are subsequently overlapped on a dichroic optic @xmath57 , and make a single pass of the ppslt crystal .
optimum frequency conversion should occur when the confocal parameter @xmath58 of the input beams are related to the length of the nonlinear crystal @xmath52 by @xmath59 @xcite .
however , the conversion efficiency is only weakly related to @xmath60 , and focusing the beams too tightly ( high @xmath60 ) puts a tighter constraint on the overlap of the beam waists in the axial direction , which is technically challenging to achieve .
instead we opt for a more relaxed @xmath58=32 mm ( @xmath60=1.56 ) for both input beams .
note that @xmath58 is the confocal parameter of the beam inside the crystal , which is quite different to that of the beam in air due to the high refractive index of the material .
our lens set - up allows for some adjustment of our chosen @xmath60 value , but we observe the conversion efficiency to be relatively insensitive to this parameter . after the crystal , dichroic mirrors ( @xmath61 in fig .
[ fig : fig1](b ) ) are used to separate the red and ir beams , and the unused ir power is dumped .
the sfg relative conversion efficiency as a function of temperature @xmath62 is predicted to follow a sinc@xmath32 function : @xmath63 the measured temperature dependence of the conversion efficiency is plotted in fig .
[ fig : fig2](a ) along with the expected curve from eq .
( [ eqn : temp1 ] ) and eq .
( [ eqn : temp2 ] ) .
we observe a feature with fwhm@xmath641@xmath51c , so the 0.1@xmath51c temperature resolution of our temperature controller is only just sufficient to stabilize the temperature to the top of the peak . the power dependence of the sfg process is shown in fig .
[ fig : fig2](b ) .
the output power @xmath65 is broadly linear with the product of the input powers @xmath66 and @xmath67 .
we obtain a maximum power of @xmath65=1.6w of @xmath29 light from the sfg process .
a least - squares linear fit to the data in fig .
[ fig : fig2](b ) yields the normalized conversion efficiency @xmath68 , @xmath69 where @xmath70 are the beam powers of @xmath71 and @xmath52 is the length of the ppslt crystal .
this measured value is favourable in comparison to other reported cw conversion efficiencies for ppslt , which are typically @xmath360.3@xmath72 to 0.4@xmath72w@xmath73cm@xmath73 @xcite .
however , it is noticeably worse than the normalized sfg conversion efficiencies achieved with the more widely used ppln , which are typically @xmath362.8@xmath72w@xmath73cm@xmath73 @xcite .
the lower nonlinear coefficient @xmath75 of lt does not account for the total difference in conversion efficiency , which could be due to the greater difficulty of growing and poling stoichiometric lithium tantalate @xcite . in the future ,
mgo doped stoichiometric ln could be a good compromise if the higher conversion efficiency is required , as the doping offers some level of protection against the damage mechanisms discussed above @xcite . .
measured data are shown in red ( a temperature independent background due to non - phase - matched shg has been removed ) , with the prediction from eq .
( [ eqn : temp1 ] ) and eq .
( [ eqn : temp2 ] ) shown in black dashes .
a small temperature offset ( 0.55@xmath51c ) has been added to the prediction plot to match the experimental data .
( b ) output power ( @xmath65 ) of the sfg process for a range of input powers ( @xmath76 ) .
data points are shown as red circles , and the blue line shows the least - squares linear fit to the data . ]
the majority of the light generated from the sfg process ( @xmath29 ) is coupled into a 10 m long high power , polarization maintaining , single - mode optical fiber , with @xmath2265% efficiency to a commercial frequency doubling unit ( toptica photonics ) .
a nonlinear crystal optimized for second harmonic generation using type - i phase matching is housed in a hermetically sealed bow - tie cavity .
the unit contains a resonant electro - optic modulator ( eom ) to apply sidebands to the @xmath29 light for pound - drever - hall ( pdh ) locking @xcite of the shg cavity length by feedback to a piezo - mounted mirror . for a well - optimized system
we have measured conversion efficiencies of @xmath3630% , producing 230mw of uv light from 780mw of red light . during typical operation
we obtain @xmath22100mw of uv power across a tuning range of 318.0 nm to 319.3 nm .
the alignment of the shg cavity requires re - optimization when the uv frequency is changed by more than @xmath36100ghz ; this is a relatively quick and simple procedure .
a tunable optical cavity is used to transfer the frequency stability of the 7.5khz wide 5s@xmath32@xmath33s@xmath25 @xmath34 5s5p@xmath19p@xmath20 689 nm transition in @xmath77sr [ fig .
[ fig : fig1](a ) ] to the uv laser system .
an ecdl operating at 689 nm is locked to a resonant mode of the optical cavity , and then the cavity length is stabilized to the 689 nm sub - doppler resonance feature in a thermal beam of sr .
the stable reference cavity can then be exploited to lock the frequency of the light at @xmath29 [ fig .
[ fig : fig1](b ) ] .
full details of the optical cavity design and construction are given in @xcite .
briefly , the optical transfer cavity is 10.5 cm long and consists of a zerodur spacer with a ring - shaped piezo between the spacer and each high reflectance mirror .
one of the cavity piezos is a large - stroke piezo for long term drift correction , whist the other is a short - stroke piezo for faster length corrections @xcite .
the cavity is housed in a temperature stabilized vacuum chamber at @xmath3627@xmath51c and @xmath3610@xmath78 mbar .
the finesse is @xmath79@xmath80@xmath2 ( fwhm=35khz ) at 689 nm and @xmath81@xmath80@xmath2 ( fwhm=13khz ) at 638 nm , as measured by cavity - ring - down spectroscopy .
the 689 nm laser is locked to the cavity using the pdh technique , and a portion of its output is used to generate an atomic error signal using sub - doppler fluorescence spectroscopy in a sr atomic beam . for this , the light is frequency modulated using a double - passed acousto - optic modulator and retro - reflected through the atomic beam .
the resulting fluorescence is measured on a high gain photodiode , followed by demodulation using a lock - in amplifier to derive the error signal .
feedback is provided to both cavity piezos with different bandwidths , stabilizing the cavity length to the atomic reference signal such that the 689 nm laser is locked on resonance . to lock the uv laser frequency ,
a small portion of the @xmath29 light is picked off from the main beam path and overlapped with the 689 nm beam on a dichroic filter ( @xmath82 in fig .
[ fig : fig1](b ) ) providing alignment into the same optical cavity .
the wavelength of the @xmath29 light is locked to a resonant mode of the atom - referenced cavity using the pdh technique . a fast servo - loop controller ( toptica photonics )
provides feedback to both the diode current and piezo voltage of the @xmath27 seed laser , with an overall loop bandwidth of @xmath36200khz .
the transfer cavity lock therefore both narrows the linewidth of the uv light as well as providing absolute frequency stability .
different modulation frequencies are used for the two pdh locks in order to prevent cross - talk between them . since the atom - referenced cavity mode is not necessarily at a frequency that is useful for driving the rydberg transitions , we use a fiber - coupled wideband eom ( jenoptik , pm635 ) to offset - lock the laser frequency from that of the cavity mode .
we use the `` electronic sideband '' technique @xcite , where a tunable sideband is locked to the cavity at a detuning @xmath83 from the main laser frequency .
the carrier frequency of the eom ( @xmath83=0.1mhz
5ghz ) is phase modulated at a frequency @xmath84=8mhz to produce the modulation sidebands required for pdh locking [ fig .
[ fig : fig1](b ) ] .
demodulation at @xmath84 yields a standard pdh error signal , locking the laser at a frequency @xmath83 away from the cavity mode .
the uv laser frequency can be stepped or scanned while locked to the cavity by varying @xmath83 @xcite .
this allows us to manipulate the laser frequency with high precision during an experimental sequence , such as for the spectrum shown in fig .
[ fig : fig4](b ) .
the laser remains locked ( or relocks very quickly ) for step sizes within the capture range of the pdh lock ( @xmath361mhz in our case ) .
we characterize the long - term frequency instability of the 638 nm and 689 nm lasers using heterodyne beat measurements with a gps - referenced optical frequency comb ( toptica photonics ) .
the fiber laser - based frequency comb has a difference frequency generation design that eliminates the carrier - envelope offset frequency .
the tenth harmonic of the 80mhz repetition rate is locked to an ultra - low noise oven - controlled oscillator ( oxco ) , which in turn is locked to a 10mhz gps - disciplined quartz oscillator ( gpsdo ) ( jackson labs fury ) for long - term ( @xmath850.1s ) stability .
independent measurements carried out by toptica photonics indicate that the performance of the frequency comb is limited by the rf - reference for timescales larger than the inverse of the locking bandwidth ( @xmath36 few kilohertz ) @xcite .
each cw laser is overlapped with the corresponding section of the frequency comb and the beat note between the laser and the comb is measured on a fast photodiode .
repeated measurements of the beat frequency ( @xmath86 ) are made over a period of a few hours using a zero - deadtime counter ( tektronix fca3003 ) . from these measurements
we calculate the overlapping allan deviation @xcite of the fractional frequency error @xmath87 , where @xmath88 is the mean beat frequency and @xmath89 is the laser frequency ( e.g. @xmath36434.829thz for the 689 nm laser ) .
the results are shown in fig .
[ fig : fig3 ] , along with the manufacturer s data for the gpsdo .
the allan deviations for the 638 nm and 689 nm lasers are in reasonable agreement , which is expected because they are locked to the same optical reference cavity .
a significant difference between these two measurements would indicate one or both of the lasers were not well locked to the transfer cavity . for @xmath90s ,
the beat frequency has less noise than the gpsdo , reflecting the much lower noise of the oxco .
above @xmath910.1s the instability of the optical frequency comb should follow that of the gpsdo @xcite . instead
, the measurements reveal excess frequency noise on both the cw lasers .
we attribute this noise to the performance of the transfer cavity length servo to the atomic reference [ fig .
[ fig : fig1](b ) ] .
the low atomic fluorescence rate ( @xmath92khz ) imposes a low bandwidth on the cavity length servo ( @xmath3610hz ) , which combined with a relatively high cavity drift rate due to piezo relaxation ( up to @xmath361mhz / s ) , leads to significant fluctuations in the lock .
nevertheless , fig .
[ fig : fig3 ] shows that the fractional frequency instability of both lasers systems is @xmath93 for all measurement times between @xmath1s and @xmath85@xmath2s , corresponding to frequency deviations of just @xmath3635khz for the uv laser and @xmath3615khz for the 689 nm laser . for measurement times
@xmath94s , the frequency deviation measurements eventually become limited by the short - term jitter of the frequency comb , so this method can not be used to infer the short - term linewidth .
instead , we estimate the short - term frequency noise of the lasers relative to the optical transfer cavity by looking at the spectrum of the in - loop pdh error signal .
note that this measurement gives a relative frequency instability , rather than the absolute instability provided by the optical frequency comb , since the resonance frequency of the cavity may fluctuate due to acoustic vibrations and electrical noise on the mirror piezos .
however , these cavity fluctuations will be relatively small at the timescales of interest for these measurements ( @xmath01ms ) .
the in - loop error signal of the 638 nm laser is converted into frequency deviations from the cavity mode center by dividing through by the pdh discriminator gradient @xmath95 , where @xmath96 is the peak to peak voltage of the unlocked error signal and @xmath97 is the cavity mode fwhm .
the overlapping allan deviation yields a relative fractional frequency instability of @xmath98 for all timescales between 2@xmath50s and 1ms , which corresponds to frequency deviations of @xmath04khz in the uv .
similar analysis of the 689 nm laser in - loop error signal is detailed in @xcite , where we estimate the short - term linewidth to be @xmath361.5khz
. therefore if the short - term fluctuations of the transfer cavity length are small , then the in - loop error signal analysis of the locked lasers shows short - term laser frequency fluctuations to be less than the natural linewidths of the 689 nm and rydberg transitions ( @xmath6410khz ) . the long - term analysis with the gps referenced
frequency comb shows frequency deviations outside of the natural linewidths , and so in future we plan to improve the long - term performance by locking to an ultra - stable optical reference cavity of the type used in optical atomic clocks . despite this , due to doppler and power broadening effects , the measured frequency instability is still sufficient to perform high resolution spectroscopy of rydberg states , as we show in the next section .
to demonstrate the versatility of this laser system , including the high tunability and narrow linewidth , we performed spectroscopy of a range of triplet rydberg states in @xmath77sr .
we use a two - step excitation scheme at 689 nm and 319 nm , as shown in fig .
[ fig : fig1](a ) .
rydberg spectroscopy was carried out in a laser cooled sample of @xmath36@xmath99 @xmath77sr atoms .
a detailed description of the apparatus can be found in @xcite .
briefly , atoms from a thermal beam of sr are slowed , cooled and trapped in a `` blue '' mot using the 5s@xmath32@xmath33s@xmath25 @xmath34 5s5p@xmath33p@xmath20 transition at 461 nm , where they have a temperature of @xmath365mk .
they are subsequently loaded into a `` broadband red '' mot using the 689 nm intercombination line . in the broadband mot ,
the 689 nm light is frequency modulated to broaden its spectrum by @xmath364mhz , which ensures efficient transfer from the relatively hot blue mot . at the end of the broadband cooling stage
the atoms typically have a temperature of @xmath100k and a density of @xmath101@xmath102 . to locate individual rydberg states , we perform continuous single - mode 24ghz scans of the uv laser frequency while the broadband mot is operating .
the uv laser is unlocked from the reference cavity , and scanned by piezo tuning of the @xmath27 ecdl cavity length .
the single - mode tuning range is ultimately limited by the @xmath103 ghz bandwidth of the sfg process , although in practice the mode - hop free tuning range of the ecdl imposes a slightly lower limit .
when the uv frequency matches a transition frequency , atoms are resonantly excited from the 5s5p@xmath19p@xmath20 state to a rydberg state .
some of these rydberg atoms spontaneously ionize , and we count the resulting ions using a multi - channel plate ( mcp ) and a fast oscilloscope .
results of such scans in the region @xmath104 and @xmath105 are shown in fig .
[ fig : fig4](a ) , where the rydberg states appear as sharp peaks in the ion signal .
we identify the rydberg states by using the rydberg - ritz parameters given in @xcite to extrapolate from previous measurements at lower principal quantum numbers @xmath10 @xcite . a small electric field ( @xmath361vcm@xmath73 ) applied during the excitation sequence , broadens and stark splits the @xmath19s@xmath20- and @xmath19d@xmath106-states as well as allowing excitation to the @xmath19p@xmath106-states .
both effects are clearly visible in the @xmath107 data . and @xmath105 .
the intensities used for the excitation are @xmath3620mw@xmath108 and @xmath361w@xmath108 for the 689 nm and uv beams respectively , and each beam is larger than the size of the atom cloud .
the variation in signal height across the scan is largely due to depletion of atoms from the mot and is not an indication of transition strength .
the frequency axis of the scans is calibrated on a 10mhz resolution wavemeter ( high finesse ) .
the offset frequencies are are @xmath36940.649thz ( @xmath104 ) and @xmath36941.616thz ( @xmath105 ) .
( b ) high resolution scan of the @xmath109 transition .
the solid blue line shows the voigt profile fit .
the intensities used for the excitation are @xmath360.1mw@xmath108 and @xmath36250mw@xmath108 for the 689 nm and uv beams respectively , and each beam is larger than the size of the atom cloud .
the central frequency of this feature is at @xmath36939.274thz . ] with the laser system locked to the transfer cavity ( as described in section [ sec : laserlock ] ) we can perform rydberg spectroscopy with @xmath010khz - level precision . here
, atoms are released from the broadband mot , before being excited to the 5s37s@xmath19s@xmath20 rydberg state using resonant 689 nm and 319 nm light .
the 689 nm excitation beam counter - propagates at an angle of 30@xmath51 to the uv beam , and both lasers are pulsed on simultaneously for 300@xmath50s . a quantization magnetic field ( @[email protected] ) applied during the excitation pulse
separates the @xmath111 states . following the rydberg excitation , a 20@xmath50s pulse of 408 nm laser light
is used to autoionize the rydberg atoms and an electric field is applied to direct the ions towards the mcp @xcite .
we repeat each measurement 10 times in the same atomic sample , after which the mot is reloaded and the uv laser frequency stepped to the next detuning value by changing the eom offset frequency @xmath83 .
we take the spectrum in 50khz steps , tuning the laser frequency up over the feature , then back down again .
the results are shown in fig .
[ fig : fig4](b ) .
the lineshape is very well described by a voigt profile ( reduced @xmath112=1.1 ) with an overall fwhm of @xmath113khz . from the fit
we extract the fwhm of the inhomogeneous ( gaussian ) @xmath114khz and homogeneous ( lorentzian ) @xmath115khz contributions .
both are in agreement with our expected doppler broadening and power broadening respectively .
the fit yields the line center with a precision of @xmath365khz .
approximate transition frequencies are given for the observed states in the caption for fig .
[ fig : fig4 ] . in the future ,
absolute frequency measurements at the @xmath116 level will be possible with the use of the gps - referenced optical frequency comb .
however , this will require a detailed analysis of the electric and magnetic field induced line shifts , which is beyond the scope of this paper .
in conclusion , we have developed a solid - state laser system for generating high power , narrow - linewidth uv light , which can be tuned across the range 316.3 nm
317.7 nm and 318.0 nm
the rydberg states shown in fig .
[ fig : fig4 ] span a frequency range of over @xmath117thz , providing a clear illustration of the wide tunability of the laser design . at the same time
, we demonstrate broad single - mode tuning range ( 24ghz ) suitable for atomic spectroscopy , and precision control of the absolute frequency , enabling rydberg spectroscopy with kilohertz precision .
the long - term frequency instability of the laser system ( @xmath035khz for timescales between @xmath1s and @xmath2s ) could be reduced by using an ultra - low drift optical reference cavity , and either an improved atomic spectroscopy set - up or the frequency comb as a long - term absolute reference . with these modifications ,
an absolute frequency instability below @xmath118khz should be attainable .
the laser output power is limited by the conversion efficiency of the ppslt crystal .
although we obtain one of the highest conversion efficiencies so far reported for this material , replacement with mgo doped ppln could provide a substantial increase in uv power .
the precision and versatility of this laser system make it ideal for future sr rydberg experiments , including high precision spectroscopy of the rydberg states and their interactions , and experiments involving rydberg dressed potentials .
the all solid - state design makes this an attractive replacement for complex frequency - doubled dye laser systems .
the authors would like to thank thomas puppe , russell kliese and alexander sell from toptica photonics for their work on developing and installing the optical frequency comb , and chunyong li at durham university for maintaining it .
we also thank ian hill from the national physical laboratory for comments on the manuscript . financial support
was provided by epsrc grant ep / j007021/ and eu grants fp7-ict-2013 - 612862-hairs and h2020-fetproact-2014 - 640378-rysq .
the data presented in this paper are available for download @xcite . | we present a solid - state laser system that generates over 200mw of continuous - wave , narrowband light , tunable from 316.3 nm 317.7 nm and 318.0 nm
319.3 nm .
the laser is based on commercially available fiber amplifiers and optical frequency doubling technology , along with sum frequency generation in a periodically poled stoichiometric lithium tantalate crystal .
the laser frequency is stabilized to an atomic - referenced high finesse optical transfer cavity . using a gps - referenced optical frequency comb we measure a long term frequency instability of @xmath035khz for timescales between @xmath1s and @xmath2s . as an application
we perform spectroscopy of sr rydberg states from @xmath3 81 , demonstrating mode - hop - free scans of 24ghz . in a cold atomic sample
we measure doppler - limited linewidths of 350khz . 99 j. j. bollinger , j. d. prestage , w. m. itano , and d. j. wineland , `` laser - cooled - atomic frequency standard , '' phys . rev . lett . * 54 * , 1000 ( 1985 ) .
h. hachisu , k. miyagishi , s. g. porsev , a. derevianko , v. d. ovsiannikov , v. g. palchikov , m. takamoto , and h. katori , `` trapping of neutral mercury atoms and prospects for optical lattice clocks , '' phys .
rev .
lett . *
100 * , 053001 ( 2008 ) .
m. d. lukin , m. fleischhauer , r. cote , l. m. duan , d. jaksch , j. i. cirac , and p. zoller , `` dipole blockade and quantum information processing in mesoscopic atomic ensembles , '' phys .
rev .
lett . * 87 * , 037901 ( 2001 ) .
j. d. pritchard , d. maxwell , a. gauguet , k. j. weatherill , m. p. a. jones , and c. s. adams , `` cooperative atom - light interaction in a blockaded rydberg ensemble , '' phys .
rev .
lett . * 105 * , 193603 ( 2010 ) .
t. wilk , a. ga@xmath4tan , c. evellin , j. wolters , y. miroshnychenko , p. grangier , and a. browaeys , `` entanglement of two individual neutral atoms using rydberg blockade , '' phys . rev
. lett . * 104 * , 010502 ( 2010 ) .
l. isenhower , e. urban , x. l. zhang , a. t. gill , t. henage , t. a. johnson , t. g. walker , and m. saffman , `` demonstration of a neutral atom controlled - not quantum gate , '' phys .
rev .
lett .
* 104 * , 010503 ( 2010 ) . y. o. dudin , l. li , f. bariani , and a. kuzmich , `` observation of coherent many - body rabi oscillations , '' nat .
phys .
* 8 * , 790 - 794 ( 2012 ) .
r. p. abel , c. carr , u. krohn , and c. s. adams , `` electrometry near a dielectric surface using rydberg electromagnetically induced transparency , '' phys .
rev .
a * 84 * , 023408 ( 2011 ) .
j. a. sedlacek , a. schwettmann , h. k@xmath5bler , r. l@xmath6w , t. pfau , and j. p. shaffer , `` microwave electrometry with rydberg atoms in a vapour cell using bright atomic resonances , '' nat .
phys .
* 8 * , 819 - 824 ( 2012 ) .
g. lochead , d. boddy , d. p. sadler , c. s. adams , and m. p. a. jones , `` number - resolved imaging of excited - state atoms using a scanning autoionization microscope , '' phys .
rev .
a * 87 * , 053409 ( 2013 ) .
m. saffman and t. g. walker , `` creating single - atom and single - photon sources from entangled atomic ensembles , '' phys . rev .
a * 66 * , 065403 ( 2002 ) . c. h. greene , a. s. dickinson , and h. r. sadeghpour , `` creation of polar and nonpolar ultra - long - range rydberg molecules , '' phys .
rev .
lett .
* 85 * , 2458 ( 2000 ) . c. boisseau , i. simbotin , and r. c@xmath7t@xmath8 , `` macrodimers : ultralong range rydberg molecules , '' phys .
rev .
lett . *
88 * , 133004 ( 2002 ) .
v. bendkowsky , b. butscher , j. nipper , j. p. shaffer , r. l@xmath6w , and t. pfau , `` observation of ultralong - range rydberg molecules , '' nature * 458 * , 1005 - 1008 ( 2009 ) . b. j. desalvo , j. a. aman , f. b. dunning , t. c. killian , h. r. sadeghpour , s. yoshida , and j. burgd@xmath6rfer , `` ultra - long - range rydberg molecules in a divalent atomic system , '' phys .
rev .
a * 92 * , 031403(r ) ( 2015 ) .
n. henkel , r. nath , and t. pohl , `` three - dimensional roton excitations and supersolid formation in rydberg - excited bose - einstein condensates , '' phys .
rev .
lett .
* 104 * , 195302 ( 2010 ) .
f. cinti , p. jain , m. boninsegni , a. micheli , p. zoller , and g. pupillo , `` supersolid droplet crystal in a dipole - blockaded gas , '' phys .
rev .
lett . * 105 * , 135301 ( 2010 ) .
l. i. r. gil , r. mukherjee , e. m. bridge , m. p. a. jones , and t. pohl , `` spin squeezing in a rydberg lattice clock , '' phys .
rev .
lett . * 112 * , 103601 ( 2014 ) .
t. keating , r. l. cook , a. m. hankin , y .- y .
jau , g. w. biedermann , and i. h. deutsch , `` robust quantum logic in neutral atoms via adiabatic rydberg dressing , '' phys .
rev . a * 91 * , 012337 ( 2015 ) .
a. k. mohapatra , t. r. jackson , and c. s. adams , `` coherent optical detection of highly excited rydberg states using electromagnetically induced transparency , '' phys .
rev .
lett . * 98 * , 113003 ( 2007 ) .
j. deiglmayr , m. reetz - lamour , t. amthor , s. westermann , a. l. de oliveira , and m. weidem@xmath5ller , `` coherent excitation of rydberg atoms in an ultracold gas , '' opt . commun . * 264 * , 293 - 298 ( 2006 ) .
p. thoumany , t. h@xmath9nsch , g. stania , l. urbonas , and th .
becker , `` optical spectroscopy of rubidium rydberg atoms with a 297 nm frequency - doubled dye laser , '' opt .
lett . * 34*(11 ) , 1621 - 1623 ( 2009 ) .
a. m. hankin , y. -y .
jau , l. p. parazzoli , c. w. chou , d. j. armstrong , a. j. landahl , and g. w. biedermann , `` two - atom rydberg blockade using direct 6s to @xmath10p excitation , '' phys .
rev . a * 89 * , 033416 ( 2014 ) . c. l. vaillant ,
m. p. a. jones , and r. m. potvliege , `` long - range rydberg - rydberg interactions in calcium , strontium and ytterbium , '' j. phys .
b * 45*(13 ) , 135004 ( 2012 ) .
v. d. ovsyanikov , a. derevianko , and k. gibble , `` rydberg spectroscopy in an optical lattice : blackbody thermometry for atomic clocks , '' phys .
rev .
lett . * 107 * , 093003 ( 2011 ) .
b. j. desalvo , j. a. aman , c. gaul , t. pohl , s. yoshida , j. burgd@xmath6rfer , k. r. a. hazzard , f. b. dunning , and t. c. killian , `` rydberg - blockade effects in autler - townes spectra of ultracold strontium , '' arxiv:1510.08032 .
r. mukherjee , j. millen , r. nath , m. p. a. jones , and t. pohl , `` many - body physics with alkaline - earth rydberg lattices , '' j. phys .
b * 44*(18 ) , 184010 ( 2011 ) .
j. millen , g. lochead , and m. p. a. jones , `` two - electron excitation of an interacting cold rydberg gas , '' phys .
rev .
lett .
* 105 * , 213004 ( 2010 ) .
j. millen , g. lochead , g. r. corbett , r. m. potvliege , and m. p. a. jones , `` spectroscopy of a cold strontium rydberg gas , '' j. phys .
b * 44*(18 ) , 184001 ( 2011 ) .
r. beigang , k. l@xmath5cke , d. schmidt , a. timmermann , and p. j. west , `` one - photon laser spectroscopy of rydberg series from metastable levels in calcium and strontium , '' phys .
scripta * 26 * , 183 - 188 ( 1982 ) .
t. manthey , t. m. weber , t. niederpr@xmath5 m , p. langer , v. guarrera , g. barontini , and h. ott , `` scanning electron microscopy of rydberg - excited bose - einstein condensates , '' new j. phys . * 16 * ( 8) , 083034 ( 2014 ) .
a. c. wilson , c. ospelkaus , a. p. vandevender , j. a. mlynek , k. r. brown , d. leibfried , and d. j. wineland , `` a 750-mw , continuous - wave , solid - state laser source at 313 nm for cooling and manipulating trapped @xmath11be@xmath12 ions , '' appl .
phys .
b * 105*(4 ) , 741 - 748 ( 2011 ) .
r. j. rengelink , r. p. m. j. w. notermans , and w. vassen , `` a simple 2w continuous - wave laser system for trapping ultracold metastable helium atoms at the 319.8 nm magic wavelength , '' arxiv:1511.00443 .
l. e. myers and w. r. bosenberg , `` periodically poled lithium niobate and quasi - phase - matched optical parametric oscillators , '' ieee j. quantum elect . *
33*(10 ) , 1663 - 1672 ( 1997 ) .
a. bruner , d. eger , m. b. oron , p. blau , m. katz , and s. ruschin , `` temperature - dependent sellmeier equation for the refractive index of stoichiometric lithium tantalate , '' opt . lett . * 28*(3 ) , 194 - 196 ( 2003 ) .
d. s. hum , r. k. route1 , g. d. miller , v. kondilenko , a. alexandrovski , j. huang , k. urbanek , r. l. byer , and m. m. fejer , `` optical properties and ferroelectric engineering of vapor - transport - equilibrated , near - stoichiometric lithium tantalate for frequency conversion , '' j. appl .
phys .
* 101 * , 093108 ( 2007 ) .
k. kitamura , y. furukawa , s. takekawa , m. nakamura , a. alexandrovski , and m. m. fejer , `` optical damage and light - induced absorption in near - stoichiometric litao@xmath13 crystal , '' lasers and electro - optics , 2001 .
cleo 01 .
technical digest .
summaries of papers presented at the conference on ; 02/2001 , 138 - 139 ( 2001 ) . y. s. kim and r. t. smith , `` thermal expansion of lithium tantalate and lithium niobate single crystals , '' j. appl .
phys . * 40 * , 4637 ( 1969 ) .
g. d. boyd and d. a. kleinman , `` parametric interaction of focused gaussian light beams , '' j. appl .
phys . * 39 * , 3597 , ( 1968 ) . s. sinha , d. s. hum , k. e. urbanek , l. yin - wen m. j. f. digonnet , m. m. fejer , and r. l. byer , `` room - temperature stable generation of 19 watts of single - frequency 532-nm radiation in a periodically poled lithium tantalate crystal , '' j. lightwave technol . * 26*(24 ) , 3866 - 3871 ( 2008 ) .
m. katz , r. k. route , d. s. hum , k. r. parameswaran , g. d. miller , and m. m. fejer , `` vapor - transport equilibrated near - stoichiometric lithium tantalate for frequency - conversion applications , '' opt .
lett . *
29*(15 ) , 1775 - 1777 ( 2004 ) .
d. l. hart , l. goldberg , and w. k. burns , `` red light generation by sum frequency mixing of er / yb fibre amplifier output in qpm linbo@xmath13 , '' electron .
lett . * 35*(1 ) , 52 - 53 ( 1999 ) .
y. furukawa , k. kitamura , s. takekawa , a. miyamoto , m. terao , and n. suda , `` photorefraction in linbo@xmath13 as a function of [ li]/[nb ] and mgo concentrations , '' appl .
. lett . * 77 * , 2494 ( 2000 ) .
r. w. p. drever , j. l. hall , f. v. kowalski , j. hough , g. m. ford , a. j. munley , and h. ward , `` laser phase and frequency stabilization using an optical resonator , '' appl .
phys .
b * 31*(31 ) , 97 - 105 ( 1983 ) . d. boddy doctoral thesis , `` first observations of rydberg blockade in a frozen gas of divalent atoms , '' durham university http://etheses.dur.ac.uk/10740/ ( 2014 ) .
y. li , t. ido , t. eichler , and h. katori , `` narrow - line diode laser system for laser cooling of strontium atoms on the intercombination transition , '' appl .
phys .
b * 78*(3 ) , 315 - 320 ( 2004 ) . j. i. thorpe , k. numata , and j. livas , `` laser frequency stabilization and control through offset sideband locking to optical cavities , '' opt .
express * 16*(20 ) , 15980 - 15990 ( 2008 ) .
t. puppe , toptica photonics ag , lochhamer schlag 19 , 82166 graefelfing ( munich ) , germany ( personal communication , 2015 ) .
f. l. walls and e. s. ferre - pikal , `` frequency standards , characterization , '' wiley encyclopedia of electrical and electronics engineering * 12 * , 767 - 775 ( 1999 ) .
j. r. rubbmark and s. a. borgstr@xmath6 m , `` rydberg series in strontium found in absorption by selectively laser - excited atoms , '' phys .
scr . * 18*(4 ) , 196 - 208 ( 1978 ) .
j. a. armstrong , j. j. wynne , and p. esherick , `` bound , odd - parity @xmath14 = 1 spectra of the alkaline earths : ca , sr , and ba , '' j. opt
. soc
. am .
* 69*(2 ) , 211 - 230 ( 1979 ) .
e. m. bridge , n. c. keegan , a. d. bounds , d. boddy , d. p. sadler , and m. p. a. jones , `` a tunable cw uv laser with @xmath035khz absolute frequency instability for precision spectroscopy of sr rydberg states : supporting data , '' doi:10.15128/1v53jw96n . |
the nearest - neighbors lattice gas analogy of onsager s solution of the two - dimensional ising model enabled one to understand physical implications of the spontaneous breaking of the particle - hole symmetry on the existence and critical properties of high - density liquid and low - density vapor phases @xcite .
as concerns continuum fluids in thermal equilibrium , widom and rowlinson @xcite introduced a simple model of identical particles ( molecules ) living in an infinite @xmath10-dimensional space of points @xmath11 . there is a sphere of radius @xmath12 and volume @xmath13 around the center of each molecule .
the potential energy @xmath14 associated with a given configuration @xmath15 of @xmath16 molecules is defined by @xmath17 , \ ] ] where @xmath18 denotes the volume covered by the corresponding @xmath16 ( in general penetrating ) spheres and @xmath19 is some energy constant . due to the obvious inequalities @xmath20 , the potential energy is bounded as follows @xmath21 the lower bound ensures a correct thermodynamics , the upper bound tells us that the short - range forces among molecules are purely attractive .
the model is studied within the grand canonical ensemble characterized by the dimensionless inverse temperature @xmath22 and the particle fugacity @xmath1 .
the corresponding particle density is given by @xmath23 ; @xmath1 is normalized so as to be asymptotically equal to @xmath24 in the ideal gas limit @xmath25 .
the symmetry of the widom - rowlinson ( wr ) fluid , whose spontaneous breaking is responsible for the existence of liquid and vapor phases , is hidden in the original formulation .
it becomes transparent after mapping ( in a thermodynamic sense ) the wr model onto a wr mixture of two kinds of molecules @xmath26 interacting in a pairwise manner @xmath27 , where the particles of the same species do not interact whereas the unlike species interact with a hard - core repulsion at distances smaller than @xmath28 , @xmath29 let us consider the case of equivalent species fugacities @xmath30 . at very low @xmath1
, the system behaves like the ideal gas with just one pure ( mixed ) phase with equivalent species densities @xmath31 . at very large @xmath1 , since unlike molecules experience an infinitely strong repulsion , the mixed phase suffers from packing effects which are substantially reduced in a demixed phase with a single majority component .
consequently , the @xmath32 symmetry is broken and the wr mixture can exist in two different homogeneous ( i.e. translationally invariant ) pure phases : the @xmath33-rich phase with @xmath34 or the @xmath35-rich phase with @xmath36 .
the two phases become equivalent ( @xmath37 ) at the `` demixing '' critical point @xmath38 . for dimensions
@xmath39 , the proof of the existence of more than one pure thermodynamic phase for sufficiently large @xmath1 was given by ruelle @xcite using the peierls contour method .
ruelle s proof was generalized by lebowitz and lieb @xcite to the case when @xmath40 ( @xmath41 ) is large positive but not infinite .
integral equation theories for the pair correlation functions of the wr mixture were developed in refs .
@xcite .
the multicomponent generalization of the wr @xmath32 mixture consists in considering molecules of @xmath0 different types @xmath42 with the same fugacity @xmath1 .
the molecules interact in a pairwise manner , the only interaction is the hard - sphere repulsion between any two particles of unlike species like in ( [ interaction ] ) .
it was shown @xcite that in dimensions @xmath39 the wr model with any finite number of components @xmath0 exhibits the demixing phase transition at some @xmath8 ; in a pure demixed phase , the homogeneous density of just one of the components is dominant , say @xmath43 .
a hard ( hyper-)cube version of the @xmath0-component wr model was studied in the limit of infinite dimensionality @xmath44 @xcite ; the calculation of thermodynamic functions within the second virial coefficient is exact in that limit .
it turns out that for @xmath45 the transition from the mixed phase at small values of @xmath1 to the demixed phase at large values of @xmath1 is preempted by solidification at intermediate values of @xmath1 , @xmath46 . in the corresponding crystal phase ,
all species are equivalent ( @xmath47 with @xmath24 being the total density of molecules ) , but the density @xmath48 varies periodically in space , i.e. the translational symmetry is broken .
the origin of this phenomenon is purely entropic : for large @xmath0 it pays the system to create a periodic structure of alternating dense and sparse regions , where the particles in the dense regions are less restricted by the hard - core repulsions coming from particles in the sparse regions . in the lattice version of the multicomponent wr model @xcite
, each lattice site @xmath49 can be either empty @xmath50 or singly occupied by a particle of type @xmath51 , all particles having the same fugacity @xmath52 .
the potential energy of a state configuration @xmath53 reads as @xmath54 , where the interaction potential between nearest - neighbor sites @xmath55 is given by @xmath56 the number density of species @xmath42 at site @xmath49 will be denoted by @xmath57 , the total density of particles at site @xmath49 by @xmath58 .
the phase diagram of the lattice wr model is similar to that of the continuous wr model . in dimensions @xmath39 and for any finite number of components @xmath0 ,
the lattice wr model exhibits the demixing phase transition at some @xmath8 ; in a pure demixed phase , the site - independent density of just one of the components is dominant , say @xmath59 .
the effect of varying exclusion diameter between different species was studied in @xcite .
when the number of components @xmath0 is equal to or larger than some minimum @xmath3 , an entropy - driven crystal phase exists for @xmath46 ; the transition at @xmath7 is always of second order . in the crystal phase ,
the average total particle densities on the even and odd sublattices are unequal , @xmath60 , while the average densities of the species @xmath61 are the same within a given sublattice , i.e. @xmath62 and @xmath63 .
the rigorous upper bound @xmath64 derived for the square lattice @xcite was surprisingly large .
the exact solution for the bethe lattice of coordination @xmath65 gives @xmath66 ^ 2 $ ] which would suggest more realistic @xmath67 for @xmath68 , while the monte - carlo ( mc ) simulations for the square lattice imply @xmath69 @xcite .
the extension of the hard - core exclusion to next - to - nearest - neighbors leads to analogous phases whose numbers and characters depend on the specific lattice geometry @xcite .
the infinite hard - core potential is a mathematical simplification of real inter - particle interactions . in this paper , we generalize the multicomponent wr model by replacing the hard - core exclusion between unlike species by an arbitrary repulsion .
namely , the infinity in the interaction potential ( [ intpot ] ) between unlike species on the nearest - neighbor sites is replaced by @xmath70 . the corresponding interaction boltzmann factor
@xmath71 lies in the interval @xmath72 $ ] .
the case @xmath73 @xmath74 corresponds to the standard multicomponent wr model .
the opposite extreme case @xmath75 @xmath76 is equivalent to the non - interacting @xmath0-component lattice gas of singly occupied lattice sites , with the trivial grand partition function for a lattice of @xmath16 sites : @xmath77 we shall concentrate on the neighborhood of the wr point @xmath73 , i.e. on large but finite repulsions , and study fundamental effects of nonzero @xmath78 on the phase diagram .
the model is solved exactly for the bethe lattice with an arbitrary coordination number @xmath65 .
the numerical calculations , based on the corner transfer matrix renormalization group ( ctmrg ) , are performed for the two - dimensional square lattice .
the results for @xmath9 indicate that the second - order phase transitions from the disordered gas to the demixed phase become of first order , for an arbitrarily small positive @xmath78 . the results for @xmath6 show that
, as the repulsion weakens , the region of crystal phase diminishes itself . for weak enough repulsions , the direct transition between the crystal and demixed phases changes into a separate pair of crystal - gas and gas - demixed transitions
; this is an example of a disorder - disorder reentrant transition via an ordered crystal phase .
if the repulsion between unlike species is too weak , the crystal phase disappears from the phase diagram .
a reentrant fluid - solid - fluid transition was observed in previous studies of systems with soft cores and two repulsive interaction ranges , like the two - scale ramp potential @xcite or the square well square shoulder model @xcite .
the computer simulations of the purely repulsive ramp potential @xcite show a structural anomaly , namely the melting of the solid phase when the pressure is increased along an isotherm .
this phenomenon was detected also in stillinger s gaussian core model @xcite and the antiferromagnetic ising model with a nearest - neighbor interaction and a staggered mean field @xcite , and later on in models with one - scale interactions @xcite .
the structural anomaly is shared by real physical systems like water , silica or phosphorus @xcite .
different definitions of the structural anomaly in fluids , which can lead to very different results , was discussed in @xcite . in all mentioned papers ,
the fluid - solid and solid - fluid transitions are discontinuous ( of first order ) , with regions of the phase coexistence . on the other hand ,
the disorder - crystal and crystal - disorder transitions of the generalized wr model are always continuous ( of second order ) .
the paper is outlined as follows . in sec .
ii , we present the exact solution of the model on the bethe lattice with the coordination number @xmath65 . for the two - dimensional square lattice ,
the ctmrg technique is explained in sec .
b brings the numerical results .
conclusions are given in sec .
we consider the bethe lattice of ( locally equivalent ) sites deep inside a tree , with coordination number @xmath65 .
the present calculation is based on the exact solution of an `` inverse problem '' for simply connected lattice structures @xcite ; for details in the @xmath73 case , see ref .
@xcite .
every site @xmath49 of the bethe lattice is an articulation point of multiplicity @xmath65 .
site @xmath49 can be either empty , @xmath79 , or occupied by one of the particles of @xmath0 different types , @xmath42 .
the corresponding set of fugacities reads as @xmath80 and @xmath81 for @xmath42 . for a given statistical model with two - site interactions between nearest - neighbor sites ,
we calculate the mean particle densities @xmath82 , constrained by @xmath83 for each lattice site @xmath49 .
the direct problem , find @xmath82 for given @xmath84 , is nonlocal .
the inverse problem , find @xmath84 for prescribed @xmath82 , is local in the sense that @xmath85 depends on @xmath57 at the same site and on @xmath86 at the nearest - neighbor sites @xmath87 . according to @xcite ,
the inverse solution for the bethe lattice can be constructed from local inverse problems for nearest - neighbor pairs of sites .
* let @xmath88 ( @xmath89 ) be the set of fugacities at site @xmath49 @xmath90 which produce the given density profiles @xmath57 and @xmath91 at nearest - neighbor sites @xmath55 .
the fugacities at the original bethe lattice are then expressible as @xmath92^{q-1 } \prod_{j=1}^q z_i^{\langle i , j\rangle}(\sigma ) .\ ] ] * similarly , denoting by @xmath93 the one - site and by @xmath94 the two - site grand partition function , the grand potential on the whole lattice is given by @xmath95 in the two - site problems , the expected symmetry breaking of the particle densities must be reflected via an analogous symmetry breaking in the corresponding fugacities . in the crystal phase
, there are two alternating even and odd sublattices . for the particle states
@xmath42 , we set @xmath96 for even sites and @xmath97 for odd sites .
in the two - site @xmath55 problem , we have @xmath98 and @xmath99 , @xmath100 for @xmath42 .
the two - site grand partition function @xmath101 reads as @xmath102 where the first term comes from two empty sites , the second one from one empty site and the other occupied by an arbitrary particle , the third one from two particles in the same state and the fourth one from two particles in different states .
the corresponding one - site particle densities are given by @xmath103 after some algebra , these equations determine @xmath104 and @xmath105 as functions of the particle densities as follows @xmath106+\sqrt{d } } { 2(1-m\rho_1)[1+j(m-1 ) ] } , \nonumber \\ & & \\ z_2 & = & \frac{m(\rho_1+\rho_2)-1+(\rho_2-\rho_1)[1+j(m-1)]+\sqrt{d } } { 2(1-m\rho_2)[1+j(m-1 ) ] } , \nonumber \\ & & \end{aligned}\ ] ] where the plus sign of the square root of the discriminant @xmath107 ^ 2 \nonumber \\ & & + 4\rho_1 \rho_2 ( m-1)(1-j)[1+j(m-1)]\end{aligned}\ ] ] is fixed by the condition @xmath108 for @xmath109 .
for @xmath110 , we get @xmath111 } \nonumber \\ & & + \frac{2[1+j(m-1 ) ] + m(\sqrt{d}-1)}{2(1-m\rho_1)(1-m\rho_2)[1+j(m-1 ) ] } .\end{aligned}\ ] ] the original lattice fugacity in ( [ origfugacity ] ) does not depend on site @xmath49 for all particle states @xmath42 , @xmath81 . using the density constraint ( [ densityconstr ] )
, we obtain one equation for each of two sublattices : @xmath112 the grand - potential per site @xmath113 is determined from ( [ omega ] ) as @xmath114 .\ ] ] introducing the new variables @xmath115 and @xmath116 , eqs .
( [ 1 ] ) and ( [ 2 ] ) can be written as @xmath117 they always have a trivial solution with @xmath118 , which corresponds to the disordered phase of equivalent even and odd sublattices , @xmath119 . a nontrivial solution @xmath120 exists if @xmath121 $ ] , where the lower and upper bounds are given by the equation @xmath122 @xcite : @xmath123 with @xmath124 the corresponding critical fugacities read @xmath125^q } \frac{1-\sqrt{e}}{1+\sqrt{e } } \left ( \frac{(q-2)/q - \sqrt{e}}{1-\sqrt{e } } \right)^q , \nonumber \\ & & \label{zcl } \\ z_c^u & = & \frac{m^{q-1}}{[1+j(m-1)]^q } \frac{1+\sqrt{e}}{1-\sqrt{e } } \left(\frac{(q-2)/q + \sqrt{e}}{1+\sqrt{e } } \right)^q .
\nonumber \\ & & \label{zcr } \end{aligned}\ ] ] the value of the nontrivial @xmath126 changes continuously from 0 at @xmath127 to some nonzero symmetry - broken value @xmath128 in the interval @xmath129 and finally goes again continuously to 0 at @xmath130 . for @xmath131 , the disordered phase with the trivial @xmath118 takes place .
the crystal phase can exist only if the discriminant @xmath132 .
this means that , for the fixed coordination number @xmath65 and the boltzmann factor @xmath133 , the number of components @xmath0 must be equal or larger than some minimum @xmath3 , @xmath6 , given by @xmath134 note that @xmath135 just at @xmath136 .
equivalently , for the fixed coordination number @xmath65 and the number of components @xmath137 , the crystal phase exists only if the boltzmann factor @xmath138 , where the maximum value @xmath139 is given by @xmath140 in the limit of @xmath141 , we have the asymptotic expansion @xmath142 in the demixed phase , all sites are equivalent but one of the components , say @xmath143 , is dominant .
this means that the density of 1-particles @xmath144 is larger than @xmath145 for all remaining particle states @xmath146 . in the two - site @xmath55 problem , we set @xmath98 , @xmath147 for particles of type 1 and @xmath148 for all particles of type @xmath146 .
the two - site grand partition function is given by @xmath149 the corresponding particle densities are given by @xmath150 considering in ( [ origfugacity ] ) @xmath81 for all sites @xmath49 and particle states @xmath61 , we get @xmath151 the grand - potential per site is obtained in the form @xmath152 .\ ] ] the above set of nonlinear equations can be solved only numerically .
the trivial disordered solution @xmath153 always exists .
if more real solutions exist , the one with the minimal @xmath154 dominates .
let @xmath155 denote the `` transition '' number of components , such that the direct transition from the disordered to the demixed phase is of second order for @xmath156 and of first order for @xmath157 .
for the bethe lattice of the coordination number @xmath65 , we have the trivial value @xmath158 independent of @xmath78 . plane for the generalized wr model on the bethe lattice with the coordination number @xmath68 and the number of components @xmath159.,scaledwidth=40.0% ] in fig . [
fig : bethe ] , we present the phase diagram in the @xmath160 plane for our generalized wr model on the bethe lattice with the coordination number @xmath68 and the number of components @xmath159 . note that from eq .
( [ defm0 ] ) we have @xmath161 for @xmath73 and @xmath69 for @xmath162 , i.e. there is no crystal phase for @xmath159 components if @xmath163 .
the second - order transition line from the disorder to crystal phases , given by formula ( [ zcl ] ) , is depicted by the solid curve . the second - order transition line from the crystal to disorder phases , given by formula ( [ zcr ] ) ,
is depicted by the dashed curve ; along this line the system exhibits the reentrant phenomenon .
the transitions from the crystal or disordered phases to the demixed phase take place along the dash - dotted curve .
it is seen that for small @xmath164 , the successive order in which the phase transitions take place is basically the same as in the standard wr model with @xmath73 . for weak enough repulsions @xmath165 , the direct transition between the crystal and demixed phases changes into a separate pair of crystal - gas and gas - demixed transitions
; this is an example of a disorder - disorder reentrant transition via an ordered crystal phase .
if the repulsion between unlike species is too weak @xmath163 , the crystal phase disappears from the phase diagram .
on the square lattice , the free energy @xmath166 can be decomposed into four corner transfer matrices @xmath167 , each representing the boltzmann weight of a quadrant of the lattice system @xmath168 @xcite .
the fourth power of the corner transfer matrix can be then interpreted as the density matrix @xmath169 .
the concept of the renormalization can be applied to this density matrix @xcite .
the ctmrg method combines the corner transfer matrix representation of the density matrix , the density matrix renormalization and an accurate approximation of the free energy for a large scale system in terms of an iterative numerical calculation @xcite . during the process
, the space of states is truncated .
the dimension of the truncated space is denoted by @xmath170 ; the larger value of @xmath170 is used , the more accurate results are obtained .
we consider a square of lattice sites with dimension @xmath171 .
there is a central row and @xmath16 additional rows on both sides , so @xmath172 where @xmath16 counts for the number of iterations of the transfer matrix .
when the density matrix @xmath173 is obtained for a sufficiently large system , we calculate the expectation values of microscopic variables at the center of the system , which represents bulk thermodynamics deep inside the system . for example , the spontaneous particle density for our crystal phase is obtained as follows @xmath174 ( -1)^{i+j}\rho^{(n ) } \right\ } , \ ] ] where @xmath175 are the coordinates of the central point and the function @xmath176 if the central site is empty and @xmath177 otherwise .
this expression reproduces correctly @xmath178 as the sublattices @xmath33 and @xmath35 difference @xmath179 . as the order parameter for the demixed phase
, we consider @xmath180 where @xmath181 is the occupation of the particle state @xmath61 and the angle @xmath182 .
the quantity @xmath183 vanishes in the disordered phase with @xmath184 and attains a positive value if one of the components is dominant . in the limit @xmath185 , @xmath186 for any value of @xmath187 and @xmath188 $ ] . in order to detect the position of the phase transitions , special quantities exhibiting singular behavior near the phase transition , like the specific heat
, are usually used . here
, we use the von neumann entropy , defined as @xmath189 close to a second - order critical point , the von neumann entropy can be expressed as @xcite @xmath190 where @xmath191 is the central charge and @xmath192 is the correlation length of the particle system .
consequently , the von neumann entropy has a logarithmic divergence at the critical point . at a first - order transition point
, it exhibits a discontinuity . as concerns critical exponents , we shall calculate the plot of the spontaneous density around the critical point to obtain the exponent @xmath193 .
moreover , in two dimensions and at the critical point , we shall study the finite - size scaling of the particle density @xmath194 to deduce the exponent @xmath195 @xcite .
using , e.g. , the numerical logarithmic derivative @xmath196-\ln[m(l)]}{\ln(l+2)-\ln(l ) } , \ ] ] we can extract the critical exponent @xmath195 as the limit @xmath197 .
components and a large finite repulsion @xmath198 .
the solid and dashed curves correspond to fixed and free boundary conditions , respectively.,scaledwidth=40.0% ] and @xmath198.,scaledwidth=40.0% ] the mc simulations for @xmath73 in ref .
@xcite indicate that at the direct disorder - demixed phase transition @xmath38 , there is no jump in the density for @xmath199 ( so the transition is of second - order ) while there is a jump in density for @xmath200 . in other words , @xmath201 for the standard wr model on the square lattice . for @xmath9 particle components ,
an arbitrarily large ( but finite ) repulsion between different species causes that the disorder - demixed transition becomes of first order .
this fact is documented in fig . [ fig : entropy ] on the plot of the von neumann entropy versus the fugacity for the generalized wr model with @xmath9 components and a large finite repulsion @xmath198 .
the solid ( black ) curve connects data evaluated with boundary conditions fixed at the particle state dominant in the demixed phase while the dashed ( red ) curve connects data for free boundary conditions .
we see that the two curves calculated with distinct boundary conditions produce a hysteresis inside which the disorder and demixed phase coexists , as is usual in the case first - order phase transitions @xcite .
the corresponding plot of the order parameter @xmath183 ( [ md ] ) is presented in fig .
[ fig : magnetization ] . for @xmath202 components
, the second - order phase transition takes place also for small @xmath203 . versus the fugacity @xmath1 for the wr model with @xmath204 particle components
the solid curve corresponds to an infinite @xmath205 repulsion between unlike species , the dashed curve corresponds to a finite repulsion @xmath206 .
dimension of truncated space of states is @xmath207.,scaledwidth=40.0% ] ( left fragment ) and the demixed order parameter @xmath183 ( right fragment ) for the widom - rowlinson model with @xmath204 components .
the solid and dashed curves correspond to @xmath205 and @xmath206 , respectively .
@xmath208.,scaledwidth=40.0% ] the mc simulations for @xmath73 in ref .
@xcite indicate that the standard wr model on the square lattice exhibits the crystal phase at and beyond @xmath69 particle components . in fig .
[ fig : tetra1 ] , we plot by solid curve the von neumann entropy @xmath209 versus the fugacity @xmath1 for that standard @xmath73 wr model with @xmath204 components .
there are two singularities .
the logarithmic one on the left indicates the continuous ( second - order ) disorder - crystal phase transition , the right one indicates the discontinuous ( first - order ) crystal - demixed phase transition . for the generalized wr model with @xmath206 ( dashed curve )
, there are three singularities .
the one in the middle corresponds to the reentrant transition from the crystal to gas phase .
the corresponding plots of the order parameters @xmath178 ( [ mc ] ) and @xmath183 ( [ md ] ) are presented in fig .
[ fig : tetra2 ] .
plane for the generalized wr model with @xmath204 ( open symbols ) and @xmath210 ( full symbols ) components .
@xmath211.,scaledwidth=40.0% ] in fig .
[ fig : tetra4 ] , the phase diagrams in the @xmath160 plane are presented for the generalized wr model with @xmath204 ( open symbols ) and @xmath210 ( full symbols ) components .
the disorder - crystal and crystal - disorder phase transitions are represented by circles , the crystal - demixed and disorder - demixed phase transitions are denoted by squares .
the phase diagrams resemble the one for the bethe lattice in fig .
[ fig : bethe ] .
for which the crystal phase exists in the generalized wr model on the square lattice with @xmath0 components .
the dashed curve corresponds to the exact bethe result ( [ jmax ] ) for the coordination number @xmath68 .
the solid curve is the ( inverse ) polynomial fit of the numerical ctmrg data ( solid circles ) which converges to @xmath212 as @xmath213 .
@xmath211.,scaledwidth=40.0% ] for the generalized @xmath0-component wr model on the bethe lattice with the coordination number @xmath65 in sec .
ii , we have defined the maximal repulsion boltzmann factor @xmath139 , for which the crystal phase exists . for @xmath68 ,
the exact formula ( [ jmax ] ) implies @xmath214 the corresponding dependence of @xmath139 on the number of components @xmath0 for the bethe lattice with @xmath68 is represented in fig .
[ fig : octo2 ] by dashed curve .
the solid curve is the ( inverse ) polynomial fit of the numerical ctmrg data for the square lattice ( solid circles ) . in the asymptotic limit @xmath213
, the fit implies that @xmath215 components .
the dependence of @xmath216 ( [ ceff ] ) on the system size @xmath217 .
the solid curve stands for the standard @xmath73 wr model at @xmath218 , the dashed curve corresponds to @xmath219 and finally the dash - dotted curve corresponds to @xmath220 . as @xmath221 ,
all three curves converge to the central charge @xmath222 of the ising universality class .
@xmath223.,scaledwidth=40.0% ] at the critical point , the correlation length of the particle system @xmath192 diverges in the bulk and @xmath224 for a finite system of characteristic length @xmath217 . according to eq .
( [ eq : cko ] ) , defining @xmath225 the central charge @xmath191 of the critical model is obtained as the limit @xmath226 . in
what follows , we shall restrict ourselves to the wr model on the square lattice with @xmath159 components . for @xmath73 , there is the only critical point @xmath218 at which the disorder - crystal phase transition takes place . for @xmath206
, there is the disorder - crystal phase transition at @xmath227 and the reentrant crystal - disorder phase transition at @xmath228 . in fig .
[ fig : octo4 ] , we plot the dependence of @xmath216 on the system size @xmath217 for the above three critical points : the solid curve stands for @xmath229 , the dashed curve corresponds to @xmath219 and the dash - dotted curve corresponds to @xmath220 . it is seen that as @xmath217 goes to infinity , all three curves converge to the value @xmath222 which is the central charge of the ising universality class .
note that due to relatively large values of @xmath217 we have to increase the dimension of the truncated space to @xmath223 .
components . the dependence of @xmath230 ( [ eq : etaeff ] ) on the system size @xmath217 .
the solid , dashed and dash - dotted curves stand for the @xmath229 , @xmath219 and @xmath220 critical points , respectively .
all three curves converge to the ising value @xmath231 .
@xmath223.,scaledwidth=40.0% ] to confirm the ising universality class , we have determined the critical index @xmath195 by studying the convergence of @xmath230 , given by formula ( [ eq : etaeff ] ) , as @xmath221 . the results for the wr model with @xmath159 components at the three considered critical points are presented in fig .
[ fig : octo3 ] . as
before , the solid , dashed and dash - dotted curves stands for the @xmath229 , @xmath219 and @xmath220 critical points , respectively . as @xmath221 ,
all three curves converge close to the 2d ising value @xmath231 .
components and @xmath206 .
the plot of @xmath232 as the function of @xmath233 at the critical points @xmath228 ( left data ) and @xmath227 ( right data ) .
the linear dependence confirms the ising critical exponent @xmath234 .
@xmath235.,scaledwidth=40.0% ] as concerns the crystal order parameter @xmath178 , we anticipate the ising behavior @xmath236 close to the critical temperature @xmath237 .
since by definition the fugacity @xmath238 with @xmath239 being the chemical potential , it holds that @xmath240 .
we can rewrite the relation ( [ ising ] ) as follows @xmath241 for the generalized wr model with @xmath159 components and @xmath206 , the plot of @xmath232 as the function of @xmath233 is drawn in fig .
[ fig : octo5 ] ; the left data set corresponds to the critical point @xmath228 and the right data set corresponds to the critical point @xmath227 .
we see that in both cases the plot is linear which confirms the ising - like behavior ( [ ising ] ) with the critical index @xmath242 .
the knowledge of two critical exponents @xmath195 and @xmath193 determines all remaining critical indices via the scaling relations @xcite .
the lattice version of the multicomponent wr model acquired a great deal of interest because it exhibits , besides the usual demixed phase transition , also the crystal phase driven by entropy .
the initial extremely large rigorous estimates of the minimum ( number of components to have the crystal ) @xmath3 were later replaced by much smaller values , 4 for the bethe lattice with coordination @xmath68 and 7 for the 2d square lattice @xcite , making the wr model of practical interest . in the original wr model ,
the nearest - neighbor interaction of unlike species corresponds to an infinite hard - core potential , with the corresponding boltzmann factor @xmath73 . in real physical systems with a a finite repulsion potential ,
@xmath78 is always positive . in this paper
, we generalized the multicomponent wr model to an arbitrary repulsion @xmath188 $ ] .
the model was studied in the region close to @xmath73 , namely by using its exact solution on the bethe lattice ( with the coordination number @xmath68 ) and the numerical ctmrg technique for the square lattice .
the consideration of nonzero @xmath78 has two fundamental effects on the phase diagram . for @xmath2 , there is a direct transition between the disordered and demixed phases . in the original wr model with @xmath73 ,
the transition is of second order for @xmath243 .
it was shown here that for @xmath9 the consideration of a very small value of @xmath198 changes this transition to the first - order one , see figs .
[ fig : entropy ] and [ fig : magnetization ] . for @xmath202 , the second order of the disorder - demixed transition remains unchanged for small @xmath78 .
the second fundamental effect concerns the crystal phase . in the original wr model with @xmath73 , if @xmath6 there are two phase transitions as the fugacity @xmath1 goes from 0 to @xmath4 : the disorder - crystal one at @xmath244 and the crystal - demixed one at @xmath38 . in the generalized wr model with @xmath6 ,
there exists certain interval of @xmath78-values in which the system undergoes three phase transitions : the disorder - crystal one at @xmath245 , the crystal - disorder one at @xmath130 and finally the disorder - demixed one at @xmath38 .
this reentrant disorder - disorder phenomenon is interesting not only from an academic point of view .
if the repulsion between unlike species is too weak , the crystal phase disappears from the phase diagram . for all
studied critical points , the generalized wr model belongs to the ising universality class with the central charge @xmath222 and the critical indices @xmath242 and @xmath231 . | in the lattice version of the multicomponent widom - rowlinson ( wr ) model , each site can be either empty or singly occupied by one of @xmath0 different particles , all species having the same fugacity @xmath1 .
the only nonzero interaction potential is a nearest - neighbor hard - core exclusion between unlike particles . for @xmath2 with
some minimum @xmath3 dependent on the lattice structure , as @xmath1 increases from 0 to @xmath4 there is a direct transition from the disordered ( gas ) phase to a demixed ( liquid ) phase with one majority component at @xmath5 . if @xmath6 , there is an intermediate ordered `` crystal phase '' ( composed of two nonequivalent even and odd sublattices ) for @xmath1 lying between @xmath7 and @xmath8 which is driven by entropy .
we generalize the multicomponent wr model by replacing the hard - core exclusion between unlike particles by more realistic large ( but finite ) repulsion .
the model is solved exactly on the bethe lattice with an arbitrary coordination number .
the numerical calculations , based on the corner transfer matrix renormalization group , are performed for the two - dimensional square lattice .
the results for @xmath9 indicate that the second - order phase transitions from the disordered gas to the demixed phase become of first order , for an arbitrarily large finite repulsion . the results for @xmath6 show that
, as the repulsion weakens , the region of crystal phase diminishes itself . for weak enough repulsions , the direct transition between the crystal and demixed phases changes into a separate pair of crystal - gas and gas - demixed transitions
; this is an example of a disorder - disorder reentrant transition via an ordered crystal phase .
if the repulsion between unlike species is too weak , the crystal phase disappears from the phase diagram .
it is shown that the generalized wr model belongs to the ising universality class . |
ap and bp stars are intermediate - mass main sequence stars that display strong photospheric abundance anomalies .
in particular , iron peak elements can be strongly overabundant compared to solar : fe by + 1 dex and cr up to + 3 dex .
si can also be overabundant by + 1 dex and rare earth elements , such as nd and eu , can be even more strongly over abundant in the star s photosphere .
in addition to these strong peculiarities , ap and bp stars also possess strong magnetic fields ( @xmath1 kg ) that are ordered on large scales ; these are predominantly dipolar in structure .
this is remarkable since intermediate mass stars do not possess the deep convective envelopes needed to power a conventional dynamo , as seen in lower mass stars .
there certainly exists a link between the strong magnetic fields and the observed chemical peculiarities . in rough terms , the abundance anomalies are a result of the competing effects of radiative levitation and gravitational settling , producing a chemically stratified atmosphere
. elements with a local accumulation near the surface of the star appear overabundant .
the magnetic field is thought to provide additional stability to the atmosphere allowing horizontal inhomogeneities and vertical stratification to persist .
however , this picture is still incomplete .
the details of the stratification process and the effects of the magnetic field are not understood .
as well , the origin of the magnetic field remains a mystery .
perhaps most glaringly , the evolution of magnetic fields and chemical anomalies in ap / bp stars during the main sequence phase remains unknown .
significant effort has gone into examining specific ap and bp stars : modelling chemical abundances , magnetic fields , and non - uniform distributions of elements both vertically and horizontally across a star s surface . despite a growing sample of
well modelled ap / bp stars , our ability to describe general trends as these stars age is limited .
determining the age of an isolated star to a high degree of accuracy can be difficult .
moreover , ap / bp stars display abnormal photometric properties , which can greatly exacerbate errors in the dating of a star .
( see bagnulo et al .
2006 for further discussion of these difficulties . )
consequently , detailed analysis of ap / bp stars ( often referred to as simply ap stars ) in open clusters can be of great benefit .
the large sample of normal stars in an open cluster enables one to determine a much more accurate age then is possible for a lone field ap star , which can then be applied to all stars in the cluster . with accurate ages and detailed properties , a proper description of the evolution of ap stars
can be obtained .
a few papers have dealt with chemical abundances analysis of intermediate mass stars in open clusters .
burkhart and coupry ( 1998 ) examined the praesepe open cluster and found a number of am stars .
varenne and monier ( 1999 ) investigated a large sample of stars in the hyades cluster and reported abundances for normal a and f stars , as well as a number of chemically peculiar stars .
sttz et al .
( 2006 ) looked at 5 stars in the open cluster ic 2391 and found some evidence for chemical peculiarity in one . despite this beginning ,
much work remains before we can properly describe the evolution of ap stars .
this paper focuses on one cluster in particular , the 220 myr old ngc 6475 .
membership is confirmed and a detailed abundance analysis is presented for four probable ap members of this cluster , as well as for one normal b member .
ngc 6475 ( m7 ) is a well - known open cluster in the constellation scorpius .
the cluster has a heliocentric distance of @xmath2 pc ( robichon et al .
1999 ) . a logarithmic age of @xmath3 y was determined by meynet et al .
( 1993 ) using the cluster ubv colour - magnitude diagram
. this cluster age was confirmed , and the age uncertainty derived , by fitting isochrones to the h - r diagram positions of the cluster stars .
photometric data for the cluster stars was obtained from the webda on - line database ( mermilliod , 1995 ) . in general this list
was not checked to confirm membership , however a few of the more important objects ( eg .
red giants ) were investigated .
isochrones were generated by linear interpolation of the tabulated evolutionary model calculations of schaller et al .
( 1992 ) , for standard mass loss rate and a metallicity z=0.020 . with a focus on
the 2 confirmed red giant members of ngc 6475 , the age of the cluster was constrained to be @xmath3 logarithmic years . as a comparison , kharchenko et al .
( 2005 ) derive ages for nearly 200 clusters , including ngc 6475 .
their method involves fitting isochrones to individual cluster members and then determining an average cluster age .
kharchenko et al .
determined a logarithmic age of 8.22 for ngc 6475 and they quote a literature value ( from a private communication by a. v. loktin in 2004 ) of 8.48 .
these two values are in agreement with our values , within the total estimated error bar .
there have been several modern studies of the cluster s apparent motion : van leeuwen ( 1999 ) and robichon et al .
( 1999 ) both used hipparcos data to derive the cluster proper motion , while dias et al .
( 2001 ) employed tycho-2 data .
van leeuwen ( 1999 ) reports proper motion in right ascension @xmath4 masyr@xmath5 , proper motion in declination @xmath6 masyr@xmath5 , and parallax @xmath7 mas .
robichon et al .
( 1999 ) report @xmath8 masyr@xmath5 , @xmath9 masyr@xmath5 , and @xmath10 mas .
dias et al . (
2001 ) report @xmath11 masyr@xmath5 and @xmath12 masyr@xmath5 .
these values are all in mutual agreement .
cluster radial velocity was investigated by gieseking ( 1985 ) , who found @xmath13 kms@xmath5 , and by robichon et al .
( 1999 ) , who found @xmath14 kms@xmath5 .
five stars in ngc 6475 were analysed in this study : hd 162305 ( b9 ) , hd 162576 ( b9p ) , hd 162725 ( b9p ) , hd 320764 ( a6 ) , and hd 162817 ( b9 ) .
the spectral classifications are from renson s catalogue ( 1991 ) . proper motion and parallax data
were found for hd 162576 , hd 162725 , and hd 162817 in the hipparcos catalogue ( esa 1997 ) .
additional proper motion data were found for all five stars in the tycho-2 data base ( hg et al .
additional proper motion data for all five stars were found in bastian and rser s positions and proper motions - south catalogue ( rser & bastian , 1993 ) . although consistent with the results of the former studies , the ppm - s data are significantly less precise and will not be considered further .
the kinematic data for individual stars , which are consistent with the cluster motion discussed in sect .
2.1 , are reported in table [ other_kin ] .
radial velocities were reported by gieseking ( 1985 ) for all 5 stars . additionally , radial velocities were measured from our own observed spectra . during the synthetic spectrum
fitting procedure radial velocity was considered to be a free parameter .
the quoted errors take into account noise in the observed spectrum , uncertainties in @xmath0 , and unmodelled blended lines .
the measurements of gieseking are in reasonable agreement with our own radial velocities , and all are consistent with the cluster radial velocities from gieseking ( 1985 ) and robichon et al .
( 1999 ) based on this analysis , we conclude that all five stars are high probability members of ngc 6475 .
cccccccc & & & gieseking & this work + star & @xmath15 masyr@xmath5 & @xmath16 masyr@xmath5 & @xmath15 masyr@xmath5 & @xmath16 masyr@xmath5 & @xmath17 masyr@xmath5 & @xmath18 kms@xmath5 & @xmath18 kms@xmath5 + hd 162305 & @xmath19 & @xmath20 & @xmath21 & @xmath21 & @xmath22 & @xmath23 & @xmath24 + hd 162576 & @xmath25 & @xmath26 & @xmath27 & @xmath28 & @xmath29 & @xmath30 & @xmath31 + hd 162725 & @xmath32 & @xmath33 & @xmath34 & @xmath35 & @xmath36 & @xmath37 & @xmath38 + hd 320764 & @xmath39 & @xmath40 & @xmath21 & @xmath21 & @xmath41 & @xmath42 & @xmath43 + hd 162817 & @xmath44 & @xmath45 & @xmath46 & @xmath47 & @xmath48 & @xmath49 & @xmath50 + ngc 6475 & @xmath51 & @xmath52 & @xmath53 & @xmath54 & @xmath55 & @xmath13 & @xmath21 +
three reddening - corrected methods were used to calculate the effective temperature of the four probable ap stars .
uvby@xmath56 and geneva photometry were obtained for each star from the general catalogue of photometric data ( mermilliod at al .
a first determination of the effective temperature and surface gravity were obtained using templogg ( sttz et al .
this software package attempts to automatically correct for reddening .
a second determination was obtained using geneva photometry and the calibration of north and nicolet ( 1990 ) , assuming their [ m / h](z ) = 1 coefficients .
finally , the hauck and north ( 1992 ) recalibration of the north and nicolet ( 1990 ) calibration , adapted specifically for ap stars , was used .
this gave a third effective temperature , although no corresponding @xmath57 value .
these methods make use of the reddening free _ x _ and _ y _ parameters .
the resulting values from the three calibrations were then averaged to give the adopted temperature and surface gravity .
the standard deviations were adopted as the uncertainties of the average values .
the uncertainty values for our effective temperatures may be overly optimistic .
we have implicitly assumed that the calibrations used are independent , and that there are no systematic errors in the resulting temperatures .
additionally , we have only sampled a small number of calibrations .
a more conservative estimate might be @xmath58 k in @xmath59 .
however , lacking evidence that our uncertainties are incorrect , we will proceed with the derived @xmath59 values reported in table [ basic_properties ] . for the normal b9.5 star hd 162817 , templogg and the calibration of north and nicolet ( 1990 ) were used to determine atmospheric properties .
( the calibration of hauck and north was not appropriate for this presumably chemically normal star . ) as an additional check of the adopted @xmath59 and @xmath57 values , synthetic atlas9 balmer line profiles with corresponding temperatures and gravities were compared with the observed h@xmath60 profiles .
good agreement was found .
spectrum synthesis for hd 162817 with these values also provides a better fit than models adjusted in @xmath59 by 500 k or in @xmath57 by 0.5 . in particular
, lines produced by different ionization states of the same element could be more accurately fit at this temperature and @xmath57 .
the stars were placed on an h - r diagram based on their effective temperature and bolometric corrected luminosity .
the cluster parallax from robichon et al .
( 1999 ) and the bolometric correction of balona ( 1994 ) were used to compute the luminosities .
the h - r diagram is presented in figure [ h - r ] , showing the derived positions of the stars as well as the cluster isochrone ( 8.35 log years ) and evolutionary tracks from 2 to 4 @xmath61 ( schaller et al .
1992 ) . from the diagram we estimate masses for the five stars , which are shown in table [ basic_properties ] .
radii for the stars are estimated from effective temperature and luminosity , and included in table [ basic_properties ] .
the fraction of the main sequence lifetime elapsed for each star , @xmath62 , is also given in table [ basic_properties ] .
the @xmath62 values are calculated using the cluster s age and the evolutionary track corresponding to the star s mass , with uncertainties taking into account the uncertainty on the stellar mass and cluster age . finally , we determined an age for each star based on its h - r diagram position .
these ages are fully consistent with the cluster age derived above .
the uncertainties of these values are based on the range of isochrones that pass through the rectangle described by the star s position error bars .
departures from standard mass loss and z=0.02 are not considered .
while the error bars on some of the stars are small , one must note that these objects are nearing the terminal age main sequence ( tams ) line , where the isochrones strongly diverge .
thus , for some of our stars , we derive an unusually precise age based on h - r diagram position . despite this relative precision
, we consider the cluster age more definitive , and use it where possible .
if we were to use a more conservative uncertainty on @xmath59 of 500 k , the uncertainties on @xmath63 would increase by about 0.01 , mass would be approximately 0.05 @xmath61 more uncertain , radius 0.1 @xmath64 more uncertain , and @xmath62 0.05 more uncertain
. the h - r diagram age would be 0.05 log yr more uncertain for our more precise limits , up to 0.2 log yr more uncertain for our least precise values . in principle
, the cluster age can be used to constrain the h - r diagram location of cluster members , providing accurate masses and luminosities .
however , the divergence of isochrones toward the tams leads to large uncertainties in the derived properties of more massive cluster members using this method . in the case of our sample of five stars
this method did not provide any advantage .
the h - r diagram positions of the stars are already well constrained , thanks to the relatively precise cluster parallax .
while no more precise , the luminosities and masses determined using the cluster age are in agreement with the values presented in table [ basic_properties ] .
, based on the calculations of schaller et al .
( 1992 ) for standard mass loss and z=0.02 .
evolutionary tracks are labelled by mass in @xmath61 .
, width=326 ] cccccccc star & temperature ( k ) & @xmath57 & @xmath63 & mass ( @xmath61 ) & radius ( @xmath64 ) & @xmath62 & h - r age ( log yr ) + hd 162305 & @xmath65 & @xmath66 & @xmath67 & @xmath68 & @xmath69 & @xmath70 & @xmath71 + hd 162576 & @xmath72 & @xmath73 & @xmath74 & @xmath75 & @xmath76 & @xmath77 & @xmath78 + hd 162725 & @xmath79 & @xmath80 & @xmath81 & @xmath82 & @xmath83 & @xmath77 & @xmath84 + hd 320764 & @xmath85 & @xmath86 & @xmath87 & @xmath88 & @xmath89 & @xmath90 & @xmath91 + hd 162817 & @xmath92 & @xmath93 & @xmath94 & @xmath95 & @xmath96 & @xmath97 & @xmath98 +
the observed spectra employed in this investigation were obtained within the context of the paranal observatory project ( pop ) program ( eso ddt program i d 266.d-5655 ) , using the uves high - resolution spectrograph at eso - vlt .
this project aims to collect a library of high resolution spectra which are readily available to the public ( bagnulo et al . , 2003 ) .
a large number of field stars have been observed by the uves - pop , as well as stars in two open clusters : ngc 6475 and ic 2391 . in total about 80 cluster stars and
about 330 field stars have been observed .
the spectra cover about 300 nm to 1000 nm , with two small gaps , at a resolving power of approximately 80000 .
the majority of the data reduction was performed by an automated pipeline developed for the uves instrument .
parameters were tailored for the pop , using the `` average extraction '' method instead of the `` optimal extraction '' method .
this pipeline automatically considers bias and flat field corrections , and wavelength calibration .
the pipeline also attempts to correct for bad ccd pixels and subtracts the background sky from the spectrum . for a further description of the pipeline see ballester et al .
once downloaded from the pop internet site , the individual spectral orders were continuum normalized with the aid of the continuum routine in the image reduction and analysis facility ( iraf ) using a 5th to 10th order legendre polynomial depending on the order of the spectrum .
the individual orders were then merged , and the correction to the heliocentric rest frame was calculated using the iraf rvcorrect routine .
the uves pop spectra are provided in both merged and unmerged ` 2d ' forms , however we found the unmerged form to be more useful .
the merged spectra display an artifact , apparently due to the merging process : there is a slight ripple in the continuum level .
thus we prefer the unmerged ` 2d ' spectra which contain the original unmerged echelle orders .
the observed spectra were modelled using the zeeman2 ( landstreet 1988 ; wade et al .
2001 ) spectrum synthesis code .
chemical abundances and projected rotational velocities @xmath0 were determined by direct fitting of the synthesized spectra . for each of the five stars ,
several spectral regions were examined .
uniform vertical and horizontal abundance distributions were assumed . while a uniform horizontal distribution is almost certainly not the case for some elements in some stars , with only one observation per star , it is impossible to accurately model surface distributions . for all ap / bp stars ,
a dipolar photospheric magnetic field was assumed , with an intensity of 1 kg , oriented with the positive magnetic pole pointing directly along the line - of - sight toward the observer .
no positive detections of the magnetic fields of the ap star members of ngc 6475 are available in the literature .
bagnulo et al .
( 2006 ) obtain null results for the longitudinal fields of hd 162305 , hd 162725 and hd 320764 with 1@xmath99 error bars of about 50 g. however , the longitudinal magnetic fields of ap stars are variable , and have a typical rms of only about 300 g ( bohlender & landstreet 1990 )
. given the small number of magnetic observations ( just one for hd 162725 and hd 320764 , and two for hd 162305 ) the stars could quite probably have been observed when the visible field configuration produced a longitudinal magnetic field below the detection threshold .
we therefore assume that these are typical " magnetic ap stars with surface field strengths of about 1 kg . inclusion of this weak magnetic field introduces magnetic desaturation , and therefore decreases slightly ( by about 0.05 dex ) the abundances inferred from most lines . for the normal star hd 162817
it was assumed that there was no magnetic field .
microturbulence was also assumed to be absent in the atmosphere of this star . while this may not be strictly correct for a star with @xmath59 10000 k ,
there are no discrepancies between our synthetic and observed spectra of the type one would expect if there were a significant microturbulence .
both weaker and stronger lines of the same species are well fit with a single abundance value . however , due to the higher projected rotational velocities of this star and our program stars , few ( if any ) weak lines ( on the linear part of the curve of growth ) are actually detected and modelled .
therefore undetected microturbulence may exist , but would have an impact , at most , at the level of our uncertainty in abundance .
for example a 1 km / s microturbulence would decrease our abundance values for this star by at most 0.1 dex . for all program stars except hd 320764
, microturbulence was assumed to be suppressed , either directly or indirectly , by the presence of a magnetic field .
therefore , no microturbulence was assumed in computing the line profiles of hd 162305 , hd 162576 , and hd 162725 . for hd 320764
a 1 kms@xmath5 microturbulence was assumed . in this star
, it proved impossible to determine a microturbulence observationally .
the very large rotational broadening makes finding the necessary mix of weak and strong lines of the same species impossible .
however , for a normal a star with a temperature of 8800 k we expect there to be a small microturbulence ( landstreet 1998 ) , so a conservative value of 1 kms@xmath5 was chosen .
input atomic line data was extracted from the vienna atomic line database ( kupka et at .
an `` extract stellar '' request was used with the @xmath59 and @xmath57 values from table [ basic_properties ] . 1 dex overabundances of si , cr , fe , and 3 dex overabundances all rare earth elements were used to ensure a complete line list . a 2 kms@xmath5 microturbulence was used to simulate a weak magnetic field . only lines with a central line depth
greater then 0.01 , as a fraction of the continuum , were used .
the list of lines and atomic data employed is provided in table 3 ( available only in electronic form ) .
the general spectrum fitting strategy involved iteratively synthesizing a spectrum , examining the fit by eye , and then adjusting the input model parameters accordingly . in some more complicated cases , for stars with line profile shapes which depart significantly from the rotationally - broadened model ( probably due to surface abundance non - uniformities ) , correspondence between model and observed equivalent widths was also verified .
uncertainties on the final @xmath0 and abundance values are based on the dispersion of values determined from many absorption lines .
if few absorption lines of an element are present in the observed spectrum , the uncertainty includes a conservative estimate for the change in abundance necessary to shift the synthesized spectrum well above the noise level , and any possible normalization errors .
thus , uncertainty values presented from spectrum synthesis correspond to a @xmath100 confidence level . the primary motivation for examining hd 162817 was to check that our fitting procedure produced the expected , approximately solar , abundance values for a normal late b - type star .
hd 162817 is a normal b9 star with no known peculiarities and v magnitude of 6.11 . the observed spectrum for this star is of good quality , with a peak signal - to - noise ratio of approximately 280 .
the line profiles of this star are moderately broad , and profiles of unblended lines show no obvious asymmetries . a solar abundance atlas9 model atmosphere ( kurucz 1993 ) with a temperature of 10000 k and @xmath57 of 3.5
was used to model the spectrum of this star . a projected rotational velocity ( @xmath0 ) of @xmath101 kms@xmath5was found .
the mean best fit chemical abundances , derived for 12 elements , are reported in table [ final_ab ] .
illustrative regions of the observed spectrum , compared with the best fit synthetic spectrum , are shown in figure [ 162817_spec ] . for most elements ,
the derived abundances of hd 162817 are consistent with solar values .
in particular , iron peak elements such as cr , normally overabundant in ap stars , display solar abundances .
this is not a surprising result , but it confirms the accuracy of our fitting method .
hempel & holweger ( 2003 ) determined abundances for a number of late b stars including hd 162817 .
they first determined the effective temperature @xmath59=9190 k and @xmath57=3.15 , based on strmgren photometry and the calibration of napiwotzki et al .
( 1993 ) and assumed a microturbulent velocity of @xmath102 kms@xmath5 .
notably , this effective temperature is not consistent with the spectral classification of b9 .
they derived lte abundances , in dex relative to the sun , of o ( 1.1 ) , mg ( 0.06 ) , si ( 0.12 ) , ca ( -0.53 ) , fe ( 0.29 ) , and sr ( -0.07 ) .
our abundances for mg , si , and fe are consistent with theirs within @xmath103 .
we derive no sr abundance for comparison . our ca abundance ( @xmath104 dex above solar ) is greater than that of hempel & holweger by @xmath105 dex .
hempel & holweger provide no uncertainty on their value , but if their uncertainty is similar to ours , then the values may in fact be consistent within @xmath103 .
additionally , their ca abundance value is derived from only one line , reducing our confidence in the value somewhat .
hempel & holweger use ca atomic data extracted from vald , as do we , so the log _ gf _ values used should be the same .
the ca ii k line at 3934 that hempel & holweger use has been known to display the effects of stratification in many ap stars ( babel , 1994 ) , however as hd 162817 appears to be a normal b star this not likely to be the source of the discrepancy . additionally , the ca ii k line is a particularly strong line , which leaves the slight possibility of some undiagnosed non - lte effect reducing the accuracy of an abundance based solely on this line .
hempel & holweger , using the o i @xmath106 triplet , derive a non - lte abundance of 0.63 dex above solar .
this is well above our abundance of solar @xmath107 dex , based on o i @xmath108 . to investigate this discrepancy
, we redetermined our abundances using hempel & holweger s temperature and @xmath57 resulting in an increase in our abundance by about 0.2 dex .
we then determined an abundance for the o i @xmath106 triplet ( assuming lte ) and obtained an abundance 1.3 dex over solar .
there appears to be a large nlte effect in this feature , likely responsible for much of the discrepancy we observe .
if this were fully taken into account , along with the correct temperature , then our results would likely be much more similar .
hd 162576 is classified as b9p sicr , with a v magnitude of 6.99 and has an approximate rotation period of 3.43 days ( catalano & renson 1998 ) .
the catalogue of renson et al .
( 1991 ) indicates that this is a highly probable chemically peculiar star .
the observed spectrum for hd 162576 is of good quality , with a peak signal - to - noise ratio of 300 , and sharp lines that are relatively smooth and unblended .
unfortunately a segment of the spectrum in the red , from about 5750 to about 6700 , was of unusually poor quality and hence unavailable for the purposes of modelling .
a temperature of 10500 k and a surface gravity of 4.0 were adopted for the solar abundance atlas9 model atmosphere of hd 162576 .
we derived a projected rotational velocity of @xmath109 kms@xmath5 and mean chemical abundances for 17 elements , reported in table [ final_ab ] .
a sample segment of spectrum with its best fit model is shown in figure [ 162576_spec ] .
clear overabundances of si , cr , and mn can be seen while o , mg , and ca are underabundant .
the cores of both the cr and ti lines in this spectrum show some structure .
ti lines are significantly deeper on the blue side of the line then the red .
the cr lines display a slight rise at the very centre of the line .
this subtle structure is likely the result of non - uniform surface abundance distributions of these elements , and their modelling is beyond the scope of this paper .
hd 162725 is a b9p sicr star with a v magnitude of 6.42 and a rotation period of 4.459 days ( renson & catalano 2001 ) .
renson et al .
( 1991 ) list this star as a well - established chemically peculiar star .
this is another sharp - lined star with a high ( 320 ) signal - to - noise ratio
. line profiles of some elements show obvious line profile structure and asymmetry .
a temperature of 9750 k and a surface gravity of 3.5 were adopted for the model atmosphere .
one notable feature of the spectrum is a distinct asymmetry in the chromium lines .
there is a clear depression on the blue side of the line and a peak near line centre .
this suggests that there may be a strongly inhomogeneous distribution of cr across the surface of this star .
our uniform - abundance model provides a relatively poor fit to these complex profiles . in order to derive the abundance of cr
, we verified that the equivalent widths of the observed and synthesized spectral lines were in good agreement .
this technique of matching equivalent widths was used for fe as well , as lines of this element also appear to display slightly asymmetric line profiles .
the resulting mean abundances for 15 elements are reported in table [ final_ab ] .
a sample section of spectrum is shown in figure [ 162725_spec ] .
hd 162725 exhibits strong overabundances of iron peak elements , as well as some rare earth elements such as nd and eu .
si appears overabundant , however o , mg and ca are under abundant .
hd 162305 is classified as a b9 star with a v magnitude of 7.81 .
renson et al .
( 1991 ) list this star as having a doubtful chemically peculiar nature .
it was included in our study , as a likely ap star , due to its large @xmath110 photometric value ( maitzen & floquet , 1981 ) and unusual line profiles .
our observation of this star has a very good peak signal - to - noise ratio of 420 .
hd 162305 has broader lines then the previous 2 stars , and again asymmetries are clearly present in the high - resolution spectrum .
the strong asymmetries are most obvious in the cr lines , there is a large flux increase in the centre of the line profile and both blue and red wings are significantly deeper .
other elements also display asymmetries : fe lines tend to be deeper in the blue wing and ti lines rise significantly in the centre of the core .
an effective temperature of 10500 k and surface gravity of 4.0 were adopted for the atlas9 model atmosphere .
the notably broader lines of this star imply a significantly greater projected rotational velocity .
the calculated @xmath0 of @xmath111 kms@xmath5is roughly three times larger for this star than for the previous two .
since the total amount of light absorbed is independent of rotational broadening , the line depth will decrease as the line width grows . as a result , line cores in the spectrum of hd 162305 are closer to the noise level
. additionally , any asymmetries in the line profile , due to inhomogeneous distributions of elements across the surface of the star , will be much more prominent in the spectrum of a faster rotating star .
these two effects combine to increase the difficulty of the fitting procedure for hd 162305 .
fitting @xmath0 was accomplished by focusing on the wings of a large number of lines ; the core was largely neglected due to line profile structure . in order to determine element abundances , we once again fit the equivalent widths of lines with weak blending .
the best fit abundances are reported in table [ final_ab ] and a sample section of spectrum is shown in figure [ 162305_spec ] . in the photosphere of hd 162305 ,
cr and mn are strongly overabundant , as is nd , while o and mg are underabundant .
the fifth star , hd 320764 , is classified as a6 , and has a v magnitude of 8.93 .
renson et al .
( 1991 ) consider this star to be of a doubtful chemically peculiar nature .
this star was included in our study , as possibly peculiar , do to its larger @xmath110 photometric value ( maitzen & floquet , 1981 ) .
the spectrum of hd 320764 exhibits the broadest lines and a signal - to - noise ratio of 320 . despite the high signal - to - noise ratio ,
the shallow lines of this star make noise more of a problem then in the other stars investigated . a temperature of 8750 k and surface gravity of 4.0 were adopted for the model atmosphere .
a 1 kms@xmath5 microturbulence was used in the spectrum synthesis . if we were to increase this value to 2 kms@xmath5
, this would result in a decrease in most abundances by about 0.2 dex .
the @xmath0 determined by spectrum synthesis is 225 kms@xmath5 .
such a large @xmath0 makes modelling very challenging due to the very shallow and heavily blended lines . in the spectrum of a star with lines
this broad and shallow , continuum normalization becomes a large potential source of error and must be dealt with carefully .
no clear asymmetries are seen in this spectrum . the fitting procedure focused on both the core and wings of the lines , although in this case drawing a distinction between wing and core is not very useful .
it is very unusual for an ap star to have a @xmath0 value this large .
modelling of the spectrum of this star shows near solar abundances for cr , fe , ti and ni with no clear overabundances of rare earth elements .
the abundance of ca appears marginally greater than solar , however only three lines were used , all of which are particularly sensitive to microturbulence .
if we were to use a microturbulence of 2 kms@xmath5 the abundance would fall to @xmath112 dex , well within uncertainty of solar . due to this sensitivity and the uncertainty in the microturbulence of hd 320764 ,
the ca abundance presented here is somewhat more uncertain then that of other elements .
although uncertain , the abundances derived for hd 320764 provide no evidence to suggest that the star is chemically peculiar .
the derived abundances are reported in table [ final_ab ] and a sample section of spectrum is shown in figure [ 320764_spec ] .
ccccccc & hd 162576 & hd 162725 & hd 162305 & hd 320764 & hd 162817 & solar + @xmath59 ( k ) & @xmath72 & @xmath79 & @xmath113 & @xmath114 & @xmath115 & + mass ( @xmath61 ) & @xmath116 & @xmath117 & @xmath118 & @xmath119 & @xmath120 & + @xmath0 ( kms@xmath5 ) & 28 @xmath121 3 & 32 @xmath121 5 & 85 @xmath121 5 & 225 @xmath121 10 & 79 @xmath121 3 & + c & -5.3 @xmath121 0.2 & & & -3.7 @xmath121 0.4 & & -3.48 + o & -3.5 @xmath121 0.3 & -3.8 @xmath121 0.2 & -3.8 @xmath121 0.2 & -3.5 @xmath121 0.3 & -3.1 @xmath121 0.1 & -3.17 + ne & & -3.5 @xmath121 0.5 & -3.5 @xmath121 0.4 & & & -3.92 + mg & -5.5 @xmath121 0.1 & -5.2 @xmath121 0.2 & -5.2 @xmath121 0.2 & -4.8 @xmath121 0.3 & -4.45 @xmath121 0.1 & -4.42 + si & -3.7 @xmath121 0.2 & -4.1 @xmath121 0.4 & -4.3 @xmath121 0.3 & -4.4 @xmath121 0.4 & -4.6
@xmath121 0.5 & -4.45 + ca & -6.55 @xmath121 0.2 & -6.8 @xmath121 0.4 & -5.1 @xmath121 0.5 & -5.0 @xmath121 0.5 & -5.4 @xmath121 0.4 & -5.64 + sc & -9.1 @xmath121 0.2 & & & & & -8.83 + ti & -7.0 @xmath121 0.3 & -7.6 @xmath121 0.3 & -7.3
@xmath121 0.3 & -6.8 @xmath121 0.4 & -7.0 @xmath121 0.3 & -6.98 + v & & & & & @xmath122 -7.7 & -8.04 + cr & -5.0 @xmath121 0.3 & -4.5 @xmath121 0.3 & -5.1 @xmath121 0.2 & -6.0 @xmath121 0.4 & -6.2 @xmath121 0.3 & -6.33 + mn & -5.1 @xmath121 0.2 & -4.8 @xmath121 0.3 & -5.4 @xmath121 0.2 & -5.5 @xmath121 1.0 & -5.9 @xmath121 0.5 & -6.61 + fe & -4.2 @xmath121 0.3 & -3.7 @xmath121 0.3 & -4.0 @xmath121 0.2 & -4.5 @xmath121 0.3 & -4.3 @xmath121 0.2 & -4.50 + co & & & & -7.1 @xmath121 0.7 & @xmath122 -6.5 & -7.08 + ni & -5.6 @xmath121 0.5 & -5.6 @xmath121 0.5 & -5.9 @xmath121 0.3 & -5.9 @xmath121 0.5 & -5.3 @xmath121 0.5 & -5.75 + ba & -9.7 @xmath121 0.3 & -9.0 @xmath121 0.4 & -8.3 @xmath121 0.7 & @xmath122 -9 & -10.0 @xmath121 0.5 & -9.87 + la & @xmath122 -8.4 & @xmath122 -8.5 & @xmath122 -8.5 & @xmath122 -9.5 & & -10.83 + ce & @xmath122 -8 & & & @xmath122 -9 & & -10.42 + pr & & @xmath122 -8.3 & & @xmath122 -8 & & -11.29 + nd & -8.4 @xmath121 0.3 & -8.0 @xmath121 0.3 & -8.5 @xmath121 0.7 & @xmath122 -8.5 & & -10.50 + sm & @xmath122 -8.5 & & & @xmath122 -9 & & -10.99 + eu & @xmath122 -9.5 & -9.5 @xmath121 0.5 & @xmath122 -9 & @xmath122 -9 & & -11.49 +
chemical abundances as well as projected rotation velocity have been determined for five confirmed members of ncg 6475 ( m7 ) .
the derived chemical abundances are presented graphically in figure [ total_ab ] .
three stars are clearly classical ap / bp stars , and the normal b9 star has been confirmed to have solar abundances .
however , the supposed ap star hd 320764 appears to be a chemically normal a star .
abundances and abundance upper limits have been found for a wide range of elements from c to eu , with typical @xmath103 uncertainties ( generally derived from the dispersion of abundances determined from multiple absorption lines ) of 0.3 dex .
the chemically normal b star hd 162817 displays abundances that are within uncertainty of solar for essentially all elements .
the three confirmed ap stars all show strong overabundances ( as compared to the sun and as compared to hd 162817 ) of cr and mn .
only marginal overabundances are found for si , fe and ni , while ti is solar or marginally underabundant .
the rare earths nd and eu are also overabundant when detected in the ap stars .
the light elements o and mg are slightly underabundant .
hd 320764 displays abundances that are within uncertainty of solar for all elements detected , except for ca and mn .
mn is nearly within uncertainty of solar and based only on one spectral window .
ca is also nearly within uncertainty of solar and the lines used are particularly sensitive to the choice of microturbulence , making the abundance somewhat less certain than that of other elements studied . the structure and strong asymmetries observed in many lines of hd 162576 , hd 162725 and hd 162305 support the conclusion that these stars are magnetic ap / bp stars .
complex line profiles including many double - peaked profiles imply complicated surface abundance distributions .
the abundances presented in this paper therefore represent averages of these complex distributions over the observable hemisphere of star at the time of observation .
additional high resolution spectra of these three stars are necessary if the surface distributions are to be mapped . the overabundances of cr and eu are marginally greater for the cooler ap star hd 162725 than for the two hotter ap stars , consistent with the cooler temperatures seen in srcreu ap stars when compared to si ap stars in the field ( wolff 1967 ) .
the values of cr overabundance seen , and the trend with temperature , are consistent with the pattern observed by ryabchikova et al .
there is no clear relationship between overabundance of elements and rotation within the three chemically peculiar stars .
however , it is notable that the chemically normal stars on average have significantly higher @xmath0 then the ap stars .
this is consistent with published results , for example abt & morrell ( 1995 ) .
hd 162576 and hd 162725 have identical age and initial composition , as well as the same @xmath0 , mass , and their derived effective temperatures are only marginally different ( @xmath123 k ) .
a period analysis shows that the stars have similar periods at 3.43 and 4.459 days respectively ( catalano & renson 1998 and renson & catalano 2001 ) . with our adopted uncertainties ,
the abundances of most elements are very similar in these two stars .
there are some marginally detectable differences : si appears to be less abundant in hd 162725 then in hd 162576 , while ba is more abundant .
cr and fe may be more abundant in hd 162725 as well .
the similarities in derived abundances between these two stars suggest that age , mass , rotation speed , and possibly magnetic field strength may be enough to uniquely determine a set of chemical peculiarities present in a late - b star .
the slight differences in abundance between the two stars may be related to the difference in rotation period , phase of rotation , or possibly magnetic field , between the two stars .
hd 162305 and hd 162576 have identical age , initial composition , effective temperatures , and masses , but @xmath0 differs by more than a factor of 3 .
these stars display many similarities , but there are some significant abundance differences ( e.g. @xmath124 dex for ca , @xmath125 dex for ba ) . assuming that the true rotational velocities of these stars differ by a similar factor , it could be that these differences reflect the effects of rotational mixing .
hd 162305 is also nearly identical to hd 162725 except in @xmath0 .
both stars have identical age and initial chemical composition , as well as very similar masses and temperatures .
there are many similarities between these two stars , as with hd 162576 , but there are also some significant differences , particularly the @xmath126 dex difference in ca , and smaller differences in cr and mn .
a study of the rotation periods of the five stars in this paper would be very useful , as it would allow for much more definite comments on the effect of rotational mixing in this sample of stars .
even more remarkably , hd 162305 and hd 162817 ( our chemically normal comparison star ) have formally identical @xmath0 and only marginally different effective temperatures , as well as identical age and initial chemical composition . however , these stars display strong differences in the photospheric abundances of chemical elements ( e.g. @xmath127 dex for cr , @xmath128 dex for ba ) .
presumably the difference results from the presence of a magnetic field , but this begs the question : what distinguishes the presence of magnetic field in co - eval , co - environmental stars with similar masses , temperatures and rotation speeds ? some major questions about ap stars remain .
in particular how their magnetic fields , rotation and chemical abundances evolve over time is not well understood .
temporal changes in rotation speed and abundance patterns potentially can provide valuable insights into the underlying mechanisms producing the abundance anomalies and magnetic fields seen in ap stars . in that context , this paper serves as a first step in a larger effort to examine trends in the evolution of ap stars .
the use of cluster stars allows us , in principle , to remove the competing effects of age and environment , and allows for precise determination of mass and evolutionary state , particularly for stars in the earlier half of their lifetime . through the use of a number of clusters with different ages
, one can construct a picture of evolutionary trends .
thus the detailed chemical abundances of all ap stars , within a single cluster , presented in this study provides important groundwork for a description of the formation and evolution of abundances of ap stars .
we would like to thank eso and the pop for collecting the spectra used in this project and making them publicly available .
this research has made use of the webda database , developed by j .-
c . mermilliod and maintained by e. paunzen at the institute for astronomy of the university of vienna .
cpf and gaw acknowledge support from the academic research programme of the canadian department of national defence .
cpf , gaw and jdl acknowledge discovery grant support from the natural sciences and engineering research council of canada .
abt , h. a. , & morrell , n. i. 1995 , apjs , 99 , 135 babel , j. 1994 , a&a , 283 , 189 ballester , p. , modigliani , a. , boitquin , o. , et al .
2000 , messenger , 101 , 31 balona , l. a. 1994 , mnras , 268 , 119 bagnulo , s. , jehin , e. , ledoux , c. , et al .
2003 , messenger , 114 , 10 bagnulo , s. , landstreet , j.d . , mason , e. , et al .
2006 , a&a , 450 , 777 bohlender , d. a. , & landstreet , j. d. 1990 , mnras , 247 , 606 burkhart , c. & coupry , m. f. 1998 , a&a , 338 , 1073 catalano , f. a. & renson , p. 1998 , a&as , 127 , 421 dias , w. s. , lpine , j. r. , & alessi , b. s. 2001 , a&a , 376 , 441 esa 1997 , the hipparcos and tycho catalogues , esa sp-1200 gieseking , f. 1985 , a&as , 61 , 75 grevesse , n. , & sauval , a. j. 1998 , space sci .
, 85 , 161 hauk , b. , & north p. 1993
, a&a , 269 , 403 hempel , m. , & holweger , h. 2003 , a&a , 408 , 1065 hg , e. , fabricius , c. , & makarov , v. v. , et al .
2000 , a&a , 355 , l27 kharchenko , n. v. , piskunov , a. e. , rser , s. , schilbach , e. , & scholz , r .- d .
2005 , a&a , 438 , 1163 kupa , f. , piskunov , n. e. , ryabchikova , t. a. , stempels , h. c. , & wiess , w. w. 1999 , a&as , 138 , 119 kurucz , r. 1993 , cdrom model distribution , smithsonian astrophys .
landstreet , j. d. 1988 , apj , 326 , 967 landstreet , j. d. 1998 , a&a , 338 , 1041 van leeuwen , f. 1999 , a&a , 341 , l71 maitzen , h. m. & floquet , m. 1981 , a&a , 100 , 3 mermilliod , j .- c . , mermilliod , m. , & hauck , b. 1997 , a&as , 124 , 349 mermilliod , j .- c .
1995 , information and on - line data in astronomy , ed .
d. egret & m.a .
albrecht ( kluwer academic press , dordrecht ) , 127 meynet , g. , mermilliod , j .- c . , & maeder , a. 1993 , a&as , 98 , 477 napiwotzki , r. , schnberner , d. , & wenske , v. 1993 , a&a , 268 , 653 north , p. , & nicolet , b. 1990 , a&a , 228 , 78 ryabchikova , t. , nesvacil , n. , weiss , w. w. , kochukhov , o. , sttz , ch .
2004 , a&a , 423 , 705 renson , p. & catalano , f. a. 2001 , a&a , 378 , 113 renson , p. , gerbaldi , m. , & catalano , f. a. 1991 , a&as , 89 , 429 robichon , n. , arenou , f. , mermilliod , j .- c . , & turon , c. 1999 , a&a , 345 , 471 rser , s. , & bastian , u. 1993 , bidcs , 42 , 11 schaller , g. , schaerer , d. , meynet , g. , & maeder , a. 1992 , a&as , 96 , 269 sttz , ch .
, bagnulo , s. , jehin , e. , et al .
2006 , a&a , 451 , 285 sttz , ch . , nendwich , j. , rogers , n.y .
2002 , templogg.v2 http://ams.astro.univie.ac.at/templogg/ varenne , o. , & monier , r. 1999 , a&a , 351 , 247 wade , g. a. , bagnulo , s. , kochukov , o. , et al .
2001 , a&a , 375 , 265 wolff , s. c. 1967 , apjs , 15 , 21 | the connection between age , rotation and chemical abundance of magnetic ap stars is poorly understood . using open clusters ,
we are able to study samples of stars that are both co - eval and co - environmental . by determining rotation and chemical abundance for ap star members of clusters with various ages ,
the variations of these properties as a function of age and environment can be derived .
all four probable ap star members of the open cluster ngc 6475 , as well as one normal late b star , were studied using detailed spectrum synthesis of high resolution uves - pop spectra .
probable cluster membership was confirmed for all five stars , however chemical abundance anomalies only appear to be present in spectra of three . projected rotational velocity and chemical abundances for 21 elements ranging from c to eu are presented for the 5 stars . in the three peculiar stars we find overabundances of si ,
cr , mn , fe and rare earths such as nd , characteristic of ap stars .
the set of chemically peculiar stars show fairly homogeneous abundance tables , however notable differences exist for a few elements .
there also exist appreciable differences in the @xmath0 and main sequence evolutionary stage of the chemically peculiar stars .
this may hint at the underlying processes giving rise to the observed abundance anomalies . with this first detailed study of chemical abundances of a complete sample of magnetic ap / bp stars in an open cluster ,
we have initiated an exploration of the environmental and evolutionary influence on chemical peculiarity .
[ firstpage ] to be added ... |
state of art calculations of four - nucleon @xmath1 scattering have been recently presented in refs .
@xcite for all possible reactions initiated by @xmath2 , @xmath3 , @xmath4 , @xmath5 and @xmath6 below three - body breakup threshold .
realistic two - nucleon @xmath7 interactions based on meson theory like av18 @xcite , cd bonn @xcite and inoy04 @xcite or chiral effective field theory ( eft ) @xcite are used between pairs together with the coulomb repulsion between the protons .
no approximations were used in the solution of the four - body scattering equations beyond the usual partial - wave decomposition and the discretization of integration variables .
the results presented are fully converged vis - a - vis the included partial waves as well as the number of mesh points used for the discretization of all continuous variables .
some observables we obtain are described quite well by all interaction models , some scale with the three - nucleon @xmath8 binding energy , and some show large deviations from the data .
therefore the next step in our understanding of @xmath9 observables in terms of the underlying forces between nucleons requires the inclusion of a @xmath10 force .
there are three distinct ways for doing this : a ) add a static two - pion - exchange irreducible @xmath10 force @xcite to the underlying @xmath11 forces ; however , in this approach these two forces are not really consistent with each other .
b ) use @xmath12 force models based on chiral eft @xcite to guaranty consistency between the @xmath11 and @xmath10 forces ; however , for a realistic description , the expansion up to at least next - to - next - to - next - to leading order ( n3lo ) is required , which for the @xmath10 force is not yet available .
c ) extend the purely nucleonic model to allow the explicit excitation of a nucleon @xmath13 to a @xmath0 isobar as was carried out in ref .
@xcite for the @xmath11 and @xmath10 systems ; this approach yields effective many - nucleon forces consistent with the underlying @xmath11 force , but does not fully satisfy chiral symmetry , much like a ) .
the studies of the @xmath10 system reveal that all these different approaches lead to qualitatively similar results . in the @xmath9 system the first one a )
was already applied to @xmath2 @xcite and @xmath3 @xcite scattering . in the present paper ,
following the work on the @xmath10 system performed in refs.@xcite , we use the last approach c ) to study all @xmath9 reactions below three - body breakup threshold . in the @xmath10 system
the excitation of a single nucleon to a @xmath0 isobar yields an effective @xmath10 force with components of fujita - miyazawa type @xcite and much richer structures in a reducible form ; beside pion @xmath14 exchange , the @xmath10 force has contributions of shorter range due to the exchange of heavier mesons . in the @xmath9 system
an effective @xmath9 force arises that also has parts of shorter range than @xmath15 exchange .
the paper introduces the dynamics chosen for the extended description of the @xmath9 system in section [ sec : dyn ] .
it discusses the effects of @xmath0-isobar excitation in the form of @xmath10 and @xmath9 forces on the @xmath9 bound state in section [ sec : bs ] and on low - energy @xmath9 scattering observables in section [ sec : scatt ] .
conclusions are given in section [ sec : concl ] .
the description of the @xmath9 system is given in a hilbert space consisting of two sectors as depicted in fig .
[ fig : hs ] ; the first sector @xmath16 is purely nucleonic , and in the second sector @xmath17 one nucleon is replaced by a @xmath0 isobar of mass @xmath18 mev .
the restriction to hilbert sectors with one @xmath0 at most has a strong physics motivation .
the single @xmath0 isobar , when coupled to explicit pion - nucleon states , mediates the @xmath19 resonance in pion - nucleon scattering ; it also mediates single - pion production in @xmath11 scattering where single - pion production is the dominant inelastic channel up to about 500 mev in the @xmath11 center of mass ( c.m . ) system , i.e. , far beyond the two - pion threshold .
thus , the adopted hilbert space is sufficient for a further extension to the intermediate - energy pionic channels where a hilbert sector with two pions appears to be dynamically suppressed .
the nucleons in the sector @xmath16 are fully antisymmetrized .
the sector @xmath17 does not has a physics life on its own , but is included only through its coupling to @xmath16 .
that coupling is symmetric in all nucleons .
though the @xmath0 isobar is physically distinct from the nucleons , only wave - function components , antisymmetrized in all four baryons , nucleons and the @xmath0 isobar , have to be considered .
thus , faddeev - yakubovsky bound - state equations in the symmetrized form of ref .
@xcite and alt , grassberger and sandhas ( ags ) scattering equations @xcite in the symmetrized form of refs .
@xcite can be used . four - baryon hilbert space considered .
it consists of a purely nucleonic sector @xmath16 and a sector @xmath17 in which one nucleon is turned into a @xmath0 isobar , indicated by a thick line . ]
the dynamics is specified by a hermitian hamiltonian @xmath20 with instantaneous two - baryon potentials as indicated in fig .
[ fig : h ] for the @xmath9 system .
the hamiltonian acts in both hilbert sectors @xmath16 and @xmath17 and couples them .
the hermitian - conjugate of the component ( b ) is not shown separately . when limited to the @xmath11 system , the hamiltonian of fig .
[ fig : h ] ( a ) - ( c ) , i.e. , its respective components @xmath21 , @xmath22 and @xmath23 , reduces to the potential cd bonn + @xmath0 , a realistic coupled - channel two - baryon potential , fitted in ref .
@xcite to the elastic @xmath11 data
. the hamiltonian component of fig .
[ fig : h ] ( d ) , corresponding to the @xmath11 potential in the presence of a @xmath0 isobar , is not constrained by @xmath11 data .
a reasonable choice is the purely nucleonic cd bonn potential @xcite which we used in our previous @xmath10 calculations .
however , we found that the results for @xmath10 observables depend extremely weakly on the parametrization of the potential in fig .
[ fig : h ] ( d ) . even choosing it to be zero has no visible consequences on the description of @xmath10 observables ; e.g. , the calculated @xmath10 binding energy changes by 20 kev only .
therefore we choose the hamiltonian component of fig .
[ fig : h ] ( d ) to be zero in our @xmath9 calculations .
that choice is an assumption on unknown dynamics , but it also yields a technical simplification .
the solution of the @xmath9 equations remains exact .
four - baryon hamiltonian .
the dashed horizontal lines indicate potentials . ]
the symmetrized equations for the faddeev - yakubovsky amplitudes @xmath24 of the @xmath9 bound state are [ eq : fy ] @xmath25 is the free four - particle green s function and @xmath26 the two - baryon transition matrix
. the operators @xmath27 obtained from @xmath28 are the symmetrized ags operators for the @xmath29 and @xmath30 subsystems and @xmath31 is the permutation operator of particles @xmath32 and @xmath33 . the equations suffice for calculating the binding energy . the step from the faddeev - yakubovsky amplitudes @xmath24 to the bound state wave function
is not yet carried out .
the corresponding equations for @xmath9 scattering and the description of the screening and renormalization method to include the coulomb interaction are given in refs .
@xcite , and for that reason are not repeated here .
full four - body calculations are carried out in section [ sec : bs ] for the @xmath9 bound state and in section [ sec : scatt ] for selected @xmath9 reactions .
the dynamics is based on the coupled - channel potential cd bonn + @xmath0 ; the purely nucleonic cd bonn potential serves as reference for isolating the full @xmath0-isobar effect on the considered observables .
however , a split of the total @xmath0-isobar effect into separate contributions is highly desirable for understanding the physics of the results . for this goal
a sequence of incomplete calculations is also done .
the dynamic input , the coupled - channel two - baryon transition matrix is calculated correctly in all its components @xmath34 , @xmath35 , @xmath36 , and @xmath37 , but is only partially included in the following incomplete calculations : \(1 ) only the purely nucleonic component @xmath34 of the two - baryon transition matrix is retained .
the lowest order @xmath0 contribution to the dynamics , kept in this calculation , is shown in fig .
[ fig:2ndisp ] .
it renders the @xmath11 interaction less attractive off - shell .
this is the well known effect of @xmath11 dispersion .
\(2 ) only the two - baryon transition matrix components @xmath34 , @xmath35 , and @xmath36 are retained .
the most important @xmath0 contribution to the dynamics , kept in this calculation in addition to the @xmath11 dispersion , is of fujita - miyazawa ( fm ) type shown in fig .
[ fig : fm ] together with higher order @xmath10 force contributions ; the sample process on the right - hand side of fig .
[ fig : fm ] occurs due to two - body contributions contained in @xmath36 .
however , according to ref .
@xcite , those higher order @xmath10 force contributions should be far less important .
thus , the second incomplete calculation , when compared to the first one , is a reasonable estimation for the effective @xmath10 force of the fujita - miyazawa type .
\(3 ) in the third incomplete calculation all @xmath9 force effects are attempted to be eliminated while keeping all @xmath10 force effects .
the two - baryon transition matrix component @xmath37 , contained in the @xmath29 subsystem transition operator @xmath38 , generates higher order ( h.o . ) @xmath10 force contributions like those in fig .
[ fig:3nfho ] , in which one spectating nucleon is interaction - free . in addition , due to
the purely nucleonic intermediate states in @xmath37 , even particular iterations of the @xmath10 fujita - miyazawa force are generated like the one also shown in fig .
[ fig:3nfho ] , in which all four baryons are involved in the interaction process .
but @xmath37 is also the source for the effective @xmath9 force , whose corresponding lowest order contributions are shown in fig .
[ fig:4nf ] .
the clean elimination of the @xmath9 force is achieved by using the full @xmath37 component in the calculation of @xmath38 , but the modified part @xmath39 when the transition matrix acts immediately before / after the permutation operator @xmath40 in the iteration process of eq . where the particular transition matrix to be modified occurs explicitly .
the subtraction of @xmath41 ensures the presence of the purely nucleonic intermediate state between two successive @xmath10 transition operators @xmath38 , which act , due to the permutation @xmath40 , in different @xmath10 subsystems .
it therefore eliminates all @xmath9 force contributions for which fig .
[ fig:4nf ] gives lowest order examples .
thus , when comparing this calculation to the previous incomplete calculation ( 2 ) and to the full calculation , the effects of h. o. @xmath10 force contributions and of the @xmath9 force are estimated separately . lowest order contribution to the @xmath11 dispersion . ]
effective @xmath10 force of the fujita - miyazawa type ( left side ) and an example for a higher order @xmath10 force ( right side ) that is included together . ]
examples for higher order @xmath10 force processes .
the first two diagrams show contributions to the @xmath10 force mediated by the two - baryon transition matrix component @xmath37 contained in @xmath38 ; one nucleon stays uninvolved . in the last diagram all
four nucleons interact ; the process is the iteration of the @xmath10 fujita - miyazawa force ; it is due to the purely nucleonic intermediate states in @xmath37 . ]
lowest order effective @xmath9 force contributions . ]
in the present study the binding energy for the @xmath9 bound state is calculated .
other properties , such as the charge radius or the charge form factor , are not determined yet .
only total isospin @xmath42 states are included and isospin averaging is performed for the two - baryon transition matrix . in table
[ tab : eb ] we collect the results for @xmath10 and @xmath9 binding energies .
the inclusion of the @xmath0 isobar increases the corresponding binding energies but is unable to reproduce the experimental values .
obviously , many - nucleon forces , not accounted for by the @xmath0 isobar , make a rather significant contribution to @xmath10 and @xmath9 binding energies .
table [ tab : eb ] also splits up the total @xmath0-isobar effect into separate contributions obtained through incomplete calculations as discussed in section [ sec : deff ] .
( 1 ) the @xmath11 dispersion turns out to be massive in the @xmath9 bound state with @xmath43 mev ; it arises mainly from the dispersion in the @xmath44 @xmath11 partial wave .
( 2 ) the @xmath10 force contribution of the fujita - miyazawa type @xmath45 mev is also quite large .
the increase by the factor of 4.5 compared to the @xmath10 bound state is understandable in terms of the different multiplicity with which the @xmath10 force contributes : one in the @xmath10 bound state and four in the @xmath9 .
the observed factor of @xmath46 comes from the fact that @xmath47 , being a denser system than @xmath48 or @xmath49 , squeezes out more binding from the underlying force .
( 3 ) the contribution of the higher order @xmath10 force terms mediated by the diagonal @xmath50 potential @xmath23 is @xmath51 mev comparable to the one of the fujita - miyazawa type . the size of these h. o. terms depends on the strength of the @xmath52-meson exchange in @xmath23 that is not really constrained by elastic @xmath11 data ; it could get constrained by the data in pionic channels coupling to @xmath11 channels above the inelastic threshold .
the @xmath52-meson exchange strength of the cd bonn + @xmath0 potential was chosen to yield more binding in the @xmath10 system , and therefore its contribution to the binding energy of @xmath47 is quite sizable as well .
an alternative realistic coupled - channel potential with a weaker @xmath52-meson in @xmath23 was developed in ref .
@xcite . in the appendix
we show the differences relative to cd bonn + @xmath0 and the corresponding predictions for @xmath10 and @xmath9 binding energies .
( 4 ) finally , in contrast to the complete @xmath10 force contribution @xmath53 mev , the contribution arising from the effective @xmath9 force @xmath54 mev is indeed rather small .
l*3@r & @xmath55 & @xmath56 & @xmath57 + cd bonn & 8.00 & 7.26 & 26.18 + cd bonn + @xmath0 & 8.28 & 7.53 & 27.10 + exp & 8.48 & 7.72 & 28.30 + @xmath58 & -0.51 & -0.48 & -2.80 + @xmath59 & 0.50 & 0.48 & 2.25 + @xmath60 & 0.29 & 0.27 & 1.30 + @xmath61 & & & 0.17 +
the @xmath2 and @xmath3 scattering is dominated by the total isospin @xmath62 states while deuteron - deuteron @xmath63 scattering by the @xmath42 states ; the @xmath4 and @xmath5 reactions involve both @xmath42 and @xmath62 states and are coupled to @xmath6 in @xmath42 .
all those reactions below three - body breakup threshold were calculated in refs .
@xcite . here
we study the @xmath0-isobar effect on the low - energy @xmath9 scattering observables . in fig .
[ fig : nt ] we study the energy dependence of the total @xmath2 cross section .
the @xmath0-isobar excitation increases the @xmath10 binding energy and through scaling improves the description of the data around threshold
. however , there is quite a significant non - beneficial @xmath0 effect in the region of the resonance which is strongly driven by @xmath2 relative @xmath64 waves . in table
[ tab : nt ] we split this effect into @xmath11 dispersion , @xmath10 and @xmath9 force contributions , the last of which we find to be negligible .
furthermore , we split the total @xmath2 cross section @xmath65 into the @xmath66- and @xmath64-wave contributions @xmath67 and @xmath68 . as can be seen in table
[ tab : nt ] , the @xmath0 effects on the @xmath49 binding energy @xmath69 and the @xmath66-wave cross section are correlated by scaling in the same way as it has been observed in ref .
@xcite , i.e. , @xmath67 decreases when @xmath69 increases .
in contrast , there is no such a correlation in @xmath64 waves where both @xmath11 dispersion and effective @xmath10 force decrease the cross section while having opposite effects on @xmath69 .
this is different from @xmath10 scattering where the @xmath0 effect becomes visible , scaling aside , only at rather high energy , beyond 50 mev in the center of mass ( c.m . ) system . at lower energies the individual @xmath0 contributions are not negligible , but very often cancel each other to a large extent .
much smaller effect on @xmath65 in the resonance region is observed in refs .
@xcite where urbana ix @xmath10 force is added to av18 .
total cross section for @xmath2 scattering as function of neutron lab energy calculated with the cd bonn ( dashed curve ) and cd bonn + @xmath0 ( solid curve ) potentials .
experimental data are from ref . @xcite . ]
l*5@r & @xmath69 & @xmath67 & @xmath68 & @xmath65 & @xmath70 + cd bonn & 8.00 & 0.975 & 1.308 & 2.283 & 0.364 +
cd bonn + @xmath0 & 8.28 & 0.958 & 1.172 & 2.130 & 0.345 + exp & 8.48 & & & 2.450 & + @xmath11 dispersion & -0.51 & 0.036 & -0.075 & -0.039 & -0.055 + @xmath10 force ( fm ) & 0.50 & -0.035 & -0.058 & -0.094 & 0.022 + @xmath10 force ( h.o . ) & 0.29 & -0.017 & -0.004 & -0.021 & 0.014 + @xmath9 force & & @xmath71 & @xmath71 & @xmath71 & @xmath71 + in fig .
[ fig : p3he ] we study the observables of @xmath3 scattering at 5.54 mev proton lab energy .
this reaction is related to @xmath2 by charge symmetry that is broken only by the coulomb interaction and hadronic charge dependence .
the charge - symmetric @xmath0 effect is therefore very similar in both reactions .
it reduces the @xmath3 differential cross section at forward and backward angles increasing the discrepancy with data .
it is small and non - beneficial for the proton analyzing power @xmath72 , while the @xmath3 spin correlation coefficients remain described quite satisfactorily .
similar effects have also been observed in ref .
@xcite using the urbana ix @xmath10 force @xcite . in the last column of table [
tab : nt ] we split the @xmath0 effect into @xmath11 , @xmath10 and @xmath9 contributions for the maximum values of @xmath72 in @xmath2 scattering which is closely related to @xmath3 @xmath72 .
the effects of the @xmath11 dispersion and effective @xmath10 force are quite sizable , about -15% and 10% , respectively , but partially cancel each other . a similar canceling
was observed in the @xmath10 @xmath72 @xcite though there the @xmath10 force effect was larger than the @xmath11 dispersion , in contrast to the @xmath9 system .
differential cross section , proton analyzing power @xmath73 and @xmath3 spin correlation coefficient @xmath74 of @xmath3 scattering at 5.54 mev proton lab energy as functions of c.m .
scattering angle .
curves as in fig .
[ fig : nt ] .
the data are from ref .
@xcite for the differential cross section and from ref .
@xcite for the spin observables . ] the elastic differential cross section in the coupled @xmath5 and @xmath4 reactions correlates to some extent with that of @xmath3 and @xmath2 scattering and is similarly reduced by the @xmath0 excitation at forward and backward angles .
in contrast , the effect is much weaker for the @xmath75 transfer cross section as shown in fig .
[ fig : pt - nh ] .
the @xmath0 effect on @xmath72 in this reaction is consistent with the findings of ref .
@xcite where increasing the @xmath10 binding energy moves the predictions away from the data .
the @xmath0 effect is tiny for the elastic @xmath6 cross section as shown in fig .
[ fig : dd ] , but is visible for the deuteron tensor analyzing powers which , however , are very small . differential cross section and proton analyzing power of the @xmath75 reaction at 6 mev proton lab energy .
curves as in fig .
[ fig : nt ] .
the cross section data are from ref .
@xmath72 data are from ref .
@xcite . ] differential cross section and deuteron tensor analyzing power @xmath76 of the elastic @xmath6 scattering at 3 mev deuteron lab energy .
curves as in fig .
[ fig : nt ] .
the cross section data are from ref .
@xcite and @xmath76 data are from ref . @xcite . ] as we found in ref .
@xcite the observables of the two charge - symmetric transfer reactions @xmath77 and @xmath78 correlate to some extent with the @xmath10 binding energy and with the deuteron @xmath79-state probability .
that former correlation is reflected in fig .
[ fig : dd - ptnh ] where the inclusion of the @xmath0-isobar excitation brings the theoretical predictions closer to the data for the differential cross section , but has a smaller effect on the analyzing powers .
differential cross section and deuteron analyzing powers of the @xmath77 and @xmath78 reactions at 3 mev deuteron lab energy .
curves as in fig .
[ fig : nt ] .
the cross section data are from refs .
@xcite ( squares ) and @xcite ( circles ) . analyzing power data are from ref .
@xcite for @xmath77 and from ref .
@xcite for @xmath78 . ]
the technical apparatus , developed in refs .
@xcite for the solution of the @xmath9 bound state and scattering equations , is employed and extended .
the extension covers the use of a realistic coupled - channel potential allowing for the excitation of a single nucleon to a @xmath0 isobar .
the @xmath0 isobar mediates effective @xmath11 , @xmath10 and @xmath9 forces , consistent with each other .
a procedure for isolating the @xmath0-isobar effects of @xmath11 , @xmath10 and @xmath9 nature on observables is given and used to study different dynamic mechanisms .
_ technically _ , this paper demonstrates that @xmath9 calculations with realistic @xmath11 , @xmath10 and @xmath9 forces are feasible .
the coulomb repulsion between protons is included
. fully converged results for the @xmath9 binding energy and for the @xmath9 scattering observables below three - body breakup threshold are obtained .
_ physicswise _ , this paper shows for the first time that , within the present model space , @xmath9 force effect on nuclear observables is much smaller than the @xmath10 force effect . that fact is shown in the framework of @xmath0-mediated effective many - body forces for the @xmath9 bound state and low - energy scattering observables , and is a very valuable confirmation of the traditional belief in a hierarchic order for the importance of many - nucleon forces
. however , the inclusion of the @xmath0 isobar is unable to resolve the long - standing discrepancies with the experimental data , neither for the @xmath2 total cross section in the resonance region nor for @xmath72 in @xmath3 scattering .
nevertheless , in addition to possible differences resulting from model dependence , some differences between the work of lazauskas & carbonell @xcite , the pisa group @xcite and the present calculations need to be sorted out in the near future . while our findings seem to coincide with those of the pisa group for @xmath3 vis - a - vis the effect of the @xmath10 force , the calculations by lazauskas & carbonell indicate that the urbana ix @xmath10 force , when added to av18 @xmath11 force , bears almost no effect on the total cross section @xmath65 in the @xmath2 resonance region .
given that @xmath2 and @xmath3 only differ by the coulomb interaction and small charge dependent terms in the @xmath11 force , one does not expect such a different behavior between @xmath2 and @xmath3 when @xmath10 forces are added .
_ dynamically _ , our calculations are based on the two - baryon coupled - channel potential cd bonn + @xmath0 @xcite , which fits the deuteron properties and @xmath11 elastic scattering data as well as the best @xmath11 potentials @xcite . as standard in the description of nuclear structure and scattering ,
the @xmath0 isobar is assumed to be a stable particle of fixed mass without subthreshold corrections arising from the @xmath0-generated @xmath19 @xmath80 resonance .
this assumption is a crude and in principle unnecessary simplification of the @xmath0 isobar s dynamic structure , not allowing a direct application of the coupled - channel potential to pionic reactions .
this fact is the reason why the @xmath81 and especially the @xmath82 parts of the two - baryon potential in its present form are not sufficiently constrained by the @xmath11 data ; their full determination requires the data of @xmath83 dynamics .
we recall that all nuclear potentials are indeterminate to some extent , e.g. , @xmath11 potentials with respect to their short - range behavior and their amount of nonlocality .
however we are especially concerned about the indeterminacy of the @xmath82 part of the employed two - baryon coupled - channel potential , since it is responsible for the higher - order @xmath10 force and for the @xmath9 force , the focus of this paper .
fortunately , that dynamic indeterminacy does not change our physics conclusion in any form : exploiting that model dependence of the potential by using , besides cd bonn + @xmath0 , also its alternative cd bonn + @xmath84 described in the appendix , the @xmath9 force effect on binding energy remains much smaller than the @xmath10 force effect and for the scattering observables it is completely negligible .
finally , completing the list of all possible shortcomings of the employed two - baryon coupled - channel potential , the @xmath11 potential in the presence of a @xmath0 isobar , encountered in the hamiltonian underlying our calculations , is constrained only by the data of @xmath85 dynamics ; however , different choices for that part of the potential appear inconsequential for all studied observables according to our findings in the @xmath10 system .
_ in a longer range vision _
, the extension of the hamiltonian of this paper to cover also pionic reactions is quite possible , pushing the descriptions of @xmath10 and @xmath9 scattering to intermediate energies and thereby decreasing the model dependence inherent in the chosen force model with @xmath0-isobar excitation .
furthermore , we hope for a consistent derivation and tuning of purely nucleonic @xmath11 , @xmath10 and @xmath9 forces in the framework of chiral eft ; their subsequent application to the @xmath9 observables , studied in this paper , would be a challenging enterprise and a wonderful alternative to our present work .
we thank r. lazauskas and m. viviani for the discussion of @xmath10 force effects .
is supported by the fundao para a cincia e a tecnologia ( fct ) grant sfrh / bpd/34628/2007 and a.c.f . in part by the fct grant pocti / isfl/2/275 .
this paper works predominantly with the two - baryon coupled - channel potential cd bonn + @xmath0 , derived in ref .
its particular feature is a strong @xmath52-meson coupling @xmath86 for the direct @xmath87 component in @xmath50 states @xmath88 coupled to the nucleonic partial wave @xmath89 . that coupling is undetermined within sizable limits
; it was chosen to obtain more binding in the @xmath10 bound state , nevertheless allowing for the optimal fit of @xmath90 to the elastic @xmath11 scattering data .
this appendix presents selected results for an alternative coupled - channel potential cd bonn + @xmath84 @xcite with weaker @xmath91 that , after refitting the parameters of the two @xmath52 mesons in the nucleonic part @xmath21 of the potential , still allows for an optimal description of @xmath11 data with @xmath90 . for completeness in table [ tab : cdbdx ]
we give the changed @xmath52-meson parameters of the cd bonn + @xmath84 potential in the @xmath89 partial wave ; other parameters have the same values as for cd bonn + @xmath0 and are given in ref .
@xcite .
table [ tab : ebx ] shows the changes in the binding energy of @xmath49 , @xmath48 and @xmath47 arising for cd bonn + @xmath84 . the @xmath0-isobar effect is weaker than for cd bonn + @xmath0 , turning even non - beneficial for @xmath47 .
whereas the @xmath11 dispersive and @xmath10 fujita - miyazawa effects , i.e. , @xmath58 and @xmath59 , remain practically unchanged , the higher order @xmath10 and @xmath9 force contributions , i.e. , @xmath60 and @xmath61 , being much more sensitive to the @xmath87 component , get strongly reduced , thereby changing the complete @xmath0-isobar effect considerably .
this is a measure of our model dependence for 0.17 @xmath98 0.03 .
l*3@r & @xmath55 & @xmath56 & @xmath57 + cd bonn & 8.00 & 7.26 & 26.18 + cd bonn + @xmath84 & 8.05 & 7.31 & 25.89 + exp & 8.48 & 7.72 & 28.30 + @xmath58 & -0.51 & -0.48 & -2.78 + @xmath59 & 0.50 & 0.48 & 2.20 + @xmath60 & 0.06 & 0.05 & 0.26 + @xmath61 & & & 0.03 + | the four - nucleon bound state and scattering below three - body breakup threshold are described based on the realistic coupled - channel potential cd bonn + @xmath0 which allows the excitation of a single nucleon to a @xmath0 isobar .
the coulomb repulsion between protons is included . in the four - nucleon system the two - baryon coupled - channel potential yields effective two- , three- and four - nucleon forces , mediated by the @xmath0 isobar and consistent with each other and with the underlying two - nucleon force .
the effect of the four - nucleon force on the studied observables is much smaller than the effect of the three - nucleon force .
the inclusion of the @xmath0 isobar is unable to resolve the existing discrepancies with the experimental data .
, , four - nucleon , bound state , scattering , @xmath0-isobar , many - nucleon forces 21.45.+v , 21.30.-x , 24.70.+s , 25.10.+s |
iron nitrides , especially in iron - rich phases , have been under intense research due to the strong ferromagnetism and interest in its physical origin @xcite .
the difficulty in obtaining a single phase has been a long - standing problem for ferromagnetic iron nitrides , to hinder fundamental understanding of intrinsic physical properties @xcite . recently , the successful epitaxial growth of single - phase ferromagnetic @xmath0-fe@xmath1n has been reported on various substrates , which helps to comprehend a crucial role for the hybridization between fe and n states in the ferromagnetism of @xmath0-fe@xmath1n @xcite .
the robust fe - n bonding also renders an fe@xmath2n layer strongly two - dimensional @xcite , which possibly facilitates a layer - by - layer stacking of @xmath0-fe@xmath1n on metals .
this contrasts with the case of elemental 3@xmath5 transition metals ( tms ) deposited on 3@xmath5 tm substrates , in which inevitable atom intermixing and exchange of constituents prevent the formation of ordered overlayers @xcite .
therefore , the investigation into the electronic and magnetic states of @xmath0-fe@xmath1n atomic layers can not only elucidate the layer-/site - selective electronic and magnetic states of @xmath0-fe@xmath1n , but unravel the origin of the strongly thickness - dependent physical properties in a thin - film limit of 3@xmath5 tm ferromagnets @xcite . + here , we report two growth modes of @xmath0-fe@xmath1n / cu(001 ) depending on preparation methods .
the scanning tunneling microscopy / spectroscopy ( stm / sts ) observations indicated a successful growth of ordered trilayer @xmath0-fe@xmath1n , without extra nitrogen bombardment onto the existing structures .
x - ray absorption spectroscopy / magnetic circular dichroism ( xas / xmcd ) measurements revealed the thickness dependence of the magnetic moments of fe atoms , the origin of which was well explained by the first - principles calculations .
based on an atomically - resolved structural characterization of the system , the layer - by - layer electronic and magnetic states of the @xmath0-fe@xmath1n atomic layers have been understood from both experimental and theoretical points of view .
a clean cu(001 ) surface was prepared by repetition of sputtering with ar@xmath3 ions and subsequent annealing at 820 k. iron was deposited at room temperature ( rt ) in a preparation chamber under an ultrahigh vacuum ( uhv ) condition ( @xmath6 torr ) , using an electron - bombardment - type evaporator ( efm , focus ) from a high - purity fe rod ( 99.998 % ) .
the stm measurements were performed at 77 k in uhv ( @xmath7 torr ) using electrochemically etched w tips .
the differential conductance d@xmath8/d@xmath9 was recorded for sts using a lock - in technique with a bias - voltage modulation of 20 mv and 719 hz .
the xas and xmcd measurements were performed at bl 4b of uvsor - iii @xcite in a total electron yield ( tey ) mode .
the degree of circular polarization was @xmath10 , and the x - ray propagation vector lay within the ( 110 ) plane of a cu(001 ) substrate .
all the xas / xmcd spectra were recorded at @xmath11 , with external magnetic field @xmath12 up to @xmath13 t applied parallel to the incident x - ray .
the symmetry and quality of the surface were also checked by low energy electron diffraction ( leed ) in each preparation chamber .
first - principles calculations were performed within the density functional theory in the local density approximation @xcite , using a self - consistent full - potential green function method specially designed for surfaces and interfaces @xcite .
( color online ) topography and atomic structure of the monolayer @xmath0-fe@xmath1n on cu(001 ) .
( a ) topographic image ( 100@xmath1450 nm@xmath15 , sample bias @xmath16 , tunneling current @xmath17 ) of the monolayer @xmath0-fe@xmath1n on cu(001 ) .
white lines represent step edges of the cu(001 ) terraces .
color contrast is enhanced within each terrace .
( b ) close view ( [email protected] nm@xmath15 , @xmath18 , @xmath19 ) of the surface fe@xmath2n layer .
the dimerization of fe atoms is indicated by encirclement .
( c ) leed pattern obtained with an incident electron energy of 100 ev .
( d ) bulk crystal structure of @xmath0-fe@xmath1n .
a dotted parallelogram represents an fe@xmath2n plane .
( e ) atomic structure of the monolayer @xmath0-fe@xmath1n on cu(001 ) .
( f ) schema illustrating @xmath20 reconstruction in the surface fe@xmath2n layer of @xmath0-fe@xmath1n .
arrows indicate the shift of the fe atoms from an unreconstructed @xmath21 coordination ( dotted circles ) . for ( d ) to ( f ) , large blue ( yellow ) and small red spheres represent fe ( cu ) and n atoms , respectively.,width=325 ] monolayer fe@xmath2n on cu(001 ) was prepared prior to any growth of multilayer @xmath0-fe@xmath1n by the following cycle : n@xmath3 ion bombardment with an energy of 0.5 kev to a clean cu(001 ) surface , subsequent fe deposition at rt , and annealing at 600 k. note that the monolayer fe@xmath2n is identical to fe@xmath1n on cu(001 ) in a monolayer limit , and thus referred to as also monolayer @xmath0-fe@xmath1n hereafter . a topographic image of the sample after one growth cycle is shown in fig .
[ fig1](a ) . the monolayer @xmath0-fe@xmath1n is formed on the cu terraces at @xmath22 0.85 ml coverage .
an atomically - resolved image of that surface displayed in fig . [ fig1](b ) reveals a clear dimerization of the fe atoms , typical of ordered @xmath0-fe@xmath1n on cu(001 ) @xcite .
a leed pattern of the surface is shown in fig .
[ fig1](c ) , which exhibits sharp spots with the corresponding @xmath20 symmetry .
it is known that @xcite the topmost layer of the @xmath0-fe@xmath1n on cu(001 ) always consists of the fe@xmath2n plane in a bulk fe@xmath1n crystal shown in fig . [ fig1](d ) . a schematic model of the monolayer @xmath0-fe@xmath1n
is given in fig .
[ fig1](e ) , composed of a single fe@xmath2n plane on cu(001 ) .
accordingly , the surface fe@xmath2n plane takes reconstruction to the @xmath20 coordination @xcite , in which the fe atoms dimerize in two perpendicular directions as illustrated in fig . [ fig1](f ) .
+ ( color online ) topography of the bilayer @xmath0-fe@xmath1n dot on cu(001 ) .
( a ) topographic image ( 120@xmath1460 nm@xmath15 , @xmath23 , @xmath17 ) of the monolayer ( darker area ) and dot - like bilayer @xmath0-fe@xmath1n on cu(001 ) .
white lines represent step edges of the cu(001 ) terraces .
color contrast is enhanced within each terrace .
( b , c ) upper panels : atomically - resolved topographic images ( 7@xmath143 nm@xmath15 , @xmath24 ) taken at ( b ) @xmath23 and ( c ) @xmath25 .
lower panels : height profiles measured along lines indicated in the upper panels .
( d ) proposed atomic structure of the bilayer - dot @xmath0-fe@xmath1n on cu(001 ) .
large blue ( yellow ) and small red spheres correspond to fe ( cu ) and n atoms , respectively.,width=325 ] after repeating the growth cycles , we found a new structure different from the monolayer @xmath0-fe@xmath1n .
figure [ fig2](a ) displays the surface after two growth cycles in total , namely , another cycle of the n@xmath3 ion bombardment , fe deposition , and annealing onto the existing monolayer @xmath0-fe@xmath1n surface .
then , the surface becomes mostly covered with the monolayer @xmath0-fe@xmath1n , which contains a small number of bright dots . for a structural identification of these dots , we measured atomically - resolved topographic images and line profiles at different @xmath26 as shown in fig .
[ fig2](b ) and [ fig2](c ) .
the dot structure imaged at @xmath27 reveals the dimerization of the fe atoms as the monolayer @xmath0-fe@xmath1n surface .
this indicates that the topmost part of the dot consists of the reconstructed fe@xmath2n . at positive @xmath28 of + 0.1 v , in contrast
, the dot is recognized as a single protrusion both in the topographic image and line profile , while the surrounding monolayer @xmath0-fe@xmath1n still shows the fe dimerization .
this implies the different electronic structure of the dot compared to the monolayer @xmath0-fe@xmath1n , which comes from the difference in a subsurface atomic structure .
+ the observed height difference between the dot and the monolayer @xmath0-fe@xmath1n ranges from 4 to 10 pm depending on @xmath26 .
these values are in the same order of a lattice mismatch between the bulk crystals of the @xmath0-fe@xmath1n / cu(001 ) ( 380 pm ) and cu(001 ) ( 362 pm ) @xcite , but an order of magnitude smaller than the lattice constant of the @xmath0-fe@xmath1n / cu(001 ) .
this suggests that the topmost layer of the dot is not located above the monolayer @xmath0-fe@xmath1n surface , but shares the fe@xmath2n plane with .
furthermore , the bright dot is composed of only four pairs of the fe dimer as imaged in fig .
[ fig2](b ) , indicating that the difference in the atomic and/or electronic structures is restricted within a small area . considering the above ,
it is most plausible that one fe atom is embedded just under the surface n atom at the dot center , and thus a bilayer @xmath0-fe@xmath1n dot is formed as schematically shown in fig .
[ fig2](d ) .
this structure corresponds to a minimum unit of the bilayer @xmath0-fe@xmath1n on cu(001 ) .
+ ( color online ) topographic images ( 15@xmath1415 nm@xmath29 ) of the surface after repetition of ( a ) two and ( b ) three growth cycles .
the set point is ( @xmath30 ) = ( + 0.25 v , 5.0 na ) for ( a ) and ( + 0.1 v , 3.0 na ) for ( b).,width=325 ] this bilayer dot formed clusters by a further repetition of the growth cycles .
figure [ fig3](a ) shows an enlarged view of the iron - nitride surface after two growth cycles .
the coverage of the dot is estimated to be @xmath22 5 % of the entire surface .
another growth cycle onto this surface led to an increase in a dot density up to @xmath22 40 % , as shown in fig . [
fig3](b ) .
however , further repetitions of the cycles resulted in neither a considerable increase in the dot density nor the formation of a continuous bilayer film .
this can be attributed to an inevitable sputtering effect in every growth cycle : an additional n@xmath3 ion bombardment to the existing surface not only implanted n@xmath3 ions but also sputtered the surface , which caused the loss of the iron nitrides already formed at the surface , as well as the increase in the surface roughness .
+ to compensate this loss of surface fe atoms by the sputtering effect , we also tried to increase the amount of deposited fe per cycle .
nonetheless , the number of fe atoms , which remained at the surface after annealing , did not increase possibly because of the thermal metastability of fe / cu systems @xcite .
the isolated fe atoms without any bonding to n atoms were easily diffused and embedded into the cu substrate during the annealing process . as a result , only the imperfect bilayer @xmath0-fe@xmath1n was obtained through this method .
+ ( color online ) topography of the trilayer @xmath0-fe@xmath1n film on cu(001 ) .
topographic images ( 100@xmath14100 nm@xmath15 ) after ( a ) two and ( b ) three cycles of the fe deposition under n@xmath2 atmosphere and subsequent annealing onto the monolayer @xmath0-fe@xmath1n on cu(001 ) .
the setpoint is @xmath31 , @xmath32 for ( a ) and -0.05 v for ( b ) .
white lines indicate step edges of the cu terraces .
color contrast is enhanced within each terrace .
( c ) atomically - resolved topographic image ( 4@xmath144 nm@xmath15 , @xmath33 ) of the trilayer @xmath0-fe@xmath1n surface .
an inset represents a leed pattern of the sample shown in ( b ) , obtained with an incident electron energy of 100 ev .
( d ) height profile measured along the line indicated in ( b ) .
( e ) xas edge jump spectra of the trilayer ( solid ) and monolayer ( dotted ) samples at the fe and cu @xmath34 edges .
the intensity is normalized to the cu edge jump .
( f ) atomic model expected for the trilayer @xmath0-fe@xmath1n on cu(001 ) .
blue ( yellow ) large and red small spheres represent fe ( cu ) and n atoms , respectively.,width=325 ] multilayer @xmath0-fe@xmath1n films were obtained by the following procedure .
first , the monolayer @xmath0-fe@xmath1n was prepared on cu(001 ) as above .
then , 2 ml fe was deposited under n@xmath2 atmosphere ( 5.0@xmath1410@xmath35 torr ) pressure , far below the parameters at which fe started to be evaporated .
this indicates the ionization of the n@xmath2 molecules and/or n atoms around the evaporator possibly by thermal electrons created inside it .
then , the n@xmath3 and n@xmath36 ions could reach to the surface together with the evaporated fe atoms , or iron nitride was already formed before landing .
] at rt , and the sample was annealed at 600 k. figures [ fig4](a ) and [ fig4](b ) show topographic images after two and three above mentioned cycles , respectively . in the images ,
the coverage of new bright area , different from the imperfect bilayer dot , monotonously increases with repeating the cycles . a close view of that new surface is displayed in fig .
[ fig4](c ) , revealing the dimerized ( or even @xmath21-like dot ) structures .
because a leed pattern shown in the inset of fig .
[ fig4](c ) exhibits the @xmath20 symmetry without extra spots , the topmost layer of this surface is composed of the reconstructed fe@xmath2n plane @xcite .
therefore , these observations suggest that the new area would consist of @xmath0-fe@xmath1n other than both of the monolayer and bilayer dot . + in order to determine the structure of this newly obtained @xmath0-fe@xmath1n , a typical height profile of the surface was recorded as shown in fig . [
fig4](d ) .
it is clear that the new structure is higher than both the cu surface and the surface including the monolayer / dot - like bilayer @xmath0-fe@xmath1n .
this suggests that the new area is composed of @xmath0-fe@xmath1n thicker than bilayer .
quantitative information on the thickness of the new structure could be obtained from fe @xmath37 edge jump spectra shown in fig .
[ fig4](e ) , whose intensity is roughly proportional to the amount of surface / subsurface fe atoms .
the sample prepared in the same procedure as that shown in fig .
[ fig4](b ) reveals an edge jump value of 0.32 , while the monolayer @xmath0-fe@xmath1n 0.12.-fe@xmath1n structures , was embedded into the cu substrate during annealing , at least several nms ( probing depth in the tey mode ) below the surface . ]
considering that the new area occupies @xmath22 60 % of the entire surface as deduced from fig .
[ fig4](b ) , the thickness of this @xmath0-fe@xmath1n must be less than quadlayer to meet the experimental edge jump value of 0.32 ( see appendix [ jumptoml ] ) .
hence , the newly obtained structure is identified as a trilayer @xmath0-fe@xmath1n film .
an atomic structure expected for the trilayer @xmath0-fe@xmath1n on cu(001 ) is presented in fig .
[ fig4](f ) .
the growth without any ion bombardment to the monolayer surface possibly stabilizes the subsurface pure fe layer , which could promote the formation of the trilayer @xmath0-fe@xmath1n film in a large area .
+ finally , let us mention another growth method of the @xmath0-fe@xmath1n film .
we previously report a possible layer - by - layer growth of the @xmath0-fe@xmath1n atomic layers on cu(001 ) , by the n@xmath3 ion bombardment with a relatively low energy of 0.15 kv @xcite .
this soft implantation of n@xmath3 ions successfully avoids extra damage to the existing @xmath0-fe@xmath1n structures during the repetition of the growth cycles .
the reported different electronic / magnetic states could then originate from the difference in the fabrication processes .
another finding is that , in the current study , only the monolayer and trilayer @xmath0-fe@xmath1n could be obtained in a continuous film form .
this implies that an fe@xmath2n - layer termination would be preferable through the present methods , possibly due to the metastability of an interface between cu and pure fe layers @xcite .
+ ( color online ) surface electronic structures of the @xmath0-fe@xmath1n on cu(001 ) .
experimental d@xmath8/d@xmath9 spectra recorded above the trilayer ( solid ) and monolayer ( dotted ) @xmath0-fe@xmath1n surfaces are presented .
the d@xmath8/d@xmath9 intensity is arbitrary .
a stm tip was stabilized at @xmath16 , @xmath38 and 7.0 na for the trilayer and monolayer surfaces , respectively .
gray lines are guide to the eye.,width=226 ] the surface electronic structures of @xmath0-fe@xmath1n showed large dependence on the sample thickness .
figure [ fig5 ] displays experimental d@xmath8/d@xmath9 spectra measured on the surfaces of the trilayer and monolayer @xmath0-fe@xmath1n .
the peaks located at @xmath39 + 0.20 , + 0.55 , and + 0.80 v , mainly originating from the unoccupied states in the down - spin band characteristic of fe local density of states ( ldos ) , are observed for both the trilayer and monolayer surfaces . a significant difference between the spectra is a dominant peak located around @xmath40 observed only for the trilayer surface .
this peak possibly originates from the ldos peak located around @xmath41 , calculated for the fe atoms not bonded to n atoms in the subsurface fe layer [ corresponding site of fe4 shown in fig .
[ fig7](b ) ] .
because of the @xmath42 orbital character , this peak could be dominantly detected in the sts spectrum for the trilayer surface . thus , the appearance of this additional peak could support the different subsurface structure of the trilayer sample , especially , the existence of the subsurface fe layer proposed above .
+ the entire electronic and magnetic properties of the sample , including both surface and subsurface information , were investigated by using xas and xmcd techniques at the fe @xmath43 absorption edges .
figure [ fig6](a ) shows xas ( @xmath44 ) and xmcd ( @xmath45 ) spectra under @xmath46 of the trilayer and monolayer samples in the grazing ( @xmath47 ) and normal incidence ( @xmath48 ) . here , @xmath49 denotes a x - ray absorption spectrum with the photon helicity parallel ( antiparallel ) to the fe 3@xmath5 majority spin , and an incident angle @xmath50 is defined as that between the sample normal and incident x - ray
the trilayer ( monolayer ) sample was prepared in the same procedure as that shown in fig .
[ fig4](b ) [ fig .
[ fig1](a ) ] .
it is clear that the xmcd intensity is larger in the trilayer one , indicating an enhancement of magnetic moments with increasing thickness .
+ for a further quantitative analysis on the magnetic moments , we applied xmcd sum rules @xcite to the obtained spectra and estimated spin ( @xmath51 ) and orbital ( @xmath52 ) magnetic moments separately .
note that the average number of 3@xmath5 holes ( @xmath53 ) of 3.2 was used in the sum - rule analysis , which was estimated by comparing the area of the experimental xas spectra with that of a reference spectrum of bcc fe / cu(001 ) ( @xmath54 ) @xcite .
the thickness dependence of the @xmath51 and @xmath52 values is summarized in fig . [ fig6](b ) .
the value of @xmath51 increases monotonously with increasing the fe @xmath34-edge jump value , namely , an average sample thickness , and finally saturates at @xmath55/atom in the trilayer sample ( corresponding edge jump value of 0.32 ) .
the change in @xmath52 is not so systematic relative to @xmath51 , however , the @xmath52 values seem to be enhanced in the grazing incidence .
this implies an in - plane easy magnetization of the @xmath0-fe@xmath1n atomic layers on cu(001 ) , also consistent with the previous reports on the @xmath0-fe@xmath1n thin films on cu(001 ) @xcite .
figure [ fig6](c ) shows magnetization curves of the monolayer sample , whose intensity corresponds to the @xmath56-peak xas intensity normalized to the @xmath57 one .
the curve recorded in the normal incidence shows negligible remanent magnetization . on the other hand ,
that in the grazing one draws a rectangular hysteresis loop , which confirms the in - plane easy magnetization .
the coercivity of the monolayer sample is estimated to be @xmath22 0.05 t at 8.0 k , larger than @xmath22 0.01 t for 5 ml fe / cu(001 ) @xcite , @xmath22 1 mt for 5 ml fe / gaas(100)-(4@xmath146 ) @xcite and the 30 nm thick @xmath0-fe@xmath1n film @xcite at rt .
+ ( color online ) layer - by - layer electronic states of the @xmath0-fe@xmath1n atomic layers on cu(001 ) . calculated layer - resolved dos projected to each 3@xmath5 orbital of the ( a ) monolayer and ( b ) trilayer @xmath0-fe@xmath1n on cu(001 ) .
the dos in the up-(down- ) spin band is shown at upper ( lower ) panels .
note that the states with @xmath58 and @xmath59 orbitals are degenerated for the fe3 and fe4 sites in ( b).,width=325 ] c|>p3.0em|>p3.0em|>p3.25em|>p3.25em|>p3.5em|>p3.5em & & & + & fe1 & fe2 & fe3 & fe4 & fe5 & fe6 + monolayer & 1.1 & 1.1 & -&-&-&-
+ trilayer & 1.8 & 1.8 & 2.0 & 3.0 & 0.62 & 0.62 the observed thickness dependence of the magnetic moments can be well understood with a help of first - principles calculations .
figures [ fig7](a ) and [ fig7](b ) show layer - resolved dos of the monolayer and trilayer @xmath0-fe@xmath1n on cu(001 ) , respectively . here ,
non - equivalent fe sites in each layer are distinguished by different numbering .
in particular , the fe atoms at the fe3 ( fe4 ) site in the trilayer @xmath0-fe@xmath1n correspond to those with ( without ) a bond to n atoms .- fe@xmath1n , ( fe1 , fe2 ) and ( fe5 , fe6 ) in the trilayer one is just a switch of the orbital assignment between @xmath58 and @xmath59 . therefore , the dos of fe2 in the monolayer @xmath0-fe@xmath1n , fe2 and fe6 in the trilayer one is not presented here . ] in table [ tab1 ] , calculated values of an atomic magnetic moment @xmath60 , corresponding to @xmath61 + @xmath62 along the easy magnetization direction , are also listed . in the monolayer case ,
the calculated @xmath60 is 1.1 @xmath63/atom , which is in perfect agreement with the experimental value .
this supports an ideal atomic structure of our monolayer sample .
+ interestingly , the value of @xmath60 for the fe atoms in the monolayer @xmath0-fe@xmath1n is more than 1.5 times smaller than that in the topmost layer of the trilayer one ( 1.83 @xmath63/atom ) . in comparison with the dos shown at the top of fig .
[ fig7](b ) , the impact of the hybridization with the cu states on the fe dos can be seen in fig .
[ fig7](a ) : first , the dos in the up - spin band , especially with @xmath42 and @xmath58 orbitals , becomes to have a tail toward a higher - energy side across the @xmath64 .
this change deviates the 3@xmath5 electrons in the up - spin band from a fully - occupied nature .
moreover , the spin asymmetry of the occupied 3@xmath5 electrons , the difference between the electron occupation into each spin band normalized by the sum of them , reduces especially for the dos with @xmath65 , @xmath42 and @xmath58 orbitals .
these changes could decrease @xmath66 of the fe atoms .
note that the similar reduction in the magnetic moments of 3@xmath5 tms due to the hybridization with cu states is reported , for example , in ref . .
+ then , by comparing two different fe@xmath2n interfaces with the cu substrate , it turns out that @xmath60 of the monolayer @xmath0-fe@xmath1n ( 1.1 @xmath63/atom ) is almost twice compared to that of the trilayer one ( 0.62 @xmath63/atom ) . in the monolayer case
, the fe@xmath2n layer faces to a vacuum and the fe atoms are under reduced atomic coordination .
this results in the narrower band width , and thus the dos intensity increases in the vicinity of @xmath64 .
accordingly , a larger exchange splitting can be possible and the spin asymmetry of the occupied 3@xmath5 electrons increases as shown in fig .
[ fig7](a ) , compared to the interfacial fe@xmath2n layer of the trilayer @xmath0-fe@xmath1n [ bottom panel of fig . [ fig7](b ) ] .
this leads to larger magnetic moments at the surface . as a result , the competition between the enhancement at the surface and the decrease at the interface would make @xmath60 values quite layer - sensitive .
+ in the subsurface fe layer of the trilayer @xmath0-fe@xmath1n , the value of @xmath60 becomes largest due to the bulk coordination of the fe atoms . especially the fe atoms not bonded to the n ones possess @xmath60 of 3.0 @xmath63/atom , which is comparable to the values of fe atoms at the same site in the bulk @xmath0-fe@xmath1n @xcite .
consequently , by averaging the layer - by - layer @xmath60 values of the trilayer @xmath0-fe@xmath1n , the total magnetic moment detected in the xmcd measurement is expected to be 1.7 @xmath67/fe , with the electron escape depth taken into account ( see appendix [ jumptoml ] ) .
considering the composition expected to the trilayer sample , this value can well explain the experimental one of @xmath22 1.5 @xmath67/fe .
+ the theory also demonstrates the direction of an easy magnetization axis .
the in - plane easy magnetization of our @xmath0-fe@xmath1n samples was confirmed by the magnetization curves as well as the incidence dependence of the @xmath52 value .
in contrast , the pristine ultrathin fe films , which form either fct or fcc structures on cu(001 ) , show uncompensated out - of - plane spins over a few surface layers @xcite . this shift of magnetic anisotropy by nitridation can be understood from the orbital - resolved fe dos shown in figs . [ fig7](a ) and [ fig7](b ) . unlike the pure fe / cu(001 )
system @xcite , the occupation of 3@xmath5 electrons in states with out - of - plane - oriented orbitals ( @xmath68 ) is considerably larger than that with in - plane - oriented ones ( @xmath69 ) .
this could make @xmath52 prefer to align within a film plane , resulting in the in - plane magnetization of the system @xcite .
in conclusion , we have conducted a detailed study on the growth , electronic and magnetic properties of the @xmath0-fe@xmath1n atomic layers on cu(001 ) .
the ordered trilayer film of @xmath0-fe@xmath1n can be prepared by the fe deposition under n@xmath2 atmosphere onto the existing monolayer surface . on the other hand ,
the repetition of the growth cycles including the high - energy n@xmath3 ion implantation resulted in the imperfect bilayer @xmath0-fe@xmath1n .
the stm and sts observations revealed the change in the surface topography and electronic structures with increasing the sample thickness .
the xas and xmcd measurements also showed the thickness dependence of the spectra , and the corresponding evolution of the @xmath51 values .
all the thickness dependence of the electronic and magnetic properties is well explained by the layer - resolved dos calculated using the first principles .
structural perfection of the system makes it possible to fully comprehend the layer - by - layer electronic / magnetic states of the @xmath0-fe@xmath1n atomic layers .
this work was partly supported by the jsps grant - in - aid for young scientists ( a ) , grant no .
16h05963 , for scientific research ( b ) , grant no . 26287061 , the hoso bunka foundation , shimadzu science foundation , iketani science and technology foundation , and nanotechnology platform program ( molecule and material synthesis ) of the ministry of education , culture , sports , science and technology ( mext ) , japan .
y. takahashi was supported by the grant - in - aid for jsps fellows and the program for leading graduate schools ( merit ) .
a.e . acknowledges funding by the german research foundation ( dfg grants no .
er 340/4 - 1 ) .
the escape probability of electrons from inside a sample to a vacuum depends on the depth at which the electrons are excited . for a numerical interpretation of the xas edge jump ,
the following factors should be mainly considered in principle : the penetration length of an incident x - ray ( @xmath70 ) and electron escape depth ( @xmath71 ) , both energy - dependent . in the case of a few atomic layers of 3@xmath5 transition metals ,
the attenuation of the incident x - ray intensity is almost negligible because @xmath70 is orders of magnitude longer than the sample thickness @xcite .
therefore , in the present case , only the electron escape probability at the depth @xmath72 from the surface , namely , a factor of @xmath73 is taken into account . as for the @xmath71 value of fe , 17 was tentatively assumed in our analysis , which is experimentally determined for fe thin films @xcite .
then , based on the experimental fe ( n ) edge jump values of 0.12 ( 0.015 ) , those for the full - coverage dot - like bilayer , trilayer , and quadlayer @xmath0-fe@xmath1n on cu(001 ) are calculated as summarized in table [ tablejump ] .
.[tablejump ] experimental and calculated fe ( n ) edge jump values for the monolayer , dot - like bilayer , trilayer , and quadlayer @xmath0-fe@xmath1n on cu(001 ) . in the calculation , each @xmath0-fe@xmath1n is assumed to have the atomic structure presented in the text and fully cover the entire surface .
for the quadlayer one , an fe@xmath2n / fe@xmath2/fe@xmath2n / fe@xmath2/cu(001 ) structure is assumed . [ cols="^,^,^,^,^ " , ] 51ifxundefined [ 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 \doibase http://dx.doi.org/10.1016/s0304-8853(99)00429-1 [ * * , ( ) ] link:\doibase 10.1103/physrev.112.751 [ * * , ( ) ] \doibase http://dx.doi.org/10.1063/1.358156 [ * * , ( ) ] \doibase http://dx.doi.org/10.1063/1.344689 [ * * , ( ) ] \doibase http://dx.doi.org/10.1063/1.112552 [ * * , ( ) ] @noop * * , ( ) \doibase http://dx.doi.org/10.1063/1.1392980 " [ * * , ( ) ] link:\doibase 10.1103/physrevb.70.115417 [ * * , ( ) ] @noop * * , ( ) \doibase http://dx.doi.org/10.1063/1.1586790 [ * * , ( ) ] link:\doibase 10.1103/physrevb.73.172410 [ * * , ( ) ] http://scitation.aip.org/content/aip/journal/jap/117/19/10.1063/1.4921431 [ * * , ( ) ] link:\doibase 10.1039/c3ra47385f [ * * , ( ) ] link:\doibase 10.1103/physrevb.55.7904 [ * * , ( ) ] link:\doibase 10.1103/physrevb.60.14382 [ * * , ( ) ] link:\doibase 10.1103/physrevb.68.205413 [ * * , ( ) ] link:\doibase 10.1103/physrevb.56.r4398 [ * * , ( ) ] link:\doibase 10.1103/physrevb.56.5100 [ * * , ( ) ] link:\doibase 10.1103/physrevb.55.3708 [ * * , ( ) ] link:\doibase 10.1103/physrevb.50.13467 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.72.3112 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.77.743 [ * * , ( ) ] link:\doibase 10.1103/physrevb.54.4075 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.103.267202 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/s0301-0104(02)00724-3
[ * * , ( ) ] http://stacks.iop.org/1347-4065/47/i=4r/a=2132 [ * * , ( ) ] http://link.aps.org/doi/10.1103/physrevb.45.13244 [ * * , ( ) ] http://stacks.iop.org/0953-8984/13/8587 [ * * , ( ) ] http://stacks.iop.org/0953-8984/27/i=43/a=435202 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.95.136102 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.116.056802 [ * * , ( ) ] link:\doibase 10.1103/physrevb.69.121404 [ * * , ( ) ] link:\doibase 10.1103/physrevb.75.125422 [ * * , ( ) ] link:\doibase 10.1103/physrevb.49.5599 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/0039-6028(94)90442-1 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/0039-6028(95)00024-0 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevb.81.035422 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.70.694 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.68.1943 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.75.152 [ * * , ( ) ] link:\doibase 10.1103/physrevb.58.890 [ * * , ( ) ] link:\doibase 10.1103/physrevb.26.6186 [ * * , ( ) ] link:\doibase 10.1103/physrevb.53.9204 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.58.2126 [ * * , ( ) ] link:\doibase 10.1103/physrevb.54.15937 [ * * , ( ) ] link:\doibase 10.1103/physrevb.39.865 [ * * , ( ) ] link:\doibase 10.1103/physrevb.59.6421 [ * * , ( ) ] | growth , electronic and magnetic properties of @xmath0-fe@xmath1n atomic layers on cu(001 ) are studied by scanning tunneling microscopy / spectroscopy and x - ray absorption spectroscopy / magnetic circular dichroism .
a continuous film of ordered trilayer @xmath0-fe@xmath1n is obtained by fe deposition under n@xmath2 atmosphere onto monolayer fe@xmath2n
/ cu(001 ) , while the repetition of a bombardment with 0.5 kev n@xmath3 ions during growth cycles results in imperfect bilayer @xmath0-fe@xmath1n .
the increase in the sample thickness causes the change of the surface electronic structure , as well as the enhancement in the spin magnetic moment of fe atoms reaching @xmath4/atom in the trilayer sample .
the observed thickness - dependent properties of the system are well interpreted by layer - resolved density of states calculated using first principles , which demonstrates the strongly layer - dependent electronic states within each surface , subsurface , and interfacial plane of the @xmath0-fe@xmath1n atomic layers on cu(001 ) . |
the relation between temperature @xmath13 and luminosity @xmath15 is a sensitive diagnostic of structural regularity in clusters of galaxies . because the luminosity is governed by the mass of gas in the intracluster medium ( icm ) while the temperature is determined by the total , gravitating cluster mass , the @xmath14 relation can be used as a tool to probe variations in gas fraction the ratio of icm gas to total mass in these systems .
evidence for a correlation between these basic observational quantities has existed since the early days of astronomy ( mitchell 1977 , 1979 ; mushotzky 1978 ; edge & stewart 1991 , david 1993 ) . despite its relative maturity , a key characteristic of this relation was only recently established .
fabian ( 1994 ) demonstrate that the departure of a cluster from the mean @xmath16 relation is correlated with the strength of the emission associated with its cooling flow core .
the picture resulting from this work is of a mixed population which is likely to be time variable .
fabian speculate that cooling flows are a recurrent phenomenon associated with a secular instability in the icm plasma which periodically is interrupted and reset by strong merging encounters .
periodic mergers of varying strength are a natural feature of hierarchical clustering models of structure formation , and numerical simulations suggest that cooling flow features can be erased by mergers ( garasi , burns & loken 1997 ) .
however , the theoretical picture remains incomplete because of the complexities of the physics operating in the cluster core .
regardless of the exact physical mechanisms responsible , the empirical fact is that the cluster population can be broadly classed into two categories by making a cut in cooling flow strength @xmath17 . in this paper , we present and analyze the luminosity temperature relation from a nearly homogeneous data set of weak cooling flow clusters , where @xmath18 defines weak .
the data set is described in 2 , with particular attention paid to systematic errors in determination of both @xmath15 and @xmath13 . in 3 , we present the @xmath14 relation and show that these data place stringent limits on gas fraction variations within this sub population .
imaging data taken from the literature allow us to estimate absolute values of the gas fraction at fixed density contrast in 4 .
we compare results for two different binding mass estimation methods , and emphasize the systematic uncertainty introduced by this choice .
hubble constant dependencies are displayed in the paper via @xmath19 and we assume @xmath20 when quoting numerical values throughout the paper .
we analyse an archival data set , part of which was assembled and discussed by arnaud ( 1994 ) .
a prime objective is to limit both statistical and systematic errors in @xmath21 and @xmath13 , in order to obtain an accurate assessment of the intrinsic dispersion about the mean @xmath16 relation .
[ cols= " < , < , < , < , < , < , < , < , < " , ] the trend of rising gas fraction with increasing radius ( lower @xmath22 ) is apparent with both mass estimation methods , and reflects the more extended nature of the gas density compared to the assumed or inferred dark matter profiles .
gas fraction values are consistently higher in the bm approach , due to the smaller , inferred total masses noted above . as a result , significantly different sample means can be quoted depending on the mass method and extent of radial coverage of the dataset .
compare , for example , @xmath23 for the vt at @xmath9 with @xmath24 for the bm at @xmath25 .
the range becomes wider when one considers subsets of the data defined by temperature .
all the panels in figure ( [ fbarplot ] ) display a trend of decreasing gas fraction at lower @xmath13 , but only the vt method at @xmath9 is deemed statistically significantly ( @xmath26 ) under our @xmath27 error hypothesis . due to the robustness of the bi
weight mean estimator , this conclusion is valid with or without the inclusion of the principal outlier a1060 .
it is perhaps also worth noting that the discrepancy between a1060 and the mean of the cool subsample is smaller under the vt approach .
the value of @xmath28 for the hot subsample in the vt case is consistent with the @xmath29 confidence region @xmath30 ( @xmath20 ) determined by evrard ( 1997 ) from the combined samples of white & fabian ( 1995 ) and david , jones & forman ( 1995 ) .
only 3 of the 26 clusters in the combined study have @xmath8 ( and these have fairly uncertain gas fractions which make negligible contribution to the sample mean ) , so comparison with our high@xmath13 subsample is reasonable .
the agreement between these two studies is not entirely trivial . although there is nearly @xmath31 overlap with our subsample
6 of our 14 are in common with white & fabian we employ here different temperature estimates in many cases ( though all are consistent within the quoted errors ) and we normalize the gas mass differently .
a less obvious trend in the data is the larger variance in gas fractions in the bm case compared to the vt .
this is more apparent in the @xmath13dependent subsamples , since the trend of mean gas fraction with temperature mimics variance in the full vt sample .
the lower scatter in the vt method confirms theoretical predictions that the @xmath5 parameter acts as an extra source of noise in the bm mass estimates ( emn ) .
an increase in scatter is also seen at larger radii / lower densities , consistent with the numerical model predictions and interpreted simply as reflecting the longer relaxation timescales in the outer regions of clusters .
there is strong evidence from the slope of the @xmath14 relation and from the behavior of the image fall - off parameter @xmath5 that cluster structure varies systematically with temperature .
whether the virial gas fraction varies with cluster temperature is a question whose answer is both model and scale dependent .
the model dependency we highlight is the choice of total mass estimation method , while the chosen density contrast sets the scale .
for example , the applied at @xmath32 produces similar mean values in high and low temperature subsamples @xmath33 and @xmath34 , respectively whereas the virial theorem at @xmath35 yields significantly different values in these subsamples @xmath36 against @xmath37 , respectively ( see table 2 and figure [ fbarplot ] , all values scale as @xmath38 ) .
that the implications of the observed luminosity temperature relation are model dependent can be seen by considering the behavior of terms with temperature dependence contained in the defining relation , equation ( [ lxmod2 ] ) .
the structure factor @xmath39 appears to increase with temperature , but this effect is not solely responsible for the steepening of the @xmath14 relation .
the dependence with @xmath13 of the other factors depends on the total mass estimation method . in the vt approach , while @xmath40 scales classically as @xmath41 , @xmath42 increases with @xmath13 . in the bm case , while @xmath43 is found to be nearly constant , the slope of the @xmath44 relation is steeper than @xmath45 , due to the extra @xmath5 factor and its increase with @xmath13 ( equation ( [ massrat ] ) ) .
both cases result in a steepening of the @xmath46 term to a slope greater than the canonical @xmath45 .
the temperature dependence of this term contributes about equally with @xmath39 to the steepening of the @xmath14 relation .
allan & fabian ( 1998 ) take a different approach in analysis of a cluster dataset combining asca spectroscopy with rosat imaging .
they fit asca data to a multi - component spectral model which includes emission from a central cooling flow ( johnstone 1992 ) , absorption from galactic hydrogen , intrinsic hydrogen absorption , and a superimposed background plasma at temperature @xmath13 . compared to simpler fits without cooling flow and intrinsic absorption ,
the best fit temperatures in their complex fits are higher , and the best fit slope of the @xmath14 relation is shallower , @xmath47 .
allen & fabian emphasize that this slope is consistent with the straightforward theoretical expectation of @xmath48 ; however , note that the range of slopes allowed in our fit does overlap the error range quoted above ( derived from a change @xmath49 in their @xmath50 analysis ) .
the approximate nature of our error analysis limits the conclusions which can be drawn regarding gas fraction variations .
a definitive analysis awaits uniform treatment of a large , homogeneous dataset .
there is some supporting evidence for deficient gas fractions in poor groups ( hickson , 1997 and references therein ) , but there also appears to be larger scatter at low temperatures ( , ponman 1996 ) and observational uncertainties on the icm mass remain large in these systems .
in particular the separation of the icm emission from the x ray emission linked to individual galaxies is ambiguous and difficult with present instruments ( mahdavi 1997 , mulchey and zabludoff 1998 ) .
the narrow scatter in the @xmath14 relation of weak cooling flow clusters ( @xmath51 ) indicates that the structure of the intracluster medium in this majority sub class is very regular . under a conservative assumption of no correlation between virial gas fraction and internal , structural variations ( @xmath52 )
, we place upper limits on the _ rms _ percentage variation in virial gas fraction of @xmath53 for the entire sample of 24 clusters , and @xmath2 for the homogeneous ginga subset of 18 clusters .
these limits are conservative because the imaging data supports the point of view that gas in low @xmath13 clusters is more extended than that of high @xmath13 clusters ( figure [ betaplot ] ) .
mohr & evrard ( 1997 ) find the same result with a non parametric analysis of images .
they also derive limits on gas fraction variations similar to those found here based on the tightness of the size temperature ( st ) relation .
the agreement is non trivial because the st analysis is essentially _ local _ , being based on an differential isophotal area , while the @xmath14 relation analysis is _ global _ , based an the integrated emission of the entire cluster atmosphere .
both analyses are consistent with idea that energy input from galactic winds has led to extended gas atmospheres in low @xmath13 clusters ( david , forman & jones 1990,1991 ; white 1991 ; metzler & evrard 1997 ; cavaliere 1997 ; ponman 1996 ) . in wind models ,
gas loss is correlated with structural extension of the gas , leading to @xmath54 . the inferred variance in gas fraction derived from equation ( [ delta_f2 ] ) could be substantially smaller than the limits phrased above if winds have played an important evolutionary role .
numerical simulations of clusters used by mohr & evrard ( 1997 ) exhibit less than 5% fractional variation in their virial gas fractions , and a similar value is seen in models with galactic winds ( metzler & evrard 1997 ) . the expected `` cosmic variance '' in virial gas fractions is thus quite small .
comparison of a number of different simulation techniques paints a similar picture .
frenk ( 1998 ) find agreement at the 5% level ( @xmath55 deviation ) for the virial gas fraction of a single cluster determined by twelve independent gas dynamic methods .
this indicates that systematic uncertainties in the tested numerical simulation techniques are small for the simplest case of a single phase icm .
however , it is important to note that the quoted simulations ignore galaxy formation and its attendant multi
phase , gas structure . introducing galaxy formation via cooling and
star formation is likely to increase the scatter in icm gas fraction values , and it will be interesting to see whether viable models can do so without violating the observational limits presented here .
evidence for trends in @xmath56 with cluster temperature is ambiguous .
a key uncertainty lies in estimation of the total , binding mass .
the approach produces masses which are characteristically smaller than the virial theorem method as currently calibrated by numerical simulations . for a cluster with @xmath57 ,
the masses inferred at @xmath4 differ substantially , @xmath58 .
this different behavior , coupled with the temperature dependence of @xmath5 , leads to differing conclusions on whether the mean gas fraction varies with cluster temperature .
for the same reason there is some ambiguity in the origin of the steepening of the @xmath14 relation .
the fact that the gas is more concentrated in high - kt cluster explains at least half of the effect , directly through the structure factor in the x ray luminosity . in the approach ,
the temperature dependence of the icm structure can fully account for the @xmath14 relation slope , via the additional dependence of the total mass with @xmath5 . on the other hand , in the vt approach
, the gas mass fraction increases with @xmath13 and this effect contributes about equally to the @xmath14 relation slope increase .
both effects , lower gas fraction and inflated gas distribution in low t systems , are expected in models incorporating the effects of galactic winds .
independent methods for estimating virial masses are crucial , and weak gravitational lensing provides an important approach .
allen ( 1997 ) provides evidence that weak lensing and mass estimates are consistent , but the present errors at radii between @xmath59 or @xmath60 are too large to discriminate between the bm and vt approaches .
we suggest that weak lensing analysis of a small sample of moderate redshift clusters selected to have @xmath61 from imaging could settle the score , if individual measurement uncertainties could be kept below a few tens of percent at @xmath62 .
future missions which provide simultaneous imaging and spectroscopic capability ( axaf , xmm , astro - e ) will provide improvements over the current dataset
. the ideal analysis would be one of a flux limited sample with central , cooling flow regions excised ( markevitch 1998 ) , using a nearly homogeneous dataset of temperatures , luminosities and surface brightness maps derived from the same telescope .
such data would minimize variance caused by instrumental uncertainties and open the door to very high precision measurements of intracluster gas fractions .
finally , small scatter in virial gas fractions lends support to arguments against an einstein desitter universe based on cluster baryon fractions and primordial nucleosynthesis ( white 1993 ; evrard 1997 ) . the standard counterargument to this idea
has been that the physics of the icm is more complicated than is assumed in most analytic or numerical models
. however , if additional physics beyond gravity , shock heating and galactic wind input is at work in cluster icm atmospheres , then the extra processes involved must conspire to keep the scatter in the observed @xmath14 and st relations small . since @xmath15 and @xmath63 ( the `` size '' in the st relation ) are manifestly different measures ( global _ vs. _ local ) , it is not obvious that interesting physical mechanisms , such as a strongly multi phase icm ( gunn & thomas 1996 ) , can be added in such a way as to preserve the current agreement between observations and current models incorporating the effects of galactic winds on the icm .
we would like to thank dr koyama for useful correspondence on the ginga observations of virgo and dr yamashita for providing us with the ginga results on a262 and a2319 .
we thank alain blanchard for a careful reading of the manuscript .
this work was supported by nasa through grant nag5 - 2790 and by the cies and cnrs at the institut dastrophysique in paris .
aee is grateful to the members and staff of the iap for the kind hospitality extended during a sabbatical visit . | we re - examine the luminosity temperature relation using a nearly homogeneous data set of 24 clusters selected for statistically accurate temperature measurements and absence of strong cooling flows . the data exhibit a remarkably tight power
law relation between bolometric luminosity and temperature with a slope @xmath0 . with reasonable assumptions regarding cluster structure ,
we infer an upper limit on fractional variations in the intracluster gas fraction @xmath1
. a strictly homogeneous ginga subset of 18 clusters places a more stringent limit of @xmath2 . imaging data from the literature are employed to determine absolute values of @xmath3 within spheres encompassing density contrasts @xmath4 and 200 with respect to the critical density . comparing binding mass estimates based on the virial theorem ( vt ) and the hydrostatic , ( bm )
, we find a temperature dependent discrepancy in @xmath3 between the two methods caused by systematic variation of the outer slope parameter @xmath5 with temperature .
mean values ( for @xmath6 ) range from @xmath7 for cool ( @xmath8 ) clusters using the vt at @xmath9 to @xmath10 for hot ( @xmath11 ) clusters using the bm at @xmath12 .
there is evidence that cool clusters have a lower mean gas fraction than hot clusters , but it is not possible to assess the statistical significance of this effect in the present dataset .
the @xmath13 dependence of the icm density structure , coupled with the increase of the gas fraction with @xmath13 in the vt approach , explains the steepening of the @xmath14 relation . the small variation about the mean gas fraction within this majority sub population of clusters presents an important constraint for theories of galaxy formation and supports arguments against an einstein
desitter universe based on the population mean gas fraction and conventional , primordial nucleosynthesis .
the apparent trend of lower gas fractions and more extended atmospheres in low temperature systems are consistent with expectations of models incorporating the effects of galactic winds on the intracluster medium .
+ # 1to
0pt#1 = 1 a&a # 1 6 cosmology : theory cosmology : observation clusters : galaxies : general dark matter |
discovery of the universe accelerated expansion @xcite stirs up theoretical analysis aimed at explanation of the observational data . on the one hand ,
most of approaches are based on exploration of the new kinds of substance ( quintessence , chaplygin s gases , phantom field , ghost condensate , and so on ) or of the new concepts of space - time ( extra - dimensions , holographic principle , string multiple vacuum ) @xcite . on the other hand
, it was proposed that such an explanation can be based on an assumption that universe or its constituents are inhomogeneous @xcite .
the extreme point of view @xcite is to consider universe as inhomogeneous at the super - hubble scales and to assume that we live at the edge of `` hubble - bubble '' and thereby can observe either accelerated or decelerated expansion of universe , which is not described by the friedmann equations .
also , one can assume that the accelerated expansion is produced by an internal motion of matter on a smaller scale ( like a galaxy formation one @xcite ) if the corresponding averaging procedure is defined for the universe scale factor @xcite .
nevertheless , it is still attractive to consider the qft vacuum in a non - stationary space - time ( i.e. in an expanding universe ) as the source of the unverse acceleration .
this can be achieved by exploring of two scales of universe : @xmath0 and @xmath1 . here
@xmath0 is the `` size '' of universe ( of the order of its curvature or its inverse hubble constant for a flat space - time ) , @xmath2 is the planck mass associated with an appearance of new physics . whereas the direct ( `` naive '' ) consideration results in the huge value of vacuum density @xmath3 , the dimensional regularization of divergent expressions in a manner of minkowski qft gives the tiny value @xmath4 @xcite .
in contrast to these approaches , there is a number of works @xcite obtaining @xmath5 that is consistent with the observed value of the universe acceleration .
@xcite is simply an observation that the schwarzschild radius @xmath6 of universe with the `` mass '' @xmath7 has not to exceed the universe `` size '' @xmath0 .
this excludes the `` naive '' value @xmath3 and results in @xmath5 as an upper limit .
other works are based on the zeldovich s idea @xcite assuming that the source of the universe acceleration is not in itself the huge qft vacuum energy but some another quantity , namely , the square root of dispersion of the vacuum energy fluctuations .
although the direct uv cut - off on the planck level in the friedmann equation does not provide the observed value of the acceleration parameter ( and is inconsistent with @xcite ) , one can use it in the equation of motion of the universe scale factor @xcite .
it is found , that the effective cosmological constant results from the vacuum fluctuations of the massless scalar field and its value is concistent with @xmath8 .
the point is that , in contrast to the friedmann equation which is held only up to some constant in such an approach , the equation of motion contains difference of the potential and kinetic energies of the field oscillators . in minkowski space - time this difference is exactly zero , because the virial theorem for an oscillator states that the kinetic energy is equal to the potential one in the virial equilibrium .
however , in the expanding universe the difference is proportional to the hubble constant squared .
this last tiny quantity is multiplied by the planck - order momentum cut - off squared that gives true order of the acceleration parameter .
the aim of this work is to analyze a number of fundamental fields , which can contribute to the cosmological constant , and to investigate a possibility of exact determination of the uv cut - off from the observed value of the acceleration parameter .
in the case of the homogeneous isotropic flat universe filled with the different species of matter , lagrangian has the form @xcite : @xmath9 where the metric @xmath10 is assumed and @xmath11 is the planck mass .
matter lagrangians are @xmath12 @xmath13 where @xmath14 is the spinor field and @xmath15 is the scalar field and three dimensional volume of integration is assumed to be unity .
let us turn to the fourier - transform @xmath16 that leads to the scalar field lagrangian in the form @xmath17 in the quasi - classical picture , where the universe scale factor is considered classically whereas the scalar field is quantum , equations of motion are : @xmath18 here @xmath19 denotes averaging over the vacuum state .
quantization of the scalar field @xcite @xmath20 leads to the operators of creation and annihilation with the commutation rules @xmath21=1 $ ] .
the complex functions @xmath22 are @xmath23 for @xmath24 . in the general case they satisfy the relations
@xcite : @xmath25 eqs .
( [ rel ] ) admit the formal wkb solution @xcite @xmath26 where function @xmath27 satisfies the equation @xmath28 adiabatic approximation consists in setting @xmath29 . at first
, we shall consider the massless case @xmath30 in which the equation of motion contains exact difference of the kinetic and potential energies of the field oscillators .
this difference is zero in the vacuum state of minkowski space . in the expanding universe it takes the form
@xmath31 when the adiabatic approximation for the function @xmath32 is used . using ( [ 15 ] ) in ( [ sys ] ) yields the equation of motion for the universe scale factor in the form @xmath33 coming from the summation over @xmath34 ( in the general case it has to include summation over particle species ) to the integration , we have @xmath35 where @xmath36 is the cutting parameter of the physical momentums and @xmath37 is the present day ( @xmath38 ) scale factor . finally , we come to the equation @xmath39 where the density of classical matter ( dust ) @xmath40 and the number @xmath41 of species of scalar particles are introduced . here
@xmath42 should be read as denoting the matter relative density and @xmath43 .
consideration of eq .
( [ 14 ] ) at @xmath38 allows determining the uv cut - off @xmath44 of physical momentum from the observed present day value of the universe acceleration @xmath45 now let us return to the massive particles . in this case there is no an exact difference of the kinetic and potential energies in the equation of motion because the multiplier @xmath46 appears in eq .
( [ sys ] ) .
subtraction of the main terms corresponding to the minkowski space - time can be used in this case as a primitive renormalization .
let us remind that for @xmath24 @xmath47 the renormalization consists in rewriting of the equation of motion in the form @xmath48 where @xmath49 the simplest way of further consideraton is to assume some particular dependence @xmath50 and to calculate the quantity @xmath51 . as it was done in @xcite
, one can choose the dependence @xmath52 which can be roughly associated with our present universe because it gives the acceleration parameter @xmath53 and @xmath54 ( here @xmath55 and @xmath56 is the hubble constant ) .
calculation of the scalar field contribution to the universe acceleration up to second order in @xmath57 gives @xmath58 where replacement of summation by integration over the momentum @xmath59 is proceeded and @xmath60 .
one can see , that the mass terms are suppressed by the multiplier @xmath61 and are negligible , if one assumes that the mass of a particle is mach smaller than the planck mass .
thus , one can conclude that the mass of the scalar particles has no impact .
to consider the fermions in an expanding universe , we keep to the earlier work @xcite and the works @xcite , where preheating of universe was considered ( see also @xcite ) .
after decomposing in the complete set of modes @xmath62 , the fermionic lagrangian ( [ ff ] ) takes the form @xmath63 the equations of motion are @xmath64 where quasi - classics is implied and dots denote matter ( dust ) , scalar field , photon field and other terms .
fermion field is quantized as @xmath65 where the bispinor is : @xmath66 @xmath67 and @xmath68 .
the bispinor @xmath69 is expressed as @xmath70 , where the symbol @xmath71 denotes the transpose vector and @xmath72 . the functions @xmath73 satisfy the equations @xmath74 eq .
( [ 1 ] ) reflects the time - dependence dictated by the bispinor equation of motion .
( [ rel1 ] ) is required to satisfy the anti - commutator relation @xmath75 , so that the operators @xmath76 satisfy @xmath77 and the similar relation is valid for @xmath78 . also , one has to note that the left hand side of eq .
( [ rel1 ] ) is the integral of motion of eq .
( [ 1 ] ) .
( [ 1 ] ) admits the wkb solution : @xmath79 where the function @xmath27 satisfies the equation : @xmath80 it should be noted , that eq .
( [ wkf ] ) gives @xmath81 at @xmath30 and this can be seen directly from eqs .
( [ 1 ] ) , ( [ rel1 ] ) giving @xmath82 . in this case , the elementary calculations lead to @xmath83 thus , according to ( [ rr ] ) the massless fermions do not contribute to the universe acceleration . to analyze the massive fermions ,
let us assume the particular dependence @xmath50 given by ( [ aeta ] ) and calculate the quantity @xmath84 contained in eq .
( [ rr ] ) .
one can see that after subtraction of the first term ( which is equivalent to the `` renormalization '' ( [ renorm ] ) in the case of the scalar field ) , the remaining terms are suppressed by the multiplier @xmath85 in comparison with the acceleration term created by the vacuum of the massless scalar field .
it seems that the negligible contribution of the massive terms is a general feature .
indeed , the quantity @xmath86 is of the forth - order in mass . from the dimension arguments
, it can be proportional to @xmath87 , @xmath88 , @xmath89 and two last terms are negligible in comparison with the first one at @xmath90 .
note , that the term @xmath91 is forbidden by the time - reversal invariance arising from the fluctuation origin ( i.e. the vacuum fluctuations ) of the quantity under consideration .
as one can see , that the massless fermions do not contribute to the universe acceleration .
the deep - laid reason ( see for instance @xcite ) is conformality of the fermionic lagrangian if @xmath30 .
other conformal lagrangians correspond to both photonic and scalar fields with the @xmath92 term for the latter @xcite . for shortness
it is sufficiently to consider only the last lagrangian , which takes the form : @xmath93 where @xmath94 .
one can see , that the field oscillators are uncoupled from the universe scale factor in the case of @xmath30 and , thereby , do not contribute to the equation of motion for it .
equivalent conclusion , that the massless conformal scalar field does not produce a vacuum polarization has been done in @xcite .
let us turn to the pure gravity .
small oscillations of the graviton oscillators near an equilibrium point can be considered on the basis of the perturbation theory @xcite .
it is known that the scalar and vector modes can not produce free oscillations without the matter fields @xcite . on the other hand , it is possible to expect that the matter fields are uncoupled from the scalar and vector perturbations at the planck frequencies , that is the frequencies providing main contribution to the vacuum energy within the framework of our approach .
thus , the remaining perturbations are the tensor ones existing in an empty space .
let us represent metric in the form , thus the quadratic terms on @xmath95 are kept in the metric . ]
@xmath96 where @xmath97 .
all components of the perturbation @xmath98 containing at least one zero index are equal to zero and the spatial components are represented as @xmath99^j_n\,\phi_{\bm k}(\eta)+\bigl[\mathcal y_{\bm k}\bigr]^j_n \,y_{\bm k}(\eta)\biggr)e^{i \bm k
\bm r}. \label{m2}\ ] ] the symmetric tensors @xmath100 and @xmath101 describing two possible polarizations have the following properties : @xmath102=tr[\mathcal y_{\bm k}]=0 $ ] , @xmath103=tr[\phi_{\bm k}\,\phi_{-\bm k}]=1 $ ] , @xmath104 .
for instance , when the @xmath34-vector is directed along @xmath105-axis , these tensors look as : @xmath106 for an arbitrary direction , they can be obtained from these expressions through the so(3)-transformation . the next step is to substitute eqs .
( [ metr ] ) , ( [ m2 ] ) in the gravity action @xmath107 , where @xmath108 according to @xcite .
finally , in the second - order in @xmath109 , @xmath110 we come to @xmath111 thus the tensor wave contributes to the vacuum energy as two scalar fields .
in the case of the massless particles and when the conformal time is used , it is shown that the equation of motion of the universe scale factor contains the difference of potential and kinetic energies of field oscillators .
calculation of the mean value of this quantity over the vacuum state leads to the universe accelerated expansion .
the observed value of the universe acceleration allows determining the uv cut - off of the physical momentum given by eq .
( [ cut ] ) , which contains the dust density and the number of species of scalar fields @xmath41 .
let us make some notes about the number of species of scalar particles . at first sight
, there appear to exists a number of the scalar particles like pions . however , the momentum @xmath112 of a virtual particle appears in the loop diagram and the main contribution arises from @xmath112 of the planck mass order . under the circumstances ,
the composite particles decompose into their constituents .
thus , @xmath41 should be read as a number of the fundamental scalar fields . in a minimal variant of the standard model
there is the su(2 ) duplet of complex scalar fields @xcite , i.e. four scalar degrees of freedom .
adding two degrees of freedom of the tensor gravitational wave , one comes to @xmath113 and the substitutions of @xmath53 and @xmath114 in eq .
( [ cut ] ) lead to the uv cut - off of the physical momentum : @xmath115 .
still , the scale of the grand unfed theory @xmath116 gev seems closer to the planck scale and , hence there are ten scalar degrees of freedom for su(5 ) ( i.e. total twelve ) . then @xmath117 .
scalar sectors of both the string theory and the supersymmetry one providing a number of the scalar superpartners of fermions demand a deeper insight .
let us remind that photons , fermions and conformally coupled scalars do not contribute to the universe acceleration .
the masses of all particles gives a negligible contribution after separation of the term corresponding to the minkowski space .
40 perlmutter s et al 1999 astrophys .
j * 517 * 565 riess a et al 1998 astron .
j * 116 * 1009 copeland e j , sami m and tsujikawa s _ preprint _ hep - th/0603057 kolb e w ,
matarrese s , notari a , and riotto a 2005 phys .
d * 71 * 023524 buchert t , larena j and alimi j - m 2006 class.quant.grav .
* 23 * 6379 rsnen s _ preprint _ astro - ph/0607626 birrell n d and davis p c w 1982 _ quantum fields in curved space _
( cambridge , univ . press ) onemli v k , woodard r p 2004 phys.rev .
* 107301 parker l and raval a 1999 phys . rev .
d * 60 * 063512 caldwell r r , komp w , parker l and vanzella d a t 2006 phys .
d * 73 * 023513 cohen a g , kaplan d b and nelson a e 1999 phys .
rev . lett . * 82 * 4971 padmanabhan t 2005 class .
. grav . * 22 * l107 padmanabhan t and singh t p 1987 class .
4 * 1397 bombelli l , koul r k , lee j - h and sorkin r d 1986 phys.rev .
d * 34 * 373 sredinski m 1993 phys
lett . * 71 * 666 brustein r , eichler d , foffa s and oaknin d h 2002 phys .
d * 65 * 105013 gurzadyan v g and xue s - s 2003 mod .
a * 18 * 561 yarom a and brustein r 2005 nucl .
b * 709 * 391 elizalde e , nojiri s , odintsov s d and wang p 2005 phys .
d * 71 * 103504 zeldovich y b 1967 pisma zh .
* 6 * 1050 [ jetp lett . * 6 * 316 ] cherkas s l , kalashnikov v l _ preprint _
gr - qc/0604020 cherkas s l , kalashnikov v l _ preprint _
astro - ph/0611795 hartle j b , hawking s w 1983 phys
* 2960 cherkas s l and kalashnikov v l 2006 grav . cosmol .
* 12 * 126 parker l 1971 phys .
d * 3 * 346 greene p b and kofman l 2000 phys .
d * 62 * 123516 peloso
m and sorbo l 2000 jhep * 0005 * 016 maamache m and lakehal h 2004 europhys . lett . * 67 * 695 zhang z - g 2006 phys .
* 74 * 218 berestetskii v b , lifshitz e m and pitaevskii l p 1982 _ quantum electrodynamics _
( oxford , pergamon press ) cherkas s l 1994 proc .
belarus , ser .
fiz .- mat .
, * 2 * 70 [ in russian ] prokopec t and woodard r p 2004 am.j.phys .
* 72 * 60 zeldovich ya b and starobinsky a a 1972 zh .
fiz . * 61 * 2161 [ jetp * 34 * ( 1972 ) 1159 ] lifshits e m and khalatnikov i m 1963 usp . fiz .
nauk , * 80 * 391 [ sov
* 6 * ( 1964 ) 496 ] landau l d and lifshitz e m 2000 the classical theory of fields ( oxford : butterworth - heinemann ) commins e d and buksbaum p h 1983 _ weak interactions of leptons and quarks _
( cambrige , university press ) | it is shown that use of the equation of motion of the universe scale factor allows calculation of the contribution of the vacuum fluctuations to the acceleration of the universe expansion .
renormalization of the equation is needed only in the case of massive particles . under a known number of the different kinds of fundamental fields ,
this provides the determination of momentum of the ultraviolet cut - off from the observed value of acceleration . |
the resonance @xmath10 is catalogued as a four star resonance in the pdg @xcite .
the @xmath11 is also catalogued as a three star resonance and has been advocated by many groups , the latest ones @xcite .
however , the @xmath11 is not `` seen '' in the gwu analysis of @xcite nor in former analyses of this group @xcite . in a recent unified chew - mandelstam
said analysis of pion photoproduction data the resonance is not needed again @xcite , and it is also not included in the maid analysis of photoproduction data @xcite where only four star resonances are considered , but it is one of the resonances included in the analysis of the bonn - gatchina group @xcite and shows more clearly in the analysis of the @xmath12 data in @xcite .
one common conclusion from @xcite is the strong coupling of the @xmath11 state to the @xmath2 ( d - wave ) channel , something not intuitive nor expected from ordinary quark models .
we should note that , as discussed in @xcite there could be some reason to miss some resonances in the analysis if the resonances are too wide ( @xmath13500 mev ) or they posses a small ratio ( br@xmath14 ) to the channel under consideration . in view of the current discussion about these two resonances , and particularly the doubts casted on the existence of the @xmath11
, we have done a different analysis , purely theoretical , although paying attention to known data on the @xmath15 channel .
the work consist on taking four coupled channels , also considered in @xcite , the @xmath0 ( s - wave ) , @xmath1 ( d - wave ) , @xmath2 ( s - wave ) and @xmath2 ( d - wave ) .
we look in the region of @xmath16 mev and take the interaction between these channels from the local hidden gauge lagrangians @xcite , then solve the bethe salpeter equation in coupled channels and from this fully unitary approach we find two resonances dynamically generated by the interaction of these channels , which are naturally associated to the @xmath4 and @xmath5 resonances .
the freedom of the theory in the choice of subtraction constants to regularize the loops is used to fit the data on the @xmath1 ( d - wave ) amplitude .
the study provides the poles and the residues of the resonances and their coupling to the different channels , widths and partial decay widths .
a fair agreement is obtained with phenomenology and the two resonances appear as poles in the complex plane . in particular
a relatively strong coupling of the @xmath5 to the @xmath2 ( d - wave ) suggested in @xcite is confirmed .
the results that we find are also illustrative on why some analyses do not find a compelling need for the @xmath5 resonance and we clarify the situation by comparing the amplitudes obtained in this case with the @xmath8 amplitude in the scalar - isoscalar sector , where the @xmath17 and @xmath6 resonances appear .
the presence of a first resonance with a width that makes it overlap with a second one with the same quantum numbers has consequences on the second resonance and the amplitude with these quantum numbers , producing a zero of the @xmath1 or @xmath8 amplitudes , respectively , in the vicinity of the second resonance , which makes its coupling to these channels very weak and makes difficult the identification of the @xmath5 resonance in the @xmath1 channel .
we follow the formalism of the hidden gauge interaction of vector mesons @xcite . for the interaction of the vectors among themselves
we have the three vector lagrangian @xmath18 where @xmath19 is the su(3 ) matrix for the nonet of the @xmath20 @xmath21 and @xmath22 , with @xmath23=93 mev . in the same way , the coupling of the vectors to pseudoscalar mesons
is given by @xmath24v^{\nu}\rangle \ , \label{lagrvpp}\ ] ] where here @xmath25 is the su(3 ) matrix of the pseudoscalar mesons , @xmath26 the lagrangian for the coupling of the vector to the baryon is given by @xmath27\rangle + \langle \bar{b}\gamma_{\mu}b \rangle \langle v^{\mu}\rangle \right ) \ , \label{lbbv}\ ] ] where @xmath28 is now the su(3 ) matrix of the baryon octet @xmath29 for the transitions @xmath30 we need the vpp lagrangian of eq .
( [ lagrvpp ] ) and one of the two pseudoscalar mesons is exchanged between the external vp and the baryon .
the coupling of the pseudoscalar to the baryon for the two su(3 ) octets is given by @xmath31\rangle + \frac{d}{2 } \langle \bar{b}\gamma_{\mu } \gamma_5 \left\lbrace u^{\mu } , b \right\rbrace \rangle \label{lbbp}\ ] ] where @xmath32 , @xmath33 . a generalization for the case of @xmath34
is given below .
in a nonrelativistic approximation that we follow , eq . ( [ lbbp ] ) provides a vertex @xmath35 where @xmath36 is the incoming momentum of the @xmath37 , and other charge combinations are trivially derived using isospin symmetry . in practice one substitutes @xmath38 by @xmath39 where @xmath40 , empirically determined .
similarly for the @xmath34 transition we take @xmath41 where the minus sign stems from the phase convention @xmath42 of isospin , with an empirically value of @xmath43 .
the operator @xmath44 is the transition spin operator from spin @xmath45 to @xmath46 normalized such that @xmath47 with @xmath48 written in spherical basis and @xmath49 the clebsch - gordan coefficient . as shown in ref .
@xcite , the leading term of the @xmath50 interaction involves the three vector vertex of eq .
( [ l3v ] ) , with one vector meson exchanged , and the coupling of this exchanged vector to the baryon , given by eq .
( [ lbbv ] ) . the potential provided by this term , keeping the dominant @xmath51 term in eq .
( [ lbbv ] ) , is given by @xmath52 where @xmath53 are the energies of the incoming and outgoing vector mesons .
the result of eq .
( [ kernel ] ) with the @xmath54 factor for the polarization of the vector mesons stems from considering the three momentum of the external vectors small with respect to the mass of the vector mesons @xcite .
the @xmath55 coefficients can be found in appendix a of ref .
@xcite , where the subindex @xmath56 and @xmath57 correspond to the different channels for all the states of isospin and strangeness .
this potential has been used as the input of the bethe - salpeter equation to study the scattering matrix , @xmath58^{-1}\ , v \label{eq : bethe}\ ] ] where g is the loop function of a vector meson and a baryon which is calculated in dimensional regularization , as shown in ref .
@xcite , is given by @xmath59 \right\ } \
, \label{eq : gpropdr}\end{aligned}\ ] ] with @xmath60=800 mev a regularization scale and @xmath61 the subtraction constant , depending on the channel . in the cases where the iteration of the bethe - salpeter equation includes the @xmath20 meson or @xmath62 , which have relatively large widths , a convolution of the loop function @xmath63 with the mass distribution is needed .
so the loop function with the convolution for the case of the @xmath20 meson would be @xmath64 where @xmath65 is normalized with @xmath66 considering the width of the @xmath20 , @xmath67=149.4 mev , the @xmath68 function is energy dependent and is given in ref .
@xcite as @xmath69 with @xmath70 for the @xmath20 using @xmath71 where @xmath72 is the kllen function and @xmath73 is the nominal width of the @xmath20 .
similarly , for the case of the @xmath62 : @xmath74 where @xmath65 is normalized with @xmath75 where for the width of the @xmath62 we take @xmath76=120.0 mev . once the scattering matrix is evaluated , some peaks appear that can be associated to states .
next step is to find the poles associated to those peaks , in order to obtain the couplings of these states to the different channels .
the method used is to search poles in the second riemann sheet , changing the momentum @xmath36 to @xmath77 in the analytical formula of the @xmath63 function when @xmath78 is over the threshold of the corresponding channel .
using this method one can find poles , as @xmath79 , where the real part correspond to the mass of the resonance and the imaginary part is half of the width of this state .
however , the convolution of the @xmath63 function eventually can make the pole disappear in channels with the @xmath20 meson . in this case
one can study the amplitude in the real axis using that near the peak the t matrix will be as @xmath80 where @xmath81 is the position of the maximum and @xmath73 the width at half - maximum .
the constants @xmath82 and @xmath83 are the couplings of the resonances to the channels @xmath56 , @xmath57
. then one can take the diagonal channel and obtain @xmath84 where the coupling @xmath82 has an arbitrary phase .
with one coupling determined , we can obtain the other ones from the @xmath85 matrices using eq .
( [ eq : polematrix ] ) , given by @xmath86 once , we obtain the couplings of the resonances for each channel , we can calculate the partial decay widths using the equation @xmath87 in the case of the decay channel @xmath0 , both resonances are under the threshold .
so the momentum used in the previous equation should be imaginary .
experimentally the decay of the resonance to @xmath0 is observed because the width of the @xmath20 is big enough to allow the decay , although the resonance is under the threshold . in order to generate this effect in our calculation of the partial decay width
, we make the convolution of the momentum with the mass of the @xmath20 .
@xmath88 here @xmath89 is the same normalization as used in eq .
( [ norm ] ) .
as we have commented in the first section , the channels involved in our study are the @xmath0 ( s - wave ) , @xmath1 ( d - wave ) , @xmath2 ( s - wave ) and @xmath2 ( d - wave ) all of them in isospin @xmath90 . in order to develop our calculation , we need the diagrams of the elastic interaction and the transitions diagrams as well . in fig .
[ fig : diagrams ] we show those diagrams . + the evaluation of the diagrams shown in fig . [
fig : diagrams ] leads us to the following vertices for the transitions ( see appendix [ app : vertices ] for details ) @xmath91 @xmath92 @xmath93 @xmath94 @xmath95 where we take @xmath40 and @xmath43 , @xmath96 and @xmath97 are the energies of the @xmath20 and the @xmath98 in the center of mass . the term of fig .
[ fig : diagrams](e ) is the kroll - ruderman term which appears in the @xmath99 transition . it has the type @xmath100 and only involves @xmath101 , providing the term @xmath102 in the bracket of eq .
( [ t3 ] ) . in order
to include the @xmath103 transitions , we use the same procedure as used in ref .
@xcite where the unknown potentials are introduced with a parameter @xmath104 and the momenta of the l=2 transition . in our case
those channels are @xmath105 , @xmath106 , @xmath107 . with the purpose of working with more suitable parameters
@xmath108 , we normalized them with the pion mass as following .
@xmath109 @xmath110 @xmath111 where @xmath112 are the momenta of each channel . with this notation
@xmath108 have no dimensions and are of the order of 0.01 .
note that the l=2 transition leads us to introduce a @xmath113 term , which at high energies have a very fast grow . to control this divergence
we introduce the blatt - weisskopf barrier - penetration factors ( see ref .
@xcite ) . in the case of @xmath103
, we have a substitution as @xmath114 here @xmath115 where @xmath116 fm .
we decided to introduce this factor normalized at the energy of 1700 mev .
let us call @xmath117 with @xmath118 the momentum for a channel @xmath56 at 1700 mev .
we introduce a factor @xmath119 defined as @xmath120 so , the potential will be now @xmath121 @xmath122 @xmath123 using this notation we obtain a more convenient framework for the fit .
in order to determine the unknown parameters @xmath108 we make a fit of the partial wave amplitude data of the @xmath1 scattering in @xmath15 of ref . @xcite . in our case
we fit both the real and imaginary parts in an energy range of 1400 to 1800 mev using the gradient fit method of @xmath124 . in the fit
we have not only the parameters of the transition with l=2 , but also the subtraction constants @xmath125 of the loop function . we know that with a regularization scale of @xmath126 , the subtraction constant have a natural size of -2 for an s - wave , but as shown in ref . @xcite
their fit gives subtraction constants for the d - wave larger than in the case of s - wave amplitudes .
for the analysis of the experimental data we need to normalize the amplitude using eq .
( 7 ) of ref .
@xcite which relates our amplitude with the experimental one .
@xmath127 where @xmath128 and @xmath129 are the baryon mass an the on - shell momentum of the specific channel . in fig .
[ fig : fit ] we can see the result of the fit , and in table [ tab : fit ] we show the parameters obtained in the fit . for the estimation of the theoretical errors we follow the criteria of ref .
@xcite where they modify the value of the parameter until the @xmath124 increases eight units ( which is the procedure to get a 68@xmath130 confidence level in the case of seven parameters ) .
( d - wave ) of ref .
@xcite . ]
.results of the parameters obtained with the fit .
the first row are the parameters and the second row their errors . [ cols="^,^,^,^,^,^,^",options="header " , ] llllcccc + & & & + & pole&mass(mev)&@xmath73(mev)&@xmath131&@xmath132&@xmath133&@xmath134 + this work & ( 1467,83 ) & & & 5.09&18.37 & 8.25 & 65.38 + pdg @xcite & ( 1510,55 ) & 1515 - 1525 & 100 - 125 & 15 - 25 & 10 - 20 & 10 - 15 & 55 - 65 + manley92 @xcite & & 1524@xmath1354 & 124@xmath1358 & 21@xmath1354 & 5@xmath1353&15@xmath1354&59@xmath1353 + manley12 @xcite & ( 1501,56 ) & [email protected]&117@xmath1351 & [email protected] & [email protected]&[email protected]&[email protected] + cutkosky79 @xcite & ( 1510,57 ) & 1525@xmath13515 & 125@xmath13525 & & & & 58@xmath1353 + vrana00 @xcite & ( 1504,56 ) & 1518@xmath1353 & 125@xmath1354 & 9@xmath1351&15@xmath1352&11@xmath1352&63@xmath1352 + toma08 @xcite & ( 1509,57 ) & 1520@xmath13510 & 125@xmath13515 & 13@xmath1355&12@xmath1354&14@xmath1355&58@xmath1358 + anisovich12 @xcite & ( 1507,56 ) & 1517@xmath1353 & 114@xmath1354 & & 19@xmath1354 & 9@xmath1352&62@xmath1353 + ardnt06 @xcite & ( 1515,57 ) & [email protected]&[email protected]&&&&[email protected] + llllcccc + & & & + & pole&mass(mev)&@xmath73(mev)&@xmath131&@xmath132&@xmath133&@xmath134 + this work & ( 1665,78 ) & & & 8.02&4.51&13.36&28.54 + pdg @xcite & ( 1700,75 ) & 1650 - 1750 & 100 - 250 & @xmath13635 & 10 - 90 & @xmath13620 & 12@xmath1355 + manley92 @xcite & & 1737@xmath13544 & 249@xmath135218&13@xmath13517&5@xmath13510&80@xmath13519&1@xmath1352 + manley12 @xcite & ( 1662,55 ) & 1665@xmath1353 & 56@xmath1358 & 38@xmath1356 & 31@xmath1359&3@xmath1352&[email protected] + cutkosky79 @xcite & ( 1660,38 ) & 1670@xmath13525 & 80@xmath13540 & & & & 11@xmath1355 + vrana00 @xcite & ( 1704,78 ) & 1736@xmath13533 & 175@xmath135133 & 7@xmath1351&11@xmath1351&79@xmath13556&4@xmath1351 + toma08 @xcite & ( 1710,78 ) & 1740@xmath13520 & 180@xmath13530 & 20@xmath13515&10@xmath1355&20@xmath13511&@xmath137 + anisovich12 @xcite & ( 1770,210 ) & 1790@xmath13540 & 390@xmath135140 & & 72@xmath13523 & @xmath13810&12@xmath1355 + as we can see in table [ tab : couplings ] , the @xmath139 couples mostly to the channel @xmath0 , which is closed for the nominal mass of the @xmath20 , but although the mass of the resonance is under the @xmath0 threshold , using eq .
( [ prhocon ] ) we can generate a momentum giving a small partial decay width . in comparison with the experiments ,
the decay width of @xmath0 to @xmath139 is smaller but of the same order of magnitude .
note that in experimental analyses one evaluates this rate subtracting the other ones from the total width . in either method
the uncertainties for this closed channel are necessarily large . in the case of @xmath140
the coupling of @xmath0 is smaller but , as we are closer to the threshold , the decay width is of the same order of magnitude as for the @xmath139 .
the result for the decay to @xmath0 is in a good agreement with experiment considering the experimental uncertainties .
the @xmath2 ( s - wave ) channel has small couplings for @xmath140 but , as we are over the threshold , the phase space is big enough to generate a moderate decay width .
the results of the branching ratios are in a fair agreement with the experiments within the large uncertainties . for this case
the pdg average has a very wide range , but the individual results are more precise .
there is a large disagreement with the result of @xcite but it is much closer to the one of @xcite by the same group .
the @xmath2 ( d - wave ) channel has the lower coupling to @xmath139 but , since it has a large momentum , the partial decay width is bigger than for the @xmath0 channel .
the branching ratio for @xmath139 agrees with experimental results , but in the case of @xmath140 , although the result is compatible with some experimental branching ratios , these results are very different and in some cases have large errors . finally , for the channel @xmath1 ( d - wave ) , since we fit the amplitude to the data , the result of the branching ratio of @xmath139 is in a very good agreement with the experimental values . on the other hand
, we get a branching ratio of the @xmath140 for the decay into @xmath141 ( d - wave ) which is in line with experimental determinations but about twice as large as the pdg average . yet ,
an inspection to table [ tab : widths2exp ] indicates that the dispersion of experimental data for the @xmath140 is quite large , for what further attention to this resonance should be most welcome .
it is interesting to compare the present results with those of the jlich group @xcite . in those works
the authors also consider @xmath0 , @xmath1 and @xmath141 channels , in s or d - waves when allowed for @xmath142 .
the dynamics used in those works in similar to the one used here up to details concerning a contact @xmath0 term , present in @xcite ( see figs . 2 and 3 ) , the kroll - ruderman term for @xmath143 no considered in @xcite and the use of form factors in those work while we use dimensional regularization . in all cases we adhere to the dynamics of local hidden gauge approach .
one novelty in our work is that the @xmath139 is dynamically generated in our approach , while in @xcite is an input resonance .
the @xmath139 was obtained as a dynamically generated state in @xcite using coupled channels and input from chiral lagrangians . in the present case we have taken only the main @xmath144 channel , but it is well known that effects of other channels can be incorporated by small changes in the subtraction constants of dimensional regularization , which in the present problem we have left as free parameters .
the other novelty is the @xmath140 that we also obtain as dynamically generated with our input .
it is interesting to remark that in the work of @xcite , with the input of non pole terms there was also a state around 1700 mev which was dynamically generated and a pole could be seen in the third riemann sheet defined there .
however when the pole terms were added , the pole in the amplitude fades away in that approach . in our work
the pole appears around 1665 mev together with the one for the @xmath139 .
the unitarization of the driving terms in all the channels that we consider , including the transition between all of them , generates in our case the two poles and their interference is responsible for the vanishing amplitude of @xmath1(d ) around 1700 mev , which is consistent with the experiment .
however , physical traces experimentally identifiable as a peak , remain in the @xmath141(d ) channel as one can see in fig.[fig : t2res ] , hence , the more intuitive picture of a peak in a cross section is also seen here but only in the @xmath1(d ) channel .
actually in @xcite , the most characteristic feature attributed to the @xmath140 was its large coupling to @xmath1(d ) .
we should also mention that there are other works that mix pseudoscalars and vectors @xcite and others that also include decuplet of baryons @xcite , this latter one invoking su(3 ) spin - isospin symmetry . in our approach
we have followed strictly the local hidden gauge formalism that has proved to be very successful in a variety of processes ( see recent review @xcite ) .
we have done a theoretical study for the meson - baryon scattering in the region of @xmath145 with @xmath146 .
we considered the standard coupled channels used in the most complete experimental analyses , @xmath0 ( s - wave ) , @xmath2 ( s - wave ) , @xmath2 ( d - wave ) and @xmath1 ( d - wave ) .
the interaction of these channels was taken from the local hidden gauge approach and the loops were regularized using dimensional regularization with subtraction constant of natural size .
these constants were varied within a moderate range to obtain a good fit to the @xmath1 ( d - wave ) data .
after this , the rest are predictions of the theory .
the first important theoretical finding is that the model obtained , after fitting exclusively the @xmath1 ( d - wave ) data , produces two poles : one around 1480 mev that we associate to the @xmath4 and another around 1670 mev that we associate to the @xmath5 .
it is worth noting that the presence of the two poles is rather solid , since they remain by making changes in the parameters that do not spoil massively the agreement with the @xmath1 ( d - wave ) experimental data .
with the model obtained we determined partial decay widths to all the channels .
we found an excellent agreement with experiments for the data on the @xmath4 and rough to fair for the @xmath5 .
yet , we noticed the large dispersions of experimental data for the @xmath5 .
the study done here gives a boost to the existence of the @xmath140 which has been questioned in some recent experimental analyses . in view of this extra support for the @xmath140 and the large dispersion of the data , further experimental studies concentrating in this energy region for the quantum numbers @xmath147 of this resonance
should be encouraged .
we would like to thank m. dring , d. m. manley , i. i. strakovsky , d. bugg and r. l. workman for a through reading of our paper and valuable comments .
this work is partly supported by the spanish ministerio de economia y competitividad and european feder funds under the contract number fis2011 - 28853-c02 - 01 , and the generalitat valenciana in the program prometeo , 2009/090 .
we acknowledge the support of the european community - research infrastructure integrating activity study of strongly interacting matter ( acronym hadronphysics3 , grant agreement n. 283286 ) under the seventh framework programme of eu .
this work is also partly supported by the national natural science foundation of china under grant 11105126 .
the diagonal transition for @xmath2 ( s - wave ) and @xmath0 ( s - wave ) are taken from the references @xcite and @xcite respectively .
so we need to evaluate the transition potential of fig .
[ fig : rhonpidel ] .
when constructing the kroll - ruderman term , we shall get the contact term of the type of @xmath148 .
we must evaluate this matrix element @xmath149 where @xmath150 as @xmath151 is part of @xmath152 , we get that in the spherical basis @xmath153 then the matrix element becomes @xmath154 where the reduced matrix element is chosen to be 1 by construction , and from the clebsch - gordan coefficient we get a @xmath155 .
finally we get that @xmath156 hence , for spin @xmath46 of @xmath0 this operator is unity 1 . getting the kroll - ruderman term
, we must substitute @xmath157 we need to write the states of isospin basis in charge basis , with our sign convention @xmath158 .
@xmath159 for the transition @xmath160 ( s - wave ) we have the vertices of fig .
[ fig : rhonpidelall ] . for evaluating those vertices
we need both lagrangians , one for the vpp vertex and the other one for the lower @xmath34 vertex .
the first diagram has a vertex @xmath161 the other coefficients of the diagrams for the different charge combinations are indicated in fig .
[ fig : rhonpidelall ] .
the lower vertex has an isospin coefficient for @xmath34 , giving a transition vertex for the first diagram @xmath162 the other isospin coefficients of the lower vertices are shown in fig .
[ fig : rhonpidelall ] . combining the two vertices and the pion propagator
, we get the transition potential of the first diagram . @xmath163 as mentioned before , we need to add the kroll - ruderman term as @xmath148 , @xmath164
so in the s - wave for a @xmath98-exchange we can sum the momenta products as @xmath165 and we have shown for @xmath166 that @xmath167 the sum of the pion propagator and the kroll - ruderman for the first diagram gives us @xmath168 we need also to evaluate the s - wave and d - wave mixing , @xmath0 is in @xmath101 but @xmath2 can be in @xmath101 and @xmath103 since there is enough momentum for @xmath139 and @xmath140 . with @xmath0 below threshold there is no need to worry about @xmath103 for @xmath0 .
we want to work with @xmath0(s ) , @xmath2(s ) and @xmath2(d ) , for this we shall also include the @xmath172 transition .
let us evaluate the transition @xmath173 thought an intermediate state @xmath2 as shown in fig [ fig : rhonboxs](a ) .
let us take the third component of spin @xmath45 and @xmath102 for @xmath89 and @xmath20 respectively , to get @xmath174 since the result does not depend on the third component . with this choice , both polarizations
@xmath151 and @xmath175 will be the same and we have neglected the momentum @xmath176 .
the terms of the four vertex are @xmath177 we can put together the polarization terms and the isospin transition operators and use that @xmath178 for the polarization terms we have in the loop integral @xmath179 as @xmath180 .
now , all together we have @xmath181 including both @xmath101 and @xmath103 contributions . for s - wave we got @xmath182 for one pion exchange ( see eq.([eq : swaveq2 ] ) ) so , the box equivalent operator with only s - wave will be @xmath183 subtracting this from the total transition ( eq . ( [ eq : totalds ] ) )
we get the d - wave transition in the box @xmath184 so finally we obtain that for the d - wave transition of one pion exchange the contribution is @xmath182 , the same one as for the s - wave transition .
all the other terms are the same but without the kroll - ruderman factor which only comes in @xmath101 .
thus , @xmath185 in the case of @xmath186 ( d - wave ) transition , the s - wave for @xmath46 does not exist , so whatever comes out will be d - wave .
we use the same procedure as for @xmath2 , we chose the third component of spin to be @xmath46 . the diagram is shown in fig [ fig : rhonboxs](b ) and the operators of the vertices are @xmath187 in the same way @xmath188 and for the polarization we have the same result as eq .
( [ eq : pola ] ) .
so we get @xmath189 the result for each vertex in the @xmath186 ( d - wave ) is @xmath190 . next we must do the isospin combination and the charge factors for each vertex as shown in fig .
[ fig : rhonpinall ] . @xmath194 looking at the diagrams , the isospin factor for the transition @xmath186 ( d - wave ) will be @xmath195 finally we have @xmath196 j. beringer _
[ particle data group collaboration ] , phys . rev .
d * 86 * , 010001 ( 2012 ) .
d. m. manley and e. m. saleski , phys .
d * 45 * , 4002 ( 1992 )
. m. batinic , s. ceci , a. svarc and b. zauner , phys .
c * 82 * , 038203 ( 2010 ) .
a. v. anisovich , r. beck , e. klempt , v. a. nikonov , a. v. sarantsev and u. thoma , eur .
j. a * 48 * , 15 ( 2012 ) [ arxiv:1112.4937 [ hep - ph ] ] .
r. a. arndt , w. j. briscoe , i. i. strakovsky and r. l. workman , phys .
c * 74 * , 045205 ( 2006 ) [ nucl - th/0605082 ] .
r. a. arndt , w. j. briscoe , i. i. strakovsky , r. l. workman and m. m. pavan , phys .
c * 69 * ( 2004 ) 035213 [ nucl - th/0311089 ] .
r. l. workman , m. w. paris , w. j. briscoe and i. i. strakovsky , phys .
c * 86 * , 015202 ( 2012 ) [ arxiv:1202.0845 [ hep - ph ] ] .
r. l. workman , r. a. arndt , w. j. briscoe , m. w. paris and i. i. strakovsky , phys .
c * 86 * , 035202 ( 2012 ) [ arxiv:1204.2277 [ hep - ph ] ] .
d. drechsel , o. hanstein , s. s. kamalov and l. tiator , nucl .
a * 645 * , 145 ( 1999 ) [ nucl - th/9807001 ] ; l. tiator , d. drechsel , s. kamalov , m. m. giannini , e. santopinto and a. vassallo , eur .
phys . j. a * 19 * , 55 ( 2004 ) [ nucl - th/0310041 ] ; d. drechsel , s. s. kamalov and l. tiator , eur .
j. a * 34 * , 69 ( 2007 ) [ arxiv:0710.0306 [ nucl - th ] ] .
a. v. anisovich , r. beck , e. klempt , v. a. nikonov , a. v. sarantsev and u. thoma , arxiv:1205.2255 [ nucl - th ] .
u. thoma , m. fuchs , a. v. anisovich , g. anton , r. bantes , o. bartholomy , r. beck and y. .beloglazov _ et al .
_ , phys .
b * 659 * , 87 ( 2008 ) [ arxiv:0707.3592 [ hep - ph ] ] .
r. a. arndt , w. j. briscoe , m. w. paris , i. i. strakovsky and r. l. workman , chin .
c * 33 * , 1063 ( 2009 ) [ arxiv:0906.3709 [ nucl - th ] ] .
m. bando , t. kugo , s. uehara , k. yamawaki and t. yanagida , phys .
lett . * 54 * , 1215 ( 1985 ) .
m. bando , t. kugo and k. yamawaki , phys .
* 164 * , 217 ( 1988 ) .
m. harada and k. yamawaki , phys .
* 381 * , 1 ( 2003 ) [ hep - ph/0302103 ] .
u. g. meissner , phys .
* 161 * , 213 ( 1988 ) .
e. oset and a. ramos , eur .
phys . j. a * 44 * , 445 ( 2010 ) [ arxiv:0905.0973 [ hep - ph ] ] .
e. oset , a. ramos and c. bennhold , phys .
b * 527 * ( 2002 ) 99 [ erratum - ibid .
b * 530 * ( 2002 ) 260 ] [ nucl - th/0109006 ] .
l. s. geng and e. oset , phys .
d * 79 * , 074009 ( 2009 ) [ arxiv:0812.1199 [ hep - ph ] ] .
l. roca , s. sarkar , v. k. magas and e. oset , phys .
c * 73 * , 045208 ( 2006 ) [ hep - ph/0603222 ] .
d. m. manley , r. a. arndt , y. goradia and v. l. teplitz , phys .
d * 30 * , 904 ( 1984 ) .
r. e. cutkosky , c. p. forsyth , r. e. hendrick and r. l. kelly , phys .
d * 20 * , 2839 ( 1979 ) .
m. shrestha and d. m. manley , phys .
c * 86 * , 055203 ( 2012 ) [ arxiv:1208.2710 [ hep - ph ] ] .
t. p. vrana , s. a. dytman and t. s. h. lee , phys .
rept . * 328 * , 181 ( 2000 ) [ nucl - th/9910012 ] .
m. doring , c. hanhart , f. huang , s. krewald and u. -g .
meissner , nucl .
a * 829 * , 170 ( 2009 ) [ arxiv:0903.4337 [ nucl - th ] ] .
d. ronchen , m. doring , f. huang , h. haberzettl , j. haidenbauer , c. hanhart , s. krewald and u. -g .
_ , arxiv:1211.6998 [ nucl - th ] .
k. p. khemchandani , h. kaneko , h. nagahiro and a. hosaka , phys .
d * 83 * , 114041 ( 2011 ) [ arxiv:1104.0307 [ hep - ph ] ] .
k. p. khemchandani , a. martinez torres , h. kaneko , h. nagahiro and a. hosaka , phys .
d * 84 * , 094018 ( 2011 ) [ arxiv:1107.0574 [ nucl - th ] ] .
k. p. khemchandani , a. martinez torres , h. nagahiro and a. hosaka , phys .
d * 85 * , 114020 ( 2012 ) [ arxiv:1203.6711 [ nucl - th ] ] .
d. gamermann , c. garcia - recio , j. nieves and l. l. salcedo , phys .
d * 84 * , 056017 ( 2011 ) [ arxiv:1104.2737 [ hep - ph ] ] .
e. oset , a. ramos , e. j. garzon , r. molina , l. tolos , c. w. xiao , j. j. wu and b. s. zou , int .
j. mod .
e * 21 * , 1230011 ( 2012 ) [ arxiv:1210.3738 [ nucl - th ] ]
. s. sarkar , b. -x .
sun , e. oset and m. j. vicente vacas , eur .
j. a * 44 * , 431 ( 2010 ) [ arxiv:0902.3150 [ hep - ph ] ] . | using an interaction extracted from the local hidden gauge lagrangians , which brings together vector and pseudoscalar mesons , and the coupled channels @xmath0 ( s - wave ) , @xmath1 ( d - wave ) , @xmath2 ( s - wave ) and @xmath2 ( d - wave ) , we look in the region of @xmath3 mev and we find two resonances dynamically generated by the interaction of these channels , which are naturally associated to the @xmath4 and @xmath5 .
the @xmath5 appears neatly as a pole in the complex plane .
the free parameters of the theory are chosen to fit the @xmath1 ( d - wave ) data .
both the real and imaginary parts of the @xmath1 amplitude vanish in our approach in the vicinity of this resonance , similarly to what happens in experimental determinations , what makes this signal very weak in this channel .
this feature could explain why this resonance does not show up in some experimental analyses , but the situation is analogous to that of the @xmath6 resonance , the second scalar meson after the @xmath7 in the @xmath8(d - wave ) amplitude .
the unitary coupled channel approach followed here , in connection with the experimental data , leads automatically to a pole in the 1700 mev region and makes this second @xmath9 resonance unavoidable . |
according to the shell model , the @xmath14 single - particle orbital is well separated from the @xmath15 , @xmath16 and other higher - energy orbitals , giving rise to the well - established @xmath17 magic numbers .
a less pronounced subshell closure for @xmath18 is also expected , as a smaller energy gap typically occurs between the @xmath14 level and the lower @xmath19-shell model states .
however , collectivity is evident for most nuclei with @xmath20 , and rotational - like yrast structures have been observed in the even - even , @xmath21 isotopes @xmath22sr and @xmath23zr @xcite .
the @xmath24 and @xmath25 shell gaps appear to be more robust for neutron - rich nuclei .
both @xmath26zr@xmath27 and @xmath28zr@xmath29 exhibit features at low excitation energy that support the presence of a subshell gap at @xmath30 @xcite .
this quasi - magic gap at @xmath30 disappears with the addition of neutrons beyond @xmath31 , and such a behavior was attributed to the strong , attractive monopole interaction between @xmath14 protons and @xmath15 neutrons @xcite .
evidence for magicity at @xmath32 for neutron - rich nuclei is gathered mainly from measurements on @xmath33ni@xmath34 and adjacent nuclei .
the energy of the first @xmath35 state [ @xmath36 in @xmath33ni is 2033 kev @xcite , much higher than the @xmath37 values for the even - even neighbors @xmath38ni@xmath39 and @xmath40ni@xmath41 .
the ratio @xmath42/@xmath37 for the even - even ni isotopes reaches a local minimum at @xmath33ni , another indication for added stability at @xmath43 .
the low transition probability for excitation to the first @xmath44 state in @xmath33ni @xcite , relative to neighboring isotopes , is also in line with a double - magic character for this nucleus and the `` goodness '' of the subshell closure at @xmath32 for @xmath45ni .
however , it seems that the subshell gap at @xmath32 may be quickly reduced with the removal or addition of protons . as an example , the systematic variation of @xmath37 energies as a function of neutron number for the @xmath45ni , @xmath10fe , and @xmath12cr isotopes is presented in fig .
[ fig : systematics ] .
the @xmath37 values for the @xmath10fe and @xmath12cr isotopes decrease with increasing neutron number , even through @xmath32 .
@xcite identified the first @xmath44 state in @xmath38fe@xmath34 from the @xmath0 decay of @xmath38mn .
the low value @xmath37 = 573 kev was taken as an indication of possible collectivity near the ground state , and a deformation parameter @xmath46 was deduced from the grodzins relation @xcite .
the decreasing trend in @xmath37 values for the neutron - rich fe isotopes continues in @xmath33fe@xmath41 , where the @xmath47 level with energy 517 kev was identified by adrich _
_ @xcite with in - beam @xmath4-ray spectroscopy following @xmath48 knockout . for the even - even @xmath45ni , @xmath10fe and @xmath12cr isotopes.,scaledwidth=50.0% ]
the systematic decrease in @xmath37 values is even more pronounced in the cr isotopic chain ( fig .
[ fig : systematics ] ) .
deformation parameters @xmath49 and @xmath50 have been deduced for the ground states of @xmath51cr@xmath5 and @xmath52cr@xmath39 , respectively @xcite , based on their @xmath37 values .
large quadrupole deformation lengths have also been reported for @xmath53cr based on results of proton inelastic scattering experiments @xcite . however , the medium - spin structure of @xmath51cr does not display features characteristic of a rotational pattern @xcite .
although data for the low - energy structure of @xmath54cr@xmath34 are still lacking , the trend of a lowering of excitation energies for the neutron - rich @xmath12cr isotopes at and beyond @xmath32 is predicted @xcite .
the strength of the subshell closure at @xmath25 depends on the magnitude of the gap between the @xmath55 and @xmath14 neutron orbitals . the strong monopole interaction between protons in the @xmath56 shell and the neutrons in the @xmath14 orbital decreases when
@xmath56 protons are removed , causing the narrowing of the @xmath57-@xmath14 energy separation @xcite .
in addition , the first two nilsson substates of the @xmath14 orbital are steeply downsloping with increasing deformation . combined , these effects can create a situation in which the occupation of the deformed levels is energetically preferential to the occupation of the spherical states , leading to increased deformation and erosion of the @xmath32 subshell closure for @xmath58 .
the appearance of low - energy @xmath59 states in the odd-@xmath7 cr isotopes gives preliminary evidence that the @xmath14 neutron orbital is indeed nearing the fermi surface .
the rotational bands built on the @xmath59 state in @xmath60cr@xmath61 @xcite and @xmath62cr@xmath63 @xcite suggest the presence of prolate - deformed structures .
however , the isomeric nature of the @xmath59 level in @xmath64cr@xmath65 @xcite appears to be more indicative of modest oblate collectivity .
while the neutron - rich , even-@xmath66 @xmath10fe and @xmath12cr nuclei near @xmath32 have been investigated , scant data are available regarding the level structures of the neighboring odd-@xmath66 @xmath8mn .
the driving force leading to the lowering of the @xmath37 energies and possible onset of collectivity in the cr and fe isotopes near @xmath32 should be expected to act in a similar manner in the mn nuclei . however , level structures are not available for mn isotopes with @xmath67 . in - beam @xmath4 rays
were recently measured for neutron - rich @xmath68mn produced by multi - nucleon transfer between a @xmath40zn projectile and @xmath69u target @xcite .
only a single @xmath4-ray transition was attributed to @xmath70mn@xmath39 .
five @xmath4 rays were assigned to the low - energy structure of @xmath52mn@xmath2 , complementing the low - energy level scheme that had previously been proposed for this nucleus from @xmath0 decay @xcite .
the low - energy structure of @xmath52mn is further complicated by the presence of two @xmath0-decaying states @xcite . in - beam and @xmath0-delayed @xmath4 rays have likewise been assigned to @xmath1mn@xmath5 . however , no level scheme was proposed based on the five @xmath0-delayed @xmath4 rays reported by sorlin _ et al .
the low - energy structure of @xmath1mn offers an important opportunity to characterize further the possible onset of collectivity inferred from the systematic behavior of the even - even cr and fe isotopes . the decrease in @xmath37 values for
neutron - rich cr may suggest that deformation sets in at @xmath51cr , with @xmath13 , while evidence for such deformation effects in the fe isotopes appears only at @xmath54fe , with @xmath71 . here
, we report on the low - energy structure of @xmath1mn@xmath5 , populated following the @xmath0 decay of @xmath1cr .
we also examine the systematic variation of the low - energy level densities of the odd-@xmath7 @xmath8mn isotopes and find no evidence for an early onset of collectivity at @xmath72 , as was proposed for neighboring @xmath12cr nuclei .
the @xmath0-decay properties of @xmath1cr were studied at national superconducting cyclotron laboratory ( nscl ) at michigan state university .
a 130-mev / nucleon @xmath22ge@xmath73 beam was produced by the coupled cyclotrons at nscl .
the @xmath22ge primary beam was fragmented on a 47 mg/@xmath74 be target at the object position of the a1900 fragment separator @xcite .
the secondary fragments of interest were separated in the a1900 with a 300 mg/@xmath74 al wedge located at the intermediate image of the separator .
the full momentum acceptance of the a1900 ( @xmath75% ) was used for fragment collection .
fragments were delivered to the @xmath0-decay experimental end station , which consisted of detectors from the beta counting system ( bcs ) @xcite and the segmented germanium array ( sega ) @xcite .
a stack of 3 si pin detectors with thicknesses 991 , 997 and 309 @xmath76 m , respectively , was placed upstream of the bcs and provided energy loss information for particle identification .
fragments were implanted in the 979 @xmath76m - thick double - sided silicon microstrip detector ( dssd ) of the bcs .
this detector was segmented into 40 strips on both front and back , for a total of 1600 pixels .
a total of @xmath77 @xmath1cr ions was implanted into the dssd over the course of the measurement . the particle identification spectrum is presented in fig . [ fig : cr61pid ] .
tm and @xmath78 tm .
energy loss was measured in the most upstream pin detector of the bcs .
time - of - flight was deduced from the difference in timing signals of the same pin detector and a plastic scintillator at the intermediate image of the a1900 .
an event - by - event momentum correction of the time - of - flight was performed based on the position information obtained from the plastic scintillator.,scaledwidth=50.0% ] delayed @xmath4 rays were detected by 16 ge detectors from sega , arranged in two concentric circles around the bcs detectors . the @xmath4-ray peak detection efficiency varied from 20% at 100 kev to 7% at 1 mev .
the energy resolution of each ge detector was measured to be better than 3.5 kev for the 1.3-mev @xmath4 ray in @xmath51co .
implanted @xmath1cr fragments were correlated with their subsequent @xmath0 decays by requiring the presence of a high - energy implantation event in a single pixel of the dssd , followed by a low - energy @xmath0 event in the same or any of the eight neighboring pixels .
the decay curve for @xmath1cr - correlated @xmath0 decays given in fig .
[ fig : halflife ] was generated by histogramming the differences between absolute time stamps for implantation and correlated decay events .
the curve was fitted with a single exponential decay combined with an exponential growth and decay of the short - lived daughter , @xmath1mn , whose half - life was taken to be @xmath79 ms as adopted in ref .
a constant background was also included as a free parameter in the fit .
a half - life of @xmath80 ms was deduced for the ground - state @xmath0 decay of @xmath1cr .
this new value compares favorably with a previous measurement by sorlin _
_ @xcite of @xmath81 ms , but is more than @xmath82 shorter than the @xmath83 ms value deduced earlier by ameil _
_ @xcite . cr , based on fragment-@xmath0 correlations .
data were fitted with a single exponential decay and exponential growth and decay of the short - lived daughter , @xmath1mn ( @xmath84 ms ) .
a constant background was also considered in the fit.,scaledwidth=50.0% ] the @xmath0-delayed @xmath4-ray spectrum for the decay of @xmath1cr is shown in fig .
[ fig : gamma_cronly ] .
the spectrum covers the energy range 0 to 2.5 mev , and the observed @xmath4 rays correspond to events that occurred within the first 3 s following a @xmath1cr implantation .
seven transitions have been assigned to the decay of @xmath1cr and are listed in table [ tab : gammas ] .
the peaks observed with energies of 207 and 629 kev in fig . [ fig : gamma_cronly ] are known transitions in the decay of the @xmath1mn daughter @xcite .
the 1028- and 1205-kev transitions in the decay of the grand - daughter @xmath1fe @xcite are also seen in fig .
[ fig : gamma_cronly ] .
the peaks at 355 , 535 , 1142 and 1861 kev correspond in energy to the transitions previously observed with low statistics by sorlin _
_ @xcite . however , the transition observed in ref .
@xcite at 1134 kev was not apparent in the spectrum of fig .
[ fig : gamma_cronly ] .
the three additional @xmath4-ray transitions at 157 , 1497 , and 2378 kev are assigned for the first time to the decay of @xmath1cr . coincidentally , two transitions with energies 155 and 355 kev had previously been assigned to the @xmath0 decay of @xmath52cr @xcite .
observation of these transitions in fig .
[ fig : gamma_cronly ] suggests that they are associated with levels in @xmath1mn , and their observation in ref .
@xcite may be evidence for @xmath0-delayed neutron decay of the @xmath52cr ground state .
it is worth noting that the in - beam @xmath4-ray spectra for @xmath1mn and @xmath52mn obtained by valiente - dobn _ et al . _
@xcite include transitions of energies 157 and 155 kev , respectively , supporting the present assignment of the 157-kev @xmath4-ray transition in fig .
[ fig : gamma_cronly ] to the decay of @xmath1cr
. -delayed @xmath4 rays following the decay of @xmath1cr .
known @xmath4 rays from the decay of the daughter @xmath1mn and the granddaughter @xmath1fe are indicated by the filled circles and squares , respectively .
transitions assigned to the decay of @xmath1cr are marked by their energy in kev.,scaledwidth=50.0% ] .energies and absolute intensities of delayed @xmath4 rays assigned to the decay of @xmath1cr .
the initial and final states for those transitions placed in the proposed @xmath1mn level scheme of fig .
[ fig : decayscheme ] are also indicated .
[ cols= " > , > , > , > , > " , ] the proposed decay scheme for levels in @xmath1mn populated following the @xmath0 decay of @xmath1cr is presented in fig .
[ fig : decayscheme ] .
the @xmath0-decay @xmath85 value was taken from ref .
absolute @xmath4-ray intensities were deduced from the number of observed @xmath1cr @xmath4 rays , the @xmath4-ray peak efficiency , and the number of @xmath1cr implants correlated with @xmath0 decays , as derived from the fit of the decay curve in fig .
[ fig : halflife ] . the two-@xmath4 cascade involving the 1142- and 355-kev transitions was confirmed by @xmath86 coincidence relationships ( see fig . [
fig : gamma_gamma ] ) .
however , the ordering of the two transitions is not uniquely determined .
the arrangement of fig .
[ fig : decayscheme ] was based on the absolute intensities of the 1142- and 355-kev transitions .
no evidence was found for @xmath4 rays in coincidence with the 157-kev ground - state transition within the statistical uncertainty of the measurement .
two counts in the 355-kev coincidence spectrum of fig .
[ fig : gamma_gamma](b ) suggested placement of the 535-kev line in cascade from a higher - lying level at 2032 kev .
this tentative placement is represented in fig .
[ fig : decayscheme ] by a dashed line .
mn level scheme populated following the @xmath0 decay of @xmath1cr .
the number in brackets following a transition energy is the absolute @xmath4-ray intensity .
the @xmath85 value was taken from ref .
observed coincidences are represented by filled circles .
absolute @xmath0-decay intensities and apparent log @xmath87 values to each state in @xmath1mn are given on the left - hand side of the figure .
the dashed state at 2032 kev is tentatively placed as described in the text.,scaledwidth=50.0% ] coincidence spectra for : ( a ) 157 kev ; ( b ) 355 kev ; and ( c ) 1142 kev @xmath4 rays .
the spikes observed below 200 kev in ( b ) and ( c ) represent 3 counts each in a single channel at low energy , and an absence of width uncharacteristic of a true @xmath4-ray coincidence.,scaledwidth=50.0% ] apparent @xmath0-decay feeding to levels in @xmath1mn was deduced from the absolute @xmath4-ray intensities .
the deduced values , along with apparent log @xmath87 values , are also given in fig .
[ fig : decayscheme ] .
the observed @xmath0 branches all have apparent log @xmath87 values between 4 and 6 , consistent with allowed transitions .
the 1142-kev level is deduced to have a small @xmath0 branching with a large error , and may not be directly populated by the @xmath0 decay of @xmath1cr .
a ground - state spin and parity assignment of @xmath88 was adopted for @xmath1mn , by runte _
@xcite , based on the systematic trends of ground state @xmath89 values for the less neutron - rich , odd-@xmath7 mn isotopes .
the first excited state at 157 kev has been tentatively assigned @xmath90 from the in - beam @xmath4-ray results @xcite . allowed @xmath0 feeding to the lowest two states
would then limit the @xmath89 quantum numbers of the @xmath1cr ground state to values of @xmath91 or @xmath92 .
@xcite tentatively assigned @xmath93 to the ground state of @xmath1cr on the basis of the observed @xmath0-decay properties of the progenitor , @xmath1v , and such an assignment is adopted in fig .
[ fig : decayscheme ] .
it is likely that the 1142-kev state has low spin , either @xmath94 or @xmath95 .
the level at 1142 kev is apparently weakly fed by @xmath0 decay , was not identified in the yrast structure of @xmath1mn @xcite , and does not depopulate to the @xmath92 level at 157 kev within the statistical certainty of this measurement .
the 1497-kev level is tentatively assigned a @xmath89 of @xmath96 or @xmath91 because of the apparent allowed @xmath0-decay branch and the favorable competition between the two depopulating @xmath4 rays that suggests both transitions have m1 multipolarity .
the proposed levels at excitation energies of 1861 and 2378 kev are apparently fed by allowed @xmath0 decay as well , and can take @xmath89 values in a range from @xmath96 to @xmath92 .
the possible increase in collectivity inferred at low energy for the neutron - rich @xmath12cr and @xmath10fe isotopes has been associated with the presence of the neutron @xmath14 single - particle orbital near the fermi surface as @xmath25 is approached . as noted in the introduction , the systematic variation in @xmath37 values ( fig .
[ fig : systematics ] ) may suggest that deformation sets in already at @xmath13 for the cr isotopes , while first evidence for such deformation effects in the fe isotopes occurs at @xmath54fe , which has @xmath71 .
therefore , it could be speculated that the low - energy structure of @xmath1mn , with @xmath13 , might already exhibit features at low energy suggestive of a change in collectivity when compared to other odd-@xmath7 mn isotopes nearer to stability .
the systematics of the known energy levels of the odd-@xmath7 mn isotopes with @xmath97 can be found in fig .
[ fig : oddamn ] . the levels shown below 2 mev for @xmath62mn
are taken from @xmath0 decay of @xmath62cr @xcite , in - beam @xmath4-ray spectroscopy @xcite , and the ( d,@xmath98he ) transfer reaction @xcite .
@xmath0 decay apparently populates seven excited states below 2 mev , as well as the @xmath62mn ground state .
the states directly fed by @xmath0 decay are all assumed to have negative parity , since the @xmath62cr parent has ground state @xmath99 quantum numbers .
the negative parity yrast states observed in @xmath62mn were populated by a heavy - ion induced fusion - evaporation reaction , and resulted in tentative spin assignments to levels up to @xmath100 .
the ( d,@xmath98he ) transfer reaction provided unambiguous @xmath101 assignments to the negative parity states below 2 mev at 84 , 851 , and 1837 kev , based on angular distribution data .
positive parity was deduced for the excited states with energies 1753 and 1965 kev .
based on the spectroscopic strengths for proton pickup @xcite , the first @xmath92 , @xmath102 , and @xmath103 levels in @xmath62mn were deduced to carry @xmath104% of the summed single - particle strength from shell model expectations .
valiente - dobn _ et al . _
@xcite studied neutron - rich mn isotopes by in - beam @xmath4-ray spectroscopy following deep inelastic collisions of a @xmath40zn beam on a @xmath69u target .
negative - parity yrast states were identified up to @xmath105 in @xmath64mn@xmath106 and @xmath107 in @xmath1mn@xmath5 . only a single transition , with energy 248 kev , was assigned to the structure of @xmath70mn@xmath39 .
the non - yrast levels in @xmath108mn shown in fig . [ fig : oddamn ] were identified following @xmath0 decay .
the @xmath0 decay of @xmath64cr to levels in @xmath64mn was most recently reported by liddick _
_ @xcite . the ground state spin and parity of the parent @xmath64cr ,
tentatively assigned as @xmath109 , restricts the range of states in @xmath64mn accessible by allowed @xmath0 decay to those with @xmath110 .
the ground state of @xmath64mn is known to have @xmath93 @xcite .
the non - yrast @xmath89 values for @xmath64mn were inferred from the @xmath0- and @xmath4-decay patterns .
@xmath8mn isotopes , along with the results of shell model calculations with the gxpf1a and @xmath111 interactions .
only levels below 2 mev are presented .
in addition , the shell model results shown are limited to the three lowest energy levels calculated for @xmath112 , and the yrast levels for @xmath113 and @xmath114 .
levels with assumed negative parity are shown as solid lines , and those with positive parity are presented as dotted lines .
spins are given as @xmath115 and , in nearly all cases , are tentative as discussed in the text.,scaledwidth=50.0% ] the negative parity yrast structures below 2 mev in odd-@xmath7 @xmath116mn exhibit little variation as a function of neutron number .
these levels are consistent with shell model calculations reported in ref . @xcite employing the gxpf1a , kb3 g and @xmath111 effective interactions .
the authors of ref .
@xcite do note , however , that the proper ordering of the @xmath117 and @xmath118 levels at @xmath119 mev is reproduced for all three isotopes only by the @xmath111 interaction . the experimental energy gap between the ground state , with @xmath93 , and the @xmath92 first excited state exhibits a regular increase from @xmath62mn ( 83 kev ) to @xmath120mn ( 272 kev ) @xcite .
both the gxpf1a and @xmath111 shell model interactions reproduce the observed trends well . the shell model calculations in ref .
@xcite were extended to include non - yrast states below 2 mev .
the calculations were performed with the code antoine @xcite and the gxpf1a @xcite and @xmath111 @xcite effective interactions .
the full @xmath121 model space was utilized for the calculations with the gxpf1a interaction .
the valence space for the calculations with the cross shell interaction @xmath111 was limited to 2 neutron excitations from the @xmath121 shell to the @xmath14 orbital . as compared to ref .
@xcite , the truncation of the basis was necessary to reach convergence for the higher density of states at lower spin for the most neutron - rich mn isotopes considered here .
however , in those cases where the calculations were completed with both 2 and 6 neutron excitations allowed from the @xmath121 shell , similar results were obtained .
the low density of levels below 1 mev is persistent in @xmath116mn , and follows the shell model results with the gxpf1a effective interaction well .
the only excited state in this energy range is the @xmath92 level , which is expected to carry a reduced @xmath56 single - particle component with the addition of neutrons @xcite .
the low - spin level density below 1 mev is shown to increase at @xmath1mn in the shell model results with the @xmath111 interaction . here
, the influence of the @xmath14 neutron orbital becomes apparent at @xmath13 .
however , this is not borne out in the observed level structure .
the regular behavior of the low - energy levels of the odd-@xmath7 @xmath8mn isotopes through @xmath13 follows the similar trend observed in the even - even @xmath10fe isotopes , where the possible onset of collectivity at low energy is not evident until @xmath71 is reached . although the present experiment did not report any new information on even-@xmath7 mn isotopes , for the sake of completeness we include them in the discussion .
only a few excited states in the odd - odd @xmath53mn are established , based on @xmath0-decay studies .
two @xmath0-decaying states are known in both isotopes , but the location and ordering of the proposed @xmath122 and @xmath123 states is not established in @xmath52mn@xmath2 @xcite .
good agreement was noted between shell model results with the gxpf1 interaction and both the low - energy structure and @xmath0-decay properties of @xmath51mn@xmath65 @xcite .
the low - energy structure of @xmath52mn built on the @xmath122 @xmath0-decaying state shows marked similarity to that in @xmath51mn , albeit the excited @xmath44 and @xmath122 states are both shifted @xmath124 kev lower in energy with the addition of two neutrons . although valiente - dobn _ et al .
_ reported a number of in - beam @xmath4 rays associated with the depopulation of yrast levels in both @xmath51mn and @xmath52mn , no level structures were proposed due to the lack of coincidence data in the literature @xcite .
a key feature yet to be identified in the neutron - rich , odd - odd mn isotopes is the location of the negative - parity levels that could signify the presence of the intruding @xmath14 neutron orbital .
again , the systematic variation of the ( few ) excited levels established in the odd - odd mn isotopes through @xmath125 does not suggest a sudden onset of collectivity , in line with the trend established for the yrast states of even - even @xmath10fe isotopes .
the @xmath0 decay of @xmath1cr has been studied to extract details on non - yrast excited states in the daughter nucleus @xmath1mn .
a more accurate half - life of @xmath80 ms was deduced for the ground - state of @xmath1cr , consistent with the previous measurement by sorlin _
_ @xcite . the low - energy level scheme for @xmath1mn ,
deduced following @xmath0 decay of @xmath1cr , features five new excited states above a 1-mev excitation energy .
however , the low - energy structure below 1 mev does not resemble that expected from the shell model results that consider the 1@xmath126 orbital .
the low density of states below 1 mev is a consistent feature observed in the odd-@xmath7 mn isotopes with @xmath127 , and follows the results of shell model calculations in the @xmath121 model space , without a need to invoke neutron excitations into the @xmath14 orbital .
there is also little variation in the yrast structures of these odd-@xmath7 mn isotopes up to @xmath128 , as reported by valiente - dobn _ et al .
no compelling evidence was found for a possible onset of collectivity in the low - energy structures of the mn isotopes through @xmath125 .
this observation is consistent with the behavior of the even - even , @xmath9 @xmath10fe isotopes , but not with that of the even - even @xmath11 @xmath12cr isotopes , where possible onset of collectivity at low energy has been suggested from the e(@xmath47 ) energy at @xmath13 .
scant data are available for both yrast and non - yrast levels in the @xmath8mn isotopes beyond @xmath13
. such data will be critical to evaluate the role of the neutron @xmath14 orbital in defining the structural properties of the neutron - rich mn isotopes .
the authors thank the nscl operations staff for providing the primary and secondary beams for this experiment and the nscl @xmath4 group for assistance in setting up the ge detectors from sega .
this work was supported in part by the national science foundation grant no .
phy-06 - 06007 , the u.s .
department of energy , office of nuclear physics , under contracts de - ac02 - 06ch11357 and de - fg02 - 94er40834 , and the polish academy of sciences grant 1po3b 059 29 .
hlc acknowledges support from the natural science and engineering research council ( nserc ) of canada .
lister , m. campbell , a.a .
chishti , w. gelletly , l. goettig , r. moscrop , b.j .
varley , a.n .
james , t. morrison , h.g .
price , j. simpson , k. connel , and o skeppstedt , phys .
lett . * 59 * , 1270 ( 1987 ) .
fischer , c.j .
lister , d.p .
balamuth , r. bauer , j.a .
becker , l.a .
bernstein , m.p .
carpenter , j. durell , n. fotiades , s.j .
freeman , p.e .
garrett , p.a .
hausladen , r.v.f .
janssens , d. jenkins , m. leddy , j. ressler , j. schwartz , d. svelnys , d.g .
sarantites , d. seweryniak , b.j .
varley , and r. wyss , phys .
lett . * 87 * , 132501 ( 2001 ) .
r. broda , b. fornal , w. krlas , t. pawlat , d. bazzacco , s. lunardi , c. rossi - alvarez , r. menegazzo , g. de angelis , p. bednarczyk , j. rico , d. de acua , p.j .
daly , r.h .
mayer , m. sferrazza , h. grawe , k.h .
maier , and r. schubart , phys .
lett . * 74 * , 868 ( 1995 ) .
o. sorlin , s. leenhardt , c. donzaud , j. duprat , f. azaiez , f. nowacki , h. grawe , zs .
dombrdi , f. amorini , a. astier , d. baiborodin , m. belleguic , c. borcea , c. bourgeois , d.m .
cullen , z. dlouhy , e. dragulescu , m. grska , s. grvy , d. guillemaud - mueller , g. hagemann , b. herskind , j. kiener , r. lemmon , m. lewitowicz , s.m .
lukyanov , p. mayet , f. de oliveira santos , d. pantalica , yu .- e .
penionzhkevich , f. pougheon , a. poves , n. redon , m.g .
saint - laurent , j.a .
scarpaci , g. sletten , m. stanoiu , o. tarasov , and ch .
theisen , phys .
lett . * 88 * , 092501 ( 2002 ) .
n. bree , i. stefanescu , p.a .
butler , j. cederkll , t. davinson , p. delahaye , j. eberth , d. fedorov , v.n .
fedosseev , l.m .
fraile , s. franchoo , g. georgiev , k. gladnishki , m. huyse , o. ivanov , j. iwanicki , j. jolie , u. kster , th .
krll , r. krcken , b.a .
marsh , o. niedermaier , p. reiter , h. scheit , d. schwalm , t. sieber , j. van de walle , p. van duppen , n. warr , d. weisshaar , f. wenander , and s. zemlyanoy , phys .
c * 78 * , 047301 ( 2008 ) .
m. hannawald , t. kautzsch , a. whr , w.b .
walters , k .-
kratz , v.n .
fedoseyev , v.i .
mishin , w. bhmer , b. pfeiffer , v. sebastian , y. jading , u. kster , j. lettry , h.l .
ravn and the isolde collaboration , phys .
82 * , 1391 ( 1999 ) .
p. adrich , a.m. amthor , d. bazin , m.d .
bowen , b.a .
brown , c.m .
campbell , j.m .
cook , a. gade , d. galaviz , t. glasmacher , s. mcdaniel , d. miller , a. obertelli , y. shimbara , k.p .
siwek , j.a .
tostevin , and d. weisshaar , phys .
c * 77 * , 054306 ( 2008 ) .
o. sorlin , c. donzaud , f. nowacki , j.c .
anglique , f. azaiez , c. bourgeois , v. chiste , z. dlouhy , s. grvy , d. guillemaud - mueller , f. ibrahim , k .-
kratz , m. lewitowicz , s.m .
lukyanov , j. mrasek , yu .- e .
penionzhkevich , f. de oliveira santos , b. pfeiffer , f. pougheon , a. poves , m.g .
saint - laurent , and m. stanoiu , eur .
j. a * 16 * , 55 ( 2003 ) .
n. aoi , h. suzuki , e. takeshita , s. takeuchi , s. ota , h. baba , s. bishop , t. fukui , y. hashimoto , h.j .
ong , e. ideguchi , k. ieki , n. imai , h. iwaski , s. kanno , y. kondo , t. kubo , k. kurita , k. kusaka , t. minemura , t. motobayashi , t. nakabayashi , t. nakamura , t. nakao , m. niikura , t. okumura , t.k .
ohniski , h. sakurai , s. shimoura , r. sugo , d. suzuki , m.k .
suzuki , m. tamaki , k. tanaka , y. togano , and k. yamada , nucl .
phys . * a805 * , 400c ( 2008 ) .
n. aoi , e. takeshita , h. suzuki , s. takeuchi , s. ota , h. baba , s. bishop , t. fukui , y. hashimoto , h.j .
ong , e. ideguchi , k. ieki , n. imai , m. ishihara , h. iwasaki , s. kanno , y. kondo , t. kubo , k. kurita , k. kusaka , t. minemura , t. motobayashi , t. nakabayashi , t. nakamura , t. nakao , m. niikura , t. okumura , t.k .
ohnishi , h. sakurai , s. shimoura , r. sugo , d. suzuki , m.k .
suzuki , m. tamaki , k. tanaka , y. togano , and k. yamada , phys .
lett . * 102 * , 012502 ( 2009 ) .
s. zhu , a.n .
deacon , s.j .
freeman , r.v.f .
janssens , b. fornal , m. honma , f.r .
xu , r. broda , i.r .
calderin , m.p .
carpenter , p. chowdhury , f.g .
kondev , w. krlas , t. lauritsen , s.n .
liddick , c.j .
lister , p.f .
mantica , t. pawlat , d. seweryniak , j.f .
smith , s.l .
tabor , b.e .
tomlin , b.j .
varley and j. wrzesiski , phys .
c * 74 * , 064315 ( 2006 ) .
appelbe , c.j .
barton , m.h .
muikku , j. simpson , d.d .
warner , c.w .
beausang , m.a .
caprio , j.r .
cooper , j.r .
novak , n.v .
zamfir , r.a.e .
austin , j.a .
cameron , c. malcolmson , j.c .
waddington , and f.r .
xu , phys .
c * 67 * , 034309 ( 2003 ) .
a.n . deacon , s.j .
freeman , r.v.f .
janssens , f.r .
xu , m.p . carpenter , i.r .
calderin , p. chowdhury , n.j .
hammond , t. lauritsen , c.j .
lister , d. seweryniak , j.f .
smith , s.l .
tabor , b.j .
varley , and s. zhu , phys .
b * 622 * , 151 ( 2005 ) .
freeman , r.v.f .
janssens , b.a .
brown , m.p .
carpenter , s.m .
fischer , n.j .
hammond , m. honma , t. lauritsen , c.j .
lister , t.l .
khoo , g. mukherjee , d. seweryniak , j.f .
smith , b.j .
varley , m. whitehead , and s. zhu , phys .
c * 69 * , 064301 ( 2004 ) .
valiente - dobn , s.m .
lenzi , s.j .
freeman , s. lunardi , j.f .
smith , a. gottardo , f. della vedova , e. farnea , a. gadea , d.r .
napoli , m. axiotis , s. aydin , d. bazzacco , p.g .
bizzeti , a.m. bizzeti - sona , g. benzoni , d. bucurescu , l. corradi , a.n . deacon ,
g. de angelis , e. fioretto , b. guiot , m. ionescu - bujor , a. iordachescu , s. leoni , n. mrginean , r. mrginean , p. mason , r. menegazzo , d. mengoni , b. million , g. montagnoli , r. orlandi , f. recchia , e. sahin , f. scarlassara , r.p .
singh , a.m. stefanini , d. steppenbeck , s. szilner , c.a .
ur , b.j .
varley , and o. wieland , phys .
c * 78 * , 024302 ( 2008 ) .
l. gaudefroy , o. sorlin , c. donzaud , j.c .
anglique , f. azaiez , c. bourgeois , v. chiste , z. dlouhy , s. grvy , d. guillemaud - mueller , f. ibrahim , k .-
kratz , m. lewitowicz , s.m .
lukyanov , i. matea , j. mrasek , f. nowacki , f. de oliveira - santos , yu .- e .
penionzhkevich , b. pfeiffer , f. pougheon , m.g .
saint - laurent , and m. stanoiu , eur .
phys . j. a * 23 * , 41 ( 2005 ) .
o. sorlin , c. donzaud , l. axelsson , m. belleguic , r. braud , c. borcea , g. canchel , e. chabanat , j.m .
daugas , a. emsallem , m. girod , d. guillemaud - mueller , k .-
kratz , s. leenhardt , m. lewitowicz , c. longour , m.j .
lopez , f. de oliveira santos , l. petizon , b. pfeiffer , f. pougheon , m.g .
saint - laurent , and j.e .
sauvestre , nucl .
a669 * , 351 ( 2000 ) .
f. ameil , m. bernas , p. armbruster , s. czajkowski , ph .
dessagne , h. geissel , e. hanelt , c. kozhuharov , c. miehe , c. donzaud , a. grewe , a. heinz , z. janas , m. de jong , w. schwab , and s. steinhuser , eur .
phys . j. a * 1 * , 275 ( 1998 ) .
e. runte , k .-
gippert , w .- d .
schmidt - ott , p. tidemand - petersson , l. ziegeler , r. kirchner , o. klepper , p.o .
larsson , e. roeckl , d. schardt , n. kaffrell , p. peuser , m. bernas , ph .
dessagne , m. langevin , and k. rykaczewski , nucl .
* a441 * , 237 ( 1985 ) .
puttaswamy , w. oelert , a. djaloeis , c. mayer - bricke , p. turek , p.w.m .
glaudemans , b.c .
metsch , k. heyde , m. waroquier , p. van isaker , g. wenes , v. lopac , and v. paar , nucl .
* a 401 * , 269 ( 1983 ) .
liddick , p.f .
mantica , r. broda , b.a .
brown , m.p .
carpenter , a.d .
davies , b. fornal , m. horoi , r.v.f .
janssens , a.c .
morton , w.f .
mueller , j. pavan , h. schatz , a. stolz , s.l .
tabor , b.e .
tomlin , m. wiedeking , phys .
c * 72 * , 054321 ( 2005 ) .
liddick , p.f .
mantica , b.a .
brown , m.p .
carpenter , a.d .
davies , m. horoi , r.v.f .
janssens , a.c .
morton , w.f .
mueller , j. pavan , h. schatz , a. stolz , s.l .
tabor , b.e .
tomlin , and m. wiedeking , phys .
c * 73 * , 044322 ( 2006 ) . | @xmath0 decay of the @xmath1cr@xmath2 ground state has been studied .
a new half - life of @xmath3 ms has been deduced , and seven delayed @xmath4 rays have been assigned to the daughter , @xmath1mn@xmath5 .
the low - energy level structure of @xmath1mn@xmath5 is similar to that of the less neutron - rich @xmath6mn nuclei .
the odd-@xmath7 @xmath8mn isotopes follow the systematic trend in the yrast states of the even - even , @xmath9 @xmath10fe isotopes , and not that of the @xmath11 @xmath12cr isotopes , where a possible onset of collectivity has been suggested to occur already at @xmath13 . |
in this material we present some details of calculation , and some additional features of the behavior of the edge majorana fermion in the considered systems , as a function of the governing parameter @xmath10 .
consider of the hamiltonian @xmath9 , where @xmath117 @xmath4 are operators of the projections of spin 1/2 situated at the @xmath5-th site , @xmath6 are coupling constants , and @xmath7 are coupling constants for the impurity , situated at the site @xmath8 . for simplicity of the consideration let us add the spin @xmath12 at the left edge of the chain with the coupling @xmath13 , so that we study the hamiltonian @xmath118 instead of @xmath15 .
the average value @xmath119 can be written as @xmath120 , where the density matrix is determined as usually @xmath121 , with @xmath122 , where @xmath84 is the temperature .
then it is easy to check that @xmath123 \nonumber \\ & & \equiv 2{\rm tr}_{n+2}(\rho_m s_0^xs_{-1}^x ) \ , \end{aligned}\ ] ] where we used the subscripts @xmath124 and @xmath125 to emphasize that the traces are taken with respect to eigenstates of the system consisting of @xmath124 or @xmath125 spins , respectively , and @xmath126 , where @xmath127 . the last equality uses the fact that the average of the operator , linear in @xmath128 , with the hamiltonian , quadratic in operators of @xmath129 and @xmath130 , is zero .
to find eigenfunctions and eigenvalues of the system with the hamiltonian @xmath19 , we use the jordan - wigner transformation to fermion operators , @xmath131 with @xmath20 ( @xmath21 ) being standard dirac creation ( destruction ) fermionic operator @xmath132 , @xmath133 , where the anticommutator is determined as @xmath134 . in that representation
we have @xmath135 \nonumber \\ & & + i'[d_1^{\dagger}a_0 + d_0^{\dagger}a_1 ] + j ' [ d_1^{\dagger}d_0^{\dagger } + d_0d_1 ] \nonumber \\ & & + \sum_{n=1}^{n-1}[i(d_n^{\dagger}d_{n+1 } + d_{n+1}^{\dagger}d_n ) \nonumber \\ & & + j(d_n^{\dagger}d_{n+1}^{\dagger } + d_{n+1}d_n ) ] \biggr)\ .
\label{hmdirferms}\end{aligned}\ ] ] then we use the unitary transformation @xmath136 where @xmath55 s are quantum numbers , which parameterize all eigenstates of the diagonalized hamiltonian .
these quantum numbers can describe extended ( band ) states .
besides , there is a possibility of localized states , caused by @xmath56 , @xmath57 , and @xmath58 .
let us define @xmath59 , i.e. , transfer to majorana modes @xmath60 .
then we can write the stationary schrdinger equation with @xmath19 in the co - ordinate space . from that equation for @xmath137
we have @xmath138 where @xmath139 are the energies ( we drop subscripts @xmath55 for simplicity ) .
on the other hand , for the sites @xmath140 we have two sets of equations , different from eqs . ( [ eqpqn ] ) .
the one , which depends on @xmath10 , is @xmath141 and the one , which does not depend on @xmath10 , is @xmath142 we have two disconnected systems of equations in finite differences .
it has two sets of eigenfunctions .
the first set of solutions describes nonzero @xmath61 for even @xmath5 , and nonzero @xmath62 for odd @xmath5 ( all other @xmath63 s and @xmath64 s are zeros ) .
the second set of solutions describes nonzero @xmath61 for odd @xmath5 , and nonzero @xmath62 for even @xmath5 ( others are zeros ) . for each
set we look for extended ( band ) states with @xmath143 , where @xmath144 is the quasimomentum of the eigenstate , and for a localized state , which wave function decays exponentially with the distance from the edge of the chain .
the solution of eqs .
( [ eqs1 ] ) and ( [ eqs2 ] ) is as follows .
for the extended states ( with quasimomenta @xmath145 in the limit @xmath25 ) we get ( @xmath146 ) @xmath147[(i^2-j^2)\sin(2 km ) \nonumber \\ & & + ( i - j)^2\sin[2k(m-1 ) ] ] \nonumber \\ & & -\varepsilon_k^2 ( i'-j')^2\sin[2k(m-1 ) ] ) \ ,
\nonumber \\ & & p_{2m , k}^{(1 ) } = \frac{2}{\sqrt{\pi}x_k^{(1)}}[2(\varepsilon_k^2 -h^2)(i - j ) \sin ( 2 km ) \nonumber \\ & & -(i'-j')^2(i\cos k\sin [ k(2m-1 ) ] \nonumber \\ & & -j\sin k \cos[k(2m-1 ) ] ) ] \ .
\label{pqk1}\end{aligned}\ ] ] for @xmath148 we obtain @xmath149 and @xmath150 here we use the following notation @xmath151 the energies of extended states are @xmath152 . for the localized state
( with @xmath70 ) we get ( @xmath146 ) @xmath153 for @xmath148 the localized eigenstates are @xmath154 , @xmath155 .
the parameter @xmath156 ( @xmath157 is the localization radius ) is @xmath68 } \nonumber \\ & & \times \biggl(4h^2 + ( i'-j')^2 -2(i^2 + j^2 ) \nonumber \\ & & -[(2h - i - j)^2+(i'-j')^2-(i - j)^2]^{1/2 } \nonumber \\ & & \times[(2h+i+j)^2+(i'-j')^2-(i - j)^2]^{1/2}\biggr ) \ . \label{r1s}\end{aligned}\ ] ] this state exists if @xmath69 [ ( 2h+i+j)^2+(i'-j')^2-(i - j)^2]>0 $ ] .
notice that @xmath70 , i.e. , localized state decays with the distance from the edge of the chain . even for the homogeneous case @xmath71 , @xmath72 for @xmath73
such a localized mode exists at @xmath56 . for the second set of extended states we have ( @xmath146 ) @xmath158 \nonumber \\ & & + j\sin k \cos[k(2m-1 ) ] ) ] \
, \nonumber \\ & & p_{2m-1,k}^{(2 ) } = - \frac{2\varepsilon_k}{\sqrt{\pi}x_k^{(2 ) } } [ ( i^2-j^2)\sin(2 km ) \nonumber \\ & & + [ ( i+j)^2-(i'+j')^2]\sin[2k(m-1 ) ] ] \ .
\label{pqk2odd}\end{aligned}\ ] ] for @xmath148 we obtain @xmath159 and @xmath160 here we use @xmath161 for the second set of localized states ( @xmath162 ) we obtain ( @xmath163 ) @xmath164 while for @xmath148 the solution has the form @xmath165 , @xmath166 .
we use @xmath167^{-1/2 } , \ \nonumber \\ & & r_{(2)}^2= \frac{(i^2-j^2)}{[(i'+j')^2-(i - j)^2 ] } \ . \label{r2}\end{aligned}\ ] ] for @xmath71 and @xmath72 there is no second localized state .
the energies of the localized states are @xmath66 ^ 2 - j^2[r_{(1,2 ) } - r^{-1}_{(1,2)}]^2 \label{eps12s}\ ] ] as the function of the strength of the applied local voltage ( magnetic field , tunneling ) @xmath10 for @xmath109 , @xmath96 and @xmath168 at @xmath169 .
the dashed ( blue ) line shows the contribution from extended states , the dotted ( red ) line describes the contribution from the localized mode , and the solid ( black ) line is the total value . ] here we also present several figures which describe the behavior of the average value for the edge majorana fermion operator @xmath170 and its local susceptibility @xmath108 for the considered quantum chains . as the function of the strength of the applied local voltage ( magnetic field , tunneling ) @xmath10 for the homogeneous chain @xmath173 , @xmath96 at @xmath169 ( top ) and @xmath174 ( bottom ) .
the dashed ( blue ) line shows the contribution from extended states , the dotted ( red ) line describes the contribution from the localized mode , and the solid ( black ) line is the total value.,title="fig : " ] as the function of the strength of the applied local voltage ( magnetic field , tunneling ) @xmath10 for the homogeneous chain @xmath173 , @xmath96 at @xmath169 ( top ) and @xmath174 ( bottom ) .
the dashed ( blue ) line shows the contribution from extended states , the dotted ( red ) line describes the contribution from the localized mode , and the solid ( black ) line is the total value.,title="fig : " ] the reader can see that no principal difference between the homogeneous and non - homogeneous cases exists .
the upper panel of fig .
[ sxy1 t ] manifests the behavior of @xmath102 for the homogeneous case at low temperatures , @xmath175 .
the lower panel of fig .
[ sxy1 t ] manifests the behavior of @xmath102 for the same homogeneous case but at high temperatures , @xmath176 . of the homogeneous chain like in fig .
[ sxy1 t ] at low temperatures @xmath175 ( top ) and at high temperatures @xmath176 ( bottom).,title="fig : " ] of the homogeneous chain like in fig .
[ sxy1 t ] at low temperatures @xmath175 ( top ) and at high temperatures @xmath176 ( bottom).,title="fig : " ] as expected , lower values of the temperature yield sharper behavior of the susceptibility at small values of @xmath10 . finally , fig .
[ chixy00112 t ] shows the temperature behavior of the local magnetic susceptibility for the chain with @xmath109 , @xmath177 , @xmath168 and @xmath178 for several values of the applied local field ( voltage , tunneling ) . as the function of the temperature for several values of the strength of the applied local voltage ( magnetic field , tunneling ) @xmath10 for the chain with @xmath109 , @xmath177 , @xmath168 and @xmath178 .
the solid ( red ) line describes the case @xmath100 ; the dashed ( black ) line shows @xmath110 case ; the dotted ( blue ) line describes the case @xmath179 , and the dashed - dotted ( green ) line shows @xmath180 case . ]
one can see that @xmath181 and/or @xmath182 ( which is related to the biaxial magnetic anisotropy in the spin system , or nonzero pairing in the toy kitaev model ) removes the @xmath183 divergency of the local susceptibility of the edge majorana mode .
also , for the cases with nonzero @xmath184 and/or @xmath185 the temperature behavior of @xmath186 is almost monotonic : @xmath187 is finite at low temperatures and decays with the growth of @xmath84 .
f. wilczek , nature physics * 5 * , 614 ( 2009 ) .
ivanov , phys .
lett . * 86 * , 268 ( 2001 ) ; s. das sarma , m. freedman , and c. nayak , _ ibid . _ * 94 * , 166802 ( 2005 ) a.yu .
kitaev , ann.phys.(ny ) * 303 * , 2 ( 2003 ) .
kitaev , phys .
usp . * 44 * , 131 ( 2001 ) .
n. read and d. green , phys .
b * 61 * , 10267 ( 2000 ) ; n.b .
kopnin and m.m .
salomaa , phys .
b * 44 * , 9667 ( 1991 ) .
j. alicea , rep .
75 * , 076501 ( 2012 ) ; c.w.j .
beenakker , annu .
rev . condens .
matter phys .
* 4 * , 113 ( 2013 ) ; t.d .
stanescu and s. tewari , arxiv:1302.5433 ( 2013 ) .
l. fu and c.l .
kane , phys .
lett . * 100 * , 096407 ( 2008 ) ; phys . rev .
b * 79 * , 161408(r ) ( 2009 ) . r.m .
lutchyn , j.d .
sau , and s. das sarma , phys .
. lett . * 105 * , 077001 ( 2010 ) ; y. oreg , g. refael , and f. von oppen , phys . rev . lett . * 105 * , 177002 ( 2010 ) . v. mourik _ et al .
_ , science * 336 * , 6084 ( 2012 ) ; a. das _
et al . _ , nature phys . * 8 * , 887 ( 2012 ) ; m.t .
deng _ et al .
_ , nano lett .
* 12 * , 6414 ( 2012 ) ; l.p .
rokhinson , x. liu , and j.f .
furdyna , nature phys .
* 8 * , 795 ( 2012 ) ; j.g .
rodrigo _ et al .
_ , arxiv:1302.0598 ( 2013 ) .
j. liu _ et al .
lett . * 109 * , 267002 ( 2012 ) ; e.j.h .
lee _ et al .
109 * , 186802 ( 2012 ) ; d.i .
pikulin _ et al .
_ , new j. phys . * 14 * , 125011 ( 2012 ) .
kleiner and v.m .
tsukernik , fiz .
* 6 * , 332 ( 1980 ) ( in russian )
[ sov . j. low temp .
* 6 * , 158 ( 1980 ) ] ; v.z .
kleiner , ph.d .
thesis iltpe , kharkov , ( unpublished ) ( 1982 ) .
see , e.g. , a.a .
zvyagin _ finite - size effects in correlated electron systems : exact results _ , imperial college press , london , 2005 , and references therein .
lieb , t.d .
schultz , and d.c .
mattis , ann .
* 16 * , 407 ( 1961 ) .
bardyn and a. imamoglu , phys .
* 109 * , 253606 ( 2012 ) ; i. carusotto _ et al . _ ,
lett . * 103 * 033601 ( 2009 ) . b. van heck , f. hassler , a.r .
akhmerov , and c.w.j .
beenakker , phys .
b * 84 * , 180502(r ) 2011 ; b. van heck _ et al . _ , new j. phys .
* 14 * , 035019 ( 2012 ) ; f. hassler and d. schuricht , new j. phys .
* 14 * , 125018 ( 2012 ) .
see , e.g. , a.a .
zvyagin , phys .
b * 85 * , 134435 ( 2012 ) and references therein . see , e.g , v. kataev _ et al . _ ,
86 * , 2882 ( 2001 ) ; k.m .
kojima _ et al .
_ , phys . rev .
b * 70 * , 094402 ( 2004 ) ; r. klingeler _ et al . _ ,
b * 72 * , 184406 ( 2005 ) . | several scenarios for realization of edge majorana modes in quantum chain systems : spin chains , chains of josephson junctions , and chains of coupled cavities in quantum optics , are considered .
for all these systems excitations can be presented as superpositions of a spinless fermion and a hole , characteristic for majorana fermion .
we discuss the features of our exact solution with respect to possible experiments , in which edge majorana fermions can be directly observed when studying magnetic , superconducting , and optical characteristics of such systems .
majorana fermions ( mf ) are particles , identical to own antiparticles
. they may appear as elementary neutral particles , or emerge as quasiparticles in many - body systems @xcite . during last years mf , besides being of fundamental interest of their own , have attracted great attention as the basis for potential application in topological quantum computation @xcite .
the search for mf is among the most prominent tasks for modern physicists . during last few years
a great progress has been achieved in such a search in condensed matter physics .
obviously , we can not expect mf to exist in ordinary metals , because excitations , electrons , considered as quasiparticles there , and their counterparts , holes ( which linear combination would correspond to the mf ) , can destruct each other : they carry opposite charges .
hence , the search in different , non - standard systems of fermions with special properties , where mf can exist as emergent non - trivial excitations , is necessary .
superconducting systems seemingly provide a basis for such states , because elementary excitations there are superpositions of electrons and holes .
however , for conventional superconductors with , e.g. , @xmath0-wave pairing , those superpositions of electrons and holes carrying _ opposite _ spin are different from majorana s construction .
then it follows that for a system of spinless fermions with pairing , like e.g. , model superconductors with @xmath1-pairing in one dimensional ( 1d ) systems @xcite , or with @xmath2-pairing in 2d ones @xcite , mf can emerge . among the most known predicted candidates for mf existence
are topological insulators @xcite , and semiconducting quantum wires @xcite , where pairing can be achieved by interfacing them with an ordinary superconductor .
the modern `` state of art '' of theoretical predictions for realizations of such systems has been recently reviewed , e.g. , in @xcite . while recent papers @xcite claim that they have observed zero bias anomalies in the tunneling conductance of normal conducting and superconducting systems , which can be explained by the presence of zero energy mf , very recent publications mention that in those experiments the spatial resolution could be not enough to detect mf , and that disorder can result in zero bias features @xcite even for non - topological system ( where mf are absent ) .
that is why , proposals for realization of direct observations of mf are highly desirable . in the present work we consider several scenarios for the _ direct _ observation of edge mf in quantum chains , which can be realized in quantum magnetic , superconducting , and optical systems . for all these systems excitations can be presented as superpositions of spinless fermions and holes , the hallmark of mf .
we choose 1d systems , because exact theoretical results can be obtained there , which is very important for comparison with experiment , and due to the significant success in fabrication and manipulation of quasi-1d materials in recent years .
we propose to use an external parameter , which _ directly governs _ the behavior of the edge mf in those quantum chains . to set the stage ,
we start with the consideration of the spin-1/2 chain , which hamiltonian is @xmath3 here @xmath4 are operators of the projections of spin 1/2 at the @xmath5-th site , @xmath6 ( @xmath7 ) are coupling constants for the host ( impurity situated at the site @xmath8 ) .
to realize the manifestation of edge mf in observable characteristics , we propose to study the system with the hamiltonian @xmath9 .
the local field @xmath10 , acting at the edge site of the chain , can be realized if the spin chain system neighbors a ferromagnet , which is magnetized along the @xmath11 axis .
let us ( formally ) add the spin @xmath12 at the left edge of the chain with the coupling @xmath13 , to study the hamiltonian @xmath14 instead of @xmath15 @xcite .
we see that @xmath16 \equiv 2{\rm tr}_{n+2}(\rho_m s_0^xs_{-1}^x)$ ] , where @xmath17 ( @xmath18 ) is the density matrix with the hamiltonian @xmath15 ( @xmath19 ) .
it means that to obtain the average value of the operator of edge spin projection with @xmath15 , we can calculate the one for the pair correlation function with @xmath19 .
after the jordan - wigner transformation with dirac creation ( destruction ) fermionic operators @xmath20 ( @xmath21 ) we get @xmath22 + \sum_{n=1}^{n-1}(i[d_n^{\dagger}d_{n+1 } + d_{n+1}^{\dagger}d_n ] \nonumber \\ & & + j[d_n^{\dagger}d_{n+1}^{\dagger } + d_{n+1}d_n])]\ , \label{hmdirferm}\end{aligned}\ ] ] where @xmath23 , @xmath24 .
in what follows we consider the limit @xmath25 ( semi - infinite chain ) .
eq .
( [ hmdirferm ] ) is , in fact , the hamiltonian of the inhomogeneous kitaev toy model @xcite ( the hamiltonian of the homogeneous kitaev toy model has the same form as the fermionic representation for the hamiltonian of the xy spin-1/2 chain introduced in ref . ) with @xmath26 , @xmath27 is the hopping parameter of spinless electrons , and @xmath28 , @xmath29 is the induced superconducting ( sc ) gap , or the @xmath1-wave pairing amplitude of the 1d topological superconductor @xcite , or a quantum wire @xcite with zero chemical potential of electrons and with inhomogeneities of hopping amplitudes and gaps near the edge of the chain .
zero chemical potential in kitaev s model permits the topological superconductivity , i.e. , the weak pairing regime , in which the size of cooper pair is infinite ( see below ) . the model eq .
( [ hmdirferm ] ) can also describe the 1d system of coupled cavities with strong in - cavity photon - photon repulsion and nonlinear photon driving @xcite in the cavity quantum electrodynamics .
there necessary re - definitions are @xmath30 , where @xmath31 is the magnitude of the photon driving , and @xmath32 , where @xmath33 is the tunneling amplitude for photon hopping between nearest neighbor cavities .
the term with @xmath10 describes the interaction of the edge cavity with the light @xcite .
our model is related to photons being in resonance with cavities .
it has been also pointed out recently that kitaev s model can be realized in 1d arrays of josephson junctions @xcite : the chain of sc islands coupled via strong josephson junctions to a common ground superconductors .
each island contains a pair of mf at the endpoints of a semiconductor nanowire .
the parameters of our hamiltonian are related to the one of the _ inhomogeneous _ array of josephson junctions as : @xmath34 , where @xmath35 is the tunnel coupling of individual electrons between sc islands : @xmath36 , where @xmath37 is the tunneling amplitude due to the aharonov - casher interference caused by the effective capacitance coupling between two islands ( @xmath38 is the electron charge , and @xmath39 is the induced charge , where @xmath40 is the capacitance to a common back gate at voltage @xmath41 with respect to the ground superconductor ) . finally , @xmath42 , where @xmath43 is the charging energy . @xmath44 and
@xmath45 can be tuned through the inhomogeneous gate voltage at each sc island .
we can also consider the term with the boundary field @xmath10 in @xmath15 as andreev s tunneling .
then we introduce mf as @xmath46 , @xmath47 , with @xmath48 , which satisfy anticommutation relations @xmath49 ( @xmath50 ) . in mf eq .
( [ hmdirferm ] ) reads @xmath51c_{b , n}c_{a , n+1 } + [ j - i ] \nonumber \\ & & \times c_{a , n}c_{b , n+1 } ) + 2hc_{a,-1}c_{b,0 } + ( j'+i')c_{b,0}c_{a,1 } \nonumber \\ & & + ( j'-i')c_{a,0}c_{b,1 } \bigr ] .
\label{hmmajferm}\end{aligned}\ ] ] without the interaction with the ( artificial ) spin at the site
@xmath52 the term in the hamiltonian @xmath15 , which describes the action of the edge field @xmath10 , has the form @xmath53 , i.e. , it is _ linear _ in mf operator .
hence , the parameter @xmath10 governs the behavior of the edge mf .
the formal introduction of the spin at site @xmath52 to the hamiltonian @xmath19 is related to the addition of the new ( artificial ) mf ( cf .
refs . ) , interacting with the linear edge mf .
the total term , proportional to @xmath10 in @xmath19 , becomes quadratic in mf . to diagonalize the hamiltonian @xmath19 we use the unitary transformation @xmath54 , where @xmath55 s are quantum numbers , which parameterize all eigenstates of the diagonalized hamiltonian .
these quantum numbers can describe extended ( band ) states .
besides , there is a possibility of localized states , caused by @xmath56 , @xmath57 , and @xmath58 .
let us define @xmath59 , i.e. , the transfer to mf @xmath60 .
we obtain _ two _ sets of eigenstates .
the first set of solutions describes nonzero @xmath61 for even @xmath5 , and nonzero @xmath62 for odd @xmath5 ( all other @xmath63 s and @xmath64 s are zeros ) .
the second set of solutions describes nonzero @xmath61 for odd @xmath5 , and nonzero @xmath62 for even @xmath5 ( others are zeros ) .
the details of calculations , and the eigenfunctions @xmath61 and @xmath62 , of the hamiltonian are presented in supplemental material .
the energies of the extended ( band ) states for both sets are @xmath65 . as for the localized modes ,
their energies can be written as @xmath66 ^ 2 - j^2[r_{(1,2 ) } - r^{-1}_{(1,2)}]^2 , \label{eps12}\ ] ] where @xmath67 play the role of the localization radii .
we get for the localized state of the first set of eigenfunctions @xmath68 }
\nonumber \\ & & \times \biggl(4h^2 + ( i'-j')^2 -2(i^2 + j^2 ) \nonumber \\ & & -[(2h - i - j)^2+(i'-j')^2-(i - j)^2]^{1/2 } \nonumber \\ & & \times[(2h+i+j)^2+(i'-j')^2-(i - j)^2]^{1/2}\biggr ) \ . \label{r1}\end{aligned}\ ] ] this state exists if @xmath69 [ ( 2h+i+j)^2+(i'-j')^2-(i - j)^2]>0 $ ] .
notice that @xmath70 , i.e. , the localized state decays with the distance from the edge of the chain . even for the homogeneous case @xmath71 , @xmath72 for @xmath73
such a localized mode exists at @xmath56 . for the second set we obtain @xmath74 $ ] .
it does not depend on @xmath10 .
it is easy to check that for @xmath71 and @xmath72 such a localized state does not exist .
the ground state wave function @xmath75 ( @xmath76 ) can be written as @xmath77|0\rangle$ ] , where the wave function of cooper - like pairs is @xmath78 .
for the considered model(s ) we have @xmath79 ( see supplemental material ) , hence kitaev s topological arguments @xcite are valid for the considered model(s ) .
it means that the models are in the topologically nontrivial weak pairing phase . for extended states
mf are coupled at adjacent sites of the chain ( with superscripts @xmath80 and @xmath81 ) , and the edge of the chain produces the unpaired mf .
the parameter @xmath10 helps us to realize such a mf in observable characteristics .
it is important that in the case of periodic boundary conditions , e.g. , in the 1d topological superconductor ring , such unpaired mf is combined with the one at the other edge of the chain @xcite into the highly non - local dirac fermion . equally important
, the energy of such isolated mf can become nonzero , e.g. , @xmath10-dependent . without inhomogeneities edge
mf become zero modes in the limit @xmath25 .
nonzero edge field @xmath10 , actually , removes the degeneracy of the chain , cf .
ref . .
using the obtained total set of eigenvalues and eigenfunctions ( see supplemental material ) we can calculate any average characteristic of the considered model .
for example , for @xmath82 we have @xmath83 where the thermal averaging with the density matrix , determined by the hamiltonian @xmath19 , is performed ( @xmath84 is the temperature ) .
we also get @xmath85 i.e. , @xmath86 is determined by the first set of eigenstates ( because the contribution of the second set is zero ) , while @xmath87 is determined by the second set of eigenstates ( zero contribution from the first set ) .
the average value @xmath88 with the hamiltonian @xmath15 is equal to @xmath89 with the hamiltonian @xmath19 , i.e. , in such a way , by observing @xmath90 in the spin chain one can _ directly _ observe the average value of the mf operator . notice that @xmath91 . also , we obtain @xmath92 valid for any @xmath10 and @xmath84 ( for zero chemical potential in kitaev s model ) .
each of the obtained observables is determined by extended and localized states .
these average values can be related to the characteristics of kitaev s model @xcite , the chain of coupled cavities with strong in - cavity photon - photon repulsion and nonlinear photon driving @xcite , and the chain of sc islands coupled via strong josephson junctions to a common ground superconductors @xcite .
the term , proportional to @xmath10 in @xmath15 , i.e. , the edge mf , for kitaev s model and the model of josephson junctions is related to the edge charge , caused by the local applied potential , or to andreev s tunneling . for the quantum optics model the term , proportional to @xmath10 , describes the state of the cavity at the edge of the chain ( e.g. , the magnitude of the photon of light , proportional to the light absorption by the edge cavity ) . in the table 1 we list possible realizations of edge mf in considered systems .
there @xmath93 is the charge of edge sc island @xcite , and @xmath94 ( @xmath95 ) are the destruction ( creation ) operators for the photon in the edge cavity @xcite .
.edge majorana modes ( emm ) in quantum chains [ cols="^,^,^,^ " , ] [ table : realiz ] , @xmath96 and @xmath97 at @xmath98 .
the dashed ( blue ) line shows the contribution from extended states , the dotted ( red ) line describes the contribution from the localized mode , and the solid ( black ) line is the total value . ]
so , the presence of the edge mf can be seen from the features of temperature- and @xmath10-dependent behavior of @xmath99 .
in fact , we see that the parameter @xmath10 governs the behavior of the edge mf
. for @xmath100 we have @xmath101 as it must be . for @xmath72 and @xmath71 the localized state exists due to nonzero @xmath10 .
fig .
[ sx11ht07 ] shows the behavior of @xmath102 .
the latter is the average value of the edge mf operator for the chain of josephson junctions as a function of the strength of the local applied voltage , and for the chain of cavities in quantum optics as a function of tunneling / pumping . for the spin chain
, @xmath102 describes the local magnetic moment at the edge of the chain as a function of the local field . at small @xmath10
the average value is determined by the contribution from the extended ( band ) states , while at large @xmath10 it is determined by the localized excitation .
the edge mf ( as well as the localized state ) exists even for @xmath103 for @xmath56 ( i.e. , for kitaev s model in the absence of pairing , @xmath104 ) , due to the pairing caused by @xmath10 itself . for @xmath105 at small values of @xmath10
the average value of the local mf operator shows @xmath106 behavior . for smaller values of @xmath107 the region of @xmath10 appears ,
in which the contribution of the localized mode is zero .
similar features can be also seen in the behavior of the local susceptibility with respect to @xmath10 , @xmath108 .
for instance , temperature dependences of the local susceptibility for several values of the strength of the local applied potential ( magnetic field , tunneling ) are shown in fig .
[ chi11th000063535 ] . at @xmath100
the local susceptibility diverges ( for @xmath105 ) , while at nonzero @xmath10 it manifests the non - monotonic temperature behavior : first it grows with @xmath84 at low temperatures , gets the maximum value ( which becomes lower with the growth of @xmath10 ) , and then decays with temperature .
as the function of the temperature for @xmath109 , @xmath96 and @xmath97 the dotted ( blue ) line corresponds to the strength of the applied local voltage ( magnetic field , tunneling ) @xmath100 , the solid ( green ) line shows @xmath110 case , the dashed - dotted ( red ) line describes @xmath111 case ( where the contribution from the localized state appears , see fig .
[ sx11ht07 ] ) , and the dotted ( black ) line shows @xmath112 case . ] such a behavior of the edge mf can be observed in a spin chain with the help of , e.g. , nuclear magnetic resonance ( nmr ) . in nmr experiments with spin chains
the shift of the resonance position is proportional to the local susceptibility @xcite .
we expect similar results to persist in the case of any spin-1/2 antiferromagnetic chain with the `` easy - plane '' magnetic anisotropy ( with or without in - plane anisotropy , which is important for experimental realization in spin chain materials ) with the local magnetic field applied in - plane .
for example , spin chain materials with magnetic ions cu@xmath113 or v@xmath114 ( spin 1/2 ) often exhibit magnetic anisotropy about 5 - 10 % , and finite spin chains can be realized via substitution of nonmagnetic ions instead of magnetic ones @xcite .
single crystals of quasi-1d magnetic materials are necessary for the realization of the effect , because in powders spin chains can be directed randomly .
the local field can be caused by the proximity effect of a ferromagnet , neighboring to the spin chain , with the value of @xmath10 governed by the distance to that ferromagnet .
one can realize in - plane direction of @xmath10 by rotation of the ferromagnet .
then the local magnetic susceptibility at the edge of the spin chain can be measured via the nmr shift .
worth noting that luttinger liquid approach can not in principle describe localized states , which affect the behavior of edge mf ; however , it can describe the low-@xmath10 behavior , determined by extended states of the chain . for the chain of josephson junctions
such a characteristic can be observed when studying the charge of the edge island as a function of the voltage , applied locally to the edge of the chain @xcite , and temperature , or the tunneling andreev conductance .
finally , in quantum optics the edge mf can be detected by measuring the state of the probe cavity ( or the edge cavity ) as a function of the tunneling amplitude @xcite .
we expect similar effects for the edge mf on the opposite side of the finite chain .
for the extended states of the latter one can replace @xmath115 with integer @xmath116 . in summary , we have proposed the way of direct observation of the edge mf in several realizations in quantum chains , where excitations can be presented as superpositions of spinless fermions and holes , the necessary condition for mf : in `` easy - plane '' spin-1/2 chains with in - plane polarized magnetic field , applied to the edge of the chain ; in the chain of josephson junctions , and in the chain of cavities in quantum optics with the tunneling of photons to the edge cavity . as we have shown , such an edge mf can be observed at nonzero temperatures in experiments on dc or ac josephson currents in chains of superconducting islands , nonlinear quantum optics , and quantum spin chain materials , as the local characteristic of the edge under the action of the governing parameter , @xmath10 , which _ directly _ affects the edge mf .
support from the institute for chemistry of the v.n .
karasin kharkov national university is acknowledged . |
in order to fix ideas we considered first a simple model @xcite for a soft confined atom obeying a schrdinger equation in @xmath2 dimensions of the form @xmath3 where @xmath4 is an attractive central potential with coulomb and confining terms . if we assume a wave function of the form @xmath5 then we find the radial eigenequation implies @xmath6 if we now choose @xmath7 , @xmath8 , @xmath9 we obtain the following family of exact solutions @xmath10 the lowest radial excitations of the familiar coulomb and oscillator problems are recovered from the special cases @xmath11 or @xmath12 such specific exact solutions allow for analytical reasoning and explorations . in addition , the explicit results provide relevant test problems for the complementary approaches that must be used to complete the solution space .
the question of the existence of exact solutions and the methods for finding them are therefore an important part of the overall task . as in the choice of the function @xmath13 in the simple illustration above , it is often the case in this context that exactness has something to do with polynomials .
thus part of the paper involves the issue of when an ordinary differential equation admits polynomial solutions . as we shall see ,
the ` asymptotic iteration method ' @xcite plays a role both in the construction of exact solutions and also in finding approximations for arbitrary values of the problem parameters . in the present work we study the exact and approximate solutions of the schrdinger eigenproblem generated by a confined soft - core coulomb potential in @xmath14-dimensions , where @xmath41 as we have discussed above , for the cases we consider ,
the spectrum of this problem is discrete , all eigenvalues are real and simple , and they can be arranged in an increasing sequence @xmath42 . the paper is organized as follows . in section [ sec2 ] , we set up the schrdinger equation for the potential and discuss the correspondence second - order differential equation . in section [ method ] ,
we present our method of solution that relies on the analysis of polynomial solutions of the differential equation @xmath43 and different variants of this general differential - equation class .
we discuss in particular necessary and sufficient conditions on the equation parameters for it to have polynomial solutions .
a brief review of the asymptotic iteration method ( aim ) is presented in section [ aim ] . in section [ app1 ] , the exact and approximate solutions for the problem
are discussed , based on the results of section [ sec2 ] ; and approximate solutions are found for arbitrary potential parameters @xmath44 and @xmath45 by an application of aim .
an analysis of the corresponding exact and approximate solutions for the pure confined coulomb case @xmath46 is presented in section [ app2 ] .
the ` hard confinement ' case , that is to say when the same system confined to the interior of an impenetrable spherical box of radius @xmath47 is discussed in section [ spec ] . in each of these sections ,
the results obtained are of two types : exact analytic results that are valid when certain parametric constraints are satisfied , and accurate numerical values for arbitrary sets of potential parameters . 0.1true in
in this section , we consider the @xmath14-dimensional radial schrdinger equation for @xmath1 : @xmath48u_{nl}^d(r)=e_{nl}^d u_{nl}^d(r),\quad \beta>0,\quad a\in(-\infty,\infty),\quad 0<r<\infty , \quad u(0 ) = 0.\ ] ] we note first that the differential equation ( [ sch ] ) has one regular singular point at @xmath49 with exponents given by the roots of the indicial equation @xmath50 and an irregular singular point at @xmath51 . for large @xmath52 , the differential equation assumes the asymptotic form @xmath53u_{nl}^d(r)\approx 0\ ] ] with an asymptotic solution @xmath54 the roots @xmath55 of eq.([indicial ] ) , namely , @xmath56 determine the behaviour of @xmath57 as @xmath52 approaches @xmath58 , only @xmath59 is acceptable , since only in this case is the mean value of the kinetic energy finite @xcite .
thus , the exact solution of ( [ sch ] ) will assume the form @xmath60 where we note that @xmath61 as @xmath62 .
on insertion of this ansatz wave function into ( [ sch ] ) , we obtain the differential equation for @xmath63 as @xmath64\,r^2+(4 a b^2 - \beta c^2 - 4 b^2 \beta\,e + 2 b c ( k-1 ) + 4 b^3 \beta k)\,r+2b\beta\,c(k-1 ) \right)f_n(r)=0.\end{aligned}\ ] ] in the next section , we study the polynomial solutions of this differential equation which itself lies within a larger class of differential equations given by @xmath65 where @xmath66 and @xmath67 are real constants for @xmath68 and @xmath69 . 0.1true in
the necessary condition ( @xcite , theorem 6 ) for polynomial solutions @xmath70 of the second - order linear differential equation is @xmath71 provided @xmath72 .
the polynomial coefficients @xmath73 then satisfy the four - term recurrence relations @xmath74 the proof of follows from an application of the frobenius method .
we note that the recurrence relations can be written as a system of linear equations in the unknown coefficients @xmath75 , @xmath76 given by @xmath77 where @xmath78 thus , for zero - degree polynomials @xmath79 and @xmath80 , we must have @xmath81 , thus , in addition to the necessary condition @xmath82 , the following two conditions become sufficient @xmath83 for the first - degree polynomial solution , @xmath84 and @xmath85 , we must have @xmath86 and @xmath87 , thus , in addition to the necessary condition @xmath88 or @xmath89 it is also required that the following two @xmath90-determinants simultaneously vanish @xmath91 for the second - degree polynomial solution , @xmath92 and @xmath93 for @xmath94 , it is necessary that @xmath95 and @xmath96 , from which we have the necessary condition @xmath97 along with the vanishing of the two @xmath98-determinants @xmath99 for the third - degree polynomial solution , @xmath100 and @xmath93 for @xmath101 , we then have the necessary condition @xmath102 along with the vanishing of the two @xmath103-determinants , @xmath104 and @xmath105 for the fourth - degree polynomial solution ( @xmath106 ) , @xmath107 and @xmath93 for @xmath108 , we then have the necessary condition @xmath109 along with the vanishing of the two @xmath110-determinants , @xmath111 and @xmath112 similar expressions for higher - order polynomial solutions can be easily obtained .
the vanishing of these determinants can be regarded as the sufficient conditions under which the coefficients @xmath113 and @xmath114 of eq . can be expressed in terms of the other parameters . 0.1true in
the asymptotic iteration method ( aim ) is an iterative algorithm originally introduced @xcite to investigate the analytic and approximate solutions of the differential equation @xmath115 where @xmath116 and @xmath117 are @xmath118differentiable functions .
a key feature of this method is to note the invariant structure of the right - hand side of ( [ aim_eq ] ) under further differentiation . indeed ,
if we differentiate with respect to @xmath52 , we obtain @xmath119 where @xmath120 and @xmath121 further differentiation of equation , we obtain @xmath122 where @xmath123 and @xmath124 thus , for @xmath125 and @xmath126 derivative of ( [ aim_eq ] ) , @xmath127 , we have @xmath128 and @xmath129 respectively , where @xmath130 from and , we have @xmath131 clearly , from if @xmath132 , the solution of ( [ aim_eq ] ) , is a polynomial of degree @xmath40 , then @xmath133 .
further , if @xmath134 , then @xmath135 for all @xmath136 . in an earlier paper @xcite , we proved the principal theorem of aim , namely 0.1 in given @xmath137 and @xmath138 in @xmath139 the differential equation ( [ aim_eq ] ) has the general solution @xmath140\ ] ] if for some @xmath141 @xmath142 where @xmath143 and @xmath144 are given by .
0.1true in recently , it has been shown @xcite that the termination condition is necessary and sufficient for the differential equation to have polynomial - type solutions of degree at most @xmath40 , as we may conclude from eq .. the application of aim to a number of problems has been outlined in many publications . the applicability of the method is not restricted to a particular class of differentiable functions ( e.g. polynomials or rational functions ) , rather , it can accommodate any type of differentiable function . the fast convergence of the iterative scheme depend on a suitable choice for the starting values of @xmath145 and the correct asymptotic solutions near the boundaries @xcite .
comparing equation with and using parameters given by @xmath146 the exact solution of assumes the following form @xmath147 where @xmath148 and @xmath40 counts the number of zeros of @xmath149 , hence the number of nodes in the wave function solution .
the coefficients @xmath150 can be easily evaluated using the four - term recurrence relations , @xmath151 using the necessary condition @xmath152 the potential parameters @xmath44 and @xmath45 satisfy sufficient conditions according to the following scenarios : for a zero - degree polynomial solution , @xmath153 , if @xmath154 , the ground - state solution of equation is given by @xmath155 with ground - state eigenenergy @xmath156 in the next section , we shall focus on the case of @xmath157 which corresponds to the coulomb potential perturbed by an added polynomial in @xmath52 . in the rest of this section
, we shall assume @xmath158 for a first - degree polynomial solution , @xmath159 , @xmath160 and the exact solution wave function of equation reads @xmath161 subject to the following two conditions related the potential parameters @xmath162 since , by assumption @xmath163 and @xmath164 , @xmath165 and the polynomial solution @xmath166 has no roots , in which case @xmath167 represent a ground - state solution of the schrdinger equation subject to the parameters @xmath168 , @xmath169 and @xmath170 satisfying the conditions given by . in summary , the exact solutions of schrdinger s equation @xmath171u_{0\ell}^d(r)=e_{0\ell}^d u_{0\ell}^d(r),\quad \beta>0,\quad 0<r<\infty.\ ] ] is explicitly given by @xmath172 in figure [ fig1 ] , we display the un - normalized ground - state solution using for @xmath173 and different values of @xmath174 . and
different values of @xmath174 . ] for the rest of the spectrum we use the asymptotic iteration method as described in section [ aim ] , starting with @xmath175 where @xmath176 corresponds to the exact solution , equation yields the second - order differential for @xmath13 as @xmath177 hence , we may initiate aim with @xmath178 the question is then to find the initial value @xmath179 that stabilizes the computation of the termination - condition roots . to this end , we take the highest of the absolute values among all the roots of @xmath180 which yields @xmath181 , henceforth we shall fix @xmath179 at @xmath182 for all of our numerical computations . in table
[ table : tab1 ] , we report our results from aim for first 12 decimal places . the eigenvalue reported in table [ table : tab1 ] were computed using maple version 16 running on an ibm architecture personal computer and we have chosen a high - precision environment . in order to accelerate our computation we have written our own code for a root - finding algorithm instead of using the default procedure solve of _
maple 16_. the results of aim may be obtained to any degree of precision , although we have reported our results to only the first twelve decimal places , .the first few eigenenergies @xmath183 of schrdinger equation .
the initial value utilize aim is @xmath184 .
the subscript @xmath185 refers to the number of iteration used by aim .
+ [ cols="^,<,^,<",options="header " , ] [ table : difdn6 ]
in this work exact and approximate solutions of schrdinger s equation with softcore coulomb potentials under hard and soft confinement were found .
these problems generate an interesting class of differential equation that goes beyond the classical problems which have solutions of hypergeometric type . in this paper
the problems were analyzed as special cases of a very general scheme for the study of linear second - order differential equations with polynomial coefficients that admit polynomial solutions .
necessary and sufficient conditions are derived for the existence of such solutions .
the methods presented in this work allow us to obtain compact algebraic expressions for the exact analytical solutions .
these are then verified by the asymptotic iteration method . in cases where the parametric conditions for exact polynomial solutions are not met ,
the asymptotic iteration method is employed directly to find highly accurate numerical solutions . in this work ,
the asymptotic iteration method served two purposes .
the first was to confirm the validity of the sufficient conditions obtained analytically .
the second is to provide approximate solutions to the eigenvalue problems , whether potential parameters are specially restricted or freely chosen . for both purposes
, the method proves to be extremely effective and provides very accurate results .
it is also clear from the present work that the method and the analytic expressions obtained for the different classes of the differential equations can be easily adapted to study other eigenproblems appearing in theoretical physics .
partial financial support of this work under grant nos .
gp3438 and gp249507 from the natural sciences and engineering research council of canada is gratefully acknowledged by us ( respectively rlh and ns ) .
d. bessis , e. r. vrscay and c. r. handy , _
hydrogenic atoms in the external potential @xmath187 : exact solutions and ground - state eigenvalue bounds using moment methods , _ j. phys .
a : math . gen . * 20 * ( 1987 ) 419 .
f. caruso , j. martins , and v. oguri , _ solving a two - electron quantum dot model in terms of polynomial solutions of a biconfluent heun equation _ , ann .
phys . * 347 * , 130 ( 2014 ) , arxiv:1308.0815 ( december 2013 ) .
b. laut and g. marcilhacy , _ on the schrdinger equations of rotating harmonic , three - dimensional and doubly anharmonic oscillators and a class of confinement potentials in connection with the biconfluent heun differential equation _ ,
a : math . gen .
* 19 * ( 1986 ) 3527 .
b. laut , g. marcilhacy , r. pons , and j. skinazi , _ on the connection problem for some schrdinger equations in relation to the biconfluent heun differential equation _ siam j. math
* 21 * ( 1990 ) 793798 .
e. ovsiyuk , m. amirfachrian , and o. veko , _ on schrdinger equation with potential @xmath188 and the bi - confluent heun functions theory _ , nonl .
. sys . * 15 * , 163 ( 2012 ) , arxiv:1110.5121 ( october 2011 ) . | analytic and approximate solutions for the energy eigenvalues generated by a confined softcore coulomb potentials of the form @xmath0 in @xmath1 dimensions are constructed . the confinement is effected by linear and harmonic - oscillator potential terms , and also through ` hard confinement ' by means of an impenetrable spherical box .
a byproduct of this work is the construction of polynomial solutions for a number of linear differential equations with polynomial coefficients , along with the necessary and sufficient conditions for the existence of such solutions .
very accurate approximate solutions for the general problem with arbitrary potential parameters are found by use of the asymptotic iteration method .
# 1 # 1eq.([#1 ] ) cuqm-149 |
quantum turbulence in counterflow of he ii has left a mystery : the transition between two different kinds of superfluid turbulence , namely , the t1-t2 transition @xcite .
recent developments in visualization techniques have enabled us to study this mystery . however
, our theoretical understanding has not yet been developed enough for two chief reasons : 1 .
the difficulty in taking proper account of the inhomogeneity of the system and the geometry of the channel in the formulation and 2 . the need to solve the coupled dynamics of the two fluids to understand the phenomena fully . in this paper , in order to address the first reason we perform a numerical analysis of the vortex dynamics by prescribing realistic nonuniform profiles to the flow of the normal fluid .
the present work reveals characteristic inhomogeneous structure and dynamics of the vortex tangle that has not been reported in preceding works with a uniform flow profile @xcite @xmath1 @xcite .
we do not address directly the t1-t2 transition .
if we can address the coupled dynamics of the two fluids , we will be able to reveal the t1-t2 transition , and we are going to report on this work in the near future .
this study is motivated by the recent observation by marakov _
_ @xcite of the flow profiles of the normal fluid in thermal counterflow in a square channel .
they have observed the flow profiles of the normal fluid by following the motion of seeded metastable @xmath2 molecules by a laser - induced - fluoresence technique @xcite .
then in a thermal counterflow they prepare thin @xmath2 molecular tracer lines perpendicular to the flow created by femtosecond - laser field ionization . above 1
k the molecules interact strongly with the normal fluid and trace the motion .
thus how the tracer lines move reveals the flow profile of the normal fluid .
first , they presented evidence that the flow of the normal fluid is indeed turbulent at relatively large velocities @xcite . in their recent studies
they observe the novel transition of the flow profile of the normal fluid @xcite . at relatively small velocities ,
the flow of the normal fluid remains laminar , whereas the superfluid is already turbulent . for heat current @xmath3 , where the value of @xmath4 obtained from the experiment is @xmath5@xmath6 , an initially straight tracer line deforms to a nearly parabolic shape , indicating a laminar poiseuille velocity profile . for @xmath7 , where the value of @xmath8 obtained from the experiment is @xmath9 ,
the outer part of the tracer line becomes flattened , which is called the tail - flattened velocity profile .
such a laminar profile has never been reported in classical fluid dynamics , and it must be characteristic of he ii , namely , a result of the interaction between the normal fluid and the superfluid . in this work we study a numerical simulation of the vortex dynamics under these laminar profiles of the normal fluid .
liquid @xmath0he shows superfluidity below a @xmath10 point of @xmath11 . in a two - fluid model
one considers superfluid @xmath0he as an intimate mixture of normal fluid and superfluid components : the normal fluid has a viscosity , whereas the superfluid is inviscid .
the velocities of the normal fluid and the superfluid component are denoted by @xmath12 and @xmath13 , respectively , and the densities are denoted by @xmath14 and @xmath15 , respectively .
the total density @xmath16 is almost independent of temperature .
the ratio of the two fluids depends on temperature , and @xmath17 increases as temperature decreases .
the velocity fields obey different two equations of motion @xcite , and they are basically decoupled .
the idea of quantized circulation in superfluid helium was suggested by onsager @xcite and has been observed experimentally by vinen @xcite . in a superfluid
any rotational motion is sustained only by quantized vortices , which have the quantum circulation @xmath18 , where @xmath19 is planck s constant and @xmath20 is the mass of a @xmath21 atom .
the elementary excitations forming the normal fluid are strongly scattered by these quantized vortices .
thus , if there is a relative velocity between the normal fluid and the quantized vortices , a frictional force works between them , being called a mutual friction force . in the superfluid turbulent state the mutual friction term is added to the equations of motion of the two fluids , so that the two fluids become coupled .
thermal counterflow in he ii , which is internal convection of the two fluids , has been studied well as a prototype of quantum turbulence .
one end of the channel is connected with a he ii bath and the other end is closed . upon heating the closed end , the normal
fluid flows to the cooler side to transfer heat .
the superfluid flows oppositely to the warmer end to satisfy mass conservation : @xmath22 where the integral is performed over the cross section of the channel .
thus a relative velocity @xmath23 occurs between the two fluids , where the overline denotes the spatial average over the channel cross section .
when the counterflow velocity exceeds a critical value , a self - sustaining tangle of quantized vortices appears to form superfluid turbulence .
the measurements @xcite show that the vortex line density @xmath24 follows the relation @xmath25 where @xmath26 is a parameter depending on temperature and @xmath27 is a small parameter .
the scheme for understanding quantum turbulence from the vortex dynamics was addressed by vinen . by assuming homogeneous turbulence he estimated the vortex growth by using a dimensional analysis and modeled the decay process phenomenologically @xcite .
he showed that the dynamics of the vortex tangle is described by vinen s equation @xmath28 where @xmath29 and @xmath30 are temperature - dependent parameters .
if the vortex tangle is in a steady state , @xmath31 , resulting in eq .
( [ vinen ] ) .
schwarz has investigated counterflow quantum turbulence using the vortex filament model @xcite .
the observable quantities obtained in his calculation agree with the experimental results for the steady state of vortex tangles .
however , his simulation could sustain the steady vortex tangle only through some artificial procedure .
et al . _ argued that this comes from the localized induction approximation ( lia ) in which the interaction between vortices is neglected @xcite . by performing the full biot - savart calculation they removed the difficulty and
successfully obtained the steady state consistent with the experimental results .
we note that in these numerical works the system was assumed to be uniform and the physics coming from the inhomogeneity was missed ; i.e. , the effects of the channel walls and the resulting flow profile were missed .
counterflow quantum turbulence is known to depend on the aspect ratio of the cross section of the channel @xcite . in low - aspect - ratio channels ,
the system has two turbulent states .
an increase in the counterflow velocity is observed to change the laminar state to the first turbulent state t1 , and next to the second turbulent state t2 .
melotte and barenghi @xcite suggested that the transition from t1 to t2 is caused by the transition of the normal fluid from laminar to turbulent .
they addressed the issue of the stability of the normal fluid with the relation to the t1-t2 transition .
in high - aspect - ratio channels , the counterflow exhibits only a single turbulent state t3 , but there has been little information about what t3 is .
the recent visualization experiments have made important breakthroughs for these issues . by using micron - sized solid hydrogen tracers the maryland group @xcite succeeded in visualizing quantized vortices in superfluid @xmath0he .
then they applied the technique to thermal counterflow to confirm directly the behavior of the two - fluid model @xcite .
et al . _ have studied a counterflow channel containing a cylinder and showed that macroscopic eddies appear both downstream and upstream of the cylinder @xcite .
la mantia _
have investigated the lagrangian dynamics of micrometer - sized solid particles in thermal counterflow and obtained the normalized probability functions of the particle velocities and accelerations @xcite .
recently , visualization of normal fluid flow in thermal counterflow has been performed .
_ @xcite found that the normal fluid in a thermal counterflow has two characteristic laminar flows , as described in the second paragraph of this introduction .
recent developments in cryogenic visualization techniques have shed light on the problems of inhomogeneous quantum turbulence .
we mention previous numerical studies of thermal counterflow in he ii with a nonuniform profile .
samuels @xcite studied the dynamics of quantized vortices in laminar circular pipe flow with normal fluid poiseuille flow and superfluid flat flow in the same direction .
aarts _ et al . _
@xcite studied the vortex tangle in thermal counterflow by assuming a normal - component velocity profile ether flat or parabolic in variously shaped channels with the lia .
@xcite@xmath1@xcite studied thermal counterflow between two parallel plates by using the prescribed poiseuille and turbulent profiles for the normal fluid flow . a vortex tangle obtained by baggaley _ et al .
_ @xcite concentrated near the channel walls and showed characteristic inhomogeneity .
the simulation generated quantum turbulence under laminar poiseuille flow and turbulent flow of the normal fluid . by examining results
, they argued that their results supported the scenario proposed by melotte and barenghi .
a better understanding of t1 and t2 states would be obtained by studying the flow in a low - aspect - ratio channel where all boundaries are solid except for the flow direction .
this is because the t1 and t2 states are actually observed in low - aspect - ratio channels and another turbulent state t3 is observed in high - aspect - ratio channels @xcite .
in this paper we performed a numerical analysis of thermal counterflow in a square channel .
it is necessary to address the coupled dynamics of the two fluids to understand the t1-t2 transition . in a typical model of the coupled dynamics
one would suppose that the superfluid is described by the vortex filament model , the normal fluid is described by the navier - stokes equation , and they are coupled through mutual friction @xcite .
however , it is difficult to solve fully the coupled dynamics . as the first essential step
, we address the t1 state where the normal fluid must be laminar .
the preliminary results for hagen - poiseuille flow was reported in the previous study @xcite .
the normal fluid is prescribed to be two types of realistic flow : hagen - poiseuille flow and tail - flattened flow . in both profiles ,
an inhomogeneous vortex tangle is obtained as a statistically steady state , where a boundary layer of the superfluid appears near the channel walls . under hagen - poiseuille flow
the vortex tangle shows a characteristic space - time oscillation .
tail - flattened flow was just observed by marakov _
@xcite , and its nature is unknown .
the present simulation reveals that tail - flattened flow is some intermediate state between hagen - poiseuille flow and uniform flow .
the contents of this paper are as follows .
section ii describes the formulation of this work . in sec .
iii we show results of our simulation with hagen - poiseuille flow . in sec .
iv we show results of our simulation with tail - flattened flow .
section v is devoted to a discussion .
section vi presents our conclusions .
here we describe the formulation of a vortex filament model and its numerical analysis .
the equation of motion of a vortex filament is first explained , then the velocity fields of the normal fluid are described and the conditions of our simulation are noted .
the physical quantities necessary for characterizing the vortex tangle are also introduced . in a vortex filament model @xcite a quantized vortex
is represented by a filament passing through a fluid and has a definite vorticity .
this approximation is very suitable in he ii , because the core size of a quantized vortex is much smaller than any other characteristic length scale . at zero temperature
the vortex filament moves with the superfluid velocity @xmath32 where @xmath33 is the velocity field produced by vortex filaments , @xmath34 is that produced by solid boundaries , and @xmath35 is the applied uniform flow of the superfluid .
the velocity field @xmath33 is given by the biot - savart law @xmath36 the filament is represented in parametric form as @xmath37 , where @xmath38 refers to a point on the filament and the integration is performed along the filament .
this work addresses the full biot - savart integral @xcite .
the velocity field @xmath34 is obtained by a simple procedure ; it is just the field produced by an image vortex that is constructed by reflecting the filament into the surface and reversing its direction . according to eq .
( [ mass ] ) we set @xmath39 in thermal counterflow , where @xmath40 . by taking into account mutual friction ,
the velocity of a point @xmath41 on the filament is given by @xmath42 , \label{vortex}\ ] ] where @xmath43 and @xmath44 are the temperature - dependent coefficients , and the prime denotes derivatives of @xmath45 with respect to the coordinate @xmath46 along the filament .
the velocity @xmath12 of the normal fluid is prescribed to be two profiles , namely , the hagen - poiseuille profile @xmath47 and the tail - flattened @xmath48 profile .
we suppose that the flow direction is along the @xmath49 coordinate and that @xmath50 are coordinates are normal to @xmath49 , namely , @xmath51 ; the center of the channel is @xmath52 . in a rectangular channel
the hagen - poiseuille profile is given by @xmath53 \frac { \cos(m \pi y / 2 a ) } { m^3 } , \end{split } \label{poi}\ ] ] where @xmath54 is a normalization factor and @xmath55 are halves of the channel width along the @xmath56 axes , respectively @xcite . , normalized by the mean value .
( a ) profile with @xmath57 .
the color bar shows the value of @xmath58 .
( b ) profiles along @xmath59 with various values of @xmath19 .
, scaledwidth=35.0% ] we realize the tail - flattened profile by composing the expression@xmath60 , \label{fla}\ ] ] where @xmath61 $ ] refers to the larger value of the two arguments , and @xmath62 is a fitting parameter , which determines the flattened area .
figure [ fla_profile](a ) shows the profile of tail - flattened flow over the channel cross section , and fig .
[ fla_profile](b ) shows the profiles along @xmath59 with various @xmath19 values .
@xmath58 is equal to @xmath63 for @xmath64 , and @xmath58 becomes uniform for @xmath65 corresponding to simulation by adachi @xmath66 @xcite .
the increase in @xmath19 makes @xmath58 more uniform , as shown in fig . [
fla_profile](b ) .
the simulations are performed under the following conditions .
we discretize the vortex lines into a number of points held at a minimum space resolution of @xmath67 .
integration in time is achieved using a fourth - order runge - kutta scheme with time resolution @xmath68 .
the computing box is @xmath69 .
periodic boundary conditions are used along the flow direction @xmath49 , whereas solid boundary conditions are applied to the channel walls at @xmath70 and @xmath71 .
the solid boundaries are assumed to be smooth completely , thus neither vortex pinning nor surface drag occurs .
we reconnect two vortices artificially , when the vortices approach each other more closely than @xmath72 .
we eliminate vortices that are shorter than @xmath73 .
temperatures @xcite are @xmath74 , @xmath75 , and @xmath76 .
the density of the normal fluid @xcite is @xmath77 at @xmath78 , @xmath79 at @xmath80 , @xmath81 at @xmath82 .
the initial state consists of eight randomly oriented vortex rings of radius = @xmath83 . in this subsection
we introduce the physical quantities used in this paper .
the vortex line density is defined as @xmath84 where the integral is performed along all vortices in the sample volume @xmath85 .
the anisotropy of the vortex tangle is represented by dimensionless parameter @xcite : @xmath86 d\xi.\ ] ] here @xmath87 represents the unit vector parallel to the flow direction .
if the vortex tangle is completely isotropic , @xmath88 .
if the tangle consists entirely of curves lying in the plane normal to the flow direction , @xmath89 . to analyze the inhomogeneity of the vortex tangle , we divide the computational box into subvolumes by using a uniform @xmath90 cartesian mesh
the vortex line density is defined at a subvolume as the local vortex line density @xmath91 where the integral is performed along all vortices in the subvolume @xmath92 .
the local anisotropic parameters are defined as @xmath93 d\xi.\ ] ] these local quantities are always averaged over the flow direction @xmath49 .
thus these are functions of position @xmath94 normal to the flow direction , namely , @xmath95 and @xmath96 .
in this section we present results of our simulation under the hagen - poiseuille flow expressed by eq .
( [ poi ] ) .
we analyze physical quantities to discuss the dynamical and statistical properties of the quantum turbulence .
we discuss the time development of @xmath24 , and we show that the inhomogeneous vortex tangle is sustained by space - time oscillation .
this is characteristic of the present case , which can not appear in uniform counterflow @xcite . under hagen - poiseuille flow with @xmath97 .
, scaledwidth=40.0% ] figure [ poi_vld_19k.eps ] shows the time evolution of @xmath24 .
the temperature is @xmath97 , and the mean velocity of the normal fluid is @xmath98 and @xmath99 .
the vortex tangle develops to the statistically steady state , even if the counterflow is nonuniform .
fluctuations in @xmath24 are much larger than those in uniform counterflow @xcite .
the amplitude of the fluctuation increases with @xmath100 .
the large fluctuation would imply that this system is unstable . to investigate the origin of the large fluctuation , we focus on the configuration of the vortex tangle .
the large fluctuation of @xmath24 is attributable to a space - time oscillation coming from the inhomogeneous dynamics of the vortex tangle . in the following the details of the oscillation
are explained . and @xmath101 , corresponding to the results of fig . [ poi_vld_19k.eps ] .
these show one cycle of the large oscillation , while snapshots from ( a ) to ( d ) correspond a local minimum ( @xmath102 ) , the middle of the increase ( @xmath103 ) , a local maximum ( @xmath104 ) , and the middle of the decrease ( @xmath105 ) , respectively . , scaledwidth=45.0% ] figure [ tangles ] shows the space - time pattern of the vortex tangle at @xmath106 and @xmath107 .
the period of the space - time oscillation is about @xmath108 , and the oscillation consists of four stages ( a)(d ) . in fig .
[ tangles](a ) , corresponding to a local minimum of @xmath24 , the vortices are dilute , and the vortices remain only near the solid walls .
then the vortices near the walls invade the central region in fig .
[ tangles](b ) .
these vortices make numerous reconnections in the central region subject to the large counterflow in fig .
[ tangles](c ) .
hence @xmath24 increases significantly to a local maximum . eventually , in fig .
[ tangles](d ) , the hagen - poiseuille flow excludes the vortex tangle from the central region toward the solid walls .
thus the vortices in the central region become dilute , and they are absorbed by the solid boundaries . even if the vortex tangles of ( b ) and ( d ) have the almost same values of @xmath24 , the vortex configurations are quite different .
then the vortex tangle returns to the stage of fig .
[ tangles](a ) .
the vortex tangle repeats the periodic motion , resulting in the large oscillation of fig .
[ poi_vld_19k.eps ] .
the vortex tangle sustained by this mechanism is more dilute than that in the case of uniform counterflow for the same @xmath109 and @xmath110 .
we investigate the statistics of the quantum turbulence in a statistically steady state . in this subsection
every physical quantity is averaged temporally over the statistically steady states and spatially over the computational box .
typical experiments using second sound attenuation have observed values integrated over the channel .
the statistical values have been extensively investigated both experimentally and numerically .
we discuss the properties of hagen - poiseuille normal fluid flow by comparing the statistical values with those obtained in previous studies of uniform counterflow . .
the dashed lines refer to the fitting of the data for each temperature .
, scaledwidth=40.0% ] . line density coefficients @xmath26 , fitting parameter @xmath27 and anisotropic parameter @xmath111 . given are numerical results @xmath112 , @xmath113 and @xmath27 obtained in this work under hagen - poiseuille flow and @xmath114 and @xmath115 obtained by adachi _
@xcite , which is a simulation subject to uniform counterflow under a periodic boundary condition .
[ cols="^,^,^,^,^,^ " , ] [ gamma2 ] by using these results we calculate @xmath116 .
table [ gamma2 ] lists the values of modified @xmath26 for hagen - poiseuille flow , @xmath117 , and for tail - flattened flow , @xmath118 .
the values of @xmath119 are larger than the values of @xmath26 , because the reduction in @xmath120 becomes greater with increasing @xmath100 .
the difference between @xmath26 and @xmath119 becomes smaller as temperature decreases .
this is because @xmath121 increases with @xmath109 , so that the contribution from the reduction in @xmath120 becomes smaller .
the present work addresses only the case in which the aspect ratio of the channel is unity , but the change of the ratio should seriously affect the profile of the vortex line density .
the aspect ratio of the channel plays an important role in the density and the profile of the vortex tangle .
as long as the counterflow velocity is fixed , the increase of the aspect ratio from unity reduces the counterflow velocity gradient along the long side of the cross section .
thus the exclusion of the vortex tangle shown in figs .
[ tangles](d ) does not occur so much along the long side of the cross section .
hence the vortex line density should increase compared to the vortex line density in a low - aspect - ratio channel .
it would be meaningful to consider how the aspect ratio affects the normal fluid , though it is prescribed in this formulation .
a linear stability analysis of the navier - stokes equation shows that the critical reynolds number for turbulence transition of a viscous fluid increases significantly with decreasing aspect ratio @xcite .
we expect that the normal fluid in counterflow could remain laminar in a low - aspect - ratio channel even if the superfluid becomes turbulent and the vortex tangle disturbs the normal fluid , which may correspond to the t1 state . to understand the t1 state , therefore , studies in a low - aspect - ratio channel will be indispensable . considering the report @xcite by marakov _
_ we would expect the t1-t2 transition to be a transition of the normal fluid from tail - flattened flow to turbulent flow . here
we discuss the t1-t2 transition under this assumption .
the nature of the t1 state can be understood by using the results of tail - flattened flow obtained in this work
. the t1 state will be the about homogeneous quantum turbulence . under tail - flattened flow ,
the vortex line density increases with flat parameter @xmath19 , becoming saturated at @xmath122 .
this suggests that the vortex line density does not jump at the transition from tail - flattened flow to turbulent flow . according to experiments @xcite , at the t1-t2 transition
the vortex line density does not jump , in agreement with our results .
the value of @xmath123 obtained in this work is larger than @xmath26 of the t1 state obtained in the experiment .
the origin of this difference is unknown , but addressing this issue by using coupled equations for the two fluids would resolve this discrepancy .
in this study , we have investigated counterflow quantum turbulence with nonuniform flows of the normal fluid using the vortex filament model .
the velocity field of the normal fluid was prescribed to consist of two nonuniform profiles , and the solid boundary condition is applied to the walls of a square channel .
first , we conducted a simulation with hagen - poiseuille flow .
the vortex line density has a fluctuation with a large amplitude .
we investigated the configuration of the vortex tangle in the oscillation , and it was revealed that one cycle of the oscillation consists of four characteristic states .
the relation @xmath124 is almost satisfied , but there are quantitative discrepancies between the results of this work and those found in uniform counterflow .
the anisotropy found here is larger than in the case of uniform counterflow .
then we investigated the inhomogeneity of the vortex tangle .
the quantized vortices concentrate near the channel walls , and the vortex tangle becomes more isotropic near the solid boundaries .
the region near the solid boundary is a boundary layer of the superfluid , which was also found by baggaley _
_ @xcite@xmath1@xcite .
we investigated the superfluid velocity field .
it is found that a velocity field @xmath33 opposite to the applied flow @xmath35 is created by the inhomogeneous vortex tangle .
thus the superfluid flow becomes smaller than the applied one .
second , we conducted a simulation with tail - flattened flow .
we showed that the vortex line density increases significantly as the flat parameter @xmath19 increases , becoming saturated at @xmath122 .
we confined ourselves to @xmath57 to be consistent with the experiments .
the value of @xmath26 was found to be larger than those obtained from simulations with uniform counterflow , and the anisotropy was found to be slightly larger than in the case of uniform counterflow .
the inhomogeneity is significantly reduced in comparison to the case of hagen - poiseuille flow . in the center of the channel
the vortex line density becomes slightly lower , and the vortex tangle becomes slightly anisotropic .
this is because the hagen - poiseuille profile remains in the central region of the tail - flattened profile .
the velocity field @xmath33 is different from that in the case of hagen - poiseuille flow , where the reduction of the central region becomes greater .
the present simulation revealed that tail - flattened flow is some intermediate state between hagen - poiseuille flow and uniform flow .
when we focus on the physics quantities averaged over the whole volume , we find that they are similar to those of uniform counterflow .
however , if we look at the spatial dependence of these values , they show inhomogeneous behavior .
these considerations suggest that some _ in situ _ observation is necessary to understand tail - flattened flow .
99 j. t. tough , in _ progress in low temperature physics _ , edited by d. f. brewer ( north - holland , amsterdam , 1982 ) , vol .
k. w. schwarz , phys .
b * 38 * , 2398 ( 1988 ) .
h. adachi , s. fujiyama , and m. tsubota , phys . rev .
b * 81 * , 104511 ( 2010 ) .
a. marakov , j. gao , w. guo , s. w. van sciver , g. g. ihas , d. n. mckinsey , and w. f. vinen , phys .
b * 91 * , 094503 ( 2015 ) .
w. guo , j. d. wright , s. b. cahn , j. a. nikkel , and d. n. mckinsey , phys .
lett . * 102 * , 235301 ( 2009 ) .
w. guo , s. b. cahn , j. a. nikkel , w. f. vinen , and d. n. mckinsey , phys .
lett . * 105 * , 045301 ( 2010 ) .
r. j. donnelly , in _
quantized vortices in helium ii _ , edited by a. m. goldman _
( cambridge university press , cambridge , england , 1991 ) .
l. onsager , nuovo cimento suppl .
* 6 * , 279 ( 1949 ) .
w. f. vinen , proc .
london ser .
a * 260 * , 218 ( 1961 ) .
w. f. vinen , proc .
london ser .
a * 242 * , 493 ( 1957 ) .
d. j. melotte and c. f. barenghi , phys . rev .
lett . * 80 * , 4181 ( 1998 ) .
g. p. bewley , d. p. lathrop , and k. r. sreenivasan , nature ( london ) * 441 * , 588 ( 2006 ) .
m. s. paoletti , r. b. fiorito , k , r. sreenivasan , and d. p. lathrop , j. phys .
jpn . * 77 * , 111007 ( 2008 ) .
t. zhang and s. w. van sciver , nat .
* 1 * , 36 ( 2005 ) . m. la mantia and l. skrbek , phys . rev .
b * 90 * , 014519 ( 2014 ) .
d. c. samuels , phys .
b * 46 * , 11714 ( 1992 ) .
r. g. k. m. aarts and a. t. a. m.
de waele , phys .
b * 50 * , 10069 ( 1994 ) . a. w. baggaley and s. laizet , phys .
fluids * 25 * , 115101 ( 2013 ) . a. w. baggaley and s. laurie , j. low temp .
phys . * 178 * , 35 ( 2014 ) .
d. kivotides , phys . rev .
b * 76 * , 054503 ( 2007 ) .
s. yui and m. tsubota , j. phys .
ser . * 568 * , 012028 ( 2014 ) .
k. w. schwarz , phys .
b * 31 * , 5782 ( 1985 ) . , edited by r. w. johnson ( crc press , boca raton , 1998 ) . c. f. barenghi , r. j. donnelly , and w. f. vinen , j. low temp . phys .
* 52 * , 189 ( 1983 ) .
r. k. childers and j. t. tough , phys .
b * 13 * , 1040 ( 1976 ) .
r. t. wang , c. e. swanson , and r. j. donnelly , phys .
b * 36 * , 5240 ( 1987 ) . j. maurer and p. tabeling , europhys .
43 * , 29 ( 1998 ) .
s. r. stalp , l. skrbek , and r. j. donnelly , phys .
lett . * 82 * , 4831 ( 1999 ) .
t. tatsumi and t. yoshimura , j. fluid mech .
* 212 * , 437 ( 1990 ) . | we perform a numerical analysis of counterflow quantum turbulence of superfluid @xmath0he with nonuniform flows by using the vortex filament model . in recent visualization experiments nonuniform laminar flows of the normal fluid , namely , hagen - poiseuille flow and tail - flattened flow , have been observed .
tail - flattened flow is a novel laminar flow in which the outer part of the hagen - poiseuille flow becomes flat . in our simulation , the velocity field of the normal fluid
is prescribed to be two nonuniform profiles .
this work addresses a square channel to obtain important physics not revealed in the preceding numerical works . in the studies of the two profiles we analyze the statistics of the physical quantities . under hagen - poiseuille flow
, inhomogeneous quantum turbulence appears as a statistically steady state .
the vortex tangle shows a characteristic space - time oscillation . under tail - flattened flow ,
the nature of the quantum turbulence depends strongly on that flatness .
vortex line density increases significantly as the profile becomes flatter , being saturated above a certain flatness .
the inhomogeneity is significantly reduced in comparison to the case of hagen - poiseuille flow .
investigating the behavior of quantized vortices reveals that tail - flattened flow is an intermediate state between hagen - poiseuille flow and uniform flow . in both profiles we obtain a characteristic inhomogeneity in the physical quantities , which suggests that a boundary layer of superfluid appears near a solid boundary .
the vortex tangle produces a velocity field opposite to the applied superfluid flow , and , consequently , the superfluid flow becomes smaller than the applied one . |
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.