article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
explaining the anomalous normal state properties of the so - called pseudogap regime of the underdoped cuprate superconductors is widely regarded as an essential step toward understanding the basic physics of these materials and unlocking the mechanism of their superconductivity@xcite .
perhaps the most striking aspect of these is the universal , or scaling , behavior , first identified in measurements of their temperature - dependent uniform magnetic susceptibility@xcite , and since found in knight shift , transport , and entropy measurements . in the present article we present a detailed review of scaling behavior in the underdoped cuprates that extends previous analyses of its manifestations in both static and low frequency dynamic behavior as well as that seen in inelastic neutron scattering ( ins ) experiments .
our review updates our earlier analysis@xcite and the results presented in norman _ _ et al.__@xcite , and complements the recent review of gap behavior presented in hfner _ _
et al.__@xcite .
we find that from zero hole doping until planar doping levels of @xmath5 are reached , the scaling behavior seen by probes of magnetic behavior reflects the presence of a spin liquid whose behavior maps onto the theoretical monte carlo calculations of the 2d heisenberg model of localized @xmath6 spins@xcite for most of its temperature domain .
we use this to extract the doping dependence of the strength , @xmath1 of the spin liquid component and the effective interaction , @xmath2 between the remnant localized spins that compose it ; for @xmath7 , @xmath8 $ ] for both the 2 - 1 - 4 and 1 - 2 - 3 materials , while , to first approximation , @xmath9 , where @xmath10 is their interaction at zero doping level .
a careful analysis of the nmr experiments on both classes of materials makes it possible to identify a quantum critical point at a doping level , @xmath11 that represents a phase transition from short range to long range order in the spin liquid .
it leads to dynamic @xmath12 scaling behavior for a wide range of doping levels that extends up to @xmath13 , the temperature at which the static susceptibility is maximum , corresponding to an antiferromagnetic correlation length of order unity .
we find that the extent of this scaling behavior is different for the 2 - 1 - 4 and 1 - 2 - 3 materials : for the former it persists down to temperatures of order the superconducting transition temperature ; for the latter it cuts - off at @xmath14 , a temperature that it considerably greater than @xmath15 over most of the doping range .
in addition to the spin liquid , whose properties dominate the low frequency magnetic response , bulk susceptibility measurements reveal the presence of a second component , a fermi liquid that makes a temperature independent , but doping dependent contribution to this quantity for temperatures greater than @xmath14 and doping levels of @xmath16 upwards .
we present a simple interpretation of the two fluid description of these coupled liquids@xcite in terms of the incomplete hybridization of the cu @xmath17 and o @xmath18 bands ; the spin liquid corresponds to the unhybridized @xmath19 component , while the fermi liquid has a large fermi surface as a result of the @xmath20 hybridization .
we derive the strength of the fermi liquid component , which goes as @xmath21 $ ] and so is proportional to @xmath22 , and show how the presence of the spin liquid is incompatible with the single band hubbard and zhang - rice approximations .
we conclude that experiment has now provided the answer to the question of the physical mechanism responsible for the remarkable pseudogap behavior seen in the underdoped 1 - 2 - 3 materials ( @xmath23 , say ) . when the present analysis is combined with the recent arpes experiments@xcite and
the stm measurements of the davis@xcite and yazdani@xcite groups , a simple physical picture emerges . in the `` normal state '' , for temperatures above @xmath24 , one has two quasi - independent components : a spin liquid of localized cu spins described by the 2d heisenberg model , whose strength and effective interaction become weaker as the doping level increases ; and a ( non - landau ) fermi liquid with a large fermi surface whose strength increases with doping and whose transport properties are determined primarily by its coupling to the spin liquid . at @xmath14
the system makes a transition to a remarkable new quantum state of matter : a state that possesses a single d - wave like gap , with a maximum gap value of order @xmath25 , that only becomes superconducting at the typically much lower superconducting transition temperature , @xmath15 .
the physical mechanism for the transition at @xmath14 ( and subsequently at @xmath15 ) in the 1 - 2 - 3 materials is magnetic because the scale of @xmath14 and the gap is set by the effective interaction between the localized spins in the spin liquid .
matters are somewhat different for the 2 - 1 - 4 materials and we speculate as to why this is the case . our review is organized as follows . in section 2
we review the literature on experimental measurements and corresponding analyses that indicate universal scaling behavior . in section 3 we introduce the phenomenological two - fluid model and use it to analyze existing magnetic and thermodynamic measurements on the bulk spin susceptibility , the entropy , and the spin fluctuation spectrum revealed in nuclear magnetic resonance and inelastic neutron scattering experiments . in section 4
we present our conclusions concerning the interaction between the fermi liquid quasiparticles and the spin liquid excitations , discuss the similarities and differences between the 2 - 1 - 4 and 1 - 2 - 3 materials , and consider the constraints imposed by experiment on microscopic theories of their high - temperature superconductivity .
we present our conclusions in section 5 .
the observation of scaling in the cuprates is not new@xcite .
not long after the discovery of the cuprate superconductors@xcite , universal behavior was identified in the magnetic properties of the 2 - 1 - 4 materials by johnston@xcite through an analysis of his measurements of the bulk spin susceptibility ; his analysis was later confirmed by nakano _ _ et al.__@xcite and oda _ _ et al.__@xcite .
the johnston - nakano scaling analysis was subsequently extended by wuyts _ _ et al.__@xcite to the knight shift measurements of alloul _ _ et al.__@xcite in the 1 - 2 - 3 family ; more recently it has been shown to be applicable to the 1 - 2 - 4 and several other members of the 1 - 2 - 3 family by curro _ _ et al.__@xcite and the authors@xcite .
a number of other experiments also indicate scaling and data collapse .
these include electronic heat capacity measurements@xcite for which an analysis of the magnetic entropy found scaling behavior , quantum critical ( qc ) scaling behavior in nmr copper nuclear spin - lattice relaxation rates@xcite , limits on the @xmath26-linear behavior of resistivity@xcite , scaling of hall resistivity@xcite , @xmath27 scaling in inelastic neutron scattering@xcite , and finite - size scaling in the insulating cuprates@xcite .
recently , scaling behavior has been discovered in the doping - dependence of arpes ( angle - resolved photoemission ) and stm experiments on the 2 - 2 - 1 - 2 members of the 1 - 2 - 3 family ; similar characteristic temperatures set the scale for the appearance of fermi arcs@xcite , and `` normal state '' gap behavior@xcite . in this section
we review the above experiments and find that the characteristic scaling temperature first identified by johnston , the temperature , @xmath28 , at which the temperature and doping dependent bulk magnetic susceptibility reaches its maximum value , provides the common thread that links these together .
the scaling behavior of the temperature dependent bulk spin susceptibility , @xmath29 , was discovered empirically by johnston@xcite , who showed that an excellent collapse ( fig.[johnstonfig ] ) of his experimental data on five samples of la@xmath30sr@xmath31cuo@xmath32 , in which the doping level @xmath33 ranged from @xmath34 to @xmath35 , could be obtained if @xmath36 had the following scaling form : @xmath37 f(t / t^m(x ) ) , \label{john1}\ ] ] where @xmath22 is the doping level , @xmath38 is a doping - dependent , temperature independent term , @xmath39 is the maximum value of @xmath40 for a given doping level , and @xmath28 is the doping - dependent temperature at which @xmath29 is maximum .
johnston concluded that the scaling parameter @xmath13 depends only on the hole doping level , @xmath22 , in the plane and that the scaling function @xmath41 is the same as that calculated for the 2d heisenberg model in its spin liquid regime ( i.e. at temperatures above the nel ordering temperature ) . in this model @xmath42 , where @xmath10 is the nearest neighbor exchange coupling between localized cu spins .
the temperature independent @xmath38 was assumed to include @xmath22-independent core and van vleck contributions to @xmath40 , and an @xmath22-dependent fermi liquid contribution that grew somewhat slower than linearly with increased doping @xmath22 .
sublattice susceptibility @xmath43 vs @xmath44 for five different samples of la@xmath30sr@xmath31cuo@xmath32 with effective doping levels @xmath45 , @xmath46 , @xmath47 , @xmath48 , and @xmath35 , compared with theoretical 2d heisenberg calculations and the weiss molecular - field ( mft ) prediction .
reproduced with permission from @xcite.,scaledwidth=75.0% ] to explain the scaling behavior of eq.([john1 ] ) , johnston@xcite introduced a doping - dependent 2d heisenberg exchange constant , @xmath49 , and the ratio @xmath50 where @xmath51 is the result obtained for the 2d heisenberg model for a given @xmath49 .
he found that @xmath52 gradually decreases from @xmath53 to @xmath54 ( fig.[johnstonr ] ) . and the ratio @xmath55 vs. hole doping @xmath56 .
@xmath57 is shown for comparison .
the solid curves are guides to the eye . reproduced with permission from @xcite.,scaledwidth=75.0% ]
oda _ _ et al.__@xcite confirmed johnston s scaling law for the bulk spin susceptibility in sr- and ba - doped la@xmath58cuo@xmath59 , and extended its applicability to the 2212 family of cuprates , bi@xmath58sr@xmath60ca@xmath61cu@xmath58o@xmath62 .
the experimental results for both families displayed excellent data collapse to the theoretical 2d heisenberg curve .
like johnston , oda _ et al .
_ found that the weight of the heisenberg - like contribution decreases with hole doping . to account for this decrease , oda _ et al .
_ introduced an effective magnetic moment , @xmath63 @xmath64 is defined in terms of @xmath28 and @xmath65 as @xmath66 they found that both @xmath28 and @xmath67 decrease linearly with doping , @xmath68 the magnitude of @xmath67 is that expected from the sum rule for a homogeneous material , @xmath69 but we caution the reader that homogeneity may not be present in the underdoped regime .
both the doping dependence of the @xmath22-dependent fermi liquid part and the temperature dependent spin liquid component were found by oda _
_ to be in good agreement with earlier results of johnston@xcite ; like johnston they found that the temperature - dependent scaling part of the bulk spin susceptibility disappears at some critical doping value ; according to eq.([odda ] ) , that is @xmath70 .
another early study of bulk susceptibility in la@xmath71sr@xmath72cuo@xmath59 was performed by yoshizaki _
@xcite , who give a plot of @xmath73 , without attempting johnston scaling . since @xmath74 does not match the value of the exchange coupling @xmath10 in the insulator , yoshizaki _ et al .
_ found a jump in @xmath28 at the mi boundary .
further confirmation of johnston scaling of the form eq.([john1 ] ) for the bulk spin susceptibility in the 2 - 1 - 4 family was obtained by nakano _
@xcite ( fig .
[ nakanofig ] ) , who demonstrated an excellent data collapse for a number of samples of la@xmath71sr@xmath72cuo@xmath59 , both in the underdoped region and that close to and beyond optimal doping .
_ arrived at their scaling law by assuming the presence of additional temperature - dependent terms : an impurity curie term @xmath75 and a linear term @xmath76 in the underdoped and overdoped regimes .
they did not attempt to fit the 2d heisenberg model calculations ; an empirical scaling data collapse was constructed instead , with results that were in agreement with oda _ _ et al.__@xcite . versus @xmath77 for the superconducting samples of la@xmath71sr@xmath72cuo@xmath59 .
the arrow shows the value of @xmath78 predicted by the 2d heisenberg model calculations .
the solid line is the universal curve .
the inset shows that the universal curve decreases more rapidly for @xmath79 than that obtained by johnston@xcite ( dashed line ) .
reproduced with permission from @xcite , scaledwidth=75.0% ] an alternative form of scaling for the bulk spin susceptibility was proposed by levin and quader@xcite .
similar to the studies reviewed above , they suggested separation of the bulk spin susceptibility in two components , the temperature - independent fermi liquid component and the scaling component .
however , in their model the scaling component originates from the contribution of a separate itinerant band .
levin and quader@xcite took the 2d density of states in the following form : @xmath80 where the index @xmath81 corresponds to a large hole band , which produces the usual fermi liquid term in the bulk spin susceptibility .
the chemical potential @xmath82 lies very close to the top of the second band @xmath83 and enters that band at @xmath84 .
the contribution of the band @xmath83 to the bulk spin susceptibility becomes temperature - dependent due to thermally activated carriers and is a universal function @xmath85 , where @xmath86 is a fermi liquid parameter .
the parameter @xmath87 is thus an analog of @xmath28 .
the two - band model produced a reasonably successful data fit for the bulk susceptibility in tlsr@xmath58(lu@xmath88ca@xmath72)cu@xmath58o@xmath89 for both the underdoped and the overdoped regimes . to summarize , the various independent measurements of the bulk susceptibility in the 2 - 1 - 4 and 1 - 2 - 3 materials can now be seen to be consistent with one another and with the picture first set forth by johnston : that in the underdoped regime , for dopings between @xmath90 and @xmath91 , and @xmath92 , one has two independent contributions to the bulk spin susceptibility .
one comes from a 2d heisenberg spin liquid with a doping - dependent effective interaction , @xmath93 ; the second represents a fermi liquid contribution whose strength increases as the doping level is increased .
as we shall see , the fall - off of the rescaled spin susceptibility below the heisenberg spin liquid value at @xmath94 appears to be a universal property of the spin liquid in the cuprates .
the knight shift seen in nmr experiments is a measurement of the bulk spin susceptibility at a particular nuclear site@xcite , so that measurements of the knight shift for different nuclei serve to supplement direct measurements of the bulk susceptibility .
most of the early analysis of the nmr data used a single - component mila - rice - shastry ( mrs ) description@xcite for which the justification was the observation by alloul _
_ et al.__@xcite and takigawa _ _ et al.__@xcite that the @xmath95 , @xmath96 and @xmath97 knight shifts in yba@xmath58cu@xmath98o@xmath99 and yba@xmath58cu@xmath98o@xmath100 have the same anomalous temperature dependence ( fig.[mptfig ] ) .
mrs proposed a hyperfine hamiltonian that described the coupling of a single magnetic component formed by the system of planar cu@xmath101 spins and holes mainly residing on the planar copper sites to the various nuclei .
most earlier knight shift experiments@xcite confirmed this one - component zhang - rice@xcite singlet picture , which is basically correct for the parent insulator .
however , as first noted by walstedt _ _ et al.__@xcite in connection with spin - lattice relaxation rate measurements , and discussed in detail below , the single component description turns out to lead to a number of contradictions in the doped materials , and requires modification .
cu@xmath98o@xmath100 plotted vs the temperature @xmath26 .
the values for the static planar susceptibility @xmath102 are given on the right - hand scale .
different knight shifts and the static spin susceptibility have the same temperature dependence . reproduced with permission from @xcite.,scaledwidth=75.0% ] cu@xmath59o@xmath62 versus @xmath103 , showing the applicability of johnston / nakano scaling and the relevance of the 2d heisenberg model@xcite .
the solid line is taken from ref .
@xcite , the dotted green line is the calculated susceptibility of the spin-1/2 2d heisenberg antiferromagnet@xcite and the data points are taken from ref.@xcite .
the arrow points to the zero - temperature prediction for the 2d heisenberg antiferromagnet . reproduced with permission from @xcite.,scaledwidth=75.0% ] just as was the case for the bulk spin susceptibility , the temperature - dependent knight shift data for different doping levels and different families of cuprates displays heisenberg model - type scaling .
the @xmath97 knight shift data from alloul _ _ et al.__@xcite for different doping levels in the 1 - 2 - 3 family has been scaled to a single curve with very good data collapse by wuyts _ _ et al.__@xcite , although no comparison to the 2d heisenberg model or the nakano _ _ et al.__@xcite bulk spin susceptibility scaling curve was provided .
recently curro _ _ et al.__@xcite and the authors@xcite have shown that the two are consistent ; as may be seen in fig.[currofig ] , the nmr knight shift follows very well the johnston - nakano bulk susceptibility scaling form and the heisenberg model for the 1 - 2 - 4 member of the 1 - 2 - 3 family .
spin liquid scaling behavior determined by @xmath28 has thus been shown to be universal in the cuprates .
what then is the direct experimental evidence from knight shift measurements for the presence of two distinct components ?
one material for which a contradiction with the simple one component model was found is yba@xmath58cu@xmath59o@xmath62 , where a more precise measurement@xcite of the copper knight shift data in magnetic fields parallel to the c - axis , @xmath104 , found that it displays a rather unusual temperature dependence , one that is different from that seen for the bulk spin liquid susceptibility .
a second example comes from quite recent measurements by haase _
_ et al.__@xcite of the planar and apical oxygen knight shifts for la@xmath105sr@xmath106cuo@xmath59 . as we discuss in a later section ,
these indicate that while the temperature - dependent part of the knight shift on different nuclei is indeed the same , there is a temperature - independent part of the knight shift above the superconducting temperature , @xmath15 , that is different for the planar copper and the planar and apical oxygen nuclei .
the key signature of this effect , the deviation of various knight shifts from the same temperature dependence at temperatures below @xmath15 , is clearly visible in fig.[mptfig ] .
these contradictions suggest that the original mrs hyperfine hamiltonian has to be modified to include these new effects , and that the one - component spin dynamics and the zhang - rice singlet picture@xcite fails at moderate doping levels@xcite , a topic to which we return in section 3 . in the mrs hamiltonian that is described in detail in the following section , wave vector dependent form factors arise from the presence of a transferred hyperfine interaction between the cu spins and the probe nucleus . for probe nuclei other than copper ,
these vanish at the commensurate wave vector @xmath107 , so that , for example , an in - plane oxygen nucleus will feel little of a spin response that is peaked at the commensurate wave vector .
( that this should be the case is obvious if one recalls that such oxygen are located midway between copper nuclei , and for an antiferromagnetic array of copper spins , the nearest neighbor spins would cancel one another out in their influence on the oxygen site . ) the striking difference of the temperature dependence and magnitude of the spin - lattice relaxation on the @xmath95 , @xmath96 , and @xmath97 nuclei led millis , monien , and pines ( mmp)@xcite to the conclusion that a localized or nearly localized spin component of the dynamic magnetic response function must be strongly peaked at the commensurate wave vector , as might be expected if one were close to an antiferromagnetic instability .
indeed , closer study shows that even a slight deviation from commensurability within the mmp approach based on the mrs one - component model will have a significant impact on oxygen relaxation rates that is not seen experimentally@xcite .
thus , an incommensurate peak structure for @xmath108 , such as has been inferred from inelastic neutron scattering ( ins ) experiments@xcite on the 2 - 1 - 4 materials , is inconsistent with this approach .
one way to get around this difficulty is to introduce additional transferred hyperfine interactions@xcite ; a second way , proposed by slichter@xcite , is to note that unlike nmr , ins is a global probe of spin excitations , so that a suitable domain structure ( regions of commensurate near - antiferromagnetic behavior , separated by domain walls ) would give rise to the apparent incommensuration inferred from the ins experiments .
we adopt this explanation in what follows . the spin fluctuation response function proposed by mmp
was that appropriate to any spin liquid near a commensurate antiferromagnetic instability : @xmath109 where the peak susceptibility takes the form , @xmath110 with @xmath81 as the magnetic correlation length , and @xmath111 as a temperature - independent constant .
the copper nmr @xmath112 and @xmath113 relaxation rates provide a direct measure of the strength and character of the spin liquid response function , measuring as they do the momentum - integrated imaginary and real part of the spin response function , @xmath114@xcite .
one finds : @xmath115 imai _ _ et al.__@xcite in an early scaling analyses ( fig .
[ imaifig ] ) of the copper relaxation rates for the 2 - 1 - 4 family demonstrated experimentally the existence of a universal high temperature limit for @xmath112 that is temperature- and doping- independent : @xmath116 which according to eq.([t1mmp ] ) means that at high temperatures the spin fluctuation energy , @xmath117 must be proportional to t. this universal high - temperature behavior of @xmath112 was explained in terms of the qc ( quantum critical)@xcite behavior seen in the sigma - model@xcite that , strictly speaking , is only applicable to the parent insulating compound .
indeed the empirical finding of a temperature- and doping independent @xmath112 in the 2 - 1 - 4 family of materials extends to temperatures much higher than those at which such a theoretical explanation would apply . measured by nqr for la@xmath71sr@xmath72cuo@xmath59 ( @xmath118 , @xmath119 , @xmath120 , @xmath121 ) .
the inset shows a semilogarithmic plot of @xmath122 ( in units of @xmath123 ) vs @xmath124 for the clean sample of la@xmath58cuo@xmath125 and for la@xmath58cuo@xmath126 .
the solid curve is the best fit to the theoretical prediction of 2d heisenberg model .
reproduced with permission from @xcite , scaledwidth=75.0% ] on temperature in la@xmath71sr@xmath72cuo@xmath59@xcite for different doping levels represents a set of parallel lines.,scaledwidth=75.0% ] two other high - temperature forms for the relaxation rates based on @xmath12 and @xmath127 dynamical scaling were therefore suggested@xcite on the basis of the non - linear sigma model@xcite and general scaling arguments , and verified experimentally . in @xmath12
scaling , @xmath117 and @xmath81 are related by @xmath128 so the presence of @xmath12 qc scaling in the underdoped materials leads to the simple results : @xmath129 on the other hand , for @xmath127 ( mean field ) scaling one has @xmath130 so that if it is present one finds @xmath131 the nmr experimental results of curro _ _ et al.__@xcite show that both forms of scaling are present in the material , yba@xmath58cu@xmath59o@xmath62 . at high temperatures , for @xmath132 , eq.([z2sc ] ) is valid and one has mean field behavior , while below @xmath13 the spin spectrum displays @xmath12 qc behavior down to a temperature @xmath133 , where the @xmath134 measurements suggest that a gap opens up in the spin liquid spectrum .
we return to this finding below .
the relaxation rates for other nuclei , such as oxygen or yttrium , also display anomalous ( i.e. non - korringa ) behavior , although in much milder form . a modified korringa - type scaling for oxygen @xmath135
was suggested by mmp : @xmath136 additional scaling forms have been proposed for the copper relaxation rates at lower temperatures . in particular , the authors@xcite have observed recently that @xmath137 reaches its universal high - temperature behavior with a different @xmath22-dependent offset ( fig.[kitaokafig ] ) , @xmath138 where the constant @xmath139 is universal , while @xmath140 changes linearly with doping , and its variation suggests the existence of a qc point in the spin liquid at @xmath141 , where @xmath142 .
eq.([prlcoll ] ) shows excellent data collapse for the 2 - 1 - 4 and the 1 - 2 - 3 families .
an alternative form of scaling has been suggested to apply at moderate to low temperatures , @xmath143 , by gorkov and teitelbaum ( gt)@xcite who found excellent data collapse ( fig.[gtscaling ] ) for the following scaling form : @xmath144 where @xmath145 is a universal function for all high - t@xmath146 materials , while @xmath147 varies with @xmath22 .
the suggested scaling form follows from the decomposition of relaxation rate into two different processes . the x - dependence of @xmath148(fig.[gtparam ] ) looks rather unusual@xcite , since the proposed empirical disorder - driven relaxation rate , @xmath148 decreases with increased doping @xmath22 .
the proposed gt scaling is assumed to be the result of intrinsic phase separation , and the pseudogap temperature @xmath149 is proposed as a measure of the onset of such behavior . for yba@xmath58cu@xmath98o@xmath100
overlaid with that for yba@xmath58cu@xmath59o@xmath62 and la@xmath105sr@xmath106cuo@xmath59 . in the lower right corner : the offset values for different compounds . reproduced with permission from @xcite , scaledwidth=75.0% ] vs sr content @xmath22 for lsco ( relative to that for yba@xmath58cu@xmath59o@xmath62 ) .
the line is a visual guide .
reproduced with permission from @xcite.,scaledwidth=75.0% ] if scaling is present , it is natural to look for the possible presence of finite size effects , as cho _ _
et al.__@xcite have done in the lightly doped la@xmath71sr@xmath72cuo@xmath59 material .
they found very good agreement with that expected for phase separation and domain formation on the insulating side of the mott transition .
however , there was no direct observation of the domain sizes of the different phases .
according to finite - size scaling theory@xcite , the nel temperature for a domain of size @xmath150 takes the form , @xmath151 where @xmath152 in mean field theory .
the experimental data can also be fit well by the expression @xmath153 with @xmath154 .
_ conclude that @xmath155 which means that the width of the domain wall is @xmath22 independent .
they found that in the antiferromagnetic regime the scaling law takes the form : @xmath156 here @xmath157 is the correlation length in the pure heisenberg model , @xmath158 inelastic neutron scattering ( ins ) experiments enable one to explore the extent to which 2d heisenberg model captures the momentum dependence and behavior at higher frequencies of the spin liquid as its properties are altered by doping .
there is by now a vast body of literature that includes several recent reviews@xcite .
our focus here is on the extent that the doped spin liquid continues to exhibit the quantum critical behavior inferred from the ultra low frequency dynamic properties measured in the nmr experiments described in the previous sections , and the ways in which departures from quantum critical behavior emerge as the frequency is increased into the multi - time range or one goes into the gapped normal or superconducting state .
as we shall see , one of the most striking features that emerges with doping is the observation of peaks whose positions at lower frequencies reflect a doping - dependent incommensuration or discommensuration that indicates dynamic stripe formation ; a second is the appearance of resonances and spin gaps in the normal state . throughout this section
we will be concerned , as we were in our discussion of nmr experiments , with possible universal behavior that is common to the 1 - 2 - 3 and 2 - 1 - 4 families .
perhaps the most direct evidence of quantum critical scaling was provided by the ins measurements of the local ( integrated ) dynamic spin susceptibility .
the remarkable results obtained in insulating and metallic low doped samples indicate the presence of the @xmath27 scaling expected in the vicinity of a @xmath12 quantum critical point . among early neutron scattering experiments that exhibit some form of scaling behavior
those of particular interest are on lightly doped yba@xmath58cu@xmath98o@xmath159 by birgeneau _
@xcite , lightly doped la@xmath71sr@xmath72cuo@xmath59 by keimer
_ et al . _
@xcite , and yba@xmath58cu@xmath98o@xmath160 by sternlieb__et al .
_ _ @xcite .
_ et al.__@xcite and birgeneau _ _ et al.__@xcite fit their experimental results for the local ( integrated ) spin susceptibility to the expression @xmath161 the data collapse is good ( fig.[keimerfig ] ) , but this form does not quite exhibit @xmath27 scaling because of the @xmath26-independent amplitude @xmath162 . .
the solid line is the function @xmath163 $ ] with @xmath164 and @xmath165 .
reproduced with permission from @xcite.,scaledwidth=75.0% ] sternlieb _ et al . _
@xcite subsequently found that true @xmath27 scaling exists for the local spin susceptibility in the underdoped material yba@xmath58cu@xmath98o@xmath160 .
their experimental results ( fig .
[ sternliebfig ] ) can be fit to the following simple scaling form : @xmath166 with a deviation from scaling that occurs at progressively lower temperatures , as the frequency @xmath167 increases
. for a fixed energy transfers converges to a universal function of @xmath168 at high temperatures .
the arrows show the temperature above which the scaling behavior occurs .
( bottom ) the corresponding scattering function @xmath169 vs @xmath168
. reproduced with permission from @xcite.,scaledwidth=75.0% ] in more recent experiments , bao _ _ et al.__@xcite studied experimentally the destruction of 2d antiferromagnetic order in @xmath170-doped la@xmath58cuo@xmath59 .
since holes are loosely bound by @xmath170 impurities , the doped material remains an insulator even when long - range order is destroyed at @xmath171 .
however , unlike the superconducting la@xmath71(sr , ba)@xmath72cuo@xmath59 , there are no additional complications due to the presence of mobile doped holes , making possible a cleaner measurement of the spin dynamics .
bao _ _ et al.__@xcite found a commensurate energy spectrum of spin excitations at @xmath171 , where the long - range order is not present , and a characteristic quantum critical @xmath27 scaling for the local spin susceptibility ( fig .
[ baosc ] ) .
the scaling function , however , is different from that obtained by sternlieb__et .
al.__@xcite in particular , bao _ _
et al.__@xcite found significant deviations of the scaling function from linearity , with @xmath172 and @xmath173 below @xmath174 this scaling changes to one with a constant energy scale @xmath175 , @xmath176 where @xmath177 scaling is valid for la@xmath58cu@xmath178li@xmath179o@xmath59 in the high temperature qc regime the solid line is the scaling function , eq.([baoscf ] ) .
( b ) @xmath27 scaling becomes invalid in the low temperature regime .
( c ) a new scaling for the low temperature regime , with a constant energy scale @xmath180 . the solid line is the scaling function eq.([baoscqd ] ) . reproduced with permission from @xcite.,scaledwidth=75.0% ] stock _ _ et al.__@xcite recently confirmed the low - frequency @xmath27 scaling for @xmath181 in oxygen ordered ortho - ii yba@xmath58cu@xmath98o@xmath182 superconductor found earlier in this material in the oxygen - disordered state by birgeneau _ _ et al.__@xcite .
_ fit their expression to the form eq.([keimsc ] ) , with only a linear term in @xmath27 present .
they find that the @xmath27 scaling breaks down at higher frequencies , @xmath183 , since the amplitude @xmath162 in eq.([keimsc ] ) becomes temperature - dependent .
keimer _ _ et al.__@xcite find evidence for finite size scaling , since a good fit to their experimental results can be obtained with the following expression for the correlation length : @xmath184 with @xmath185 , where @xmath150 is the size of the domain for finite size scaling .
eq.([necore ] ) was later generalized to finite frequencies @xmath167 by aeppli _
_ et al.__@xcite , who fit their results for the correlation length in la@xmath186sr@xmath187cuo@xmath59 ( fig.[aepplisc ] ) to the following form : @xmath188 at various fixed energy transfers plotted against @xmath189 .
the solid line corresponds to a @xmath12 qcp .
the graph in the upper right shows how the peak response depends on @xmath190 .
the inset in the upper left shows the 3d space defined by @xmath191 phase space probed by the @xmath192 sample , and the solid circle represents the nearby qcp .
reproduced with permission from @xcite.,scaledwidth=75.0% ] inelastic neutron scattering experiments provide a direct measurement of the spectrum of spin excitations , and thus of the spin wave velocity and exchange couplings in the antiferromagnetic insulator@xcite .
we shall see that the detailed measurements of the spectrum of the spin excitations in underdoped high - temperature superconductors that have recently been carried out for both the 2 - 1 - 4@xcite and the 1 - 2 - 3@xcite families provide evidence for its universality .
[ [ incommensurationdiscommensuration ] ] incommensuration / discommensuration + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + early neutron measurements on both the 2 - 1 - 4 and the 1 - 2 - 3 families of high - temperature superconductors revealed an incommensurate spectrum of spin excitations at low frequencies .
cheong _ _ et al.__@xcite and mason _ _ et al.__@xcite discovered in their inelastic neutron scattering measurements on the 2 - 1 - 4 family of materials that the position of the low - energy spin fluctuation peak shifts from @xmath193 in the parent insulating compound to @xmath194 and @xmath195 in the underdoped superconductor .
they found that the incommensurability @xmath196 in the underdoped regime grows approximately linearly with increased hole concentration @xmath22 , @xmath197 , at least for @xmath198 .
for @xmath199 , the incommensurability @xmath200 saturates at a finite value .
incommensurate peaks in the spin response function @xmath201 at low frequencies were also discovered in the bilayer system yba@xmath58cu@xmath98o@xmath159@xcite .
as mook _ _ et al.__@xcite have first shown , the incommensurate peaks in the bilayer system are located at the same positions as in the single - layer la@xmath71sr@xmath72cuo@xmath59 system at similar level of hole doping .
thus , they concluded that the band structure , which is very different in these two families of materials , plays a minor role in the spectrum of spin excitations at low frequencies . in what follows
we will refer to these peaks as reflecting incommensurate behavior , although , as noted earlier , for these results to be consistent with nmr , these should rather be regarded as reflecting discommensuration , or dynamic stripe order .
wakimoto _ _ et al.__@xcite and matsuda _ _ et al.__@xcite extended the measurements of @xmath201 in la@xmath71sr@xmath72cuo@xmath59 to the spin glass regime , @xmath202 .
they found that while the low - energy spin excitations remain incommensurate with the same @xmath197 as in the metallic phase , the positions of the incommensurate peaks are rotated by @xmath203 relative to those in the metal , as shown in fig.[matsudafig ] .
unlike in metallic phase , the diagonal stripe structure for the insulating spin glass phase is seen as bragg peaks in elastic scattering measurements and thus is * static * at low temperatures@xcite .
moreover , the incommensurability of the spin fluctuations disappears at higher frequencies or temperatures , above a certain energy threshold for the incommensurate structure , which we shall see is directly related to the spin gap .
an interesting linear scaling relationship between the incommensurability @xmath200 and the superconducting temperature , @xmath204 , was suggested by yamada _
_ et al.__@xcite . which , they find , holds extremely well for the 2 - 1 - 4 family of materials in the underdoped regime , @xmath205 , with @xmath206 .
balatsky and bourges@xcite and dai _ _ et al.__@xcite found that yamada scaling is also applicable to the 1 - 2 - 3 family of materials , with @xmath207 and @xmath208@xcite .
since the maximum of @xmath15 is different in different families of cuprates , while the incommensurability is universal and depends only on hole doping , @xmath209 is different for different families of materials .
dependence of the splitting of the incommensurate peaks @xmath210 in la@xmath71sr@xmath72cuo@xmath59 in tetragonal reciprocal lattice units .
open circles indicate the data for the inelastic incommensurate peaks reported by yamada _
_ et al.__@xcite .
filled circles and square are the data for elastic incommensurate peaks reported by wakimoto _
_ et al.__@xcite .
the broken and solid lines correspond to @xmath211 .
the insets show the configuration of incommensurate peaks in the insulating phase ( diagonal stripe ) and the superconducting phase ( collinear stripe ) reproduced with permission from @xcite.,scaledwidth=75.0% ] stock _ _ et al.__@xcite considered the applicability of yamada scaling to both 2 - 1 - 4 and 1 - 2 - 3 families of high - temperature superconductors .
they plotted the incommensurability @xmath196 as a function of @xmath212 , a quantity proportional to the number of holes in the plane in the underdoped regime ( see fig.[stockfiga ] ) .
their plot thus confirms for the 2 - 1 - 4 materials the conclusion of mook _
_ et al.__@xcite that the incommensurability of low - energy spin fluctuations in cuprate superconductors depends only on hole concentration in the plane , and not a particular material or band structure . as it relates to the incommensurability @xmath196 in yba@xmath58cu@xmath98o@xmath159 and la@xmath71sr@xmath72cuo@xmath59 .
reproduced with permission from @xcite.,scaledwidth=75.0% ] [ [ high - energy - spin - excitations - and - the - resonance - peak ] ] high - energy spin excitations and the resonance peak + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + early ins studies of high - frequency spin fluctuations were limited due to poor resolution .
many of these measurements were initially done on the 1 - 2 - 3 family of materials , where the resonance peak@xcite enhances spin response at higher energies .
the resonance peak in the 1 - 2 - 3 family of materials corresponds to a sharp enhancement of the intensity of commensurate spin excitations at high frequencies near the frequency , @xmath213 , that depends on hole doping @xmath22 .
the observation then raised the question of whether the peak was a consequence of spin gap in the fermionic excitations or was an intrinsic property of the spin liquid in the vicinity of @xmath193 . at first the resonance
was only observed in the superconducting state .
since the resonance peak was not observed in the inelastic neutron scattering experiments on the 2 - 1 - 4 family , this led to an explanation that this feature most likely arises as a result of the coupling of spin excitations to fermions with a d - wave gap or d - wave like gap in the normal state .
indeed , dai _ _ et al.__@xcite found that the resonance peak first appears above @xmath15 at @xmath149 , which decreases with increased hole doping , the resonance energy @xmath213 increases with increased doping and tracks the doping dependence of @xmath204 , as shown in fig.[daireson ] .
however , more detailed neutron scattering studies revealed that the resonance corresponds to a special frequency at which high - energy spin excitation spectrum becomes commensurate .
as we have seen above , the spin excitation spectrum depends only on hole doping , and not on a particular material , a conclusion that casts doubt on the quasiparticle gap explanation of the resonance .
inelastic neutron scattering studies near the resonance frequency were first done by bourges _ _ et al.__@xcite in the yba@xmath58cu@xmath98o@xmath182 material .
_ found that the commensurate resonance peak was broadening in momentum , both above and below the resonance frequency @xmath214 .
based on their findings , bourges__et al .
_ _ suggested that the spin excitations disperse at high frequencies in a way that is similar to that expected for spin waves in the parent insulating compound .
_ et al.__@xcite later studied the momentum dispersion of spin excitations near the resonance peak in yba@xmath58cu@xmath98o@xmath215 in more detail .
they found evidence for two modes that meet near the resonance energy , one that opens downwards and gives rise to incommensurate spin excitations at low energies , the other , a new mode that opens upwards in energy , giving rise to spin wave - like dispersive excitations at high energies .
_ found that the two modes meet at the frequency @xmath216 , slightly above the characteristic resonance frequency @xmath217 for their material .
more recent ins experiments on yba@xmath58cu@xmath98o@xmath182@xcite , yba@xmath58cu@xmath98o@xmath160@xcite and yba@xmath58cu@xmath98o@xmath218@xcite , however , suggest that the high energy mode and the low energy mode meet at the resonance frequency . and the resonance energy @xmath219 as a function of @xmath15 in yba@xmath58cu@xmath98o@xmath159
. the open circles and open squares are temperatures at which @xmath220 reaches broad maximum .
the open diamonds show the pseudogap temperature @xmath14 determined from nmr measurements@xcite .
the filled circles correspond to @xmath15 and @xmath14 , where the resonance first appears in ins measurements .
filled squares are from fong _ _ et al.__@xcite .
horizontal error bars are superconducting transition widths .
the solid lines are guides to the eye . reproduced with permission from @xcite.,scaledwidth=75.0% ]
cu@xmath98o@xmath182 with respect to commensurate @xmath193 position at @xmath221 .
the filled circles are from two gaussian fits along the @xmath222 $ ] and @xmath223 $ ] directions above the resonance energy and from @xmath222 $ ] direction below the resonance energy .
the open circles represent the positions of the incommensurate peaks found in experiments conducted at chalk river .
the open squares represent the peak position that results from a polar average around @xmath224 position .
the solid lines schematically represent the dispersion of the insulating compound as measured by hayden _
_ et al.__@xcite .
the dashed lines are fits of the high - energy dispersion to linear spin wave theory for @xmath225 .
the horizontal bars show the q - width observed for three optic scans at constant energy . reproduced with permission from @xcite.,scaledwidth=50.0% ] a typical energy spectrum of the spin excitations as measured by stock _ _ et al.__@xcite in a detwinned ortho - ii sample of yba@xmath58cu@xmath98o@xmath182 is shown in fig.[stockfigc ] .
_ et al.__@xcite had found earlier for a detwinned sample of yba@xmath58cu@xmath98o@xmath160 , stock _ et al .
_ found two , rather than four incommensurate peaks below the resonance frequency @xmath226 .
this result strongly suggested a one - dimensional character for the spin fluctuations , consistent with the formation of dynamic stripes@xcite .
the incommensurability @xmath196 observed at low frequencies decreased with increased energy transfer and disappeared at @xmath227 .
the spin excitations started to disperse outward again at high frequencies , @xmath228 .
stock _ _ et al.__@xcite found the dispersion for acoustic spin excitations to be isotropic and similar to the spin waves in the insulator , as shown in fig.[stockfigb ] ; they fit their results to a gapped spectrum , @xmath229 with @xmath230 and a spin wave velocity @xmath231 , which is close to the value @xmath232 obtained from the slope of high energy spin excitations .
the spin wave velocity is dramatically reduced from its value in the insulator , @xmath233@xcite .
stock _ et al .
_ also measured the width of the ring of high energy spin wave excitations and found that it increased linearly with frequency at low temperatures .
the low - energy spectral weight was dominated by the resonance peak at @xmath226 , which had an asymmetric shape , with a quick drop - off of intensity for @xmath228 and a much slower reduction at @xmath234 .
stock _ _ et al.__@xcite , however , estimated that the resonance represented only 3 % of the total spectral weight of spin excitations ; most of its spectral weight would appear only at higher energies .
similar to spin waves in the insulating compound , @xmath235 approached a constant at high energies @xmath228 , @xmath236 , somewhat reduced from a similar value for the insulator , @xmath237@xcite .
a very similar dispersion for spin excitations was found by reznik__et al.__@xcite in a twinned sample of yba@xmath58cu@xmath98o@xmath218 .
they observed four incommensurate peaks dispersing inward with increased frequency at frequencies below the resonance frequency @xmath238 and becoming commensurate at the resonance frequency .
the dispersion of spin excitations was cut off below the resonance frequency at a characteristic spin gap frequency @xmath239 .
_ concluded that their data was consistent with the isotropic gapped spin wave - like excitations dispersing outward at frequencies above the resonance frequency .
as had stock _
et al . _ ,
reznik _ et al .
_ found the @xmath240-integrated intensity to be approximately constant at high frequencies @xmath228 , as expected for spin wave - like excitations . because of resolution problems , reznik _ et al .
_ were not able to calibrate their data in absolute units , or estimate the spin wave velocity .
not all ins studies have found spin wave - like excitations with a reduced effective exchange coupling @xmath10 at high energies .
for example , hayden _ _ et al.__@xcite performed measurements in a twinned sample of yba@xmath58cu@xmath98o@xmath160 with effective hole doping @xmath241 close to the @xmath242 stripe ordering instability . as had other groups , hayden _ et al .
_ observed incommensurate spin excitations that dispersed inward up to resonance frequency @xmath243 .
however , their measurements at frequencies above @xmath214 indicated a strikingly different picture . instead of the gapped spin waves found by other groups , hayden _ et al . _ resolved four peaks at @xmath228 rotated by 45 degrees that dispersed outward in energy .
their measured dispersion for spin excitations is consistent with that observed earlier in stripe - ordered materials , such as la@xmath71ni@xmath72cuo@xmath59@xcite and la@xmath244ba@xmath245cuo@xmath59@xcite .
hayden _ _ et al.__@xcite found little dispersion of the rotated incommensurate peaks at high energies , @xmath246 .
this led them to conclude that the high - energy spectrum of spin excitations they measured was inconsistent with commensurate gapped spin wave spectrum observed by other groups . and @xmath247 energy transfers and integrated @xmath248 along the energy axis in yba@xmath58cu@xmath98o@xmath182 .
at @xmath247 the ring of scattering shows the intersection of the cone of spin wave dispersion emanating from the @xmath249 position with a constant energy surface . within statistics
the velocity is isotropic .
the intensity is represented by false color . reproduced with permission from @xcite.,scaledwidth=50.0% ] , open squares : @xmath250 , filled squares : @xmath251 , filled circles : @xmath252 , open circles : @xmath253 .
`` peak - dip - hump '' structure in the integrated susceptibility @xmath254 suggesting that the magnetic response has two components .
the inverse correlation length @xmath255 shows significant broadening at high energies .
reproduced with permission from @xcite.,scaledwidth=50.0% ] high energy spin excitations have also been studied in several materials belonging to the 2 - 1 - 4 family .
the question of the renormalization of the effective exchange coupling @xmath10 and the spin wave velocity @xmath256 of the insulator by doped carriers was first investigated by hayden _ _ et al.__@xcite for la@xmath186sr@xmath187cuo@xmath59 in high - energy transfer inelastic neutron scattering experiments .
they found that the spectrum of spin excitations at high energies fits the heisenberg model well , with an effective exchange coupling @xmath10 that is only mildly reduced from its value in the insulator : for this material , @xmath257 , @xmath258 , values that are somewhat reduced from the corresponding values in the insulator , @xmath259 , @xmath260 .
they concluded that quantum fluctuations increase , but that at the high energies they study , @xmath10 and @xmath256 do not get strongly renormalized by doped holes .
however , the spectral weight of their measured high - energy spin excitation decreases very strongly from its value at @xmath118 ; the peak of that weight for @xmath261 was found@xcite to be at @xmath262 , significantly below the corresponding peak for the insulating compound .
the details of the spin excitation spectrum at low energies , however , could not be resolved .
more recent high resolution inelastic neutron scattering studies of the stripe - ordered compound , la@xmath244ba@xmath245cuo@xmath59@xcite , and the optimally doped superconductor , la@xmath263sr@xmath264cuo@xmath59@xcite , reveal the details of the spin excitation spectrum at low energies . in their measurements on stripe - ordered sample of la@xmath244ba@xmath245cuo@xmath59 tranquada _
_ et al.__@xcite found two different branches of spin excitations that meet at a characteristic frequency @xmath265 , an energy spectrum that is similar to that found by other groups , and a sharp feature in the spectrum of spin excitations at @xmath266 .
their results are consistent with the behavior seen earlier in other stripe - ordered compounds , such as la@xmath71ni@xmath72cuo@xmath59 , where the spectrum of spin excitations was measured earlier@xcite . in particular , as found by hayden _ _ et al.__@xcite and expected on a stripe model@xcite , tranquada _
_ found four incommensurate peaks rotated by 45 degrees that disperse outward at high energies @xmath267 .
they fit their data to a model of charge stripes separated by 2-leg heisenberg ladders with @xmath268 , a value somewhat reduced from its value @xmath269@xcite in the insulator .
recent high - resolution neutron scattering studies by christensen__et al.__@xcite and vignolle _
_ et al.__@xcite of la@xmath263sr@xmath264cuo@xmath59 provide further important details on the universality of the spin wave spectrum in high - temperature superconductors .
_ et al.__@xcite studied the spin excitation spectrum of this material at low frequencies @xmath270 and concluded that the spectrum of incommensurate spin excitations is dispersive inwards and maps onto the spectrum of acoustic spin excitations observed in yba@xmath58cu@xmath98o@xmath271 , a material with similar planar hole doping . using this analogy , they concluded that spin excitations in la@xmath263sr@xmath264cuo@xmath59 should become commensurate at a frequency @xmath272 .
however , they did not see an associated resonance peak .
_ et al.__@xcite recently extended these studies to higher frequencies ; they found the spectrum of spin excitations shown in fig.[vignsp ] , with a high - frequency component emerging for @xmath225 . both @xmath10 and @xmath256
do get strongly renormalized with doping : the high - energy excitation spectrum they observe corresponds to commensurate gapped spin waves with an effective coupling @xmath273 , a value that is dramatically reduced from its value @xmath274 in the insulator .
below @xmath275 they find the usual@xcite attenuated incommensurate structure .
their analysis of the unusual double peak structure in the spectral weight suggests the spin excitation spectrum separates into two components - incommensurate attenuated spin excitations for @xmath276 , and commensurate gapped spin waves for @xmath225 .
they find that the spectral weight at high energies , @xmath277 is roughly 1/3 of the spectral weight observed in the parent insulating compound . in summary ,
the energy spectrum of spin excitations as probed by inelastic neutron scattering is approximately universal , i.e. , for different materials , it reflects only the hole doping @xmath22 .
in particular , the positions of incommensurate peaks at low frequencies are the same in both 2 - 1 - 4 and 1 - 2 - 3 families of materials at similar hole doping levels .
the spectrum of spin excitations at all frequencies also turns out to be the same at a similar doping level , at least for the yba@xmath58cu@xmath98o@xmath271 and la@xmath263sr@xmath264cuo@xmath59 pair .
the energy spectrum at high energies is consistent overall with what is expected for gapped spin waves , and measurements on different materials reveal a suppression of the effective exchange constant @xmath2 and the spin wave velocity @xmath278 with increased doping , as well as a suppression of the spectral weight of high - energy spin excitations .
however , experiments in materials with the hole doping close to @xmath279 found a rotated peak structure at high energies , similar to that expected from static stripe ordering .
overall , the energy spectrum of spin excitations in different families of materials is consistent with the picture of phase separation and fluctuating stripe order that first appears below @xmath14 .
the specific heat measurements and electronic entropy analysis of loram _ _ et al.__@xcite provide another important constraint on the normal state behavior of underdoped yba@xmath58cu@xmath98o@xmath280 . by fitting their data to fermions that are assumed to have a `` normal state '' energy gap of the bcs d - wave form , loram _ _ et al.__@xcite found that at @xmath281 , well above the superconducting temperature , @xmath15 , their hypothesized gap , @xmath282 , increases linearly as the doping is decreased , from being @xmath283 zero in a near optimally - doped sample ( y = 1 ) to 200k in an underdoped sample with y = 0.7 .
it is clear from their entropy data that their proposed gap size is related to @xmath28 and @xmath2 , that is : @xmath284 in a subsequent paper , loram _ _ et al.__@xcite argued that @xmath285 and the @xmath286 knight shift @xmath287 are proportional to each other , with a wilson ratio for nearly free electrons , @xmath288 the peculiar @xmath26-linear behavior of the electrical resistivity@xcite that has been observed near optimal doping in all families of cuprate superconductors , represents one of their main unresolved puzzles .
it has been explained as representing marginal fermi liquid behavior arising from the proximity of optimally doped materials to a quantum critical point@xcite , with concomitant @xmath27 scaling@xcite .
however , in the underdoped materials , the electrical resistivity shows strong deviations from linearity in @xmath26 , below a temperature @xmath289 .
moreover , scaling with a characteristic temperature @xmath13 has been clearly seen in other transport measurements , with the best data collapse being that observed in the hall measurements of hwang _
_ et al.__@xcite , who use the following scaling form : @xmath290 where @xmath291 is a universal function and @xmath292 and @xmath293 are doping - dependent functions(fig .
[ hwangfig ] ) . ) for la@xmath71sr@xmath72cuo@xmath59 with @xmath294 plotted rescaled as @xmath295/r_h^*$ ] vs @xmath296 .
inset : the parameters @xmath297 and @xmath298 vs sr composition @xmath22 .
the temperature @xmath14 agrees with @xmath13 obtained from the maximum in the bulk spin susceptibility .
reproduced with permission from @xcite , scaledwidth=75.0% ] the hall data for yba@xmath58cu@xmath98o@xmath299 from ito _ _ et al.__@xcite and carrington _ _
et al.__@xcite is analyzed in wuyts _
_ et al.__@xcite , who find the following scaling forms ( figs .
[ wanfig ] , [ wnfig ] ) for the hall angle , @xmath300 , and the number , @xmath301 , lead to excellent data collapse : @xmath302 @xmath303 wuyts _ _ et al.__@xcite extended the @xmath26-linear scaling of resistivity near optimal doping to the underdoped regime .
their analysis of ito _
et al._s@xcite resistivity data for the yba@xmath58cu@xmath98o@xmath299 family shows that for temperatures @xmath26 above @xmath28 the resistivity is universal and linear ( fig .
[ wrfig ] ) , and is similar to the optimally doped samples : @xmath304 the doping dependence of @xmath28 inferred from these scaling forms for the transport coefficients is consistent with its direct magnetic measurements .
divided by @xmath305 vs square of reduced temperature @xmath306 for 15 sets of thin film data together with reported data for oxygen - deficient and co - doped ybco single crystals .
@xmath307 agrees with the temperature obtained by rescaling @xmath286 knight shift . reproduced with permission from @xcite.,scaledwidth=75.0% ]
divided by @xmath308 vs reduced temperature @xmath309 for 15 sets of thin film data together with reported data for oxygen - deficient and co - doped ybco single crystals .
@xmath307 agrees with the temperature obtained by rescaling @xmath286 knight shift . reproduced with permission from @xcite.,scaledwidth=75.0% ]
vs scaled temperature @xmath309 for 15 sets of thin film data together with reported data for oxygen - deficient and co - doped ybco single crystals .
@xmath307 agrees with the temperature obtained by rescaling @xmath286 knight shift . reproduced with permission from @xcite.,scaledwidth=75.0% ]
levin and quader analyzed scaling of the hall and transport@xcite data based on the two - band model eq.([eqlq ] ) with degenerate ( @xmath81 ) and nondegenerate ( @xmath83 ) carriers .
they found that the @xmath310 dependence of the cotangent of the hall angle and the deviations from it can be fit to their model if two lifetimes are introduced for two different bands , @xmath311 and @xmath312 .
their two - band model predicts three - parameter scaling for the planar resistivity in the form : @xmath313 where @xmath314 is the energy scale analogous to @xmath14 .
a more recent alternative analysis@xcite of the high - temperature hall data@xcite for the underdoped la@xmath71sr@xmath72cuo@xmath59 family has its basis on the idea of phase separation .
it indicates that the hall data can be understood if one assumes an activated form for the carrier concentration , @xmath315 with a doping - dependent @xmath316 that stays linear in @xmath22 up to @xmath317 , above which strong deviations from linearity are observed .
the behavior of @xmath318 was found to be consistent with that measured in photoemission experiments . according to ref.@xcite
, there is crossover behavior when the number of activated carriers becomes approximately equal to the number of doped carriers @xmath316 ; this occurs at a temperature @xmath319 that is consistent with characteristic temperatures inferred from other measurements . at a proposed candidate qcp , @xmath320 ,
the energy gap for activated carriers , @xmath318 , goes to zero@xcite .
one of the earliest universal relations that emerged in the high-@xmath15 field is the linear uemura relation@xcite between the superfluid density @xmath321 , provided by penetration depth measurements , and the superconducting temperature , @xmath15 , valid in the underdoped regime .
this relation provides an important constraint on the number of electrons that become superconducting .
recently@xcite it has been shown ( fig .
[ uemurafig ] ) that a modified uemura scaling relation , @xmath322 holds for all cuprate superconductors . ) versus the product of d.c .
conductivity ( @xmath323 ) and the superconducting transition temperature ( @xmath15 ) for various cuprate superconductors .
@xmath323 is measured just above the transition . within error ,
all of the data fall on the same universal ( dashed ) line with slope of unity , defined by eq.([uem ] ) ; the dotted lines are from @xmath324 reproduced with permission from @xcite.,scaledwidth=75.0% ] enormous progress in resolution has been made in angle - resolved photoemission spectroscopy ( arpes ) experiments in the past decade so that arpes has now emerged as one of the best experimental probes of the cuprates@xcite .
one of the most important contributions of arpes was the detection of the anisotropic normal state gap in the fermionic spectrum , first by marshall
_ _ et al.__@xcite , loeser _ _ et al.__@xcite , and then by ding _ _ et al.__@xcite in bi@xmath58sr@xmath58cacu@xmath58o@xmath325 . by examining their data ,
marshall _ et al .
_ concluded that the normal state gap has two energy scales .
they identified the low energy scale gap ( @xmath326 ) by the location of the leading - edge midpoint .
this is a clear gap in the fermionic spectrum that has a @xmath17-wave - like momentum dependence , with gapless fermi surface arcs near the nodal regions .
the high energy scale ( @xmath327 ) gap appears as a broad incoherent feature in the spectrum near the @xmath328 point , where the low - energy spectral weight is strongly suppressed .
the low energy leading - edge gap first appears below @xmath149 , a temperature that has the same doping dependence as @xmath329 observed in other measurements .
however , @xmath14 inferred from arpes measurements@xcite in bi@xmath58sr@xmath58cacu@xmath58o@xmath325 turns out to be significantly lower than @xmath13 , @xmath330 .
both the measured magnitude and the @xmath331-dependence of the normal state leading edge gap are very similar to that of the d - wave gap in the superconducting state ; the normal state gap smoothly evolves into a superconducting gap below @xmath15 .
the doping dependence of the normal state d - wave - like gap amplitude @xmath318 tracks that of @xmath149@xcite .
the ratio of @xmath149 and @xmath28 varies for different families of materials .
detailed arpes investigations of the la@xmath71sr@xmath72cuo@xmath59@xcite and other families of cuprate superconductors@xcite find the same general form of the spectrum and the same linear doping dependence of the energy gap amplitude @xmath332 as that found in the bi@xmath58sr@xmath58cacu@xmath58o@xmath325 family of materials . following these and other arpes results , damascelli _ _ et al.__@xcite suggest that the value of @xmath149 and @xmath318 is determined by the maximum value of @xmath15 for a given material , @xmath333 , rather than exchange coupling @xmath10 : @xmath334 the high energy incoherent feature of the arpes spectra also has a d - wave like dispersion@xcite .
it is often claimed to be a remnant of the antiferromagnetic insulator@xcite , since it exhibits the same dispersion@xcite along the @xmath335-@xmath328 and @xmath224 - @xmath328 directions , which are only equivalent in the reduced antiferromagnetic brillouin zone , and is similar to the arpes spectra observed in undoped antiferromagnetic insulators .
the fact that the normal state gap closes non - uniformly in momentum space with increased temperature or doping has been well established for some time@xcite .
recently kanigel _ _ et al.__@xcite investigated the temperature- and doping- dependence of the low energy normal state gap in several samples of bi2212 and discovered a remarkable scaling relation : the anisotropy of the normal state gap and the length of the gapless fermi arcs near the nodal region for these different samples depend only on a single parameter , a reduced temperature @xmath336 ( fig . [ campscal ] ) .
et al . _ also found that the length of gapless fermi arcs near the nodes above the superconducting temperature @xmath15 is linear in @xmath337 , while below @xmath15 the fermi arcs develop the usual d - wave superconducting gap with point nodes .
$ ] , where @xmath338 is the symmetrized intensity at the gap energy at the fermi point labeled by angle @xmath339 , and @xmath340 is that at the fermi energy at the same point .
different symbols correspond to different @xmath337 .
the gray area represents the straight section of the fermi surface ( b ) variation of the arc length with respect to reduced temperature @xmath296 . on the y axis ,
0 % is the node and 100 % is the antinode . reproduced with permission from @xcite.,title="fig:",scaledwidth=75.0% ] $ ] , where @xmath338 is the symmetrized intensity at the gap energy at the fermi point labeled by angle @xmath339 , and @xmath340 is that at the fermi energy at the same point .
different symbols correspond to different @xmath337 .
the gray area represents the straight section of the fermi surface ( b ) variation of the arc length with respect to reduced temperature @xmath296 . on the y axis ,
0 % is the node and 100 % is the antinode . reproduced with permission from @xcite.,title="fig:",scaledwidth=75.0% ] is the superconducting gap that develops at @xmath15 in the nodal fermi arc region the same as the normal state gap , or is it a different gap ?
recently millis@xcite argued that this important question has not been resolved for years , since different experimental studies yield conflicting results .
an experimental observation of two different gaps in the underdoped regime will confirm that the normal state gap in the underdoped cuprates is driven by a different , competing nonsuperconducting order , not pairing fluctuations@xcite .
since the superconducting gap in the nodal region is extremely small , answering this question in arpes requires a very high resolution study .
so far arpes studies have proved to be conflicting as well@xcite .
for example , a recent leading edge gap arpes study of tanaka _ _ et al.__@xcite on three different samples of bi2212 found that the superconducting gap in the nodal arc region had a doping dependence different from that of @xmath149 , and scaled with @xmath204 instead . on the other hand , valla _ _ et al.__@xcite reported that the energy gap in a non - superconducting sample of la@xmath244ba@xmath341cuo@xmath59 , where the superconducting order is suppressed by charge ordering , has essentially the same simple d - wave form and magnitude as in the superconducting sample at higher doping concentration , thus yielding crucial support to the single gap scenario .
all stm studies find strong nanoscale inhomogeneity , in which gap magnitudes are observed to vary strongly on the nanometer length scale , while the recent stm measurements of alldredge _ _ et al.__@xcite , gomez _ _ et al.__@xcite , and boyer _ _ et al.__@xcite provide detailed evidence on gap formation in the underdoped 2212 and 2201 families of cuprates . according to gomez@xcite , in the overdoped regime locally
there is only one gap that has a d - wave symmetry .
the distribution of the energy gaps is , nevertheless , very inhomogeneous , indicating the presence of disorder .
as the temperature is raised above @xmath15 , the total spatial area of ungapped regions increases until the whole sample becomes ungapped .
the local temperature @xmath342 at which the normal state gap first appears varies strongly in space . nevertheless , the ratio of the local gap maximum @xmath343 in @xmath331-space to the local ordering temperature @xmath342 turns out to be universal for different spatial regions and very large@xcite , @xmath344 , much greater than the d - wave bcs limit of @xmath345 , indicating strong coupling . in the underdoped regime ,
gomez _ et al .
_ first see the spatially inhomogeneous formation of the local gaps below @xmath14 , defined as the temperature at which 90 % of the sample becomes gapped .
below @xmath15 , gomez _ et al .
_ observe the formation of a smaller kink inside the larger local gap , corresponding to the onset of superconductivity .
they conclude that in the underdoped regime there are two energy scales , with the lower energy scale corresponding to the onset of phase coherence .
boyer _ _ et al.__@xcite adopted a different approach in their analysis of their stm measurements on an overdoped sample , a member of bi2201 family , ( bi@xmath346pb@xmath89)@xmath58sr@xmath58cuo@xmath159 ( t@xmath347 ) . since inhomogeneity is unaffected by the onset of superconductivity at @xmath15 , boyer _
look for the signature of superconductivity , the emergence of the subgap kink at t@xmath146 , by removing the effective background of the high - temperature stm spectra , which is largely unaffected by the onset of superconductivity , from the low - temperature spectra . as a result of this subtraction
, they find a second , homogeneous superconducting gap , @xmath348 that forms at @xmath349 .
their analysis thus strongly suggests that the subgap kink at @xmath15 , also seen by gomez _
_ , rather than being an onset of phase coherence , corresponds to the opening of superconducting gap at @xmath15 .
alldredge _ _ et al.__@xcite have studied the tunneling density of states in momentum space .
their analysis indicates that there is only one gap in the underdoped materials .
_ find that a large effective anisotropic scattering rate @xmath350 is needed to fit their data in the underdoped regime .
their effective scattering rate becomes non - zero at @xmath351 , and increases approximately linearly with decreased doping , similarly to the normal state energy gap .
thus , they conclude , they have effectively two kinds of quasiparticles in the underdoped regime : quasiparticles near the @xmath328 antinodes in momentum space that are incoherent and almost localized , and coherent quasiparticles in the nodal region .
the observations of data collapse reviewed in the previous section provide two very important constraints on any theory of the underdoped cuprate superconductors .
they indicate the presence of two distinct fluids whose composition might be expected to vary with doping : a spin liquid containing localized cu spins with a doping - dependent effective interaction and a fermi liquid whose transport properties differ markedly from those of a landau fermi liquid . in this section
we develop a two - fluid framework for analyzing these experiments and use it to extract their consequences .
we present as well the details of our earlier analysis@xcite of the magnetic measurements of two fluid behavior .
our approach is inspired by the success of the two - fluid phenomenology developed for the 1 - 1 - 5 family of heavy electron materials@xcite .
for these and other heavy electron materials containing a kondo lattice of localized @xmath352-electrons coupled to a conduction band , a two - fluid phenomenological model of the hybridization of the @xmath352 electron localized spins with those in the conduction band describes very well the emergence of a hybridized non - landau heavy electron fermi liquid that coexists with unhybridized local f - electrons and conduction electrons .
the emergence of two components in the bulk spin susceptibility and the knight shift in the cuprates can be understood by writing the total spin of the system as a sum of the localized @xmath17-electron and @xmath18-hole spins , @xmath353 where @xmath354 are the positions of copper @xmath17-electrons , @xmath355 the positions of oxygen p - holes .
quite generally , the coupling between these spins gives rise to three contributions to spin susceptibility@xcite , @xmath356 where @xmath357 represents the contribution from the localized cu spins , @xmath358 represents the part of the oxygen @xmath18-band that is not hybridized , while @xmath359 corresponds to the magnetic response of the hybridized quasiparticles with a large fermi surface : @xmath360 the fermi liquid contribution to spin susceptibility thus arises from @xmath361 and @xmath358 .
only * the position * of oxygen holes is important for the above decomposition ; whether or not they form a single hybridized band is irrelevant .
the three parts are present in an explicit two - band model of walstedt _ _ et al.__@xcite that produces correct expressions for the relaxation rates in terms of @xmath358 , @xmath361 , and @xmath357 .
it differs from the standard mila - rice - shastry / mmp@xcite ionic model that assumes a single spin degree of freedom residing on the copper site , so that @xmath362 and the hybridized @xmath363 band description of millis and monien@xcite , in which although @xmath358 , @xmath361 , and @xmath357 are taken to be non - zero , all three components are assumed to track the spin susceptibility of a single hybridized @xmath363 band .
the scaling results of the previous section tell us that @xmath357 must maintain its local character throughout much of the phase diagram , for how else could the bulk susceptibility and the low frequency magnetic response map onto the 2d heisenberg model ? to describe this we therefore write : @xmath364 where @xmath1 is the fraction of the @xmath365 ( and the total ) response function that retains its local spin character , while the remaining portion of the spin - spin response function describes the response of a ( non - landau ) fermi liquid with the large fermi surface that results from hybridization : @xmath366 \chi_{fl}(\bm{q } , \omega ) .
\label{pdpart}\ ] ] we thus arrive at the two fluid description , @xmath367 \chi_{fl}(\bm{q } , \omega ) , \label{dyn}\ ] ] that we will use in our subsequent analysis to extract from experiment the doping dependence of both @xmath368 and @xmath1 .
the spin liquid contribution completely determines the low frequency dynamic magnetic spin susceptibility measured by nmr on copper nuclei .
it corresponds to the scaled spin response of the 2d heisenberg model for localized copper spins , which in the low frequency limit can be written in a general phenomenological form proposed by millis , monien , and pines@xcite , modified to include the possibility of propagating spin wave excitations @xcite : @xmath369 where @xmath370 is the gap in spin excitation spectrum .
we note that in general damping in the heisenberg model is caused by the spin wave scattering , so that the relaxational frequency , @xmath117 , must be , in general , frequency - dependent in the underdoped cuprates .
we further note that if one assumes heisenberg model scaling , then , from dimensional arguments , @xmath371 is proportional to @xmath372 .
the walstedt _
_ generalization of the mila - rice - shastry hyperfine hamiltonian for cu site can be written as : @xmath373 + ^{63}i_{\rho } d \sum_j \bm{s}^p_{\rho}(\bm{r}_j ) \right\ } \label{63hyp}\ ] ] here @xmath374 corresponds to the nearest neighbor cu sites , @xmath375 while @xmath376 is a new transferred hyperfine constant for the spins sitting on oxygen .
the sum index @xmath377 goes over the oxygen sites neighboring the copper site .
similarly , one can write the hyperfine hamiltonian for the oxygen nuclei : @xmath378 here @xmath379 is the transferred hyperfine coupling constant for localized cu spins and @xmath380 is the contact hyperfine interaction for the oxygen nucleus ; the sum index @xmath374 goes over the nn copper sites . making use of @xmath381 where @xmath382 is the strength of the external magnetic field and @xmath383 is the electron magnetic moment
, we get , from the above hamiltonians : @xmath384 here @xmath385 are the corresponding nuclear gyromagnetic ratios . on introducing the dynamic spin susceptibilities , @xmath386 where @xmath387
, we obtain the walstedt _
_ expression for relaxation rates : @xmath388 where @xmath389 is the direction of magnetic field , @xmath390 is the nuclear site . for @xmath357 one
gets the usual mmp form factors : @xmath391 @xmath392 @xmath393 the form factors for the mixed p - d contribution are : @xmath394 @xmath395 @xmath396 @xmath397 finally , for the p - p part of spin susceptibility : @xmath398 @xmath399 @xmath400 apart from @xmath401 , the form factors for the p - d and p - p contributions vanish at @xmath402 : @xmath403 in what follows we focus on the behavior of planar dynamic and static magnetic response and the thermodynamic behavior , as measured by the inelastic neutron scattering , nmr , heat capacity and bulk spin susceptibility .
we now apply our two - fluid description to an analysis of the bulk spin susceptibility and knight shift measurements in la@xmath71sr@xmath72cuo@xmath59 and yba@xmath58cu@xmath98o@xmath159 families of cuprate superconductors . for the static susceptibility , the two components in eq.([dyn ] )
can be written in the following form : @xmath404 where @xmath405 here @xmath406 and @xmath407 are temperature- and doping - independent van vleck and core contributions .
@xmath408 is usually taken to be diamagnetic , due to a large core contribution@xcite .
as noted earlier , @xmath368 follows very well the calculated@xcite bulk spin susceptibility for the heisenberg model with a doping - dependent exchange constant @xmath93 .
thus , the spin liquid contribution to static magnetic response can be written as @xmath409 where @xmath410 is the maximum value of the spin liquid susceptibility and @xmath411 is a universal function of @xmath103 .
sr@xmath72cuo@xmath59 from nakano _ _ et al.__@xcite to the heisenberg model numerical calculations of makivi and ding@xcite .
the maximum for spin susceptibility is reached at @xmath412 .
deviations from the heisenberg model results are observed for @xmath413 .
the inset shows the numerical results for the correlation length@xcite , that demonstrate that @xmath414 at temperature @xmath415.,scaledwidth=75.0% ] a comparison of the universal data collapse curve for the bulk spin susceptibility in la@xmath71sr@xmath72cuo@xmath59 family found by nakano _ _
et al.__@xcite to the heisenberg model calculations@xcite is shown in fig.[hna ] .
we see that the universal function of nakano__et al . _
_ deviates significantly from the 2d heisenberg model results at low temperatures , @xmath413 .
as noted earlier , the scaling analysis of nakano _ _ et al.__@xcite differs from that of johnston@xcite , because it includes at lower doping levels a curie term @xmath75 , and a linear term @xmath416 in the static spin response .
while the inclusion of these terms may be justified , it introduces a number of new parameters that are , in a sense , unnecessary since the universal curve in the insulating compound must agree with the 2d heisenberg model results .
we therefore omit these terms in our analysis , and we assume , in agreement with recent photoemission studies@xcite , that the fermi liquid contribution , @xmath417 , is temperature - independent above @xmath14 , but could , for the 1 - 2 - 3 materials become temperature - dependent below the temperature , @xmath14 , at which a gap starts to form in the quasiparticle energy spectrum , leading to the formation of the fermi arcs . to determine the doping dependence of the parameters in eq.([static ] ) , we analyzed the bulk spin susceptibility data@xcite for metallic underdoped la@xmath71sr@xmath72cuo@xmath59 . with a constant contribution @xmath418 subtracted , we find , in agreement with the earlier results@xcite , the data collapse to the 2d heisenberg model curve , shown in fig .
[ knan1 ] , where we see that the data collapse continues below the temperature @xmath419 at which one no longer finds agreement with the heisenberg model .
sr@xmath72cuo@xmath59 from nakano _ _ et al.__@xcite to the heisenberg model numerical calculations of makivi and ding@xcite .
the maximum for spin susceptibility is reached at @xmath412 . while the data collapse remains reasonably good even at low temperatures , large deviations from 2d heisenberg model prediction
are observed for @xmath420.,scaledwidth=75.0% ] a similar scaling analysis for the @xmath95 knight shift data in 1 - 2 - 3 materials is shown in fig.[knan ] .
knight shift@xcite in yba@xmath58cu@xmath98o@xmath159 , compared to the scaling function obtained by nakano _ _
et al.__@xcite for the bulk spin susceptibility in la@xmath71sr@xmath72cuo@xmath59 .
@xmath13 is the temperature at which the knight shift has a maximum .
, scaledwidth=75.0% ] the doping dependence of @xmath421 that follows from the analysis of the bulk susceptibility and knight shift data , is shown in fig .
[ nak ] for both la@xmath71sr@xmath72cuo@xmath59 and yba@xmath58cu@xmath98o@xmath159 families .
@xmath28 falls linearly with increased doping @xmath22 for both materials .
for la@xmath71sr@xmath72cuo@xmath59 , and doping levels less than @xmath422 , @xmath423 that follows from our analysis of the nakano _
_ et al.__@xcite bulk spin susceptibility data for la@xmath71sr@xmath72cuo@xmath59 , and ( b ) the knight shift data for yba@xmath58cu@xmath98o@xmath159,title="fig:",scaledwidth=75.0% ] that follows from our analysis of the nakano _ _
et al.__@xcite bulk spin susceptibility data for la@xmath71sr@xmath72cuo@xmath59 , and ( b ) the knight shift data for yba@xmath58cu@xmath98o@xmath159,title="fig:",scaledwidth=75.0% ] our analysis of the bulk susceptibility data in la@xmath71sr@xmath72cuo@xmath59 also enables us to extract the doping dependence for the fermi liquid and spin liquid components .
the results for @xmath65 and @xmath424 given by eqs.([fmax]),([static ] ) are shown in fig.[flcomp ] . and
@xmath424 , from our analysis of the bulk susceptibility measurements in la@xmath71sr@xmath72cuo@xmath59 . here
@xmath424 here also includes a diamagnetic constant , @xmath408 , hence the non - zero offset.,scaledwidth=75.0% ] the doping dependence for the fractional occupation , @xmath1 , of the spin liquid state can be determined from @xmath28 and @xmath65 .
the dimension of @xmath40 is @xmath425 .
thus , since the universal curve for @xmath426 coincides with the numerical calculation for the 2d heisenberg antiferromagnet with @xmath427 , we can write : @xmath428 where @xmath411 is a universal function for the heisenberg antiferromagnet . using eqs([scbulk ] ) and ( [ fmax ] ) , the doping dependence of @xmath1 is then determined by the product @xmath429 : @xmath430 from our data analysis , we find that for the doping levels available for our analysis , @xmath1 also decreases linearly with @xmath22 , @xmath431 as shown in fig .
[ fofx ] , so that @xmath28 and @xmath1 roughly track each other . , from our analysis of the bulk susceptibility measurements in la@xmath71sr@xmath72cuo@xmath59@xcite.,scaledwidth=75.0% ]
[ flcomp ] also shows the doping dependence for the fermi liquid component , @xmath432 .
we find that the fermi liquid component increases linearly with doping in the metallic regime , in agreement with eq.([static ] ) .
the offset for this linear dependence is due to the presence of @xmath408 , which is diamagnetic @xcite .
since @xmath1 also decreases linearly with @xmath22 , we conclude that @xmath433 in eq.([static ] ) has only mild , if any , doping - dependence .
we now examine the knight shift data in more detail . as noted in section 2 , most early knight shift experiments on different nuclei found the same temperature - dependent contribution to spin susceptibility@xcite , although some deviations from this universality ( within error bars ) can be seen on the universal plot below the superconducting temperature @xmath15@xcite .
the linear @xmath434 and @xmath435 plots above @xmath15 , observed in many early knight shift experiments , then provide a measurement of the hyperfine couplings to single copper spin .
these measurements had been regarded as a key proof of an effective one - component model .
however , the cu knight shift measurements on the 1 - 2 - 4 material by suter _ _ et al.__@xcite and the recent apical oxygen knight shift measurements in la@xmath105sr@xmath106cuo@xmath59 by haase _ _ et al.__@xcite , tell a different story , and we consider these now . the knight shift measurements of suter _
et al . _ were done for an external magnetic field in the c direction , where , because of the well - known accidental cancellation@xcite of the @xmath357 form factor , @xmath436 , in eq.([k63eq ] ) , within the single - component approximation , the knight shift for fields along the c - axis would be of purely orbital origin .
within the two - fluid description , we see from eq.([k63eq ] ) that one can , in addition , probe fermi liquid behavior through an isotropic transferred hyperfine interaction @xmath376 that arises from the hybridization of planar oxygen p - orbitals and copper s - orbitals .
so to the extent that the fermi liquid susceptibility becomes temperature dependent as a result of a quasiparticle gap opening up at @xmath14,one would expect to see this in @xmath104 . as may be seen in fig .
[ chippd ] , where we have extracted the behavior of @xmath437 as a function of t from their knight shift measurements , this is just what was found by suter _ _
et al.__@xcite .
the anomalous fermi liquid behavior begins at a temperature , @xmath438 for yba@xmath58cu@xmath98o@xmath62 , an onset temperature that is consistent with the onset of fermi arc behavior at a comparable value of @xmath14 seen in the arpes measurements on underdoped 2212 materials . in their recent re - analysis of the planar @xmath439cu and @xmath440o knight shift experiments on la@xmath105sr@xmath106cuo@xmath59 that was accompanied by new apical oxygen knight shift experimental results , haase _ _ et al.__@xcite have found convincing experimental evidence for the existence of both a fermi liquid and a spin liquid component in this material .
they show that while all nuclei see the same temperature - dependent spin liquid component , the different nuclei see as well a component that is independent of temperature above the superconducting transition temperature , @xmath15 , but displays the expected d - wave signature below @xmath15 , and hence must be associated with a fermi liquid component , rather than representing an orbital or van vleck core term .
quite importantly , they combine their results in the superconducting state ( that include correcting for any meissner shielding effects ) and the normal state for magnetic fields in and perpendicular to the cuo planes with equations ( [ k63eq ] ) and ( [ k17eq ] ) , supplemented by an equivalent expression for the apical oxygen site , to determine all the relevant hyperfine couplings .
they show these can be used to determine the strength and temperature dependence in the normal and superconducting states of all three components , @xmath357 , @xmath359 , and @xmath358 .
we refer the interested reader to their paper for the details of their findings , which place important constraints on any description of the hybridization of the cu @xmath17 electrons with the oxygen @xmath18 band . as a function of @xmath103 , inferred from @xmath441 measurements@xcite in yba@xmath58cu@xmath98o@xmath62.,scaledwidth=75.0% ] while further knight shift measurements under conditions favorable to observing the fermi liquid component are highly desirable , we believe it is reasonable to conclude that the two - fluid analysis of knight shift in heavy electron materials@xcite extends naturally to high-@xmath15 superconductors , with @xmath442 and @xmath443 eq . ( [ hsopar ] ) can then be combined with the results of haase _ et al .
_ , and our previously determined value of @xmath1 , to determine the doping - independent quantity @xmath433 for the 2 - 1 - 4 materials .
the results of this analysis can be checked by extending the haase _
analysis to other doping levels . in analyzing the nmr relaxation rates
, we shall focus on the @xmath95 spin - lattice relaxation rate , @xmath112 , and the spin - echo - decay time , @xmath113 . as discussed in the beginning of the section
, there are different components of the spin response . however , due to the large peak of the spin - liquid part at @xmath107 , and the corresponding non - vanishing form factors for @xmath95 nuclei , the copper relaxation rates primarily probe the low - frequency properties of the spin - liquid contribution that arises from @xmath444 part of the spin susceptibility . in our modification of mmp@xcite both @xmath361 and @xmath358
correspond to the contribution of fermions with a large hybridized fermi surface , or pieces of such a fermi surface , in agreement with the knight shift measurements of haase _ _ et al.__@xcite , and therefore have a negligible effect on copper relaxation rates . an alternative possibility ,
that we mention only briefly , is that the two components seen in the bulk measurements are a result of microscopic phase separation and formation of dynamic stripes .
surprisingly , while the transition points that we obtain from the two - component analysis of bulk measurements coincide with the well - known region of phase separation in la@xmath58cuo@xmath445@xcite , the two - component nature is evident in bulk measurements at temperatures well above the phase separation temperature region .
where these can both be carried out , measurements of @xmath95 spin - lattice relaxation and spin - echo decay rates indicate the presence of two different dynamic scaling regimes that lie above and below @xmath13@xcite . above @xmath13 one is in a mean - field , @xmath127 scaling regime , in which the relaxational frequency , @xmath117 varies as @xmath446 ; below @xmath13 it varies as @xmath447 because the spin liquid has entered the @xmath12 dynamic scaling regime@xcite expected for the quantum critical ( qc ) regime of a 2d antiferromagnet . as discussed in ref.@xcite
, one can describe the spin dynamics using the quantum non - linear sigma model , or spin wave theory@xcite .
the resulting qc scaling theory@xcite for the spin liquid without long - range order gives a linear dependence of the correlation length on temperature , @xmath448 where @xmath449 is the spin wave velocity , and the offset @xmath450 goes to zero at @xmath451 , a quantum critical point for the spin liquid that marks the onset of long - range order .
a similar linear dependence on @xmath26 can be expected also for @xmath452@xcite , @xmath453 with an offset @xmath140 that measures the distance from the proposed quantum critical point , while @xmath454 is a universal coefficient .
quantum critical theory applies in the universal regime , @xmath455 , thus , well below @xmath13 .
scaling of the form seen in eq.([qct1 t ] ) is self - evident from the low - temperature @xmath112 data on la@xmath71sr@xmath72cuo@xmath59 material of ohsugi _ _ et al.__@xcite(fig.[kitaokafig ] ) .
when they plot the product @xmath137 vs temperature in the metallic underdoped regime , they find a set of parallel lines for their different doping levels that extend from @xmath456 down to a low - temperature upturn in @xmath457 that is near @xmath15 .
we plot in figs.[t1tla],[t1ty ] the scaling behavior for @xmath112 in the 2 - 1 - 4 and the 1 - 2 - 3 materials .
the @xmath22 dependence of the off - set , @xmath140 , of the low - temperature linear behavior in @xmath457 points to a critical point in the spin liquid at @xmath11 , separating long range order from short range order , a result that was also suggested earlier in an analysis of the high - temperature @xmath112 data@xcite .
in la@xmath71cuo@xmath59@xcite the offset @xmath140 shown in the inset depends linearly on @xmath22 , and points to a qcp in the spin liquid.,scaledwidth=75.0% ] in yba@xmath58cu@xmath98o@xmath159 .
the offset @xmath140 shown in the inset depends linearly on @xmath22 , and points to a qcp in the spin liquid.,scaledwidth=75.0% ] the long range magnetic order at @xmath458 can be a 2d antiferromagnet or a spin glass@xcite .
while the details of the microscopic theory at low doping may vary , the experimental data imply the presence of a linear spin wave - type excitation spectrum in the spin liquid with a temperature - dependent gap , @xmath459 , that is controlled by its distance from the quantum critical point at @xmath11 .
the spin wave excitation spectrum has the following form : @xmath460 here @xmath256 is the spin wave velocity . in the @xmath12 qc regime
the spin gap @xmath461 tracks @xmath462 . for @xmath463
the gap saturates at low temperatures to @xmath464 at the crossover to the qd gapped spin liquid regime .
we note that the origin of the saturation of the correlation length is largely irrelevant for data analysis - an energy gap can appear in the 2d quantum spin liquid@xcite , or in 1d stripes of spin liquid@xcite due to dimensional crossover . in the latter case
the correlation length will saturate at the size of the domain or dynamic stripe order , @xmath465 .
therefore , experiment does not distinguish between these possibilities .
the qc - qd crossover could explain the sharp upturn in @xmath457 at low temperatures observed in nmr experiments on 1 - 2 - 3 materials at temperatures well above @xmath15 that is not found in experiments on the 2 - 1 - 4 materials .
since @xmath466 , and becomes small as one approaches @xmath467 , while @xmath468 increases linearly from @xmath11 , the zero - temperature spin liquid gap @xmath469 will have a bell - shaped form similar to @xmath204 .
the gapping of the spin liquid leads to an exponential decrease of damping with temperature in the spin liquid .
the detailed properties of the spin liquid can be extracted directly from the nmr @xmath112 relaxation rates measurements using the ansatz for the correlation length @xmath470 , or , where these exist , from both @xmath112 and @xmath113 measurements , as it was shown some time ago by the authors@xcite . in the @xmath12 qc regime , @xmath471 where all parameters are taken from eq.([slpar ] ) ; these equations differ from those in ref .
@xcite by the presence of an extra factor @xmath1 in eq.([slpar ] ) .
a combination of these two measurements@xcite gives a handle on the spin wave velocity @xmath256 : @xmath472 the coefficients of proportionality are given by the hyperfine hamiltonian .
as easily seen from the above equations , @xmath473 the high - temperature scaling for @xmath112 , @xmath474 , that was seen by imai _ _ et al.__@xcite in la@xmath71sr@xmath72cuo@xmath59 , but is not present in the 1 - 2 - 3 materials , at first sight seems to imply @xmath22-independent exchange integral , @xmath10 , or @xmath475 , which contradicts the bulk susceptibility data for this material .
however , a closer look shows there is no contradiction .
as we have seen in section 3.1 , @xmath1 and @xmath28 both depend linearly on @xmath22 .
thus , on general grounds , one can write : @xmath476 the imai _
_ et al.__@xcite scaling then follows from eq.([tstca ] ) , since one can approximately write : @xmath477 in general , the empirical scaling of eq.([empir ] ) is somewhat accidental , since the low - energy properties of the spin liquid can be renormalized by interaction with fermionic quasiparticle excitations .
this renormalization appears to be small for the 2 - 1 - 4 materials but appears to be important for the 1 - 2 - 3 materials , where the ohsugi _ _ et al.__@xcite results and the imai _
_ et al.__@xcite high - temperature scaling of @xmath112 are not observed .
@xmath1 , @xmath28 , and @xmath478 can in principle be independently determined from nmr experiments .
since @xmath113 has only been measured for the yba@xmath58cu@xmath98o@xmath159 family , we plot the doping dependence of @xmath475 for these materials in fig .
[ cx ] , assuming _ commensurate _ local spin response . from nmr experiments in yba@xmath58cu@xmath98o@xmath159,scaledwidth=75.0% ] it is evident that @xmath478 decreases approximately linearly with doping .
however , a comparison of @xmath478 with @xmath28 shows that the two quantities do not track each other , violating the empirical scaling of eq.([empir ] ) . while the doping dependence for both is linear , it is evident from fig.[ctst ] that they are not proportional to one another .
the violation of the imai _
_ et al.__@xcite scaling relation by this family of materials is therefore to be expected .
vs @xmath28 from nmr experiments in yba@xmath58cu@xmath98o@xmath159,scaledwidth=75.0% ] the doping dependence of the parameter @xmath479 , that follows from nmr @xmath112 measurements only , is shown in figs [ calpha],[caltst ] .
we see that the empirical scaling relation eq.([alpaf ] ) holds reasonably well for the la@xmath71sr@xmath72cuo@xmath59 family , so that the doping dependence of this parameter also yields the doping dependence of @xmath278 . from nmr experiments in la@xmath71sr@xmath72cuo@xmath59,scaledwidth=75.0% ] vs @xmath28 from nmr experiments in la@xmath71sr@xmath72cuo@xmath59,scaledwidth=75.0% ]
we now turn to an analysis of thermodynamic measurements of the electronic entropy@xcite .
these are quite difficult to carry out , since the phonon contribution accounts for 98 - 99 % of the measured value of the entropy , and must be subtracted to a high degree of accuracy to get the electronic contribution .
the subtraction procedure is a complicated process , involving subtracting data for a reference material , such as yba@xmath58cu@xmath98o@xmath480@xcite , and making additional corrections for the modification of the phonon dos with doping .
the reference material is usually an insulator , either the parent compound , or a compound in which the fermi liquid is suppressed with impurities such as zn .
the difficulty with such a subtraction procedure for analyzing the data is that the spin liquid contribution exists even when the fermi liquid is suppressed .
thus , a consistent data analysis requires that the reference material be known . for this reason
we restrict our attention to the data of ref.@xcite , where the reference material is yba@xmath58cu@xmath98o@xmath480 .
the spin liquid contribution of the parent compound is subtracted from the electronic entropy in this analysis .
nevertheless , at moderate temperatures @xmath143 it remains comparatively small .
our analysis of the data of ref.@xcite shows that the entropy displays scaling and two component behavior in the form : @xmath481 with a scaling function @xmath482 that corresponds to the theoretically calculated entropy for the heisenberg model@xcite . in yba@xmath58cu@xmath98o@xmath159 .
the entropy data is from ref.@xcite in the intermediate doping regime , where the spin liquid part of the entropy is not subtracted . @xmath28 and @xmath1 obtained from this analysis are consistent with our analysis of the nmr knight shift and the bulk spin susceptibility .
the solid line is the integrated numerical result for the heat capacity of the 2d heisenberg model taken from ref .
@xcite.,scaledwidth=75.0% ] we have integrated the numerical results for the heat capacity of the heisenberg model from ref.@xcite to get the expected entropy .
we show the results of this analysis and the data collapse on the heisenberg scaling curve in fig.[entr ] .
we find that at small doping scaling becomes progressively worse , as might be expected because of the subtraction procedure followed .
to understand our result , we note that on general grounds , one can write the entropy in the form : @xmath483 where @xmath484 is the configurational entropy of the fluctuating stripe or domain order discussed in the next subsection . the configurational part , however , is not seen in the data factorization eq.([entrfact ] ) we believe , the reason for this is that the configurational contribution of the fluctuating stripes could also be linear in temperature@xcite .
indeed , the domain walls can be treated as solitons , with the hamiltonian @xmath485 where @xmath486 is the direction along the stripe , while @xmath487 is the displacement of the wall perpendicular to its equilibrium direction ; @xmath488 is the linear tension of the domain wall ( soliton ) .
this model maps on a 1d fermion model@xcite , and the resulting configurational entropy has a linear form , @xmath489 where @xmath490 is the mean distance between solitons . such a linear contribution to the entropy is indistinguishable from the fermi liquid contribution .
our analysis also casts light on the issue of intrinsic spatial inhomogeneity@xcite in the pseudogap state@xcite .
the issue is how to reconcile the commensurate spin susceptibility peaks required by the spin liquid and the fit to the nmr data@xcite , with the incommensurate peaks seen in inelastic neutron scattering experiments@xcite .
the oxygen relaxation rate is known to be very small compared to copper , with a very different modified korringa - type temperature dependence .
an attempt to explain it within the one - component model with incommensurate spin liquid peaks@xcite in la@xmath105sr@xmath106cuo@xmath59 leads to an oxygen relaxation rate that is a factor of three too large compared with actual experimental data , and has the wrong temperature dependence .
similar inconsistencies occur with relaxation rates on other nuclei , such as @xmath491 or @xmath286 in yba@xmath58cu@xmath98o@xmath159 .
the contradiction can not be resolved if one assumes two different components sitting on oxygen and copper .
indeed , since the knight shift on different nuclei has an identical temperature dependence@xcite , it requires the existence of a transferred hyperfine coupling from the copper@xcite . such a transferred hyperfine interaction assumed by mmp
will , however , induce a `` leakage '' spin liquid contribution from the spin liquid in the two - component picture for any incommensurate spin liquid response .
the resolution was suggested some time ago by slichter@xcite , who pointed out that if , as a result of coulomb interaction between the holes , regions that were hole rich formed domain walls between regions in which one had nearly antiferromagnetic commensurate behavior , the two kinds of experiments were compatible . this issue has been independently confirmed and clarified by tranquada__et al.__@xcite , who found that the spin wave dispersion seen in cuprate superconductors by inelastic neutron scattering is consistent with commensurate spin response of a spin ladder , while the incommensurate structure appears as a result of the fluctuating stripe order .
such a scenario implies finite - size scaling , and saturation of the antiferromagnetic correlation length at low temperatures at a value of the order of the size of the domain . estimating the size of the domain from the discommensuration seen in neutron scattering experiments@xcite in la@xmath186sr@xmath187cuo@xmath59 , @xmath492 , @xmath493 , we can write : @xmath494 thus , @xmath495 where @xmath496 is the lattice spacing .
aeppli _ _ et al.__@xcite find the saturated correlation length in la@xmath71sr@xmath72cuo@xmath59 at a doping level of @xmath192 is @xmath497 a value that is very close to our estimate of @xmath150 .
we conclude that both magnetic probes imply the presence of a fluctuating striped or other domain order .
apart from the issue of incommensurate peaks seen in ins experiments and its possible resolution in terms of a fluctuating domain order , the spin wave spectrum and the correlation length observed in inelastic neutron scattering experiments is completely consistent with our proposed two - fluid scenario , as we show below .
we return now to the magnetic scaling seen in the inelastic neutron scattering experiments which , like the @xmath95 nmr relaxation rates , probe spin liquid behavior .
we begin with the experiments of sternlieb _ _ et al.__@xcite that imply the presence of a true @xmath27 scaling for the local spin susceptibility in the underdoped material yba@xmath58cu@xmath98o@xmath160 .
the data collapse observed experimentally can be fit to the following simple scaling form : @xmath166 with deviations from scaling that occur at progressively lower temperatures , as the frequency @xmath167 increases .
this scaling confirms the high - temperature scaling seen in nmr@xcite . to clarify this point
, we plot both experimental results in the same units in fig .
[ insnmr ] .
the quantity measured in neutron scattering experiments analogous to @xmath452 is a function of the local spin response , @xmath498 , typically given in counts / minute .
we therefore fix the absolute value of that response to be consistent with high - temperature nmr .
cu@xmath98o@xmath160 , @xmath499 , scaled to nmr @xmath457 in la@xmath71sr@xmath72cuo@xmath59 and yba@xmath58cu@xmath98o@xmath159.,scaledwidth=75.0% ] both experiments reveal linear in temperature @xmath117 , a signature of magnetic quantum critical behavior . at finite frequencies
this regime extends to lower temperatures , as one may expect on general grounds .
the presence of @xmath12 quantum critical behavior is verified further by the ins measurements of the temperature dependence of the correlation length seen in la@xmath186sr@xmath187cuo@xmath59@xcite .
_ et al.__@xcite show that @xmath500,\ ] ] where @xmath501 , @xmath502 . for the qc
scaling in the @xmath503-model@xcite , @xmath504 on the other hand , the aeppli _ _ et al.__@xcite result gives , if @xmath81 is measured in lattice constants : @xmath505 thus , for @xmath261 we find @xmath506 a value that is much less than that found for the insulator value @xmath507 , and is qualitatively consistent with the behavior of the effective @xmath10 inferred from the bulk spin susceptibility measurements .
another comparison of theory with experiment on the spin liquid phase diagram comes from the neutron scattering measurements of the magnetic correlation length . for the heisenberg spin liquid , the correlation length @xmath508 at @xmath415 . on using the results for @xmath81 obtained by aeppli _
, we find : @xmath509 in good agreement with the result , @xmath510 , at @xmath192 doping obtained using eq.([tmresu ] ) . a further important point that follows from the results of aeppli _ _ et al.__@xcite concerns the breakdown of scaling , which should occur at a universal scale @xmath511 , such that @xmath512 where @xmath513 is the temperature of the breakdown of @xmath27 scaling seen in the neutron scattering experiments at a fixed frequency @xmath167 .
nmr experiments just measure @xmath511 directly .
we analyze the data of sternlieb _ _ et al.__@xcite using this framework to see if it is consistent with the results shown in fig .
[ st1 ] . according to aeppli _
_ , the breakdown of scaling frequency / temperature can be fit to the form : @xmath514 from this analysis we find , for ybacuo@xmath160 : @xmath515 or @xmath516 overall , this experimental picture is self - consistent , although the experimental finding of the simple form for the breakdown eq .
( [ break ] ) is rather surprising .
@xmath28 corresponds to a crossover at @xmath517 to the strongly correlated phase in the spin liquid .
@xmath518 corresponds to the appearance of fluctuating stripe order @xmath519 marks the spin liquid crossover from the quantum critical ( qc ) to quantum disordered ( qd ) , or gapped spin liquid , regime in the stripe phase .
@xmath520 is the spin liquid crossover from the qc to renormalized classical ( rc ) regime .
( b ) the spin liquid phase diagram for yba@xmath58cu@xmath98o@xmath159 displays similar crossovers.,title="fig:",scaledwidth=75.0% ] sr@xmath72cuo@xmath59 .
@xmath28 corresponds to a crossover at @xmath517 to the strongly correlated phase in the spin liquid .
@xmath518 corresponds to the appearance of fluctuating stripe order @xmath519 marks the spin liquid crossover from the quantum critical ( qc ) to quantum disordered ( qd ) , or gapped spin liquid , regime in the stripe phase .
@xmath520 is the spin liquid crossover from the qc to renormalized classical ( rc ) regime .
( b ) the spin liquid phase diagram for yba@xmath58cu@xmath98o@xmath159 displays similar crossovers.,title="fig:",scaledwidth=75.0% ] we conclude this section with a brief discussion of phase diagram for the spin liquid that is consistent with the experimental constraints discussed above .
our proposed phase diagram@xcite is shown in fig.[phased ] .
the phase transition or a crossover at a doping level @xmath91 is that from the conducting fermi liquid on the right - hand side to pseudogap matter on the left - hand side , in which a portion of the quasiparticle fermi surface has lost its low - frequency spectral weight , becoming localized in an insulating spin liquid that coexists with the remaining fermi liquid component .
the two interacting components are present at all temperatures in the underdoped regime , @xmath521 . at high temperatures @xmath522
the interaction leads to a modification of @xmath10 , the effective exchange interaction in the spin liquid , while the fermi liquid component remains intact . as the temperature is lowered , the growth of correlation length in the spin liquid results in stronger renormalization of the fermi liquid component , and the formation of the fermi arcs at temperatures @xmath413 .
the antiferromagnetic correlation length also becomes finite at low temperatures , which results in a gap for the spin excitation spectrum at @xmath523 .
the quantum critical point at @xmath524 corresponds to a quantum transition between two different states of the spin liquid .
we now return to the question of how close the measured spin liquid maps onto the heisenberg spin liquid .
while there is a one to one correspondence for the static spin susceptibility above @xmath525 , the difference becomes large at lower temperatures , where the correlation length @xmath81 and the spin liquid gap @xmath461 saturate at a finite value .
presumably it is this difference that is responsible for the mild departure in concentration dependences deduced for @xmath1 and @xmath28 .
given the presence of both a spin liquid and a fermi liquid in the normal state of underdoped superconductors , it is natural to consider their interaction as being responsible for their anomalous normal state properties and the transition to a gapped quantum state at @xmath14 and to superconductivity at @xmath15 .
we begin by considering the extent to which the observed scaling behavior reflects that interaction , and what might be some of its other consequences . and cold regions away from the hot spots , scaledwidth=75.0% ] from an examination of the experimental results presented in the previous sections we draw the following conclusions : \(i ) the scaling with @xmath13 of various transport properties of the non - landau fermi liquid tells us that the transport properties in the normal state of the cuprate superconductors are dominated by the magnetic coupling between the quasiparticles and the spin liquid excitations \(ii ) in the 1 - 2 - 3 materials , the scaling of @xmath14 and @xmath526 , the d - wave energy gap found below @xmath14 for quasiparticles located in the `` hot '' regions of the fermi surface , and below @xmath15 for the remaining `` cold '' or nodal quasiparticles , tells us that the physical mechanism for their gaps and subsequent superconductivity must be magnetic .
\(iii ) as may be seen in fig.[hotcold ] , the hot regions seen in the arpes experiments are defined by their proximity to points on the fermi surface that are separated by commensurate wavevectors , @xmath523 ; their physical location tells us that the magnetic interactions responsible for the gap must be peaked at @xmath527 . this is to be expected if the `` glue '' for the quasiparticle interaction is provided by the spin liquid excitations of the heisenberg model that exhibit the mmp form with a strong peak at @xmath527 , since their coupling to the fermi liquid yields an induced quasiparticle interaction that is likewise peaked at @xmath527 .
moreover , in agreement with the early models of spin - fluctuation - induced superconductivity@xcite the resulting energy gaps will be d - wave .
\(iv ) the penetration - depth measurements of the doping - dependence of the of the superconducting condensate tell us that where present in the underdoped materials , the spin liquid does not participate in their superconductivity .
\(v ) the fact that the scaling relations below @xmath13 for a pure heisenberg model are much better obeyed in the 2 - 1 - 4 materials than in the 1 - 2 - 3 materials tells us that the influence of the hot quasiparticles on the spin liquid is considerably weaker in the 2 - 1 - 4 materials .
\(vi ) since the spin liquid excitations in the two classes of materials are quite similar , these differences plausibly arise from band - structure effects@xcite that reduce the importance of hot spots on their fermi surface .
\(vii ) these last conclusions concerning the relative strengths of the coupling of hot quasiparticles to the spin liquid and their spin - fluctuation - induced quasiparticle interaction in the 1 - 2 - 3 and 2 - 1 - 4 materials provide a simple physical explanation of some of the major differences between their physical behavior of the two classes of materials .
it is why the 2 - 1 - 4 materials show no indications of either a spin fluctuation gap or a quasiparticle gap opening up at @xmath14 , and may be related to their significantly lower superconducting transition temperatures .
\(viii ) there are many indications that `` stripy '' behavior is more prevalent in the 2 - 1 - 4 materials ; this may be due to the fact that the fluctuating domains become well established only when fairly strong af correlations are present .
for the 1 - 2 - 3 materials gaps open up at @xmath94 , so the maximum af correlation length one expects to find is @xmath528 lattice spacings ; for the 2 - 1 - 4 materials , on the other hand , one finds af correlation lengths @xmath529 lattice spacings near optimal doping and still longer correlation lengths are to be expected for the underdoped materials .
\(ix ) experiment indicates that at low temperatures @xmath530 , charge fluctuations , domains , and tendencies toward phase separation all play an important role , although the two - component nature is present in the underdoped regime at all temperatures , and is therefore intrinsic .
\(x ) our two - fluid model provides a natural way to develop a phenomenological description of stripe formation at low frequencies , since all that is required is to let @xmath1 vary periodically in space and time .
we conclude this section with a brief discussion of some necessary elements of a microscopic description of the spin and fermi liquids .
we note first that for the spin liquid , the af critical point at @xmath531 is responsible for the observed @xmath532 qc behavior .
the spin liquid component at temperatures @xmath533 is well described to first approximation by the standard bosonic @xmath534 hertz - millis theory of quantum critical behavior@xcite . for @xmath535
one is outside the critical region , in the vicinity of weak - coupling gaussian fixed point , where fluctuation corrections are not important .
the standard damping due to its interaction with fermions gives rise to the @xmath536 mean field ornstein - zernike / mmp form . in the critical region
@xmath534 theory becomes equivalent to the qnl@xmath503 model@xcite ( the two theories differ by an irrelevant operator ) .
however , at low correlation lengths the irrelevant operator of the @xmath534 theory plays a crucial role , inducing a crossover to the mean field gaussian regime@xcite with @xmath537 for @xmath538 version@xcite .
such @xmath532 to @xmath536 mean field crossover can not be obtained in the quantum non - linear @xmath503-model with fermionic damping@xcite .
there appear to be two kinds of departures from the simple heisenberg description .
the first concerns the static bulk susceptibility that falls below the currently calculated values for the heisenberg model for @xmath539 . because this fall - off is essentially the same in both the 2 - 1 - 4 and 1 - 2 - 3 materials , it is likely not brought about by spin - fermion coupling , since that coupling is quite different for the two materials .
it could represent an inadequacy in the calculations of the expected behavior of the heisenberg model at these low temperatures .
the second concerns the opening up of a gap in the low frequency spin liquid excitation spectrum at @xmath193 that is seen as a minimum in @xmath540 at @xmath541 in the nmr experiments on the 1 - 2 - 4 material and at @xmath542 in the yba@xmath58cu@xmath98o@xmath100 material .
it is plausible to conclude that it reflects a reduction in quasiparticle damping brought about by the onset of the hot quasi - particle gapped behavior at @xmath14 .
however , it would then seem that the transition to this new quantum state is not universal , i.e. does not always occur at a fixed fraction of @xmath13 . indeed these nmr - determined onset temperatures , @xmath14 , scale , to first approximation with @xmath15 ; thus for the 1 - 2 - 4 material one has @xmath543 , while for yba@xmath58cu@xmath980@xmath100 one finds @xmath544 .
based on these estimates , and those obtained from the arpes and stm experiments on 2212 members of the 1 - 2 - 3 family , our candidate phase diagram for the 1 - 2 - 3 materials is given in fig .
[ phased]b .
it differs from the corresponding diagram for the 214 materials in the presence of a `` high temperature '' transition at @xmath14 to the gapped quantum state . in considering microscopic models for the evolution of fermi liquid behavior with doping , we begin by recalling that high - t@xmath146 superconductors are doped charge - transfer insulators@xcite .
thus , an adequate starting point for a description of the cuo@xmath58 planes follows from their chemistry , and is the three - band hubbard model composed of 3 d@xmath545 orbitals on the cu site and 2 p@xmath546 orbitals on the o sites@xcite .
the two key parameters of the three - band hubbard model are are the charge - transfer energy @xmath547 , and the copper on - site coulomb repulsion , @xmath548 . in the limit of strong p - d hybridization and very strong on - site cu coulomb interaction ,
the model gives rise to 5 different bands : an antibonding band , further split into upper hubbard band ( cu @xmath549 configuration ) , and lower hubbard band ( cu @xmath550 configuration ) by the coulomb interaction , a bonding band , further split into zhang - rice ( zr)@xcite singlet and triplet bands , and the non - bonding oxygen band . following anderson@xcite , most theoretical studies have considered a simpler one band hubbard model , which has fewer parameters and assumes that the low - energy physics of the electronic states in the vicinity of the charge - transfer gap is described well by taking a relevant low - energy subset of this complex band structure , the zr singlet band and the upper hubbard band , separated by @xmath551 of the order of the charge - transfer gap .
numerical studies of the three - band hubbard model@xcite support such reduction .
the two low - energy bands of the effective single - band hubbard model are strongly correlated hybridized copper - oxygen bands , with the upper hubbard band having mostly oxygen character@xcite .
however , on the basis of numerical lda calculations , it has been suggested that a single - band tight binding model may not provide a sufficiently good qualitative description of the many bands involved@xcite .
moreover , we have seen that @xmath357 contains the spin liquid component localized on the copper sites , while @xmath361 and @xmath358 pick up contributions from the holes on the oxygen sites .
the difference in relevant form factors could potentially explain many anomalies observed for nmr relaxation rates@xcite and inelastic neutron scattering experiments , as we have discussed in the previous section . the analogy with two - fluid model for heavy fermions@xcite can be better understood from an alternative approach to three - band hubbard model@xcite , that is perturbative in p - d hybridization , @xmath552 . in case of exactly one hole per cu and strong hubbard coulomb on - site repulsion @xmath548 , the low - energy hamiltonian for the three - band hubbard model reduces to the heisenberg spin hamiltonian with exchange coupling @xmath553 since the charge excitations have a gap @xmath554 .
numerical calculations for the three - band hubbard model@xcite extract the value @xmath555 , consistent with that observed in neutron scattering experiments .
the oxygen holes and the nearest @xmath6 spin interact via a strong kondo exchange interaction , which can be obtained in perturbation theory in @xmath556 , @xmath557 the perturbative analysis of the three - band hubbard model reduces it to the spin - fermion model via the schrieffer - wolff@xcite transformation , @xmath558 here the index @xmath374(@xmath377 ) runs over cu ( o ) sites , @xmath559 , @xmath560 are creation and annihilation operators , respectively , for holes in oxygen p bands .
the kondo interaction is , in general , non - local@xcite .
the hamiltonian eq.([sfmodel ] ) is the kondo - heisenberg model , known for the periodic anderson model in many physical systems , including magnetic semiconductors@xcite , heavy fermions ( the kondo lattice model)@xcite , and high - t@xmath146 superconductors@xcite .
the heavy fermion materials and high-@xmath15 materials correspond to different limits of the model eq.([sfmodel ] ) . in heavy fermion superconductors
the kondo interaction @xmath561 is initially comparatively small , and gets renormalized by the conduction band as the temperature is lowered , leading to the onset of coherence at @xmath14 , and the two - fluid heavy fermion behavior@xcite with a temperature - dependent fraction of spin liquid ( spin impurity contribution ) .
the high-@xmath15 superconductors , on the other hand , are in the strong - coupling zhang - rice@xcite limit of the same microscopic model , eq.([sfmodel ] ) .
the kondo interaction is very large , @xmath562 , which leads to a formation and propagation of the `` zhang - rice singlets '' , reducing the microscopic hamiltonian to an effective one - band @xmath563 model@xcite .
the projection operators of the zhang - rice singlet@xcite automatically imply that @xmath564 .
thus , as suggested by mmp@xcite , different nuclei then feel the presence of a single temperature - dependent spin component residing on copper , or @xmath357 . the zhang - rice singlet picture ( and the t - j model ) , however , inevitably gets violated at higher doping , as a single large fermi surface develops , that corresponds to antibonding p - d hybridized band . a linear decrease of effective @xmath49 with increased doping @xmath22 is `` high - energy physics '' , as is evident from , for example , an adiabatic treatment of the t - j model@xcite .
thus , the fraction of spin liquid@xcite observed in `` low energy '' experiments on the cuprate superconductors will be temperature - independent . in heavy fermion metals
it is the mixed @xmath565 contribution , @xmath566 , that gives rise to the heavy fermi liquid contribution@xcite . in high - temperature superconductors , on the other hand , one normally uses the zhang - rice singlet band and the upper hubbard band of the three - band hubbard model as lower and upper hubbard bands of the effective single - band hubbard hamiltonian .
using the analogy the heavy fermion two - fluid model@xcite and the two - fluid model for high - t@xmath146 superconductors , the fermi liquid contribution in the cuprates should arise from @xmath361 and @xmath358 parts of the bulk spin susceptibility , which , strictly speaking , indicates a violation of the zhang - rice singlet picture at high doping , given by eqs .
( [ ddpart ] ) , ( [ pdpart ] ) .
since the effective mass of carriers is not heavy in high-@xmath15 superconductors , one would expect both @xmath358 and @xmath361 contributions to be present .
this result contrasts strongly with the pure zhang - rice@xcite limit in which all contributions arise from @xmath357@xcite .
the key answer to this very important question is held by the new knight shift experiments@xcite , since then knight shifts on different nuclei should see a different temperature dependence , at least below the superconducting temperature , @xmath15 .
our review of scaling behavior in the cuprate superconductors makes evident the critical role that the localization of cu spins by strong on - site repulsive coulomb interactions plays in determining the properties of the magnetically underdoped superconductors , those whose planar hole concentrations are @xmath567 .
while there is much more in the way of details that experiment can and will provide , from additional c - axis nmr experiments on 1 - 2 - 3 materials that enable one to follow the emergence of a distinct fermi liquid contribution , to arpes , nmr , and stm experiments that fill in the details of the emergence of the gapped hot quasiparticle state at @xmath14 , we believe that the overall low - energy physical picture of a normal state described by two fluids whose interaction leads to its remarkable properties has now been firmly established .
it is equally clear that much much more remains to be done , beginning with the development of a better microscopic foundation for the physical picture that experiment has `` forced '' upon us , and an improved treatment of transport processes based upon it .
we mention here but a few of the many remaining challenges and mysteries that surround these fascinating materials : \(i ) there are striking similarities in the doping dependence of @xmath15 in the 1 - 2 - 3 and 2 - 1 - 4 materials despite the difference in their spin - fermion couplings and the magnitudes of their respective transition temperatures .
does this tell us that these differences are primarily in how the hot quasiparticles couple to spin fluctuations and that the spin - fermion couplings for the cold quasiparticles that bring on their superconductivity can somehow be scaled for these two classes of materials ?
\(ii ) although the spin liquid ceases to be a significant `` player '' for doping levels greater than , say , @xmath0 , the existence of a smooth cross - over in most physical properties of the normal state , including the superconducting transition , as one goes from magnetically underdoped to magnetically overdoped materials , suggests that its disappearance must be accompanied by the emergence of sufficiently strong antiferromagnetic correlations in the fermi liquid that enable its properties to merge smoothly with those induced by the spin liquid .
can these be detected experimentally and understood microscopically ?
might the disappearance of the spin liquid be accompanied by the emergence of a single - band hubbard model that is valid for the overdoped materials ?
\(iii ) do the gapped hot quasiparticles found below @xmath14 in the 1 - 2 - 3 materials become effectively localized and so become part of the spin liquid ?
is there any way to make their gapping more productive , in terms of their becoming part of the superconducting condensate , thereby opening the way to a superconducting transition at the considerably higher transition temperature , @xmath14 ?
\(iv ) can the lessons learned from the cuprates be transferred to other materials , and so lead over time to the discovery of still higher temperature superconductors ?
we would like to thank e. abrahams , n. curro , l.p .
gorkov , j. haase , j. schmalian , and c.p .
slichter for stimulating discussions , and acknowledge support for vb from nhmfl through the nsf cooperative agreement no .
dmr-008473 and for dp from the institute for complex adaptive matter and the us department of energy .
a. kanigel , m.r .
norman , m. randeria , u. chatterjee , s. souma , a. kaminski , h.m .
fretwell , s. rosenkranz , m. shi , t. sato , t. takahashi , z.z .
li , h. raffy , k. kadowaki , d. hinks , l. ozyuzer , and j.c .
campuzano , nature physics * 2 * , 447 ( 2006 ) .
b. keimer , r. j. birgeneau , a. cassanho , y. endoh , r. w. erwin , m. a. kastner , and g. shirane , phys .
lett . * 67 * , 1930 ( 1991 ) ; b. keimer , n. belk , r. j. birgeneau , a. cassanho , c. y. chen , m. greven , m. a. kastner , a. aharony , y. endoh , m.a .
kastner , and g. shirane , phys .
b * 46 * , 14034 ( 1992 ) .
j.c.campuzano , h.ding , m.r .
norman , h.m.fretwell , m. randeria , a. kaminski , j. mesot , t. takeuchi , t. sato , t. yokoya , t. takahashi , t. mochiku , k. kadowaki , p. guptasarma , d.g .
hinks , z. konstantinovic , z.z.li , and h. raffy , phys .
* 83 * , 3709 ( 1999 ) .
uemura , g.m .
luke , b.j .
sternlieb , h. brewer , j. f. carolan , w.n .
hardy , r. kadono , j.r .
kempton , r.f .
kiefl , s.r .
kreitzman , p. mulhern , t.m .
riseman , d.l .
williams , b.x .
yang , s. uchida , h. takagi , j. gopalakrishnan , a.w .
sleight , m.a .
subramanian , d.l .
chien , m.z .
cieplak , g. xiao , v.y .
lee , b. w. statt , c.e .
stronach , w.j .
kossler , x.h .
yu , phys .
lett . * 62 * , 2317 ( 1989 ) . | we present a detailed review of scaling behavior in the magnetically underdoped cuprate superconductors ( hole dopings less than @xmath0 ) and show that it reflects the presence of two coupled components throughout this doping regime : a non - landau fermi liquid and a spin liquid whose behavior maps onto the theoretical monte carlo calculations of the 2d heisenberg model of localized cu spins for most of its temperature domain .
we use this mapping to extract the doping dependence of the strength , @xmath1 of the spin liquid component and the effective interaction , @xmath2 between the remnant localized spins that compose it ; we find both decrease linearly with x as the doping level increases .
we discuss the physical origin of pseudogap behavior and conclude that it is consistent with scenarios in which the both the large energy gaps found in the normal state and their subsequent superconductivity are brought about by the coupling between the fermi liquid quasiparticles and the spin liquid excitations , and that differences in this coupling between the 1 - 2 - 3 and 2 - 1 - 4 materials can explain the measured differences in their superconducting transition temperatures and other properties .
_ advances in physics : _ + * universal behavior and the * + * two - component character of magnetically * + * underdoped cuprate superconductors * + victor barzykin@xmath3 and david
pines@xmath4 + @xmath3 national high magnetic field laboratory , florida state university , tallahassee , fl 32310 , usa + @xmath4 department of physics and institute for complex adaptive matter , university of california , davis , ca 95616 , usa |
the @xmath0-jet mass spectrum in @xmath67 scattering has been investigated with the zeus detector at hera .
an excess of events relative to standard model expectations has previously been reported by the h1 @xcite and zeus @xcite collaborations in neutral current deep inelastic scattering at high @xmath68 and high @xmath5 .
these events contain high - mass @xmath0-jet final states .
several models have been discussed @xcite as possible sources of these events , including leptoquark production @xcite and r - parity - violating squark production @xcite .
this paper presents an analysis of zeus data specifically aimed at searching for high mass states decaying to @xmath0-jet .
candidate events with high transverse energy , an identified final - state positron , and at least one jet are selected .
the measured energies ( @xmath69 ) and angles of the final - state positron and the jet with highest transverse momentum are used to calculate an invariant mass @xmath70 where @xmath71 is the angle between the positron and jet .
the angle between the outgoing and incoming positron in the @xmath0-jet rest frame , @xmath72 , is also reconstructed using the measured energies and angles .
no assumptions about the production process are made in the reconstruction of either @xmath73 or @xmath72 .
the search was performed using @xmath2 pb@xmath3 of data collected in the 1994 - 1997 running periods . in the following ,
expectations from the standard model , leptoquark production and r - parity - violating squark production are summarized . after a discussion of the experimental conditions ,
the analysis is described and the @xmath73 and @xmath74 distributions presented . since these distributions do not show a clear signal for a narrow resonance , limits on the cross section times branching ratio
are extracted for the production of such a state .
limits are also presented in the mass versus coupling plane which can be applied to leptoquark and squark production .
high - mass @xmath0-jet pairs , produced in the standard model ( sm ) via neutral current ( nc ) scattering , form the principal background to the search for heavy states .
this process is reviewed first .
leptoquark ( lq ) production and squark production in r - parity - violating ( @xmath75 ) supersymmetry are used as examples of physics beyond the sm that could generate the @xmath0-jet final state .
the diagrams for nc and lq processes are shown in fig .
[ fig : nclq ] .
the squark production diagrams are similar to the lq diagrams , but different decay modes are possible , as discussed below .
the kinematic variables used to describe the deep inelastic scattering ( dis ) reaction @xmath76 are @xmath77 where @xmath78 and @xmath79 are the four - momenta of the incoming and outgoing positron , respectively , and @xmath80 is the four - momentum of the incoming proton . the center - of - mass energy is given by @xmath81 .
the nc interaction occurs between the positron and a parton ( quark ) inside the proton ( see fig . [
fig : nclq ] ) .
the production of the large @xmath0-jet masses of interest requires high @xmath68 partons , where the valence quarks dominate the proton structure . in leading - order electroweak theory ,
the cross section for the nc dis reaction can be expressed as @xcite @xmath82\ ] ] with @xmath83 and @xmath84 the fine structure constant .
the contribution from the longitudinal structure function , @xmath85 , is expected to be negligible in the kinematic range considered here .
the @xmath68 dependence of the nc cross section is very steep .
in addition to the explicit @xmath86 factor , the structure functions @xmath87 and @xmath88 are dominated at large @xmath68 by valence - quark densities that fall quickly for @xmath89 .
the @xmath90 dependence of the cross section is dominated by the @xmath91 term .
the structure functions vary slowly with @xmath90 at fixed @xmath68 . the uncertainty in the nc cross section predicted by eq . [ eq : dis ]
is dominated by the uncertainty in the structure functions ( parton densities ) , and is small , about @xmath92% at the high-@xmath68 and moderate-@xmath90 ranges of this analysis @xcite .
the quantity of interest in this paper is the @xmath0-jet cross section , which is sensitive to qcd corrections .
the uncertainty arising from these corrections has been estimated to be small for this analysis . for dis or lq events produced via the diagrams shown in fig .
[ fig : nclq ] ( i.e. assuming no qed or qcd radiation ) , the mass of the @xmath93 system is related to @xmath68 via @xmath94 and @xmath95 is related to @xmath90 via @xmath96 the steeply - falling @xmath68 and @xmath90 dependences of dis events will therefore produce distributions falling sharply with mass and peaking towards @xmath97 .
leptoquark production is an example of new physics that could generate high - mass @xmath0-jet pairs .
the set of leptoquarks with @xmath98-invariant couplings has been specified @xcite . only lqs with fermion number @xmath99 are considered here , where @xmath100 and @xmath101 denote the lepton and baryon number , respectively .
these leptoquarks are listed in table [ tab : lq ] together with some of their properties .
the @xmath102 lqs have higher cross sections in @xmath103 scattering than @xmath104 scattering since in the @xmath67 case a valence quark can fuse with the positron .
@xmath105 in principle , additional lq types can be defined @xcite which depend on the generations of the quarks and leptons to which they couple . only lqs which preserve lepton flavor and which couple to first - generation quarks
are considered in this analysis . as shown in fig .
[ fig : nclq ] , leptoquark production can generate an s - channel resonance provided @xmath106 .
contributions to the @xmath67 cross section would also result from u - channel exchange and interference of lq diagrams with photon and @xmath107 exchange .
the cross section in the presence of a leptoquark can be written as @xmath108 the first term on the right - hand side of eq .
[ eq : terms ] represents the sm contribution discussed previously .
the second ( third ) term arises from the interference between the sm and u - channel ( s - channel ) lq diagram , and the fourth ( fifth ) term represents the u - channel ( s - channel ) lq diagram alone .
the additional contributions to the sm cross section depend on two parameters : @xmath109 , and @xmath110 or @xmath111 , the coupling to @xmath112 and quark
. leptoquarks of well - defined helicity ( @xmath113 ) are assumed for simplicity in the limit - setting procedure , and one species of lq is assumed to dominate the cross section .
the @xmath74 dependence varies strongly for the different terms : it is flat for scalar - lq production in the s - channel , and for vector - lq exchange in the u - channel , while it varies as @xmath114 for vector - lq production in the s - channel or scalar - lq exchange in the u - channel .
the interference terms produce a @xmath74 dependence which is steeper due to the sharply - peaking @xmath74 distribution in nc dis .
in general , the s - channel term dominates the additional contributions to the sm cross section if @xmath106 , the coupling @xmath115 is small , and the lq is produced from a quark rather than an antiquark . however , there are conditions for which the other terms can become significant , or even dominant @xcite , leading to important consequences for the expected mass spectra and decay angular distributions .
the u - channel and interference terms can not produce a resonance peak in the mass spectrum and the angular distributions from such terms can behave more like those of nc deep inelastic scattering .
limits are presented in this paper for narrow - width lq and under conditions for which the s - channel term dominates .
the width of a lq depends on its spin and decay modes , and is proportional to @xmath109 times the square of the coupling . in the narrow - width approximation ,
the lq production cross section is given by integrating the s - channel term @xcite : @xmath116 where @xmath117 represents the spin of the lq , @xmath118 is the quark density evaluated at @xmath119 and with the scale @xmath120 . in the limit - setting procedure ( sect .
[ sec : limits ] ) , this cross section was corrected for expected qed and qcd ( for scalar lq only ) radiative effects .
the qcd corrections @xcite enhance the cross section by 20 - 30% for the @xmath102 lq considered here .
the effect of qed radiation on the lq production cross section was calculated and was found to decrease the cross section by 5 - 25% as @xmath109 increases from @xmath121 gev . in the supersymmetry ( susy )
superpotential , @xmath122-parity - violating terms of the form @xmath123
@xcite are of particular interest for lepton - hadron collisions . here , @xmath124 , @xmath125 , and @xmath126 denote left - handed lepton and quark doublets and the right - handed down - type quark - singlet chiral superfields , respectively . the indices @xmath127 , @xmath128 , and @xmath78 label their respective generations . for @xmath129 , which is the case for @xmath130 collisions
, these operators can lead to @xmath131- and @xmath132-type squark production .
there are 9 possible production couplings probed in @xmath67 collisions , corresponding to the reactions @xcite @xmath133 for production and decay via the @xmath134 coupling , squarks behave like scalar leptoquarks and the final state is indistinguishable , event by event , from standard model neutral and charged current events .
however , as for the scalar leptoquarks , the angular distributions of the final - state lepton and quark will be different and this fact can be exploited in performing searches .
limits derived for scalar lq production can then be directly related to limits on squark production and decay via @xmath135 .
in addition to the yukawa couplings , gauge couplings also exist whereby @xmath136 can decay by radiating a neutralino or chargino which can subsequently decay .
the final - state signature depends on the properties of the neutralino or chargino .
the search for such decay topologies from a squark is outside the scope of this analysis .
in 1994 - 97 , hera operated with protons of energy @xmath137 gev and positrons of energy @xmath138 gev . the zeus detector is described in detail in @xcite .
the main components used in the present analysis were the central tracking detector ( ctd ) positioned in a 1.43 t solenoidal magnetic field and the uranium - scintillator sampling calorimeter ( cal ) .
the ctd was used to establish an interaction vertex with a typical resolution along ( transverse to ) the beam direction of @xmath139 cm .
it was also used in the positron - finding algorithm that associated a charged track with an energy deposit in the calorimeter .
the cal was used to measure the positron and hadronic energies .
the cal consists of a forward part ( fcal ) , a barrel part ( bcal ) and a rear part ( rcal ) , with depths of @xmath140 interaction lengths , respectively .
the fcal and bcal are segmented longitudinally into an electromagnetic section ( emc ) , and two hadronic sections ( hac1,2 ) .
the rcal has one emc and one hac section .
the cell structure is formed by scintillator tiles ; cell sizes range from @xmath141 @xmath142 ( femc ) to @xmath143 @xmath142 at the front face of a bcal hac2 cell .
the light generated in the scintillator is collected on both sides of the module by wavelength - shifter bars , allowing a coordinate measurement based on knowledge of the attenuation length in the scintillator .
the light is converted into an electronic signal by photomultiplier tubes .
the cells are arranged into towers consisting of @xmath144 emc cells , a hac1 cell and a hac2 cell ( in fcal and bcal ) .
the transverse dimensions of the towers in fcal are @xmath145 @xmath142 .
one tower is absent at the center of the fcal and rcal to allow space for passage of the beams .
the outer boundary of the inner ring of fcal towers , used to define a fiducial cut for the jet reconstruction , defines a box of @xmath146 @xmath142 . under test beam conditions , the cal has energy resolutions of @xmath147 for positrons hitting the center of a calorimeter cell and @xmath148 for single hadrons , where energies are in gev . in the zeus detector ,
the energy measurement is affected by the energy loss in the material between the interaction point and the calorimeter . for the events selected in this analysis ,
the positrons predominantly strike the bcal , while the jets hit the fcal .
the in - situ positron - energy resolution in the bcal has been determined to average @xmath149 while the jet - energy resolution in the fcal averages @xmath150 .
the jet - energy resolution was determined by comparing reconstructed jet energies in the calorimeter with the total energy of the particles in the hadronic final - state using monte carlo simulation , and therefore includes small contributions from the jet - finding algorithm . in the reconstruction of the positron and jet energies ,
corrections were applied for inactive materials located in front of the calorimeter and for non - uniformities in the calorimeter response @xcite . for the high energies important in this analysis ,
the overall energy scale is known to @xmath151% for positrons in bcal and @xmath152% for hadrons in fcal and bcal .
the electromagnetic energy scale was determined by a comparison with momentum measurements in the central tracking detector ( using lower - energy electrons and positrons ) .
its linearity was checked with energies reconstructed from the double angle ( da ) method @xcite .
the hadronic - energy scales in the fcal and bcal were determined by using transverse - momentum balance in nc dis events .
the angular reconstruction was performed using a combination of tracking and calorimeter information . from monte carlo studies ,
the polar - angle resolutions were found to be @xmath153 mrad for positrons and approximately ( @xmath154 ) mrad for jets with energies above @xmath155 gev .
the luminosity was measured from the rate of the bremsstrahlung process @xmath156
@xcite , and has an uncertainty of @xmath157% .
the zeus coordinate system is right - handed and centered on the nominal interaction point , with the @xmath158 axis pointing in the direction of the proton beam ( forward ) and the @xmath159 axis pointing horizontally toward the center of hera .
the polar angle @xmath160 is defined with respect to the @xmath158 axis .
the events of interest with large @xmath0-jet mass contain a final - state positron at a large angle and of much higher energy than that of the incident positron beam , as well as one or more energetic jets .
the only important sm source of such events is nc scattering with large @xmath161 .
other potential backgrounds , such as high transverse - energy ( @xmath162 ) photoproduction , were determined to be negligible .
the following requirements selected events of the desired topology : * a reconstructed event vertex was required in the range @xmath163 cm . * the total transverse energy , @xmath164 ,
was required to be at least @xmath165 gev . *
an identified @xcite positron was required with energy @xmath166 gev , located either in the fcal or bcal .
the positron was required to be well - contained in the bcal or fcal and not to point to the bcal / fcal interface , at approximately @xmath167 .
positrons within @xmath168 cm of the boundary between adjacent bcal modules , as determined by tracking information , were also discarded to remove showers developing in the wavelength - shifter bars . * a hadronic jet with transverse momentum @xmath169 gev , located in a region of good containment , was required .
the jets were reconstructed using the longitudinally - invariant @xmath170-clustering algorithm @xcite in the inclusive mode @xcite . only jets with a reconstructed centroid outside the inner ring of fcal towers were considered . in events where multiple jets were reconstructed ,
the jet with highest transverse momentum was used .
after all cuts , 12% of the events had more than one jet , both in the data and monte carlo simulation ( see below ) .
the @xmath162 cut , the jet - containment cut and the positron - containment cut define the available kinematic region for further analysis , as shown in fig .
[ fig : acceptance ] .
the jet containment cut , in particular , limits the values of @xmath171 that can be measured at the highest @xmath0-jet masses .
because most such events have @xmath171 near @xmath151 , the acceptance for nc dis events ( with @xmath172 gev ) falls below @xmath173% for masses beyond 220 gev . in the region allowed by the cuts shown in fig .
[ fig : acceptance ] , the acceptance is typically @xmath174% .
a total of @xmath175 events remained after applying all cuts , compared to @xmath176 events predicted by the nc monte carlo simulation based on the measured luminosity of @xmath2 pb@xmath3 ( the sources of uncertainty on the expected number of events are described in sect .
[ sec : systematics ] ) .
the @xmath162 distributions for data and nc simulation are compared in fig .
[ fig : compare]a .
the positron transverse - momentum ( @xmath177 ) spectrum , jet transverse - momentum ( @xmath178 ) spectrum and the ratio @xmath179 , where @xmath180 is the transverse momentum of the hadronic system , are shown in figs .
[ fig : compare](b - d ) , respectively .
the missing transverse momentum , , and the longitudinal momentum variable , @xmath181 , are compared in figs .
[ fig : compare](e , f ) .
the global properties of the events are well reproduced by the simulation .
the sm deep inelastic scattering events were simulated using the heracles 4.5.2 @xcite program with the django 6 version 2.4 @xcite interface to the hadronization programs . in heracles , corrections for initial- and final - state electroweak radiation , vertex and propagator corrections , and two - boson exchange are included .
the nc dis hadronic final state was simulated using the meps model of lepto 6.5 @xcite , which includes order @xmath182 matrix elements and models of higher - order qcd radiation . as a systematic check ,
the nc final state was simulated using the color - dipole model of ariadne 4.08 @xcite .
the cteq4 parton - distribution set @xcite was used to evaluate the expected number of events from nc dis scattering .
the leptoquark events were generated using pythia 6.1 @xcite .
this program takes into account the finite width of the lq , but only includes the s - channel diagrams .
initial- and final - state qcd radiation from the quark and the effect of lq hadronization before decay are taken into account , as are initial- and final - state qed radiation from the positron .
the generated events were input into a geant - based @xcite program which simulated the response of the zeus detector .
the trigger and offline processing requirements applied to the data were applied to the simulated events .
the luminosity of the nc monte carlo samples ranges from @xmath183 pb@xmath3 at @xmath184 gev@xmath185 to @xmath186 pb@xmath3 at @xmath187 gev@xmath185 .
the mass of each @xmath0-jet pair was reconstructed from the measured energies and angles of the positron and jet as described by eq .
[ eq : mass ] .
this formula makes no correction for the finite jet mass .
possible mass shifts and the resolutions for resonant lepton - hadron states were estimated from pythia .
narrow scalar lq events in the mass range @xmath189 gev were simulated .
the mean mass for reconstructed events was found to be within 6% of the generated value , while the peak position as determined by a gaussian fit was typically lower than the generated value by only 1% . the average mass resolution , determined from a gaussian fit to the peak of the reconstructed mass spectrum , ranged from 5.5% to 3% for masses from 150 to 290 gev .
the rms of the distribution was typically twice as large .
the positron scattering angle in the @xmath0-jet rest frame , @xmath188 , was reconstructed as the angle between the incoming and outgoing positron directions in this frame .
these directions were determined by performing a lorentz transformation using the measured positron and jet energies and angles in the laboratory frame .
the resolution in @xmath190 near @xmath191 , as determined from a gaussian fit , was @xmath192 degrading to @xmath193 as @xmath194 decreases .
the shift in @xmath190 was less than @xmath192 for both the nc mc and the leptoquark mc . in order to determine limits on leptoquark and squark production ,
the mass of the electron - hadron system was reconstructed by the constrained - mass method .
this method reconstructs the @xmath0-hadron mass as @xmath195 where @xmath196 is the sum of the energy and @xmath197 contributions from the positron and all jets satisfying @xmath198 gev and pseudorapidity @xmath199 ( with the highest @xmath200 jet required to be outside the fcal inner ring ) .
the @xmath201 cut removes contributions from the proton remnant .
the constraints @xmath202 and @xmath203 , which are satisfied by fully contained events , have been assumed in arriving at this equation . when using this mass - reconstruction method , events with measured @xmath204 gev were removed to avoid large initial - state qed radiation .
the @xmath205 method gave , on average , improved resolution over the @xmath73 method for narrow lq mc events .
the improved resolution occurred at smaller @xmath74 ( for @xmath206 the mass resolution determined from a gaussian fit to the reconstructed mass distribution for @xmath207 gev was about @xmath168% in the @xmath205 method and @xmath208% for the @xmath73 method ) ; at the larger @xmath74 values where nc dis events are concentrated , the resolutions of the two methods were similar ( about @xmath208% for @xmath207 gev ) .
the @xmath205 method relies on constraints which do not necessarily apply to a resonant state whose properties can not be anticipated in detail .
we therefore choose to use the @xmath73 method as our primary search method .
the @xmath205 method is used in the limit - setting procedure .
the reconstructed values of @xmath73 are plotted versus @xmath171 for the selected events in fig .
[ fig : cost_mej ] .
most of the events are concentrated at large @xmath74 and small mass , as expected from standard model nc scattering .
the five events indicated as open circles are from data taken in 1994 - 96 , with total luminosity @xmath210 pb@xmath3 .
they were the subject of a previous publication @xcite . in this earlier analysis ,
the kinematic variables were reconstructed with the da method .
the five events also stand out with the @xmath73 reconstruction technique .
the average value of @xmath73 for these events is @xmath211 gev , or @xmath212 gev less than the corresponding mass calculated previously via @xmath213 , where @xmath214 is the estimator of bjorken-@xmath68 calculated with the da method .
this mass shift is compatible with expectations based on resolution and initial state radiation effects . with the present luminosity of @xmath2 pb@xmath3
, 7 events are observed in the region of @xmath215 gev and @xmath216 , where @xmath217 events are expected . the @xmath73 spectrum for events with @xmath218 gev is shown in fig . [ fig : mej]a on a logarithmic scale .
the high - mass part of the spectrum is shown on a linear scale in the inset .
the predicted number of events ( @xmath219 ) from nc processes is shown as the histogram .
the ratio of the measured mass spectrum to the expectation is shown in fig .
[ fig : mej]b .
the shaded band indicates the systematic uncertainty on the expectations .
the uncertainty on @xmath219 varies with mass from @xmath212% at 100 gev up to 30% at 250 gev .
the most important uncertainties are on the energy scale and the jet position .
the nc dis cross section given in eq .
[ eq : dis ] ( neglecting @xmath85 ) can be rewritten in terms of the @xmath220 invariant mass , @xmath221 , and the polar angle of the outgoing struck quark in the laboratory frame , @xmath222 : @xmath223\;\ ; .\ ] ] the mass dependence is very steep .
in addition to the explicit @xmath224 dependence , there is also a strong suppression of high masses implicit in the structure functions .
an incorrect energy scale will produce a shift in the mass spectrum and potentially a significant error in the number of expected events at a given mass .
the dependence on the quark angle is also steep , approximately @xmath225 at small @xmath222 .
the number of events passing the jet fiducial cut is therefore strongly dependent on the accuracy of the jet position reconstruction .
the jet fiducial - volume cut requires the highest-@xmath200 jet to point outside the inner ring of fcal towers .
many distributions from data and mc were compared to search for possible systematic biases .
the dominant sources of uncertainty are itemized below in order of decreasing importance : 1 .
knowledge of the calorimeter energy scales : + the scale uncertainties discussed in sect .
[ sec : conditions ] are @xmath151% for bcal positrons and @xmath152% for hadrons , leading to an uncertainty of @xmath226% in @xmath219 at @xmath227 gev .
2 . uncertainties in the simulation of the hadronic energy flow , including simulation of the proton remnant , the energy flow between the struck quark and proton remnant , and possible detector effects in the innermost calorimeter towers : + many distributions of data and mc were compared and no important systematic differences were found .
figure [ fig : eflow ] shows the fraction of the jet energy in the inner ring of fcal towers associated with the highest @xmath200 jet as a function of @xmath228 .
this is shown for all events in fig .
[ fig : eflow]a , as well as for those with @xmath229 gev in fig .
[ fig : eflow]b . for the highest @xmath228 values considered
, this ratio is about @xmath210% .
the energy located in the innermost towers of the fcal and not associated with the highest @xmath200 jet is shown in fig .
[ fig : eflow]c , d , and compared to the mc simulation .
no large differences are seen between data and mc ( the lowest @xmath230 bin in fig [ fig : eflow]d contains only five data events ) .
the innermost towers of the fcal have a larger uncertainty in the energy scale than the rest of the fcal owing to their slightly different construction and proximity to the beam .
the energy in these cells has been varied by @xmath231% . as a test of the simulation of the forward energy flow
, the ariadne mc has been used instead of the lepto mc . these tests yielded variations in @xmath219 of 13% at @xmath232 gev .
3 . uncertainty in the parton density functions : + the parton density functions were estimated as in @xcite , and led to an uncertainty of 5% in @xmath219 at @xmath232 gev .
4 . uncertainties in the acceptance : + the alignment of the fcal was determined to better than @xmath92 mm , and various jet position reconstruction algorithms were compared .
these studies yielded an uncertainty of 2% in @xmath219 .
5 . uncertainties in the energy resolution functions .
+ these were studied by comparing tracking information with calorimeter information for individual events , as well as by comparing different reconstruction methods .
the mc energies were smeared by additional amounts to represent these uncertainties , leading to a variation of less than @xmath92% in @xmath219 .
other uncertainties include positron finding efficiency , luminosity determination , vertex simulation , multijet production rates , and hadronization simulation .
these were found to be small in comparison to the items listed above .
the overall systematic uncertainty was obtained by summing the contributions from all these sources in quadrature . the data in fig .
[ fig : mej ] are in good agreement with the sm expectations up to @xmath233 gev .
some excess is seen at higher masses . for @xmath229 gev
, @xmath234 events were observed in the data , while @xmath235 events are expected
. a careful study of individual events in this mass region uncovered no signs of reconstruction errors .
rather , the events always contain clear examples of a high - energy positron ( typically @xmath236 gev ) near @xmath237 and a high - energy jet ( typically @xmath238 gev ) in the forward direction ( 2 events have a second jet , in accord with nc dis monte carlo expectations ) .
the distributions shown in fig .
[ fig : compare ] for all the data are restricted to the events with @xmath229 gev in fig .
[ fig : compare_210 ] . whereas the shapes of the distributions are similar , the data lie systematically above the mc , which is normalized to the integrated luminosity .
the events with large @xmath73 have characteristics similar on average to nc dis events .
in particular , the @xmath239 projection of the events with @xmath229 gev is shown in fig .
[ fig : cost ] and compared to the mc expectations for neutral current dis ( solid histogram ) .
the expectations for narrow s - channel scalar and vector lq production are also shown for comparison . for @xmath102 lqs with @xmath240
, the u - channel and interference terms would not significantly affect these expectations .
the shape of the data and nc mc @xmath74 distributions are qualitatively similar , peaking at high values of @xmath190 . in summary
, there is some excess of events with @xmath229 gev above the standard model predictions .
the probability of observing such an excess depends strongly on possible systematic biases .
the most important of these are biases in the energy scales . as a test
, many mc experiments were generated where the jet energy scale was shifted by + 2% and the electron energy scale by + 1% .
a window of width @xmath241 , where @xmath242 is the mass resolution at mass @xmath73 , was moved over the accessible mass range . for each simulated experiment ,
the number of observed events within the mass window was compared with the nominal expectations as a function of @xmath73 , seeking the excess which gave the largest statistical significance .
the same procedure was applied to the data . as a result , it was found that 5% of the simulated experiments would observe , somewhere in the mass spectrum , an excess of statistical significance at least as large as the one found in the data .
the excess is therefore not statistically compelling .
furthermore , the events have the characteristics of neutral current scattering .
limits are therefore set on the production of narrow scalar or vector states , as discussed below .
limits are set on the production cross section times branching ratio into positron+jet , @xmath243 , for a narrow scalar or vector state . for definiteness , limits on coupling strength versus mass for @xmath102 leptoquarks
are presented , as well as limits on @xmath244 versus mass for scalar states coupling to @xmath245 or @xmath246 quarks , such as @xmath75 squarks .
the limits are extracted for @xmath247 , allowing the use of the narrow - width approximation assumed in eq .
[ eq : nwa ] .
the @xmath205 mass reconstruction method was used to set limits as described in sect .
[ sec : reconstruction ] .
the positron fiducial cuts were removed since this method is less sensitive to the positron - energy measurement , while the cut @xmath248 gev was applied to reduce radiative effects . the mass spectrum reconstructed with this technique is shown in fig .
[ fig : mcj]a . in total , 8026 events passed all selection cuts while 7863 events are predicted by the mc .
the leptoquark mc described in sect . [ sec : mc ] was used to determine the event selection efficiency and the acceptance of the fiducial cuts , as well as to estimate the mass resolution .
this mc and the nc background simulation were used to calculate an optimal bin width , @xmath249 , for each @xmath205 , and optimal @xmath250 range , @xmath251 , to obtain on average the best limits on lq couplings .
the bin widths were typically 20 gev .
the values of @xmath252 for setting limits range from @xmath253 to @xmath254 for vector leptoquarks with masses between @xmath189 gev , and from @xmath255 to @xmath254 for scalar leptoquarks in the same mass range .
the mass spectrum after applying the optimal @xmath74 cut for the scalar search is shown in fig .
[ fig : mcj]b .
no significant deviations from expectations are seen after applying this cut .
the 95% confidence level ( cl ) limits on @xmath243 were obtained directly from the observed number of data events with @xmath256 in the particular mass window @xcite .
the procedure described in @xcite was extended to include the systematic uncertainties in the numbers of predicted events .
this was found to have negligible effect on the limits .
the limits for a narrow scalar or vector state are shown in fig .
[ fig : csl_mejet ] .
these limits lie between @xmath151 and @xmath255 pb as the mass increases from @xmath8 to @xmath257 gev .
the 95% cl exclusion limits for different species of lq are given in the coupling versus mass plane in fig .
[ fig : lqlimits ] .
the limits exclude leptoquarks with coupling strength @xmath258 for masses up to 280 gev for specific types of f=0 leptoquarks .
the h1 collaboration has recently published similar limits @xcite . in fig .
[ fig : lqlimits ] , the zeus results are compared to recent limits from opal . at lep
@xcite , sensitivity to a high - mass lq arises from effects of virtual lq exchange on the hadronic cross section . the hera and lep limits are complementary to tevatron limits @xcite , which are independent of the coupling @xmath259 .
the limits by d0 ( cdf ) extend up to @xmath260 gev for a scalar lq with 100% branching ratio to @xmath93 .
the d0 limits are shown as vertical lines in fig .
[ fig : lqlimits ] .
the tevatron limits for vector lqs are model dependent @xcite , but are expected to be considerably higher than for scalar lqs .
the zeus limits presented in fig . [ fig : lqlimits ] can also be applied to any narrow state which couples to a positron and a @xmath245 or @xmath246 quark and with unknown branching ratio to @xmath0-jet(s ) .
these states correspond to the leptoquark types as labelled in the figure . for these states ,
the limits are on the quantity @xmath244 .
examples of scalar states for which these limits apply are @xmath75-squarks ( e.g.,the limit on the @xmath261 lq can be read as a limit on the @xmath262 r - parity - violating coupling ) .
data from @xmath2 pb@xmath3 of @xmath263 collisions at a center - of - mass energy of 300 gev have been used to search for a resonance decaying into @xmath0-jet .
the invariant mass of the @xmath0-jet pair was calculated directly from the measured energies and angles of the positron and jet .
this approach makes no assumptions about the production mechanism of such a state .
the observed mass spectrum is in good agreement with standard model expectations up to @xmath0-jet masses of about @xmath4 gev . above this mass ,
some excess is seen .
the angular distribution of these events is typical of high-@xmath5 neutral current events and does not give convincing evidence for the presence of a narrow scalar or vector state . by applying restrictions on the decay angle to optimize sensitivity to a narrow state in the presence of nc background
, limits have been derived on the cross section times decay branching fraction for a scalar or vector state decaying into positron and jet(s ) .
these limits can be interpreted , for example , as limits on leptoquark or r - parity - violating squark production .
limits on the production of leptoquarks and squarks are presented in the coupling strength versus mass plane . at a coupling strength @xmath7 ,
new states are ruled out at 95% confidence level for masses between @xmath8 and @xmath9 gev .
we thank the desy directorate for their strong support and encouragement , and the hera machine group for their diligent efforts .
we are grateful for the support of the desy computing and network services .
the design , construction and installation of the zeus detector have been made possible by the ingenuity and effort of many people from desy and home institutes who are not listed as authors .
it is also a pleasure to thank w. buchmller , r. rckl and m. spira for useful discussions .
j. l. hewett , ` single squark production at hera via r parity violating interactions ' , proc . of the 1990 dpf summer study on high - energy physics , snowmass , colorado , usa , eds .
e. berger , world scientific ( 1992 ) | the @xmath0-jet invariant mass spectrum produced in the reaction @xmath1 has been studied at a center - of - mass energy of 300 gev .
the data were collected using the zeus detector operating at the hera collider , and correspond to an integrated luminosity of @xmath2 pb@xmath3 .
the observed mass spectrum is in good agreement with standard model expectations up to an @xmath0-jet mass of @xmath4 gev . above this mass ,
some excess is seen .
the angular distribution of these events is typical of high-@xmath5 neutral current events and does not give convincing evidence for the presence of a narrow scalar or vector state .
limits are presented on the product of cross section and branching ratio for such a state and are interpreted as limits on leptoquark or r - parity - violating squark production .
specific leptoquark types are ruled out at @xmath6% confidence level for coupling strength @xmath7 for masses between @xmath8 and @xmath9 gev .
-1.5 cm -0.3 cm -0.3 cm 16.cm 0.cm 0.3 cm plus0.05 cm minus0.05 cm desy
00 - 023 + february 2000 + -1.cm -0.3 cm -0.3
cm 16.cm 0.cm 0.3 cm plus0.05 cm minus0.05 cm the zeus collaboration j. breitweg , s. chekanov , m. derrick , d. krakauer , s. magill , b. musgrave , a. pellegrino , j. repond , r. stanek , r. yoshida + _ argonne national laboratory , argonne , il , usa _ @xmath10 m.c.k .
mattingly + _ andrews university , berrien springs , mi , usa _
g. abbiendi , f. anselmo , p. antonioli , g. bari , m. basile , l. bellagamba , d. boscherini@xmath11 , a. bruni , g. bruni , g. cara romeo , g. castellini@xmath12 , l. cifarelli@xmath13 , f. cindolo , a. contin , n. coppola , m. corradi , s. de pasquale , p. giusti , g. iacobucci , g. laurenti , g. levi , a. margotti , t. massam , r. nania , f. palmonari , a. pesci , a. polini , g. sartorelli , y. zamora garcia@xmath14 , a. zichichi + _ university and infn bologna , bologna , italy _
@xmath15 c. amelung , a. bornheim , i. brock , k. cobken , j. crittenden , r. deffner , h. hartmann , k. heinloth , e. hilger , p. irrgang , h .-
p .
jakob , a. kappes , u.f .
katz , r. kerger , e. paul , h. schnurbusch , + a. stifutkin , j. tandler , k.ch .
voss , a. weber , h. wieber + _ physikalisches
institut der universitt bonn , bonn , germany _
@xmath16 d.s .
bailey , o. barret , n.h .
brook@xmath17 , b. foster@xmath18 , g.p .
heath , h.f .
heath , j.d .
mcfall , d. piccioni , e. rodrigues , j. scott , r.j .
tapper + _ h.h .
wills physics laboratory , university of bristol , bristol , u.k .
_ @xmath19 m. capua , a. mastroberardino , m. schioppa , g. susinno + _ calabria university , physics dept.and infn , cosenza , italy _ @xmath15 h.y .
jeoung , j.y .
kim , j.h .
lee , i.t .
lim , k.j .
ma , m.y .
pac@xmath20 + _ chonnam national university , kwangju , korea _
@xmath21 a. caldwell , w. liu , x. liu , b. mellado , s. paganis , r. sacchi , s. sampson , f. sciulli + _ columbia university , nevis labs .
, irvington on hudson , n.y . , usa _
@xmath22 j. chwastowski , a. eskreys , j. figiel , k. klimek , k. olkiewicz , k. piotrzkowski@xmath23 , m.b .
przybycie , p. stopa , l. zawiejski + _ inst . of nuclear physics ,
cracow , poland _
@xmath24 l. adamczyk , b. bednarek , k. jele , d. kisielewska , a.m. kowal , t. kowalski , m. przybycie , + e. rulikowska - zarbska , l. suszycki , d. szuba + _ faculty of physics and nuclear techniques , academy of mining and metallurgy , cracow , poland _
@xmath24 a. kotaski + _ jagellonian univ . ,
dept . of physics ,
cracow , poland _
@xmath25 l.a.t .
bauerdick , u. behrens , j.k .
bienlein , c. burgard@xmath26 , k. desler , g. drews , , u. fricke , f. goebel , p. gttlicher , r. graciani , t. haas , w. hain , g.f .
hartner , d. hasell@xmath27 , k. hebbel , k.f .
johnson@xmath28 , m. kasemann@xmath29 , w. koch , u. ktz , h. kowalski , l. lindemann@xmath30 , b. lhr , m. milite , t. monteiro@xmath23 , m. moritz , d. notz , f. pelucchi , m.c .
petrucci , m. rohde , p.r.b .
saull , a.a .
savin , , f. selonke , m. sievers , s. stonjek , e. tassi , g. wolf , u. wollmer , + c. youngman , + _ deutsches elektronen - synchrotron desy , hamburg , germany _ c. coldewey , h.j .
grabosch , , a. meyer , , p.b . straub + _ desy zeuthen , zeuthen , germany _ g. barbagli , e. gallo , p. pelfer + _ university and infn , florence , italy _ @xmath15 g. maccarrone , l. votano + _ infn , laboratori nazionali di frascati , frascati , italy _
@xmath15 a. bamberger , a. benen , s. eisenhardt@xmath31 , p. markun , h. raach , s. wlfle + _ fakultt fr physik der universitt freiburg i.br .
, freiburg i.br .
, germany _
@xmath16 p.j .
bussey , a.t .
doyle , s.w .
lee , n. macdonald , g.j .
mccance , d.h .
saxon , l.e .
sinclair , + i.o .
skillicorn , r. waugh + _ dept . of physics and astronomy , university of glasgow , glasgow ,
u.k .
_ @xmath19 i. bohnet , n. gendner , u. holm , a. meyer - larsen , h. salehi , k. wick + _ hamburg university , i. institute of exp .
physics , hamburg , germany _
@xmath16 d. dannheim , a. garfagnini , i. gialas@xmath32 , l.k .
gladilin@xmath33 , d. kira@xmath34 , r. klanner , e. lohrmann , g. poelz , f. zetsche + _ hamburg university , ii .
institute of exp .
physics , hamburg , germany _
@xmath16 r. goncalo , k.r .
long , d.b .
miller , a.d .
tapper , r. walker + _ imperial college london , high energy nuclear physics group , london , u.k .
_ @xmath19 u. mallik + _ university of iowa , physics and astronomy dept . ,
iowa city , usa _
@xmath10 p. cloth , d. filges
+ _ forschungszentrum jlich , institut fr kernphysik , jlich , germany _ t. ishii , m. kuze , k. nagano , k. tokushuku@xmath35 , s. yamada ,
y. yamazaki + _ institute of particle and nuclear studies , kek , tsukuba , japan _
@xmath36 s.h .
ahn , s.b .
lee , s.k .
park + _ korea university , seoul , korea _
@xmath21 h. lim , i.h . park , d. son + _ kyungpook national university , taegu , korea _
@xmath21 f. barreiro , g. garca , c. glasman@xmath37 , o. gonzalez , l. labarga , j. del peso , i. redondo@xmath38 , j. terrn + _ univer .
autnoma madrid , depto de fsica terica , madrid , spain _
@xmath39 m. barbi , f. corriveau , d.s .
hanna , a. ochs , s. padhi , m. riveline , d.g .
stairs , m. wing + _ mcgill university , dept . of physics , montral , qubec ,
canada _
@xmath40 @xmath41 t. tsurugai + _ meiji gakuin university , faculty of general education , yokohama , japan _
v. bashkirov@xmath42 , b.a .
dolgoshein + _ moscow engineering physics institute , moscow , russia _
@xmath43 r.k .
dementiev , p.f .
ermolov , yu.a .
golubkov , i.i .
katkov , l.a .
khein , n.a .
korotkova , + i.a .
korzhavina , v.a .
kuzmin , o.yu .
lukina , a.s .
proskuryakov , l.m .
shcheglova , a.n .
solomin , + n.n .
vlasov , s.a .
zotkin + _ moscow state university , institute of nuclear physics , moscow , russia _ @xmath44 c. bokel , m. botje , n. brmmer , j. engelen , s. grijpink , e. koffeman , p. kooijman , s. schagen , a. van sighem , h. tiecke , n. tuning , j.j .
velthuis , j. vossebeld , l. wiggers , e. de wolf + _ nikhef and university of amsterdam , amsterdam , netherlands _
@xmath45 d. acosta@xmath46 , b. bylsma , l.s .
durkin , j. gilmore , c.m .
ginsburg , c.l .
kim , t.y .
ling + _ ohio state university , physics department , columbus , ohio , usa _
@xmath10 s. boogert , a.m. cooper - sarkar , r.c.e .
devenish , j. groe - knetter@xmath47 , t. matsushita , o. ruske , + m.r .
sutton , r. walczak + _ department of physics , university of oxford , oxford u.k . _ @xmath19 a. bertolin , r. brugnera , r. carlin , f. dal corso , u. dosselli , s. dusini , s. limentani , m. morandin , m. posocco , l. stanco , r. stroili , c. voci + _ dipartimento di fisica dell universit and infn , padova , italy _ @xmath15 l. iannotti@xmath48 , b.y .
oh , j.r .
okrasiski , w.s .
toothacker , j.j .
whitmore + _ pennsylvania state university , dept . of physics ,
university park , pa , usa _
@xmath22 y. iga + _ polytechnic university , sagamihara , japan _
@xmath36 g. dagostini , g. marini , a. nigro + _ dipartimento di fisica , univ .
la sapienza and infn , rome , italy _ @xmath49 c. cormack , j.c .
hart , n.a .
mccubbin , t.p .
shah + _ rutherford appleton laboratory , chilton , didcot , oxon , u.k .
_ @xmath19 d. epperson , c. heusch , h.f .- w .
sadrozinski , a. seiden , r. wichmann , d.c .
williams + _ university of california , santa cruz , ca , usa _
@xmath10 n. pavel + _ fachbereich physik der universitt - gesamthochschule siegen , germany _
@xmath16 h. abramowicz@xmath50 , s. dagan@xmath51 , s. kananov@xmath51 , a. kreisel , a. levy@xmath51 + _ raymond and beverly sackler faculty of exact sciences , school of physics , tel - aviv university , +
tel - aviv , israel _
@xmath52 t. abe , t. fusayasu , k. umemori , t. yamashita + _ department of physics , university of tokyo , tokyo , japan _ @xmath36 r. hamatsu , t. hirose , m. inuzuka , s. kitamura@xmath53 , t. nishimura + _ tokyo metropolitan university , dept . of physics , tokyo , japan _
@xmath36 m. arneodo@xmath54 , n. cartiglia , r. cirio , m. costa , m.i .
ferrero , s. maselli , v. monaco , c. peroni , m. ruspa , a. solano , a. staiano + _ universit di torino , dipartimento di fisica sperimentale and infn , torino , italy _ @xmath15 m. dardo + _ ii faculty of sciences , torino university and infn - alessandria , italy
_ @xmath15 d.c .
bailey , c .- p .
fagerstroem , r. galea , t. koop , g.m .
levman , j.f .
martin , r.s .
orr , s. polenz , a. sabetfakhri , d. simmons + _ university of toronto , dept . of physics , toronto , ont .
, canada _ @xmath55 j.m .
butterworth , c.d .
catterall , m.e .
hayes , e.a .
heaphy , t.w .
jones , j.b .
lane , b.j .
west + _ university college london , physics and astronomy dept . ,
london , u.k .
_ @xmath19 j. ciborowski , r. ciesielski , g. grzelak , r.j .
nowak , j.m .
pawlak , r. pawlak , b. smalska , + t. tymieniecka , a.k .
wrblewski , j.a .
zakrzewski , a.f .
arnecki + _ warsaw university , institute of experimental physics , warsaw , poland _
@xmath24 m. adamus , t. gadaj + _ institute for nuclear studies , warsaw , poland _
@xmath24 o. deppe , y. eisenberg@xmath51 , d. hochman , u. karshon@xmath51 + _ weizmann institute , department of particle physics , rehovot , israel _ @xmath56 w.f .
badgett , d. chapin , r. cross , c. foudas , s. mattingly , d.d .
reeder , w.h .
smith , a. vaiciulis@xmath57 , t. wildschek , m. wodarczyk + _ university of wisconsin , dept . of physics ,
madison , wi , usa _
@xmath10 a. deshpande , s. dhawan , v.w .
hughes + _ yale university , department of physics , new haven , ct , usa _
@xmath10 s. bhadra , c. catterall , j.e .
cole , w.r .
frisken , r. hall - wilton , m. khakzad , s. menary , w.b .
schmidke + _ york university , dept . of physics , toronto , ont . ,
canada _
@xmath55 @xmath58 now visiting scientist at desy + @xmath59 also at iroe florence , italy + @xmath60 now at univ . of salerno and infn napoli , italy + @xmath61 supported by worldlab , lausanne , switzerland + @xmath62 pparc advanced fellow + @xmath63 also at university of hamburg , alexander von humboldt research award + @xmath64 now at dongshin university , naju , korea + @xmath65 now at cern + @xmath66 now at barclays capital plc , london + @xmath27 now at massachusetts institute of technology , cambridge , ma , usa + @xmath28 visitor from florida state university + @xmath29 now at fermilab , batavia , il , usa + @xmath30 now at sap a.g . , walldorf , germany + @xmath31 now at university of edinburgh , edinburgh , u.k
. + @xmath32 visitor of univ . of crete ,
greece , partially supported by daad , bonn - kz .
+ @xmath33 on leave from msu , supported by the gif , contract i-0444 - 176.07/95 + @xmath34 supported by daad , bonn - kz .
+ @xmath35 also at university of tokyo + @xmath37 supported by an ec fellowship number erbfmbict 972523 + @xmath38 supported by the comunidad autonoma de madrid + @xmath42 now at loma linda university , loma linda , ca , usa + @xmath46 now at university of florida , gainesville , fl , usa + @xmath47 supported by the feodor lynen program of the alexander von humboldt foundation + @xmath48 partly supported by tel aviv university + @xmath50 an alexander von humboldt fellow at university of hamburg + @xmath51 supported by a minerva fellowship + @xmath53 present address : tokyo metropolitan university of health sciences , tokyo 116 - 8551 , japan + @xmath54 now also at universit del piemonte orientale , i-28100 novara , italy + @xmath57 now at university of rochester , rochester , ny , usa + [ cols= " > , < " , ] -1.5 cm -0.3 cm -0.3 cm 16.cm 0.cm 0.3 cm plus0.05 cm minus0.05 cm |
principal component analysis ( pca ) , also known as the karhunen - love transform or the hotelling transform , is a well - known method in machine learning , signal processing and statistics @xcite .
pca essentially builds an orthogonal transform to convert a set of observations of possibly correlated variables into a set of linearly uncorrelated variables .
pca has been used for manifold description and dimensionality reduction in a wide range of applications because of its simplicity , energy compaction , intuitive interpretation , and invertibility .
nevertheless , pca is hampered by data exhibiting nonlinear relations . in this paper , we present a nonlinear generalization of pca that , unlike other alternatives , keeps all the above mentioned appealing properties of pca . in recent years
, several dimensionality reduction methods have been proposed to deal with manifolds that can not be linearly described ( see @xcite for a comprehensive review ) : the approaches proposed range from local methods @xcite , to kernel - based and spectral decompositions @xcite , neural networks @xcite , and projection pursuit methods @xcite . however , despite the advantages of nonlinear methods , classical pca still remains the most widely used dimensionality reduction technique in real applications .
this is because pca : 1 ) is easy to apply , 2 ) involves solving a convex problem , for which efficient solvers exist , 3 ) identifies features which are easily interpretable in terms of original variables , and 4 ) has a straightforward inverse and out - of - sample extension . the above properties , which are the base of the success of pca , are not always present in the new nonlinear dimensionality reduction methods due either to complex formulations , to the introduction of a number of non - intuitive free parameters to be tuned , to their high computational cost , to their non - invertibility or , in some cases , to strong assumptions about the manifold .
more plausibly , the limited adoption of nonlinear methods in daily practice has to do with the lack of feature and model interpretability . in this regard ,
the usefulness of data description methods is tied to the following properties : 1 .
_ invertibility of the transform .
_ it allows both characterizing the transformed domain and evaluating the quality of the transform . on the one hand
, inverting the data back to the input domain is important to understand the features in physically meaningful units , while analyzing the results in the transformed domain is typically more complicated ( if not impossible ) . on the other hand ,
invertible transforms like pca allow the assessment of the dimensionality reduction errors as simple reconstruction distortion .
_ geometrical interpretation of the manifold .
_ understanding the system that generated the data is the ultimate goal of manifold learning .
inverting the transform is just one step towards knowledge extraction .
geometrical interpretation and analytical characterization of the manifolds give us further insight into the problem .
ideally , one would like to compute geometric properties from the learned model , such as the curvature and torsion of the manifold , or the metric induced by the data .
this geometrical characterization allows to understand the latent parameters governing the system .
it is worth noting that both properties are scarcely achieved in the manifold learning literature .
for instance , _ spectral _ methods do not generally yield intuitive mappings between the original and the intrinsic curvilinear coordinates of the low dimensional manifold . even though a metric can be derived from particular kernel functions @xcite ,
the interpretation of the transformation is hidden behind an implicit mapping function , and solving the pre - image problem is generally not straightforward @xcite . in such cases ,
the application of ( indirect ) evaluation techniques has become a relevant issue for methods leading to non - invertible transforms @xcite .
one could argue that direct and inverse transforms can be alternatively derived from mixtures of local models @xcite .
however , the effect of these local alignment operations in the metric is not trivial . in the same way , explicit geometric descriptions of the manifold , such as the computation of curvatures , is not obvious from other invertible transforms , as autoencoders or deep networks @xcite . in this paper
, we introduce the principal polynomial analysis ( ppa ) , which is a _ nonlinear generalization of pca _ that still shares all its important properties .
ppa is computationally easy as it only relies on matrix inversion and multiplication , and it is robust since it reduces to a series of marginal ( univariate ) regressions .
ppa implements a volume - preserving and invertible map .
not only the features are easy to interpret in the input space but , additionally , the analytical nature of ppa allows to compute classical geometrical descriptors such as curvature , torsion and the induced metric at any point of the manifold .
applying the learned transform to new samples is also as straightforward as in pca .
preliminary versions of ppa were presented in @xcite , and applied to remote sensing in @xcite .
however , those conference papers did not study the analytical properties of ppa ( volume preservation , invertibility , and model geometry ) , nor compared with approaches that follow similar logic like nl - pca .
the proposed ppa method can be motivated by considering the conditional mean of the data .
in essence , pca is optimal for dimensionality reduction in a mean square error ( mse ) sense _ if and only if _ the conditional mean in each principal component is constant along the considered dimension .
hereafter , we will refer to this as the _ conditional mean independence _ assumption .
unfortunately , this symmetry requirement does not apply in general , as many datasets live in non - gaussian and/or curved manifolds .
see for instance the data in fig .
[ fig2 ] ( left ) : the dimensions have a nonlinear relation even after pca rotation ( center ) .
in this situation , the mean of the second principal component given the first principal component can be easily expressed with a parabolic function ( red line ) .
for data manifolds lacking the required symmetry , nonlinear modifications of pca should remove the residual nonlinear dependence .
following the previous intuition , ppa aims to remove the condition mean . left panel in fig .
[ fig2 ] shows the input @xmath0 data distribution , where we highlight a point of interest , @xmath1 .
ppa is a sequential algorithm ( as pca ) that transforms one dimension at each step in the sequence .
the procedure in each step consists of two operations .
the _ first operation _ looks for the best vector for data projection .
even though different possibilities will be considered later ( section 2.3 ) , a convenient choice for this operation is the leading eigenvector of pca .
figure [ fig2][middle ] shows the data after this projection : although the linear dependencies have been removed , there are still relations between the first and the second data dimensions .
the _ second operation _ consists in subtracting the conditional mean to every sample .
the conditional mean is estimated by fitting a curve predicting the residual using the projections estimated by the first operation .
this step , composed of the two operations above , describes the @xmath2-dimensional data along _ one _ curvilinear dimension through ( 1 ) a projection score onto certain leading vector , and ( 2 ) a curve depending on the projection score .
ppa differs from pca in this second operation because it bends the straight line into a curve , thus capturing part of the nonlinear relations between the leading direction and the orthogonal subspace .
since this example is two - dimensional , ppa ends after one step .
however , when there are more dimensions , the two - operations are repeated for the remaining dimensions . at the first step ,
the @xmath3-dimensional information still to be described is the departure from the curve in the subspace orthogonal to the leading vector .
this data of reduced dimension is the input for the next step in the sequence
. the last ppa dimension will be the @xmath4 residual which , in this example , corresponds to the residuals in the second dimension .
the paper is organized as follows .
section 2 formalizes the forward ppa transform and analytically proves that ppa generalizes pca and improves its performance in dimensionality reduction .
the objective function of ppa , its restrictions , and its computational cost are then analyzed .
section 3 studies the main properties of ppa : jacobian , volume preservation , invertibility , and metric . in section 4
we discuss the differences between ppa and related work . in section 5
, we check the generalization of mahalanobis distance using the ppa metric , and its ability to characterize the manifold geometry ( curvature and torsion ) . finally , we report results on standard databases for dimensionality and redundancy reduction .
section 6 concludes the paper .
additionally , the appendix details a step - by - step example of the forward transform .
in this section , we start by reviewing the pca formulation as a deflationary ( or sequential ) method that addresses one dimension at a time . this is convenient since it allows to introduce ppa as the generalization that uses polynomials instead of straight lines in the sequence . given a @xmath2-dimensional centered random variable @xmath1 , the pca transform , @xmath5 , maps data from the input domain , @xmath6 , to a response domain , @xmath7 .
pca can be actually seen as a sequential mapping ( or a set of concatenated @xmath8 transforms ) .
each transform in the sequence explains a single dimension of the input data by computing a single component of the response : @xmath9 and hence the pca transformation can be expressed as : @xmath10 . here
vectors , @xmath11 , and transforms , @xmath12 , refer to the @xmath13-th step of the sequence .
each of these elementary transforms , @xmath14 , acts only on part of the dimensions of the output of the previous transform : the residual , @xmath15 .
subscript @xmath16 refers to the input data so @xmath17 .
this sequential ( deflationary ) interpretation , which is also applicable to ppa as we will see later , is convenient to derive most of the properties of ppa in section 3 . in pca ,
each transform @xmath14 : ( 1 ) @xmath18 , which is the projection of the data coming from the previous step , @xmath15 , onto the unit norm vector @xmath19 ; and ( 2 ) @xmath20 , which are the residual data for the next step , obtained by projecting @xmath15 in the complement space : @xmath21 where @xmath22 is a @xmath23 matrix containing the remaining set of vectors . in pca , @xmath19 is the vector that maximizes the variance of the projected data : @xmath24 \ } } , \label{eq : orthoerror}\ ] ] where @xmath25 represents the set of possible unit norm vectors .
@xmath26 can be any matrix that spans the subspace orthonormal to @xmath19 , and its rows contain @xmath27 orthonormal vectors .
accordingly , @xmath19 and @xmath28 fulfil : @xmath29 which will be referred to as the _ orthonormality relations _ of @xmath19 and @xmath28 in the discussion below . in the @xmath13-th step of the sequence ,
the data yet to be explained is @xmath30 .
therefore , truncating the pca expansion at dimension @xmath13 implies ignoring the information contained in @xmath11 so that the dimensionality reduction error is : @xmath31 = \mathbb{e } [ \|{{\bf x}}_{p } \|_2 ^ 2 ] .
\label{pcaerror}\ ] ] pca is the optimal linear solution for dimensionality reduction in mse terms since eq
. implies minimizing the dimensionality reduction error in eq . due to the orthonormal nature of the projection vectors @xmath19 and @xmath28 .
ppa removes the conditional mean in order to reduce the reconstruction error of pca in eq . . when the data fulfill the _ conditional mean independence _
requirement , the conditional mean at every point in the @xmath19 direction is zero . in this case
, the data vector goes through the means in the subspace spanned by @xmath28 , resulting in a small pca truncation error .
however , this is not true in general ( cf . fig .
[ fig2 ] ) and then the conditional mean @xmath32 \neq 0 $ ] . in order to remove the conditional mean @xmath33 from @xmath11
, ppa modifies the elementary pca transforms in eq . by subtracting an estimation of the conditional mean , @xmath34 : @xmath35 assuming for now that the leading vector , @xmath19 , is computed in the same way as in pca
, ppa only differs from pca in the second operation of each transform @xmath14 ( cf .
. however , this suffices to ensure the superiority of ppa over pca
. we will refer to this particular choice of @xmath19 as the _ pca - based solution _ of ppa . in section 2.3 , we consider more general solutions to optimize the objective function at the cost of facing a non - convex problem . in any case , and independently of the method used to choose @xmath19 , the truncation error in ppa is : @xmath36 .
\label{ppaerror}\ ] ] [ [ estimation - of - the - conditional - mean - at - step - p . ] ] estimation of the conditional mean at step @xmath13 .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the conditional mean can be estimated with any regression method @xmath37 . in this work
, we propose to estimate the conditional mean at each step of the sequence using a polynomial function with coefficients @xmath38 and degree @xmath39 .
hence , the estimation problem becomes : @xmath40 which , in matrix notation is @xmath41 where @xmath42 , and @xmath43^\top$ ] .
note that when considering @xmath44 input examples , we may stack them column - wise in a matrix @xmath45 . in the above
mentioned _ pca - based solution _ , the @xmath13-th step of the ppa sequence starts by computing pca on @xmath46 .
then , we use the first eigenvector of the sample covariance as leading vector @xmath19 , and the remaining eigenvectors as @xmath28 . these eigenvectors are orthonormal ; if a different strategy is used to find @xmath19 , then @xmath28 can be chosen to be any orthonormal complement of @xmath19 ( see section 2.3 ) . from the projections of the @xmath44 samples onto the leading vector ( i.e. from the @xmath44 coefficients @xmath47 with @xmath48 ) , we build the vandermonde matrix @xmath49 , by stacking the @xmath44 column vectors @xmath50 , with @xmath48 . then , the least squares solution for the matrix @xmath51 of coefficients of the polynomial is : @xmath52 where @xmath53 stands for the pseudoinverse operation .
hence , the estimation of the conditional mean for all the samples , column - wise stacked in matrix @xmath54 , is : @xmath55 and the residuals for the next step are , @xmath56 .
summarizing , the extra elements with respect to pca are a single matrix inversion in eq . and the matrix product in eq . .
also note that the estimation of the proposed polynomial is much simpler than fitting a polynomial depending on a natural parameter such as the orthogonal projection on the curve , as one would do according to the classical principal curve definition @xcite . since the proposed objective function in eq
does not estimate distortions orthogonal to the curve but rather those orthogonal to the leading vector @xmath19 , the computation of the projections is straightforward and decoupled from the computation of @xmath51 . the proposed estimation in eq .
consists of @xmath27 separate univariate problems only : this means that ppa needs to fit @xmath27 one - dimensional polynomials depending solely on the ( easy - to - compute ) projection parameter @xmath18 . since lots of samples @xmath57 are typically available , the estimation of such polynomials is usually robust .
the convenience of this decoupling is illustrated in the step - by - step example presented in the appendix .
since we compute @xmath51 using least squares , we obtain three important properties : the ppa error does not depend on the particular selection of the basis @xmath28 if it satisfies the orthonormality relations in eq . . + * proof : * using different basis @xmath58 in the subspace orthogonal to @xmath19 is equivalent to applying an arbitrary @xmath59 rotation matrix , @xmath60 , to the difference vectors expressed in this subspace in eq . : @xmath61 $ ] .
since @xmath62 , the error is independent of this rotation , and hence independent of the basis .
the ppa error is equal to or smaller than the pca error .
+ * proof : * the ppa eqs . and
reduce to pca eq . in the restricted case of @xmath63 .
since , in general , ppa allows for @xmath64 , this implies that @xmath65 .
even though the superiority of ppa over pca in mse terms is clearer when taking @xmath19 as in pca , this property holds in general .
if a better choice for @xmath19 is available , it would reduce the error while having no negative impact in the cost function , since it is independent from the basis @xmath28 chosen ( see _ property 1 _ above ) .
ppa reduces to pca when using first degree polynomials ( i.e. straight lines ) .
+ * proof : * in this particular situation ( @xmath66 ) , the first eigenvector of @xmath46 is the best direction to project onto @xcite .
additionally , when using first degree polynomials , @xmath67 is very simple and @xmath68 can be computed analytically . plugging this particular @xmath68 into eq . , it is easy to see that @xmath63 since the data is centered and @xmath18 is decorrelated of @xmath69 . therefore , when using straight lines @xmath51 vanishes and ppa reduces to pca . [ cols="^,^,^ " , ]
this paper has been partially supported by the spanish mineco under project tin2012 - 38102-c03 - 01 and by the swiss nsf under project pz00p2 - 136827 .
figure [ con_cucharita ] presents a step - by - step example to illustrate how the sequence of ppa curvilinear components and projections are computed on a manifold of well - defined geometry : an helix embedded in a 3d space corrupted with additive gaussian noise which is a usual test case in principal curves @xcite .
data ( in gray ) were sampled from the same helix as in section [ helice ] and noise with standard deviation 0.3 . since @xmath70 ,
ppa consists of a sequence of two transforms ( see eq . ):
@xmath71 ( first row in fig .
[ con_cucharita ] ) and @xmath72 ( second row ) .
a representative sample is highlighted throughout the transform . in this example
we use the pca - based solution .
therefore , the leading vector @xmath73 is the first eigenvector ( biggest eigenvalue ) of the covariance matrix of @xmath74 . in the example
, @xmath73 ( or pc1 , in orange ) , and the vectors pc2 and pc3 ( in green and blue respectively ) constitute the basis @xmath75 .
the first ppa component , @xmath76 , is the projection of the data onto the first leading vector , @xmath77 , in eq .
( orange dots and the circle for the highlighted sample ) . the conditional mean , @xmath78 , is shown decomposed in two subspaces in the top center panel . we will call @xmath79 the conditional mean in the subspace spanned by @xmath73 and pc2 ( green dots ) , and
let @xmath80 be the conditional mean in the subspace spanned by @xmath73 and pc3 ( blue dots ) .
it is obvious the strong non - linear dependence of the conditional mean with @xmath76 , i.e. given the value of @xmath76 ( @xmath73 axis -black line- ) it is easy to predict the value of the data in the orthogonal subspaces ( blue and green dots ) using a non - linear function .
fitting the first ppa polynomial in @xmath81 dimensions with regard to the parameter @xmath76 is equivalent to fitting the polynomials in the @xmath0 subspaces in the center plot ( simple univariate regressions ) .
the polynomials in the @xmath0 subspaces have the coefficients @xmath82 $ ] , and equivalently , @xmath83 ; which are the rows of the matrix @xmath84 .
polynomial coefficients are easy to fit by constructing the vandermonde matrix of degree @xmath85 using @xmath76 , @xmath86^\top$ ] and applying eq . .
this ensures the best fitting in least squares terms .
then , we estimate @xmath79 ( and correspondingly @xmath80 ) using @xmath76 and the weights , eq . :
@xmath87 in the top center panel , the estimated conditional mean , @xmath88^\top$ ] , is represented by the curve ( red ) , while the curve projected in the bottom plane ( green ) and the curve projected in the vertical plane ( blue ) represent the conditional means in the respective subspaces ( @xmath89 and @xmath90 ) .
once the polynomial has been fitted , we can remove @xmath91 from each sample ( second line in eq . ) obtaining the residuals ( departures from the conditional mean ) represented in the top right plot ( yellow dots ) . summarizing the process in the top row ,
the transform @xmath71 , the first principal polynomial ( red curve ) accounts for the first curvilinear dimension of the data .
after @xmath71 , we have @xmath92 dimensions yet to be explained : @xmath93 , at the top right and bottom left plots . the second row of fig .
[ con_cucharita ] reproduces the same steps in the reduced dimension residual : projection onto the first pc in the bottom left plot ( orange dots ) , fitting the polynomial ( in this case , the best cross - validation solution was a second order polynomial , represented by the curve ( red ) in the bottom center plot , and removing the conditional mean so that the residuals ( yellow dots ) are aligned , and projected in the orthogonal subspace .
j. arenas - garca , k. petersen , g. camps - valls , and l.k .
kernel multivariate analysis framework for supervised subspace learning : a tutorial on linear and kernel multivariate methods .
, 30(4):1629 , 2013 .
l. ronan , r. pienaar , g. williams , e.t .
bullmore , t.j .
crow , n. roberts , p.b .
jones , j. suckling , p.c .
intrinsic curvature : a marker of millimeter - scale tangential cortico - cortical connectivity ?
, 21(5 ) : 351366 , 2011 . | this paper presents a new framework for manifold learning based on a sequence of principal polynomials that capture the possibly nonlinear nature of the data . the proposed principal polynomial analysis ( ppa )
generalizes pca by modeling the directions of maximal variance by means of curves , instead of straight lines .
contrarily to previous approaches , ppa reduces to performing simple univariate regressions , which makes it computationally feasible and robust .
moreover , ppa shows a number of interesting analytical properties .
_ first _ , ppa is a volume - preserving map , which in turn guarantees the existence of the inverse .
_ second _ , such an inverse can be obtained in closed form .
invertibility is an important advantage over other learning methods , because it permits to understand the identified features in the input domain where the data has physical meaning .
moreover , it allows to evaluate the performance of dimensionality reduction in sensible ( input - domain ) units .
volume preservation also allows an easy computation of information theoretic quantities , such as the reduction in multi - information after the transform .
_ third _ , the analytical nature of ppa leads to a clear geometrical interpretation of the manifold : it allows the computation of frenet - serret frames ( local features ) and of generalized curvatures at any point of the space . and _ fourth _
, the analytical jacobian allows the computation of the metric induced by the data , thus generalizing the mahalanobis distance .
these properties are demonstrated theoretically and illustrated experimentally .
the performance of ppa is evaluated in dimensionality and redundancy reduction , in both synthetic and real datasets from the uci repository . |
the foundations of our present understanding of _ advection - dominated accretion _ were laid out in a series of papers by narayan & yi ( 1994 , 1995a , b , hereafter ny94 , ny95a , ny95b ) , abramowicz et al .
( 1995 ) and chen et al .
( 1995 ) , although some ideas were anticipated much earlier by ichimaru ( 1977 ) .
the specific abbreviation adaf , which stands for _ advection - dominated accretion flow _ , was introduced by lasota(1996 ) and has become standard in the field .
we review here the application of adafs to accreting black holes ( bhs ) , and in 4 we also consider their application to accreting neutron stars .
the energy equation per unit volume of an accretion flow can be written compactly in the form @xmath4 where @xmath5 is the density , @xmath6 is the temperature , @xmath7 is the entropy per unit mass , @xmath8 is time , @xmath9 is the flow velocity , and @xmath10 and @xmath11 are the heating and cooling rates per unit volume .
this equation states that the rate at which the entropy per unit volume of the gas increases is equal to the heating rate minus the cooling rate .
since any entropy stored in the gas is advected with the flow , the left - hand side of equation ( 1 ) may be viewed as the effective _ advective cooling
_ rate @xmath12 .
equation ( 1 ) can then be rearranged to give @xmath13 which states that the heat energy released by viscous dissipation is partially lost by radiative cooling and partially by advective cooling .
the standard thin accretion disk model ( shakura & sunyaev 1973 ; novikov & thorne 1973 ; frank , king & raine 2002 ) corresponds to the case when the accreting gas is _ radiatively efficient _ , so that we have @xmath14 since the gas cools efficiently , the sound speed is much less than the local keplerian speed @xmath15 and the disk is geometrically thin .
also , the disk radiates about a tenth of the rest mass energy of the accreting gas , the precise fraction depending on the bh spin ( see shapiro & teukolsky 1983 ) .
an adaf corresponds to the opposite regime . here
the gas is _
radiatively inefficient _ and the accretion flow is underluminous .
thus , an adaf is defined by the condition @xmath16 some papers in the literature define an adaf as a flow that corresponds exactly to a self - similar solution described in ny94 ( see 3.1 ) .
this is a needless restriction . in our opinion , it is more fruitful to employ the general definition of an adaf as given in equation ( [ adafdef ] ) , and the flow is only marginally radiatively inefficient ( see fig .
4 ) . nevertheless , even here , the two solutions are very distinct from each other . ] .
there are two distinct regimes of advection - dominated accretion .
the first , the one that we focus on in this article is when the accreting gas is very tenuous and has a long cooling time ( ny94 ; ny95b ; abramowicz et al .
this regime is sometimes referred to as a riaf a `` radiatively inefficient accretion flow '' and is defined by the condition @xmath17 here @xmath18 is the cooling time of the gas and @xmath19 is the accretion time .
the second regime of advection - dominated accretion is when the particles in the gas have no trouble cooling , but the scattering optical depth of the accretion flow is so large that the radiation is unable to diffuse out of the system .
this radiation - trapped regime was briefly discussed by begelman ( 1979 ) and was developed in detail by abramowicz et al .
( 1988 ) in their `` slim disk '' model .
the defining condition for this regime of adafs is @xmath20 where @xmath21 is the diffusion time for photons .
the present review is devoted exclusively to the adaf / riaf form of accretion ( see narayan , mahadevan & quataert 1998b ; kato , fukue & mineshige 1998 ; lasota 1999a , b ; quataert 2001 ; narayan 2002 , 2005 ; igumenshchev 2004 ; done , gierlinsky & kubota 2007 for reviews emphasizing various aspects of adafs ) .
we will henceforth drop the modifier riaf and refer to these flows simply as adafs . as explained above ,
an adaf is by definition very different from the standard thin accretion disk .
correspondingly , it is characterized by very distinct observational signatures .
both adafs and thin disks have well - known counterparts in nature .
in an adaf , since most of the energy released by viscous dissipation is retained in the gas , the pressure is large , and so is the sound speed : @xmath22 , where the keplerian velocity @xmath23 is equal to @xmath24 with @xmath25 being the radius in schwarzschild units , @xmath26 .
the large pressure has several immediate consequences .
first , the accretion flow becomes geometrically thick , with a vertical height @xmath27 of order the radius @xmath28 ( an adaf may be viewed as the viscous rotating analog of spherical bondi accretion ) .
second , the flow has considerable pressure - support in the radial direction , so the angular velocity becomes sub - keplerian .
third , the radial velocity of the gas is relatively large : @xmath29 , where @xmath30 ( see below ) is the standard dimensionless viscosity parameter ( shakura & sunyaev 1973 ) .
fourth , the large radial velocity leads to a short accretion time : @xmath31 , where @xmath32 is the free - fall time .
finally , the large velocity and large scale height cause the gas density to be very low , and so the cooling time is very long and the medium is optically thin .
the above properties are nicely illustrated in the self - similar adaf solution derived by ny94 , in which all quantities in the accretion flow behave as power - laws in radius .
( a similar solution was obtained by spruit et al .
1987 in a different context . )
ny94 obtained a general solution for arbitrary viscosity parameter @xmath33 , adiabatic index @xmath34 , and advection parameter corresponds to a fully cooling - dominated ( no advection ) flow and @xmath35 corresponds to a fully advection - dominated ( no radiative cooling ) flow . ]
@xmath36 . in the limit @xmath37 ( a good approximation ) and @xmath38 ( radiatively very inefficient flow )
, the solution simplifies to @xmath39v_k = -0.53 \alpha v_k , \\
\omega & = & \left[{2(5/3-\gamma)\over3(\gamma-5/9)}\right]^{1/2}\omega_k = 0.34 \omega_k , \\
c_s & = & \left[{2(\gamma-1)\over3(\gamma-5/9)}\right]^{1/2}v_k = 0.59 v_k.\end{aligned}\ ] ] the numerical coefficients on the right correspond to @xmath40 , a reasonable choice ( see quataert & narayan 1999 ) .
the great virtue of the above self - similar solution is that it is analytic and provides an easy and transparent way of understanding all the key properties of an adaf .
its biggest deficiency is that it is scale - free , which means that it is inappropriate near the inner or outer boundary of the flow .
therefore , for detailed work , one must use global solutions of the adaf equations that satisfy appropriate boundary conditions ( e.g. , abramowicz et al .
1996 ; narayan , kato & honma 1997c ; chen , abramowicz & lasota 1997 ; manmoto , mineshige & kusunose 1997 ; popham & gammie 1998 ; manmoto 2000 ) . calculating global solutions
is somewhat involved and one may wish to employ some short - cuts ( e.g. , yuan , ma & narayan 2008 ) . on the other hand ,
even the exact global solutions are somewhat limited since they solve a set of height - integrated equations . for greater realism
, one might wish to work directly with numerical simulations ( e.g. , goldston , quataert & igumenshchev 2005 ; noble et al .
2007 ) .
the adaf solution is gas pressure dominated . since @xmath41 , this means the gas temperature is nearly virial . under normal conditions , gas at such a high temperature
will radiate copiously , especially at small radii where the temperature can approach @xmath0 k. thus , in order to have an adaf , the accreting gas generally has to be a _ two - temperature plasma _
( at least at small radii ) , with electron temperature @xmath42 much less than the ion temperature @xmath43 ( ny95b ) .
( the only way to avoid this condition is by having an extremely low accretion rate below about @xmath44 of the eddington rate . )
typical adaf models have the two temperatures scaling as @xmath45 in order for gas in an adaf to be two - temperature , there must be weak coupling between electrons and ions . models generally assume that the coupling occurs via coulomb collisions , which is inefficient at the densities under consideration .
begelman & chiueh ( 1988 ) investigated whether plasma instabilities might enhance the coupling and drive the plasma rapidly to a single temperature ; this would be problematic for the adaf solution .
however , they were unable to identify a clear mechanism .
for normal mass accretion rates , the electrons in a two - temperature adaf will have a thermal energy distribution ( but not necessarily the ions , see mahadevan & quataert 1997 ) .
early work on two - temperature adafs assumed that viscous heating acts primarily on the ions ; for instance , ny95b took the ratio of electron heating to total heating , @xmath46 , to be zero , while esin , mcclintock & narayan ( 1997 ) assumed @xmath47 .
however , an adaf does not _ require _ @xmath46 to be this small .
because of a degeneracy in model parameters ( quataert & narayan 1999 ) , it is possible to have a viable adaf model with larger values of @xmath46 , provided the mass loss parameter @xmath48 , defined in 3.6 , is adjusted .
more recent adaf models ( e.g. , yuan , quataert & narayan 2003 ) typically assume @xmath49 .
various attempts have been made to estimate the value of @xmath46 from first principles by considering the effects of reconnection ( bisnovatyi - kogan & lovelace 1997 ; quataert & gruzinov 1999 ) or mhd turbulence ( quataert 1998 ; blackman 1999 ; medvedev 2000 ) .
these studies do not agree on a single value of @xmath46 , but generally suggest that @xmath46 is likely to be much larger than @xmath50 .
recently , sharma et al . ( 2007 ) considered heating by the dissipation of pressure anisotropy and showed that @xmath51 .
it should be noted that , even if electrons and ions receive equal amounts of the dissipated energy ( i.e. , @xmath52 ) , the plasma can still be two - temperature .
this is because a large part of the heating in an adaf is by compression ( since the density increases inward ) .
once @xmath53 , which is the case for @xmath54 a few hundred , the electrons become relativistic and have an adiabatic index @xmath55 , whereas the ions continue to be non - relativistic with @xmath56 . since adiabatic heating by compression causes the temperature to scale as @xmath57 , the electrons heat up only as @xmath58 whereas the ions heat up as @xmath59 .
thus , even in the limiting case of @xmath60 , adafs naturally become two - temperature at small radii . for instance , in the adaf model of sagittarius a@xmath3 ( sgr a@xmath3 ) proposed by yuan et al .
( 2003 ) , the authors obtain @xmath61 close to the bh even though they assumed @xmath62 .
an important property of the adaf solution is that it is thermally stable ( ny95b ; wu & li 1996 ; kato et al .
1997 ; wu 1997 ) .
the demonstration of this property was a crucial advance .
two decades earlier , in a seminal paper , shapiro , lightman & eardley ( 1976 ) , and after them rees et al .
( 1982 ; ion tori ) , introduced the idea of a two - temperature accretion flow and derived a hot two - temperature accretion flow solution , the sle solution .
however , that solution turned out to be thermally unstable ( piran 1978 ) .
until the development of the adaf solution , and the recognition that it is different from the sle solution , no stable , hot , optically thin solution was available to model the many accretion systems whose spectra ( especially in the hard state , see below ) demand such a flow .
chen et al .
( 1995 ) and yuan ( 2003 ) have explored the relationships among the adaf / riaf , sle , adaf / slim disk and thin disk solutions .
the adaf is a full accretion solution which incorporates consistent dynamics , thermal balance , radiation physics , etc .
therefore , the radial profiles of all gas properties can be calculated self - consistently once we know the values of certain parameters : bh mass @xmath63 , accretion rate @xmath64 , viscosity parameter @xmath33 , pressure parameter @xmath65 , adiabatic index @xmath34 ( usually @xmath66 ) , viscous heating parameter @xmath46 , advection parameter @xmath67 . actually , apart from the system - specific parameters @xmath63 and @xmath64 ( which may be estimated through observation ) , most of the other parameters are constrained . under the nearly collisionless conditions expected in an adaf , the viscosity parameter is moderately enhanced relative to a collisional gas ( sharma et al .
2006 ) . in the case of dwarf novae in the hot state , smak ( 1999 )
estimates @xmath68 , while numerical simulations of the magneto - rotational instability give @xmath69 ( hawley , gammie & balbus 1996 ) .
thus we expect @xmath30 for an adaf ( typical values used in models are @xmath70 ) .
numerical simulations further suggest that magnetic fields are generally subthermal , with @xmath71 ( e.g. , hawley et al . 1996 ) , so we expect @xmath72 . by calculating the energy loss via radiation from the hot accretion flow ( synchrotron , bremsstrahlung , compton scattering ) ,
@xmath67 can be obtained self - consistently ( e.g. , ny95b ; narayan , barret & mcclintock 1997a ; esin et al .
1997 ; yuan et al .
thus , we have only one poorly constrained parameter : @xmath46 . the recent work of sharma et al .
( 2007 ) provides a serviceable prescription even for this parameter . at large radii , where the plasma is effectively one - temperature , their formula gives @xmath73 , while in the energetically important inner region , where the plasma is two - temperature , they find @xmath74 , depending on model details .
the final two parameters are discussed later : the transition radius @xmath75 , 3.5 , and the wind parameter @xmath48 , 3.6 . .
( from esin et al .
1998),width=316 ] since the earliest days of x - ray astronomy , it has been clear that bh binaries ( bhbs ) have a number of distinct spectral states ( see zdziarski & gierlinski 2004 ; mcclintock & remillard 2006 ; done et al .
the most notable among these are the luminous _ high soft state _ , or _
thermal state _ ,
the slightly less luminous _ low hard state _ , and the very under - luminous
_ quiescent state_. the thermal state is well described by the thin disk model , and a multi - color disk ( mcd ) blackbody model ( e.g. , _ diskbb _ , mitsuda et al .
1984 ; _ ezdiskbb _ , zimmerman et al . 2005 ; both available in xspec , arnaud et al . 1996 ) has been successfully used for years to model the x - ray spectra of sources in this state .
recently , fully relativistic versions of the mcd model for arbitrary bh spin ( _ kerrbb _ , li et al . 2005 ; _ bhspec _ , davis & hubeny 2006 ) have been developed , based on the relativistic thin disk model of novikov & thorne ( 1973 ) .
these models provide excellent fits to the x - ray spectra of bhbs in the thermal state ( e.g. , mcclintock et al .
2006 ; davis , done & blaes 2006 ) . whereas a satisfactory theoretical model , viz .
, the thin disk model , was established early on for the thermal state , the hard state was for many years a mystery .
figure 1 shows the spectrum of a typical bhb , gro j0422 + 32 , in the hard state ( esin et al .
the observations indicate that the accreting gas is very hot , @xmath76 kev . the gas must also be optically thin , since optically thick blackbody emission at a temperature of 100 kev would correspond to a luminosity @xmath77 for any reasonable estimate of the radiating area @xmath78 .
the most natural explanation of the emission in j0422 and other bhbs in the hard state is that it is produced by thermal comptonization .
until the adaf model was established , no accretion model could reproduce such a spectrum .
( the sle solution could , but it was unstable . ) indeed , astronomers were reduced to using empirical comptonization models ( sunyaev & titarchuk 1980 ) in which they postulated a comptonizing cloud with some arbitrary geometry and parameterized the cloud with an adjustable temperature and an optical depth ( e.g. , zdziarski et al . 1996 , 1998 ; gierlinski et al .
1997 ) .
the situation changed with the recognition of the adaf solution .
this model turned out to have the precise properties density , electron temperature , stability needed to explain the hard state .
the solid line in fig .
1 shows an adaf model of j0422 ( esin et al .
1998 ) in which the accretion rate has been adjusted to fit the spectrum ; the required rate is about a tenth of the eddington mass accretion rate @xmath79 , where @xmath80 , i.e. , it is the mass accretion rate at which a disk with radiative efficiency 0.1 would radiate at the eddington luminosity . it is gratifying that both the temperature ( which determines the position of the peak ) and the compton @xmath81-parameter ( which determines the power - law slope below the peak ) are reproduced well .
esin et al .
( 1997 , 1998 , 2001 ) present models of other bhbs in the hard state .
-0.2 in limits from asca ( narayan et al .
1997a ) , and the upper limit in the euv is derived from the absence of a heii @xmath82 line .
the solid line corresponds to an adaf model with @xmath83 .
the dashed line is a thin disk model whose @xmath64 has been adjusted to fit the optical data .
this model fits poorly in the x - ray band and is inconsistent with the euv limit .
( from narayan et al .
1997a),title="fig:",width=355 ] -1.5 in .
the radio data are from falcke et al .
( 1998 ; open circles ) and zhao et al .
( 2003 ; filled circles ) , the ir data are from serabyn et al .
( 1997 ) and hornstein et al .
( 2002 ) , and the two `` bow - ties '' in the x - ray band correspond to the quiescent ( lower ) and flaring ( higher ) data from baganoff et al .
( 2001 , 2003 ) .
the solid line is an adaf model of sgr a@xmath3 in the quiescent state .
the mass accretion rate is @xmath84 near the bh .
( from yuan et al .
2003),width=288 ] a typical accreting bh observed in the hard state has a luminosity on the order of a few percent of eddington . at much lower luminosities
, we have the quiescent state , where the spectrum becomes noticeably different .
figures 2 and 3 show observations of a quiescent bhb , v404 cyg , and a quiescent supermassive bh , the galactic center source sgr a@xmath3 .
although these spectra look very different from the one shown in fig .
1 , the adaf model is able to fit these and other observations of quiescent systems ( narayan , mcclintock & yi 1996 ; narayan et al .
1997a ; yuan et al .
all it requires is a lower value of @xmath64 , as appropriate for the lower luminosity .
the qualitative features of the spectrum , e.g. , a softening of the x - ray power - law index ( see corbel , tomsick & kaaret 2006 ) , follow naturally
. however , a caveat is in order : in some cases , the x - ray emission in quiescence may be from a jet lauched from the adaf ( 3.6 ) , rather than from the adaf itself ( e.g. , yuan & cui 2005 ) . of an accretion
flow around a bh , as a function of the eddington - scaled mass accretion rate ( hopkins , narayan & hernquist 2006b ) .
the horizontal segment between 0.01 and 0.1 of the eddington accretion rate shows a transition regime in which a part of the accretion flow is an adaf , but the radiative efficiency is still large .
it might correspond to the intermediate state , and perhaps the upper end of the hard state ( 3.5 , fig .
although this plot is based on calculations shown in fig .
11 of ny95b and fig
. 13 of esin et al .
( 1997 ) , it is still very approximate.,width=316 ] as expressed in equation ( [ adafdef ] ) , the defining property of an adaf is that it is radiatively inefficient , @xmath85 calculations show that the adaf solution is possible only for low mass accretion rates .
specifically , only when @xmath86 is the gas density low enough to permit a two - temperature plasma ( ny95b ; narayan 1996 ; esin et al .
1997 ) . near the critical luminosity @xmath87 or critical mass accretion rate @xmath88 at which the adaf solution first becomes viable
, the radiative efficiency is , by continuity , not very different from that of a thin disk : @xmath89 .
however , with decreasing @xmath64 , the efficiency decreases .
very roughly , we estimate ( fig .
4 ) @xmath90 ( we note that for the different prescription for @xmath46 used by sharma et al . 2006 , @xmath91 falls rapidly only for @xmath92 . )
the extreme inefficiency of an adaf at very low accretion rates is critical for understanding the peculiar properties of accreting bhs in the quiescent state ( as first discussed by narayan et al .
1996 for bhbs and narayan et al .
1995 for supermassive bhs ) .
we do not discuss this topic here , but point the reader to other reviews for details ( e.g. , quataert 2001 ; narayan 2002 , 2005 ) .
the quiescent state has also played a major role in our efforts to test for the presence of an event horizon in bhbs .
this is a key topic of 4 .
( based on esin et al .
the adaf is represented by the hatched ellipses , with the intensity of hatching indicating the density of the hot gas .
the horizontal lines represent a standard thin disk .
the lowest panel shows the quiescent state ( qs ) , which corresponds to a very low mass accretion rate ( say @xmath93 ) , a weak jet ( 3.6 ) , a low radiative efficiency ( fig .
4 ) , and a large transition radius ( fig . 6 ) .
the second panel from the bottom shows the hard state ( hs ) , where the mass accretion rate is higher ( @xmath94 ) but still below the critical rate @xmath88 , the jet is stronger , the radiative efficiency is somewhat larger , and the transition radius is smaller .
the second panel from the top shows the intermediate state ( is ) , where @xmath95 , the jet is even stronger , the radiative efficiency is high @xmath96 , and the transition radius is fairly close to the isco .
finally , the top panel shows the thermal state ( ts ) , where there is no adaf , the thin disk extends down to the isco , @xmath89 , and there is no jet .
esin et al .
( 1997 ) included a tentative proposal for the so - called very high state ( see also done et al .
2007 ) , now called the steep power - law state ( spl , mcclintock & remillard 2006 ) , but we do not include this still - mysterious state here.,width=355 ] based on the properties of the adaf solution discussed above , narayan ( 1996 ) proposed a simple model for understanding the spectral states of accreting bhs .
this picture was developed in detail by esin et al .
the basic idea is illustrated in fig .
the key parameter that determines the spectral state of an accreting bh is @xmath64 .
when @xmath97 , only the thin disk solution is available , and so the accretion flow is in the form of a thin disk all the way down to the innermost stable circular orbit ( isco ) .
the system is then in the thermal state , and its spectrum is well - described by the mcd model .
once @xmath64 falls below @xmath88 , both the thin disk and adaf solutions become viable ( at least at small radii ) .
now accretion continues as a thin disk at radii larger than a transition radius , @xmath98 , but the flow switches to an adaf at smaller radii .
when the transition from a pure thin disk ( thermal state ) to a disk - plus - adaf configuration first occurs , i.e. , when @xmath64 is just below @xmath88 , the adaf is very small in size and we have a more - or - less radiatively efficient flow , with @xmath89 .
this corresponds to the so - called _
intermediate state_. then , at a somewhat lower @xmath64 , the adaf expands a bit and @xmath91 is modestly lower and we have the classic hard state .
finally , when @xmath64 is much lower than @xmath88 , the adaf becomes much larger , @xmath99 , with @xmath100 ( narayan et al .
1996 , 1997a ; menou , narayan & lasota 1999b ) .
this is the quiescent state .
the above paradigm has proved durable ( see zdziarski & gierlinski 2004 ; done et al .
in particular , considerable evidence has accumulated that the thin disk retreats from the innermost stable circular orbit ( isco ) to a large radius in the quiescent state .
the evidence is strongest in transient bhbs and cvs .
the spectra of quiescent bhbs show absolutely no sign of any soft blackbody - like x - ray emission from a thin disk at small radii ( narayan et al .
1996 , 1997a ; mcclintock , narayan & rybicki 2004 ) .
timing properties also indicate a large a large size for the adaf ( hynes et al .
2003 ; shahbaz et al . 2005 ) . in additional , theoretical arguments
indicate that the thermal - viscous disk instability , which causes the transient behavior in these systems , is incompatible with observations unless the disk is severely truncated in the quiescent state ( lasota , narayan & yi 1996b ; hameury , lasota & dubus 1999 ; lasota 2001 , 2008 ; dubus , hameury & lasota 2001 ; yungelson et al .
there are also indications from the time delay between the optical and x - ray outbursts in the bhb gro j165540 ( orosz et al .
1997 ; hameury et al .
1997 ) that the cool disk is truncated at a large radius in the quiescent state . in the case of supermassive bhs in the quiescent state
, there is no feature in the spectrum that might be associated with a thin disk , suggesting that there is no disk at all ; examples are sgr a@xmath3 ( narayan et al .
1998a ; narayan 2002 ) and m87 ( di matteo et al . 2000 , 2003 ) .
intermediate luminosity agn do exhibit optical emission from a disk , but the `` big blue bump '' is much less pronounced than in high - luminosity agn ( ho 1999 ) ; this suggests that the disk is truncated at a radius @xmath101 and the interior is filled with an adaf ( gammie , narayan & blandford 1999 ; quataert et al .
incidentally , the thermal - viscous disk instability does not appear to operate in agn disks ( menou & quataert 2001 ; hameury , lasota & viallet 2007 ) . in the more luminous hard state ,
again , there is considerable spectral evidence that the disk is truncated at a transition radius outside the isco , and that the inside is filled with an adaf - like hot flow .
the most spectacular example is the bhb xte j1118 + 480 , for which observations carried out in the hard state had unprecedented spectral coverage .
the observations are fit well with an adaf model , with a transition radius at @xmath102 ( esin et al .
the model even fits the complicated timing behavior of the source ( yuan , cui & narayan 2005 ) .
it is hard to imagine that the same data could be explained with any model in which a cool disk extends down to the isco .
done et al .
( 2007 ) review spectral observations of a number of other bhbs where again the data require a truncated disk .
nemmen et al .
( 2006 ) show that both the spectrum and the double - peaked balmer line profile of the liner source ngc 1097 are consistent with a disk truncated at a few hundred @xmath103 . in an interesting study of cyg x1 , gilfanov , churazov & revnivtsev ( 1999 ;
see cui et al .
1999 ; done et al . 2007
; for discussions of other sources ) found that , as the characteristic frequency in the variability spectrum of the source increases , the power - law tail in the spectral energy distribution steepens and the amplitude of the reflection component in the spectrum increases .
this is exactly what one expects when the transition radius between the outer cool disk and the inner hot adaf varies . with decreasing transition radius , ( i ) the noise frequency (
which is likely related in some fashion to the keplerian frequency at the transition radius ) should increase , ( ii ) the hot medium should be cooled more effectively by soft photons from the disk , giving a steeper power - law tail , and ( iii ) the solid angle subtended by the cool disk at the adaf should increase , and there should be a larger reflection component .
zdziarski , lubinski & smith ( 1999 ) have noted that a correlation between spectral slope and reflection is commonly seen in both bhbs and agn . a few bhbs in the hard state have been found to show a soft blackbody - like component in their spectra ( balucinska - church et al .
1995 ; di salvo et al .
2001 ; miller et al .
2006a , b ; ramadevi & seetha 2007 ; rykoff et al . 2007 ) .
this could be interpreted as evidence that the thin disk extends all the way down to the isco , not at the isco ] .
some authors have noted that it is possible for a thin disk to evaporate to an adaf at a relatively large radius and for the hot gas to then re - condense into a thin disk at small radii ( rozanska & czerny 2000 ; liu et al .
2007 ; mayer & pringle 2007 ) . such models might explain the occurrence of a soft spectral component in some hard state sources .
however , we note that the soft component typically has only 10% of the total observed luminosity .
it is hard to understand how a radiatively efficient thin disk located at the isco could be such a minor contributor to the emitted radiation .
this difficulty is highlighted in the work of dangelo et al .
they point out that , in any model of the hard state , there will be considerable interaction between the hot gas which produces the hard x - rays and cool gas that may be present in a conventional disk .
the interaction will occur via x - ray irradiation ( unless one has large outward beaming , which seems unlikely , e.g. , narayan & mcclintock 2005 ) as well as particle bombardment . using a prototype model for ion bombardment
( deufel et al . 2002 ; spruit & deufel 2002 ; dullemond & spruit 2005 ) , dangelo et al . (
2008 ) show that a weak soft component in the x - ray spectrum arises quite naturally when the thin disk is truncated at @xmath104 .
the model fits the observations surprisingly well ; in fact , the same model would predict a much stronger soft component , and would strongly disagree with the observations , if the cool disk were to extend down to the isco .
, plotted as a function of the eddington - scaled accretion luminosity @xmath105 , for a sample of bhbs and low - luminosity agn ( yuan & narayan 2004 ) .
the individual estimates of @xmath75 are obtained by fitting spectral observations and are very uncertain .
nevertheless , there seems to be a trend of increasing transition radius with decreasing luminosity.,width=297 ] one of the most vexing problems in the theory of adafs is that we do not have a robust method of estimating the location of the transition radius @xmath75 . from the earliest studies ( e.g. , ny95b ; meyer & meyer - hofmeister 1994 )
, it has been plausibly argued that @xmath75 increases with decreasing @xmath64 .
however , reliable predictions of the exact dependence have proved difficult , although a number of studies have come up with semi - quantitative results ( meyer & meyer - hofmeister 1994 ; honma 1996 ; rozanska & czerny 2000 ; meyer , liu & meyer - hofmeister 2000 ; spruit & deufel 2002 ; mayer & pringle 2007 ) .
yuan & narayan ( 2004 ) tried to use observations to deduce the run of @xmath75 with accretion luminosity .
figure 6 shows their results .
another issue that has been discussed recently is hysteresis in the transition between the thermal state and the hard state .
specifically , with increasing @xmath64 , the hard state survives up to fairly high luminosities @xmath106 , whereas with decreasing @xmath64 , the transition from the thermal state to the hard state occurs at a much lower luminosity @xmath107 ( miyamoto et al . 1995 ; maccarone & coppi 2003 ; zdziarski et al .
2004 ; done et al .
meyer - hofmeister , liu & meyer ( 2005 ) have provided a plausible explanation .
their disk evaporation model , coupled with compton - cooling of the hot electrons in the adaf , naturally produces a hysteresis in the location of the transition radius .
the brightest systems in the hard state ( @xmath108 ) are difficult to model with the standard adaf model . a variant of the adaf solution a natural extension of the model called the luminous hot flow ( lhaf ; yuan 2001 , 2003 ; yuan & zdziarski 2004 ; yuan et al . 2007 ; see also machida , nakamura & matsumoto 2006 ) , looks promising for modeling these sources .
ny94 , ny95a discovered an unexpected property of the adaf solution : _ the accreting gas in an adaf has a positive bernoulli parameter _
, which is defined as the sum of the kinetic energy , potential energy and enthalpy , @xmath109 a positive bernoulli constant means that the gas is not bound to the bh .
( this is not surprising , since the gas is not losing energy through radiation . ) therefore , the above authors suggested that adafs should be associated with strong winds and jets ( see also meier 2001 ) .
strong outflows have been seen in numerical simulations of adafs .
the first indications came from 2d and 3d hydrodynamic simulations ( stone , pringle & begelman 1999 ; igumenshchev & abramowicz 2000 ; igumenshchev , abramowicz & narayan 2000 ) , but it was soon confirmed in mhd simulations as well ( stone & pringle 2001 ; hawley & balbus 2002 ; igumenshchev , narayan & abramowicz 2003 ; pen , matzner & wong 2003 ; machida , nakamura & matsumoto 2004 ; mckinney & gammie 2004 ; igumenshchev 2004 ) .
apart from producing a gas - dominated , large - scale outflow , mhd simulations of adafs also have a second distinct outflow component along the axis in the form of a collimated , poynting - dominated , relativistic jet ( mckinney 2005 , 2006 ) .
the original suggestion of ny94 that adafs would have outflows and jets has thus been confirmed by these simulations .
nevertheless , some authors have disputed a connection between a positive bernoulli parameter and an outflow since it is possible to come up with explicit models that have a positive bernoulli parameter but no outflow ( paczyski 1998 ; abramowicz et al .
2000 ) .
observational evidence for the association of nonthermal relativistic jets with adafs has accumulated in recent years with the discovery of radio emission in virtually every bhb in the hard state ( corbel et al .
2000 ; fender 2001 ; fender , belloni & gallo 2004 ; fender & belloni 2004 ) .
the radio emission is generally too bright to be produced by thermal electrons in the accretion flow .
it is therefore very likely to come from nonthermal electrons in a jet ; in fact , radio vlbi imaging has revealed a resolved jet in a few sources .
thus , it is now observationally well - established that the hard state / adaf is associated with relativistic jets . according to the discussion in 3.5
, the quiescent state also has an adaf and should have a ( weaker ) jet .
indeed , radio emission has been seen from two quiescent systems , v404 cyg ( hjellming et al .
2000 ; gallo , fender & hynes 2005 ) and a062000 ( gallo et al . 2006 ) , confirming this expectation .
how much of the x - ray emission in an adaf system comes from the accretion flow and how much from the jet ?
a strong correlation has been seen between radio and x - ray luminosity in the hard state and quiescent state ( corbel et al .
2003 ; gallo , fender & pooley 2003 ) . at first sight
this might suggest that the x - ray must also come from the jet ( e.g. , falcke , krding & markoff 2004 ) .
however , since the jet flows out of the adaf and is thus highly coupled to it , any model in which the x - ray emission is from the adaf and radio is from the jet is equally compatible with the observations . in recent times ,
several authors have come down in favor of an adaf origin for the x - rays ( e.g. , heinz & sunyaev 2003 ; merloni , heinz & di matteo 2003 ; heinz 2004 ; heinz et al .
2005 ; yuan et al .
2005 ) .
additional evidence on this issue comes from hard state spectra indicating thermal emission ( e.g. , see fig .
1 ) , which is very different from the nonthermal power - law emission one expects from a jet .
markoff , falcke & fender ( 2001 ) , among others , have argued that the x - ray emission is due to synchrotron emission from a carefully tuned distrubution of nonthermal electrons .
however , zdziarski et al . ( 2003 ) and zdziarski & gierlinski ( 2004 ) showed that the high - energy cutoff in the spectra of bhbs in the hard state , which arises naturally in a thermal adaf - like model , is very difficult to reproduce in a nonthermal synchrotron model .
a variant of the jet model assigns the x - ray emission to thermal electrons in the `` base of the jet '' ( markoff , nowak & wilm 2005 ) , but the discussion then becomes semantic .
the base of the jet is surely embedded in the underlying adaf and it is not clear that one gains anything by simply relabeling the radiation from the adaf as jet emission .
the issue is discussed in greater depth in narayan ( 2005 ) , who advocates a ` jet - adaf ' model ( as described in yuan et al . 2005 ; malzac , merloni & fabian 2004 ) , in which the high energy x - ray emission in the hard state comes from the adaf and the low - energy radio ( and infrared ) emission comes from a jet .
the situation is less clear - cut in the quiescent state , where nonthermal emission from the jet might dominate even in x - rays ( yuan & cui 2005 ; wu , yuan & cao 2007 ) .
it is interesting , however , that the most quiescent system we know , sgr a@xmath3 , shows no sign of a jet .
radio images with a resolution of @xmath110 appear to be jet - free ( shen et al . 2005 ) , while the quiescent x - ray emission is spatially resolved and appears to be from thermal gas near the bondi radius at @xmath111 , not from a jet ( baganoff et al .
2001 , 2003 ) .
turning our attention now to the extended wind that is predicted from an adaf , and seen in numerical simulations , one consequence of the wind is that , at each radius , a fraction of the accreting gas is lost from the system .
thus , the accretion rate itself varies with radius .
this is usually parameterized with an index @xmath48 such that @xmath112 .
operationally , this means that , in the self - similar regime , the gas density varies with radius as @xmath113 rather than as @xmath114 in the ny94 model .
unfortunately , apart from the rather general constraint that @xmath48 should lie between 0 ( no mass loss ) and 1 ( the limit of a cdaf , 3.7 ) , there is no good theoretical estimate of the value of @xmath48 .
we have to resort to numerical simulations or observations .
the former tends to give somewhat larger values , e.g. , @xmath115 ( pen et al .
2003 ) , while the one case where the latter approach has been tried , sgr a@xmath3 , yields a smaller value , @xmath116 ( yuan et al .
2003 ) .
another general property of adafs was highlighted by ny94 ( see also begelman & meier 1982 ) : adafs have entropy increasing inward and should be convectively unstable by the schwarzschild criterion .
numerical hydrodynamic simulations of adafs confirmed the presence of strong convection ( igumenshchev et al .
2000 ; narayan , igumenshchev & abramowicz 2000 ) and led to the development of an analytical self - similar model called the convection - dominated accretion flow ( cdaf ; narayan et al .
2000 ; quataert & gruzinov 2000 ) .
the cdaf model employs a simplified one - dimensional treatment of the accretion flow , in which all fluxes are assumed to be in the radial direction .
the resulting density varies as @xmath117 . in the language of 3.6
, a cdaf corresponds to @xmath118 .
however , technically , an idealized cdaf has no mass outflow , only an outward flow of energy by convection ; the energy is assumed to flow out into a surrounding medium . in practice , the flow is never perfectly one - dimensional ; 2d effects intrude and one expects substantial winds and mass loss as well .
a real adaf thus involves a complicated interplay between convection and winds .
magnetic fields introduce further complexity , and there has been some discussion in the literature on whether mhd adafs do or do not have real convection ( machida , matsumoto & mineshige 2001 ; balbus & hawley 2002 ; narayan et al .
2002b ; pen et al .
2003 ; igumenshchev 2004 ) .
the question has no practical significance , however , since there is general agreement that adafs have both unstable entropy gradients and strong magnetic stresses . in the particular case of a spherical , non - rotating , mhd flow ( the magnetic analog of the bondi problem ) , some complications are absent and the problem becomes relatively clean .
here one finds that convection and magnetic fields strongly influence the accretion physics , and the resulting flow is very different from the standard bondi solution ( igumenshchev & narayan 2002 ; igumenshchev 2004 , 2006 ) .
this result is likely to have significant implications for astrophysics ; for instance , isolated neutron stars accreting from the interstellar medium will be very much dimmer than one might expect based on the bondi solution ( perna et al .
2003 ) .
the adaf solution is essentially independent of the mass @xmath63 of the central bh .
that is , if length and time are scaled by @xmath63 and the accretion rate is scaled by the eddington rate ( also proportional to @xmath63 ) , then the same solution is valid for any bh mass .
therefore , any successful application of the adaf model to a bhb system has immediate consequences for supermassive bhs ( smbhs ) in an equivalent state , and vice versa .
this close connection between adafs in bhbs and adafs in agn was highlighted in narayan ( 1996 ) ; it is also empirically obvious from fig .
6 , which combines observations and models of bhbs and agn .
the first smbh to be modeled as an adaf was the galactic center source sgr a@xmath3 ( narayan et al .
this was soon followed by fabian & rees ( 1995 ) , who suggested that quiescent nuclei in nearby giant ellipticals ( e.g. , m87 ) must be accreting via adafs , and by lasota et al .
( 1996a ) , who argued that low ionization nuclear emission - line region sources ( liners , e.g. , ngc 4258 ) and low - luminosity active galactic nuclei ( llagn ) must have adafs .
both suggestions have turned out to be correct ( e.g. , reynolds et al .
1996 ; mahadevan 1997 ; gammie et al . 1999 ; quataert et al . 1999 ; di matteo et al .
2000 , 2003 ; loewenstein et al . 2001 ; ulvestad & ho 2001 ; nemmen et al . 2006 ) .
other authors have suggested that all of the following systems have adafs : fri sources ( baum , zirbel & odea 1995 ; reynolds et al .
1996 ; begelman & celotti 2004 ) , bl lac sources ( maraschi & tavecchio 2003 ) , x - ray bright optically normal galaxies ( xbongs ; yuan & narayan 2004 ) , and even some seyferts ( chiang & blaes 2003 ) .
all of these sources are relatively low - luminosity agn , where an adaf is likely to be present ( figs . 5 , 6 ) .
as discussed in 3.6 , a feature of the adaf model is the presence of outflows and jets .
this implies that adaf systems should generally be radio - loud .
this is certainly the case for the fri and bl lac sources mentioned above .
ho ( 2002 ; see also nagar et al .
2000 ; falcke et al .
2000 ) presents in fig .
5b of his paper a very interesting plot of radio loudness , defined as the ratio of the flux at 6 cm to the flux in the optical b band versus eddington - scaled luminosity , for a sample of galactic nuclei .
virtually every source in the plot with an eddington - ratio below about 0.01 is radio loud , which is perfectly consistent with our expectation that all such sources should have adafs .
moreover , the degree of radio loudness increases with decreasing eddington ratio , again consistent with our expectation that the adaf should become more and more dominant with decreasing @xmath64 . the majority of sources in ho s plot that have @xmath119 are radio - quiet , as we would expect from fig .
5 if these systems have cool disks .
however , a small minority of these bright sources do show powerful jets ( they are generally frii sources ) .
the exact nature of the accretion in these sources is unclear . a more complete plot , with updated data ,
can be found in sikora , stawarz & lasota ( 2007 ) .
the jet and the extended wind from an adaf carry with them substantial kinetic energy .
this energy will be dumped into the external medium and will have important consequences . in the context of galaxy formation , there has been discussion recently of the so - called `` radio mode '' of accretion ( croton et al .
2006 ) in which outflowing energy from an accreting smbh in the galactic nucleus heats up the surrounding medium .
this kind of agn feedback can lead to various effects such as shutting off accretion and/or star formation ( di matteo et al .
2005 ; hopkins et al .
2006a ) .
it is worth noting that this radio mode is nothing other than the adaf mode of accretion reviewed in this article .
investigators of agn feedback may find it profitable to study the considerable work that has been done on adafs over the last fifteen years .
the first bh , cygnus x-1 , was identified and established in 1972 via a measurement of its mass , which was shown to be too large for a neutron star ( ns ) .
the surest evidence for the existence of bhs continues to be through dynamical mass measurements .
we now know of 20 additional compact binary x - ray sources ( mcclintock & remillard 2006 ; orosz et al .
2007 ) with primaries that are too massive to be a ns or any stable assembly of cold degenerate matter , assuming that gr is valid .
similarly , dynamical data have established the existence of supermassive bhs , most notably in the nucleus of our milky way galacy ( schdel et al . 2002 ; ghez et al .
2005a ) and in ngc 4258 ( miyoshi et al .
1995 ) . are these compact objects genuine bhs
pockets of fully collapsed matter that are walled off from sight by self gravity and that , like a shadow , reveal no detail or are they exotic objects that have no event horizons but manage to masquerade as bhs ?
most astrophysicists believe that they are genuine bhs .
there are several reasons for this confidence , the most important being that bhs are an almost inevitable prediction of gr . however , this argument is circular because it presumes that gr is the correct theory of gravity .
furthermore , it ignores the many exotic alternatives to bhs that have been suggested .
thus , the collapsed objects that we refer to throughout as `` black holes '' are strictly speaking dynamical bh candidates .
the current evidence for bhs is not decisive , nor can dynamical measurements be expected to make it so .
we now consider some approaches aimed at establishing that these dynamical bhs are genuine .
the defining property of a bh is its event horizon .
demonstrating the existence of this immaterial surface would be the certain way to prove the reality of bhs .
unfortunately , unlike any ordinary astronomical body such as a planet or a star of any kind , it is quite impossible to detect radiation from the event horizon s surface of infinite redshift .
( hawking radiation is negligibly weak for massive astrophysical bhs . ) nevertheless , despite the complete absence of any emitted radiation , it is possible to marshal strong _ circumstantial evidence _ for the reality of the event horizon .
the fruitful approaches described below are based on comparing x - ray binary systems that contain bh primaries with very similar systems that contain ns primaries .
such investigations are motivated by the simple fact that the termination of an accretion flow at the hard surface of a ns has observational consequences .
the earliest and strongest evidence for the event horizon is based on the faintness in quiescence of bh transient systems relative to comparable ns systems . in
4.1 , we discuss the physical arguments that underpin this evidence , and in 4.2 we present the comparative luminosity data for bhs and nss .
alternative explanations for the lower luminosities of the bh systems , which do not involve the event horizon , are considered in 4.3 . in
4.4 , we present three independent and additional arguments for the existence of the event horizon , which are again rooted in comparing bh and ns x - ray binaries . in 4.5 , we discuss the extreme faintness and properties of sgr a * and the evidence that this supermassive bh has an event horizon , and in 4.6 we argue that even very exotic objects with enormously strong surface gravity ( e.g. , gravastars ) can not escape our arguments for the event horizon .
a test particle in a circular orbit at radius @xmath28 around a mass @xmath63 has , in the newtonian limit , kinetic energy per unit mass equal to @xmath120 , and potential energy equal to @xmath121 . in the context of a gaseous accretion disk ,
this means that a gas blob at radius @xmath28 retains @xmath122 of its potential energy as kinetic energy .
the remaining 50% was transformed into thermal energy during the viscous accretion of the blob to its current radius ( frank et al .
2002 ) .
we now turn to consider accretion on to the material surface of a ns .
if the mass @xmath63 at the center of an accretion disk has a surface , then the kinetic energy in the accreting gas will be converted to thermal energy in a viscous boundary layer and radiated ( frank et al .
in addition , the residual thermal energy that the gas possesses when it reaches the inner edge of the disk will also be radiated from the surface . if accretion occurs via an adaf , the luminosity from the accretion disk @xmath123 and that from the stellar surface @xmath124 will satisfy @xmath125 where @xmath126 is the radius of the stellar surface .
the accretion luminosity is small because the flow is radiatively inefficient .
therefore , essentially all the potential energy of the accreting gas remains in the gas in the form of thermal and kinetic energy .
if the central object has a surface , e.g. , it is a ns , the total luminosity we observe will be equal to @xmath124 .
however , if the object has an event horizon , i.e. , it is a bh , there will be no radiation from a stellar surface and the luminosity we observe will only be equal to @xmath123 .
thus , we expect @xmath127 , regardless of whether the accretion flow is radiatively efficient or not . in the case of a bh
, however , the radiative efficiency @xmath91 of the accretion flow is of paramount importance , and the observed luminosity becomes @xmath128 when the accretion rate is highly sub - eddington.,width=316 ] figure 7 shows schematically the luminosity difference we predict between a ns and a bh .
this plot is based on the accretion efficiency estimate shown in fig .
4 . especially at low mass accretion rates , say @xmath129 ( deep quiescent state ) , we expect a huge luminosity difference between nss and bhs .
efforts to test this prediction are described in 4.2 .
if the accretion disk is radiatively efficient , then the gas has little thermal energy when it reaches the inner edge of the disk . in this case
, we expect .
moreover , if the central mass spins too close to the `` break - up '' limit , the extra luminosity due to the surface can actually exceed @xmath130 by a large factor ( popham & narayan 1991 ) . ]
@xmath131 the above result is based on a newtonian analysis and is okay so long as the central object has a radius @xmath132 , the radius of the isco . for more compact objects , we must allow for the fact that , inside the isco , the accreting gas no longer spirals in by viscosity but free - falls in the gravitational potential of the central mass .
now we have @xmath133 the accretion luminosity is limited by the binding energy of the gas at the isco , which gives a radiative efficiency @xmath89 . on the other hand ,
the total energy budget is @xmath134 , which means that a larger fraction of the luminosity is released at the surface .
( the free - falling gas inside the isco , crashes on the surface and releases its energy in a shock . ) in the limit when @xmath126 is arbitrarily close to @xmath103 ( cf . , the discussion of gravastars in 4.6 ) , the total luminosity is equal to @xmath135 , i.e. , the entire rest mass energy of the accreting gas is converted to radiation ( 100% radiative efficiency ) .
note that all luminosities discussed here refer to measurements by an observer at infinity . combining the results in equations ( [ adafsurf ] ) , ( [ surf2 ] ) and ( [ surf3 ] ) , we see that
, regardless of the radius of the accretor and whether the accretion flow is radiatively efficient or inefficient , we expect @xmath136 although this relation is weaker than ( [ adafsurf ] ) , it can still be used in favorable cases to test for the presence of an event horizon ( 4.5 ) .
bh and ns transient binary systems are very similar in many respects , and it is reasonable to expect that their mass accretion rates and luminosities would be comparable under similar conditions . there is , however , one important qualitative difference between the two kinds of object nss have surfaces whereas bhs do not .
often , this difference is not important .
however , as discussed in 4.1 ( see eq .
[ adaflum ] ) , when accretion occurs via an adaf , a ns binary should be much more luminous than a bh binary ( ny95b )
. the difference will be especially large in the quiescent state , when the accretion flow is radiatively extremely inefficient ( fig . 7 ) .
narayan , garcia & mcclintock ( 1997b ) and garcia , mcclintock & narayan ( 1998 ) collected available x - ray data on quiescent ns and bh transients and showed that bh systems are consistently fainter than ns systems .
this was the first indication that the objects that astronomers call `` black holes '' are indeed genuine bhs with event horizons .
fainter than ns systems with similar orbital periods .
, width=316 ] lasota & hameury ( 1998 ) made the important point that it is necessary to compare quiescent ns and bh transients at similar mass accretion rates , and the surest way to achieve this is to plot luminosities as a function of the binary orbital period @xmath137 ( the relevant arguments are outlined in menou et al .
1999b ; lasota 2000 , 2008 ) .
since 1999 , this has been the standard way of plotting the data ( menou et al . 1999a ; garcia et al .
2001 ; narayan , garcia & mcclintock 2002a ; hameury et al .
2003 ; mcclintock et al .
figure 8 shows the current status of the comparison , with the luminosities of quiescent bh and ns transients plotted in ( most - appropriate ) eddington units .
9 shows the same data without the luminosities being scaled .
it is clear that , for comparable orbital periods , the bh systems are 2 to 3 orders of magnitude fainter than their ns cousins . as discussed in 4.2 , a radiatively inefficient adaf provides a natural explanation for the large luminosity deficit of the bh systems . in this model
, the gas approaches the center with a large amount of thermal energy .
a bh is dim because the bulk of this thermal energy is trapped in the advective flow , passes through the event horizon , and is lost from sight . on the other hand ,
a ns is bright because the thermal energy is radiated from its surface .
a number of alternative models have been put forward in an attempt to rationalize the luminosity differences between quiescent ns and bh systems .
notice below that some of these models generate x - ray emission without invoking accretion at all .
we first consider a recent challenge centered on the discovery of the extremely low - luminosity ns transient 1h 1905 + 00 .
next , we discuss the possibility that the bulk of the accretion power is channeled into a steady jet rather than into x - ray emission .
we then consider several diverse models that are discussed in further detail in narayan et al .
( 2002a ) . _
the case of 1h 1905 + 00 : _ jonker et al .
( 2007 ) claim that the extremely low x - ray luminosity of the ns transient and type i burst source 1h 1905 + 00 ( hereafter , h1905 ) undermines the evidence summarized in 4.2 for the existence of the event horizon .
the luminosity of this ns ( @xmath138 ; @xmath139 kpc ) , is the same as the luminosity of a062000 and several other short - period ( @xmath140 hr ) bh systems . based on this result , jonker et al .
assert that the evidence for event horizons is `` unproven . ''
however , they ignore a key point of our argument : namely , as discussed above and illustrated in figures 8 and 9 , the case for event horizons depends critically on comparing bh and ns systems with _ similar orbital periods_. we first note that the orbital period of h1905 is unknown .
more importantly , this unknown period is believed to be very short so short that there are no bh systems with comparable periods .
this conclusion is based on deep optical imaging data and the lack of a counterpart .
jonker et al .
( 2007 ) conclude that the secondary `` can only be a brown or a white dwarf , '' and that the system is probably an ultracompact binary .
hence the orbital period is likely to be tens of minutes , far less than the shortest bh binary period of 4.1 hr .
thus , there is no bh system comparable to h1905 . and there is no way to usefully predict the mass accretion rate , as jonker et al .
conclude in the paper s final sentence .
the rough trend of luminosity with period indicated in figure 9 for nss might imply a very low luminosity for h1905 , as observed , but such an extrapolation is unwarranted .
see lasota ( 2007 , 2008 ) for a more detailed discussion of this and other issues .
jonker et al . (
2007 ) additionally argue that an unknown amount of mass transfered from the secondary might be lost from the system in winds or jets .
their one example of a wind , viz .
, an extraordinary mass - loss episode observed in gro j1655 - 40 ( miller et al .
2006c ) when the source was in outburst and had a luminosity @xmath141 times the quiescent level , is quite inappropriate .
also , their mention of possible mass expulsion for quiescent nss via the propeller mechanism ( lasota & hameury 1998 ; menou et al .
1999a ) , if effective , would reduce the luminosities of nss and would only strengthen our argument .
_ jet outflows : _ fender et al .
( 2003 ) argue that transient bhs at low accretion rates ( @xmath142 ) `` should enter ` jet - dominated ' states , '' in which the majority of the accretion power drives a radiatively - inefficient jet . as we have seen in 3.6 , there is good evidence for this .
for instance , the presence of a radio jet has been reasonably well established in quiescence for a062000 ( gallo et al .
2006 ) , the closest and one of the least luminous of the bhs in question .
fender et al .
( 2003 ) appeal to the empirical result that bhs are @xmath143 times as ` radio loud ' as nss at similar accretion rates to deduce that quiescent bhs should be @xmath144 times _ less _ luminous in x - rays than quiescent nss .
their argument is a little convoluted since , if the radio and x - ray emission in quiescence come from the jet ( see 3.6 ) , one would think the bhs would be @xmath143 times _ more _ luminous , not 100 times less luminous .
therefore , naively , the jet argument only makes the discrepancy in figs . 8 and 9 more severe .
fender et al .
( 2003 ) get around this difficulty by postulating , in addition to different jet efficiencies , also different origins for the x - rays seen in quiescent bhs and nss .
it is unsatisfying that the reason for the difference in jet activity between bhs and nss `` remains unclear . '' at bottom , the authors are suggesting that in quiescence the bhs are in the ` jet - dominated ' regime and the nss are , `` if not jet - dominated , close to the transition to this regime . ''
a key uncertainty in this suggestion is whether the scaling relation that has been established for bhs between radio luminosity and jet luminosity ( @xmath145 ) is also valid for nss , as is assumed . in addition
, occam s razor suggests that the relative faintness of bhs is unlikely to be attributable to their stronger jets : in 4.5 , we show that sagittarius a@xmath3 , which is radiating in quiescence at the same level of eddington - scaled luminosity as a0620 - 00 , does not possess an energetically - important jet . _ coronal emission from bh secondaries : _
the quiescent x - ray luminosity of the bh systems has been attributed to a rotationally - enhanced stellar corona in the secondary star ( bildsten and rutledge 2000 , but see lasota 2000 ) . however , as discussed in detail by narayan et al .
( 2002a ) , the luminosities of three of the systems plotted in figures 8 and 9 exceed by a factor of 660 the maximum predicted luminosity of the coronal model ; likewise the three systems with adequate data quality show x - ray spectra that are harder / hotter than that typically seen in stellar coronae .
finally , if stellar coronae do contribute at some level , then the accretion luminosities of the bhs are even lower than our estimates , which would further strengthen the evidence for event horizons .
_ incandescent neutron stars : _ the quiescent luminosity of ns transients has been attributed to heating of the star s crust during outburst followed by cooling in quiescence .
this model likewise has problems .
( for details and references , see narayan et al .
briefly , the rapid variability of the prototypical ns transient system cen x-4 is not expected in a cooling model and implies that no more than about a third of the quiescent luminosity is due to crustal cooling .
furthermore , power - law tails that carry about half the total luminosity are observed for many ns transients ( e.g. , cen x-4 and aql x-1 ) .
these are unlikely to arise from a cooling ns surface , but could easily be produced by accretion .
finally , strong evidence for continued accretion in quiescence comes from optical variability , which is widely observed for these quiescent systems .
_ pulsar wind / shock emission : _ in another accretionless model , the ns transient switches to a radio pulsar - like mode in quiescence ( campana & stella 2000 ) . the x - ray luminosity is expected to be of order the `` pulsar shock '' luminosity @xmath146 , which is close to the observed level . both the observed power - law and thermal components of emission ( see the previous paragraph ) are naturally explained by this model : the former component is produced by the shock and the latter is radiated from the ns surface .
the lack of any significant periodicity in the quiescent emission ( in any electromagnetic band ) could be a problem for this model .
millisecond x - ray pulsars such as sax j1808.4 - 3658 do show periodicities in outburst , but that emission is clearly the result of accretion , not a pulsar wind / shock .
_ optical / uv luminosity : _ campana & stella ( 2000 ) note that the comparison shown in figs . 8 and 9 assumes the x - ray luminosity is an accurate measure of the accretion rate near the bh or ns .
they argue that the optical and uv luminosity also originates near the central object and should therefore be included in the comparison .
the non - stellar optical / uv luminosity is much greater than the x - ray luminosity .
when it is included , the difference between the bh and ns systems largely disappears .
however , as detailed in narayan et al .
( 2002a ) , there are some problems with this argument .
the level of optical / uv emission generated in the inner region of adafs is strongly suppressed by winds and convection , as evidenced by a comparison of the luminosities of nss and white dwarfs ( loeb , narayan & raymond 2001 ) .
it thus appears unlikely that the optical and uv emission is generated in the hot gas close to the accretor ( see shahbaz et al .
2005 ) . in the case of cvs
, it has been established that a large fraction of the optical / uv emission comes from the `` hot spot , '' and it is quite reasonable to expect that this is true for bh and ns systems as well ( there is some evidence to support this ; narayan et al .
further observations are needed to determine the origin of optical / uv emission in these systems .
in addition to the arguments discussed above , we briefly summarize three additional lines of evidence for the existence of event horizons .
all three are based on comparisons between bh and ns x - ray binaries .
as in the examples above , the first argument considers quiescent systems , whereas the latter two consider active states of accretion . in quiescence ,
a soft component of thermal emission is very commonly observed from the surfaces of accreting nss , which is widely attributed to either deep crustal heating ( brown , bildsten & rutledge 1998 ) or to accretion .
no such component is present in the spectrum of the quiescent bhb xte j1118 + 480 ( hereafter j1118 ) , as one would expect if the compact x - ray source is a bona fide bh that possesses an event horizon ( mcclintock et al .
2004 ) . because of the remarkably low column density to j1118 ( @xmath147 @xmath148 ) the limit on a hypothetical thermal source is very strong ( @xmath149 0.011 kev )
; it is in fact a factor of @xmath150 lower in flux than the emission predicted by the theory of deep crustal heating , assuming that j1118 has a material surface analogous to that of nss .
likewise , there is no evidence that accretion is occurring in quiescence onto the surface of j1118 , which is the mechanism often invoked to explain the far greater thermal luminosities of nss .
the simplest explanation for the absence of any thermal emission is that j1118 lacks a material surface and possesses an event horizon .
type i x - ray bursts are very common in ns x - ray binaries , but no type i burst has been seen among the bh systems . a model developed by narayan & heyl ( 2002 , 2003 ) , which reproduces the gross observational trends of bursts in ns systems , shows that , if the dynamical bhs have surfaces , they should exhibit instabilities similar to those that lead to type i bursts on nss .
remillard et al . ( 2006 ) , following earlier work by tournear et al . ( 2003 ) , measured the rates of type i x - ray bursts from a sample of 37 nonpulsing x - ray transients observed with _ rxte _ during 19962004 . among the ns sources
, they found 135 type i bursts in 3.7 ms of pca exposures ( 13 sources ) with a burst rate function consistent with the narayan & heyl model .
however , for the bh group ( 18 sources ) , they found no confirmed type i bursts in 6.5 ms of exposure .
their upper limit on the incidence of burst activity in these sources is inconsistent with the model predictions at a high statistical significance if the accretors in bhbs have solid surfaces .
the results provide strong indirect evidence for bh event horizons , and it would appear that the evidence can be refuted only by invoking rather exotic physics . likewise drawing on the extensive archive of _ rxte
_ data , done and gierlinski ( 2003 ) have examined the patterns of x - ray spectral evolution of active bh and ns sources and identified a distinct type of soft spectrum that is occasionally observed only in the bh sources .
they attribute this spectrum to thermal emission from the inner accretion disk ( corresponding to the high state discussed earlier ) .
they then argue that nss with a similar accretion rate can not exhibit such a simple , low - temperature spectrum because they would have a second component in the emission from the boundary layer where accreting matter impacts the stellar surface .
they present a thermal / nonthermal comptonization model for the boundary layer emission which has significant uncertainty , given that so many details of the accretion physics are complex and poorly understood .
nevertheless , done & gierlinski ( 2003 ) do appear to have identified a systematic difference in the x - ray spectra of accreting bhs and nss , and it is surely worth pursuing this signature in the effort to amass evidence for the reality of event horizons .
the supermassive bh in sgr a@xmath3 has a mass @xmath151 ( schdel et al .
2002 ; ghez et al .
2005a ) but an accretion luminosity of only @xmath152 ; thus the source is highly sub - eddington , @xmath153 . from our earlier discussion
, the accretion flow must be in the form of an adaf , i.e. , the radiating gas must be extremely hot and optically thin .
this expectation is confirmed by the @xmath154 k brightness temperature of the radio / millimeter emission ( shen et al .
2005 ) and the fact that most of the emission is in this band rather than at frequencies @xmath155 ( x - ray/@xmath34-ray band ) .
the adaf model has been successfully applied to sgr a@xmath3 ( narayan et al .
1995 , 1998a ; manmoto et al . 1997
; mahadevan 1998 ; yuan et al .
2003 ; to mention a few ) .
if we assume that the accretion flow is radiatively very inefficient , then it is easy to make a strong case for sgr a@xmath3 not having a surface ( narayan et al .
however , as broderick & narayan ( 2006 , 2007 ) showed , we can argue for the presence of an event horizon even without assuming an adaf .
if sgr a@xmath3 does not have an event horizon , but has a surface , then any emission from the surface will be blackbody - like . this is because we expect the surface to be optically thick .
however , the radio / millimeter emission mentioned above can not be from this surface because of its incredibly high @xmath154 k brightness temperature .
thus , the radio / millimeter radiation is from the accretion flow , and its luminosity gives a strict lower bound on @xmath123 . by equation ( [ lsurfacc ] ) then , we expect a luminosity of at least @xmath156 from the surface .
the surface radiation would be thermal and blackbody - like and should come out in the infrared ( as easily shown , given the luminosity and the likely area of the surface ) . however , there is no sign of it ! ( assuming the source has a surface ) as a function of the surface radius @xmath28 .
each limiting curve is derived from a limit on the quiescent flux of sgr a@xmath3 in an infrared band .
the hatched area at the top labeled `` typical riaf range '' corresponds to the mass accretion rate in typical adaf models of sgr a@xmath3 ( e.g. , yuan et al .
the horizontal dashed line represents the minimum accretion rate needed to power the bolometric luminosity of sgr a@xmath3 .
( from broderick & narayan 2006),width=316 ] figure 10 shows constraints on the mass accretion rate in sgr a@xmath3 .
the horizontal dashed line is the minimum accretion rate @xmath157 in sgr a@xmath3 ; this is the rate needed , even with a radiatively efficient flow ( @xmath89 ) , just to power the observed radio / millimeter radiation .
the four solid lines show the maximum mass accretion rate @xmath158 allowed if sgr a@xmath3 has a spherical surface of radius @xmath28 .
each curve corresponds to a measured limit on the steady quiescent infrared flux in a particular band ( stolovy et al .
2003 ; clenet et al .
2004 ; ghez et al .
2005b ) and provides an independent upper limit on @xmath158 .
for radii less than @xmath159 ( the upper limit on the size of sgr a@xmath3 ; shen et al .
2005 ) , we see that all four bands give upper limits on @xmath158 that are far below the minimum @xmath157 ; in fact , the discrepancy is larger than a factor of 100 in the @xmath160 band .
note that the discrepancy would be much larger if we assumed that the accretion flow is radiatively inefficient ( corresponding to the hatched region in fig .
the only way to avoid the large discrepancy illustrated in fig .
10 is to give up the assumption that sgr a@xmath3 has a surface .
obviously , if the source has an event horizon , then we do not expect any surface radiation , and there is no problem .
what if sgr a@xmath3 ejects all the accreting mass via a jet before the gas reaches the surface ? is this a viable explanation for the lack of surface emission ?
we can easily rule out this possibility . recall that the source of energy in an accretion system is gravity . at a bare minimum
, we know that matter is being accreted at a rate @xmath161 ( shown by the horizontal dashed line in fig .
10 ) in order to produce the observed radiation .
all of this mass _ has _ to fall into the potential well in order to release its energy .
if sgr a@xmath3 has a jet with a kinetic luminosity @xmath162 , the energy for this must also come from accretion .
we will then require an even larger @xmath161 , and hence a larger @xmath163 , since the accreting gas now has to power both the radiation and the jet .
the discrepancy with the observed limits on the quiescent infrared flux would then be even larger .
the arguments for the event horizon presented in 4.2 , 4.5 are based on newtonian ideas . some authors
( e.g. , abramowicz , kluzniak & lasota 2002 ) have questioned whether the arguments might be substantially modified by strong gravity in the vicinity of the compact accretor .
buchdahl ( 1959 ) showed , for a wide class of reasonable equations of state , that the smallest radius allowed for a compact non - rotating object is @xmath164 .
an object with this limiting radius has a gravitational redshift from its surface @xmath165 . at such modest redshifts
, we do not expect the effects of strong gravity to be particularly large , and so the arguments presented in 4.2 , 4.5 will continue to hold . recently , however , a new class of solutions has been discussed , which goes variously under the name of `` gravastar '' and `` dark energy star '' ( mazur & mottola 2001 ; chapline et al . 2003 ; visser & wiltshire 2004 ; carter 2005 ; lobo 2006 ) . in this model
, the radius @xmath28 of an object of mass @xmath63 is allowed to be arbitrarily close to @xmath103 : @xmath166 .
the surface redshift can then be arbitrarily large : @xmath167 although the gravastar model is very artificial , it is nevertheless interesting to ask whether such a model , which has no event horizon , can explain the observations described in 4.2 , 4.5 ( abramowicz et al . 2002 ) .
broderick & narayan ( 2006 , 2007 ) show that it can not .
we briefly summarize the arguments here .
a large value of @xmath168 means that any radiation emitted from the surface is redshifted greatly before it reaches the observer at infinity .
this looks like an easy way of hiding the surface luminosity .
however , a simple energy conservation argument shows otherwise . assuming steady state , in the limit as @xmath169 , the total luminosity observed at infinity , @xmath170 ,
_ must _ be equal to @xmath135 ( the rate of accretion of rest mass energy ) .
moreover , the effective radius of the source as viewed by a distant observer is @xmath171 , and so the surface radiation will be in the x - ray band for accreting bhbs ( 4.2 ) and in the infrared for sgr a@xmath3 ( 4.5 ) , exactly as in the newtonian analysis .
could the large gravitational redshift cause a large delay in the signals from the surface of the star , and could this be why we do not see the radiation ?
the extra delay due to relativity is easily estimated by considering null geodesics in the schwarzschild metric : @xmath172 since the dependence is only logarithmic , the extra delay is not significant .
for instance , even if we take @xmath173 to be comparable to the planck length @xmath174 cm ( the smallest length we can legitimately consider ) , the delay is only @xmath175 ms for a bhb and @xmath176 s for sgr a@xmath3 .
we assumed that the radiation from the surface would have a blackbody - like spectrum .
is this likely ? actually , it is virtually guaranteed as @xmath177 .
when the radius of an object is very close to the schwarzschild radius , most rays emitted from the surface are bent back on to the surface , and only a tiny fraction of rays escapes to infinity .
the solid angle corresponding to escape is @xmath178 . in this limit
, the surface behaves just like a furnace with a pinhole ( the textbook example of a blackbody ! ) , and so the escaping radiation is certain to have a nearly perfect blackbody spectrum .
could the surface luminosity escape in the form of particles rather than radiation ? by the blackbody argument given above , whatever escapes must be in thermodynamic equilibrium at the temperature @xmath179 observed at infinity . for the sources of interest to us , @xmath180 ev ( sgr a@xmath3 ) and kev ( bhbs ) .
in thermodynamic equilibrium , the only particles with any significant number density at these temperatures are neutrinos .
even if we include all three species of neutrinos , the radiation flux is reduced by a factor of only 8/29 ( broderick & narayan 2007 ) .
this is a small correction compared to the large discrepancies @xmath181 that we described in
4.2 , 4.5 .
finally , we note that our arguments for the event horizon are based on a steady state assumption .
specifically , we assume that the surface luminosity is proportional to the average mass accretion rate on the surface . in the gravastar model , in particular , one can imagine scenarios in which steady state is not reached .
however , broderick & narayan ( 2007 ) show that this explanation can be ruled out , at least with current gravastar models .
we thus conclude that strong gravity effects are unable to weaken our arguments for the presence of an event horizon in quiescent bhbs and in sgr a@xmath3 .
we must look to more conventional astrophysical explanations .
the only idea that we find somewhat plausible is that the radiation we observe from quiescent x - ray binaries and sgr a@xmath3 is not produced by accretion at all , but by some other process .
this would undercut all our arguments since we assume , as an article of faith , that the radiation we observe is powered by gravity through accretion .
given the success of the accretion paradigm in explaining a vast body of observations on a variety of bh systems in many different spectral states , it seems rather extreme to abandon the idea of accretion in just those particular sources where we find evidence for the presence of an event horizon .
the aim of this article is two - fold : to give an updated and current account of the adaf model , with an emphasis on applications to accreting bhs ( 3 ) , and to review the considerable body of evidence presently available for the presence of event horizons in astrophysical bhs ( 4 ) .
observations of bh binaries ( bhbs ) and active galactic nuclei ( agn ) indicate that , at luminosities of a few percent or less of eddington , accretion occurs via a very different mode than the standard thin accretion disk . at these luminosities ,
sources have hard x - ray spectra , quite unlike the soft blackbody - like spectra seen in more luminous sources .
prior to the establishment of the adaf model , observations in the hard state were modeled in an ad hoc way using empirical thermal comptonization models . in the mid-1990s ,
the adaf solution was shown to have precisely the densities , temperatures , radiative ( in)efficiencies and stability required to provide a physical description of the observations . moreover , since the adaf model is essentially mass - independent , observations of bhbs and agn are both explained with more - or - less the same adaf model .
the model gives satisfactory results for a wide range of luminosity , from about @xmath182 down to @xmath183 of eddington ( below which there are no observations ) .
the defining characteristic of an adaf is that the gas is not radiatively efficient .
a significant fraction of the energy released by viscous dissipation is retained in the accreting gas and advected with the flow .
the trapped thermal energy causes the accreting gas to be weakly bound to the bh .
based on this property , it was predicted already in the earliest adaf papers that sources in the adaf state would have strong winds and jets .
nonthermal radio emission has been detected in recent years from bhbs in the hard state and quiescent state , as well as from all agn with luminosities below @xmath184 of eddington .
these sources have spectra consistent with the adaf model , thus confirming a strong connection between adafs and jets / outflows .
a currently active topic of research is the role of agn feedback on galaxy formation .
research in this area can now draw on the extensive literature on adafs . of special interest in this work
are those accreting objects for which there exist estimates or constraints on the two system specific parameters : bh mass @xmath63 and mass accretion rate @xmath185 .
specifically , we are referring here to saggitarius a@xmath3 and a selected sample of quiescent bh and ns x - ray binaries .
we have featured these two examples because they provide strong observational evidence for the existence of the event horizon . in the case of the binaries
, one can make a plausible argument that , for comparable orbital periods , the mass accretion rates and luminosities of both types of systems should be comparable . remarkably , however , the bh systems are observed to be dimmer by a factor of @xmath186 .
we review a wide range of unsatisfactory attempts to explain this large luminosity difference .
in contrast , the adaf model provides an entirely straightforward explanation for the faintness of quiescent bhs : a ns must radiate the trapped thermal radiation that is advected with the accretion flow and rains down on its surface , whereas the bh hides the energy behind its event horizon .
additional evidence for the event horizon is provided in further comparisons of nss and bhs in outburst : the bhs lack type i x - ray bursts , and they lack a distinctive boundary - layer component of emission .
both of these properties are expected if the bhs possess event horizons , but hard to explain otherwise .
the supermassive bh in sgr a@xmath3 is extraordinarily quiescent .
the observations strongly support the existence of an adaf , and it is consequently easy to provide a compelling argument for an event horizon .
however , for sgr a@xmath3 , one can make an even stronger argument for the lack of a hypothetical material surface without at all invoking an adaf .
the large radio / millimeter accretion luminosity of sgr a@xmath3 , which has a brightness temperature @xmath187 k , obviously can not be emitted by an optically - thick surface , and so it must be radiated from the accretion flow .
this establishes a hard lower limit on the accretion luminosity and @xmath64 .
meanwhile , high angular resolution radio observations constrain the radius of the surface to be @xmath188 .
this constraint and the lower limit on @xmath185 predict a near - ir flux of thermal , blackbody - like surface emission that is far above the observed limits .
the obvious explanation is that there is no material surface , only an event horizon . as an added bonus ,
it is energetically impossible to explain away the lack of surface emission by appealing to mass loss in a jet .
the case for event horizons in both sgr a@xmath3 and in the stellar - mass bhs is robust against appeals to strong gravity or the leading models of exotic stars .
first , gr effects will be mild for nearly all conventional models of degenerate stars , whose radii are restricted to be @xmath189 and whose surface redshifts are therefore @xmath190 .
secondly , even for an exotic star ( a gravastar ) with an extraordinary surface redshift of a million or more , the full accretion luminosity from its surface will be delivered as x - rays ( in binaries ) or infrared ( in sgr a@xmath3 ) to a distant observer .
thus , extreme redshifts have practically no effect on the argument for the event horizon . to play on carl sagan s famous comment
, there will always be an absence of _ direct _ evidence for the event horizon , but this surely can not be taken as evidence of its absence in nature . on the contrary , and with a pun in mind , many indicators show that the event horizon is an inescapable reality . | as the luminosity of an accreting black hole drops to a few percent of eddington , the spectrum switches from the familiar soft state to a hard state that is well - described by a distended and tenuous advection - dominated accretion flow ( adaf ) .
an adaf is a poor radiator , and the ion temperature can approach @xmath0 k near the center , although the electrons are cooler , with their temperature typically capped at @xmath1 k. the foundational papers predicted that the large thermal energy in an adaf would drive strong winds and jets , as later observed and also confirmed in computer simulations . of chief interest , however , is the accreting gas that races inward .
it carries the bulk of the accretion energy as stored thermal energy , which vanishes without a trace as the gas passes through the hole s event horizon .
one thus expects black holes in the adaf regime to be unusually faint .
indeed , this is confirmed by a comparison of accreting stellar - mass black holes and neutron stars , which reside in very similar transient x - ray binary systems .
the black holes are on average observed to be fainter by a factor of @xmath2 .
the natural explanation is that a neutron star must radiate the advected thermal energy from its surface , whereas a black hole can hide the energy behind its event horizon .
the case for an event horizon in sagittarius a@xmath3 , which is immune to caveats on jet outflows and is furthermore independent of the adaf model , is especially compelling .
these two lines of evidence for event horizons are impervious to counterarguments that invoke strong gravity or exotic stars . , |
the laughlin wave function @xcite is a paradigm as a variational wave function @xcite .
its construction is simple , yet its implication is rich . it not only explains the most robust fractional quantum hall effects @xcite , but also embodies many exotic notions like fractionally charged excitations , fractional statistics , and composite - fermions / bosons @xcite .
its simplicity and fertileness make it very intriguing , and significant efforts have been dedicated to its nature @xcite . besides the conventional way of studying correlation functions , a more recent approach is studying the entanglement in the wave function @xcite . in ref .
@xcite , bipartite entanglement entropy in the laughlin wave function has been studied both analytically and numerically .
a noteworthy result is that , by studying the scaling behavior of the entanglement entropy , the topological entropy corresponding to the so - called `` total quantum dimension '' @xcite is extracted , confirming the fact that the laughlin wave function is a topologically ordered state .
however , a most natural way of quantifying the entanglement in the laughlin wave function has not yet been explored so far .
it is well known that unlike the integer quantum hall effect , for the fractional quantum hall effect , the coulomb interaction between the electrons is essential . for a non - interacting fermionic system ,
the multi - particle eigenstates are very simple they are slater determinants @xcite , the simplest kind of fermionic wave function satisfying the antisymmetry condition . for a generic interacting fermionic system ,
the interaction mixes up the slater determinants and a generic eigenstate @xmath2 can no longer be written in the form of a slater determinant @xcite , no matter how the single - particle orbitals are chosen .
specifically , as we shall prove below , the laughlin wave function can not be written as a single slater determinant .
now , from the point of view of approximation theory @xcite , a natural question is , how close can it be approximated by a slater determinant ?
this question leads to the notion of the optimal slater approximation of a fermionic wave function @xcite , namely , the slater determinant which maximize the overlap @xmath3 where @xmath4 denotes a slater determinant .
denote the maximal value of @xmath5 as @xmath6 .
it takes values from the interval @xmath7 $ ] .
apparently , the larger @xmath6 is , the closer the original wave function @xmath2 is to a slater determinant , and it is reasonable to say that the less the fermions are entangled with each other . in particular , if @xmath8 , the wave function is a slater determinant and the entanglement is zero .
therefore , a measure of entanglement can be defined as @xmath9 this is a geometric measure of entanglement in the sense that it involves only the notion of inner product between two states in a hilbert space .
it is a natural fermionic generalization of the geometric measure of entanglement for spin systems @xcite , or more precisely , composite systems whose hilbert space is the tensor product of that of its components .
the point is that , for spin systems , a generic wave function has no symmetry and one uses separable states to approximate it ; while here for a system of identical fermions , the target wave function is always antisymmetric , and hence the simpler wave function used to approximate it should also be antisymmetric .
we thus need to antisymmetrize a separable state and in doing so we get a slater determinant if not zero .
we note that while the geometric entanglement of spin systems have been extensively studied ( e.g. , see refs .
@xcite and references therein ) , it is much less discussed for identical particles , although some works on symmetric spin states @xcite can be translated into the bosonic language too .
the reason might be that the extra symmetry or antisymmetry constraint makes the optimization problem seemingly more complicated @xcite . despite this difficulty , however
, recently some progress has been made on the fermionic case @xcite .
in particular , in ref .
@xcite , an efficient algorithm to construct the optimal slater approximation of an arbitrary fermionic wave function was brought up .
it is this algorithm that enables us to study the geometric entanglement in the laughlin wave function . before proceeding to do so
, we note that the advantage of the geometric entanglement to bipartite entanglements is that , it is intrinsic , because it involves no partition , neither of the space nor of the particles . by avoiding partitioning the space ( the modes ) ,
it is representation - independent ; by avoiding partitioning the particles into two groups , it treats identical particles identically , i.e. , on an equal footing .
in contrast , the schmidt decomposition based bipartite entanglement depends on the partition of the system , which is inevitably arbitrary .
the algorithm we will utilize was already detailed in ref . @xcite . here
we sketch it briefly for the sake of completeness , and tailor it somehow for the specific problem .
the laughlin wave function of @xmath10 electrons , in the dimensionless form , is ( @xmath11 ) @xmath12 here @xmath13 is a positive odd integer and @xmath14 is a normalization factor . only the @xmath15 states are of interest
the @xmath16 state is simply a slater determinant @xcite .
we shall focus on the @xmath17 and @xmath18 states .
this is on the one hand due to the limitation of numerical resources and on the other hand because of the fact that the states with higher values of @xmath19 are less good as variational wave functions @xcite . by construction ,
each particle in the laughlin wave function ( [ lwf ] ) is in the lowest landau level .
more specifically , @xmath20 here @xmath21 is an @xmath22-tuple with @xmath23 . in the second line ,
we have introduced the single - particle orbitals ( @xmath24 ) @xmath25 which are an orthonormal basis of the lowest landau level . by ( [ lwf2 ] ) , we see that @xmath26 is supported by the @xmath27-dimensional single - particle hilbert space of @xmath28 actually , it is readily seen that these first @xmath29 @xmath30 s are also the natural orbitals @xcite of @xmath31 and each of them has a nonzero population . because an @xmath22-particle slater - determinant wave function has exactly @xmath22 populated natural orbitals ( each of population 1 )
, we see that the laughlin wave function with @xmath32 is not a slater determinant .
the fact that @xmath26 is supported by a finite dimensional single - particle hilbert space greatly facilitates the numerical calculation .
the @xmath22-electron hilbert space is spanned by the @xmath33 slater determinants @xmath34 with @xmath35 being an @xmath22-tuple .
the laughlin wave function can be expanded in terms of these slater determinants as @xmath36 this expansion ( to determine the coefficients @xmath37 ) is highly nontrivial .
but fortunately , some efficient algorithms , which we employ here , have been given in ref .
@xcite .
a generic slater determinant is supported by an @xmath22-dimensional subspace @xmath38 of @xmath39 .
suppose @xmath40 is an orthonormal basis of @xmath41 .
they are linearly related to the orthonormal basis @xmath42 of @xmath43 by a @xmath44 matrix @xmath45 , i.e. , @xmath46 the condition that @xmath47 be orthonormal requires that @xmath48 the @xmath49 identity matrix .
the slater determinant constructed out of @xmath47 is @xmath50 choosing another basis of @xmath41 will yield the same wave function up to a global phase .
the strategy to maximize the modulus of the overlap between @xmath51 and the laughlin wave function @xmath26 is very simple .
we start from some initial orbitals @xmath47 chosen randomly , and then fix all but one of them and try to optimize it with respect to the rest orbitals .
for example , let us fix orbitals @xmath52 and try to find an optimal @xmath53 with respect to them .
we note that the inner product between @xmath4 and @xmath26 can be written as ( @xmath54 ) @xmath55 here from the first line to the second line , we have used the fact that the wave function @xmath56 is antisymmetric and thus the @xmath57 terms summing over @xmath58 are all equal actually . from the second line to the third line , the integrals over @xmath59 were performed , and we have defined a single - particle state @xmath60 equation ( [ inner1 ] ) is in the form of the inner product of @xmath53 and @xmath61 . by the cauchy - schwarz inequality , we immediately see that the optimal @xmath62 should be proportional to @xmath63 . here an accidental and fortunate fact is that , @xmath61 is orthogonal to all @xmath64 because of the antisymmetry property of @xmath26 .
hence , when we update @xmath62 , the orthonormal condition ( [ mm ] ) between the @xmath65 s is still maintained .
the problem is then reduced to calculating @xmath61 . using ( [ expansion ] ) and ( [ gvsf ] ) , it is straightforward to get the compact formula @xmath66 we can then calculate @xmath63 using the laplace expansion @xcite for the determinant , and update the orbital @xmath53 , or more specifically , update the first column of @xmath67 . by doing so , the overlap @xmath5 increases .
now the point is that , while @xmath53 has been chosen as the optimal one with respect to the rest @xmath68 orbitals , @xmath69 is not necessarily optimized with respect to its accompanying orbitals , namely @xmath70 .
therefore , we can turn to @xmath71 and update it in a similar way .
this procedure can be repeated by sweeping across the @xmath22 orbitals ( the @xmath22 columns of @xmath67 ) in a circular way .
the overlap increases after each update and surely will converge as it is upper bounded by unity . to make sure that we hit the global maximum instead of being trapped by some local one , the program has to be run dozens of times with different initial values of @xmath45 . in fig .
[ m3m5 ] , we present the geometric entanglement @xmath72 against the number of electrons , both for @xmath17 and @xmath73 . each data point is obtained with at least 30 different initial values of @xmath45 , and the columns of @xmath45 are updated at least for 40 rounds . out of the final values of @xmath5 , the maximal one is then taken as @xmath74 .
we note that the @xmath17 case is surprisingly regular .
first , all the points lie very close to a straight line .
this linear behavior is not obvious at all , as we know that for the maximally entangled dicke state , in which half of the spins are up and half are down , @xmath75 is not linear in the number of spins but logarithm @xcite .
second , even if @xmath75 is found to be linear asymptotically in the large @xmath10 limit ; in the small @xmath10 limit , deviation from this behavior is expected .
that is , there should be some crossover regime .
however , the linear behavior extends well down to the smallest possible value of @xmath76 .
the @xmath73 case is less regular .
however , again there is no apparent crossover regime at the lower limit of @xmath77 .
the pattern of the data points suggests using the linear least squares approximation to fit the data .
that is , @xmath78 we get for @xmath17 , @xmath79 and for @xmath73 , @xmath80 in ref .
@xcite , it is found that for some topologically ordered systems such as the toric code , double semion , color code , and quantum double models , the geometric entanglement , like the bipartite entanglement based on space partition , consists of a term proportional to @xmath22 plus a constant term called the topological entropy .
if this is also the case for the laughlin wave function , we would have the constant term @xmath81 as @xmath82 for @xmath17 , and @xmath83 for @xmath84 .
however , this is not the case as shown by ( [ m3 ] ) and ( [ m5 ] ) .
actually , even the trend is in the wrong direction .
in conclusion , we have used an iterative algorithm to find the optimal slater approximation of the laughlin wave function and in turn to calculate the geometric entanglement in it .
to the best of our knowledge , this is the first time that the geometric entanglement in a strongly interacting fermionic system , which is of great physical interest , is calculated .
the importance of the laughlin wave function warrants such a study .
we find that the geometric entanglement @xmath75 scales linearly with the electron number @xmath10 .
while this is not so surprising , it is unexpected that the linear behavior extends well down to the lower limit of @xmath85 . at the time being , a full understanding of this finding is still lacking .
but , it is in alignment with the fact that the laughlin wave function has a very short correlation length ( it is on the order of the magnetic length ; or unity by our notation ) @xcite , which means that the @xmath85 case is already close to the thermodynamic limit .
the linear behavior of the geometric entanglement prompted us to compare the constant term with the expected topological entropy . as it turns out
, they do not agree .
this is in contrast to ref .
@xcite , where the topological entropy can be successfully extracted from the scaling behavior of the geometric entanglement .
our numerical result as a negative case then indicates that the relation between the geometric entanglement and the topological entropy is subtle .
this is reasonable , as the geometric entanglement involves no spatial partition , to which the topological entropy pertains , at least in its original derivation @xcite .
we are grateful to k. yang , f. pollman , and t .- c .
wei for stimulating discussions .
this work is supported by the fujian provincial science foundation under grant no .
2016j05004 , nsf of china under grant no . 11504061 , the china postdoctoral science foundation , and the foundation of lcp .
note that we are talking about the generic case .
interaction does not necessarily mean the total wave function can not be written as a slater determinant .
for example , for @xmath22 fermions in @xmath86 orbitals , the wave function is always a slater determinant , although superficially it is the linear superposition of @xmath86 slater determinants .
see appendix a in ref .
@xcite or lemma 1 in ref .
@xcite for more details . | we study numerically the geometric entanglement in the laughlin wave function , which is of great importance in condensed matter physics .
the slater determinant having the largest overlap with the laughlin wave function is constructed by an iterative algorithm .
the logarithm of the overlap , which is a geometric quantity , is then taken as a geometric measure of entanglement .
it is found that the geometric entanglement is a linear function of the number of electrons to a good extent .
this is especially the case for the lowest laughlin wave function , namely the one with filling factor of @xmath0 .
surprisingly , the linear behavior extends well down to the smallest possible value of the electron number , namely , @xmath1 .
the constant term does not agree with the expected topological entropy . |
recent discovery of the superconductivity in mgb@xmath0@xcite with relatively high transition temperature ( @xmath3k ) has invoked renewed interests in @xmath4-bonded materials .
electronically , the system is a @xmath5 electron system on layered honeycomb lattice , which immediately reminds us of graphite intercalation compounds ( gic ) such as lic@xmath6@xcite or kc@xmath7@xcite .
while the gic is considered to be a conventional superconductor with @xmath8k , mgb@xmath0 has an unusually high @xmath9 for @xmath4-bonded materials ( with a recent exception of c@xmath10-fet structure@xcite ) .
recently , choi _
et al._@xcite have used an ab - initio pseudopotential density functional theory to solve eliashberg s equation numerically , and have reproduced @xmath11k , isotope - effect exponent @xmath12@xcite , and have obtained a gapful bcs pairing , which is consistent with experimental results such as specific heat@xcite , tunneling and photoemission spectra@xcite , penetration depth@xcite , and the raman spectra@xcite . thus gic and
mgb@xmath0 both seem to be mainly phonon - mediated superconductors .
however , to realize a high @xmath13 , electron repulsion should not stand in the way of the phonon mechanism , so the question may be paraphrased : can the electron repulsion stand away from or possibly even assist the phonon - mediated pairing . on a more positive side , superconductivity from electron - electron repulsion itself is fascinating in many ways , but there are many open questions . while there is a growing consensus that high-@xmath13 cuprates
may be related to the electron correlation , we are only beginning to understand the link between the underlying band structure and the way in which the electron - mechanism superconductivity appears@xcite .
indeed , the way in which the superconductivity occurs is sensitively affected by the shape of the fermi surface .
recently , two of the present authors proposed@xcite that multi - band systems should open a new possibility of much higher @xmath9 , where a fully gapped bcs gap function can appear when the fermi surface consists of disconnected pieces , while the usual wisdom dictates that the repulsion - originated superconductivity should have , as in the cuprates , a strongly anisotropic gap function with nodes .
they have conceived and demonstrated that , for some two - dimensional lattice models , the pair scattering ( the coulombic matrix elements that scatter pairs of electrons across the fermi surface ) can occur across the pockets , which gives rise to a gap function that changes sign across the pockets with the same sign within each pocket .
so the purpose of the present paper is a combination of the above two motivations .
namely , we study whether the _ inter - band nesting _ can favor superconductivity arising from electron repulsion in a three - dimensional system , by taking the repulsive hubbard model on a stack of honeycomb lattices as a prototype . there
, the honeycomb , a typical non - bravais lattice , provides two pieces of the fermi surface arising from the two bands , while the stacking can provide a natural nesting along that direction .
so , if the pair scattering across the two bands along the nesting vector works favorably , we can expect a pairing from the interband nesting with the bcs gap with opposite signs across the two bands .
so we have studied the systematic dependence of the solution of eliashberg s equation with the multi - band fluctuation exchange approximation on the shape of fermi surface by changing the band filling and the third - direction hopping .
the presence of a strong inter - band nesting is found to indeed give rise to gap functions that change sign across the two bands , but , if we turn to the symmetry within each band , there exist two , nearly degenerate modes that behave respectively like s- and d - waves .
this implies ( a ) the electron repulsion can assist an s - wave like pairing when the phonon - mechanism pairing exists as a dominant mechanism .
we have further found that ( b ) when the electron repulsion is strong enough , the d - wave is realized , which should be , for a group - theoretic reason , a time - reversal broken linear combination of two symmetries with _ point _ nodes in the gap . according to the band calculation@xcite ,
the fermi surface of mgb@xmath0 consists of two tubular networks having a boron-@xmath14 character along with two cylinders of the boron-@xmath15 characters . as has been stressed by several authors@xcite ,
the nesting between these two @xmath5 bands are quite good .
although @xmath16 bands are considered to be important in that the electron - phonon interaction is much stronger in this band , here we concentrate on the @xmath17 bands in order to focus on the effects of the fermi surface nesting . the gic , on the other hand , has a much smaller inter - layer transfer energy ( @xmath18 ) , so that the fermi surface is a cylinder of carbon @xmath5 character with no dominant nesting vectors .
so in the latter part of the paper we shall cover both mgb@xmath0- and gic - situations in a systematic variation of @xmath18 and band filling .
the systematic study also serves to explore how the inter- and intra - band nesting compete in realizing superconductivity .
now let us start with the case corresponding to mgb@xmath0 .
spin - fluctuation - mediated superconductivity is here studied with the fluctuation exchange approximation ( flex ) developed by bickers _
et al._@xcite .
the model is a 3d two - band hubbard model with the repulsion @xmath19 on layered honeycomb lattice , @xmath20 where @xmath19 is the hubbard repulsion .
the essential ingredient here is the non - bravais lattice having a and b sublattices , which has two bands within a layer .
here we assume a non - staggered layer stacking as realized in mgb@xmath0 and gic ( fig.[lattice ] ) , which also depicts the inter - layer hopping @xmath18 .
hereafter we take the intra - layer hopping @xmath21 .
the noninteracting band dispersion for 3d honeycomb is @xmath22}.\ ] ] in the two - band flex@xcite , green s function @xmath23 , self - energy @xmath24 , spin susceptibility @xmath25 , and the gap function @xmath26 all become @xmath27 matrices , such as @xmath28 , where @xmath29 with @xmath30 being the matsubara frequency for fermions .
the self - energy is given by @xmath31 where the fluctuation - exchange interaction @xmath32 is @xmath33_{\alpha\beta}\!\!\!+\frac{1}{2}u^2\!\!\left[\frac{\chi^{{\rm irr}}(q)}{1+u\chi^{{\rm irr}}(q)}\right]_{\alpha\beta}\nonumber\\ & & -u^2\chi^{\rm irr}_{\alpha\beta}(q)\end{aligned}\ ] ] with @xmath34 here we denote @xmath35 with @xmath36 being the matsubara frequency for bosons , and @xmath37 the number of @xmath38-points on a mesh .
with dyson s equation , @xmath39_{\alpha\beta}=\left[g^{0}(k)^{-1}\right]_{\alpha\beta}+\sigma_{\alpha\beta}(k),\label{equend}\ ] ] where @xmath40 is the bare green s function @xmath41_{\alpha\beta}$ ] with @xmath42 the bare energy , we have solved eqns.([equbegin])-([equend ] ) self - consistently .
@xmath9 may be obtained from eliashberg s equation ( for the spin - singlet pairing ) , @xmath43 where @xmath26 is the gap function , and the pairing interaction @xmath44 is given as @xmath45_{\alpha\beta}\!\!\!-\frac{1}{2}u^2\!\!\left[\frac{\chi^{{\rm irr}}(q)}{1+u\chi^{{\rm irr}}(q)}\right]_{\alpha\beta}\nonumber\\ & & + u\delta_{\alpha\beta}.\end{aligned}\ ] ] @xmath9 is determined as the temperature at which maximum eigenvalue @xmath46 becomes unity .
the susceptibility , @xmath47_{\alpha\beta},\ ] ] may be expressed as diagonalized components , @xmath48^{2}+\left|\chi_{{\rm ab}}\right|^{2 } } .\ ] ] throughout this study , we take @xmath49 @xmath38-point meshes , and the matsubara frequencies @xmath50 from @xmath51 to @xmath52 with @xmath53 , which gave converged results .
let us first discuss the spin structure . we have fitted the shape of the fermi surface to that obtained by the band calculation@xcite to have @xmath54 at @xmath55 , where @xmath1 is band filling ( @xmath56 for half filling)@xcite .
the fermi surface , fig.[mgb2-fermi - chi ] , consists of two sets of tubular networks corresponding to the bonding and anti - bonding @xmath5 bands .
the spin - susceptibility @xmath57 displayed in the same figure shows a sharp peak around @xmath58 reflecting a good nesting along the @xmath59 direction ( arrows in fig.[mgb2-fermi - chi ] ) .
figure [ mgb2-phi ] shows the gap function obtained from eliashberg s equation .
the solutions having the largest @xmath46 are gapless @xmath60 and @xmath61 , which are degenerate .
as expected from kuroki - arita@xcite , a _ gapful _
@xmath62 does exist , although its @xmath46 is slightly smaller than that for the d - wave .
we have called the former solutions ` d - wave ' in that the gap function changes sign as @xmath63 azimuthally ( i.e. , within each band ) , while the latter gap ` s - wave ' in that it does not . as a hallmark that the scattering of the ( intra - band ) pairs across the inter - band nesting is exploited , all these solutions indeed have @xmath64 we can confirm in eliashberg s eq.([eliashberg ] ) that the sign change across the two bands works favorably in increasing @xmath46 if we note @xmath65 ( peaked around @xmath66 ) and a relation for multiband green s function @xmath67 which gives @xmath68 .
the fact that the dominant ` d - wave ' solutions are doubly degenerate can be understood by a group theoretical argument@xcite , in which these solutions belong to @xmath69 representation for the honeycomb system ( while ` s - wave ' to @xmath70 ) . the true gap function below @xmath9 to maximize
the gap should be a linear combination of the two d - waves , @xmath71 which breaks the time - reversal symmetry .
this combination has _
point nodes _ on the fermi surface , which is curious but natural as evident from fig.[d+id ] which superposes two sets of nodal planes to show how the nodal planes intersect each other along some lines for the doubly degenerate function and how these lines in turn intersect the closed fermi surface .
the reason why ` s - wave ' is only subdominant may be traced back to the fermi surface , which is rather extended in @xmath38 space in this particular case , so that there is an appreciable contribution from the intra - band pair scattering that the d - wave can exploit .
if we turn to the temperature dependence of @xmath46 in fig.[su - mgb2 ] , we can see that @xmath46 is significantly smaller than unity even for @xmath72 .
this implies that the spin fluctuation alone is not strong enough to realize the d - wave superconductivity in this temperature range for the band filling and @xmath18 taken here . on the other hand , the @xmath46 for the gapful ` s
- wave ' pairing is seen to have nearly the same magnitude as that of the gapless d - wave , although @xmath46 is again small .
thanks to the absence of nodes on the fermi surface , this one has a gapful pairing , which is eligible for assisting the phonon - mediated pairing if the electron - phonon interaction is considered on top of the electron - electron interaction@xcite .
so we conclude that inter - band spin fluctuations can work cooperatively with intra - band phonons to realize a gapful superconductivity .
let us depart from mgb@xmath0 to move on to the strong coupling regime in search of superconductivity from electron repulsion alone .
in general , pairing instability mediated by spin fluctuations in 3d systems is definitely weaker than that in 2d systems@xcite .
this has been shown in a flex study for the hubbard model by three of the present authors@xcite , who have identified its reason in the @xmath73 space volume fraction of the effectively attractive pair scattering region that is much smaller in 3d than in 2d .
furthermore , 3d systems have a strong tendency toward various magnetic orders , so that to identify the 3d systems that favor superconductivity from electron repulsion becomes a challenging problem . here
we have optimized the pairing instability in the layered honeycomb lattice by varying the inter - layer hopping @xmath18 and the band filling @xmath1 .
namely , we have searched for sets of parameter values that give large values of @xmath46 _ without _ encountering antiferromagnetic instability at low temperatures .
the resulting best parameter set is found to be @xmath74 ( inset of fig.[u8-n1.15-tz0.7-fermi - chi ] ) , for which the fermi surface and the ( inverse ) spin susceptibility are shown in fig.[u8-n1.15-tz0.7-fermi - chi ] . in fig.[su - u8-n1.15-tz0.7 ] we can see that the estimated @xmath75 ( @xmath76 ) .
the pairing symmetry is again @xmath77 .
apart from the above optimization , we have systematically explored how the inter - band nesting affects the pairing in the 3d honeycomb lattice .
since we wish to separate out the effect of the band filling and the hopping in the @xmath59 direction , we have done this along two paths displayed in fig.[diagram ] .
the first path starts from the parameter values corresponding to the @xmath17 bands in mgb@xmath0 , while the second path includes the parameter regime ( @xmath78 ) which corresponds to gic except for the value of @xmath19 .
the result along the first path is shown in fig.[su - u1.5-tz0.65 ] , where the @xmath1-dependence of @xmath79 is plotted for @xmath80 at @xmath55 .
we can see that the pairing instability becomes weaker as @xmath1 is increased , which is natural since the inter - band nesting becomes degraded along this path .
the result along the second path is displayed in fig.[su - u8-n1.2 ] , where @xmath46 is plotted as a function of @xmath18 for @xmath81 at @xmath55 .
here we have adopted a rather large @xmath82 , because we want to have sizeable @xmath46 over a wide range of @xmath18 including the case of bad nesting . there , we have covered both the layered case ( @xmath83 ) and a quasi-1d case ( @xmath84 ) . as indicated in fig.[su - u8-n1.2 ] , antiferromagnetism occurs ( i.e. , @xmath85 ) when the inter - band nesting becomes too strong as @xmath86 both from below and from above . before the transition to antiferromagnetism occurs @xmath46 increases both from below and from above , indicating that inter - band 3d nesting is effective , the layered case @xmath87 is more advantageous than the case of 2d @xmath88 and quasi-1d @xmath89 on the present path .
in conclusion , we have studied the possibility of spin - fluctuation mediated superconductivity in 3d honeycomb lattice systematically .
we have shown that if we take the parameter set corresponding to the @xmath17 bands in mgb@xmath0 , the spin fluctuation favors the gapful pairing , which suggests that the electron correlation can help the phonon in forming the cooper pairing .
experimentally , the electron repulsion acting constructively may be confirmed if some phase - sensitive method can detect the gap function having opposite signs in two @xmath5-bands .
when strong enough , the electron repulsion alone will give rise to a @xmath90-wave pairing , with a time - reversal broken @xmath77 symmetry associated with degenerate representations in the non - bravais lattice with peculiar point nodes on the fermi surface .
numerical calculations were performed at the computer center and the issp supercomputer center of university of tokyo .
this study is in part supported by a grant - in - aid for scientific research from the ministry of education of japan .
j. nagamatsu , n. nakagawa , t. muranaka , y. zenitani , and j. akimitsu , nature , * 410 * , 63 ( 2001 ) .
n. a. w. holzwarth , s. rabii , and l. a. girifalco , phys .
b * 18 * , 5190 ( 1978 ) .
t. inoshita , k. nakao , and h. kamimura , j. phys .
. jpn . * 43 * , 1237 ( 1977 ) .
d. p. divincenzo and s. rabii , phys .
b * 25 * , 4110 ( 1982 ) .
j. h. schn , c. kloc , and b. batlogg , science * 293 * , 2432 ( 2001 ) . h.
j. choi , d. roundy , h. sun , m. l. cohen , and s. g. louie , cond - mat/0111182 ( 2001 ) .
d. g. hinks , h. claus , and j. d. jorgensen , nature , * 411 * , 457 ( 2001 ) .
s. l. budko , g. lapertot , c. petrovic , c. e. cunningham , n. anderson , and p. c. canfield , phys .
. lett . * 86 * , 1877 ( 2001 ) .
y. wang , t. plackowski , and a. junod , physica c * 355 * , 179 ( 2001 ) .
f. bouquet , r. a. fisher , n. e. phillips , d. g. hinks , and j. d. jorgensen , phys .
* 87 * , 047001 ( 2001 ) .
t. takahashi , t. sato , s. souma , t. muranaka , and j. akimitsu , phys .
86 * , 4915 ( 2001 ) .
s. tsuda , t. yokoya , t. kiss , y. takano , k. togano , h. kitou , h. ihara , and s. shin , phys .
lett . * 87 * , 177006 ( 2001 ) .
g. karapetrov , m. iavarone , w. k. kwok , g. w. crabtree , and d. g. hinks , phys .
lett . * 86 * , 4374 ( 2001 ) .
p. szab , p.
samuely , j. kamark , t. klein , j. marcus , d. fruchart , s. miraglia , c. marcenat , and a. g. m. jansen , phys .
lett . * 87 * , 137005 ( 2001 ) .
f. manzano , a. carrington , and n. e. hussey , phys .
lett . * 88 * , 047002 ( 2002 ) .
x. k. chen , m. j. konstantinovi , and j. c. irwin , d. d. lawrie , and j. p. franck , phys .
lett . * 87 * , 157002 ( 2001 ) .
r. arita , k. kuroki and h. aoki , j. phys .
soc . jpn . *
69 * , 1181 ( 2000 ) ; phys .
b * 60 * , 14585 ( 1999 ) .
k. kuroki , and r. arita , phys .
b * 64 * , 024501 ( 2001 ) .
j. kortus , i. i. mazin , k. d. belashchenko , v. p. antropov , and l. l. boyer , phys .
lett . * 86 * , 4656 ( 2001 ) .
k. yamaji , j. phys .
. jpn . * 70 * , 1476 ( 2001 ) .
n. furukawa , j. phys .
* 70 * , 1483 ( 2001 ) .
n. e. bickers , d. j. scalapino , and s. r. white , phys .
. lett . * 62 * , 961 ( 1989 ) .
n. e. bickers and d. j. scalapino , ann .
* 193 * , 206 ( 1989 ) .
t. dahm and l. tewordt , phys .
b * 52 * , 1297 ( 1995 ) .
m. langer , j. schmalian , s. grabowski , and k. h. bennemann , phys .
lett . * 75 * , 4508 ( 1995 ) .
s. koikegami , s. fujimoto , and k. yamada , j. phys .
. jpn . * 66 * , 1438 ( 1997 ) .
h. kontani and k. ueda , phys .
lett . * 80 * , 5619 ( 1998 ) .
@xmath91 is appropriate here because @xmath92 ev [ k. voelker , v. i. anisimov , and t. m. rice , cond - mat/0103082 ] for @xmath93 orbitals in mgb@xmath0 , while @xmath19 for 2p orbitals should be less than @xmath94 ev as estimated for the o2p orbitals in the high-@xmath13 cuprates . m. sigrist and k. ueda , rev .
phys . * 63 * , 239 ( 1991 ) . here
we assume that phonons have long wavelength , so that they mainly mediate attractive pairing interaction within each band .
p. monthoux and g. g. lonzarich , phys .
b * 59 * , 14598 ( 1999 ) . | in order to study whether the inter - band nesting can favor superconductivity arising from electron - electron repulsion in a three - dimensional system , we have looked at the repulsive hubbard model on a stack of honeycomb ( i.e. , non - bravais ) lattices with the flex method , partly motivated by the superconductivity observed in mgb@xmath0 . by systematically changing the shape of fermi surface with varied band filling @xmath1 and the third - direction hopping , we have found that the pair scattering across the two - bands is indeed found to give rise to gap functions that change sign across the bands and behave as an s- or d - wave within each band .
this implies ( a ) the electron repulsion can assist _ gapful _ pairing when a phonon - mechanism pairing exists , and ( b ) the electron repulsion alone , when strong enough , can give rise to a d - wave - like pairing , which should be , for a group - theoretic reason , a time - reversal broken d@xmath2id with _ point nodes _ in the gap . 2 |
the first neutron star was discovered by bell and hewish in 1967 . since then
, the neutron star has spawned the topics of research in numerous studies in astronomy and nuclear physics .
a neutron star can be treated as a huge nucleus composed of neutrons , protons , and leptons @xcite .
the properties of neutron stars , like mass and radii , can be precisely measured with the gradual development of astronomical observation technology .
this development has promoted theoretical studies on the nuclear many - body theory @xcite , especially on the properties of nuclear matter at high densities , where high density refers to densities above 3 times the nuclear saturation density in the core of a neutron star . in the tolman - oppenheimer - volkoff ( tov ) equation @xcite , which is derived by solving the einstein equations for a spherically symmetric and time invariant system ,
the properties of a neutron star are fixed by the equation of state ( eos ) of neutron star matter , which is a charge neutral system in @xmath1 equilibrium .
the eoss of neutron star matter have been widely studied using various nuclear many - body theories , by the application of density functional theory ( dft ) and _ ab initio _ methods in nuclear physics @xcite .
the effective nucleon - nucleon ( @xmath0 ) interactions are adopted in dft , which are determined by fitting the empirical saturation properties of nuclear matter and the ground states of finite nuclei .
there are usually two schemes used in dft , i.e. , relativistic and non - relativistic approximation .
the most popular nuclear many - body theory in non - relativistic dft is the skyrme - hartree - fock theory @xcite , which was constructed based on the point - coupling @xmath0 interaction . in the relativistic version ,
covariant dft was proposed based on the one - boson - exchange picture @xcite .
the eoss of dft around the saturation density region can be constrained very well and are kept consistent within different effective @xmath0 interactions . however , their behaviors in the high density region , especially in the cores of neutron stars , show variability , which generate widely different predictions of the properties of neutron stars . with the discoveries of massive neutron stars ,
whose masses are around @xmath4 , many dft interactions are eliminated @xcite . on the other hand , in the state - of - the - art nuclear many - body methods , via _
ab initio _ calculations , a realistic @xmath0 interaction is used that is obtained by reproducing the @xmath0 scattering data .
the most popular realistic @xmath0 potentials are constructed by the meson - exchange picture @xcite and chiral effective field theory @xcite .
the @xmath0 potentials , based on chiral low - momentum expansions , are soft cores at short distances due to pion exchange ; whereas , in the meson - exchange potential , such as the bonn potential , the @xmath5 and @xmath6 meson exchanges are essential in the core region of the @xmath0 interaction , which generate very strong repulsive core .
there are also other high - momentum effects at short distance , such as the strong tensor force , besides strong repulsion @xcite which is related with the nucleon inner structure @xcite in the realistic @xmath0 interaction .
since 1950 , many microscopic many - body methods have been proposed to take such high momentum contributions into account .
_ used the hyper - netted chain - summation technique in the non - relativistic variational method @xcite and the av18 potential @xcite to calculate the eos of neutron rich matter @xcite with a three - body force .
krastev and sammarruca adopted the relativistic brueckner - hartree - fock ( rbhf ) @xcite theory to study the properties of neutron stars with bonn potentials @xcite .
et al . _ also compared their results with the calculations made by akmal using the auxiliary field diffusion monte carlo method @xcite .
discussed the properties of neutron star with @xmath0 interactions from chiral effective filed theory and constrained the neutron star mass - radius relation in the framework of renormalization group theory @xcite .
all of these values of eoss obtained from _ ab initio _ calculations agree well with each other in the high density region and generate similar properties of massive neutron stars . for neutron rich matter , the tensor effect in realistic @xmath0 interaction is suppressed largely in the @xmath7 isospin channel @xcite .
therefore , it is enough to explicitly consider only the short - range correlation in neutron rich matter .
based on this motivation , a relativistic central variational ( rcv ) method was proposed by including a central correlation jastrow function @xcite , which was inspired by the series works by panda _
et al . _
@xcite . in the rcv method ,
the solutions of the relativistic hartree - fock ( rhf ) method @xcite were chosen as the trial wave function of the total system .
the saturation properties obtained by the rcv method are @xmath8 different from the empirical data , likely because of the lack of tensor correlation in symmetric nuclear matter . however , the eos of pure neutron matter within the bonn potentials can reproduce the results from rbhf theory very well
. the solutions of the rhf method in nuclear matter are constructed by spinors with plane wave functions , the summation of spin and isospin for which can be easily and systematically evaluated via the feynman trace technology .
therefore , the calculation in the rcv method for neutron - rich matter is essential and economical , when compared with other time - consuming _ ab initio _ methods such as non - relativistic variational methods and the rbhf theory in asymmetric nuclear matter .
furthermore , the calculation procedures of the _ ab initio _ aforementioned methods , when they were applied to neutron star matter in @xmath1 equilibrium , are complicated , therefore , the equations of @xmath1 equilibrium are usually solved with various approximations instead of self - consistence treatments @xcite .
hence , the intention of this work is to apply the rvc method to calculate the eoss of neutron star matter in @xmath1 equilibrium self - consistently and use these eoss to study the properties of neutron stars with realistic @xmath0 interaction , bonn potentials .
in addition , we compare our results with the calculations of other _ ab initio _ methods and discuss the role of central correlation in neutron stars . in sect .
2 , we give a basic theoretical formulation of the rcv method . in sect .
3 , numerical results are presented on the properties of neutron stars and are compared with other calculations . in sect .
4 , we give a summary of the present work .
we firstly show the hamiltonian of the bonn potential @xcite , which is constructed based on the one - boson - exchange potential . it is defined as a sum of one - particle amplitudes of six bosons , consisting of @xmath9 and @xmath10 pseudovectors , @xmath11 and @xmath12 scalars , and @xmath6 and @xmath5 vector mesons , h&=&^a_it_i+^a_i , jv_i , j&=&d^3(x)(-i+m_n)(x)&&+_i=,,,,,d^3d^3(x)(x)(x)(x ) , where @xmath13 corresponds the nucleon field .
@xmath14 is the transfer momentum between two nucleons and @xmath15 are the masses of nucleons and mesons , respectively .
the @xmath16 matrices represent the vertex between nucleons and mesons , which are depicted below : & & _ ( 1,2)=-g^2_,&&_(1,2)=-g^2__1_2,&&_(1,2)=-()^2(/q_5)_1(/q_5)_2,&&_(1,2)=-()^2(/q_5)_1(/q_5)_2_1_2,&&_(1,2)=g^2__(1)^(2),&&^v_(1,2)=g^2__(1)^(2)_1_2,&&^t_(1,2)=()^2q_^(1)q^_(2)_1_2,&&^vt_(1,2)=i()_(2)^q_(1)_1_2 , where @xmath17 $ ] is an antisymmetric tensor gamma matrix , and the tensor coupling part between @xmath5 meson and nucleon has been neglected because the value of @xmath18 is negligible .
meanwhile , a monopole form factor should be considered , f_i(q^2)= , for each meson - nucleon vertex denoted by @xmath19 .
all coupling constants and cut - off momenta @xmath20 were determined by fitting the @xmath0 scattering data .
in the rcv method , one introduces a central correlation function of the wave function of the rhf theory as the trial wave function of the nuclear matter system @xcite , to treat the strong short - range repulsion , |=f| , where @xmath21 is the rhf wave function , and the correlation factor @xmath22 is chosen to be a product of two - body correlation functions @xmath23 ,
f=^a_i < jf(r_ij ) . here , @xmath23 is jastrow correlation function @xcite .
the total energy density with the correlation function is obtained as , _ c&=&=||&=&t+t_c+v .
where the explicit form of the correlated hamiltonian @xmath24 appears as & = & ^a_it_i+^a_i , j_ij&=&^a_it_i+^a_i , j\{f^(r_ij)[t_i+t_j+v_ij]f(r_ij)-(t_i+t_j)}. furthermore , the kinetic energy density part is not only related with the original one - body kinetic operator @xmath25 , t=^k_f_0p^2dp[pp+m_nm ] , but also the two - body operator with the jastrow correlation function , @xmath26 , t_c&=&c_b^k_f_0p^2dp[pp+m_nm]&-&^k_f_0p^2dpp^2dp\{[pp(p)+2m_n m(p)]i(p , p)+pp(p)j(p,p ) } , where the first and second terms correspond to the direct ( hartree ) and exchange ( fock ) contributions of @xmath26 , respectively and @xmath27 $ ] .
the isospin degeneracy is @xmath28 for symmetric nuclear matter and @xmath29 for pure neutron matter . @xmath30 and @xmath31 are defined as p(p)= , m(p)=. the potential energy density @xmath32 will be split into the direct term @xmath33 and exchange term @xmath34 .
the pseudo - vector mesons do not provide their contributions in the hartree approximation .
we have the result of @xmath33 as , v_d&=&-(^p_s+^p_s)^2-(^p_s-^n_s)^2&&+(^p_b+^n_b)^2+(^p_b-^n_b)^2 , where @xmath35 and @xmath36 are the scalar and nucleon vector densities , respectively , given as , ^p(n)_s&=&^k^p(n)_f_0p^2dpm(p),^p(n)_b&=&^k^p(n)_f_0p^2dp .
the proton and neutron cases are distinguished by the superscript @xmath37 .
the exchange contribution of @xmath11 meson @xmath38 as an example is given by , v^_e=^k_f_0pdppdp . here , the moment dependent functions @xmath39 and @xmath40 are various angular integrals related with the meson - exchange potentials , which are listed in the appendix of ref .
the contributions of other mesons can be expressed in similar forms .
finally , the total energy density is written as , _
c = t+t_c+v_d+_iv^i_e .
usually , several free parameters , @xmath41 , will appear in the jastrow function @xmath42 . in the present work ,
the correlation function is chosen as , f(r)=1-(c_0+c_1r+c_2r^2+c_3r^3)e^-c_4r . where the exponential term makes @xmath42 unity at large distance .
a natural choice from the unitary property of the correlation function is a normalization constraint on @xmath23 , [ nc ] d^3 r_ij[f^2(r_ij)-1]=0 , and should go to zero for small @xmath43 because of the repulsive core of @xmath0 interaction , which lead to @xmath44 .
furthermore , we should also ensure it is a monotonously increasing property at short distance , [ gcon ] f(0)0 .
we can then calculate the binding energy of nuclear matter after determining the remaining parameters with the variational principle .
the minimal value of the total energy should appear at @xmath45 and @xmath46 with the constraint ( [ gcon ] ) to make the jastrow correlation function increase at short distance .
therefore , we can obtain the relations between @xmath47 , and @xmath48 , c_1=c_4 , c_2=. now , there is only one parameter , @xmath48 , in the actual calculation , because @xmath49 is fixed by the normalization condition of the jastrow correlation function , eq .
( [ nc ] ) .
we would like to determine @xmath48 by the variational principle with the energy density , = 0 .
more detailed formulas of the rcv method can be found in ref .
@xcite . for neutron star matter
, there are not only the nucleons , but also the leptons , like electrons and muons .
all of them exist in the neutron star with the equilibrium conditions of the chemical potentials for the @xmath1 decay , [ bteq ] _
n&=&_p+_e,_&=&_e , where the chemical potentials @xmath50 , and @xmath51 are determined by the relativistic energy - momentum relation at the fermi momentum @xmath52 , _
i&=&^i_0(k_f)+e^_i(k_f),_&= & , where , @xmath53 and @xmath54 .
@xmath55 is the zero component of the self - energy of proton or neutron .
furthermore , the nucleon density conservation and charge neutrality are imposed in neutron star matter as , & & = _ n+_p,&&_e+_=_p .
the pressure of the neutron star system can be obtained with thermodynamics relation , as p()=^2=_i =
n , p , e,_i_i- .
the stable configurations of a neutron star then can be obtained from the well known hydrostatic equilibrium equations , by tolman , oppenheimer and volkoff @xcite for the pressure @xmath56 and the enclosed mass @xmath57 , [ tov ] & = & -,&=&4r^2(r ) , where , @xmath58 is the pressure of neutron star at radius , @xmath59 , and @xmath60 is the total star mass inside a sphere of radius @xmath59 .
once the eos @xmath61 is specified , @xmath62 being the total energy density ( @xmath63 is the gravitational constant ) , for a chosen central value of the energy density , the numerical integration of eq.([tov ] ) provides the mass - radius relation of neutron star .
after solving the equilibrium conditions of chemical potentials for various particles in the neutron star matter , eq .
( [ bteq ] ) , the binding energies per nucleon as functions of density are shown in the left panel of fig . [ re ] with three bonn potentials , bonn a , bonn b , and bonn c , which were fitted by the scattering data of @xmath0 system .
the binding energy of the bonn a potential is the smallest of the three bonn potentials , which is in accordance with the conclusion of rbhf theory for symmetric nuclear matter @xcite .
the tensor component of bonn a among these three potentials is the weakest , whose @xmath64 state probability , @xmath65 , is the smallest in deuteron .
furthermore , the difference of binding energies of these three potentials in neutron star matter is obviously less than that in the symmetric nuclear matter .
the effect of the tensor force becomes weaker with increasing neutron fraction and does not play any role in pure neutron matter .
similarly , the pressure as a function of energy density for neutron star matter is required as input data , when we calculate the properties of neutron stars in the tov equation .
the pressure - energy relations with bonn potentials are given in the right panel of fig .
their behaviors are very similar to those of the binding energy and are almost identical with each other .
it demonstrates that these eoss will generate similar properties of the neutron stars . the central correlation on kinetic energy per nucleon
is presented for the neutron star matter in fig .
the variational method can be achieved based on the competition between the correlation on kinetic energy and potential .
the central correlation on kinetic energy provides a repulsive effect to prevent two - nucleon approach at high densities , whereas the one on potential gives an attractive contribution to remove the repulsion of realistic @xmath0 interactions in short - range region .
finally , the correlations on kinetic and potential energies determine the minimum total energy and explicitly confirm the variational parameters in the jastrow function .
it can be found that the correlation on kinetic energy contributes almost half of the binding energy per nucleon for neutron star matter and plays a very essential role in the rcv method . in the rcv method
there is only one independent variational parameter , @xmath48 @xcite , which is shown in fig .
[ c4 ] as a function of density with bonn a , bonn b , and bonn c potentials . from this figure
, we can see that the central correlation strength increases slowly at low density , reaches a maximum value around the normal nuclear saturation density , and starts to decrease with increasing density thereafter .
it demonstrates that the central correlation will have to consider more variables to generate the saturation density and that it becomes weaker at high density , since the distance between two nucleons is already sufficiently compressed . as a function of density for neutron star matter with bonn
a , bonn b , and bonn c potentials.,width=340 ] with the eoss from the rcv method , we obtain the properties of neutron stars by solving the tov equation with bonn potentials .
the mass - radius and mass - density relations of neutron stars within our present framework are illustrated in fig .
[ rm ] . in the left panel ,
we plot the mass - radius relations of neutron stars with bonn a , bonn b , and bonn c potentials .
the maximum masses and corresponding radii predicted are almost the same , around @xmath2 and @xmath3 km , respectively .
these results are in good agreement with the previous _ ab initio _ calculations .
the maximum masses of neutron stars are @xmath66 in non - relativistic variational method by akmal _
_ @xcite .
similarly , the maximum masses and corresponding radii of neutron stars were given as @xmath67 and @xmath68 km in the rbhf theory with bonn potentials @xcite .
it indicates that our rcv method can economically describe the neutron star matter as well as both the non - relativistic full variational method and the rbhf theory .
the maximum masses and the mass - radius relations of neutron stars with three bonn potentials have obvious distinctions , since the neutron star matter includes not only neutrons but also protons
. the tensor force should contribute to some effects at the @xmath69 channel here , especially at low density , which is shown clearly in the mass - radius curves at large @xmath70 .
in the right panel of fig .
[ rm ] , we plot the mass - density relations of neutron stars with bonn potentials .
the central densities at maximum neutron star mass are @xmath71 @xmath72 for all of three bonn potentials , which are similar to the calculations of akmal _ et al . _
@xcite , @xmath73 @xmath72 and rbhf theory , @xmath74 @xmath72 @xcite . in the low density region
, the masses of neutron stars display distinguishable behaviors .
they are almost identical with increasing densities . in table
[ pro ] , the properties of neutron stars within the present framework are listed for bonn a , bonn b , and bonn c potentials .
they are compared with the ones taken from the rbhf theory @xcite .
no matter the maximum mass or the corresponding radii and central density , our results are only @xmath75 different from the ones calculated via the rbhf theory . furthermore
, the maximum masses of neutron stars in our calculation satisfy the requirements of recent observation on massive neutron stars @xcite , around @xmath4 .
the corresponding radius is also located within the constraint region worked out by hebeler _
et al . _
@xcite .
.the properties of neutron stars ( maximum mass , corresponding radii and central density ) within present framework and compared with the ones calculated by rbhf theory for bonn a , bonn b , and bonn c potentials . [ cols="^,^,^,^,^,^,^ " , ] the fractions of various particles appearing in neutron stars , which are neutrons , protons , electrons , and muons are plotted in fig .
[ ypabc ] with the three bonn potentials . at the beginning ,
the muon is absent for the @xmath1 equilibrium conditions .
when the electron chemical potential is larger than the muon mass , the muon will appear in the neutron star matter at densities less than @xmath76 @xmath72 , which is smaller than the density of muons in the rbhf theory .
the earliest appearance of the muon in the bonn a potential around the normal saturation density . at high density
, the fraction of muons will approach that of electrons . the proton fraction in bonn a has the largest magnitude compared to the other two bonn potentials , which should be due to its smaller tensor component .
we give the proton fractions with bonn potentials alone in fig .
[ urca ] . from this figure ,
the difference of proton fractions obtained from the three bonn potentials is clearly revealed , which becomes very obviously in the intermediate density region and decreases in the high density region .
the proton fraction in our calculation continues to increase with density and is in contrast to the case in rbhf theory , where it decreases at high density and the largest fraction is about @xmath77 @xcite . in rbhf theory , to simplify the calculation , the leptons were treated with the non - relativistic approximation . at high densities , the relativistic effect becomes more important , where the fermi momentum is very high and can be comparable with the light speed . therefore , the direct urca processes related with the cooling mechanism of neutron star could not occur easily with bonn b and bonn c potentials in rbhf theory , which should be satisfied by a proton fraction larger than approximately @xmath78 .
however , in our work , the direct urca processes can be produced in all of the bonn potentials and the densities appearing in direct urca processes are located between @xmath79 @xmath72 .
these densities will lead the neutron star to cool very rapidly .
we also give the dirac effective nucleon masses in neutron star matter in fig .
[ emn ] . for proton or neutron dirac effective masses reflecting the nucleon media effect ,
their magnitudes with different potential are almost equal , which is very similar situation to the energy and pressure cases .
if we compare the proton effective mass with the neutron effective mass , it can be found that the proton effective mass is larger than the neutron one at low density .
when the density is @xmath80 @xmath72 , the splitting of proton and neutron effective masses is reversed .
this behavior is quite different from the effective mass in rhf theory , where the proton effective mass should be larger than the neutron one in all density regions @xcite .
for rbhf theory , the splitting of proton and neutron effective masses for asymmetric nuclear matter is strongly dependent on the treatment of the @xmath81matrix .
brockmann and machleidt used the single - particle potential to extract the effective nucleon masses , where the neutron effective mass is larger than the proton one in neutron - rich matter @xcite ; whereas dalen _ et al . _ adopted the projection method to distinguish the spin components in the @xmath81matrix and calculated the effective nucleon mass with these interaction components in the rhf model @xcite . in this scheme ,
the proton effective mass is larger than the neutron one . to discuss the splitting of proton - neutron effective masses in our framework
, we show the kinetic and potential contributions of the nucleon effective mass in a neutron star with bonn a potential in fig .
[ emc ] . without the central correlation function ,
the kinetic energy is a one - body operator , which does not provide any contribution to the nucleon effective mass .
once the central correlation function is included , the central correlation on kinetic energy becomes a two - body operator and contributes to the nucleon effective mass . from fig .
[ emc ] , we can find that the potential generates the negative contribution to the effective nucleon mass .
in neutron - rich matter , the neutron effective mass will obtain more negative components compared to the proton one .
therefore , in the rhf model , the proton effective mass is larger than the neutron one in neutron - rich matter .
meanwhile , the central correlation on kinetic energy has the positive contributions to effective nucleon mass .
furthermore , its effect on neutrons is much higher than on protons at high densities . with the competition between the potential and kinetic energy , the proton effective
mass is larger than the neutron one at low density and with increasing density , the neutron effective mass is larger than the proton one .
therefore , the central correlation on kinetic energy plays a very important role in the splitting of proton - neutron effective masses . for dalen _ et al .
_ , they obtained the effective nucleon mass from the potential part ; whereas brockmann and machleidt considered the splitting of effective masses from the single particle potential , which is related to kinetic energy .
this may be the reason why there are opposite conclusions between the two groups , in rbhf theory on effective nucleon mass .
the rcv method based on the framework of rhf theory was applied to study the properties of neutron stars with one - boson - exchange potentials , i.e. , bonn potentials , which were determined by fitting the nucleon - nucleon ( @xmath0 ) scattering data . in neutron - rich matter
, the tensor force has a very small effect of its isospin feature .
therefore , it is essential to take the central correlation on the strong repulsion of @xmath0 interaction at short distances for the description of neutron - rich matter .
the equation of state ( eos ) of pure neutron matter obtained by a novel _ ab initio _ calculation of rbhf theory was completely reproduced by the present rcv method .
the eoss of neutron star matter in @xmath1 equilibrium with nucleons and leptons , were self - consistently solved in the rcv method with bonn a , bonn b , and bonn c potentials .
their behaviors were almost identical for the weak effect of tensor force in neutron - rich matter , since the difference among three bonn potentials only appeared in their tensor components .
the relativistic central correlation on kinetic energy played a very important role in the process of minimizing the total binding energy , with the variational principle , and gave half of the contribution to total binding energy .
furthermore , we found that the strength of the central correlation function was the strongest at the saturation density through the variational parameters , which correspond to the saturation mechanism of symmetric nuclear matter .
the properties of neutron stars were studied with the eoss of neutron star matter by solving the tov equation .
the maximum neutron star masses and corresponding radii were around @xmath2 and @xmath3 km , respectively , using the rcv method with bonn potentials .
the central densities of neutron stars were about @xmath71 @xmath72 .
these results are in good agreement with the calculations from rbhf theory and the non - relativistic variational method .
it demonstrated that the rcv method can describe neutron - rich matter reasonably and economically , compared with the conventional _ ab initio _ calculation .
the proton fractions in neutron star matter with the three bonn potentials showed some differences . the proton fraction in bonn
a potential was the largest and bonn c the smallest .
the direct urca processes would be generated in all of these potentials above the densities , @xmath79 @xmath72 with a proton fraction larger than approximately @xmath78 .
furthermore , the splitting of proton - neutron effective masses was reversed with increasing density of neutron star matter . in the low density region ,
the proton effective mass was larger than the neutron one ; whereas these behaviors are opposite at high densities , since the central correlation on kinetic energy played a more important role at high densities compared with the correlation on potential .
this may explain the conflicting predictions about the splitting of proton - neutron masses in the two different treatments in rbhf theory .
although , we can describe the properties of neutron stars very well , which are comparable to the results obtained by the other _ ab initio _ methods , it is necessary to take the tensor correlation into account to study symmetric nuclear matter and its saturation properties to reproduce the empirical data in future .
this work was supported in part by the national natural science foundation of china ( grant no .
11375089 and grant no .
11405090 ) .
99 n. k. glendenning , _ compact stars , nuclear physics , particle physics , and general relativity _
( springer - verlag , new york , 1997 ) .
f. weber , _ pulsars as astrophysical laboratories for nuclear and particle physics _ ( iop , bristol , 1999 ) .
m. prakash , i. bombaci , m. prakash , p. j. ellis , j. m. lattimer , and r. knorren , phys . rep . * 280 * , 1 ( 1997 ) .
h. heiselberg and m. hjorth - jensen , phys . rep . * 328 * , 237 ( 2000 ) .
j. oppenheimer and g. volkoff , phys . rev .
* 55 * , 374 ( 1939 ) .
r. c. tolman , phys . rev . * 55 * , 364 ( 1939 ) .
m. dutra , o. lourenco , j. s. s martins , a. delfino , j. r. stone , and p. d. stevenson , phys .
c * 85 * , 035201 ( 2012 ) .
m. dutra , o. lourenco , s. s. avancini , b. v. carlson , a. delfino , d. p. menezes , c. providncia , s. typel , and j. r. stone , phys .
c * 90 * , 055203 ( 2014 ) .
d. vautherin and d. m. brink , phys .
c * 5 * , 626 ( 1972 ) .
m. bender , p. h. heenen , and p. g. reinhard , rev .
* 75 * , 121 ( 2003 ) .
j. r. stone and p. g. reinhard , prog .
nucl . phys . *
58 * , 587 ( 2007 ) .
b. d. serot and j. d. walecka , adv .
* 16 * , 1 ( 1986 )
. p. ring , prog .
* 37 * , 193 ( 1996 ) .
j. meng , h. toki , s. g. zhou , s. q. zhang , w. h. long , and l. s. geng , prog .
phys . * 57 * , 470 ( 2006 ) .
p. b. demorest , t. pennucci , s. m. ransom , m. s. e. roberts , and j. w. t. hessels , nature ( london ) * 467 * , 1081 ( 2010 ) .
j. antoniadis , p. c. c. freire , n. wex , t. m. tauris , r. s. lynch _
et al . _ ,
science * 340 * , 6131 ( 2013 ) .
r. machleidt , adv .
* 19 * , 189 ( 1989 ) .
e. epelbaum , h .- w .
hammer , and ulf - g .
meiner , rev .
phys . * 81 * , 1773 , ( 2009 ) .
r. machleidt and d. r. entem , phys . rep . *
503 * , 1 ( 2011 ) .
r. jastrow , phys . rev . * 98 * , 1479 ( 1955 ) .
h. toki , z. physik a * 294 * , 173 ( 1980 ) .
m. oka and k. yazaki , prog .
. phys . * 66 * , 556 ( 1981 ) .
v. r. pandharipande and r. b. wiringa , rev .
phys . * 51 * , 821 ( 1979 ) .
r. b. wiringa , v. g. j. stoks , and r. schiavilla , phys .
c * 51 * , 38 ( 1995 ) .
a. akmal , v. r. pandharipande , and d. g. ravenhall , phys .
c * 58 * , 1804 ( 1998 ) .
r. brockmann and r. machleidt , phys . rev .
c * 42 * , 1965 ( 1990 ) .
p. g. krastev and f. sammarruca , phys .
c * 74 * , 025808 ( 2006 ) . s. gandolfi , a. yu .
illarionov , k. e. schmidt , f. pederiva , and s. fantoni , phys .
c * 79 * , 054005 ( 2009 ) .
k. hebeler , j. m. lattimer , c. j. pethick , and a. schwenk , phys .
lett . * 105 * , 161102 ( 2010 ) .
k. hebeler , j. m. lattimer , c. j. pethick , and a. schwenk , astrophys .
j. * 773 * , 11 ( 2013 ) .
j. hu , h. toki , w. wen , and h. shen , phys .
b * 687 * , 271 ( 2010 ) .
y. wang , j. hu , h. toki , and h. shen , prog .
* 127 * , 739 ( 2012 ) .
j. hu , h. toki , and y. ogawa , prog .
103d02 * , ( 2013 ) .
j. hu , h. toki , and h. shen , j. phys .
g. * 38 * , 085105 ( 2011 ) .
p. k. panda , d. p. menezes , c. providncia , and j. da providncia , phys .
c * 71 * , 015801 ( 2005 ) .
p. k. panda , j. da providncia , and c. providncia , phys .
c * 73 * , 035805 ( 2006 )
. p. k. panda , c. providncia , and j. da providncia , phys .
c * 75 * , 065806 ( 2007 ) .
a. bouyssy , j. f. mathiot , n. van giai , and s. marcos , phys .
c * 36 * , 380 ( 1987 ) .
a. li , j. n. hu , x. l. shang , and w. zuo , phys .
c * 93 * , 015803 ( 2016 ) .
f. sammarruca , int . j. mod .
e * 19 * , 1259 ( 2010 ) .
e. n. e. dalen , c. fuchs , and a. faessler , eur .
j. a * 31 * , 29 ( 2007 ) .
e. n. e. dalen and h. muether , int
e * 19 * , 2077 ( 2010 ) . | the properties of neutron stars are investigated within the relativistic central variational method by using a realistic nucleon - nucleon ( @xmath0 ) interaction .
the strong repulsion of realistic @xmath0 interactions at short distances is treated by a jastrow central correlation function , whose form is completely determined through minimization of the total energy of the nuclear many - body system .
the relativistic hartree - fock wave functions are chosen as the trial wave function . in this framework ,
the equation of state of the neutron star matter in @xmath1 equilibrium is obtained self - consistently .
we further determine the properties of neutron stars via the tolman - oppenheimer - volkoff equation using bonn a , b , and c potentials .
the maximum masses of neutron stars with these realistic potentials are around @xmath2 and their corresponding radii are around @xmath3 km .
these results are in accordance with the calculations of the relativistic brueckner - hartree - fock theory with the same potentials .
furthermore , we also find that the splitting of proton - neutron effective masses will be reversed at high density in the neutron star matter , which are caused by the contribution of short - range correlation on kinetic energy . |
we consider here the collective behavior of the system of @xmath2 two - level atoms coupled to a single mode of the electromagnetic field in a resonator .
the useful form for the atom - field interaction was proposed in the rotating wave approximation ( rwa ) by tavis and cummings @xcite . in their model
@xmath1 identical two - level atoms interact via dipole coupling with a single - mode quantized radiation field at resonance , so that the hamiltonian is given by @xmath3 here @xmath4 is a frequency of the electromagnetic field and @xmath5 is the level splitting of the two - level atoms .
the operators @xmath6 are collective spin variables of @xmath1 two - level atoms .
these operators are defined as @xmath7 where @xmath8 s are pauli matrices .
they satisfy the _
su(2 ) _ algebra .
@xmath9 are the annihilation and creation operators of the field . due to historical reasons
the tavis - cummings ( tc ) model is often called the dicke model @xcite .
we concentrate here on the case of exact resonance , i.e. @xmath10 . in this case
the system exhibits a most interesting collective behavior . for simplicity ,
time will be measured in units of the coupling constant @xmath11 , i.e. we assume in the following that @xmath12 .
the one - atom version of the tc model , i.e. the jaynes - cummings model(jc ) @xcite has recently attracted much attention because of e.g. the spectacular experiments reported in @xcite . in this work the properties of a dilute flux of atoms excited to rydberg states and interacting with a single quantum mode of the electromagnetic field
were analyzed in a high - q cavity .
the results reported in @xcite demonstrate that the physical assumptions underlying the jc model can indeed be experimentally realized .
this triggered a lot of theoretical interest towards the one - atom micromaser(microlaser ) ( see @xcite and references therein ) .
the analysis of the jc model is greatly simplified by the fact that the eigenvalue problem of the jaynes - cummings hamiltonian can be solved exactly .
an algebraic approach to the solution of generalized jaynes - cummings types of models can be found in @xcite .
the approach to the tavis - cummings type models based on the algebraic bethe ansatz is formulated in @xcite .
an analytical solution of the @xmath1-atom case , i.e. the tc model , is much more complicated than that of jc model and is thus still far from completion . in this article
we analyze this model and develop an analytical technique that is applicable to a variety of problems of tc type . to begin with we notice that hamiltonian ( [ eq1 ] ) belongs to a class of operators that can be expressed in the form : @xmath13 here
@xmath14 is an analytic function of @xmath15 with real coefficients , while the operators @xmath16 satisfy the commutation relations @xmath17
= \pm a_\pm.\ ] ] we also assume the operator @xmath18 to be self - adjoint .
this means that , for any irreducible representation ( irrep ) , the operators @xmath19 must satisfy the conditions @xmath20 it is worth mentioning that hamiltonians of type ( [ eq1 ] ) are also used in description of raman and brillouin scattering , and frequency conversion , as well as parametric amplification that involves trilinear boson operators@xcite : @xmath21 the equivalence of hamiltonians ( [ eq1 ] ) and ( [ lin3 ] ) can be readily seen by applying the schwinger transformation @xmath22 the hamiltonians of the type of eq .
( [ h_g ] ) are usually analyzed by approximating them with an exactly solvable one .
the solvable hamiltonian is usually quadratic in the boson operators or linear in the generators of a classical lie algebra . in this approach the so - called parametric approximation or its variations @xcite
are often used .
the basic assumption of this approximation is that one of the quantum modes or subsystems is prepared in a highly excited ( often coherent ) state , or in a state close to the vacuum .
this approach , however , puts certain restrictions on the type of possible initial conditions as well as on the time span over which the quantum dynamics can be followed .
another type of an approach to the solution of hamiltonian ( [ h_g ] ) is based on perturbation theory for nonlinear algebras @xcite .
this approach requires however the existence of a small parameter . in the tc model
such a parameter was found in the case the atoms are in completely symmetric states ( @xmath23 ) @xcite .
the basic idea of the present work is to combine these two approaches through algebraic methods as applied e.g. in @xcite ) . to this end
we reformulate the hamiltonian in terms of an algebra that better allows the diagonalization of the hamiltonian .
this idea was already used by holstein and primakoff @xcite .
they expressed the generators @xmath6 of the _
su(2 ) _ algebra in terms of boson operators @xmath24 , @xmath25 here @xmath26 is an index that characterizes the irrep of _
su(2)_. however , in @xcite the square root in the transformation eq .
( [ h_p ] ) was in the end replaced by unity , which amounts to applying the so - called `` weak field '' approximation ( @xmath27 ) .
obviously , this approximation corresponds to zeroth order in the expansion of the problem with respect to parameter @xmath28 .
transfromation ( [ h_p ] ) has also been applied @xcite with expansion up to second order .
we consider here the case when the operators @xmath19 in eq .
( [ h_g ] ) are generators of a polynomial deformation @xmath0 of the lie algebra _
su(2 ) _ @xcite .
numerous physical applications exist for polynomially deformed algebras @xcite .
a particularly interesting , in view of the present problem , application of deformed algebras was developed by karassiov ( see in @xcite and refs . therein ) .
the method to be introduced below is an extension of karassiov s method .
we introduce here the notion of a polynomial algebra of excitations ( pae ) . in this algebra the coefficients of the structure polynomials
are @xmath29-numbers , rather than casimir operators as is typical of polynomial deformations .
we derive an exact mapping between isomorphic representations of two arbitrary pae .
we also provide a classification of isomorphic representations of polynomial algebras of excitations . in our approach classes of isomorphic representations
are specified by the multiplicity of the maximal and minimal eigenvalues of the operator @xmath30 in the given representation .
we formulate then an analytical approach that allows us to expand the hamiltonian , when expressed in terms of pae , as a perturbation series . for completely symmetric states of atoms
our results agree with those reported in @xcite .
our formalism provides however a solution of the problem for any value of @xmath26 , which allows us to discuss new physical effects in the dicke model .
the article is organized as follows . in the next section
ii we discuss the irreducible representations of pae . in the section iii
we apply the general approach to the tavis - cummings model , and in section iv construct the perturbation theory for the tc hamiltonian and solve its eigenvalue problem up to third order .
the generalized @xmath31-atom quantum rabi frequency is defined for arbitrary quantum states of the system . in the section
v we use the zero order approximation for the tc hamiltonian to calculate the intensity of spontaneous emission of atoms prepared in the state of thermal equilibrium with the resonator mode .
we show that the correlation of the atoms due to interaction with the field gives rise to the enhancement of spontaneous emission as compared to the atoms in the absence of resonator . in conclusion ,
we discuss possible further applications of the methods developed here .
technical details of the algebraic manipulations are given in the appendix .
the coefficients of the structure polynomial of a polynomially deformed algebra are usually expressed through the casimir operators of the algebra . in this section
we discuss representations of a special class ( pae ) of polynomially deformed algebras when the coefficients of the structure polynomial are @xmath29-numbers .
we denote a pae with a structure polynomial of order @xmath32 as @xmath33 .
formally @xmath33 is an associative algebra with unity , defined by three generators @xmath34 .
these generators satisfy two basic commutation relations , eq .
( [ eq2 ] ) . as can be readily seen from these commutation relations , @xmath35 = 0 $ ] .
we can thus assume that @xmath36 here @xmath37 is a structure polynomial of order @xmath32 , whose coefficients are generally complex numbers .
the terminology is chosen in analogy to the structure functions of quantum algebras ( @xmath38-deformed algebras ) @xcite , and the structure constants of the linear lie algebras .
the set of @xmath32 real roots of the structure polynomial is denoted as @xmath39 . in physical applications the operators @xmath16 of eq .
( [ eq2 ] ) often play the role of creation and annihilation operators of collective excitations .
therefore , hereafter the algebra @xmath33 will be referred to as the polynomial algebra of excitations ( pae ) of order @xmath32 .
notice that this algebra is different from the algebra @xmath0 .
all pae s are completely defined by @xmath40 c - numbers , the coefficient @xmath41 and the @xmath32 roots @xmath42 of the structure polynomial . in the case of @xmath0
, however , the structure polynomial has some coefficients in the form of casimir operators . below we consider two elementary , but important for the following , examples of pae . to begin we assume , without loss of generality , that @xmath43
indeed , in the case @xmath44 , it is always possible to renormalize the generators of @xmath33 , @xmath45 such that the commutation relations eq .
( [ eq2 ] ) remain intact . using these commutation relations
it can also be readily seen that @xmath46 as indicated above , @xmath47 and @xmath48 have the physical meaning of creation and annihilation operators of collective excitations ( quasiparticles ) , while @xmath49 is the operator for the number of excitations .
the most simple and important example of a pae of first order , @xmath50 , is provided by the well - known heisenberg - weil lie algebra , viz .
@xmath51 here @xmath52 are the usual boson operators . for the sake of simplicity , in what follows
we will denote the generators of @xmath50 by @xmath52 .
the algebra @xmath50 allows us to construct the irrep of any other pae of higher order @xmath53 as a multiple tensor product of @xmath50 .
as practical example let us describe in detail a second order pae denoted in the paper by @xmath54 .
it is relevant to the algebra _
su(2 ) _ given by the commutation relations @xmath55
= \pm s_{\pm , \quad } \left [ { s _ + , s _ - } \right ] = 2s_3.\ ] ] it is plain that @xmath56 for every irrep the casimir operator @xmath57 is equal to @xmath58 , where @xmath59 is the identity operator .
the corresponding pae of second order , @xmath54 , is constructed such that @xmath60 obviously , @xmath54 has a matrix irrep isomorphic to the irrep of _
su(2 ) _ with the same @xmath26 . notice that _ only _ in this representation of @xmath54 the condition eq.([ad_con ] ) is fulfilled .
this gives us a motivation to denote the three generators of @xmath54 as @xmath61 .
below we describe a general method to construct a realization of any pae @xmath33 through the algebra @xmath50 . choosing a root @xmath62 of the structure polynomial we can construct this realization in the form @xmath63 here @xmath64 are the usual boson operators associated with the chosen root @xmath62 of the structure polynomial @xmath37 .
the corresponding fock vectors , i.e. the eigenvectors of @xmath65 , are denoted as @xmath66 . in what follows the chosen root @xmath62 of the structure polynomial
will be referred to as the _ pivotal _ root .
notice that the product in ( [ eq9 ] ) always contains exactly @xmath32 multipliers regardless of the multiplicity of the root .
since @xmath65 has a well - defined discrete spectrum , the square root function is defined in the form of spectral decomposition .
we specify the branch of the square root by choosing @xmath67 .
it is easy to see that the condition eq.([eq3 ] ) and the commutation relations eq.([eq2 ] ) for generators of @xmath33 are satisfied .
a useful automorphism @xmath68 , @xmath69 of @xmath50 allows us to construct another realization of @xmath33 through @xmath50 , @xmath70 it is worth mentioning that transformations similar to eqs .
( [ eq9 ] ) , ( [ eq12 ] ) have been introduced earlier under the name of multiboson realizations of bose operators .
these multiboson realizations satisfy the usual boson commutation relations , @xmath71 = 1 $ ] @xcite . applying the realizations eqs.([eq9]),([eq12 ] ) in any representation of @xmath50
, we can construct a representation of @xmath33 . in the case of realization eq.([eq9 ] ) , an irreducible representation of @xmath33 is constructed through application of the operator @xmath72 to the vacuum vector @xmath73 .
the finite dimensional representation can be constructed in the case when the root @xmath74 is separated from @xmath62 by a natural number @xmath75 .
then it can be readily seen that @xmath76 .
indeed , provided that @xmath77 , it is not difficult to show that the norm of the vector @xmath78 vanishes , i.e. @xmath79 . the general construction can be exemplified by @xmath54 .
when @xmath26 is integer or half - integer , @xmath80 is integer , and @xmath54 has a finite dimensional irrep which is isomorphic to the corresponding irrep of _
su(2)_. in the case of realization eq.([eq12 ] ) , the corresponding irrep is constructed through application of the operator @xmath81 to the vacuum vector @xmath73 . using an argumentation similar to that given above
, it can be seen that the finite dimensional representation can now be constructed provided that the root @xmath82 is separated from @xmath62 by a natural number .
the meaning of transformations eqs.([eq9 ] ) and ( [ eq12 ] ) becomes now transparent .
the realization eq.([eq9 ] ) corresponds to the case when the operator @xmath72 is a creation operator , while the realization eq.([eq12 ] ) corresponds to the case when @xmath72 is an annihilation operator . since the spectrum of operator @xmath65 is a set of natural numbers and zero , the operator @xmath83 in eq .
( [ eq9 ] ) has a non - negative spectrum .
therefore , the argument of the square root function is a positive operator in the finite dimensional subspace , where the structure polynomial @xmath84 has non - negative spectrum . in this case
the operators @xmath47 and @xmath48 are hermitian conjugated . in the subspace
corresponding to the negative values of the spectrum of the structure polynomial , the argument of the square root function in eq.([eq9 ] ) has a negative spectrum .
the operators will be anti - conjugated , i.e. @xmath85 , which is not plausible .
thus , in a physical problem ( see ( [ ad_con ] ) ) , we should only consider those irrep for which the spectrum of @xmath84 is nonnegative .
the same is true for eq .
( [ eq12 ] ) .
the two relations , eqs.([eq9 ] ) and ( [ eq12 ] ) , between any two algebras @xmath33 and @xmath50 , show that there is no principal difference as to how exactly the meanings of the creation and annihilation operators of collective excitations are prescribed to the pair @xmath16 .
the important point is that the pair exists .
it is the physical problem in question which prescribes the meaning of operators @xmath16 and determines the location of the equidistant spectrum of @xmath49 on the real axis .
should one be interested in the eigenvalues of @xmath30 , to the right of the pivotal root @xmath86 , it is necessary to choose the transformation eq.([eq9 ] ) , while for the region to the left of @xmath86 , it is necessary to use the realization ( [ eq12 ] ) . as was explained above
, we choose the region such that the structure polynomial is nonnegative .
the general considerations given above can be illustrated by the @xmath54 algebra . in the case of the realization of @xmath54 connected to the pivotal root @xmath87
it is necessary to use eq.([eq9 ] ) .
this leads to the conventional holstein - primakoff representation , @xmath88 in the case of @xmath89 , we use the realization eq .
( [ eq12 ] ) , @xmath90 the spectra of @xmath91 and @xmath92 are limited from above by the value @xmath93 , while the subspace spanned by eigenvectors @xmath94 of operator @xmath49 is irreducible ( cf .
( [ h_p ] ) ) .
obviously all the realizations of pae constructed through eqs .
( [ eq9 ] ) , ( [ eq12 ] ) are fully characterized by the dimension @xmath75 of the invariant subspace and by the order @xmath95 of the left and @xmath96 of the right roots defining the corresponding irreducible representation .
thus an irrep of pae is characterized by a set of parameters @xmath97 .
such irrep we will denote by @xmath98 .
for instance , @xmath99 means the representation of @xmath50 , while the irrep of @xmath54 is @xmath100 .
an isomorphism between irreps of @xmath33 and @xmath101 that belong to the same class @xmath98 , is given by @xmath102 or by @xmath103 in eqs .
( [ eq13]),([eq14 ] ) the operator argument of the square root function should be taken after identical multipliers in the nominator and denominator are cancelled .
the pivotal roots @xmath62 and @xmath104 define a vacuum vector of the irrep .
compare these expressions with eqs.([eq9]),([eq12 ] ) .
to recapitulate : in the general case two irreducible representations @xmath98 and @xmath105 of @xmath33 and @xmath101 are isomorphic provided that @xmath106 , and @xmath107 for symmetric irreps , i.e. when @xmath108 , these conditions coincide .
the isomorphism is a consequence of the fact that , under conditions ( [ eq15 ] ) , one can choose @xmath62(@xmath109 ) in such a way that the functions under the square root in eqs.([eq13 ] ) , ( [ eq14 ] ) do not have zeros in the spectrum of operator @xmath30 , and therefore we can consider the square root as a single - valued analytic function .
this means that transformations are invertible , analytic , and therefore define an isomorphism of irreps .
the constructed transformations give us a tool to realize @xmath33 in terms of a simpler pae with the same type of irrep .
this procedure will be applied below to the tavis - cummings hamiltonian .
the interaction part of the hamiltonian eq .
( [ eq1 ] ) can be expressed in terms of third - order pae . the generators @xmath110 of this algebra are realized as @xmath111 it is plain that these generators satisfy the commutation relations eqs .
( [ eq2 ] ) .
the generators of the algebra @xmath112 commute with the operators @xmath113 hereafter we use the same notation @xmath114 both for the casimir operator and its eigenvalue , if no confusion arises .
we show below that the eigenvalues @xmath115 of the operators of eq.([eq17 ] ) parameterize the pae in question .
we thus denote this pae as @xmath116 .
the structure polynomial of @xmath116 can be expressed in the form @xmath117 the parameters of this structure polynomial are @xmath118 and its behavior as a function of @xmath119 is given in fig .
[ fig : polyn3 ] .
the structure polynomial @xmath120 of the @xmath116 algebra for @xmath121.,width=302 ] we turn next to the description of finite dimensional irrep of @xmath116 . in physical applications the parameter @xmath26 has the meaning of collective dicke index .
this index runs from @xmath122 to @xmath123 with unit steps , while @xmath114 can be any natural number including zero .
thus @xmath124 is the biggest positive root of first order . if @xmath125 , then @xmath126 ; if @xmath127 then @xmath128 ; the case @xmath129 corresponds to a root of second order . a typical plot of the structure polynomial is shown in fig .
[ fig : polyn3 ] for these three cases . a finite dimensional representation of @xmath116 ( where @xmath130 ) corresponds to the positive spectrum of @xmath131 .
the spectrum is limited from the right by @xmath124 and from the left by @xmath132 or @xmath133 .
notice that the number @xmath114 is conserved .
therefore this number is determined by the initial state .
the different values of @xmath114 and @xmath26 define different algebras @xmath116 , whose single physical finite dimensional representation we will call a _ zone_. the case @xmath125 corresponds to _ nearby _ zones .
the two largest roots are @xmath134 and @xmath135 and the irrep has the type @xmath136 .
consequently , the well - known weak - field limit corresponds to nearby zones .
the case @xmath137 corresponds to _ remote _ zones .
the two largest roots are @xmath138 and @xmath135 , and the corresponding irrep is of the type @xmath100 .
notice that the region @xmath139 is usually called the strong - field limit . in the special case @xmath140 , referred to _ intermediate _
zone , the algebra @xmath116 possesses an irrep of the type @xmath141 .
it is the only irreducible representation that principally differs from all the others . as indicated above , the simplest pae with irrep of the type @xmath142
is @xmath143 ( we use here @xmath144 to distinguish it from the ( physical ) collective index @xmath26 )
. it would be convenient to solve the eigenvalue problem in terms of the simplest algebra @xmath143 .
notice that in any finite dimensional irrep of pae characterized by @xmath142 , the structure polynomial of the algebra can be approximated by a parabolic curve .
this is shown in fig .
[ fig : polyn23 ] ( for @xmath145 ) for the structure polynomial of @xmath116 .
the larger(smaller ) @xmath114 is in comparison with @xmath93 , the better is the approximation .
however , for @xmath146 the approximation is not satisfactory ( see fig . [
fig : polyn23]b ) . in the regions where the approximation is adequate , it is then not difficult to diagonalize the operator @xmath147 , defined in terms of generators of the conventional _
su(2 ) _ algebra . the latter has a parabolic structure polynomial . in fig .
[ fig : polyn23 ] the two roots of @xmath148 are chosen to be equal to the corresponding two roots of @xmath120 . the choice of @xmath41 in @xmath148 will be explained below .
thus in the cases of nearby or remote zones , it is convenient to study the problem in terms of algebra @xmath143 . the approximation illustrated in fig .
[ fig : polyn23 ] indicates that the tc problem can be solved via an appropriate perturbation theory .
approximation of the structure polynomial @xmath120 by the parabolic @xmath148 for @xmath121 .
a ) @xmath149 ; b ) @xmath150 ; c ) @xmath151.,width=321 ] to begin with we consider the transformation of @xmath116 to @xmath143 for the case of remote zones .
the dimension of a remote zone is @xmath152 , and the algebra @xmath143 should be characterized by @xmath153 .
the finite dimensional irrep of @xmath154 is isomorphic to the corresponding irreducible representation of the atomic subsystem . for the pivotal root @xmath62
we choose the largest root that bounds the irrep of @xmath116 from the right ( the @xmath124 in ( [ eq20 ] ) ) , while as the root @xmath104 we take the @xmath132 of @xmath54 from eq .
( [ eq23 ] ) . applying
the mapping ( [ eq14 ] ) we obtain @xmath155 the spectrum @xmath156 of the operator @xmath157 belongs to the region @xmath158 , consequently the argument of the square root function in eq.([eq28 ] ) is positive in the remote zones @xmath159 .
the relations eqs.([eq28 ] ) express the generators of algebra @xmath116 as analytic function of the generators of the @xmath54 algebra .
they thus allow us to approximate the more complex algebra @xmath116 of third order by a simpler algebra of second order . before we begin to study this approximation
we construct a realization of @xmath54 in terms of boson and spin variables .
> from eqs.([eq16]),([eq17 ] ) , and ( [ eq28 ] ) it follows that in remote zones @xmath160 notice that the subspaces that correspond to remote zones do not contain the vacuum state of the field .
it is also worth mentioning that the matrix representation of operator @xmath161 is @xmath162 in any remote zone .
this operator has been considered before as a earlier in terms of phase operator @xcite .
we turn now to the nearby zones @xmath125 . for this region
the mapping of the algebra @xmath116 to the algebra @xmath143 is realized through procedure similar to that described above for remote zones .
notice that the dimension of nearby zones is @xmath163 , and therefore @xmath164 ( cf .
[ fig : polyn23]a ) .
for the nearby zones there is no simple correspondence to a physical representation like that of eq.([eq29a ] ) .
the particular choice of the pivotal root @xmath104 is unimportant and we use the same choice as before . applying eq.([eq14 ] )
we obtain @xmath165 since all the eigenvalues of the operator @xmath166 belong to the interval @xmath167 to @xmath144 , the argument of the square root function does not have zero eigenvalues in the nearby zones . the realization of @xmath143 through spin and boson variables
is then given by @xmath168 notice that the nearby zones do not contain the eigenvector @xmath169 of @xmath170 .
to clarify the structure of intermediate zone we choose as the pivotal root as the third root of @xmath120 of eq.([eq20 ] ) , and apply the transformation ( [ eq12 ] ) .
we thus obtain @xmath171 here @xmath172 are generators of the algebra @xmath50 .
in the region @xmath173 the multiplier @xmath174 in eq.([eq_r3 ] ) vanishes first .
this corresponds to remote zones . in the region
@xmath175 the multiplier that vanishes first is @xmath176 . in the intermediate zone @xmath177
we have @xmath178 in the intermediate zone we thus obtain a special realization of the tc hamiltonian , @xmath179 .
\label{eq_bmodel}\end{aligned}\ ] ] the domain of the quantum space for this hamiltonian is specified by the condition @xmath180 .
let us introduce an @xmath181-atom generalization for arbitrary values @xmath26 and @xmath114 of the well - known quantum rabi frequency such that @xmath182 for @xmath23 our definition agrees with that used in@xcite .
introducing a small parameter @xmath183 we can rewrite the realizations of @xmath184 in nearby and remote zones ( see eqs.([eq28]),([eq34 ] ) ) in the form @xmath185 the diagonalization problem for the operator eq.([eq1 ] ) can now be solved in each zone by means of perturbation theory with respect to the small parameter @xmath186 .
one can show that the eigenvalues of the argument of the square root function in eq.([eq29_new ] ) are less then unity .
hence we can expand the square with respect to @xmath186 , and find thereby for the interaction part of the hamiltonian @xmath187 .\quad \label{eq31}\end{aligned}\ ] ] in the interaction representation the hamiltonian coincides with @xmath188 , we only need to diagonalize the latter .
up to third order in @xmath186 we find that @xmath189 where the @xmath190 are terms of @xmath191-th order in @xmath186 , and are given in the appendix [ app : simtran ] . in the appendix we also show that unitary transformations @xmath192 , which bring the interaction operator into diagonal form : @xmath193 } \right\ } , \quad \tilde u \equiv \tilde u_3 \ , \tilde u_2 \ , \tilde u_1 \ , \tilde u_0 . \label{eq36}\end{aligned}\ ] ] the spectrum of the operator @xmath188 as given by eq.([eq36 ] ) agrees with the results of @xcite for the symmetric states of the atoms .
deviation of the eigenvalues of @xmath194 from their numerical values in zero ( solid line ) and second ( dashed line ) order in @xmath186 , for @xmath195.,width=302 ] energy levels of @xmath194 in the ascending order for @xmath196 , as calculated numerically ( circles ) , from eq .
( [ eq36 ] ) ( crosses ) , and as is given in @xcite ( squares ) .
a ) @xmath197 ; b ) @xmath198 .
, width=302 ] we compared the third order solution eq.([eq36 ] ) with the exact numerical diagonalization of @xmath188 and found that the result ( [ eq36 ] ) is very accurate , especially for increasing values of @xmath199 .
the results of this comparison are shown in fig .
[ fig : spec0 ] . in their original article
@xcite tavis and cummings also found an approximative analytical expression for the spectrum of the interaction operator @xmath188 .
we compare the results of ref .
@xcite with our analytical and numerical solutions in fig .
[ fig : spec2 ] .
it is evident that the tavis and cummings solution is only accurate for very large and very small values of index @xmath200 .
the fig .
[ fig : spec4 ] compares the energies calculated numerically and in accordance with the analytical solution eq .
( [ eq36 ] ) . in the intermediate region of @xmath114
the curves for nearby and remote zones overlap and coincide thus providing still satisfactory correspondence to the exact solution
. however evidently the expansion for the remote zone breaks down in the nearby zone and vice versa .
this means that the classification of zones introduced in this paper is indeed adequate . the dependence on @xmath114 of the maximal eigenvalue of @xmath188 for @xmath201 , as calculated numerically and from eq.([eq36]).,width=302 ]
finally we consider the @xmath1-atom _ dressed states_. in other words we introduce a representation in which the zero order hamiltonian is diagonal .
this representation is given by transformation @xmath202 .
for @xmath203 all the higher order terms in eq.([eq31 ] ) ( higher than zero order ) vanish , while the eigenstates of the zero order hamiltonian coincide with the dressed states@xcite of the jaynes - cummings model .
we can thus call the eigenstates of the zero order hamiltonian the _ dressed states _ of the @xmath1-atom model .
notice also that in the remote zones the rabi frequency eq.([eq_rabi ] ) does not depend on the dicke index @xmath26 .
if we consider only the zero order terms of @xmath18 , it is convenient to combine all the remote zones that have the same rabi frequency into a remote _ superzone _ , whose dimension is @xmath204 . introducing an operator for the total number of quanta in the atom - field system ( cf . the definition of @xmath114 in eq.([eq17 ] ) ) , @xmath205 we can define the remote superzone as follows . a remote superzone contains all eigenvectors of @xmath206 that have the same eigenvalue , provided it is larger than @xmath1 . in fig .
[ fig : sket ] we show the definition of the zones in the tavis - cummings model for @xmath207 .
zones structure for @xmath207 .
@xmath208 is the spectrum of the zero order hamiltonian .
@xmath206 is an eigenvalue of the corresponding operator ( [ k ] ) , and @xmath209 is a number of equivalent irreps of _
su(2 ) _ defined below.,width=302 ] it can be readily seen that , in the hilbert subspace corresponding to the remote superzone , the unitary transformation @xmath202 can be factorized into a product of @xmath1 single - particle transformations that are the dressing operators for the jaynes - cummings problem ( cf .
eq.([eq29a ] ) ) , @xmath210}.\ ] ] > from the discussion following eq.([eq29a ] ) , it follows that in each remote superzone the field operators @xmath211 and @xmath212 commute .
this means that the system behaves almost like a semiclassical one .
in the previous section we developed an algebraic approach to the tavis - cummings model .
we introduced the operators @xmath213 describing collective excitations in the atom - field system . in terms of these operators we constructed a perturbation series for the tavis - cummings hamiltonian eq.([eq1 ] ) .
the derived perturbation series gives us a tool to distinguish and classify cooperative ( multiparticle ) effects of different orders that are involved in calculations of different physical observables characterizing the atom - field system . in the preceding section we constructed a hamiltonian eq.([eq31 ] ) corresponding to the zero order approximation for the hamiltonian eq.([eq1 ] ) .
this simplified operator depends on multiparticle rabi frequency eq.([eq_rabi ] ) , which depends on the number of atoms in the cavity .
therefore the simplified hamiltonian eq.([eq31 ] ) allows to account for cooperative effects in the system . in this section
we study a contribution of cooperative effects into the rate of spontaneous emission generated by the atom - field system .
the atom - field system is assumed to be prepared in the state of thermal equilibrium .
this state is described by the canonical gibbs ensemble with the thermostat temperature @xmath214 .
this means that under the `` system '' we imply @xmath31 atoms strongly coupled to the resonator mode . under the `` thermostat ''
we imply the surrounding environment , for instance , cavity walls taken at the temperature @xmath214 . in the state of thermal equilibrium the exact atom - field density matrix should be defined using the hamiltonian eq.([eq1 ] )
. however , for the exact density matrix the analytical analysis of the transition probability , if possible at all , would be highly technically involved .
we demonstrate that nontrivial physical results for the intensity of spontaneous emission of @xmath1 two - level atoms placed inside the cavity can be already obtained for the zero order approximation of the exact hamiltonian eq.([eq31 ] ) .
the thermal state is given by @xmath215.\ ] ] here @xmath216 is a normalization factor .
we show here that the intensity of spontaneous emission of the system comprised of @xmath31 atoms _ strongly _ coupled to the resonator mode and prepared in the thermal state , can be greatly enhanced at a certain temperature .
this amplification results from high correlations in the atomic subsystem .
similar effect exists for dicke s superradiant state @xcite .
this state is prepared by a short laser pulse .
therefore the setting of the dicke s theory is quite different from our considerations of the stationary state of thermal equilibrium .
when calculating the rate of spontaneous emission ( or the intensity proportional to this quantity ) we merely follow the ideas of dicke s paper ( see e.g. @xcite ) . according to this theory ,
the rate of spontaneous emission in the system is proportional to the average of the square of the atomic dipole , viz .
@xmath217 it is convenient to calculate the average in the dressed states basis , where @xmath218.\ ] ] taking into account eq.([cas ] ) along with the fact that , for an arbitrary zone , @xmath219 it can be shown that in the dressed states basis @xmath220 finally , the intensity of spontaneous emission is given by @xmath221.\quad\end{aligned}\ ] ] here @xmath222 , and @xmath223 is the number of equivalent representations with the same @xmath26 .
let us consider the intensity per atom , i.e. @xmath224 .
this intensity consists of two terms , i.e. the first is given by a single - particle contribution @xmath225 and the second one proportional to the two - particle correlation function @xmath226 , @xmath227 .
they are found to be @xmath228 it is plain that in the absence of the cavity , the correlation function vanishes and the only contribution to @xmath229 is given by the first term , @xmath230 the contribution of any remote superzone can easily be found due to the factorization property of the operator eq.([factor ] ) .
if we denote the trace over states that belong to the same remote superzone by subscript @xmath206 , we obtain @xmath231 in fig .
[ fig : int ] we compare the spontaneous emission @xmath229 in the presence of a cavity against the intensity @xmath232 of @xmath2 atoms in the absence of cavity .
the intensity of spontaneous emission per atom ( in units of @xmath233 ) versus cavity temperature .
the curve ( 1 ) is the classical result given by eq.([isl ] ) .
the curves ( 2,3,4 ) correspond to @xmath234 , @xmath235 , @xmath236 , respectively , and @xmath237.,width=302 ] notice that if the number of atoms is big enough , the intensity of radiation exhibits a high maximum . in a cavity at low @xmath214 ,
the cluster of @xmath1 two - level atoms emits much more intensively than it does in the free space .
it should be possible to drive the system to thermal equilibrium at the temperature where the spontaneous emission exhibits maximum .
the marked amplification of spontaneous emission should be observed in cavity experiments . concluding this section , we recapitulate our main results .
we consider spontaneous emission of the system comprised of @xmath2 two - level atoms strongly coupled to the cavity mode and prepared in the state of thermal equilibrium . in the absence of the cavity the atoms in the thermal equilibrium would be uncorrelated . in this case
the spontaneous emission would be described by the conventional formula eq.([isl ] ) . for high - q resonators the strong coupling to the resonator mode
should necessarily be taken into account .
we demonstrate that in this case the intensity of spontaneous emission can be be greatly enhanced .
this phenomenon can be explained by additional correlation between atoms established by the cavity mode . to analyze the effect
analytically we have replaced the exact tavis - cummings hamiltonian eq.([eq1 ] ) by its zero order approximation derived in the previous sections .
this allowed us to represent the intensity of spontaneous emission in the simple analytical form eq.([i_spon2 ] ) .
it is appropriate to emphasize once again that the zero order approximation of the hamiltonian contains strong coupling and , thus , describes cooperative effects in the atomic subsystem .
this is the consequence of the fact that the operators @xmath213 describe collective excitations in the atom - field system .
in this work we solved the tavis - cummings problem by applying the technique of polynomially deformed algebras .
we constructed the transformations that map one polynomial algebra of operators onto another .
this allowed us to reformulate the problem in terms of a simpler algebra of second order , @xmath54 , and develop a specific perturbation theory .
our results have a significant advantage over the so - called linearization approximation , i.e. the case when the hamiltonian is linearized in terms of the algebra @xmath50 . in this latter approach a structure polynomial of a higher order
is approximated by a polynomial of first order .
this method allows only to calculate the lowest or highest eigenvalues , and the corresponding eigenstates .
it does not take into account the finite dimensionality of the representation .
the parabolic approximation developed in this work provides in this respect a significant advantage because it allows to construct a finite dimensional representation for the problem .
we were able to find analytical expressions for all the eigenvalues of the hamiltonian up to third order in the small parameter @xmath186 .
for the nearby zones we showed explicitly how the collective quantum rabi frequency depends on dicke index @xmath26 . since this index characterizes the symmetry of atomic states , the result has significant physical implications .
the dependence on atomic symmetry is revealed already in zeroth order in the perturbation expansion .
employing our methods we found an interesting new effect , amplification of spontaneous emission of thermal @xmath1-atom states due to collective effects .
we expect that this phenomenon can be observed in cavity experiments .
it is worth mentioning that the applicability of the method developed in this paper can be extended to many other problems , including bose - einstein condensation , multi - photon interactions in the micromaser , and multimode interaction of an electromagnetic field with matter .
our methods allow in particular to address the problem of collective and dressed states in the mentioned physical systems .
this work has been supported by the academy of finland ( project no .
44875 ) .
i.v . is grateful to the centre for international mobility ( cimo ) and university of jyvaskyla for financial support .
we look for similarity transformations @xmath238 that diagonalize the hamiltonian in different orders of @xmath186 , where @xmath240 . in eq .
( [ app0 ] ) ) only the terms of order @xmath191 in the small parameter @xmath186 are present . for @xmath241 ,
the term @xmath242 should be replaced by the corresponding term in eq.([eq31 ] ) .
regrouping the terms we obtain @xmath243 @xmath252,\nonumber\\ { \mathop{\rm v}\nolimits } _ 0^{\left ( 3 \right ) } = \frac{1}{2 } \left ( { \frac{\alpha } { 2 } } \right)^3 \left [ { \tilde s_x b \tilde s_x } \right ] .
\label{app2 } \end{aligned}\ ] ] it can readily be seen that the transformation @xmath254,\quad
d1 \equiv -i\left [ { \tilde s_3 \tilde s_y + \tilde s_y\tilde s_3 } \right ] \label{app6}\ ] ] diagonalizes the operators in the first order . in the diagonalization
one needs the commutators @xmath255 = b , \nonumber\\ \left [ { \left [ { \tilde s_3 , d1 } \right],d1 } \right ] = 4\tilde s_3 \left ( { \tilde s^2 - 2\tilde s_3 ^2 - \frac{1}{4 } } \right),\nonumber\\ \left [ { \left [ { \left [ { \tilde s_3 , d1 } \right],d1 } \right],d1 } \right ] = \left ( { 4\tilde s^2 - 1 } \right ) b - 8 \left [ { \tilde s_3 b + b\tilde s_3 - \tilde s_3 \tilde s_x \tilde s_3 } \right],\nonumber\\ \left [ { \tilde s_3 \tilde s_x \tilde s_3 , d1 } \right ] = 2\tilde s_3 ^2 \left ( { \tilde s^2 - 2\tilde s_3 ^2 } \right ) + \left ( { \tilde s_3 \tilde s_x } \right)^2 + \left ( { \tilde s_x\tilde s_3 } \right)^2 . \label{app6_2}\end{aligned}\ ] ] where @xmath267 . the diagonalizing transformation is then given by @xmath268 } \right\ } , \label{app13}\ ] ] where @xmath269 . keeping the terms up to third order we obtain ( see ( [ app9 ] ) ) @xmath270 .
\label{app14}\end{aligned}\ ] ] diagonalization of @xmath274 can be performed in a similar way with an operator @xmath275 $ ] .
as there are no diagonal terms in @xmath274 , which would contribute to the spectrum of the hamiltonian , we do not give here the fairly complicated form of operator @xmath276 .
the final diagonal form for the interaction @xmath188 is thus given by to recapitulate , we introduced four transformations @xmath278 , which successively diagonalize the interaction operator in the tavis - cummings hamiltonian up to third order with respect to the small parameter @xmath279 , with @xmath280 the generalized rabi frequency of eq.([eq_rabi ] ) , such that | an algebraic method is introduced for an analytical solution of the eigenvalue problem of the tavis - cummings ( tc ) hamiltonian , based on polynomially deformed _
su(2 ) _ , i.e. @xmath0 , algebras . in this method
the eigenvalue problem is solved in terms of a specific perturbation theory , developed here up to third order .
generalization to the @xmath1-atom case of the rabi frequency and dressed states is also provided .
a remarkable enhancement of spontaneous emission of @xmath1 atoms in a resonator is found to result from collective effects . |
we now have uncontroversial evidence for neutrino flavor conversion coming from `` celestial '' ( solar and atmospheric ) as well as `` laboratory '' studies with reactor and accelerator neutrinos @xcite .
oscillations constitute the only viable explanation of the data and provide the first sign of physics beyond the standard model ( sm ) .
the basic concept in terms of which to describe them is the lepton mixing matrix , the leptonic analogue of the quark mixing matrix . in its simplest @xmath0 unitary form
it is given as @xcite @xmath1 where each @xmath2 is characterized by an angle and a corresponding cp phase .
present experiments are insensitive to cp violation , hence we set all three phases to zero . in this
approximation oscillations depend on the three mixing angles @xmath3 and on the two squared - mass splittings @xmath4 and @xmath5 characterizing solar and atmospheric transitions . to a good approximation
, one can set @xmath6 in the analysis of atmospheric and accelerator data , and @xmath7 to infinity in the analysis of solar and reactor data .
the neutrino oscillation parameters obtained from a global analysis of the world s neutrino oscillation data are summarized in figs .
[ fig : global ] and [ fig : th13 ] . the left and right
panels in fig .
[ fig : global ] give the `` atmospheric '' and `` solar '' oscillation parameters , @xmath8 & @xmath7 , and @xmath9 & @xmath10 , respectively .
the dot , star and diamond indicate the best fit points of atmospheric minos and global data , respectively .
we minimize with respect to @xmath10 , @xmath9 and @xmath11 , including always all the relevant data .
similarly the `` solar '' oscillation parameters are obtained by combining solar and reactor neutrino data .
the dot , star and diamond indicate the best fit points of solar , kamland and global data , respectively .
we minimize with respect to @xmath7 , @xmath8 and @xmath11 , including always all relevant data .
current neutrino oscillation parameters , from ref .
, title="fig:",height=188 ] current neutrino oscillation parameters , from ref .
, title="fig:",height=188 ] one sees that data from artificial and natural neutrino sources are clearly complementary : reactor and accelerators give the best determination of squared - mass - splittings , while solar and atmospheric data mainly determine mixings .
constraints on @xmath12 from different parts of the global data given in ref .
@xcite.,title="fig:",height=188 ] constraints on @xmath12 from different parts of the global data given in ref .
@xcite.,title="fig:",height=188 ] fig .
[ fig : th13 ] summarizes the information on the remaining angle @xmath11 , the right panel shows how current data slightly prefer a nonzero value for @xmath11 .
since this is currently not significant , we prefer to interpret this as a weaker bound on @xmath11 ) are given for 1 dof , while the regions in fig .
[ fig : th13 ] ( left ) are 90% cl for 2 dof ] : @xmath13 a possible experimental confirmation of a non - zero @xmath11 would encourage the search for cp violation in upcoming neutrino oscillation experiments @xcite . note that all cp violating observables , such as cp asymmetries , are proportional to the small parameter @xmath14 well - determined experimentally as @xmath15
in spite of the great experimental progress summarized above , pinning - down the ultimate origin of neutrino mass remains a challenge for the next decades . to understand the pathways to neutrino masses it is important to note that , being electrically neutral , neutrino masses should , on general grounds , be of majorana - type @xcite . indeed , specific neutrino mass generation mechanisms also differ from those of charged fermions in the sm .
the latter come in two chiral species and get mass linearly in the electroweak symmetry breaking vacuum expectation value ( vev ) @xmath16 of the higgs scalar doublet .
in contrast , as shown in the left panel in fig .
[ fig : d-5-nsi ] , neutrinos acquire mass from an effective lepton number violating dimension - five operator @xmath17 ( where @xmath18 denotes a lepton doublet ) @xcite .
operators characterizing neutrino masses and non - standard neutrino interactions ( nsi ) arising , say , from the non - trivial structure of lepton mixing in seesaw - type schemes @xcite.,title="fig:",height=113 ] operators characterizing neutrino masses and non - standard neutrino interactions ( nsi ) arising , say , from the non - trivial structure of lepton mixing in seesaw - type schemes @xcite.,title="fig:",height=113 ] a natural way to account for the smallness of neutrino masses , irrespective of their specific origin , is that l - number is restored , in the absence of l - violating operator(s ) . such may be naturally suppressed either by a high - scale @xmath19 in the denominator or , alternatively
, it may involve a low - mass - scale in the numerator .
the big question is to identify which _ mechanism _ gives rise to this operator , its associated mass _ scale _ and its _ flavor structure_. gravity is often argued to break global symmetries @xcite , and could induce the dimension - five operator , with @xmath19 identified to the planck scale .
the resulting majorana neutrino masses are too small , hence the need for physics beyond the sm @xcite .
the coefficient @xmath20 could vanish due to symmetry , so that the effective operator responsible for neutrino mass is of higher dimension @xcite .
alternatively it may be suppressed by small scales , yukawa couplings and/or loop - factors @xcite . to arrange our brief discussion
i consider three options : ( i ) tree level , ( ii ) radiative , and ( iii ) hybrid mechanisms , all of which may have high- or low - scale realizations .
if lepton - number symmetry is broken spontaneously there is either an extra neutral gauge boson or a nambu - goldstone boson coupled to neutrinos , depending on whether it is gauged or not .
it is easy to construct models based on either high- or low - scale symmetry breaking , the former are more popular among theorists , because they are closer to the idea of unification .
however the most basic and general description of the seesaw is in terms of the sm 21 gauge structure @xcite .
in such a framework it is clear that the relevant scale can be large or small , depending on model details .
hence there is a fair chance that the origin of neutrino mass may be probed at accelerators like the lhc .
-0.5 cm the classic way to to generate weinberg s dimension-5 operator @xcite is the exchange of heavy fermion states with masses close to the `` unification '' scale .
.3 cm type - i and iii ( left ) and type - ii ( right ) realizations of the seesaw mechanism.,title="fig : " ] type - i and iii ( left ) and type - ii ( right ) realizations of the seesaw mechanism.,title="fig:",height=94 ] depending on whether these are 21 singlets or triplets the mechanism is called * type - i * @xcite @xcite , or * type - iii * seesaw @xcite , respectively . as seen in the right panel in fig .
[ fig : seesaw ] , the seesaw may also be induced by the exchange of heavy triplet scalars , now called * type - ii * seesaw @xcite @xcite , a convention opposite to the one used originally in @xcite .
the `` complete seesaw '' was thoroughly studied in @xcite and its perturbative diagonalization was given ref .
@xcite in a general form that may be adapted to different models .
the hierarchy of vevs required to account the small neutrino masses @xmath21 in such seesaw was studied in detail in ref .
@xcite .
the seesaw may be implemented in the type i , type ii or type iii manner , with different gauge groups and multiplet contents , with gauged or ungauged b
- l , broken explicitly or spontaneously , at a high or at a low energy scale , with or without supersymmetry .
there are so many ways to seesaw , that a full taxonomy describing all variants will probably never be written , as nature may be more imaginative than physicists . since any extended symmetry model must ultimately break to the sm , what is phenomenologically relevant is the seesaw description at the 21 level @xcite
such low - energy description is specially relevant in accurately describing low - scale variants of the seesaw mechanism , whose interest has now been revived with the coming of the lhc .
an attractive class of such schemes employs , in addition to the left - handed sm neutrinos @xmath22 , two 21 singlets @xmath23 @xcite ( for other extended seesaw schemes see , e.g. @xcite ) .
the basic lepton - number - violating parameter is small @xcite and may be calculable due to supersymmetric renormalization group evolution effects @xcite .
one may implement such schemes in the 0 framework @xcite , leaving , in addition a light @xmath24 to be probed at the lhc @xcite .
decay length versus @xmath25 for @xmath26 gev , @xmath27 , @xmath28 , and several values of @xmath29 .
the widths of the three shaded bands around @xmath30 gev correspond to the variation of the brpv parameters in such a way that the neutrino masses and mixing angles fit the required values within @xmath31 .
right : ratio of branching ratios , br@xmath32 over br@xmath33 as a function of the atmospheric angle in bilinear r parity violation @xcite.,title="fig:",width=226,height=170 ] decay length versus @xmath25 for @xmath26 gev , @xmath27 , @xmath28 , and several values of @xmath29 .
the widths of the three shaded bands around @xmath30 gev correspond to the variation of the brpv parameters in such a way that the neutrino masses and mixing angles fit the required values within @xmath31 .
right : ratio of branching ratios , br@xmath32 over br@xmath33 as a function of the atmospheric angle in bilinear r parity violation @xcite.,title="fig:",width=226,height=181 ] -0.5 cm neutrino masses may be absent at tree level and calculable @xcite , with no need for a large scale . in this case
the coefficient @xmath20 is suppressed by small loop - factors , by yukawa couplings and possibly by a small scale parameter characterizing the breaking of lepton number , leading to naturally small neutrino masses . like low - scale seesaw schemes
( see above ) radiative models open the door to phenomenology associated with the new states required to provide the neutrino mass and which could be searched for , e. g. , at the lhc .
-0.5 cm it could well be that the origin of neutrino masses is intrinsically supersymmetric .
this is the case in models with r parity violation @xcite , in which lepton number is broken together with the so - called r parity .
this may happen spontaneously , driven by a nonzero vev of an 21 singlet sneutrino @xcite , leading to an effective model with bilinear violation of r parity @xcite .
this provides the minimal way to break r parity and add neutrino masses to the mssm @xcite .
the neutrino spectrum is hybrid , with one scale ( typically the atmospheric ) generated at tree level by neutralino - exchange _ weak - scale seesaw _ , and the other scale ( solar ) induced by _
calculable _ one - loop corrections @xcite . unprotected by any symmetry
, the lightest supersymmetric particle ( lsp ) will decay . given
the scale of neutrino mass indicated by experiment these decays will happen inside typical detectors at the tevatron or the lhc @xcite but with a decay path that can be experimentally resolved , leading to a so - called displaced vertex @xcite ( left panel in fig .
[ fig : ntrl ] ) .
more strikingly , its decay properties correlate with the neutrino mixing angles . indeed , as seen in the right panel in fig .
[ fig : ntrl ] the lsp decay pattern is predicted by the low - energy measurement of the atmospheric angle @xcite .
such a prediction will be tested at the lhc , and will potentially allow a high - energy redetermination of @xmath8 .
similar correlations hold in variant models which have other supersymmetric states as lsp @xcite
as seen above current neutrino oscillation data indicate solar and atmospheric mixing angles which are unexpectedly large when compared with quark mixing angles .
this challenges our attempts to explain the flavor problem in unified schemes where quarks and leptons are related .
it has been noted that the neutrino mixing angles are approximately given by @xcite , @xmath34 there have been many schemes suggested in the literature in order to reproduce the full tri - bi - maximal pattern , or at least to predict maximal atmospheric mixing using various discrete flavor symmetry groups containing mu - tau symmetry , e. g. @xcite .
one expects the flavor symmetry to be valid at high energy scales .
deviations from tri - bi - maximal ansatz @xcite may be calculable by renormalization group evolution @xcite .
a specially simple ansatz is that , as a result of a given flavor symmetry such as a4 @xcite , neutrino masses unify at high energies @xmath19 @xcite , the same way as gauge couplings unify at high energies due to supersymmetry @xcite .
such quasi - degenerate neutrino scheme predicts maximal atmospheric angle and vanishing @xmath11 , @xmath35 leaving the solar angle @xmath9 unpredicted , but cabibbo - unsuppressed , @xmath36 if cp is violated @xmath11 becomes arbitrary and the dirac phase is maximal @xcite .
one can show that lepton and slepton mixings are related and that at least one slepton lies below 200 gev , within reach of the lhc .
the absolute majorana neutrino mass scale @xmath37 ev ensures that the model will be probed by future cosmological tests and @xmath38 searches .
rates for lepton flavour violating processes @xmath39 typically lie in the range of sensitivity of coming experiments , with br@xmath40 and br@xmath41 .
flavor is violated in neutrino propagation @xcite .
it is therefore natural to expect that , at some level , it will also show up as transitions involving the charged leptons , since these sit in the same electroweak doublet .
two basic mechanisms are : ( i ) neutral heavy lepton exchange @xcite and ( ii ) supersymmetry @xcite .
both exist in supersymmetric seesaw - type schemes of neutrino mass , the interplay of both types of contributions depends on the seesaw scale and has been analysed in @xcite . barring fine - tunings , high - scale seesaw models require supersymmetry in order to have sizeable lfv rates .
moreover , supersymmetry brings in the possibility of direct in the production of supersymmetric particles .
this will provide the most direct way to probe lfv at the lhc in high - scale seesaw models , as seen in fig .
[ fig : ntrl ] , from ref .
@xcite . [ cols="^,^ " , ] in contrast
, the sizeable admixture of right - handed neutrinos in the charged current ( rectangular nature of the lepton mixing matrix @xcite ) in low - scale seesaw schemes induces potentially large lfv rates even in the absence of supersymmetry @xcite .
indeed , an important point to stress is that lfv @xcite and cp violation @xcite can occur in the massless neutrino limit , hence their attainable magnitude is unrestricted by the smallness of neutrino masses . in fig .
[ fig : lfv - low - scale ] we display @xmath42 versus the small lepton number violating ( lnv ) parameters @xmath43 and @xmath44 for two different low - scale seesaw models , the inverse and the linear seesaw , respectively .
clearly the lfv rates are sizeable in both cases , the different slopes with respect to @xmath43 and @xmath44 follow from the fact that lnv occurs differently in the two models , @xmath45 versus the @xmath46 , respectively . versus the lepton number violation scale for two low - scale seesaw models : the inverse seesaw in top ( red color ) , the linear seesaw in bottom ( blue color ) .
in both cases , @xmath47 is fixed as @xmath48 ( continous line ) , @xmath49 ( dashed line ) and @xmath50 ( dot - dashed line ) .
the right panel shows typical correlation between mu - e convertion in nuclei and @xmath42.,title="fig:",width=264,height=188 ] versus the lepton number violation scale for two low - scale seesaw models : the inverse seesaw in top ( red color ) , the linear seesaw in bottom ( blue color ) . in both cases ,
@xmath47 is fixed as @xmath48 ( continous line ) , @xmath49 ( dashed line ) and @xmath50 ( dot - dashed line ) .
the right panel shows typical correlation between mu - e convertion in nuclei and @xmath42.,title="fig:",width=264,height=188 ] similarly one can show @xcite that in low - scale seesaw models the nuclear @xmath51 conversion rates lie within planned sensitivities of future experiments such as prism @xcite .
note that models with specific flavor symmetries , such as those in @xcite relate different lfv rates . to conclude we mention that the some seesaw schemes , like type - iii @xcite or inverse type - iii @xcite , may be directly probed at the lhc by directly producing the tev rh neutrinos at accelerators .
neutrino oscillations can not distinguish dirac from majorana neutrinos .
in contrast , lnv processes , such as @xcite hold the key to the issue .
indeed , in a gauge theory , _ irrespective of the mechanism that induces _ , it implies a majorana mass for at least one neutrino @xcite , as illustrated in fig .
[ fig : bbox ] .
+ such `` black - box '' theorem @xcite holds in any `` natural '' gauge theory , though quantitative implications are very model - dependent , for a recent discussion see @xcite .
the detection of neutrinoless double beta decay remains a major challenge @xcite .
the observation of neutrino oscillations suggests that the exchange of light majorana neutrinos will induce through the so - called _ mass - mechanism_. the corresponding amplitude is sensitive both to the majorana cp violation @xcite , and also to the absolute scale of neutrino mass , neither of which can be probed in oscillations .
together with high sensitivity beta decay studies @xcite , and with cosmic microwave background and large scale structure observations @xcite , neutrinoless double beta decay provides complementary information on the absolute scale of neutrino mass . taking into account current neutrino oscillation parameters @xcite and state - of - the - art nuclear matrix elements @xcite
one can determine the average mass parameter @xmath52 characterizing the neutrino exchange contribution to , as in fig .
42 of ref .
quasi - degenerate neutrino models @xcite give the largest possible signal . in normal hierarchy models there
is in general no lower bound on @xmath52 as there can be a destructive interference among the three neutrinos .
in contrast , the inverted neutrino mass hierarchy implies a generic `` lower '' bound for the amplitude .
specific flavor models may , however , imply a lower bound for even with normal hierarchy , as discussed in @xcite @xcite .
the best current limit on @xmath52 comes from the heidelberg - moscow experiment , for the current experimental status and perspectives , see ref .
@xcite , which should be compared with nuclear theory @xcite . | after summarizing neutrino oscillation results i discuss high and low - scale seesaw mechanisms , with or without supersymmetry , as well as recent attempts to understand the pattern of neutrino mixing from flavor symmetries .
i also mention the possibility of intrinsic supersymmetric neutrino masses in the context of broken r parity models , showing how this leads to clear tests at the lhc .
address= ahep group , instituto de fsica corpuscular , c.s.i.c .
universitat de valncia + edificio de institutos de paterna , apartado 22085 , e46071 valncia , spain + |
planets form through the assembly of refractory dust , volatile ice , and gas in disks around young stars .
the spatial distribution of volatile abundances helps determine the compositions of nascent planets that form at different disk radii .
the separation of these volatiles into their gas and ice phases is set by the balance of adsorption ( freeze - out " of gas onto solids ) , and desorption back to the gas phase .
desorption results from both thermal and non - thermal processes that can be active in different regions of protoplanetary disks .
the thermal desorption ( sublimation ) rate for a volatile is set by its energy barrier for ice desorption in concert with the grain temperature .
common volatiles like h@xmath4o , co@xmath4 , and co are characterized by very different desorption barriers . for a typical disk with midplane temperatures that increase closer to the host star
, this range of desorption energy barriers will introduce a series of distinct locations snow lines " where specific volatiles are liberated from ices into the gas phase .
the highly volatile species co and n@xmath4 have the lowest thermal desorption barriers , and so have snow lines at large disk radii where the disk is cold , @xmath330100au @xcite .
irradiation heating maintains a thermal inversion in the disk atmosphere , where temperatures increase with height ( @xmath5 ) above the midplane ( e.g. * ? ? ?
these higher surface temperatures mean that snow lines are really two dimensional snow surfaces whose boundaries extend to larger @xmath6 for larger @xmath5 .
in addition to thermal desorption , there are analogous non - thermal processes driven by interactions between ice and high - energy photons and particles , or the release of chemical energy @xcite .
ice photodesorption by ultraviolet ( uv ) radiation has garnered special interest because of the strong uv fields produced by young stars ( e.g. , * ? ? ?
* ) and the high uv desorption yields measured in laboratory experiments @xcite . in the inner disk , uv photodesorption can result in lower scale heights for snow surfaces compared to the ones expected from thermal desorption @xcite . at the low densities of the outer disk ,
uv radiation may penetrate to the cold midplane to desorb co ice well outside the expected location of the thermal co snowline @xcite .
dco@xmath2 is expected to form efficiently in cold regions where co gas is still present in the gas - phase , whether thermally or non - thermally desorbed , through deuteron transfer from the low - temperature ( t@xmath750 k ) product h@xmath4d@xmath2 to co. therefore , the expectation is that dco@xmath2 emission peaks close to the co snow surface , i.e. at the coldest disk locations where co is still abundant in the gas - phase if only thermal desorption is considered .
several studies have found an increasing dco@xmath2 abundance with radius ascribed to this process @xcite , and @xcite resolved a ring of dco@xmath2 in the disk around the herbig ae star hd 163296 , which seems closely related to the co snow line .
their proposed connection between the dco@xmath2 emission morphology and the co snowline is challenged by new disk chemistry models , where significant dco@xmath2 can be produced at temperatures above 30 k through pathways that do not involve h@xmath4d@xmath2 @xcite . however
, this complication does not eliminate the fact that dco@xmath2 _ requires _ gas - phase co to form , nor does it remove the expectation that dco@xmath2 is formed in excess if h@xmath4d@xmath2 is abundant ( i.e. , if the gas is cold ) . in this study ,
we present alma cycle 2 observations of the i m lup disk , located at a distance of 155 pc @xcite , in 1.4 mm continuum , dco@xmath2 32 , c@xmath0o 21 and h@xmath1co@xmath2 32 line emission ( 2 ) .
3 presents the observed emission morphology of dco@xmath2 and the other species . in
4 we use a disk chemistry model , applied to a generic t tauri disk structure , to interpret the observations .
we highlight the role of non - thermally desorbed co to explain the presence of dco@xmath2 emission in the outer regions of the disk in the discussion in 5 and summarize our findings in 6 .
r c c c c c dco@xmath2 j=32 & 216.1126 & -2.62 & 20.7 & @xmath8 ( -71 ) & @xmath9 + c@xmath0o j=21 & 219.5604 & -6.22 & 15.8 & @xmath10 ( -60)&@xmath11 + h@xmath1co@xmath2 j=32 & 260.2553 & -2.87 & 25.0 & @xmath12 ( 89)&@xmath13 observations of the i m lup disk were acquired with the atacama large millimeter / submillimeter array ( alma ; project code ads / jao.alma#2013.1.00226.s ) on 2014 july 6 with 31 antennas ( targeting dco@xmath2 and c@xmath0o ) and on 2014 july 17 with 32 antennas ( targeting h@xmath1co@xmath2 ) .
the antenna separations spanned baselines of 15650 m .
the total on source integration time was @xmath321 minutes in each execution .
the nearby quasars j1534@xmath143526 and j1427@xmath144206 were used for gain and bandpass calibration , respectively .
titan was observed to calibrate the absolute amplitude scale . on the july 6 execution ,
the correlator was configured to process 13 distinct spectral windows ( spws ) .
the spectral resolution @xmath15 and bandwidth for 12 of the spw were 61 khz and 59 mhz , respectively ; the remaining spw had a coarser resolution with a bandwidth of 469 mhz .
the dco@xmath2 32 line was located in spw1 , and c@xmath0o 21 in spw6 ( table [ lines ] ) . for the july 7 execution
, the correlator was configured to process 14 spws in a different part of the band 6 spectrum .
all had the same spectral resolution , 61 khz ; twelve spws had a bandwidth of 59 mhz , and the remaining two had a bandwidth of 117 mhz . the h@xmath1co@xmath2 32 line ( table [ lines ] )
was located in spw1 .
the visibility data were calibrated by alma staff .
the calibration plots were inspected , deemed adequate , and no changes to the delivered calibration were made . each individual spw was further phase and amplitude self - calibrated in casa 4.2.2 . in each spw
, the continuum was subtracted using line - free channels .
the fully calibrated visibilities were fourier inverted and cleaned using the briggs weighting scheme : the robust parameter was set 0.5 for the continuum and 1.0 for the spectral lines .
a clean mask was manually generated based on the bright c@xmath0o 21 line , and then applied to the h@xmath1co@xmath2 and dco@xmath2 lines .
the cleaned maps were restored with a synthesized beam with fwhm dimensions of @xmath16 .
the 1.4 mm continuum has an rms noise level of @xmath30.24mjy beam@xmath17 ( dynamic range limited ) .
the integrated flux density is 181@xmath1818 mjy ( which includes a 10% flux calibration uncertainty ) , consistent with previous sma measurements @xcite .
the resulting rms noise level in 0.2 km s@xmath17 binned channels is 4 - 6 mjy .
figure [ fig1 ] shows the continuum and integrated line emission maps of c@xmath0o 21 , h@xmath1co@xmath2 32 , and dco@xmath2 32 from the i m lup disk .
the continuum is centrally peaked and spatially resolved , extending out to a radius of @xmath32 ( 300 au ) , comparable to previous lower resolution observations @xcite .
the peak flux density at the disk center is 66@xmath187 mjy beam@xmath17 at 1.4 mm ( c@xmath0o and dco@xmath2 spectral setting ) and 78@xmath188 mjy beam@xmath17 at 1.1 mm ( h@xmath1co@xmath2 spectral setting ) , corresponding to a brightness temperature of @xmath36 k. integrated line emission maps ( fig .
[ fig1 ] , top panel ) were constructed by summing up emission from individually cleaned 0.2 km s@xmath17-wide channels , employing a 2@xmath19 threshold to maximize the snr .
they are shown together with the continuum 2@xmath19 contour .
the middle panel of fig [ fig1 ] presents integrated line emission maps of the same lines split up into two velocity bins around the source velocity to visualize the disk rotation . the integrated emission in these panels
have been scaled to a constant emission maximum to show the increasing importance of extended emission for dco@xmath2 compared to h@xmath1co@xmath2 , and for h@xmath1co@xmath2 compared to c@xmath0o . to visualize the radial structure of these tracers more clearly and to mitigate any bias introduced by the applied clipping threshold
, we also produced moment-0 maps without any clipping , deprojected and azimuthally averaged them into radial brightness profiles ( fig .
[ fig1 ] , bottom panels ) .
we assumed an inclination of 49 and a position angle of -35 for these calculations @xcite .
the integrated emission maps and average radial profiles reveal two significant morphological features .
first , all three molecular species show a clear central depression .
the c@xmath0o and dco@xmath2 emission each peak at a radius of 06 , or @xmath390au , and the h@xmath1co@xmath2 line peaks at 08 ( @xmath3125au ) .
second , dco@xmath2 exhibits a large second ring at 2 , near the outer edge of the continuum emission .
no such ring is observed in c@xmath0o , h@xmath1co@xmath2 or the continuum .
however , the h@xmath1co@xmath2 emission `` plateaus '' around the same radius and c@xmath0o exhibits a subtle slope change .
to interpret the observed c@xmath0o , h@xmath1co@xmath2 , and dco@xmath2 radial structures , we explored the relevant predictions of the disk chemistry model presented by @xcite , updated with deuterium chemistry that considers nuclear spin ( ortho / para ) chemistry of light species .
this model adopts a generic steady , axisymmetric keplerian disk around a t tauri star with stellar mass , radius , and effective temperature of @xmath20 @xmath21 , @xmath22 @xmath23 , and @xmath24 k , respectively , and disk mass @xmath25 m@xmath26 @xcite .
the stellar uv and x - ray luminosities are set to @xmath27 erg s@xmath17 and @xmath28 erg s@xmath17 , respectively , based on the observed spectrum of the young star tw hya @xcite .
the disk model assumes that the grains have grown up to 1 mm throughout the disk , resulting in a factor of 10 decrease in grain surface area compared to ism dust .
the gas temperature , dust temperature and density distributions of the disk are calculated self - consistently , considering various heating and cooling mechanisms @xcite .
compared to i m lup , this model employs a fainter star , which results in a colder disk , and also a lower disk mass .
the differences in temperature and density structure clearly precludes any quantitative comparison between the model and data .
however , even though the model has not been tuned to the parameters of i m lup , qualitative comparisons still provide valuable insight into the origins of the observed molecular emission structure , as long as the model results are robust to the model parameter choices . to evaluate the latter we performed a small parameter study where we varied the co binding energy , the disk density and temperature structure , and the cosmic ionization rate .
as described below the main results are not sensitive to these parameters .
we therefore conclude that the predictions from the generic disk model are useful for interpreting the observed emission pattern of the specific disk around i m lup .
the chemical composition of the model is calculated by integrating a time - dependent system of gas and grain surface rate equations for 300kyr , assuming the physical structure is static @xcite .
the gas - grain reaction network is based on the work by @xcite , supplemented with calculations that account for high - temperature gas phase reactions @xcite , x - ray mediated chemistry ( * ? ? ? * and references therein ) , deuterium chemistry ( * ? ? ? * and references therein ) , and nuclear spin state chemistry of h@xmath4 , h@xmath29 and their isotopologues @xcite .
species containing chlorine , phosphorus , or more than four carbon atoms were excluded for computational expediency .
elemental abundances are taken from @xcite .
the initial abundances are obtained by calculating the molecular evolution of a star - forming core , following @xcite .
importantly for this study , the binding energy of co in the generic disk model is set to 1150 k , its measured value for h@xmath4o ice surfaces @xcite , which results in a co midplane freeze - out temperature of @xmath326 k. if co is instead freezing out on a pure co ice , the binding energy will be lower .
we explored the effects of such a scenario by running a model with the co binding energy set to 1000 k. non - thermal desorption is included through stochastic heating by cosmic - rays @xcite , photodesorption @xcite , and reactive desorption @xcite .
the co photodesorption yield is set to @xmath30 @xcite .
uv photodesorption and cosmic ray heating are the most important non - thermal desorption mechanisms .
their relative contributions depend on a combination of adopted yields , cosmic ray ionization rate , dust properties , and uv field , which varies across the disk . the main target of this study , dco@xmath2 , forms through three different pathways in the model 1 . h@xmath4d@xmath2+co@xmath31 h@xmath4+dco@xmath2 , 2 .
hco@xmath2+d@xmath31dco@xmath2+h 3 .
ch@xmath4d@xmath2+o@xmath31dco@xmath2+h@xmath4 and other pathways related to ch@xmath4d@xmath2 .
the d and ch@xmath4d@xmath2 pathways regulate dco@xmath2 formation at elevated temperatures , i.e. in the inner ( r@xmath710 au ) disk and in the disk atmosphere at all radii .
the d pathway is generally the more important of the two in the present model .
@xcite , on the other hand , argued that the ch@xmath4d@xmath2 path is more important , probably due to a different disk ionization structure compared to our model .
exterior to 10 au , warm dco@xmath2 is a minor contribution to the total dco@xmath2 column density . instead , the h@xmath4d@xmath2 pathway dominates the dco@xmath2 production .
h@xmath4d@xmath2 forms efficiently at low temperatures due to the small ( @xmath7230 k ) zero - point energy difference between h@xmath29 and h@xmath4d@xmath2 ( the exact energy difference depends on the spin states of h@xmath4 and h@xmath4d@xmath2 ) .
figure [ fig2 ] shows the model disk temperature and density structure .
one particular difference worth highlighting between this generic model and i m lup is in the disk midplane temperature profile . in the model ,
20 k is reached at a radius of @xmath340 au , while a previous analysis of the i m lup disk suggests that 20 k is reached at a larger radius ( @xmath340100 au ) , dependent in detail on the model assumptions @xcite . as a result ,
co freeze - out in this model probably starts at smaller radii compared to the i m lup disk .
these model calculations do not include c and o isotope - specific chemistry . to make comparisons with the observations of isotopes
, we have scaled the calculated co and hco@xmath2 densities by factors of 500 and 70 , respectively , adopting the standard ism @xmath32o/@xmath0o and @xmath33c/@xmath1c , to estimate the c@xmath0o and h@xmath1co@xmath2 densities .
we expect that this is a reasonable approximation in the disk midplane ( in the disk atmosphere , isotopologue specific photodissociation rates will change the relative co isotopologue abundances @xcite ) .
figure [ fig2 ] ( middle row ) shows the radial and vertical gas - phase abundance profiles for the species that are the focus of this study , i.e. co , hco@xmath2 and dco@xmath2 , with the first two abundances scaled to correspond to the observed isotopologues .
all three molecules exhibit an abundance depression in the disk midplane exterior to 25 au due to freeze - out of co onto grain surfaces ( i.e. , the midplane co snow line ) . between 25 and 50 au
, the co abundance decreases by seven orders of magnitude , the hco@xmath2 and dco@xmath2 abundance by five orders of magnitude .
exterior to 90 au , the co , hco@xmath2 and dco@xmath2 gas - phase midplane abundances begin to increase . with decreasing density , the freeze - out rate decreases , which changes the co desorption / freeze - out balance and therefore increases the gas / solid ratio .
the accompanying decrease in column density results in increased uv penetration and therefore co ice uv photodesorption , further increasing the co gas / solid ratio .
the result is a substantial amount of co gas in the cold , outer disk midplane , corresponding to a c@xmath0o abundance of @xmath34 with respect to n@xmath35 .
this is six orders of magnitude above the abundance minimum found at 50 au , but still small compared to the total c@xmath0o abundance of @xmath36 , i.e. most co ( isotopologues ) are present in the form of ice in the outer disk midplane .
by contrast , a c@xmath0o abundance of @xmath36 is reached in the warm molecular disk layer , @xmath37 , and this is the main reservoir of c@xmath0o gas in the outer disk .
hco@xmath2 and dco@xmath2 are also present in the warm molecular disk layer , but this layer is no longer the dominant reservoir . in the case of dco@xmath2 ,
the midplane and molecular layer abundances are comparable .
the relatively high dco@xmath2 abundance in the midplane , at @xmath38 , is due to the increased efficiency of dco@xmath2 production at low temperatures . because of an increasing gas density toward the disk midplane ,
the vertically integrated dco@xmath2 column density is dominated by the midplane dco@xmath2 in the outer disk .
hco@xmath2 presents intermediate trends between co and dco@xmath2 .
we calculated disk emission spectral data cubes using the 3d monte carlo radiative transfer code lime @xcite from the model disk density , temperature and molecular abundance structures , for the observed c@xmath0o , h@xmath1co@xmath2 and dco@xmath2 lines .
figure [ fig2 ] ( bottom row ) shows the radial emission profiles ( arbitrary units ) of the c@xmath0o , h@xmath1co@xmath2 and dco@xmath2 obtained from the model spectral - image data cubes , integrated in velocity space and azimuthally averaged like the data . the c@xmath0o emission profile is centrally peaked .
it shows a radial gradient small change at @xmath325au at the onset of co freeze - out .
the effect of co gas in the midplane beyond 90 au on the co emission profile is subtle , and in practice unobservable , since the overall column density and thus emission at these radii is dominated by the abundant gas - phase co in the warmer atmosphere layers .
the hco@xmath2 and dco@xmath2 emission profiles decrease rapidly at radii of 2540au , due to the freeze - out of gas - phase co ( their formation reservoir ) .
the hco@xmath2 emission then presents a `` plateau '' out to 100 au , attributed to cosmic ray and uv photodesorption of co back into the gas phase .
the effect of the midplane co non - thermal desorption on the dco@xmath2 emission is , as expected , more dramatic .
the return of co to the gas at large radii , where low temperatures drives ann efficient deuterium fractionation chemistry , results in a second maximum ( outer ring ) peaking at 200 au .
dco@xmath2 also presents a small peak toward the central star , caused by warm deuterium fractionation chemistry , and an inner ring close to the co midplane snow line . unlike for co emission
, the presence of cold co in the outer midplane has a considerable effect on hco@xmath2 and dco@xmath2 emission , which should be observable .
based on a small grid of models the shapes of the predicted co , hco@xmath2 and dco@xmath2 radial profiles are not sensitive to specific model assumptions . in particular
the two concentric dco@xmath2 rings are reproduced when changing the co binding energy from 1150 to 1000 k , i.e. from a water to a co dominated ice environment , and when the cr ionization rate is reduced to 10@xmath39 s@xmath17 . in the latter case
uv photodesorption still maintains cold co gas in the outer disk , albeit at a lower level compared to when a high cosmic ray ionization rate is assumed .
that is , the outer ring is still present when the cosmic ray ionization is attenuated , but it may be less pronounced .
the molecular emission trends in fig [ fig2 ] thus appears to be robust as long as some non - thermal desorption process is active in the disk midplane . however , the absolute dco@xmath2 column density and predicted emission profile in the outer disk do depend on model parameters , and whether the predicted dco@xmath2 double - ring will be observable may thus vary from disk to disk .
qualitatively , the observed and predicted co , hco@xmath2 and dco@xmath2 radial profiles agree well . in particular
, the model reproduces several key features of the i m lup observations : the centrally peaked c@xmath0o emission profile , the h@xmath1co@xmath2 shoulder in the outer disk , and the dco@xmath2 inner ring and outer ring .
however , the model shows no signs of a central depression for c@xmath0o , and only a very small depression for h@xmath1co@xmath2 .
for dco@xmath2 , the model double - ring is accompanied by an emission peak at the disk center , which is also not observed .
the most surprising result from the alma observations is the clear detection of a concentric pair of emission rings in the dco@xmath2 32 spectral line , peaking at radii of @xmath390 and 300au .
based on the midplane temperature profile for the i m lup disk that was inferred from previous measurements @xcite , the inner ring may coincide with the onset of co freeze - out .
the relationship between the co midplane snowline and the inner dco@xmath2 ring location is , however , complicated by multiple dco@xmath2 formation pathways in the inner disk .
the outer dco@xmath2 ring requires the return of some co into the gas - phase at large disk radii .
this can be accomplished by either non - thermal ice desorption or a radial thermal inversion in the low density outer region .
of the two scenarios we favor the former , since it naturally explains the large dco@xmath2 enhancement and the much smaller h@xmath1co@xmath2 enhancement at large disk radii .
this could be observationally checked by observing additional dco@xmath2 and h@xmath1co@xmath2 lines that constrain the gas temperature .
the location of the outer dco@xmath2 ring coincides with the disappearance of mm - grain emission ( fig .
[ fig1 ] ) .
it may also coincide with a steepening of the gas - density profile @xcite .
the radial profile of smaller grains , which carry most of the opacity at uv wavelengths , is not known independently , but it seems reasonable to assume that it too is steepening at the same radius .
the dco@xmath2 outer ring then likely coincides with a disk location where the uv radiation penetration depth is increasing rapidly with radius .
the connection between the dco@xmath2 outer ring and uv radiation is supported by chemical modeling .
an outer dco@xmath2 ring is produced when co is non - thermally desorbed from icy grains in the cold , low - density outer disk . in the model uv photodesorption and cr heating
are both important , but the coincidence between the ring and the loss of dust opacity suggests that in the case of i m lup , uv photodesorption is the most important desorption mechanism in the outer disk . once in the gas phase , co can be chemically converted into dco@xmath2 following deuteron transfers from abundant reservoirs of h@xmath4d@xmath2 and other deuterated h@xmath29 isotopologues ( which are enhanced at such low temperatures ) .
this effect is expected to be generally present in disks and especially pronounced in disks with significant grain growth , which reduces uv opacity . in our model
we assume a uniform grain growth throughout the disk , parameterized as a reduction in total grain surface area by a factor of 10 .
further grain growth is likely present in older disks , and it may therefore be illustrative to compare the dco@xmath2 distribution in disks of different ages and with different grain properties as estimated from continuum observations @xcite .
based on the discussion above and the seemingly robust model results in [ model_results ] , it is tempting to conclude that concentric dco@xmath2 rings should be commonly observed in protoplanetary disks .
it is important to realize however , that the expected intensity of the second dco@xmath2 ring _ is _ model dependent .
@xcite , for example , finds a dco@xmath2 enhancement in the outer disk , but it is small compared to the findings in our model .
the same is true in the model by @xcite . in general
, the dco@xmath2 abundance in the outer disk will depend on the intensity and penetration depths of radiation fields ( uv , cosmic rays and x - rays ) , and the dco@xmath2 ring column density will depend on the total amount of cold molecular gas exposed to this radiation . the intensity of the dco@xmath2 ring will also depend on the excitation conditions that characterize the outer disk midplane , especially its density and temperature .
it may very well be that dco@xmath2 rings , while common , are only observable in a small sub - set of disks where disk density , temperature and radiation structures conspire to produce sufficient amounts of dco@xmath2 flux at large disk radii .
still the detection of one such outer ring in the i m lup disk demonstrates that non - thermal desorption in general , and uv photodesorption of co in particular , is important in setting the chemical structure of protoplanetary disks .
a lingering puzzle in the observations presented here is the central depression in the intensities of all the observed spectral lines ( but notably not in the continuum ) .
we consider four potential explanations for these depressions : ( 1 ) a reduction in the gas surface densities ; ( 2 ) line - of - sight absorption by foreground cloud material ; ( 3 ) unexpected chemistry ; and ( 4 ) high continuum optical depths that prevent some of the line emission from escaping the disk .
the first possibility seems unlikely , since there is no analogous signature in the continuum , and we are unaware of a physical mechanism that would systematically enhance the mm dust - to - gas ratio in the inner disk .
the second scenario , molecular cloud contamination , should produce deeper depressions ( absorption ) for the more optically thick spectral lines .
inspection of co data toward the same source show that co exhibits less visible depressions than c@xmath0o , ruling out this scenario .
chemistry can produce central depressions in hco@xmath2 and dco@xmath2 following desorption of h@xmath4o and other molecules with high proton affinity in the inner disk , resulting in efficient proton transfer from hco@xmath2 and dco@xmath2 to h@xmath4o .
chemistry may thus contribute to the h@xmath1co@xmath2 and dco@xmath2 holes .
the c@xmath0o central depression is not predicted by our or any other published disk chemistry model ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , though the conversion efficiency of co into other species across the disk is an active topic of research ( e.g. * ? ? ?
. based on the results presented so far , it seems unlikely , however , that chemistry alone can explain the entire observed effect .
the last proposed explanation is that the continuum optical depths in the disk center are high enough to block some of the line emission from the midplane .
the deeper depressions observed for more optically thin tracers , which originate closer to the midplane , and for lines emitted at higher frequencies ( especially h@xmath1co@xmath2 ) supports this scenario . but
this scenario has one significant problem : the peak continuum intensity at the disk center corresponds to a brightness temperature of only 6 k , much lower than expected for optically thick emission at these radii . the only way of reconciling this explanation with the data is through significant beam dilution .
if the optically thick continuum emission is concentrated in small scale features ( occupying an area @xmath35@xmath40 smaller than the synthesized beam ) , then this scenario may work .
that may seem unlikely , but the recent observations of the hl tau disk showing small scale structure @xcite indicate that it remains a possibility .
this hypothesis could be tested directly by very high spatial resolution continuum measurements , or less directly by observations of co and hco@xmath2 isotopologue lines at 3 mm , where the dust opacity should be reduced .
alma observations of the massive disk around i m lup have revealed a spectacular dco@xmath2 double - ringed structure .
the inner ring appears connected with the thermal co snowline , but dust opacity and multiple dco@xmath2 formation pathways complicate the interpretation . the second
, outer dco@xmath2 emission ring at 300 au is readily reproduced by non - thermal co ice desorption in the low - density , outer disk .
the ring coincides with a rapid decrease in dust opacity , which is likely accompanied by increased uv penetration toward the disk midplane .
we hypothesize that the dco@xmath2 outer ring is caused by efficient uv photodesorption of co ice at this radius . without the presence of non - thermal desorption
this outer disk region would be completely devoid of co gas and thus of co - gas - mediated chemistry . even at its most efficient , non - thermal desorption in disk mid planes
is not expected to maintain a large fraction of co in the gas - phase , @xmath41% of the total co reservoir .
the presented observations and models show , however , that the release of such trace amounts of co into the gas - phase can still produce an observable signature through its effects on dco@xmath2 production .
this dco@xmath2 can be used as a probe of the kinematics and ionization in the outer disk midplane , opening up a new window on the structures of protoplanetary disks .
we are grateful to ugo hincelin and eric herbst for providing rates and branching ratios for ortho / para chemistry .
this paper makes use of the following alma data : ads / jao.alma#2013.1.00226.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 ) , in cooperation with the republic of chile .
the joint alma observatory is operated by eso , aui / nrao and naoj .
the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc .
ki also acknowledges funding from the alfred p. sloan foundation , and the packard foundation .
is supported by a research fellowship from the japan scoiety for the promotion of science ( jsps ) .
astrochemistry in leiden is by a - erc grant 291141 chemplan and a knaw professorship prize . | in a protoplanetary disk , a combination of thermal and non - thermal desorption processes regulate where volatiles are liberated from icy grain mantles into the gas phase .
non - thermal desorption should result in volatile - enriched gas in disk - regions where complete freeze - out is otherwise expected .
we present alma observations of the disk around the young star i m lup in 1.4 mm continuum , c@xmath0o 21 , h@xmath1co@xmath2 32 and dco@xmath2 32 emission at @xmath305 resolution .
the images of these dust and gas tracers are clearly resolved . the dco@xmath2 line exhibits a striking pair of concentric rings of emission that peak at radii of @xmath306 and 2 ( @xmath390 and 300au , respectively ) .
based on disk chemistry model comparison , the inner dco@xmath2 ring is associated with the balance of co freeze - out and thermal desorption due to a radial decrease in disk temperature . the outer dco@xmath2 ring is explained by non - thermal desorption of co ice in the low - column - density outer disk , repopulating the disk midplane with cold co gas .
the co gas then reacts with abundant h@xmath4d@xmath2 to form the observed dco@xmath2 outer ring .
these observations demonstrate that spatially resolved dco@xmath2 emission can be used to trace otherwise hidden cold gas reservoirs in the outmost disk regions , opening a new window onto their chemistry and kinematics . |
the non - local character of quantum correlations is manifested by the violation of bell s inequality @xcite , and , more generally , of the so - called chsh inequality @xcite .
the experimental violation of these inequalities @xcite constitutes by now an important part of modern scientific culture .
this property and its relation to entanglement has in fact become one of the cornerstones of quantum information theory , see e.g. @xcite .
the relation between bell s inequalities and convex geometry is also well - known : it has been studied somewhat independently in the mathematical physics literature @xcite , ( * ? ? ? * appendix a ) , @xcite , @xcite and in the statistics literature @xcite . in particular , bell s inequalities are known to be related to the representation of a polytope as intersection of halfspaces .
nevertheless , the problem of finding a minimal set of bell - like inequalities is in general still open as it relates to the hard problem of finding all the facets of a polytope @xcite . the purpose of this note is to show that , in the simple but important original case of @xmath0 spin random variables , the fact that bell s inequalities provide a facets representation of the polytope of all spin correlation matrices can be proven in a direct , straightforward way using the available description of the extreme points @xcite .
this elementary derivation , which is apparently missing in the large body of literature on this topic , only works in the @xmath0 case .
in such a case , and for no other @xmath1 , the polytope is actually a simplex , so that it constitutes a _ tetrahedron _ , see figure [ tetrahedron ] .
let @xmath2 be a @xmath3-dimensional real vector space .
we denote by @xmath4 the _ affine hull _ of @xmath5 , namely the intersection of all linear manifolds in @xmath2 containing @xmath6 .
it can be shown that @xmath7 we denote by @xmath8 the _ convex hull _ of @xmath6 , namely the intersection of all convex sets containing @xmath6 .
it is easy to see that @xmath9 a convex set @xmath10 is a @xmath11-_simplex _ if @xmath12 and the points @xmath13 are _ affinely independent _ ,
i.e. @xmath14 .
necessarily , @xmath15 . if @xmath16 is a @xmath11-_simplex _ , every point @xmath17 admits a _ unique _ representation as convex combination of the points @xmath13 .
moreover , @xmath18 , the set of _ extreme points _ of @xmath16 , is indeed @xmath19 .
a set @xmath20 which is the convex hull of a finite number of points is called a _ ( convex ) polytope_. consider a family of spin random variables @xmath21 with @xmath22 and form the corresponding column vector @xmath23 , where @xmath24 , @xmath25 denoting transposition .
the random vector @xmath23 has zero expectation and takes values in @xmath26 .
let @xmath27 denote the corresponding covariance / correlation matrix .
given the @xmath28 , the question whether the corresponding matrix can be realized as the correlation matrix of spin random variables is a nontrivial one . indeed , while any symmetric , positive semidefinite matrix @xmath29 can be realized as the covariance of a gaussian random vector , the conditions on @xmath29 become much more stringent if we require realization through spin random variables . since @xcite , it is known that for three spins variables , the @xmath28 need to satisfy bell s inequalities to be realizable . actually , it was shown in @xcite that , in the case @xmath30 , bell s inequalities @xmath31 are necessary and sufficient for the matrix @xmath32 to be the correlation matrix of @xmath33 spin random variables .
notice that a correlation matrix is symmetric and has ones on the main diagonal .
it is therefore determined by the elements above the main diagonal .
hence , these matrices are in one to one correspondence with a @xmath34-dimensional euclidean space . for @xmath35 ,
let @xmath36 denote the set of correlation matrices that can be realized through spin random variables .
pitowsky @xcite and gupta ( * ? ? ?
* theorem3.1 ) showed that @xmath36 is a _
polytope_. it consists namely of all convex combinations of a finite number of correlation matrices .
they also gave an explicit description of the extreme points ( matrices ) of this set .
these are rank-@xmath37 matrices of the form @xmath38 , where the @xmath33-dimensional vector @xmath39 has entries equal to @xmath37 or @xmath40 .
notice that @xmath39 and @xmath41 generate the same correlation matrix so that there are @xmath42 different such matrices .
since a polytope simply consists of all convex combinations of its extreme points , this completely characterizes @xmath36 .
by the minkowski - weyl theorem @xcite , ( * ? ? ?
* appendix a ) , a polytope , besides being the convex hull of its extreme points ( v - representation ) , can also be dually described as the intersection of half - spaces through a system of linear inequalities ( h - representation ) .
for instance , a triangle in the plane is the intersection of three half - planes determined by the straight lines containing its sides .
it is known @xcite that the latter description for @xmath36 is related to the bell s inequalities ( [ bellsinequalities ] ) .
consider now the @xmath0 case , where ( [ bellsinequalities ] ) take the form @xmath43 the set @xmath44 is a @xmath45-simplex , namely a tetrahedron .
bell s inequalities ( [ bell1])-([bell4 ] ) provide the h - representation of the tetrahedron @xmath44 .
thus , any element in @xmath44 satisfies bell s inequalities ( [ bell1])-([bell4 ] ) .
conversely , any correlation matrix satisfying bell s inequalities belongs to @xmath44 . consider the four vectors @xmath46 by @xcite ,
the extreme points of @xmath44 are given by the four matrices @xmath47,\quad \sigma_2=v_2v_2^t=\left[\begin{matrix}+1 & -1 & + 1\\ -1 & + 1 & -1\\ + 1 & -1 & + 1\end{matrix}\right],\\ & & \sigma_3=v_3v_3^t=\left[\begin{matrix}+1 & + 1 & -1\\ + 1 & + 1 & -1\\ -1 & -1 & + 1\end{matrix}\right],\quad\sigma_4=v_4v_4^t= \left[\begin{matrix}+1 & -1 & -1\\ -1 & + 1 & + 1\\ -1 & + 1 & + 1\end{matrix}\right].\end{aligned}\ ] ] consider now a symmetric matrix @xmath29 with ones on the main diagonal .
as observed before , they can be bijectively mapped onto a @xmath45-dimensional space .
let @xmath48,\quad x=\left[\begin{matrix}1\\ \sigma_{12}\\ \sigma_{13}\\ \sigma_{23}\end{matrix}\right].\ ] ] knowledge of @xmath49 is equivalent to knowledge of @xmath29 .
define @xmath50.\ ] ] then bell s inequalities ( [ bell1])-([bell4 ] ) simply state that the vector @xmath51 belongs to the positive orthant @xmath52 , it has namely all nonnegative components . consider a symmetric , @xmath53 matrix @xmath29 with ones on the main diagonal expressed as _ linear _ combination of the @xmath54 : @xmath55 since @xmath56 and the @xmath57 have the same property , it follows that necessarily @xmath58 namely @xmath29 is in @xmath59 . from ( [ form ] ) , a simple calculation yields @xmath60=a\left[\begin{matrix}\lambda_1\\ \lambda_2\\ \lambda_3\\ \lambda_4\end{matrix}\right].\ ] ] observe that @xmath61 is involutory , namely @xmath62 .
since @xmath6 is invertible , it follows that any symmetric matrix with ones on the main diagonal may be expressed as in ( [ form ] ) .
in particular , every element in the convex hull of @xmath63 admits a unique representation .
hence , the @xmath57 are affinely independent and generate a simplex , namely a tetrahedron .
moreover , we have @xmath64=4\left[\begin{matrix}\lambda_1\\ \lambda_2\\ \lambda_3\\ \lambda_4\end{matrix}\right].\ ] ] thus bell s inequalities are _ equivalent _ to having @xmath65 . taking ( [ form ] ) and ( [ sum ] ) into consideration , we conclude that a convex combination of the @xmath54 is a correlation matrix satisfying bell s inequalities .
namely , any element in @xmath44 satisfies bell s inequalities ( [ bell1])-([bell4 ] ) .
conversely , any correlation matrix satisfying bell s inequalities belongs to the convex hull of the @xmath54 , i.e. it belongs to @xmath44 .
the above elementary argument does not extend to higher dimensions .
in particular , @xmath36 is not a simplex for @xmath67 .
for instance , we know from @xcite that also in the case @xmath68 bell s inequalities are necessary and sufficient for @xmath29 to be realizable as correlation matrixof spin random variables . there are @xmath69 such inequalities .
the set @xmath70 has eight extreme points whereas @xmath29 is determined by six elements ( those above the main diagonal ) .
but , in a @xmath71-dimensional space , a simplex can have at most @xmath72 extreme points .
thus the polytope @xmath70 is not a simplex . more generally , the equation @xmath73 relating the number of extreme points of @xmath36 and the maximum number of affinely independent points in a space of dimension @xmath34 has no solution for @xmath67 .
moreover , for @xmath74 , bell s inequalities ( [ bellsinequalities ] ) are known not to be sufficient for a correlation matrix to be the correlation matrix of spin random variables @xcite . | using results of pitowsky and gupta , we show in a direct , elementary fashion that , in the case of three spins , bell s inequalities indeed provide a representation of the tetrahedron of all spin correlation matrices as intersection of half - spaces . + + mathematics subject classification[2010]:52b12,62h20,81p40 |
cometary globules are believed to be molecular cloud condensations , which are so dense that they are not disrupted when an region expands into the molecular cloud(s ) surrounding it .
such globules are always bright rimmed and often have a cometary or tear shaped form , because the stellar wind and ionizing radiation from the central o stars ablate away the low density gas on the side facing the o stars and sweep away dust towards the tail .
cometary globules are thought to form from `` elephant trunks '' @xcite , although for compact cores surrounded by low density gas this phase can be quite short .
the rosette nebula is a good example of an region , which shows an abundance of both cometary globules and elephant trunk structures @xcite .
although elephant trunks and cometary globules have been known for a long time , the fact that these sites may be forming stars was realized only in the last few decades . the first clear confirmation that stars form in cometary globules was a result of the identification of bernes 135 in the cometary globule cg1 , one of the large cometary globules in the gum nebula , as a pre - main - sequence ( pms ) star @xcite .
the discovery of molecular outflows associated with cold iras sources without optical counterparts in three bright - rimmed globules @xcite , definitely confirmed that stars form in elephant trunks and cometary globules .
probably the most well - known example of star forming in the tip of elephant trunks is seen in the hubble poster , `` the pillars of creation '' , in the eagle nebula ( m16 ) @xcite . in this paper we examine star formation in a more nearby cometary globule , ori i-2 , located in the large region ic434 .
this is one of the three globules , in which @xcite discovered a molecular outflow . by using _
irac and mips images , 2mass data , deep r and h@xmath1 images from the nordic optical telescope ( not ) , and scuba sub - millimeter imaging we can identify all young stars in and near the globule and provide more accurate information on their physical characteristics .
@xcite were the first to notice this globule in the large region ic434 and named it orii-2 .
the region ic434 is illuminated by the trapezium like system @xmath3 ori with at least five early type stars @xcite .
the two hottest members , a and b ( o9.5v and b0.5v ) form a binary system , which allows for an accurate distance determination .
@xcite derived a distance of 334 pc assuming that they form a binary , although he also considered the possibility that it is a hierarchical triple system , in which case the distance would be @xmath4 385 pc .
@xcite , in their review of distance estimates to the @xmath3 orionis cluster , quote a slightly larger distance , 420 @xmath5 30 pc .
the morphology of orii-2 , with the bright rim facing @xmath3 ori and the tail aligned in the direction of the star suggests that the globule is at the same distance or slightly in front of @xmath3 ori . in this paper
we adopt the distance to orii-2 as 380 pc , favoring the distance estimate by @xcite .
@xcite included orii-2 in their molecular line survey of small bok globules and found it to be the densest , the most compact , and the warmest , 18 k , of all the globules in their sample .
@xcite discovered a bipolar molecular outflow in the globule .
they found that the outflow had a linear extent of 0.34 pc , a dynamical age of 5 @xmath2 10@xmath6 yr , and was centered on a low luminosity ( l = 11 ) cold iras source , iras @xmath7 , without any optical counterpart .
@xcite did a more detailed study of orii-2 in co , @xmath8co and c@xmath9o j = @xmath10 , hcn j = @xmath11 , cs j = @xmath12 and cs j = @xmath13 with higher spatial resolution and sensitivity than @xcite and @xcite .
they confirmed that the molecular outflow is bipolar and centered on the iras source , iras @xmath7 , but found it to be more compact ( 0.22 pc ) than what @xcite estimated from their low - spatial resolution observations . by examining the palomar sky survey prints and doing ccd imaging in r , i , and narrowband h@xmath1 and [ ]
, they showed that the bright rim is a diffuse region and not a reflection nebula .
they identified a faint star , star a , as a possible counterpart to the iras source .
they found that the gas temperature of the shock heated gas in the bright rim is @xmath4 25 k , while the gas temperature in the opaque ( a@xmath14 @xmath4 20@xmath15 ) core of the globule is @xmath4 12 k , similar to that of normal isolated globules .
they determined the total mass of the globule to be @xmath4 8 , with the opaque core having a mass of @xmath4 2.3 ( corrected to a distance of 380 pc ) .
orii-2 was included in the catalogue of remnant molecular clouds in the ori ob1 association by @xcite as cloud number 40a .
they also identified three fainter cometary globules `` behind '' the tail of orii-2 ( 40a ) , which they called 40b , c , and d. @xcite , who did a photographic and ccd imaging survey of the l 1630 and l 1641 molecular cloud regions in orion , identified a chain of hh objects , hh289 , to the east of ori i-2 and a tube like feature ( cavity ) protruding out on the western side of the globule , both apparently excited by the embedded iras source .
their observations show that the iras source drives a giant outflow with a projected size of @xmath16 pc to the east .
a good overview of orii-2 and other globules in the vicinity of @xmath3 orionis can be found in @xcite .
we have extracted irac and mips ( 24 & 70 ) observations from the spitzer space observatory archive ( program i d 30050 : star formation in bright rimmed clouds by fazio et al . ) .
the irac data were taken in the high dynamic readout ( hdr ) mode using a single aor ( astronomical observation request ) with a five - point dither pattern .
we have processed both the short ( 0.6 sec ) and the long ( 12 sec ) integration basic calibrated data ( bcd ) frames in each channel using the artifact mitigation software developed by sean carey and created mosaics using mopex .
these irac observations go far deeper than any previous observations of orii-2 . in the short wavelength bands we can detect point - sources down to 30 40 @xmath17jy and are slightly less sensitive in the long wavelength bands , but in regions without nebulosity
we can still detect point - sources down to @xmath4 60 @xmath17jy .
we have also created mosaics of the mips 24 and 70 bcds using mopex .
both data sets are of excellent quality .
the 70 @xmath17m - image , due to the lower angular resolution , shows only one bright point - like source embedded in the globule .
the image also shows strong emission from the hot ionized rim facing @xmath3 ori and fainter emission on the western side of the globule .
we have carried out multiframe psf photometry using the ssc - developed tool apex on all the _ spitzer _ irac images and on the mips images .
the 3.6 and 4.5 irac images show signs of saturation on bright stars . we have used combination of automated routines and eye - inspection to detect sources and extract photometry of these sources from the irac and mips images . for sources ,
which apex failed to detect at one or several wavelengths , we have used the user list option in apex to supply the coordinates for the source to extract it and perform photometry on it .
this enabled us to derive photometry for every source , which we could visually identify on any image . in the irac images
we have detected 118 , 122 , 41 and 32 sources respectively at 3.6 , 4.5 , 5.8 and 8.0 .
since the field of view of the two irac cameras do not completely overlap , we end up with a total of 125 mid - infrared sources .
we detect five of them in the 24 mips image , but only one in the 70 image .
we have crosscorrelated the sources detected in the irac and mips bands , and the 2mass point source catalogue .
we have used the following association radii : 1 for the irac images , 25 for the mips 24 image and 2 for 2mass data .
table [ tab_mirsrc ] gives the coordinates of the 125 sources identified in orii-2 together with their @xmath0-band and 2mass magnitudes , _ spitzer _ irac and mips flux densities and a preliminary classification based on selected color - color plots . in table
[ tab_mirsrc ] we have given them the prefix orii-2 , but throughout the paper we simply refer to them as mir - nn , where nn is the number of the source . out of the 125 sources 51 were found to have 2mass counterparts .
all the sources detected in the mips 24 image were also detected in the four irac bands .
we find that around 50% of the sources in the list have been detected only at 3.6 and 4.5 and have no counterparts in 2mass as well , although many of them are seen in the deep @xmath0-band image .
the optical ccd images presented here were obtained on october 26 2007 using the 2.56 m nordic optical telescope ( not ) located at el observatorio del roque de los muchachos on the island of la palma in the canary islands .
we used the andalucia faint object spectrometer and camera ( alfosc ) , which uses a thinned loral 2048 @xmath2 2048 ccd array with 15 @xmath17 m pixels giving a field of view of 6.5 @xmath2 6.5 arcmin@xmath18 with 0.187 arcseconds pixel@xmath19 .
for these observations we obtained two images with 300 seconds exposure time in bessel r ( 6500 ) separated by offsets of @xmath2090,+90 from the center of the globule in right ascension and declination , and a short 30 second exposure centered on the globule .
we also observed the globule in narrow - band h@xmath1 .
the narrow band h@xmath1 filter was a circular filter giving a more restricted field of view and centered on 6564 with a bandwidth of 33 .
here we obtained two images , each with 300 second exposure and separated by 60 in right ascension .
the observing conditions were excellent with an average seeing of @xmath4 075 .
the images were reduced with the starlink program suite ccdpack .
the transformation to the world coordinate system ( wcs ) was done using the starlink program gaia and we used more than twenty 2mass stars as astrometric reference stars .
the 2mass stars have very good astrometric accuracy and the fit to these reference stars indicate that the astrometry of the ccd images have an accuracy of @xmath21 01 for both the r and the narrow - band h@xmath1 image compared to the 2mass reference frame .
the images of the photometric standard stars were unavailable , hence the @xmath0-band image was calibrated using the 30 second exposure and seven stars in the 14 - 15@xmath15 range from the usno - a2.0 catalogue .
we estimate the photometric calibration accuracy to be @xmath4 0.1 - 0.2@xmath15 .
this calibration was transferred to the h@xmath1 image using a somewhat larger set of overlapping stars with approximately neutral colors judged from 2mass j magnitudes .
the h@xmath1 image is estimated to have a photometric accuracy no better than 0.3@xmath15 .
the r band image has a limiting magnitude of @xmath4 23.5@xmath15 and stars brighter than 15@xmath15 are partially saturated .
we did aperture photometry in r and h@xmath1 of all stars , which coincided with an irac source to within 1 .
we present only the result of the r band photometry in table [ tab_mirsrc ] .
the h@xmath1 is virtually identical , except for one star , see below .
we detect all 2mass sources in r band as well as 26 irac sources without 2mass detection .
comparison of the r and h@xmath1 photometry revealed only one h@xmath1 emission line star .
mir-52 has an h@xmath1 magnitude of @xmath4 18.5 , while it is @xmath4 19.9 in r ( table [ tab_mirsrc ] ) .
the excess seen in h@xmath1 , 1.4@xmath15 , is much larger than our photometric uncertainty .
therefore the star is definitely an h@xmath1 emission line star .
mir-52 has a strong ir excess in the mips 24 band , so that we identify it as a class ii object based on irac and mips color color diagrams , see section [ sec_ysoclass ] .
the 850 @xmath17 m and 450 @xmath17 m continuum observations were obtained with bolometer array scuba on jcmt , mauna kea , hawaii .
scuba @xcite has 37 bolometers in the long and 91 in the short wavelength array separated by approximately two beam widths in a hexagonal pattern .
the field of view of both arrays is @xmath42.3 .
both arrays can be used simultaneously by means of a dichroic beamsplitter .
the scuba observations reported in this paper were all obtained in jiggle - map mode @xcite .
on october 17 , 1997 we obtained two separate five integration maps with a chop throw of 100 in azimuth ; which unfortunately was not sufficient to completely chop outside of the globule . on december 17 , 1997
we therefore used a 120 chop with a fixed position angle of 80 measured from north ( equatorial reference frame ) .
this time we obtained three maps , each with small position offsets ( dither ) with three , eight , and five integrations .
the total integration time from both nights was therefore 3328 seconds .
the sky conditions were good on both nights .
on october 17 the atmospheric opacity measured with the caltech submillimeter observatory ( cso ) taumeter , cso @xmath22 was @xmath4 0.07 , and on december 17 @xmath23 was 0.04 - 0.05 .
the dust emission from the globule is too extended to be observed in jiggle map mode and the emission from the cold cloud core is therefore likely to be underestimated , especially at 450 @xmath17 m .
any strong emission in the off source positions was carefully blanked out in the data reduction stage .
it should not affect the morphology or photometry of the strong , compact source , which dominates the emission both at 850 and 450 @xmath17 m .
pointing corrections were derived from observations of the blazar 0528 + 134 on october 17 , and the blazar 0529 + 075 on december 17 .
the secondary calibrators crl618 and hl tau were used for flux density calibration .
we estimate the calibration accuracy to be @xmath4 10% at 850 and @xmath4 20% at 450 . the half power beam width ( hpbw )
was estimated from observations of uranus obtained in the early part of each night and was found to be @xmath4 145
15 at 850 @xmath17 m and @xmath4 8 85 at 450 @xmath17 m .
the data were reduced in a standard way using surf @xcite and starlink imaging software , i.e. , flat fielded , extinction corrected , sky subtracted , despiked , and calibrated the images in jy beam@xmath19 .
each data set was corrected for any drift in pointing between successive pointing observations and the data were added together to determine the most likely submillimeter position at 850 .
once we had derived a basic 850 astrometric image , we made gaussian fits of the compact sub - millimeter source in each data set and derived small additional ra and dec corrections to each scan ( shift and add ) to sharpen the final image to this position . since there is a small mis - alignment between the 850 and 450 arrays , we first corrected the 450 images for any pointing drifts and then did shift and add using the position derived at 850 .
we estimate the astrometric accuracy to be @xmath24 2 .
the final coadd was done by noise - weighting the data in order to minimize the noise in the final images .
the rms of the 450 image is @xmath4 0.50 mjy beam@xmath19 and @xmath4 10 mjy beam@xmath19 for the 850 image .
all the maps were converted to fits - files and exported to miriad @xcite for further analysis . in order to correct for the error lobe contribution , especially at 450
, we have deconvolved all the maps using clean with the same beam model used by @xcite . since the hpbw varies slightly from night to night , this model beam is not ideal , but it is the best we can do .
the actual beam size is probably somewhat more extended .
figure [ fig_scuba ] shows the 450 and 850 scuba images of orii-2 .
these images show a single bright source , located at the head of the globule beyond the optically bright rim .
this source , smm1 , is coincident with mir-54 , which is the only source seen in the mips 70 image . in order to derive the position and size of smm1 we have fitted a two component elliptical gaussian using the task imfit in miriad , one for the sub - mm source , and the other for the surrounding cloud .
the fit to the broader component is mainly to provide a good subtraction of the extended emission , and is not to estimate the flux density of the surrounding cloud .
the sub - millimeter position is @xmath1(2000.0 ) = 05@xmath25 38@xmath15 05098 , @xmath26(2000.0 ) = @xmath27 01 45 114 .
the integrated flux densities are 0.54 @xmath5 0.08 jy and 3.36 @xmath5 0.75 jy for 850 and 450 , respectively .
the fitted size , 179@xmath2 154 , should be considered an upper limit , since the true beam sizes were probably somewhat larger than the model beams used for deconvolving the images .
since orii-2 lies within the @xmath3 orionis cluster ( age @xmath4 3 myr ) , we expect to detect some young stars belonging to the cluster in the 5 @xmath2 5 field that we are analyzing .
stars belonging to the @xmath3 orionis cluster should be preferentially located outside the globule and generally appear older than stars that have recently formed in the globule . figure [ fig_composite ] present three - color images of orii-2 using irac 3.6 , 4.5 and 8 in the _ left _ panel and the irac 3.6 and 8 bands combined with the mips 24 channel in the _ right _ panel .
both images show that the front side of the globule , which is dominated by strong polyaromatic hydrocarbon emission ( pah - emission ) is far from smooth .
this may in part be due to rayleigh - taylor instabilities at the ionization front , but it is also due to outflows from the young stars which have recently formed in the cloud .
the outflow from mir-54 ( iras 05355 - 0146 ) , the bright red star in the irac / mips color image , is clearly seen as a cavity like structure to the east of the star and a bluish nebulosity on the western side of the star .
this outflow is very prominent in the h@xmath1 image ( figure [ fig_halpha_rband ] ) , where it shows up as a limb brightened cavity , extending @xmath4 50 from the western edge of the globule and @xmath4 1.5 from the star exciting it .
another outflow just south of the h@xmath1 outflow and approximately aligned with it , can be seen in the irac images terminating at a blue star ( unrelated to the outflow ) .
this outflow is powered by another young star , mir-52 , @xmath4 20 to the south of mir-54 . to the east both outflows overlap .
the western side of the globule is very sharp with an ionized rim . to the east
the the boundary between the globule on the surrounding region is more diffuse .
the most opaque part of the globule is just north or northwest of the iras source .
the bright star located @xmath4 43 away from mir-54 in the color image is sao 132389 ( hd 37389 ) .
this star , identified as mir-73 in table [ tab_mirsrc ] , is detected even at 24 , although it is an unreddened foreground a0 star with an @xmath28 .
the star has a radial velocity of @xmath29 km s@xmath19 , which is very different from the radial velocity ( 12 ) of orii-2 @xcite .
most of the faint stars outside the globule are field stars , although a few are young stars belonging to the @xmath3-orionis cluster .
the irac and mips data are ideal for detecting young stellar objects ; especially in obscured regions like in orii-2 , which has a visual extinction exceeding 20@xmath15 . outside the globule , where the extinction is low , search for brown dwarfs , for which the spectral energy distribution ( sed ) peaks at j , can be most efficiently performed using near - infrared data .
however , we go deeper than 2mass in the 3.6 and 4.5 irac bands .
the same is true even for the not broadband r filter .
there is therefore no particular advantage in using the 2mass data alone .
all 2mass sources in our field have also been detected by irac , at least in the two short wavelength bands .
furthermore , at the distance of the @xmath3 orionis cluster only the brightest m dwarfs are barely detectable with 2mass .
since we can see several galaxies in the deep r band image , we first analyzed the irac data to eliminate any galaxies , which could have mir colors similar to a yso .
@xcite demonstrated that ( a ) normal star - forming galaxies and narrow - line agns with increasing 5.8 and 8.0 and ( b ) broad - line agns with red , nonstellar seds , result in colors which are very similar to bona - fide ysos .
it is therefore necessary to inspect deep irac images for contamination due to the extragalactic sources . @xcite
have extensively discussed the criteria for identifying such extragalactic objects in the irac color - color diagrams .
if we use the criteria given in the appendix of their paper , we identify mir-16 , 25 , 88 , 89 , 93 , and 114 as extragalactic .
however , inspection of these sources in the r - band ccd image , which was taken in excellent seeing conditions , suggests that most of them are stellar , with the exception of mir-93 .
this source is clearly extended , and therefore it is an extragalactic source . for mir-88 , which is close to the detection limit in the r - band ,
we get no help from the r band image , it could be stellar , or it could be extragalactic .
mir-89 , which is not detected in the r band , was also identified as an extragalactic object .
however , not only irac , but also the mips colors suggest that mir-89 is a class 0/i object . this classification is also supported by its location in the eastern side of the bright rim .
we therefore only positively identify mir-93 as an extragalactic object .
figure [ fig_iracmipscol ] presents color - color diagrams of 2mass , irac and mips 24 sources detected in the orii-2 field .
we have used several criteria ( shown as dashed lines and boxes in figure [ fig_iracmipscol ] ) to identify potential pms stars using these color - color diagrams .
the most stringent classification scheme uses the irac ( [ 3.6][5.8 ] ) colors and the [ 8][24 ] irac and mips color . at 24 the reddening due to extinction is small and the photospheric colors are very close to zero for all spectral types @xcite .
therefore the [ 8][24 ] color is very sensitive to infrared excess , but of course not all young stars bright are enough to be detected at 24 . using this color - color diagram
we find four sources with infrared excess ( figure [ fig_iracmipscol ] ( _ left _ ) ) .
we identify mir-54 and mir-89 as class 0/i sources , while mir-41 and mir-52 are in the class ii regime .
figure [ fig_iracmipscol ] ( _ middle _ ) presents the [ 3.6][4.5 ] vs [ 5.8]-[8.0 ] color - color plot for the sources detected in all the irac bands .
sources with the colors of stellar photospheres are centered at ( [ 3.6][4.5],[5.8][8.0])=(0,0 ) and include foreground and background stars as well as diskless ( class iii ) pre - main sequence stars .
the box outlined in figure [ fig_iracmipscol ] ( _ middle _ ) , defines the location of class ii objects @xcite , i.e. sources whose colors can be explained by young , low - mass stars surrounded by disks .
@xcite have shown from their observations of young stars in the taurus - auriga complex that class 0/i protostars require [ 3.6][4.5]@xmath30 and [ 5.8][8.0]@xmath30 . while mir-89 clearly falls in the class 0/i region , mir-54 does not .
it is well above the horizontal line corresponding to [ 3.6][4.5]=0.7 , but the color [ 5.8][8.0 ] is slightly less than 0.7 .
since the demarcating lines primarily serve as a guidance and mir-54 is securely identified as a class 0/i object based on the irac / mips color - color plot , we identify it as a class 0/i object here as well . in this color - color diagram
five stars end up in the class ii color regime : mir-23 ( marginally ) , mir-25 , mir-41 , mir-51 , and mir-88 .
mir-51 , which shows a large excess in the [ 5.8][8.0 ] color is a class ii / i source , i.e. a flat spectrum source .
we did not pick up mir-52 , because the [ 3.6][4.5 ] color is only 0.12 , yet it is the only strong h@xmath1 emission line star in our sample ( see section [ not ] ) and definitely a pms star .
it has , however , a strong infrared excess longward of 8 , which is why it was picked up in the [ 3.6][5.8 ] vs. [ 8][24 ] color - color diagram .
@xcite also investigated color - color diagrams using the shortest irac bands , which have the highest sensitivity and which are least affected by pah emission with near - ir colors taken from the 2mass catalogue .
we have similarly used the photometry for the sources presented in table [ tab_mirsrc ] which have 2mass counterparts to plot the @xmath31[3.6 ] versus [ 3.6][4.5 ] color - color diagram ( figure [ fig_iracmipscol ] _ right _ ) .
we identify 12 potential pms stars , i.e. stars which are located outside the box demarcating stellar sources .
these sources are mir-10 , 29 , 41 , 47 , 49 , 51 , 52 , 61 , 63 , 78 , 93 , and 117 . if we account for the large errors in faint 2mass sources we find at least one more source , mir-17 , which appears to have a clear excess in the 5.8 filter .
for both mir-10 and 78 the 2mass @xmath31 fluxes have large uncertainties .
it is therefore desirable to have additional evidence of ir excess in order to justify their classification as pms objects .
while mir-10 shows a clear excess at 5.8 , mir-78 is not detected beyond 4.5 and it s sed is completely stellar with no sign of an ir excess .
thus we do not consider mir-78 to be a pms object . since mir-93 was found to be an extragalactic source , we are left with 10 potential pms stars .
of these we have already identified mir-41 , 49 , 51 and 52 as pms stars based on mir color - color plots .
the two class 0/i sources , mir-54 and 89 , and mir-88 , a potential class ii source , were not detected in @xmath31 , and hence do not appear in this color - color diagram .
neither do the sources mir-23 , and 25 , both of which were classified as pms objects in the irac color - color diagram , show @xmath31[3.6 ] or [ 3.6][4.5 ] colors expected for class ii objects .
the @xmath31[3.6 ] color is 0.3 for mir-23 , which would put it in the class ii regime , while it is @xmath4 0 for mir-25 .
the latter , however , does have a significant color excess at 8 .
not all of the stars identified in the 2mass / irac color color diagram are likely to be pms stars , they could just be heavily reddened background stars .
mir-29 , 47 , and 61 are all located in the western or southern rim of the globule , and have very blue irac colors .
they are all almost certainly heavily reddened background stars .
the same is true for mir-63 which is seen toward the opaque part of the globule .
we identify mir-63 with the star that @xcite labeled b , and proposed might be a stellar counterpart to the iras source . however , the [ 3.6 ] [ 4.5 ] color for mir-63 is extremely blue , @xmath272.4@xmath15 , which shows that the red near - ir colors are simply due to extinction .
the same is true for mir-117 , which is located east of the globule in a relatively low extinction area . for this star
the @xmath32[3.6 ] color excess is rather marginal and there is no excess in the irac bands .
it appears to be a reddened background star .
mir-10 lies outside the globule in a region of low extinction and is very red in the [ 4.5 ] [ 5.8 ] color band , and could therefore be a pms star , although it was not detected at 8 . in all we therefore identify 11 potential pms objects based on 2mass , irac and mips colors .
we classify mir-54 and 89 as class 0/i stars , while mir-10 , 17 , 23 , 25 , 41 , 49 , 51 , 52 , and 88 are class ii objects .
we also note that mir-104 , which lies just outside the southern rim of the globule , was only detected at 5.8 and 8 , suggesting that it has an infrared excess .
it is therefore a potential pms object . among these pms objects mir-10 , 17 , and
41 are located well south of the globule , while mir-25 is west of the cometary globule .
if all of these four sources are pms stars , they most likely belong to the young @xmath3-orionis cluster and are unrelated to orii-2 .
all the rest could potentially have formed in the globule . in section
[ sed_modeling ] we investigate the seds of these candidate pms objects in more detail to see whether they are young stars , and if so , what their physical properties are .
@xcite identified a parsec scale herbig - haro outflow , hh289 , to the east of the globule by analyzing images from the aat / ukst h@xmath1 survey , broadband r ( ivn ) plates and narrowband ccd images in h@xmath1 + [ ] .
their hh objects 2 5 , i.e. knots c f , are all outside the area we imaged in r and h@xmath1 .
we do detect their knot b in our h@xmath1 image , but not knot a. knot b appears very elongated and is located in the cavity wall from the red - shifted outflow from mir-54 .
@xcite s feature 1 , which coincides with the outflow cavity emerging to the sw out of the globule is also readily visible in the h@xmath1 image ( figure [ fig_halpha_rband ] ) .
although we see no trace of knot a in either r or h@xmath1 , it seems to coincide with the bluish nebulosity west of mir-54 , which is very prominent in the irac three color image ( figure [ fig_composite ] ) . in the 3.6 and 4.5 images the mir-54 nebulosity appears to be the inner part of an outflow cavity which is not visible in the mid - ir outside the bright western rim of the globule .
however , this outflow lobe is seen more prominently as a limb - brightened h@xmath1 cavity extending out another 50 from the western edge of the globule .
the total length of the outflow is therefore @xmath4 1.5 , 0.17 pc . since this outflow extends into `` empty '' space , i.e. the low density region ic434 , it is possible that the outflow extends even further , but becomes so diffuse that it is no longer visible .
it is probably the counterpart to the large herbig - haro flow , but since the density of the globule is much lower on the eastern side of the globule , it has been able to expand more freely .
however , in the irac images one can also see an outflow cavity from mir-52 protruding out of the globule .
this outflow is approximately parallel with the mir-54 outflow . it is brighter on the southern side of the cavity and has an apparent length of @xmath4 50 from mir-52 . to the east
the two outflows appear as a large evacuated hole ( cylinder ) and can not be separated .
either star could therefore drive the large hh outflow .
however , we also detect a herbig - haro like nebulosity with three knots or condensations in the r - band image @xmath4 140 east of mir-54 .
a blow - up of this nebulosity is shown in figure [ fig_halpha_rband ] , although it is not seen at all in the h@xmath1 image . normally hh objects are bright in h@xmath1 . if this is an hh object , the emission must completely be dominated by [ sii ] emission , i.e. it has to be an extreme low excitation hh object .
it is not detected by 2mass , nor is it seen in the irac images , which go deeper than 2mass .
it is , however , seen faintly in @xcite s h@xmath1 + [ sii ] image , which proves that it is real and not just an instrumental artifact .
in fact it appears quite strong in @xcite s 2.12 molecular hydrogen image , while the nebulosity is completely absent on their ivn plate .
this confirms that there is no continuum at the position of the nebulosity .
the only conclusion we can draw , is that this is a rather low excitation hh object , which is completely dominated by line emission from [ sii ] and which is also seen in vibrationally excited h@xmath33 at 2.12 .
the coordinates of the three hh knots , which are here labeled hh992 , are given in table [ tab_hhobjects ] .
this hh object is not a part of the large herbig - haro flow , hh289 , seen by @xcite .
the most likely exciting source of the hh992 complex is mir-89 , which is @xmath475 west of the hh objects .
mir-89 was identified as a class 0/i object in the color - color diagrams , but sed modeling ( section [ sed_modeling ] ) suggest that it is a class ii object . in the 3.6 and 4.5 images
one can see a limb - brightened fan - shaped nebulosity emerging to the west from the star .
this nebulosity may trace an outflow into the globule , which is roughly aligned ( p.a . @xmath4
-95 ) with the hh objects on the opposite side of the star .
in order to interpret the observed seds for the sources identified to be pms and better characterize these sources we have explored the archive of two - dimensional ( 2d ) axisymmetric radiative transfer models of protostars calculated for a large range of protostellar masses , accretion rates , disk masses and disk orientations created by @xcite .
this archive also provides a linear regression tool which can select all model seds that fit the observed sed better than a specified @xmath34 .
each sed is characterized by a set of model parameters , such as stellar mass , temperature , and age , envelope accretion rate , disk mass , and envelope inner radius .
we have used this online tool to generate models , which fit the observed seds for the 11 candidate pms objects .
seds of six of the objects could be adequately fit with accretion disk models , the other five were better fit with stellar atmospheric models , reddened by foreground extinction .
we restricted the sed fitting tool to explore only distances between 350 and 450 pc .
below we present and briefly discuss the best fit models . figure [ fig_sedfits ] shows results of detailed modeling of the observed sed in the mid - infrared ( and fir and sub - mm for mir-54 ) for all the candidate pms objects .
a major criticism against the use of these models has been the non - uniqueness of the solutions obtained from the model library .
however we find that for four out of the six pms objects , _ viz . ,
mir-51 , 52 , 54 and 89 _ the best fit models are either distinctively better in reproducing all the observed flux densities than the next few models or that the distances for other possible models do not match the distance to orii-2 .
for mir-23 we find 14 models with @xmath34@xmath35 and for mir-41 we find four models with @xmath34@xmath36 , which appear plausible . for mir-23
we find two independent models for which the inclination is varied to generate the 14 models .
for each of these two sources we have obtained the average values and variances of the fitted parameters by taking a weighted average of the apparently acceptable models in a manner similar to @xcite .
table [ tab_sedfits ] presents the parameters corresponding to the best fit models for the six sources which we fitted with accretion disk models .
@xcite presented a classification scheme which is essentially analogous to the class scheme , but refers to the actual evolutionary stage of the object based on its physical properties like disk mass and envelope accretion rate rather than the slope of its near / mid - ir sed . according to the @xcite classification scheme , mir-54 ,
is a stage 0/i object , while all the others are stage ii objects . at a first glance it is surprising to find that mir-89 , which we classified as a young class 0/i object both in the irac and irac - mips color - color diagrams , is now identified as an evolved stage ii object .
the sed model shows that it only has a minuscular envelope and there is no sign of envelope accretion .
all reasonable sed models suggest that it is a heavily extincted ( @xmath37 50@xmath15 ) , very low - mass star .
this is the sole reason why it appears as a class 0/i object in the color - color diagrams , because the observed magnitudes are not corrected for extinction in these diagrams .
the star still has a rather massive disk compared to the stellar mass , @xmath38@xmath4 0.02 , and a moderately high disk accretion rate of 3.2 10@xmath39 yr@xmath19 .
it is located in the bright rim on the eastern edge of the cometary globule orii-2 , and appears to drive an outflow seen as a fan - shaped nebulosity west of the star , and as the herbig - haro objects , hh992 , seen @xmath465 , 0.12 pc to the east of the star .
mir-51 , which we identified as a class i / ii object based on color - color diagrams , appears more similar to mir-52 ( class ii object ) and if anything slightly older , not younger .
both appear to be young stage ii objects ( @xmath40/@xmath41 10@xmath42 yr@xmath19 and @xmath43/ @xmath41 10@xmath44 ) and are embedded inside the globule .
this is consistent with mir-52 having strong h@xmath1 emission and powering an outflow .
the sed models place mir-23 and 41 as late stage ii objects .
they have no measurable envelope and very low disk accretion rates ( table [ tab_sedfits ] ) .
the models also suggest no or low foreground extinction .
these results of modeling together with the location of these two objects in the region suggest that these objects are therefore more evolved pms stars belonging to the @xmath3-orionis cluster , although we can not exclude that they could have formed in the globule at an earlier epoch .
we could not obtain any believable dust / envelope models for mir-10 , 17 , 25 , and 88 , which are all outside the globule in regions of low or moderate extinction , or for mir-49 , which is located in the most opaque part of the globule .
we therefore fitted these objects with pure stellar atmospheric models reddened by foreground extinction .
these fits are not very accurate , since we have no optical data , other than the r magnitudes .
furthermore , mir-10 , 17 and 49 were not detected at 8 , while mir-88 was not detected by 2mass .
we have however used upper limits from these bands to better constrain the fits .
* mir-10 : * at wavelengths shortward of 4.5 , the observed sed is fitted reasonably well by a stellar photosphere = 5250 k , log@xmath45z / h@xmath46$]=-0.5 , log@xmath45g@xmath46 $ ] = 0.0 and = 1.45@xmath15 .
however , the star shows an excess at 5.8 , which suggests that it could be a transitional disk object .
* mir-17 : * similar to mir-10 the 2mass and short wavelength irac data for this star can be explained by reddening . even though it also shows an ir excess at 5.8 , when compared with the sed due to stellar photosphere = 3500 k , log@xmath45z / h@xmath46$]=-0.5 , log@xmath45g@xmath46 $ ] = 5.0 and @xmath4 2@xmath15 , the 8 upper limit suggests that there is no excess at 8 .
we therefore dismiss mir-17 as a pms star .
* mir-25 : * a stellar model with @xmath47 = 3500 k , log @xmath45z / h@xmath46$]=-2.5 , log@xmath45g@xmath46 $ ] = 5.0 , and = 0@xmath15 gives a reasonable fit to the observed data
. there might be a small excess at 8 , which could continue to longer wavelengths . based on the available data
, we therefore retain mir-25 as a candidate pms star , although the evidence for dust excess in this star is rather marginal . * mir-49 : * the sed of this source is well reproduced by a stellar photosphere with @xmath47 = 3500 k and = 16@xmath15 .
mir-49 therefore appears to be a heavily reddened background star .
* mir-88 : * the stellar model with = 3500 k , log@xmath45z / h@xmath46$]=-0.5 , log@xmath45g@xmath46 $ ] = 0.0 = 6.1@xmath15 reproduces the observed sed significantly better than the accretion disk models and there is no evidence for excess emission in the irac bands .
it is therefore a reddened background star .
most of the stars , which could not be fit with accretion disk models , appear to be reddened background stars , except perhaps for mir-10 and 25 , both of which could have some long wavelength excess .
it is possible that they could be transitional disk objects belonging to the @xmath4 3 myr @xmath3 orionis cluster .
transitional disk objects are stars , for which the inner disk has been cleared out , see e.g. @xcite , and which therefore do not show any excess until 6 or 8 . for mir-54 the models completely fail to reproduce the observed sub - millimeter flux densities , irrespective of the values of any of the input parameters .
we discuss mir-54 in more detail in section [ sect_mir-54 ] .
based on our sed modeling we conclusively identify six pms stars .
all the pms stars in the orii-2 region appear to be low luminosity stars of spectral types k and m. the pms stars outside the globule have very low disk masses and disk accretion rates , 10@xmath48 to a few times 10@xmath49 yr@xmath19 ( table [ tab_sedfits ] ) .
this is not surprising since these stars are most likely members of the @xmath3-orionis cluster , which has an age of @xmath4 3 myr @xcite .
mir-54 is the mid - infrared counterpart to iras 05355 - 0146 , which was detected in all four iras bands .
mir-54 is within 13 of the nominal iras position .
it is the only source in orii-2 detected at 24 and 70 as well as in the sub - millimeter .
the mips 70 flux density for mir-54 is @xmath50 jy . the iras flux densities are much higher than what we observe with mips .
this is to be expected , since the large iras beam will include emission from the hot bright rim as well .
the luminosity for the iras source , i.e. mir-54 , has therefore been severely over estimated in the past .
we can obtain a more precise estimate of the bolometric luminosity by doing a greybody fit to the mips and scuba data .
this is illustrated in figure [ fig_greyfit ] where we show a two - component greybody fit to the observed flux densities between 24 and 850 .
based on the observed size of the source in the mid - infrared and sub - mm , we have assumed @xmath51 to be 10and 13 for the warm and the cold components , respectively .
the dust temperatures ( ) , dust emissivity indices ( @xmath52 ) and masses calculated for the best fit model are presented in figure [ fig_greyfit ] .
the total dust and mass gas , @xmath53 was derived from the fitted and @xmath52 using @xmath54 where @xmath55 is the distance in @xmath45kpc@xmath46 $ ] , s@xmath56 is the total flux ( in jy ) at frequency @xmath57 and m@xmath58 is given in @xmath45@xmath46 $ ] .
based on the greybody fitting we obtain dust temperatures of 23 k and 67 k and @xmath52 of 0.9 and 0.8 for the cold and the warm components . between 24 and sub - mm wavelengths ,
the observed sed of the source is reasonably well fit using this simple two - temperature greybody model and it is consistent with the observed 1.3 mm flux density .
however we could not derive a single radiative transfer based model , which reproduces the entire sed starting from 3.6 to the sub - mm wavebands ( figure [ fig_sedfits ] ) .
the model is particularly poor at the longer wavelengths .
however , both the greybody model and the radiation transfer based models give a luminosity of 1.31.8 and mass of 0.190.24 .
mir-54 is therefore a deeply embedded ( not detected in the near - ir ) low - mass , very young object .
it excites an h@xmath33o maser @xcite .
it is most likely the exciting source for the large parsec scale outflow , hh289 .
inside the globule this is seen as a low velocity molecular outflow @xcite .
the outflow is also visible in the irac images ( section [ sec_hh992 ] ) .
the sed modeling predicts @xmath40/@xmath4 2 10@xmath59 yr@xmath19 , which satisfies the criterion ( @xmath40/@xmath41 10@xmath60 yr@xmath19 ) used by @xcite to identify stage 0/i objects . since mir-54 powers
such a large outflow , it is probably not a class 0 ( or stage 0 ) protostar , but rather a deeply embedded , somewhat more evolved class i protostar .
we find clear evidence for triggered star formation in the cometary globule orii-2 , with four young stars embedded in the front side of the globule .
the pms object furthest away from the bright rim and approximately at the center of the globule , mir-54 , is a young class i or stage i protostar .
it is most likely responsible for the large parsec scale herbig - haro outflow , with its counter flow seen as a bright h@xmath1 outflow breaking out on the western side of the globule .
however , mir-52 , the heavily obscured h@xmath1 emission line star @xmath4 20 south of mir-54 , also appears to drive an outflow , which is approximately aligned with the outflow from mir-54 . to the east both outflows overlap , and either star could therefore be the exciting star for the large hh outflow .
mir-89 , which is located in the bright rim , is a more evolved class ii object , which most likely excites the hh object , hh992 , which we discovered in this study .
hh992 lies outside the globule , @xmath4 65 east of the star and south of the large herbig - haro outflow , hh289 .
hh992 breaks up into three knots , and must be rather low excitation , since it is not seen in the h@xmath1 image .
another candidate young star , mir-104 , which is close to the globule and near mir-89 ( figure [ fig_halpha_rband ] ) may also have formed in the globule .
however , mir-104 was only detected at 5.8 and 8 and we therefore have insufficient information to determine whether it really is a pms object .
if it is , it could have formed inside the globule before the gas was ablated away by the strong uv radiation from @xmath3 ori .
we identify two infrared excess stars , one of which is far to the south , @xmath42 from the globule and the other is seen in projection against the bright rim in the tail region of the globule .
both these sources have zero or very little foreground extinction and are probably unrelated to the globule , although we can not exclude the possibility that they could have formed in the globule in an earlier epoch of star formation .
we also find two other candidate pms stars in the field , which may have weak mid - ir excesses .
all these stars appear older , i.e. they are pre - transitional or transitional disk stars , which now are part of @xmath3-orionis cluster , which has an age of @xmath4 3 myr .
models of cometary globules can successfully explain their morphology and suggest that they form stars through radiation driven implosion ( rdi ) @xcite .
this was already qualitatively discussed by @xcite , who showed that cometary globules are active sites of star formation . in these models
cometary globules are denser cores in the molecular cloud surrounding an region , which get exposed to the uv radiation from the central ob association , when the region expands into it .
the uv flux of the ob association ionizes the external layers of the core facing the ob association and an ionization front is formed at the front surface .
the gas is heated during the ionization and the temperature increases .
the increased pressure due to the hot ionized gas drives an isothermal shock into the cloud , which compresses the neutral gas in the core . at the same time
, the ionized and heated gas at the surface of the core flows radially away from the surface and forms an evaporating layer surrounding the surface of the core . due to the curvature of the front surface of the core the shock waves preceding
the ionization front converge and collide at their focus , causing the interior of core to be compressed to much higher densities .
this causes the central part of the core to become gravitationally unstable , triggering the formation of a star .
the observations of @xcite of the isolated cometary globule brc37 in the region ic1396 appears to match this model rather well .
their observations show a string of pms objects along the symmetry axis of the globule , with the oldest being in front of the globule , and the youngest object still embedded in the head of the globule , suggesting sequential formation of young stars as the ionization front proceeds towards the tail of the globule . for some of the other cometary globules in the same region , like ic1396n ( brc38 ) , this scenario is less clear .
@xcite found an elongated clustering of x - ray sources aligned with the globule , with the youngest stars still embedded inside it , consistent with the rdi model for triggered star formation .
however @xcite , who studied the same cometary globule using deep j , h , k@xmath61 , and narrowband h@xmath33 2.12 imaging and previously published mm imaging did not find any color or age gradient in the south - north direction , i.e. in the direction of the globule . they therefore conclude that not all star formation in this globule can be explained in terms of triggering .
we can think of two possible explanations for why ic1396n does not appear to show the same type of age progression and sequential triggering as seen in brc37 .
if the x - ray sources in front of the globule are late class ii objects , i.e. transitional disk like objects , or class iii objects , they will show no color excess in the near - ir and therefore appear as field stars .
secondly , ic1396n is not a well defined isolated cometary globule with a simple head / tail structure , suggesting that it could consist of more than one cometary globule .
orii-2 , the cometary globule studied in this paper , does appear more similar to brc38 .
it shows clear evidence for triggered star formation , with the youngest object , mir-54 , being furthest away from the front side of the globule .
the two other young class ii objects are definitely older and closer to the bright rim and the ionizing star , @xmath3 orionis .
all three stars are also close to the symmetry axis of the globule , as predicted by the rdi model .
mir-89 and possibly mir-104 , which both lie on the eastern side of the globule , do not appear to have formed by focussed compression along the symmetry axis of the globule . other processes , like rayleigh - taylor instabilities in the bright rim , may also play a role and form low - mass stars like mir-89 .
a similar situation is seen in the horsehead , a bright rimmed elephant trunk structure east of @xmath3 orionis , which also shows some evidence for sequential star formation triggered by the expanding region @xcite .
these young pms stars are also low - mass stars , which appear to have formed in the bright rim by rayleigh - taylor instabilities , since hardly any of them lie close to the symmetry axis of the horsehead
. several , if not all of the young stars in orii-2 drive outflows .
it is interesting to note that all of the outflows are roughly perpendicular to the symmetry axis of the globule .
it maybe a coincidence , but other cometary globules with embedded protostars , like brc37 , show a similar behavior .
if this is generally true , it would provide additional support to the focussed rdi model , which predicts that the gas inside a cometary globule will be compressed in a ridge along the symmetry axis of the globule .
when density condensations , i.e. cores in this ridge become gravitationally unstable and collapse , the collapse would preferentially be towards the symmetry axis , therefore aligning the protoplanetary disk along the symmetry axis .
since outflows are driven by disks , one would therefore see outflows which are perpendicular to the symmetry axis of the cometary globule .
orii-2 is thus a low mass star forming cometary globule .
the star formation in this region appears to be triggered by a combination of rdi and rayleigh - taylor instabilities in the rim of the globule . in the deep r - band images for the first time we detect hh 992 , which based on its non - detection in the image
is most likely an extremely low excitation hh object .
follow - up optical and infrared spectroscopic observations to understand the true nature of hh 992 as well as mir-52 , the sole emission star in the field , are needed .
we find that many of the young stars drive outflows which are aligned perpendicular to the symmetry axis of the globule .
higher spatial resolution imaging of the outflows in orii-2 in near - infrared and millimeter would provide structural details of the outflows and lead to correct identification of the sources exciting these outflows . based on observations made with the nordic optical telescope , operated on the island of la palma jointly by denmark , finland , iceland , norway , and sweden , in the spanish observatorio del roque de los muchachos of the instituto de astrofisica de canarias .
we thank bo reipurth for constructive criticism , which helped us improve the paper .
allen , l. e. , et al .
2004 , , 154 , 36 bally , j. , et al .
2009 , , 300 , 300 beltran , m. t. , massi , f. , lopez , r. , girart , j. m. , & estalella , r. 2009 , arxiv:0902.4543 bessell , m. s. , & brett , j. m. 1988 , , 100 , 1134 bok , b. j. , cordwell , c. s. , & cromwell , r. h. 1971 , in dark nebulae , globules and protostars , proceedings of a symposium held in tucson , march 26 27 , 1970 .
b. t. lynds ( tucson : university of arizona press ) p. 33
bontemps , s. 1996 , ph.d .
thesis , caballero , j. a. 1998 , , 383,750 cernicharo , j. , bachiller , r. , duvert , g. , gonzalez - alfonso , e. , & gomez - gonzalez , j. 1992 , , 261 , 589 codella , c. , palumbo , g. g. c. , pareschi , g. , scappini , f. , caselli , p. , & attolini , m. r. 1995 , , 276 , 57 cohen , j. g. , persson , s. e. , elias , j. h. , & frogel , j. a. 1981 , , 249 , 481 gutermuth , r. a. , et al .
2008 , , 674 , 336 flaherty , k. m. , pipher , j. l. , megeath , s. t. , winston , e. m. , gutermuth , r. a. , muzerolle , j. , allen , l. e. , & fazio , g. g. 2007 , , 663 , 1069 forrest , w. j. , et al .
2004 , , 154 , 443 getman , k. v. , feigelson , e. d. , garmire , g. , broos , p. , & wang , j. 2007 , , 654 , 316 hartmann , l. , megeath , s. t. , allen , l. , luhman , k. , calvet , n. , dalessio , p. , franco - hernandez , r. , & fazio , g. 2005 , , 629,881 herbig , g. h. 1974 , , 86 , 604 hester , j. j. , et al .
1996 , , 111 , 2349 hodapp , k .- w . 1994 , , 94 , 615 hogerheijde , m. r. , & sandell , g. 2000 , , 534 , 880 holland , w. s. , et al . 1999 , , 303 , 659 ikeda , h. , et al .
2008 , , 135 , 2323 indebetouw , r. , et al .
2005 , , 619 , 931 jenness , t. , & lightfoot , j.f . 1999 , starlink user note 216.6 , rutherford appleton laboratory , particle physics & astronomy research council lefloch , b. , & lazareff , b. 1994 , , 289 , 559 mader , s. l. , zealey , w. j. , parker , q. a. , & masheder , m. r. w. 1999 , , 310 , 331 martin , r. n. , & barrett , a. h. 1978 , , 36 , 1 megeath , s. t. , et al .
2004 , , 154 , 367 meyer , m. r. , calvet , n. , & hillenbrand , l. a. 1997 , , 114 , 288 miao , j. , white , g. j. , nelson , r. , thompson , m. , & morgan , l. 2006 , , 369 , 143 mookerjea , b. , sandell , g. , jarrett , t. h. , & mcmullin , j. p. 2009
, , submitted muzerolle , j. , et al .
2004 , , 154 , 379 ogura , k. , & sugitani , k. 1998 , pub .. astron .
soc . aust .
, 15 , 91 reipurth , b. , & bouchet , p. 1984 , , 137 , l1 reipurth , b. 1983 , , 117,183 robitaille , t. p. , whitney , b. a. , indebetouw , r. , & wood , k. 2007 , , 169 , 328 robitaille , t. p. , whitney , b. a. , indebetouw , r. , wood , k. , & denzmore , p. 2006
, , 167 , 256 sandell , g. , jessop , n. , & jenness , t. 2001 , `` the scuba map reduction cookbook '' , starlink cookbook 11.2 , rutherford appleton laboratory , particle physics & astronomy research council sandell , g. , et al . 1999 , , 519 , 236 sault , r.j . , teuben , p.j . , & wright , m.c.h . , 1995 , in asp conf
77 : astronomical data analysis software and systems iv , eds .
shaw , h.e .
payne , & j.j.e .
hayes ( astronomical society of the pacific : san francisco ) , p. 433
simon , j. d. , bolatto , a. d. , whitney , b. a. , robitaille , t. p. , shah , r. y. makovoz , d. , stanimirovi , s. , barb , r. h. , & rubio , m. 2007 , , 669 , 327 stern , d. , et al .
2005 , , 631 , 163 sugitani , k. , fukui , y. , mizuni , a. , & ohashi , n. 1989 , , 342 , l87 sugitani , k. , tamura , m. , & ogura , k. 1995 , , 455 , l39 tamura , m. , itoh , y. , oasa , y. , & nakajima , t. 1998 , science , 282 , 1095 walter , f.m . , sherry , w.h . ,
wolk , s.j .
, & adams , n.r .
2008 , in _ handbook of star forming regions _ , vol .
bo reipurth , astron .
, p. 732 wouterloot , j. g. a. , & walmsley , c. m. 1986 , , 168 , 237 lccrcccccccc mir-1 & 5:37:55.30 & -1:43:39.2 & 17.5 & [email protected] & 14.00@xmath5 0.03 & 13.82@xmath5 0.06 & [email protected] & 0.71 @xmath5 0.01 & 0.48 @xmath5 0.01 & 0.32 @xmath5 0.02 &
+ mir-2 & 5:37:55.41 & -1:45:22.1 & 20.3 & & & & [email protected] & 0.13 @xmath5 0.00 & & & + mir-3 & 5:37:55.43 & -1:43:16.4 & 16.3 & [email protected] & 14.15@xmath5 0.04 & 14.14@xmath5 0.06 & [email protected] & 0.44 @xmath5 0.01 & 0.30 @xmath5 0.01 & 0.17 @xmath5 0.02 & + mir-4 & 5:37:55.47 & -1:47:24.7 & 17.2 & [email protected] & 14.05@xmath5 0.04 & 13.85@xmath5 0.05 & [email protected] & 0.59 @xmath5 0.01 & 0.43 @xmath5 0.03 & & + mir-5 & 5:37:55.55 & -1:46:04.4 & 22.1 & & & & [email protected] & 0.07 @xmath5 0.00 & & & + mir-6 & 5:37:55.93 & -1:45:56.4 & 21.2 & & & & [email protected] & 0.05 @xmath5 0.00 & & & + mir-7 & 5:37:55.94 & -1:42:33.4 & 17.3 & [email protected] & 14.13@xmath5 0.04 & 13.85@xmath5 0.05 & & 0.62 @xmath5 0.01 & & & + mir-8 & 5:37:56.30 & -1:42:55.4 & 15.2 & [email protected] & 11.81@xmath5 0.02 & 11.63@xmath5 0.02 & [email protected] & 4.81 @xmath5 0.02 & 3.03 @xmath5 0.03 & 1.78 @xmath5 0.02 & + mir-9 & 5:37:56.39 & -1:45:15.2 & 23.3 & & & & [email protected] & 0.06 @xmath5 0.00 & & &
+ mir-10 & 5:37:57.04 & -1:47:40.9 & 18.2 & [email protected] & 16.18@xmath5 0.15 & 16.81@xmath5 & [email protected] & 0.08 @xmath5 0.00 & 0.10 @xmath5 0.01 & & + mir-11 & 5:37:57.28 & -1:44:10.5 & & & & & & & & 0.17 @xmath5 0.02 & + mir-12 & 5:37:57.80 & -1:44:29.5 & 14.0 & [email protected] & 12.38@xmath5 0.02 & 12.37@xmath5 0.03 & [email protected] & 2.32 @xmath5 0.01 & 1.49 @xmath5 0.02 & 0.86 @xmath5 0.02 & + mir-13 & 5:37:57.99 & -1:47:27.4 & & & & & [email protected] & 0.04 @xmath5 0.00 & & & + mir-14 & 5:37:58.01 & -1:45:23.4 & 22.2 & & & & [email protected] & 0.05 @xmath5 0.00 & & & + mir-15 & 5:37:58.12 & -1:44:40.7 & 17.9 & [email protected] & 16.27@xmath5 0.16 & 14.96@xmath5 & [email protected] & 0.07 @xmath5 0.00 & & & + mir-16 & 5:37:58.29 & -1:42:37.4 & 19.3 & & & & & 0.13 @xmath5 0.00 & & 0.29 @xmath5 0.02 &
+ mir-17 & 5:37:58.38 & -1:47:30.3 & 20.0 & [email protected] & 16.31@xmath5 0.16 & 15.50@xmath5 0.22 & [email protected] & 0.12 @xmath5 0.00 & 0.14 @xmath5 0.01 & & + mir-18 & 5:37:58.64 & -1:47:30.6 & 21.5 & & & & [email protected] & 0.10 @xmath5 0.00 & & &
+ mir-19 & 5:37:58.77 & -1:44:56.3 & 18.1 & [email protected] & 16.34@xmath5 0.17 & 15.95@xmath5 & [email protected] & 0.07 @xmath5 0.00 & & & + mir-20 & 5:37:58.78 & -1:45:50.0 & 15.6 & [email protected] & 12.40@xmath5 0.02 & 12.19@xmath5 0.02 & [email protected] & 2.70 @xmath5 0.01 & 1.95 @xmath5 0.02 & 1.06 @xmath5 0.02 & + mir-21 & 5:37:58.86 & -1:47:28.1 & 23.1 & & & & [email protected] & 0.07 @xmath5 0.00 & & &
+ mir-22 & 5:37:59.07 & -1:44:14.2 & & & & & & & & 0.14 @xmath5 0.02 & + mir-23 & 5:37:59.13 & -1:43:54.2 & 16.9 & [email protected] & 13.77@xmath5 0.04 & 13.63@xmath5 0.05 & [email protected] & 1.16 @xmath5 0.10 & 1.04 @xmath5 0.02 & 1.50 @xmath5 0.02 & + mir-24 & 5:37:59.24 & -1:46:35.5 & & & & & [email protected] & 0.03 @xmath5 0.00 & & & + mir-25 & 5:37:59.35 & -1:45:17.9 & 16.9 & [email protected] & 14.70@xmath5 0.04 & 14.57@xmath5 0.07 & [email protected] & 0.26 @xmath5 0.01 & 0.16 @xmath5 0.01 & 0.16 @xmath5 0.02 & + mir-26 & 5:37:59.36 & -1:44:32.1 & 20.5 & [email protected] & 16.44@xmath5 0.18 & 15.54@xmath5 0.22 & [email protected] & 0.11 @xmath5 0.00 & 0.09 @xmath5 0.01 & &
+ mir-27 & 5:37:59.49 & -1:45:52.2 & 19.0 & [email protected] & 15.57@xmath5 0.10 & 15.25@xmath5 0.16 & [email protected] & 0.15 @xmath5 0.00 & 0.10 @xmath5 0.01 & & + mir-28 & 5:37:59.53 & -1:46:49.2 & & & & & [email protected] & 0.08 @xmath5 0.00 & & & + mir-29 & 5:38:00.00 & -1:44:30.6 & 18.3 & [email protected] & 15.24@xmath5 0.07 & 15.23@xmath5 0.17 & [email protected] & 0.19 @xmath5 0.00 & & &
+ mir-30 & 5:38:00.04 & -1:46:22.5 & 18.6 & & & & [email protected] & 0.06 @xmath5 0.00 & & & + mir-31 & 5:38:00.44 & -1:42:59.4 & 19.5 & & & & [email protected] & 0.05 @xmath5 0.00 & & & + mir-32 & 5:38:00.90 & -1:46:44.9 & & & & & [email protected] & 0.04 @xmath5 0.00 & & & + mir-33 & 5:38:01.03 & -1:44:33.7 & & & & & [email protected] & 0.11 @xmath5 0.00 & & & + mir-34 & 5:38:01.06 & -1:47:14.1 & 16.4 & [email protected] & 14.30@xmath5 0.04 & 14.46@xmath5 0.07 & [email protected] & 0.31 @xmath5 0.00 & 0.24 @xmath5 0.01 & 0.15 @xmath5 0.02 & + mir-35 & 5:38:01.07 & -1:44:50.8 & 22.5 & & & & [email protected] & 0.07 @xmath5 0.00 & & & + mir-36 & 5:38:01.07 & -1:43:51.1 & & & & & [email protected] & 0.05 @xmath5 0.00 & & &
+ mir-37 & 5:38:01.60 & -1:44:32.5 & & & & & [email protected] & 0.04 @xmath5 0.00 & & & + mir-38 & 5:38:01.83 & -1:45:50.1 & 15.1 & [email protected] & 11.82@xmath5 0.02 & 11.62@xmath5 0.03 & [email protected] & 4.40 @xmath5 0.02 & 4.07 @xmath5 0.03 & 3.86 @xmath5 0.03 & + mir-39 & 5:38:02.10 & -1:44:10.9 & 23.2 & & & & [email protected] & 0.04 @xmath5 0.00 & & & + mir-40 & 5:38:02.24 & -1:47:30.1 & & & & & [email protected] & 0.05 @xmath5 0.00 & & & + mir-41 & 5:38:02.33 & -1:47:40.4 & 19.1 & [email protected] & 13.49@xmath5 0.03 & 13.10@xmath5 0.03 & [email protected] & 3.21 @xmath5 0.30 & 2.88 @xmath5 0.03 & 2.55 @xmath5 0.06 & 2.40 @xmath5 0.05 + mir-42 & 5:38:02.35 & -1:43:47.3 & & & & & [email protected] & 0.05 @xmath5 0.00 & & &
+ mir-43 & 5:38:02.36 & -1:44:01.8 & 22.7 & & & & [email protected] & 0.03 @xmath5 0.00 & & & + mir-44 & 5:38:02.87 & -1:42:36.7 & 17.4 & [email protected] & 12.58@xmath5 0.02 & 12.23@xmath5 0.03 & & 3.34 @xmath5 0.02 & 2.13 @xmath5 0.04 & 1.26 @xmath5 0.03 & + mir-45 & 5:38:03.42 & -1:45:15.1 & & & & & [email protected] & 0.04 @xmath5 0.00 & & & + mir-46 & 5:38:03.57 & -1:44:15.6 & & & & & [email protected] & 0.06 @xmath5 0.00 & & & + mir-47 & 5:38:03.87 & -1:45:57.5 & 17.6 & [email protected] & 15.40@xmath5 0.08 & 15.40@xmath5 0.17 & [email protected] & 0.16 @xmath5 0.00 & & & + mir-48 & 5:38:04.10 & -1:46:25.1 & 18.5 & & & & [email protected] & 0.05 @xmath5 0.00 & & & + mir-49 & 5:38:04.33 & -1:44:37.9 & & 17.27@xmath5 & 16.68@xmath5 0.26 & 15.20@xmath5 0.16 & [email protected] & 0.54 @xmath5 0.01 & 0.43 @xmath5 0.01 & & + mir-50 & 5:38:04.43 & -1:45:50.3 & & & & & [email protected] & 0.09 @xmath5 0.00 & 0.76 @xmath5 0.01 & & + mir-51 & 5:38:04.56 & -1:45:53.0 & 17.3 & [email protected] & 14.72@xmath5 0.04 & 14.64@xmath5 0.10 & [email protected] & 0.76 @xmath5 0.10 & 1.67 @xmath5 0.37 & 2.72 @xmath5 0.03 &
+ mir-52 & 5:38:04.78 & -1:45:32.2 & 19.9 & [email protected] & 12.73@xmath5 0.04 & 12.16@xmath5 0.03 & [email protected] & 5.13 @xmath5 0.12 & 4.00 @xmath5 0.38 & 2.91 @xmath5 0.03 & 11.37 @xmath5 0.05 + mir-53 & 5:38:05.10 & -1:45:41.9 & 22.8 & & & & [email protected] & 0.02 @xmath5 0.00 & & & + mir-54 & 5:38:05.15 & -1:45:12.4 & & & & & [email protected] & 4.97 @xmath5 0.04 & 15.73 @xmath5 0.06 & 13.63 @xmath5 0.06 & 231.90 @xmath5 0.44 + mir-55 & 5:38:05.19 & -1:46:18.9 & 20.7 & & & & [email protected] & 0.06 @xmath5 0.00 & & & + mir-56 & 5:38:05.40 & -1:47:05.8 & 22.5 & & & & [email protected] & 0.06 @xmath5 0.00 & & &
+ mir-57 & 5:38:05.50 & -1:42:43.1 & 21.7 & & & & & 0.10 @xmath5 0.00 & & & + mir-58 & 5:38:05.83 & -1:42:36.3 & 14.1 & [email protected] & 11.94@xmath5 0.02 & 11.84@xmath5 0.03 & [email protected] & 3.47 @xmath5 0.02 & 2.26 @xmath5 0.03 & 1.28 @xmath5 0.02 & + mir-59 & 5:38:05.86 & -1:47:04.1 & 19.1 & [email protected] & 16.26@xmath5 0.16 & 15.70@xmath5 0.23 & [email protected] & 0.08 @xmath5 0.00 & & & + mir-60 & 5:38:05.90 & -1:43:05.7 & 19.8 & & & & [email protected] & 0.05 @xmath5 0.00 & & &
+ mir-61 & 5:38:06.06 & -1:46:00.2 & 18.2 & [email protected] & 15.34@xmath5 0.07 & 15.06@xmath5 0.13 & [email protected] & 0.20 @xmath5 0.00 & & &
+ mir-62 & 5:38:06.26 & -1:43:59.6 & & & & & [email protected] & 0.04 @xmath5 0.00 & & & + mir-63 & 5:38:06.38 & -1:45:02.1 & 20.2 & [email protected] & 16.23@xmath5 0.15 & 15.40@xmath5 0.18 & [email protected] & 0.02 @xmath5 0.00 & & & + mir-64 & 5:38:06.38 & -1:46:36.4 & 16.6 & [email protected] & 13.24@xmath5 0.02 & 12.99@xmath5 0.03 & [email protected] & 1.30 @xmath5 0.01 & 0.93 @xmath5 0.02 & 0.47 @xmath5 0.02 & + mir-65 & 5:38:06.65 & -1:46:55.5 & 20.8 & & & & [email protected] & 0.07 @xmath5 0.00 & & & + mir-66 & 5:38:06.77 & -1:47:14.8 & 21.6 & & & & [email protected] & 0.05 @xmath5 0.00 & & &
+ mir-67 & 5:38:06.84 & -1:47:03.8 & & & & & [email protected] & 0.03 @xmath5 0.00 & & & + mir-68 & 5:38:06.99 & -1:43:04.0 & 20.1 & [email protected] & 16.07@xmath5 0.14 & 15.77@xmath5 0.26 & [email protected] & 0.10 @xmath5 0.00 & & & + mir-69 & 5:38:07.25 & -1:45:59.7 & 20.7 & & & & [email protected] & 0.06 @xmath5 0.00 & & & + mir-70 & 5:38:07.27 & -1:47:13.0 & 23.3 & & & & [email protected] & 0.04 @xmath5 0.00 & & & + mir-71 & 5:38:07.64 & -1:42:46.8 & 18.5 & [email protected] & 16.15@xmath5 0.14 & 15.86@xmath5 & [email protected] & 0.07 @xmath5 0.00 & & & + mir-72 & 5:38:07.65 & -1:44:17.9 & 16.2 & [email protected] & 13.01@xmath5 0.03 & 12.76@xmath5 0.04 & [email protected] & 1.78 @xmath5 0.01 & 1.25 @xmath5 0.02 & 0.65 @xmath5 0.02 & + mir-73 & 5:38:08.01 & -1:45:07.7 & s & [email protected] & 8.55@xmath5 0.03 & 8.57@xmath5 0.02 & [email protected] & 59.6 @xmath5 0.2 & 41.4 @xmath5 0.1 & 21.8 @xmath5 0.1 & 6.32 @xmath5 0.05 + mir-74 & 5:38:08.11 & -1:42:43.5 & 20.1 & [email protected] & 16.20@xmath5 0.15 & 15.47@xmath5 0.19 & [email protected] & 0.11 @xmath5 0.00 & 0.04 @xmath5 0.01 & & + mir-75 & 5:38:08.19 & -1:43:16.0 & 16.3 & [email protected] & 14.27@xmath5 0.04 & 14.08@xmath5 0.06 & [email protected] & 0.45 @xmath5 0.01 & 0.30 @xmath5 0.01 & 0.12 @xmath5 0.02 & + mir-76 & 5:38:08.68 & -1:43:01.1 & 18.2 & [email protected] & 15.16@xmath5 0.06 & 14.80@xmath5 0.10 & [email protected] & 0.22 @xmath5 0.00 & 0.08 @xmath5 0.01 & & + mir-77 & 5:38:08.88 & -1:43:28.3 & 22.6 & & & & [email protected] & 0.03 @xmath5 0.00 & & & + mir-78 & 5:38:08.92 & -1:44:48.2 & 18.2 & [email protected] & 16.11@xmath5 0.14 & 16.81@xmath5 & [email protected] & 0.07 @xmath5 0.00 & & & + mir-79 & 5:38:09.14 & -1:45:06.7 & & & & & [email protected] & 0.05 @xmath5 0.00 & & & + mir-80 & 5:38:09.40 & -1:44:20.7 & & & & & [email protected] & 0.13 @xmath5 0.00 & & &
+ mir-81 & 5:38:09.66 & -1:47:16.8 & 17.0 & [email protected] & 14.64@xmath5 0.04 & 14.41@xmath5 0.08 & [email protected] & 0.33 @xmath5 0.00 & 0.27 @xmath5 0.01 & 0.14 @xmath5 0.02 & + mir-82 & 5:38:09.69 & -1:44:51.0 & & & & & [email protected] & 0.04 @xmath5 0.00 & & & + mir-83 & 5:38:09.78 & -1:44:42.6 & 19.4 & & & & [email protected] & 0.05 @xmath5 0.00 & & & + mir-84 & 5:38:09.93 & -1:45:47.5 & 16.2 & [email protected] & 14.19@xmath5 0.03 & 14.12@xmath5 0.06 & [email protected] & 0.44 @xmath5 0.01 & 0.30 @xmath5 0.01 & 0.21 @xmath5 0.02 & + mir-85 & 5:38:09.96 & -1:45:53.0 & & & & & [email protected] & 0.02 @xmath5 0.00 & & & + mir-86 & 5:38:09.97 & -1:47:00.6 & 20.3 & & & & [email protected] & 0.07 @xmath5 0.00 & & &
+ mir-87 & 5:38:10.01 & -1:43:35.6 & 17.8 & [email protected] & 15.94@xmath5 0.13 & 15.75@xmath5 0.25 & [email protected] & 0.10 @xmath5 0.00 & & & + mir-88 & 5:38:10.02 & -1:45:38.1 & 23.5 & & & & [email protected] & 0.35 @xmath5 0.10 & 0.12 @xmath5 0.01 & 0.15 @xmath5 0.02 & + mir-89 & 5:38:10.16 & -1:45:15.6 & & & & & [email protected] & 0.58 @xmath5 0.10 & 0.73 @xmath5 0.01 & 2.91 @xmath5 0.02 & 21.40 @xmath5 0.05 + mir-90 & 5:38:10.30 & -1:44:15.4 & & & & & [email protected] & 0.06 @xmath5 0.00 & & &
+ mir-91 & 5:38:10.31 & -1:43:44.8 & & & & & [email protected] & 0.07 @xmath5 0.00 & & & + mir-92 & 5:38:10.35 & -1:45:59.4 & & & & & [email protected] & 0.02 @xmath5 0.00 & & & + mir-93 & 5:38:10.41 & -1:45:26.4 & 19.2 & [email protected] & 15.13@xmath5 0.08 & 14.62@xmath5 0.11 & [email protected] & 0.35 @xmath5 0.00 & 0.29 @xmath5 0.01 & 0.51 @xmath5 0.02 & + mir-94 & 5:38:10.52 & -1:43:13.7 & 18.6 & [email protected] & 16.03@xmath5 0.14 & 15.54@xmath5 & [email protected] & 0.08 @xmath5 0.00 & & & + mir-95 & 5:38:10.69 & -1:44:07.4 & & & & & [email protected] & 0.06 @xmath5 0.00 & & & + mir-96 & 5:38:10.70 & -1:45:03.9 & 23.7 & & & & [email protected] & 0.05 @xmath5 0.00 & & & + mir-97 & 5:38:10.79 & -1:44:11.6 & 22.9 & & & & [email protected] & 0.06 @xmath5 0.00 & & &
+ mir-98 & 5:38:10.80 & -1:44:44.5 & 19.2 & & & & [email protected] & 0.04 @xmath5 0.00 & & & + mir-99 & 5:38:10.85 & -1:46:57.6 & & & & & [email protected] & 0.07 @xmath5 0.00 & & &
+ mir-100 & 5:38:10.85 & -1:46:52.1 & & & & & [email protected] & 0.05 @xmath5 0.00 & & & + mir-101 & 5:38:10.98 & -1:44:03.1 & & & & & [email protected] & 0.06 @xmath5 0.00 & & & + mir-102 & 5:38:11.08 & -1:46:09.5 & & & & & [email protected] & 0.04 @xmath5 0.00 & & & + mir-103 & 5:38:11.21 & -1:45:35.0 & 16.9 & [email protected] & 15.39@xmath5 0.08 & 15.15@xmath5 0.15 & [email protected] & 0.17 @xmath5 0.00 & 0.07 @xmath5 0.01 & & + mir-104 & 5:38:11.66 & -1:44:56.5 & & & & & & & 0.10 @xmath5 0.01 & 0.30 @xmath5 0.02 & + mir-105 & 5:38:11.71 & -1:46:57.4 & & & & & [email protected] & 0.05 @xmath5 0.00 & & & + mir-106 & 5:38:11.71 & -1:44:48.1 & 23.1 & & & & [email protected] & 0.04 @xmath5 0.00 & & & + mir-107 & 5:38:11.81 & -1:43:01.1 & 17.6 & [email protected] & 14.01@xmath5 0.04 & 13.80@xmath5 0.05 & [email protected] & 0.74 @xmath5 0.01 & 0.45 @xmath5 0.01 & 0.25 @xmath5 0.02 & + mir-108 & 5:38:11.97 & -1:46:38.8 & & & & & [email protected] & 0.12 @xmath5 0.00 & & &
+ mir-109 & 5:38:11.99 & -1:45:44.5 & 21.2 & & & & [email protected] & 0.03 @xmath5 0.00 & & & + mir-110 & 5:38:12.01 & -1:45:06.5 & 18.2 & [email protected] & 16.46@xmath5 0.19 & 15.72@xmath5 0.23 & [email protected] & 0.05 @xmath5 0.00 & & & + mir-111 & 5:38:12.04 & -1:43:36.6 & & & & & [email protected] & 0.04 @xmath5 0.00 & & & + mir-112 & 5:38:12.16 & -1:46:56.6 & 23.7 & & & & [email protected] & 0.06 @xmath5 0.00 & & & + mir-113 & 5:38:12.44 & -1:46:53.8 & & & & & [email protected] & 0.08 @xmath5 0.00 & & & + mir-114 & 5:38:12.62 & -1:47:07.5 & 17.5 & [email protected] & 15.29@xmath5 0.08 & 14.95@xmath5 0.12 & [email protected] & 0.20 @xmath5 0.00 & 0.16 @xmath5 0.01 & 0.13 @xmath5 0.02 & + mir-115 & 5:38:13.32 & -1:47:17.5 & 17.4 & [email protected] & 15.17@xmath5 0.09 & 14.97@xmath5 0.13 & [email protected] & 0.21 @xmath5 0.00 & 0.14 @xmath5 0.01 & & + mir-116 & 5:38:13.37 & -1:42:41.9 & 13.8 & [email protected] & 12.10@xmath5 0.03 & 12.02@xmath5 0.03 & [email protected] & 3.14 @xmath5 0.01 & 1.93 @xmath5 0.02 & 1.10 @xmath5 0.02 & + mir-117 & 5:38:13.57 & -1:43:36.1 & 16.4 & [email protected] & 14.67@xmath5 0.06 & 14.98@xmath5 0.15 & [email protected] & 0.30 @xmath5 0.00 & 0.14 @xmath5 0.01 & & + mir-118 & 5:38:13.59 & -1:44:27.8 & & & & & [email protected] & 0.04 @xmath5 0.00 & & & + mir-119 & 5:38:13.69 & -1:45:15.3 & 19.7 & & & & [email protected] & 0.07 @xmath5 0.00 & & & + mir-120 & 5:38:14.08 & -1:43:50.9 & 14.2 & [email protected] & 11.63@xmath5 0.02 & 11.51@xmath5 0.03 & [email protected] & 4.90 @xmath5 0.02 & 3.21 @xmath5 0.03 & 1.84 @xmath5 0.02 &
+ mir-121 & 5:38:14.68 & -1:44:53.1 & 23.5 & & & & [email protected] & 0.07 @xmath5 0.00 & & & + mir-122 & 5:38:14.90 & -1:46:07.9 & & & & & [email protected] & 0.07 @xmath5 0.00 & & & + mir-123 & 5:38:14.92 & -1:47:04.1 & 18.4 & [email protected] & 16.52@xmath5 0.21 & 16.35@xmath5 & [email protected] & 0.06 @xmath5 0.00 & & & + mir-124 & 5:38:15.01 & -1:45:58.6 & 20.4 & & & & [email protected] & 0.10 @xmath5 0.00 & & & + mir-125 & 5:38:15.45 & -1:45:56.1 & 20.4 & & & & [email protected] & 0.09 @xmath5 0.00 & & &
+ + crrrrrrrr mir-23 & 3110@xmath518 & [email protected] & 1.7(-4)@xmath55.3(-5 ) & 2.9(-6)@xmath59.2(-7 ) & 0.0 & 1.9(-11)@xmath56.2(-13 ) & [email protected] & [email protected] + mir-41 & 3750@xmath5184 & [email protected] & 5.0(-3)@xmath53(-3 ) & 2.0(-6)@xmath51.8(-6 ) & 0.0 & 1.4(-8)@xmath57.1(-9 ) & [email protected] & [email protected] + mir-51 & 4500 & 1.56 & 2.0(-4 ) & 3.3(-4 ) & 1.9(-7 ) & 8.2(-10 ) & 6.5 & 1.9 + mir-52 & 2980 & 0.14 & 5.2(-5 ) & 9.7(-4 ) & 1.0(-7 ) & 3.1(-11 ) & 0.24 & 4.8 + mir-54 & 3380 & 0.30 & 2.5(-3 ) & 1.9(-1 ) & 5.9(-6 ) & 1.1(-8 ) & 1.36&12 + mir-89 & 3370 & 0.24 & 4.5(-3 ) & 2.1(-9 ) & 0.0 & 3.3(-8 ) & 0.3 & 50 + | we investigate the young stellar population in and near the cometary globule orii-2 .
the analysis is based on deep nordic optical telescope @xmath0-band and h@xmath1 images , jcmt scuba 450 and 850 images combined with near - infrared 2mass photometry and mid - infrared archival _
spitzer _
images obtained with the irac ( 3.6 , 4.5 , 5.8 and 8 ) , and mips ( 24 and 70 ) instruments .
we identify a total of 125 sources within the 5@xmath25 region imaged by irac . of these sources 87
are detected in the @xmath0-band image and 51 are detected in the 2mass survey .
the detailed physical properties of the sources are explored using a combination of near / mid - infrared color - color diagrams , greybody fitting of seds and an online sed fitting tool that uses a library of 2d radiation transfer based accretion models of young stellar objects with disks .
orii-2 shows clear evidence of triggered star formation with four young low luminosity pre - main sequence stars embedded in the globule .
at least two , possibly as many as four , additional low - mass pms objects , were discovered in the field which are probably part of the young @xmath3 orionis cluster . among the pms stars which have formed in the globule
, mir-54 is a young , deeply embedded class 0/i object , mir-51 and 52 are young class ii sources , while mir-89 is a more evolved , heavily extincted class ii object with its apparent colors mimicking a class 0/i object .
the class 0/i object mir-54 coincides with a previously known iras source and is a strong sub - millimeter source .
it is most likely the source for the molecular outflow and the large parsec scale herbig - haro flow .
however the nearby class ii source , mir-52 , which is strong a h@xmath1 emission line star , also appears to drive an outflow approximately aligned with the outflow from mir-54 , and because of the proximity of the two outflows , either star could contribute .
mir-89 appears to excite a low excitation hh object , hh 992 , discovered for the first time in this study . |
xmm - newton ( jansen 2001 ) , with the huge collecting area of its mirrors and the high quantum efficiency of its epic detectors ( strder 2001 , turner 2001 ) , is the most sensitive x - ray observatory ever flown .
this is strikingly evident during slew exposures which , while yielding only at most around 15 seconds of on - source observing time , actually constitute a hard - band 2 - 10 kev survey with a limiting flux sensitivity of @xmath2erg @xmath3 s@xmath4 .
this is 5@xmath010 times deeper than all other all - sky surveys .
the soft - band 0.2 - 2 kev survey ( limiting flux sensitivity @xmath5erg @xmath3 s@xmath4 ) is comparable with the rosat pspc all - sky survey ( rass ) , offering long - term variability studies .
with the release of the first source catalogue ( xmmsl1 , released in may 2006 ) into the xmm x - ray science archive - ` http://xmm.vilspa.esa.es/ ` ` external / xmm_data_acc / xsa / index.shtml ` , we are able here to show selected highlights seen so far .
this will allow the community to see immediately what science is possible using the slew survey data and catalogue .
general results ( catalogue properties , source population properties etc . ) from the xmm - newton slew survey first catalogue ( xmmsl1 ) are described in full , along with a description of how the analysis of the slew data was performed and how the catalogue was constructed , in saxton ( in preparation ) . in the following subsections we present a concise report of the first major science results from the xmm - newton slew survey :
all of the catalogued detected slew sources were cross - correlated with several catalogues and databases , including simbad , ned , heasarc , rass etc .
counterparts were found for over 50% of the sources , and these counterparts can be found in the xmmsl1 slew catalogue .
17% of the non - extended slew sources are correlated ( within 30 ) with sources in the sloan digital sky survey ( sdss ) dr5 imaging catalogue .
this is 86% of the possible counterparts , given the isotropic distribution of the slew sources and the area of the sdss dr5 area ( 8000 sq.degrees ) @xmath0 i.e. most of the slew sources within the sdss area have sdss matches .
very large numbers ( several hundred ) of agn and galaxies were found in the slew survey , the most distant being a qso ( j0646 + 4451 ) ( or hb89 0642 + 449 ) at a redshift of z=3.4 .
its count rate of 1.2ct s@xmath4 corresponds at this redshift to a luminosity of 8.5@xmath6erg s@xmath4 ( assuming a @xmath7 power law and @xmath8=75 km s@xmath4 mpc@xmath4 ) . over 150 of the detected sources
are detected both in the soft and the hard band , and we are therefore able to construct hardness ratios for these sources . plotting the spectral hardness against the galactic latitude , we see that the very hardest sources all lie at a galactic latitude of @xmath10 , i.e. they lie in the galactic plane . of these sources , the ones that we have been able to identify are all lmxrbs .
conversely , hardly any soft sources at all are seen in the galactic plane , and of the few that are , none are extragalactic .
this is as expected due to galactic absorption , but is very encouraging to observe .
a number of extremely bright objects have been detected in the slew survey , including several lmxrbs close to the galactic centre .
three of the brightest detected are 4u 1758 - 25 , sgr x-3 & v2216 oph , all known rosat sources , with epic - pn count - rates of between 330 and 530ct s@xmath4 .
another extremely bright source detected in the slew is a potential black hole candidate , previously seen not by rosat , but by xte @xmath0 xte_j1746 - 319 .
the brightest non - galactic sources detected are in the lmc @xmath0 lmc x-2 and lmc x-3 .
note that these extremely bright sources are severely affected by pile - up , and their true count - rates would be much higher in optimum ( smaller frame - time ) observing conditions .
this fact , along with the as yet unsolved problems associated with sources moving swiftly across the detector , and the associated response matrix and effective area / psf issues make quantitative spectral analysis extremely difficult at present .
however , spectra can be extracted from the brightest sources , and rudimentary spectral analyses give satisfactory results . the spectrum extracted from v2216 oph ( 4u 1728 - 16 ) for example ( fig.[fig_spec ] ) appears wholly consistent with the source being an lmxrb , as is believed .
sec of exposure time ) of v2216 oph ( 4u 1728 - 16 ) , a believed lmxrb source towards the galactic centre .
an absorbed high@xmath9 bremsstrahlung model has been fitted to the data , and the data@xmath0model residuals are shown in the lower panel.,width=226 ] as the hard - band 2@xmath010kev slew survey is 5@xmath010 times deeper than all other previous and current all - sky surveys , it is no surprise that a large number of sources are detected in this band . comparing our results with those of the heao-1 a2 all sky survey ( piccinotti 1982 )
, we see that whereas they detect 85 sources in the high latitude sky ( excluding the smc and the lmc ) in an area totalling 65% of the sky , we detect 148 hard - band sources in similar sky regions ( though covering only 8% of the sky ) . in the hard x - ray band , therefore , the xmm - newton slew survey sees approximately 14 times more sources per unit area than the heao-1 a2 all sky survey , and this will give very useful input to the 2@xmath010 kev log - n / log - s curve in the few times 10@xmath10 to the few times 10@xmath11erg @xmath3 s@xmath4 range .
issues of eddington bias and completeness will need a careful treatment though ( see saxton ( in preparation ) ) .
some of the xmm - newton slews analysed to date are seen to recross areas of the sky covered by previous slews .
as such the possibility exists to go to fainter fluxes in certain areas of the sky , as more and more slews pass over the same area . also , variability studies within the slew survey are possible .
105 of the sources within the xmmsl1 catalogue have been seen in more than one slew , including the black hole candidate xte_j1746 - 319 @xmath0 in this particular case , no evidence for variability in flux or spectrum is seen .
conversely , a source detected in three separate slews associated with the galaxy 2masxj06100652 - 6243125 , shows evidence for a factor @xmath12 variation in flux in only 10days .
cross - correlating the xmm - newton slew catalogue with the rosat all - sky - survey catalogue ( rass ; voges 1999 ) reveals that approximately 50% of the non - extended slew sources ( with detection likelihood @xmath12 ) have rass counterparts within 30 - 0.1em .
many of the remaining sources are either x - ray hard ( and perhaps invisible to rass ) , are variable , or are in the small portions of the sky not covered by the rass .
12kev ) xmm - newton slew x - ray emission ( exposure approximately 10sec ) superimposed on a dss image of the galaxy cluster abell 2029 .
( bottom ) a radial profile of the xmm - newton slew x - ray emission from abell 2029 , compared with that from a bright point source.,title="fig:",width=321 ] 12kev ) xmm - newton slew x - ray emission ( exposure approximately 10sec ) superimposed on a dss image of the galaxy cluster abell 2029 .
( bottom ) a radial profile of the xmm - newton slew x - ray emission from abell 2029 , compared with that from a bright point source.,title="fig:",width=321 ] approximately 1% of the sources with rass counterparts show a rass - to - slew flux variability greater than a factor of 10 .
many of these are perhaps expected .
the variable star uy vir , for instance , shows a rass - to - slew flux increase by a factor of @xmath1325 , whereas the cataclysmic variable qq vul shows a decrease by about the same factor . of great interest
though , are the sources showing extreme variability where no variability was expected . a slew source coincident with the otherwise innocuous , small , elliptical , non
- active galaxy ngc3599 is seen to have increased in flux by about a factor of almost 200 since the rass .
the fact that this source is very soft allows us to consider such a huge variability in terms of flaring radiation emitted due to the tidal disruption of a star by a supermassive quiescent black hole at the centre of the galaxy .
this source and other similar high - variability , possible tidal disruption events , are discussed fully in esquej ( in preparation ) .
note that it is only when a large fraction of the sky is observed , as is the case here , that rare events such as these have a chance of being observed .
approximately 15% of the catalogued sources show some significant extension beyond the psf , and a number of these have been identified with all manner of extended sources including supernova remnants , galaxies and galaxy groups and clusters : the cross - correlations reveal that the vast majority of objects that are seen ( for which we have an object type ) are galaxies , and agn make up approximately half of the 900 or so cross - correlations with known galaxies .
several ngc and markarian galaxies are seen , including a number of starburst galaxies ( notably m83 ) and interacting galaxies ( an example being ngc4748 ) .
a large number of seyferts ( 1 and 2 ) , bllacs and qsos ( extending out to a redshift of @xmath14 ) are also detected .
of the catalogued sources that show some significant extension , many are associated with known groups and clusters .
a total of 81 abell and zwicky clusters have apparent slew counterparts , and a great many of these are seen to have large extension values .
indeed these abell / zwicky sources make up a large fraction ( approximately half ) of all the slew sources which have large values of estimated extent . in a number of the brighter cases , the size and extent of the cluster can be mapped , both on the sky and as a radial profile ( see fig.[fig_cluster ] ) .
sources associated with galaxy groups have also been detected , two examples being the ngc1566 and ngc2300 groups .
the xmm - newton slew survey is seen to be very sensitive to relatively faint clusters and extended sources in general thanks to the wide band - pass , the tight psf and the low background . at present , the estimated average cluster sensitivity limit for the slew survey is just over @xmath15erg s@xmath4 @xmath3 ( at 0.2@xmath02kev ) , calculated for a @xmath16@xmath3 , @xmath17kev cluster .
this is comparable with that of the macs survey ( ebeling 2001 ) , though the sky area covered ( @xmath115% of the sky ) is much smaller than that of macs .
as the slew survey progresses however , the area of sky covered will increase to perhaps that of macs and beyond , where more than 100 high redshift ( @xmath18 ) x - ray luminous clusters should be observable . a great many sources that were initially formally detected as individual sources were seen , once careful cross - correlations were performed , to be instead due to large galactic extended sources , i.e.supernova remnants ( snr ) . for example , a great number of these spurious point sources were seen associated with puppis a. this was not merely confined to our own galaxy , as many other sources were seen associated with other snr in the lmc ( such as n132d ) , and in the smc . as a large percentage of the whole sky has currently been covered by the xmm - newton slew ( the released catalogue covers @xmath16300 sq.degrees ) , and certain areas have been covered more than once , then we would expect certain sky areas to have quite a large coverage .
given the orbit of xmm - newton , the slewing strategy is such that a great many of the slews pass close to ( though not directly over ) the north and south ecliptic poles .
one is able to combine many slew images together to obtain large - scale maps , and an example of this is shown in fig.[fig_vela ] , where a rosat pspc image of the vela snr is shown compared with the current xmm - newton slew large - scale image of the same field .
vela , together with puppis a , the bright , extended ( @xmath131@xmath19 diameter ) snr at the north - western edge of vela , lie fairly close to the region where a large number of slews are seen to cross .
the image shows that only a few seconds of xmm - newton slew time are necessary to observe the features that are seen in the rosat image .
as the mission progresses more slews will become available to add to this picture .
in addition to this , xmm - newton slew data exists not only in the ( rosat - like ) soft x - ray band ( 0.2@xmath02kev ) , but also in the hard x - ray ( 2@xmath012kev ) band . using this hardness information , we see that puppis a appears somewhat harder than the vast majority of vela , though a small number of harder features are also seen within the main vela remnant .
the potential therefore for xmm - newton being able to map out large areas of the x - ray sky via slews such as these is excellent .
xmm - newton slew exposures yield at most only 15 seconds of on - source observing time .
the high quantum efficiency of its epic - pn detector plus the huge collecting area of its mirrors however , result in an xmm - newton slew hard - band ( 2@xmath010kev ) survey ten times deeper than all other all - sky surveys , and a soft - band ( 0.2@xmath02kev ) survey comparable with the rosat pspc all - sky survey . with the release of the xmm - newton slew survey first source catalogue ( xmmsl1 ) and data files into the xmm x - ray science archive ,
we have been able here to show the first science results from the survey .
these include : * identifications of many sources , including a large number of galaxies and agn , with the current most distant xmm - slew object ( a qso ) lying at a redshift of @xmath14 . * several extremely bright sources , some not seen by rosat , both in our galaxy , especially close to the galactic centre , and beyond . * a large number of hard ( 2@xmath010kev ) sources @xmath0 the xmm - newton slew survey is the deepest ever hard - band all - sky survey , and sees @xmath1314 times more sources per unit area than the heao-1 a2 all - sky survey . *
sources detected in more than one slew , some showing variability .
* extremely variable sources , when compared with the rass , some possibly due to the tidal disruption of stars by central supermassive black holes ( esquej et al .
, in preparation ) .
* several detections of extended sources , including galaxies , galaxy groups and especially galaxy clusters , plus also more nearby supernova remnants in our galaxy and in the magellanic clouds . * the emerging excellent ability of xmm - newton to map large areas of the x - ray sky , by co - adding several slews together .
the example of the vela snr ( plus puppis a ) indicates that a few seconds of xmm - newton time are all that are needed to reveal the x - ray features observed with rosat . | we have attempted to analyse all the available data taken by xmm - newton as it slews between targets .
this slew survey , the resultant source catalogue and the analysis procedures used are described in an accompanying paper . in this letter
we present the initial science results from the survey . to date , detailed source - searching has been performed in three x - ray bands ( soft , hard and total ) in the epic - pn 0.2@xmath012kev band over @xmath16300 sq.degrees ( @xmath115% of the sky ) , and of order 4000 x - ray sources have been detected ( @xmath155% of which have ids ) .
a great variety of sources are seen , including agn , galaxies , clusters and groups , active stars , snrs , low- and high - mass xrbs and white dwarfs .
in particular , as this survey constitutes the deepest ever hard - band 2@xmath012kev all - sky survey , a large number of hard sources are detected .
furthermore , the great sensitivity and low - background of the epic - pn camera are especially suited to emission from extended sources , and interesting spatial structure is observed in many supernova remnants and clusters of galaxies .
the instrument is very adept at mapping large areas of the x - ray sky .
also , as the slew survey is well matched to the rosat all - sky survey , long - term variability studies are possible , and a number of extremely variable x - ray sources , some possibly due to the tidal disruption of stars by central supermassive black holes , have been discovered . |
the multi - particle analogon of the operator @xmath0 is often used to describe relativistic effects in atoms and molecules , see , e.g. , @xcite . here
@xmath1 is the free dirac operator @xmath2 and @xmath3 with @xmath4 is the projection operator on the electron subspace of the free dirac operator in momentum space .
the underlying hilbert space of @xmath5 is the space of square integrable four spinors @xmath6 with @xmath7 .
the operator @xmath5 was introduced for the two particle case by brown and ravenhall @xcite ( see also @xcite ) to cure the continuum dissolution of the two particle dirac operator with coulomb interaction : since the dirac operator is unbounded form below the two particle dirac operator has the whole real line as spectrum , the eigenvalues of the one particle operator `` dissolve '' .
@xmath5 is called no - pair operator or in the description of meson models ( usually with different potentials ) `` reduced salpeter operator '' @xcite . recently ,
its multi particle analogon was used in connection with the `` stability of relativistic matter''question @xcite .
similar to the dirac operator with coulomb potential , the no - pair operator has a critical nuclear charge @xmath8 .
hardekopf and sucher @xcite observed that @xmath9 $ ] with the fine structure constant @xmath10 and investigated the ground state energy of @xmath5 numerically .
they claimed that , as for the dirac equation , the ground state energy vanishes for @xmath11 @xcite .
evans et al @xcite proved that the energy @xmath12 is bounded from below by @xmath13 , if the nuclear charge does not exceed the critical charge @xmath14 , otherwise it is unbounded .
tix @xcite improved this bound to @xmath15 .
this last result shows the difficulties to obtain accurate numerically results near the critical coupling parameter and the need for sharp bounds . in this short note
we explore the results from @xcite by using some rather trivial numerics to obtain a reliable lower bound for the ground state energy for all @xmath16 .
a short summary of the method to obtain a lower bound for @xmath17 given in @xcite is provided .
any normalized four spinor in the electron subspace of @xmath1 can in be written in momentum space as @xmath18 with the normalized pauli spinor @xmath19 , the pauli - matrices @xmath20 and @xmath21 , @xmath22^{1/2}$ ] . in @xcite it was shown that the minimizer of @xmath23 is radially symmetric , i.e. , the minimizing @xmath19 is of the form @xmath24 where @xmath25 is normalized and positive .
this yields for the energy @xmath23 @xmath26 with @xmath27 , @xmath28 , @xmath29 and @xmath30 the functions @xmath31 are legendre functions of the second kind ( see @xcite for the notation ) and appear here for the same reason as in the treatment of the non
relativistic hydrogen atom in momentum space ( * ? ? ? * problem 77 ) and @xmath32 .
because of scaling we have assumed @xmath33 in . using the schwarz inequality to introduce trial functions @xmath34 in the potential energy part of @xmath23 @xmath35 is obtained . note that the operator @xmath5 is similar to the herbst operator @xmath36 .
the ground state energy of @xmath37 is obtained by minimizing the quadratic form with @xmath38 , without the square roots and with @xmath39 instead of @xmath40 .
since no error is introduced by using the schwarz inequality when @xmath41 is proportional to the ground state wave function of , this suggests a similar trial function than the one that was used by raynal et al @xcite for the herbst operator .
they obtained excellent upper and lower bounds for the ground state of @xmath37 .
choosing the trial functions @xmath42 essentially as the fourier transform of @xmath43 , the lower bound @xmath44(p).\ ] ] was obtained in @xcite where @xmath45(p)=e(p)-\frac{\alpha z}{2 } \frac{\sqrt{p^2+\mu^2}}{\beta } \frac{\sin [ \beta\arctan ( p/\mu ) ] } { \sin[(\beta+1 ) \arctan(p/\mu ) ] } \left(1+\frac{1}{e(p)}\right ) \\ - \frac{\alpha z}{2}\left(1 - \frac{1}{e(p ) } \right ) \frac{\sqrt{p^2+\mu^2}}{\beta+2}\ , \frac{p^{-3/2}_{\beta-1/2 } \left(\mu ( p^2+\mu^2)^{-1/2 } \right)}{p^{-3/2 } _ { \beta+1/2 } \left(\mu ( p^2+\mu^2)^{-1/2 } \right ) } \end{gathered}\ ] ] with the associated legendre functions @xmath46 ( see @xcite ) . to cancel the large momentum terms in @xmath47
is chosen to fulfill @xmath48 the resulting function @xmath49 $ ] is numerically evaluated .
the values for the parameters @xmath50 and @xmath51 for some values of @xmath52 can be found in table [ tab1 ] and a plot of the energy over the coupling constant is shown in figure [ fig1 ] .
.lower bound for the energy @xmath53(p)$ ] and the corresponding parameters [ cols="<,<,<,<",options="header " , ] of the no - pair hamiltonian , width=302 ]
this work was partially supported by the european union under grant erb4001gt950214 and under the tmr - network grant fmrx - ct 96 - 0001 .
irene a. stegun .
legendre functions . in milton abramowitz and irene a. stegun , editors ,
_ handbook of mathematical functions with formulas , graphs , and mathematical tables _ , chapter 8 , pages 331353 .
dover publications , new york , 1965 . | a lower bound for the ground state energy of a one particle relativistic hamiltonian - sometimes called no - pair operator - is provided . |
even if we do not notice it , our life on the internet is influenced by recommendations .
popular web sites such as amazon , netflix and youtube attempt to facilitate our navigation by suggesting us new possibly relevant items and thus increase our satisfaction and their profits @xcite . employed recommendation algorithms range from simple variants of `` buyers who choose item a also choose item b '' @xcite in amazon to more sophisticated techniques such as the singular value decomposition @xcite . even though many users still act independently of any automated assistance , the use of recommendation is on a rise .
for example , the dvd rental company netflix estimated that 75@xmath0 of the rental choices of their users come from some form of recommendation @xcite .
the rationale behind recommendation is to match the right customers with the right products .
this task is particularly important and difficult for less popular items for which user patterns can not be easily identified .
correct matching of little popular items is crucial for e - commerce studies have shown that from 20@xmath0 to 40@xmath0 of amazon s sales do not come from the best selling items @xcite .
it has been suggested that if one ranks items according to their sales and thus constructs a so - called popularity - rank curve ( see an example in fig . [
fig : introduction]a ) , there is a long tail which comprises a large number of niche items @xcite .
these niche items enjoy a higher profit margin compared to the small profit margin determined by a more competitive market of popular items , and can even boost the sales of other items by providing a convenient one - stop outlet to users @xcite . in this respect ,
recommendation algorithms seem to be the best candidate to explore the profit that hides in the long tail .
recommendation working as intended thus contributes to : ( 1 ) increasing the diversity of recommended items , ( 2 ) distributing user attention more evenly among the items . in this case , recommendation would cause the long tail in fig .
[ fig : introduction]a to gain more weight with time .
however , figs .
[ fig : introduction]b - d demonstrate that the opposite is found in reality . despite recommendation algorithms implemented at netflix , amazon , and movielens ( a web site for movie recommendation ) , the tail of the popularity distribution becomes shorter with time ( see also the evolution of the distribution in fig .
simultaneously , the most popular items account for an increasing share of total sales .
it is an adverse effect if some items become too dominant , similar to the emergence of over - dominant species in an ecosystem , which leads to the reduction of biodiversity and ultimately may lead the loss of the equilibrium in the system .
evaluation of recommendation exclusively on the basis of accuracy - oriented metrics can not , by its nature , capture and explain this long - term behavior .
when recommendation is iterated for a small number of rounds as in @xcite , the possible feedback between user choice and the recommender system is detected .
these interactions between users and the system impact on its macroscopic properties and trend , similar to other physical systems .
a more physical approach is thus needed to gain the first insights into the puzzle posed by fig .
[ fig : introduction ] . to understand the long - term impact of recommender systems
, one has to study the co - evolution of the recommendation and the online user - item network .
the recommendation results affect the growth of the network and the change of the network meanwhile influence the future recommendation outcome .
the effect is amplified with successive recommendations . in this paper
, we investigate this issue by repeatedly applying recommendation on the user - item network .
our focus is different from currently known recommendation studies which aim at short - term metrics such as accuracy @xcite , diversity @xcite and others @xcite .
we demonstrate that the repeated use of usual recommendation algorithms makes the system reach a stationary state where user attention is concentrated on a few items instead of distributed over a broad range of items which is in agreement with the data presented in fig .
[ fig : introduction ] . in other words ,
usual recommendation algorithms ultimately narrow user choice and reduce information horizons instead of widening them .
we also observe a hysteresis phenomenon which implies that while recommendation naturally gives rise to hugely popular items , to revert this change is not possible because the uneven distribution of user attention is robust over a broad range of the recommendation algorithm s parameters .
our observations directly challenge the role of recommendation for online retailers , as well as other applications of recommendation in search engines @xcite , online social networks @xcite , news media @xcite , and even suggestions of research papers @xcite .
we finally show that before some items become too dominant , it is possible to make a compromise between recommendation accuracy and long - term effects of recommendation .
these results can provide insights and motivations for the design of a next generation of recommender systems .
* data . * the data used for the empirical study in fig . 1 and fig .
s1 is described in the supplementary information . in the following simulation ,
we use two benchmark data sets : movielens ( an online movie rating and recommendation service ) and netflix ( a dvd rental service ) .
the movielens data ( available at _ www.grouplens.org_ ) contains @xmath1 movies and @xmath2 users who have rated movies using the integer scale ranging from 1 ( worst ) to 5 ( best ) . to obtain an unweighted bipartite network ,
we represent any rating of 3 or more as a link between the respective user and item .
the resulting network contains 82520 links .
the average degree of users and items is @xmath3 and @xmath4 , respectively .
the maximum degree of users and items is 509 and 558 , respectively .
the netflix data is a subset of the original data set released for the purpose of the netflixprize ( available at _ www.netflixprize.com _
the subset contains 1014 users and 2049 movies chosen from the original data at random and all links among them ( since the input data uses the same rating scale as movielens , the threshold rating of @xmath5 is again used to determine whether a link is present or not ) .
the resulting network contains 54093 links .
the average degree of users and items is @xmath6 and @xmath7 , respectively .
the maximum degree of users and items is 409 and 604 , respectively . * models and metrics . * to model the co - evolution of user choices and the recommendations generated from the recommender systems , we construct a model of recommendation ecosystem as follows .
the real data described above are used as the initial configuration .
the network evolves through a so - called rewiring process where each link is assigned a time stamp ( the initial time stamps are chosen at random ) . in each rewiring step , the oldest link of every user is redirected to a new item and assigned with the current time ( i.e. , it becomes the user s newest link ) .
deleting the oldest links simulates the case where the recommendation results are generated based on recent historical record @xcite .
we assume that when a new target item for a currently rewired link is being chosen , the user follows recommendation with probability @xmath8 and decides independently of recommendation with the complementary probability @xmath9 .
we use the widespread item - based collaborative filtering ( icf ) as the recommendation algorithm ( see details below ) ; variants of this algorithm are employed by amazon and other major web sites @xcite . when a user follows recommendation , they select an item from their current recommendation list with probability inversely proportional to item rank in the list ( the motivation to use rank - reciprocal rather than equal probability for all listed items comes from @xcite )
. when a user acts independently of recommendation , they either choose an item at random ( which we refer to as random attachment , ra ) or they choose proportionally to item degree increased by one ( which we refer to as preferential attachment , pa ; item degree is incremented by one to make it possible for items which lose all their links to be chosen again ) .
as the network evolves , degree of each user is preserved .
on the other hand , network structure and item degree values change significantly by rewiring .
network structure can be at any moment represented by a so - called network adjacency matrix @xmath10 whose element @xmath11 is one when user @xmath12 is currently connected with item @xmath13 and zero otherwise .
we use the gini coefficient @xmath14 to measure inequality of the item popularity distribution during the network evolution .
while this quantity has been originally proposed to quantify inequality of income or wealth distribution @xcite , it has also been used in other fields @xcite .
the gini coefficient can be computed as @xmath15 where @xmath16 , the popularity of objects , has been sorted in the ascending order .
the two extreme values of the gini coefficient are @xmath17 and @xmath18 which correspond to equal popularity of all items and zero popularity of all items but one , respectively .
an increase of the gini coefficient thus corresponds to the item popularity distribution becoming more unequal .
we also used other measures , such as the link share of 1@xmath0 most popular items and the herfindahl index , to study the rewiring model .
results obtained with different inequality metrics are consistent with those obtained with the gini coefficient
. * recommendation .
* we use the widely spread item - based collaborative filtering ( icf ) as the main recommendation algorithm @xcite . in icf , the recommendation score of an item is computed based on the item s similarity with other items collected by a target user .
the score of item @xmath13 for user @xmath12 reads @xmath19 where @xmath20 is the similarity of items @xmath13 and @xmath21 , and @xmath22 are elements of the network s adjacency matrix .
we choose the following formula for item similarity @xmath23 where @xmath24 denotes the set of users who have collected item @xmath13 and @xmath16 denotes the degree of item @xmath13 .
parameter @xmath25 can be continuously adjusted in the range @xmath26 $ ] which includes three classical cases : the common neighbor similarity which simply counts the number of users who have collected both items ( when @xmath27 ) , the cosine similarity which is sometimes referred to as the salton index ( when @xmath28 ) and the leicht - holme - newman similarity ( when @xmath29 ) @xcite . due to a lack of normalization , high - degree items are favored when @xmath27 . by contrast
, low - degree items are favored when @xmath30 .
equation ( [ simil ] ) thus gives us the opportunity to gradually move from recommendations biased towards high - degree items ( when @xmath27 ) to diversity - favoring recommendations biased towards low - degree items ( when @xmath29 ) . to obtain a recommendation list for a given user ,
all items that are currently not connected with this user are sorted according to their recommendation score in a descending order and finally the top @xmath31 items are kept on the recommendation list .
we use @xmath32 here which is a common value in previous studies of recommender systems @xcite . considering only the top @xmath31 items
is motivated by the fact that users in real online systems do not have time to inspect the list of all items ranked by their recommendation score ( as further reflected in the rank - reciprocal probability in our link rewiring process ) .
* distribution of item popularity . * after a sufficient number of iteration steps in our model , the system reaches a stable state and the item degree distribution becomes stationary .
we begin our analysis by comparing the distribution of item degree in the original data with the stationary outcome of the rewiring procedure .
we assume here for simplicity that the users choose new items solely by recommendation ( i.e. @xmath33 ) .
our implementation of item - based collaborative filtering employs a user similarity metric with one parameter given by eq .
( [ simil ] ) .
three particular cases of our user similarity lead to well - known similarity measures : @xmath34 correspond to common neighbor similarity ( cn ) , cosine similarity ( cos ) , and leicht - holme - newman similarity ( lhn ) , respectively .
the present parameterization allows us to continuously tune between recommendation that favors high - degree ( when @xmath25 is small ) and low - degree ( when @xmath25 is big ) items .
this is well demonstrated by fig .
[ fig : deg_distr ] which shows the item degrees in the original data compared to those after the rewiring procedures using cn and lhn .
icf with cn ( cn - icf ) improves the popularity of the most popular items and essentially eliminates the long tail .
this inferior outcome produced by an otherwise well accepted and popular recommendation method demonstrates the potential danger of recommendation for information diversity , similar to the loss of biodiversity in an ecosystem . on the other hand
, lhn - icf strengthens the long tail but , as we shall see later , its recommendation accuracy is low . to obtain a quantitative comparison
, we measure the gini coefficient over the item degrees .
the higher the values , the more uneven the distributions , and the greater the loss of information diversity in the system . the gini coefficient corresponding to the distributions shown in fig .
[ fig : deg_distr ] are @xmath35 , @xmath36 , @xmath37 ( movielens ) and @xmath38 , @xmath39 , @xmath40 ( netflix ) for the lhn - icf , original data , and cn - icf , respectively .
the lhn - icf method leads to a remarkable equalization of the item popularity , while the cn - icf method further advances the degree heterogeneity .
here we use @xmath32 .
we show in fig .
s2 for the results when other @xmath31 values are used .
moreover , we study in fig .
s3 the case where users are influenced by the similarity constraint when selecting items from the recommendation list .
we want in this way to prevent all information from being washed away .
* user reliance on recommendation . * in practice , users do not always follow recommendations .
we thus consider the case with @xmath41 , i.e. users follow recommendation with a probability @xmath8 and otherwise ( with a probability @xmath9 ) choose an item according to preferential attachment ( as shown in fig .
s4 and s5 , the resulting behavior is similar if preferential attachment is replaced by random or real choice of items ) .
the value of @xmath8 thus characterizes users reliance on recommendation , and parameter @xmath25 controls the popularity bias of recommendations . while users recommendation lists are populated mostly with popular items when @xmath25 is large , the number of niche ( little popular ) items increases when @xmath25 is small .
we repeatedly rewire the input data and again quantify the diversity of the system by the stationary value of the gini coefficient , @xmath42 . as shown in fig .
[ fig : stationaryg]a - b for various @xmath8 values , @xmath42 is not sensitive to @xmath25 over a broad range @xmath43 $ ] where @xmath44 .
once @xmath45 , @xmath42 decreases quickly and eventually reaches values lower than those produced by preferential attachment only .
another important finding out of our expectations is that , recommendation can hurt information diversity even more than preferential attachment ( i.e. , icf can lead to higher @xmath42 than pa ) .
this is because pa is not personalized and biased on popularity , but any items may be chosen ; recommendation algorithms are personalized , biased on popularity and only the top @xmath31 items identified by the algorithms are recommended .
if the top @xmath31 items for different users significantly overlap , they will attract a large amount of links .
this strong tendency of recommendation to decrease information diversity is a joint outcome of a popularity - favoring recommendation method and the fact that the users are recommended with a short list of items which further contributes to the winner takes it all " situation .
we also consider a different recommendation algorithm @xcite which show remarkable similarity with the results for icf presented in this paper ( see results in fig .
* the impact of data density . * the high density of these two data sets enables us to adjust the density of the network by removing some links . in fig .
[ fig : stationaryg]c - d , we study the effect of data density on the resulting popularity inequality when @xmath33 ( see fig .
s7 for the results with @xmath41 ) .
we randomly remove links from the original network until the desired density is reached .
note that the last link of the users will never be removed in this process .
the effect of data density is particularly strong when recommendations are made with the diversity - favoring lhn - icf method which leads to low @xmath42 when the data density is high but does the opposite when the data density is low .
this can be explained by the presence of objects with only a few links in low - density data : once those links are rewired to other objects , the respective object can not be any more recommended and it effectively disappears from the system , thus contributing to an increased popularity inequality and a high @xmath42 . nevertheless , a decrease in the data density does not always increase gini value .
these observations may come from an interesting phenomenon , which can be shown by two different simulations . in the first case , users are randomly removed from the system and @xmath42 is found to increase with decreasing data density ( see fig .
s8(c ) ) . in the second case ,
items are removed at random , and by contrast @xmath42 decreases with decreasing data density ( see fig .
while these two scenarios look similar , the average item degree decreases in first case ( see fig .
s8(a ) ) and is preserved in the second case ( see fig .
s8(b ) ) . in other words , even though data density decreases , recommendation algorithms work equally effectively to distribute the popularity given the average item degree is preserved .
this may come from the fact that the nature of an item can be reflected by the characteristics of individual users who have collected it , the degree of an item represents the amount of available information on it . as a result , when the average item degree is preserved , recommendation algorithm works effectively since there are sufficient information on the items .
these results show that the degree of an item is not just merely related to its role in the network , but may also represents the amount of information we possess on it .
the effectiveness of recommendation algorithms are thus strongly dependent on the average item degree . on the stationary gini coefficient ; links which are not drawn based on recommendation are drawn based on preferential attachment .
( c , d ) the effect of data density on the stationary gini coefficient for different recommendation methods ( here all links are drawn based on recommendation ) .
the curve labeled original is the gini coefficient of the network after density modification and before rewiring . ]
* hysteresis . * the task of choosing the recommendation parameter @xmath25 is made more important by a pronounced hysteresis phenomenon .
[ fig : hysteresis ] shows that while a state with low inequality achieved with @xmath29 can be fully reverted to a high inequality state by changing @xmath25 to @xmath17 , the opposite is not true . in other words , once high concentration of item popularity has set in , it can only be partially corrected by the use of a diversity - favoring recommendation method .
this is also related to the amount of information available on the objects
after we lose information on some objects in a recommendation algorithm , we may not retrieve it again .
effective recommendation can not be made for these items and the popularity remains unevenly distributed , even with diversity - favoring methods . in fig .
s7 , we show that the hysteresis phenomenon exists also when @xmath41 . we thus conclude that @xmath42 depends on the system s initial condition , especially when it corresponds to highly heterogeneous item popularity
. achieved with @xmath29 and high-@xmath42 achieved with @xmath27 , in ( a ) movielens and ( b ) netflix data .
( c ) and ( d ) show the effect of @xmath25 on the stationary gini ( icf method ) under different initial configurations . ] * trade - off between diversity and accuracy . *
the previous figures show that the use of high @xmath25 can limit or even reverse the potential popularity - concentrating effect of recommendation .
however , high @xmath25 generally leads to low recommendation accuracy because it tends to recommend objects of low degree @xcite .
this motivates us to investigate the trade - off between the possibly decreased stationary gini coefficient @xmath42 and recommendation precision @xmath46 measured on the original input data before rewiring . to measure recommendation accuracy ,
we apply the standard evaluation procedure which is based on randomly dividing the network data into two parts : a training set @xmath47 comprising 90% of all links and a probe set @xmath48 comprising the rest .
the training data @xmath47 is then used to compute recommendation lists for all users .
the standard accuracy metric called _ precision _ is based on comparing these recommendation lists with the probe data @xmath48 : a good recommendation algorithm is expected to be able to reproduce a large part of @xmath48 based on @xmath47 @xcite .
if for user @xmath12 , there are @xmath49 probe entries related to this user in @xmath12 s recommendation list , we say that the recommendation precision for this user is @xmath50 . by averaging this quantity over all users with at least one entry in the probe set @xmath48
, we obtain the overall recommendation precision @xmath51 . to further remove its possible dependence on the data division , we average precision over ten independent training set - probe set divisions .
based on the training - probe sets division , one can also measure the short - term recommendation diversity which is simply the average degree of items that appear in the recommendation lists .
s9 shows the relation between the long - term gini coefficient and short - term recommendation diversity .
[ fig : sp_hy]a simultaneously plots @xmath42 and @xmath46 as a function of @xmath25 in the movielens data .
one can see that the highest precision is achieved at @xmath52 which is a point where the stationary gini decreases quickly with @xmath25 .
this gives us the possibility to lose some precision by increasing @xmath25 from its optimal value and in turn achieve a substantial decrease of @xmath42 .
this is demonstrated by fig .
[ fig : sp_hy]c where the desired value of @xmath42 is plotted on the horizontal axis : the resulting precision first decreases rather slowly from its highest value as the desired stationary gini coefficient is lowered . only when the desired @xmath42 is lower than the gini value in the original data , precision decreases sharply
the situation is less favorable for the netflix data where precision is maximized at @xmath53 which lies in the region where @xmath42 changes rather slowly with @xmath25 as shown in fig .
5b . as a result
, one has to substantially increase @xmath25 in order to achieve a significant decrease of @xmath42 .
this manifests itself in fig .
[ fig : sp_hy]d which lacks the gentle slope region seen in fig .
[ fig : sp_hy]c . nevertheless , one can limit @xmath42 to the values seen in the original movielens and netflix data by sacrificing 17% and 12% of the optimal recommendation precision , respectively .
these results show the possibility to compromise recommendation accuracy and long - term impacts on diversity for recommendation systems . in ( a ) movielens and ( b ) netflix data .
panel ( c ) and ( d ) show the relation between the desired stationary gini coefficient and the short - term recommendation precision . ]
by studying the co - evolution of the recommendations generated from the recommender systems and users choices , we demonstrate the long - term effect of recommendation on the distribution of item popularity .
this novel approach to the evaluation of recommendation performance gives us the possibility to observe new phenomena .
contrary to the common belief that recommendation helps to match niche items with users who may appreciate them and thus contribute to improving their recognition , we show that typical recommendation methods reinforce the position of already popular items at the cost of niche items .
this is particularly true when recommendation algorithms optimized for their accuracy are used because these tend to favor popular items .
our observations suggest that recommendation may divert the system to a state where a few items enjoy extraordinarily high levels of user attention .
furthermore , we found a strong hysteresis effect which implies that this state is very robust and can be hardly changed back to a state with a more even popularity distribution , even with the help of a diversity - favoring recommendation algorithm of lesser accuracy .
all these are adverse effects if one considers the interaction between users and the recommender system as an information ecosystem .
we remark that although it is not ideal for buyers to receive recommendations of common popular items , such systems may still benefit the sellers by lowering the effective number of distinct items that they need to keep in their inventory and thus reducing the costs for logistics . to the best of our knowledge ,
our model is the first one to analyze the long - term influence of recommendation on the evolution of online systems .
we tested many different choices of our model including examining different original data sets and another recommendation method , combining recommendation with random attachment instead of preferential attachment , preventing the real information from being washed out by repeated rewiring , and setting different lengths of the recommendation list .
our finding of the adverse effect of a sub - optimal recommendation system on information ecosystems in the long run is in general still valid in these cases and thus needs to be seriously considered in practice .
our work raises a number of questions which aim to further strengthen our understanding of the long term influence .
for example , the model now assumes that all users accept the recommendation with the same probability .
one can actually instead consider a scenario where experienced users search for items on their own and thus depend less on recommendation .
we measure recommendation accuracy before the rewiring process
. it would be interesting to monitor the recommendation accuracy during the network s evolution , and couple it with the rate at which users accept it ( the higher the accuracy , the higher the probability that users follow recommendation ) .
these more complicated scenarios may make the results of the model quantitatively different from the current ones .
a systematic study on these variances would be an interesting and important extension of the current model .
our findings suggest a need for a next generation of recommender systems which would take into account both short - term and long - term goals .
one might argue that the prime goal of a commercial system is to increase the profit by maximizing the recommendation accuracy and the long - term goal of enhancing or at least preserving the item diversity is secondary .
our results suggest that this is a short - sighted approach as focusing on short - term performance of recommendation may ultimately lead to a system where the long tail has been decimated together with its economic potential . to overcome this problem ,
item diversity can be enhanced by sacrificing a small fraction of recommendation s short - term accuracy in exchange for higher long - term diversity
. a detailed investigation of various approaches to study the long - term effects of recommendation as well as possible trade - offs between short- and long - term performance of recommendation are of great interest to both researchers and practitioners in the future .
this work was partially supported by the youth scholars program of beijing normal university ( grant no .
2014nt38 ) , eu fp7 grant 611272 ( project growthcom ) and by the swiss national science foundation ( grant no .
200020 - 143272 ) .
the work of chy is partially supported by the internal research grant rg 71/2013 - 2014r of hkied .
99 j. b. schafer , j. a. konstan and j. riedl , _ data min .
discov . _ * 5 * , 115 ( 2001 ) .
n . ziegler , s. m. mcnee , j. a. konstan and g. lausen , _ in proc .
14th international www conference , chiba , japan_. new york : acm .
22 - 32 ( 2005 ) .
liang , h .- j .
lai and y .- c .
ku , _ j. manage . inform .
syst . _ * 23 * , 45 ( 2007 ) j. a. konstan , b. n. miller , d. maltz , j. l. herlocker , l. r. gordon and j. riedl , _ commun .
acm _ * 40 * , 77 ( 1997 ) .
g. linden , b. smith and j. york , _ internet computing , ieee _ 7 , 76 - 80 ( 2003 ) .
g. adomavicius and a. tuzhilin , _ ieee trans .
_ , * 17 * , 734 ( 2005 ) .
g. taks , i. pilzy , b. nmeth and d. tikk , _ acm sigkdd explorations newsletter _ * 9 * , 80 - 83 ( 2007 ) .
netflix prize , http://techblog.netflix.com/2012/04/netflix-recommendations-beyond-5-stars.html .
e. brynjolfsson , y. hu and m. d. smith , _ manag .
sci . _ * 49 * , 1580 - 1596 ( 2003 ) . c. anderson , the long tail : why the future of business is selling less of more , hyperion ( 2006 ) .
j. leskovec , l. a. adamic and b. a. huberman , _ acm trans . web _ * 1 * , 5 ( 2007 ) .
h. yin , b. cui , j. li , j. yao and c. chen , _ proceedings vldb endowment _ * 5(9 ) * , 896 - 907 ( 2012 ) .
a. zeng , c. h. yeung , m .- s .
shang and y .- c .
zhang , _ europhys .
lett . _ * 97 * , 18005 ( 2012 ) .
l. l , m. matus , c. h. yeung , y .- c .
zhang , z .- k .
zhang and t. zhou , _ phys .
rep . _ * 519 * , 1 - 49 ( 2012 ) .
t. zhou , z. kuscsik , j .-
liu , m. medo , j. r. wakeling and y .- c .
zhang , _ proc .
natl acad .
sci . _ * 107 * , 4511 - 4515 ( 2010 ) .
j. l. herlocker , j. a. konstan , k. terveen and j. t. riedl , _ acm trans .
* 22 * , 5 ( 2004 ) .
l. page , s. brin , r. motwani and t. winograd , _ technical report .
stanford infolab . _
* 1999 * , 66 ( 1999 ) .
l. l , y .- c .
zhang , c. h. yeung and t. zhou , _
plos one _ * 6 * , e21202 ( 2011 ) .
h. kwak , c. lee , h. park and s. moon , _ in proc .
19th international www conference , raleigh , usa_. new york : acm .
591 - 600 ( 2010 ) .
k. sugiyama and m .- y . kan , _ in proc .
10th annual joint conference on digital libraries , gold coast , australia_. new york : acm . 29
j. bennett and s. lanning , _ in proc .
kdd cup and workshop ( acm ) , san jose , usa_. new york : acm .
35 ( 2007 ) .
we first denote by @xmath54 the rank of item @xmath13 according to their increase in degree within a year @xmath55 .
we further denote the number of active items in year @xmath55 as @xmath56 , one can find the normalized rank of item @xmath13 by @xmath57 . to ensure a fair comparison
, popularity is rescaled to keep the area under the popularity - rank curves the same for the results from different years .
this can be achieved by computing the normalized item popularity ( @xmath58 ) , given by @xmath59 , where @xmath60 is the degree of item @xmath13 in year @xmath55 .
y. koren , _ commun . of acm _
* 53 * , 89 ( 2010 ) .
r. lempel and s. moran , predictive caching and prefetching of query results in search engines .
_ in proc .
12th international www conference , budapest , hungary_. new york : acm . 29
t. joachims , l. granka , b. pan , h. hembrooke , f. radlinski and g. gay , _ acm trans .
syst . _ * 25 * , 7 ( 2007 ) . c. gini , variabilit e mutabilit
( in english : variability and mutability ) ( c. cuppini , bologna ) ( 1912 ) .
l. wittebolle , et al .
_ nature _ * 458 * , 623 - 626 ( 2009 ) .
y. asada , _ popul .
health metr .
_ * 3 * , 7 ( 2005 ) . w. halffman and l. leydesdorff , _ minerva _ * 48 * , 55 - 72 ( 2010 ) .
d. goldberg , d. nichols , b. oki and d. terry , _ commun .
acm _ * 35 * , 61 - 70 ( 1992 ) .
l. l and t. zhou , _ physica a _ * 390 * , 1150 ( 2011 ) . | recommender systems daily influence our decisions on the internet .
while considerable attention has been given to issues such as recommendation accuracy and user privacy , the long - term mutual feedback between a recommender system and the decisions of its users has been neglected so far .
we propose here a model of network evolution which allows us to study the complex dynamics induced by this feedback , including the hysteresis effect which is typical for systems with non - linear dynamics . despite the popular belief that recommendation helps users to discover new things , we find that the long - term use of recommendation can contribute to the rise of extremely popular items and thus ultimately narrow the user choice .
these results are supported by measurements of the time evolution of item popularity inequality in real systems .
we show that this adverse effect of recommendation can be tamed by sacrificing part of short - term recommendation accuracy . |
the standard model of elementary particle physics ( sm ) gives an excellent explanation of all existing experimental data . however , there are quite strong theoretical motivations to believe that the sm is an effective theory of a more fundamental theory and that the gauge couplings of @xmath16 have a common origin .
the construction of grand unified theories ( guts ) @xcite , which unify @xmath17 gauge interactions in a single non abelian group [ @xmath2 , @xmath18 , @xmath19 etc ] , give an elegant explanation of charge quantization and also unify quark - lepton families .
the idea of gut got a great support from the fact that the three gauge couplings measured at that early times were indeed unifying at energies near @xmath20 gev @xcite .
progress in measuring the strong gauge coupling and also the weak mixing angle @xmath21 with higher accuracy has ruled out the minimal @xmath2 gut [ and also minimal @xmath18 without intermediate scale ] from the viewpoint of coupling unification @xcite , @xcite .
however , the minimal supersymmetric extension of the standard model ( mssm ) and also the minimal susy @xmath2 gut @xcite(which except for gut threshold corrections both have the same pattern of running of couplings below @xmath22 ) were giving values for the @xmath23 coupling @xcite-@xcite well within the experimental limits at that time . indeed
since susy theories stabilize hierarchies , for realistic model building supersymmetry might be the best way to proceed .
it is assumed , that the susy breaking scale @xmath24 lies in a range @xmath25 gev - few tev and below this characteristic scale the theory is just the sm with minimal particle content except the higgs sector , while above the @xmath24 scale the theory is supersymmetric . despite these nice features of susy theories ,
there are various puzzles and problems , which are connected with susy guts and we will list some of them here . * ( i ) * baryon number violation is a particular feature of guts such as @xmath2 , @xmath18 . since for susy guts the unification scale @xmath26 gev is larger than for non susy guts , the gauge mediated @xmath27 nucleon decay is compatible with the latest superkamiokande ( sk ) limit @xmath28 yrs @xcite .
however , with susy there is a new source for nucleon decay through @xmath29 operators , which makes the minimal @xmath2 and @xmath18 scenarios incompatible @xcite with sk data . *
( ii ) * the unified multiplets of minimal @xmath2 lead to the wrong asymptotic mass relations @xmath30 . in the minimal @xmath18
the situation is even worse , since @xmath31 and @xmath32 is predicted . *
( iii ) * the problem of doublet - triplet ( dt ) splitting in the higgs supermultiplet still needs to be resolved . in guts ,
the mssm higgs doublets are usually accompanied by colored triplets . in order to maintain coupling unification and reasonably stable nucleons
[ triplets could induce nucleon decay through @xmath29 operators , see * ( i ) * ] , triplet components must be superheavy .
so , one should provide a natural explanation of the fact that sometimes states ( higgs doublets and colored triplets ) coming from the same gut multiplet are split with a huge mass gap @xmath33 .
* ( iv ) * the spontaneous breaking of the gut symmetry requires scalars in a high representation of the gauge group considered ; thus the superpotential , responsible for symmetry breaking , contains many unknown parameters and usually looks rather complicated . *
( v ) * the so - called @xmath0 problem exists even within the mssm .
4d superpotential couplings allow a @xmath34 term , where @xmath35 , @xmath36 are the mssm higgs doublets and @xmath37 is some mass close to the cutoff scale of the theory .
so , somehow large values for @xmath38(@xmath39 ) must be avoided . in order to have the desired electroweak symmetry breaking and a reasonable phenomenology ,
a @xmath0 term of the magnitude @xmath40 gev - few tev has to be generated in a good model ( also within guts after solution of the dt splitting problem * ( iii * ) and having succeeded to obtain @xmath41 ) .
* ( vi ) * recent atmospheric @xcite and solar @xcite neutrino sk data have confirmed neutrino oscillations .
the explanation of the atmospheric anomaly ( by a characteristic mass squared scale @xmath42 ev@xmath43 ) already forces us to step beyond the mssm and the minimal susy @xmath2 ( a neutrino mass @xmath44 ev can be generated through planck scale @xmath29 operators and can explain the solar anomaly through large angle vacuum oscillations .
however , this solution is disfavored by the sk data ) .
in order to have neutrinos with masses @xmath45 ev , the lepton number must be violated by a proper amount .
this requires considering extensions of the mssm and the minimal susy @xmath2
. it would be most welcome if the considered model would contain a source for the needed lepton number violation . *
( vii ) * very accurate measurements of @xmath23 @xcite already allow to judge whether a given gut scenario is viable or not .
two loop renormalization studies of the mssm ( with all susy particles near the @xmath46 scale ) predict @xmath47 @xcite , which contradicts the experimental @xmath48 @xcite . this situation can be improved either by pushing all susy particle masses up to the @xmath49 tev mass scale @xcite , or by some gut threshold corrections . with the latter the minimal susy @xmath2 does not give any promising results @xcite . comparing gut scenarios ,
those would be considered more attractive which , without constraining the susy particle mass spectra , give acceptable values for the strong coupling . on the theoretical side
quite a few new possibilities have been found since the early days of guts and also the introduction of susy . *
@xmath50 ) * string theory is primarily a theory of ( super)gravity but it also contains in a less unique manner matter and gauge fields .
it had an enormous effect on the taste of model builders although concrete phenomenological results are still not obvious .
we particularly mention symmetry breaking mechanisms not requiring very high higgs representations of the gut gauge groups and a natural assignment of fundamental representation to matter .
also the possibility to calculate ( in principle ! ) yukawa couplings is very impressive .
but unfortunately there is a huge and even increasing number of string vacua with ( presently ) no possibility to make a choice of one or the other except for phenomenological reasons .
still until recently @xcite it was notoriously difficult to find a string model realization implementing the sm with three generations . *
@xmath51 ) * extra dimensions : the kaluza - klein use of extra dimensions to be curled up one way or the other had several renaissances .
of course it is very tempting to obtain extra model informations from some extra dimensions - there is plenty of space in this dreamland , which can contain geometry / topology . in string
theory extra dimensions are mandatory for consistency .
a drawback then seems to be that such extra dimensions would show up only at the string scale which is normally identified with the planck scale of our gravity .
one then is led to talk about physics which presumably never will be tested in the laboratory .
recently it became a point of common interest whether the string scale might be as low as the tev scale @xcite still allowing for our gravity scale . in this case
higher dimensions should show up soon in experiments @xcite . *
@xmath52 ) * it was exciting news that dualities connect the various types of string theories @xcite .
the open string picture allows for @xmath53-branes which contain our 3-dimensional space but also allow for some extra dimensions which may be curled up or projected out in the case of intersecting branes .
this version of string theory may be particularly appropriate also in the case of singular points of divided out symmetries for an approximation by a description in local quantum field theory language since there are no winding states . in resolving problems * ( i)-(iv ) * of susy gut scenarios the orbifold constructions seem to be very promising @xcite-@xcite . in the original paper of ref .
@xcite , a five dimensional ( 5d ) @xmath3 susy @xmath2 gut on an @xmath54 orbifold was considered . due to this construction
, it turns out that the problems * ( i)*-*(iv ) * can be resolved in a very natural way for a wide class of unified models @xcite-@xcite , while * ( v)*-*(vii ) * still depend on peculiarities of the scenario considered and will be discussed in more detail below . due to specific boundary conditions , it is possible to mod out selected sub - states from a given gut representation . through this self consistent procedure ,
it is possible to obtain the desired gut symmetry breaking , nucleon stability and natural dt splitting . in the last years
, theories with extra dimensions have attracted great attention .
originally the main phenomenological motivation was the possibility to resolve the gauge hierarchy problem without supersymmetry .
it was observed @xcite , that due to sufficiently large extra dimensions , it is possible to lower the fundamental scale @xmath55 even down to a few tev ( indeed , this can be an excellent starting point for understanding the electroweak scale ) , while the 4d planck mass still has the required value @xmath56 gev . due to the large extra dimensions , newton s law could be modified at short distances where the behavior of gravity is still unknown and is studied in ongoing experiments @xcite . similarly and perhaps with a richer phenomenology @xcite , one can study the spectrum for scenarios with a string scale of a few tev @xcite .
it turned out , that the presence of extra dimensions can play a crucial role also for obtaining low scale unification of gauge couplings @xcite-@xcite through power law running @xcite .
the construction of realistic gut scenarios with low scale unification raises the hope that phenomenological implications can be detected .
however , the orbifold gut scenarios considered up to now do not allow for low scale unification @xcite , @xcite because in these settings the gut symmetry is restored at energies higher than the compactification scale .
thus full gut multiplets [ either of @xmath2 or @xmath6 will participate in the running and power law unification does not take place .
relatively low scales @xmath57 gev are also preferable for lepton number violation .
one way for obtaining unification on a scale much below @xmath10 gev is to consider either gut models with product groups or with ( intermediate ) stages of symmetry breaking - step by step compactification of more than one extra dimension . on the gut scale @xmath22
a first step compactification ( @xmath58 ) takes place and the unified group @xmath59 reduces to its subgroup @xmath60 . in the second step compactification , whose scale @xmath61 is below @xmath22 , the subgroup @xmath60 is broken . if @xmath60 is different from @xmath2 and if the states are non complete multiplets of @xmath2 , then due to their contribution to the running between @xmath62 and @xmath22 there can appear power law unification on intermediate or low scales . in this paper
we consider 5d @xmath3 susy models with orbifold compactifications .
we start our discussion with the standard model @xmath17 gauge group and an @xmath1 orbifold . in order to have a model without the phenomenological problems * ( i)*-*(vi ) * [ in this case except the * ( ii)*-*(iv ) * of course ] , we introduce a discrete @xmath11 symmetry which elegantly resolves problems * ( i ) * , * ( v ) * and replacing matter @xmath63 parity allows for some lepton number violating couplings which can generate neutrino masses .
thus , * ( vi ) * also can be resolved .
we confirm that , with the mssm states plus appropriate kaluza - klein ( kk ) excitations , successful unification holds only for @xmath64 gev .
low scale unification requires either some extensions @xcite , @xcite or the existence of specific threshold corrections @xcite .
similarly we can discuss 5d @xmath3 susy @xmath2 gut on an @xmath54 orbifold . in this setting the problems * ( ii)*-*(iv ) * are resolved naturally , while * ( i ) * , * ( v ) * will again be resolved by introducing a discrete symmetry @xmath11 . as far as the gauge coupling unification is concerned , all states including matter supermultiplets and their copies form full @xmath2 multiplets above the compactification scale . because of this , low and intermediate scale unification can not take place .
then we address the question whether power law unification is possible or not ( at low or intermediate scale ) within the orbifold gut construction .
we emphasize the possibility of a so - called step by step compactification with an intermediate gauge group , different from the @xmath2 in structure and field content .
this potentially allows for power law unification . besides the latter
a quite different and peculiar phenomenology can arise . to demonstrate this we consider pati - salam @xmath65 @xcite and flipped @xmath66 guts .
both these gauge groups are maximal subgroups of @xmath18 @xcite , @xcite and thus one could imagine that they are produced in a first step breaking of @xmath18 in six dimensions by the compactification of one dimension . within 5d @xmath3 susy @xmath8 and @xmath9 models an extension with a discrete @xmath11 symmetry is needed for a simultaneous solution of the problems * ( i)*-*(vi)*. these models involve sm singlet right handed states which are necessary for the breaking of the rank and obtaining the @xmath17 gauge group . in combination with the @xmath11 symmetry , these singlet states also play a crucial role in understanding of problems * ( i ) * , * ( v ) * , * ( vi)*. they are also tied with lepton number violation and the generation of an intermediate symmetry breaking scale .
the @xmath8 model allows to lower the unification scale not only down to intermediate scales , but even down to the multi tev region .
differently , within @xmath9 the unification scale is close to @xmath10 gev .
the models considered have some peculiar phenomenological implications testable in the future .
the paper is organized as follows . in section 2
we present the main construction principles of the models considered . in section 3 we write the needed one loop renormalization - group equations ( rge ) . using them we study gauge coupling unification within various models , in the presence of kk states .
sections 4 and 5 are devoted to the 5d orbifold @xmath3 susy @xmath17 and the @xmath2 models resp . in section 6
we discuss the issue of power law unification , within orbifold gut scenarios , and outline the ways of its realization . in sections 7 and 8
pati - salam @xmath8 and flipped @xmath67 guts resp .
are studied on an @xmath7 orbifold .
finally discussions and conclusions are presented in section 9 .
the paper contains an appendix a , in which the influence of some brane couplings on the gauge coupling running is estimated .
in this section we present our construction principles of 5d susy theories . as we will see they are divided into two categories : principles which are related to the higher dimensionality and others which deal with problems existing on the 4d level , after dimensional reduction .
* @xmath68 . 5d susy action * we start the construction with a 5d @xmath3 susy theory . from the viewpoint of 4d
( with coordinates @xmath69 ) , with the fifth coordinate @xmath70 as a parameter , it is equivalent to @xmath71 susy . @xmath71 supermultiplets can be expressed in terms of the usual 4d @xmath3 supermultiplets @xcite : a gauge supermultiplet @xmath72 contains the 4d @xmath3 vector superfield @xmath73 and the chiral superfield @xmath74 , both in the adjoint representation of the gauge group @xmath59 and depending on the fifth coordinate .
the 5d matter superfield , in 4d language , is the @xmath71 chiral supermultiplet @xmath75 , where @xmath76 is the @xmath3 chiral superfield and @xmath77 is it s conjugate -the so - called mirror ( through out the paper the mirrors will be denoted by an overline ) .
so , if @xmath78 is in some irreducible representation @xmath79 of @xmath59 , then @xmath77 will be in an antirepresentation @xmath80 of @xmath59 . under gauge transformations
one has @xmath81
e^ , e^- , where @xmath82 is a chiral superfield .
the transformation of @xmath74 in ( [ n2gaugetr ] ) reflects the 5d gauge invariance , since @xmath74 contains the fifth component of the five dimensional gauge field @xcite , @xcite .
the 5d action can be written in terms of 4d @xmath3 superfields @xcite and has the form s^(5)=d^5x(l^(5)_v+l^(5 ) _ ) , where @xmath83 d^4(_5v+^+)e^-v ( -_5v+)e^v+ _ 5e^-v_5e^v , ^(5)_=d^4^+e^-v+ e^v^+ + d^2m_+_5-+ h.c .. here @xmath84 is the field strength supermultiplet , also in the adjoint representation of @xmath59 and built from @xmath73 ( @xmath85 ) .
the last term in ( [ lagf ] ) contains the @xmath86-term of @xmath87 , which is crucial for 5d lorentz invariance : for a bosonic component @xmath88 of the @xmath78 superfield it produces the term @xmath89 which together with @xmath90 [ coming from the first coupling in ( [ lagf ] ) ] is 5d lorentz invariant .
the same happens for the fermionic components .
the @xmath91 combination is crucial for the 5d gauge invariance under ( [ n2gaugetr ] ) .
there are two supersymmetries in @xmath92 , the obvious 4d @xmath3 susy and one related by a global @xmath93 symmetry to the former one @xcite .
thus the susy transformation parameters as well as the scalar components of @xmath94 and the two spinors @xmath95 in @xmath84 and @xmath74 form doublets under this @xmath93 .
the fermionic components of @xmath94 and the bosonic components of @xmath84 and @xmath74 are @xmath93 singlets .
the @xmath3 susy theory in 5d has the advantage that there is no free superpotential .
the action is completely fixed except for the @xmath96 term in ( [ lagf ] ) which in some cases might be forbidden by orbifold parities ( see below ) .
the @xmath96 only connects fields with their mirrors .
* @xmath97 .
compactification and orbifold symmetries * since we have one extra dimension , it is important somehow to reduce the theory to the 4d one .
one can start from a @xmath98 theory , where @xmath99 is the four dimensional minkowski space - time and @xmath100 a compact circle .
equivalently , one can consider the fifth dimension as an infinite @xmath101 line and impose some periodicity @xmath102 , where @xmath63 is the radius of the circle corresponding to the characteristic compactification scale @xmath103 .
so , the theory in the fifth dimension is defined on a interval @xmath104 $ ] or equivalently on @xmath105 $ ] . on the interval
@xmath106 one can introduce discrete symmetries , and @xmath107 is the simplest one z_2 : y - y , which folds the circle .
the theory is then built on an @xmath1 orbifold . under ( [ z2orbsym ] ) all introduced fields
@xmath108 should have definite parity transformation properties @xmath109 , such that the 5d lagrangian ( [ lagv ] ) , ( [ lagf ] ) is invariant ( @xmath108 designates all gauge and matter supermultiplets we have ) .
@xmath110 and the mode expansions of states @xmath111 and @xmath112 with positive and negative parities resp . have the form _ + ( x , y)=_n=0^^(n)_+(x ) , _ -(x , y)=_n=1^^(n)_-(x ) , @xmath113 and @xmath114 are kaluza - klein ( kk ) states .
as we see , @xmath115 does not contain a zero mode .
massive kk modes have masses @xmath116 .
we have two fixed points @xmath117 and @xmath118 . with the help of the @xmath107 orbifold parity it is possible to project out some states ( assigning them negative parities ) and to achieve the breaking of supersymmetries and gauge symmetries .
if we wish to break the gauge group @xmath59 down to its subgroup @xmath60 , gauge fields @xmath119 should have negative parities , while the parities of fragments @xmath120 are positive . from ( [ lagv ] ) , it is clear that in this case @xmath121=-1 $ ] and @xmath122=+1 $ ] ( because @xmath123 changes sign under @xmath107 ) .
also , it follows from ( [ lagf ] ) that mirrors must have opposite parities .
because of all this , together with the gauge symmetry , half of the susy is broken and at the fixed points we have a @xmath124 @xmath3 susy theory with a reduced gauge group .
but we are also left with the additional zero mode states of @xmath125 . in order to avoid them ,
the orbifold symmetry can be extended to @xmath126 @xcite : by additional folding of the half circle z_2 : y - y , z_2 : y-y , where @xmath127 , one can ascribe negative @xmath128 parity to @xmath125 and @xmath128 charge for @xmath119 .
now the theory is defined on an @xmath7 orbifold and at the @xmath117 fixed point ( identified with our 4d world @xmath129-brane ) we have a 4d @xmath3 susy theory with gauge group @xmath60 .
no additional fragments of @xmath74 with zero mode wave functions emerge . of course ,
also in this case mirrors should have opposite @xmath128 parities . in the next sections we will demonstrate transparently with concrete examples how this procedure is realized .
each state has a definite @xmath126 parity @xmath130 parity @xmath131 .
therefore , under the transformations ( [ orbzs ] ) : p , p . depending on the @xmath130 parity , there are four possible mode expansions
@xmath132 @xmath133 @xmath134 @xmath135 _ (x , y)=_n=0^ ^(2n+2)_(x ) . consequently
, the masses of the appropriate kaluza - klein ( kk ) modes of @xmath136 , @xmath137 , @xmath138 and @xmath139 will be @xmath140 , @xmath141 , @xmath141 , and @xmath142 , resp .
only the @xmath143 states contain massless zero modes .
states with other parities are massive .
we emphasize again that , if we introduce states in the bulk and ascribe to them some parity @xmath144 , the mirror must carry @xmath145 parity . in this way we have 5d lorentz invariance .
this is quite different when a state is fully restricted to the brane and does not have kk excitations ( as possibly chiral matter in some cases which we will consider below ) .
if we want a state ( introduced in the bulk ) to have a zero mode component , we should assign to it ( + , + ) parity .
for all other parity choices , the states have only massive kk excitations .
* @xmath146 .
construction of the 4d theory on a brane , * additional discrete symmetries and extensions as we have already mentioned , 5d susy does not allow to have a superpotential which leads to yukawa couplings .
this enforces brane couplings in order to build a realistic phenomenology .
couplings at the @xmath117 fixed point =d y(y ) w^(4)(x , y ) , possess 4d @xmath3 supersymmetry and involve fields with zero mode wave functions .
@xmath147 includes yukawa couplings which are responsible for the generation of fermion masses .
the couplings in ( [ branesup ] ) do not violate the higher supersymmetries and gauge symmetries of the 5d bulk .
the reason for this is that the wave functions of generators which transform zero mode states to states with negative orbifold parities vanish in the 4d fixed point . in this way
the whole theory is self consistent .
as we have already mentioned in the introduction , the orbifold constructions have big advantages in resolving various puzzles connected with guts . however
, the problems * ( i ) * , * ( v)*-*(vii ) * ( mentioned in the introduction ) still remain at the 4d level , and need to be tackled . amongst them
the most urgent ones are baryon number conservation and the @xmath0 problem .
furthermore , problems emerging from matter parity violating operators should be avoided and the neutrino deficits must be explained .
we consider these problems to be severe enough to motivate us to think about some reasonable extension of the considered scenario . starting with the @xmath0 problem ,
for its solution we introduce an additional discrete symmetry @xmath11 and prescribe transformation properties to @xmath35 , @xmath36 in such a way as to forbid a direct @xmath0 term .
we also introduce singlets @xmath148 , @xmath149 which have vevs @xmath150 ( the cutoff scale ) . through @xmath151 type couplings with a proper choice of @xmath152
we obtain a @xmath0 term of the desired magnitude @xcite . in section 4
, we explicitly demonstrate how the generation of @xmath149 , @xmath148 vevs and the @xmath0 term suppression are realized .
the mssm and the minimal susy @xmath2 require @xmath148 , @xmath149 singlets , while the models @xmath153 and flipped @xmath67 automatically involve scalars being singlets of the mssm ( see sections 7 , 8) . in the mssm and susy guts ,
usually a @xmath107 @xmath63-parity is assumed , which distinguishes matter and scalar superfields and avoids baryon number and large lepton number violation . in our approach , for the same purpose we use the @xmath11 symmetry , which avoids all baryon number violating couplings which also violate @xmath63 parity . with help of the introduced @xmath11 symmetry
we also avoid @xmath29 and @xmath27 baryon number violating planck scale operators , which are otherwise allowed on the 4d level , causing unacceptably rapid nucleon decay ( @xmath27 operators become dangerous if we are dealing with low or intermediate scale theories ) .
so , from this point of view , the extension with a discrete @xmath11 symmetry turns out to be very efficient @xcite . @xcite and anomalous gauge @xmath154
@xcite symmetries have been used . in @xcite models with gauged baryon number were suggested . ]
as far as the lepton number violating couplings are concerned it is well known that the mssm and the minimal susy @xmath2 do not give sufficiently large neutrino masses and that , for accommodation of atmospheric and solar neutrino data , some extensions are needed . in our constructions
we admit some lepton number violating couplings ( which usually are absent due to @xmath63 parity ) and due to proper suppression ( with the help of the @xmath11 symmetry ) they give desirable value(s ) for the neutrino masses .
we will discuss this issue in more detail through the sections 4 , 5 and 8 . concluding this section we point out that , when using @xmath155 symmetry , one should make the corresponding charge assignments to the matter and scalar supermultiplets in such a way that the terms in ( [ lagf ] ) , allowed by orbifold symmetries , are invariant also under @xmath11 .
this means that mirrors must have opposite @xmath11 charges and if the considered scenario is a gut , the states coming from one unified multiplet should have the same transformation properties under the @xmath11 symmetry .
in this section we will present general expressions for the solutions of the one loop renormalization - group equations ( rge ) in the presence of kk excitations corresponding to one extra space like dimension , which will be needed to estimate gauge coupling unification in different scenarios . at energy scales below the compactification scale @xmath61 the one loop running of the gauge couplings @xmath156 has logarithmic form @xcite _ i^-1(_+1 ) = _
i^-1(_)- . for the standard model
the gauge groups labeled by @xmath157 correspond to @xmath158 , @xmath159 , @xmath160 resp . without intermediate scales and additional states
, the @xmath161 factors will be just those corresponding to the states of the sm or the mssm ( depending on whether the theory we are studying is supersymmetric or not ) .
assume that up to a certain mass scale @xmath162 we have the @xmath16 gauge group with the minimal content of sm / mssm .
then the couplings at @xmath162 are _
a^-1(m_i)=_a^-1(m_z)- .
labeling couplings in ( [ rge321 ] ) by @xmath163 we emphasize that we are dealing with @xmath17 gauge couplings . above the scale @xmath162
the gauge group can be different and consequently runnings should be studied according to the existing gauge group and the corresponding states .
couplings at different mass scale regions must be matched at the intermediate scale(s ) @xmath162 .
so , we will run couplings up to the unification scale @xmath22 , which we treat as the cutoff scale of a theory .
since we are considering theories with one compact dimension , above the scale @xmath62 we should include the effects of kk modes . in the concrete models considered below , at an intermediate scale @xmath162
, two gauge groups [ either two @xmath164s or @xmath165 and @xmath164 ] are reduced to the @xmath158 .
we have the boundary / matching condition _ 1 ^ -1(m_i)=^2 _ g_1 ^ -1(m_i)+ ^2 _
g_2 ^ -1(m_i ) , where @xmath167 is a group - theoretical factor determined from the pattern of @xmath158 gauge group embedding in a product group @xmath168 .
@xmath169 , @xmath170 are the couplings of the gauge groups @xmath171 , @xmath172 and above @xmath162 we will have equations of the ( [ rgelog ] ) type for them .
however , we can also write rge for the combinations ( [ match ] ) in ( [ rgelog ] ) form , where the role of @xmath173 is now to be played by a superposition of @xmath174 and @xmath175 , similar to ( [ match ] ) b_1^m_i=^2 b_g_1+^2 b_g_2 . taking all this into account , we will have _
a^-1(m_g)=_a^-1(m_z)-+_a , with _
a=_a^0+_a^kk , where _
a^0=- includes contributions from all existing zero mode states @xmath50 with mass @xmath176 .
@xmath177 comes from the contributions of kk states . in the case
that their masses are @xmath178 , we have _
=- s , s=_n=1^n_0 , where @xmath179 is a common factor of the given kk states and @xmath180 stands for the maximal number of kk states which lie below @xmath22 . for models with @xmath181 orbifold parities
, @xmath177 will have the form _ a^kk =- s_1-s_2 , where @xmath182 and @xmath183 include contributions from kk states with masses @xmath184 and @xmath185 resp .
: s_1=_n=0^n , s_2=_n=0^n . in ( [ skk ] )
, @xmath186 and @xmath187 are the maximal numbers of appropriate kk states which lie below @xmath22 , i.e. ( 2n+2)_0m_g , ( 2n+1)_0m_g .
kk states with masses larger than @xmath22 are irrelevant . for a given @xmath188
the @xmath186 and @xmath187 can be calculated from ( [ maxns ] ) .
let us note , that @xmath189 , @xmath190 and @xmath191 will be expressed by similar superpositions as @xmath192 in ( [ bsupg1g2 ] ) ,
_ 1=^2 _ g_1+^2 _
g_2 , _ 1=^2 _
if at a scale @xmath22 we impose the condition of gauge coupling unification _
1(m_g)=_2(m_g)=_3(m_g)_g , then from ( [ alga ] ) , eliminating @xmath193 and @xmath194 , we find for the strong coupling at the @xmath46 scale _ 3 ^ -1=_2 ^ -1- _ 1 ^ -1 + _
2-_1-_3 , where @xmath195 in ( [ als ] ) stands for @xmath196 .
also , from ( [ alga ] ) one can obtain = ( _ 1 ^ -1-_2 ^ -1)+ ( _ 1-_2 ) , and finally the value of the unified gauge coupling _
g^-1=_2 ^ -1-+_2
. for a given model , the values of @xmath197 can be fixed [ according to ( [ delta ] ) , ( [ delog ] ) , ( [ dekkz ] ) or ( [ dekkzz1 ] ) ] and from ( [ als ] ) one can calculate @xmath198 .
the contribution from the @xmath197s should not be too large , such that the experimental value @xcite @xmath199 is obtained .
if the contributions from @xmath197 in ( [ scale ] ) are negative and large , one can obtain a ( relatively ) low @xmath162 scale and consequently small @xmath62 , @xmath22 .
when constructing models , we should keep in the mind that the gauge couplings must remain in the perturbative regime until they reach the unification point .
for this it is enough to require a perturbative value for @xmath193 , calculated from ( [ alg ] ) . in the following ,
equations ( [ als])-([alg ] ) will be used to estimate the status of gauge coupling unification in various scenarios . of course , taking into account various threshold corrections ( from weak and gut scales or from some brane localized operators ) , these equations will have additional entries .
the relevance of such contributions will be commented below .
consider a 5d @xmath3 susy @xmath201 theory . since we do not have to break the gauge group
it is enough to introduce only one @xmath107 orbifold parity , i.e. the theory is defined on an @xmath1 orbifold .
according to the discussions of section 2 , in this way we can break half of the supersymmetries .
the field content , their orbifold parities and @xmath202 hypercharges are given in table [ t:2 ] .
we use the @xmath2 normalization y=(2 , 2 , 2 , -3 , -3 ) . at the @xmath117 fixed point
we are left with the susy @xmath17 gauge theory with zero mode states @xmath203 , @xmath204 , @xmath205 , @xmath206 , @xmath207 , @xmath35 , @xmath208 , which is just the content of the mssm . * fixed @xmath117 point brane couplings * and some phenomenology in order to build a realistic theory we write brane couplings of the ( [ branesup ] ) type .
the 4d yukawa superpotential , responsible for the generation of up - down quark and charged lepton masses , has the form ( neglecting coupling constants ) w_y = qu^ch_u+qd^ch_d+le^ch_d .
according to part * @xmath146 * of section 2 , to resolve various problems , it is useful to introduce a @xmath11 discrete symmetry . with the symmetry transformation
h_uh_de^ih_uh_d , the @xmath209 coupling is forbidden . introducing singlet states
@xmath210 , @xmath149 , @xmath149 states can be introduced in the bulk . in this case , on the 5d level they are accompanied by the appropriate mirrors with opposite orbifold parities . for us
the 4d superpotential couplings are important in which only the zero modes of @xmath148 , @xmath149 participate . ] with the transformation e^i , we get that the relevant coupling will be w_=m^n-1h_uh_d .
@xmath211 due to the transformations of @xmath212 and @xmath213 , @xmath11 acts as a @xmath214 symmetry .
if @xmath148 , @xmath215 develop vevs such that @xmath216 , by an adequate choice of @xmath152 one can obtain a properly suppressed @xmath0 term . the lowest superpotential coupling for @xmath148 , @xmath215 is w_s = m^3^n , and in the unbroken susy limit the conditions @xmath217 give @xmath218 .
after susy breaking , soft susy breaking terms should be involved . the relevant soft terms concerning @xmath148 ,
@xmath215 are v_soft(s)=m_1 ^ 2 |s|^2+m_2 ^ 2||^2 + m_3a(w_s+w^*_s ) , where @xmath219 , @xmath220 , @xmath221 are all of order of the susy scale @xmath24 and @xmath222 is a dimensionless constant . with ( [ sups ] ) , ( [ softspot ] )
one can write the total potential for @xmath148 as v(s)=|f_s|^2+|f_|^2+v_soft(s ) .
minimization of ( [ pots ] ) leads to a non zero solution for @xmath223 , @xmath224 ~ ~m^ . substituting ( [ sols ] ) into ( [ supmumssm ] )
, we obtain for the @xmath0 term m^2n-2~m .
as we see the @xmath214 symmetry gives a natural generation of the @xmath0 term ( which is independent of @xmath152 ! ) with the required magnitude .
the relevant feature is that the @xmath225 , @xmath226 in ( [ sols ] ) are expressed through the interplay of the two scales @xmath24 and @xmath37 @xcite . for @xmath227 tev , @xmath228 gev ( reduced planck mass ) and @xmath229 one has @xmath230 . for lower values of the fundamental scale @xmath37
, the desired gap between @xmath225 and @xmath37 can be obtained by a proper choice of @xmath152 .
for example , for the same value of @xmath24 , @xmath231 and @xmath232 gev , we need @xmath233 , while for @xmath234 tev no large suppression is required and one can take @xmath235 .
the introduced discrete symmetry is crucial for avoiding @xmath29 baryon number violation and also unacceptably large @xmath63-parity violating operators .
we note however that amongst the latter operators there are lepton number violating couplings which , being properly suppressed , could generate neutrino masses of the needed magnitude @xcite-@xcite . together with the yukawa couplings in ( [ yukmssm ] ) , we can therefore include the lepton number violating coupling m^n-1 ^kh_ul , which after substituting the appropriate vevs [ see ( [ sols ] ) , ( [ valmuterm ] ) ] leads to the bi - linear operator _
l h_ul , _
l~^k . due to this operator the sneutrino field can gain a vev of the order @xmath236 .
the latter produces a neutrino - neutralino mixing which leads to a neutrino mass through the see - saw type mechanism @xcite m_~o(100 gev)^2 , where @xmath237 ( assuming that there is no alignment between the superpotential and the soft susy breaking couplings ) . to have a neutrino mass @xmath238 ev , in ( [ nubilins ] ) we need @xmath239 . with a @xmath0-term @xmath240 and @xmath241 , @xmath242 , @xmath243 - @xmath244 , from ( [ bilinops ] ) we have @xmath245 which gives @xmath246 ev , indeed the order of magnitude needed for explaining the atmospheric neutrino anomaly .
the phases of @xmath35 , @xmath208 , @xmath148 , @xmath215 were not fixed by the couplings ( [ supmumssm ] ) , ( [ sups ] ) .
the couplings given above determine the transformation properties of the different states under the @xmath11 symmetry @xmath247 [ @xmath248 is the phase of state @xmath249 , and its mirror @xmath250 has opposite phase ) . due to the couplings in ( [ yukmssm ] ) , ( [ supmumssm ] ) , ( [ sups ] ) , ( [ bilins ] ) we have @xmath251 @xmath252 ( u^c)=-(q)-(h_u ) , ( e^c)=2(h_u)+k(s)-2 , = , where @xmath253 , @xmath254 , @xmath255 are undetermined .
other allowed @xmath63 parity breaking operators also violating the lepton number are ^kqd^cl , ^ke^cll . after substituting vev of @xmath148
they lead to the couplings qd^cl ,
e^cll , ~ ~^k .
these couplings induce neutrino masses at one loop , with the dominant contribution given by the @xmath256 state inside the loop , , which for @xmath257 , @xmath258 , @xmath227 tev , @xmath259 gev is evaluated as @xmath260 ev , to explain the solar neutrino puzzle either through msw ( by large or small angle , depending on which mixing scenario is realized for the fermion sector ) or large angle vacuum oscillations ( lavo ) .
this way of neutrino mass generation through properly suppressed @xmath63 parity violating operators @xcite-@xcite looks attractive since it does not require the introduction of right handed neutrinos . however , additional symmetries ( in this case @xmath11 ) are crucial @xcite , @xcite for obtaining properly suppressed neutrino masses . with the assignments ( [ zphasesmssm ] ) and taking @xmath261 , @xmath262 , @xmath263 the discrete symmetry introduced would be @xmath264 . with the phases presented , the @xmath265 and other higher order terms are allowed , but along the ( [ sols ] ) solution they are strongly suppressed in comparison to terms in ( [ sups ] ) , ( [ softspot ] ) . therefore , the analyzes above stay valid .
one can also verify that for any integer @xmath266 the baryon number violating @xmath267 operator @xmath268 is forbidden .
also , the baryon number violating @xmath29 operators qqql , , qqqh_d are not allowed .
there are also @xmath27 baryon number violating @xmath53-term operators _ d , _ d , _ d , which for low values of @xmath37 can become important and induce nucleon decay .
it is easy to check that they are also forbidden by the @xmath269 symmetry . *
unstable lsp * with the presence of @xmath63 parity violating couplings , the lsp - the lightest neutralino @xmath270 - is an unstable particle . in the scenario considered ,
the lsp three body decays mostly proceed due to the bi - linear ( [ bilinops ] ) coupling , and the lsp lifetime is ^-1_=_l^2z_^2 +^2 _ w+^4 _ w .
for the value @xmath271 gev ( dictated from the atmospheric neutrino scale ) we have @xmath272 sec .
therefore the lsp would be cosmologically irrelevant and some other candidate for cold dark matter should be found .
below the compactification scale @xmath62 the field content is just that of the mssm and the corresponding @xmath273 factors are ( b_1 , b_2 , b_3)= , 1 , -3 . above the @xmath62
scale the kk states enter into the renormalization . having kk excitations for all gauge and scalar superfields and also for @xmath274 families of bulk
matter , the @xmath275 factors , corresponding to the power law running ( [ dekkz ] ) , are @xmath276 from ( [ als])-([alg ] ) , taking into account ( [ dekkz ] ) , ( [ bmssm ] ) , ( [ kkmssm ] ) , we obtain @xmath277 @xmath278 @xmath279 where @xmath280 was taken since there is no intermediate scale . @xmath281 , @xmath282 are known with high precision and @xmath283 with some precision and thus , eq.([eq : a ] ) sets a constraint to the values of @xmath284 . with @xmath285 , @xmath286 and @xmath287
we get @xmath288 from eq.([eq : b ] ) we see that the _ unification _ scale is also constrained : @xmath289 from the definition ( [ dekkz ] ) of @xmath284 and ( [ sazs ] ) , ( [ sazmg ] ) a constraint to the values of @xmath180 , the number of kk levels , arises .
it is not difficult to see that @xmath290 are the only values of @xmath180 allowed for @xmath291 .
two examples of unification for these values of @xmath180 are n_0 , , , _ 3 = 1 , 1.3 , 1.710 ^ 16 , 0.117 , 2 , 2 , 1.310 ^ 16 , 0.119 .
the question may arise whether taking into account some threshold corrections will change the results or not .
in fact , the susy threshold corrections would introduce additional terms in ( [ eq : a])-([eq : c ] ) , which can be important for the predictions of @xmath198 .
these threshold corrections can be characterized by one threshold scale @xmath292 and change the strong coupling as @xcite @xmath293 , where @xmath294 is given in ( [ eq : a ] ) . for @xmath295 tev inequality ( [ sazs ] )
is modified to @xmath296 , which would change @xmath22 not more than by factor of @xmath129 .
as we see , there is no qualitative change , but scales can be modified slightly . because of this , throughout the paper we will not take into account this type of threshold corrections . apart from this , on the 4d fixed points @xmath3 susy invariant kinetic terms are allowed , which in general could alter the unification picture @xcite . however , if the size of extra dimension(s ) is relatively large @xmath297(@xmath298fundamental scale ) , then contributions from localized kinetic terms will be negligible @xcite .
the condition @xmath299 holds for the low scale unification scenarios considered below and is also crucial for avoiding unwanted effects from other brane operators ( see appendix a ) . *
low scale unification * by a look at the equations ( [ eq : a ] ) , ( [ eq : b ] ) we recognize that to have low scale unification it is important that the last term in eq.([eq : b ] ) is a large negative number , while the kk contributions in ( [ eq : a ] ) must be small or vanish .
thus , to have low scale unification we need some extension as to cancel the last term in ( [ eq : a ] ) and to keep the negative last term in ( [ eq : b ] ) . among a few possible extensions @xcite-@xcite , the simplest one seems to be the model of ref .
@xcite , where states @xmath300 ( @xmath301 ) were introduced .
the @xmath302 , @xmath303 states are singlets of @xmath160 , @xmath159 and carry hypercharges @xmath304 , @xmath305 , resp .
, in the @xmath306 units ( [ hipcharge ] ) . with the @xmath107 parities ( e^(1 ) , ^(2))~+ , ( e^(2 ) , ^(1))~- , only the @xmath307 , @xmath308 states will have zero modes .
the contribution of the @xmath309 states to the @xmath275 factors is then _ e ( _ 1 , _ 2 , _
3)= , 0 , 0 , while the @xmath273 factors corresponding to the logarithmic runnings get the additions _ e ( b_1 , b_2 , b_3)= , 0 , 0 .
taking these into account we have @xmath310 @xmath311 and @xmath312 where @xmath313 is the 4d mass of @xmath307 , @xmath308 written as a brane coupling . in this setting the kk modes do not contribute in ( [ eq:12 ] ) . to have a reasonable value for @xmath198 one
has to take @xmath314 .
the gut scale @xmath22 can be as low as we like . with @xmath315 , @xmath316
we have @xmath317 and from ( [ eq:13 ] ) we obtain @xmath318 tev . for @xmath319 in ( [ eq:14 ] )
the @xmath193 remains in the perturbative regime .
although the value of @xmath313 is much higher than @xmath62 , the form of the power law function @xmath284 of ( [ dekkz ] ) is not affected by the @xmath320 brane coupling . in appendix a we study the possible brane operator effects on rge and show that for @xmath321 they do not affect the expressions of ( [ dekkz ] ) , ( [ skk ] ) .
therefore the analysis carried out through eqs .
( [ eq:12])-([eq:14 ] ) remains valid .
we start our study of gut orbifold models with the 5d @xmath3 susy @xmath2 theory .
the fifth dimension is compact and is considered to be an @xmath7 orbifold .
two @xmath107 s are necessary to avoid extra zero mode states . in the notation of sect .
2 , the 4d @xmath3 gauge supermultiplet @xmath322 in @xmath17 terms splits up as v(24)=v_c(8 , 1)_0+v_su(2)_l(1 , 3)_0+v_s(1 , 1)_0 + v_x(3 , |2)_5+v_y(|3 , 2)_-5 , where subscripts are the hypercharge @xmath202 in the @xmath306 units of ( [ hipcharge ] ) . the decomposition of @xmath323 will be similar to ( [ dec24ofsu5 ] ) .
we ascribe to the fragments of @xmath322 and @xmath323 the following @xmath126 parities @xmath324 _c , _
su(2)_l , _
s~(- , - ) , _x , _
y~(+ , - ) .
with this assortment all the couplings in ( [ lagv ] ) remain invariant . from the @xmath71 susy
@xmath2 gauge supermultiplet only the @xmath3 gauge superfields of @xmath17 have zero modes .
therefore , at the @xmath117 fixed point ( brane ) we have a @xmath3 susy @xmath17 gauge theory .
the other states , including the @xmath325 , @xmath202 gauge bosons which would induce @xmath27 nucleon decay , are projected out . to have two mssm higgs doublets , one should introduce two @xmath71 supermultiplets @xmath326 , @xmath327 where @xmath60 , @xmath328 are @xmath242-plets of @xmath2 . in terms of @xmath17 h(5)=h_u(1 , 2)_-3+t ( 3 , 1)_2 , ( |5)=h_d ( 1 , |2)_3+(|3 , 1)_-2 , and similarly for @xmath328 , @xmath329 .
@xmath330 and @xmath329 are mirrors of @xmath60 and @xmath328 resp . with the following assignment of orbifold parities ( h_u , h_d)~(+ , + ) , ( h_d , h_u)~(- , - ) , ( t , )~(- , + ) , ( , t)~(+ , - ) , the states @xmath35 , @xmath331 have zero modes , which we identify with one pair of mssm higgs doublets .
as we can see , all colored triplet states are projected out and therefore will not participate in the @xmath29 nucleon decays .
all couplings in ( [ lagf ] ) are invariant under the @xmath126 symmetry except the @xmath332 type couplings , which thus are not allowed .
this means that on the 5d level the @xmath333 coupling is absent due to @xmath71 susy , while @xmath212 and @xmath334 are not allowed by the orbifold parities . concerning the matter sector in @xmath2 we have anomaly free @xmath335 , @xmath336 multiplets , one per generation .
if at the level of the 5d susy theory we wish to introduce them as bulk fields , we should embed them into the @xmath71 matter supermultiplets . per generation
we then have @xmath337 and @xmath338 , where @xmath339 and @xmath242 are mirrors of @xmath335 and @xmath336 resp . in terms of @xmath17
this reads 10=e^c(1 , 1)_-6+q ( 3 , 2)_-1+u^c ( |3 , 1)_4 , |5=l ( 1 , |2)_3+d^c ( 1 , mirrors and to project them out , one can easily realize that due to the parities ( [ z2z2ofv24 ] ) of the gauge fields , some of the states in ( [ matterdec ] ) will not have zero modes . to overcome this difficulty one can introduce copies ( @xcite and first refs . in @xcite ) @xmath340 , @xmath341 where exactly those states are allowed to have zero modes which correspond to the mssm states which come from @xmath342 and @xmath343 and
are projected out . with orbifold parity prescriptions q , l , u^c , d^c , e^c ~(+ , + ) , q , l , u^c , d^c , e^c ~(- , + ) , and opposite ones for the corresponding mirrors , it is easy to verify that now the scenario is compatible with the bulk construction since all terms ( except the mass term ) of ( [ lagf ] ) are invariant .
an alternative possibility would be to introduce fermionic states only on the @xmath117 fixed point brane .
this case can appear in string models with intersecting branes @xcite .
thus , in general one can have @xmath344 of generations living at the brane only and @xmath274 generations living also in the bulk .
for the latter case we have to introduce @xmath274 copies .
this applies not only for the @xmath2 model , but also for the other scenarios considered below .
* fixed point brane couplings at @xmath117 * and some phenomenology at the @xmath117 fixed point we are left with a susy @xmath17 gauge theory with states @xmath203 , @xmath204 , @xmath345 , @xmath346 , @xmath347 , @xmath35 , @xmath331 , which is just the field content of the mssm . since the 5d action does not provide any yukawa couplings , we should write appropriate couplings on the brane . the 4d yukawa superpotential , responsible for the generation of up - down quark and lepton masses , has the form w_y = qu^ch_u+qd^ch_d+le^ch_d .
since states with primes and without primes come from different unified multiplets , in this case we do not have any asymptotic relations between fermion masses .
this avoids the problem * ( ii ) * of fermion masses which exists in the minimal @xmath2 gut .
also , since the colored triplets are projected out , the dt splitting problem * ( iii ) * as well as the problem * ( i ) * caused by @xmath29 colored triplet exchange nucleon decay do not exist any more .
however , the problem due to general @xmath29 baryon number violating operators and the @xmath0 problem are still unresolved on the 4d level ( as in the case of the mssm ) unless some additional mechanism is applied .
one way for resolving these problems is to impose a continuous @xmath63 symmetry @xcite , @xcite , which in orbifold constructions emerges on the 4d level after compactification as an @xmath348 symmetry @xcite .
the latter can guarantee baryon number conservation , a suppressed @xmath0 term and automatic matter parity . here
and throughout the paper , alternatively , we impose a discrete @xmath11 symmetry which allows some matter parity / lepton number violating operators responsible for the generation of appropriately suppressed neutrino masses . in the spirit of sect .
4 , we also introduce @xmath148 , @xmath149 singlets . with the transformations
( [ ssym ] ) for @xmath349 and for doublets ( [ mutermmssm ] ) ( here in all couplings @xmath36 must be replaced by @xmath331 ) the relevant couplings will be precisely the same as in ( [ supmumssm ] ) and ( [ sups ] ) . through the couplings in ( [ pots ] ) the vev in ( [ sols ] ) is obtained and consequently a @xmath0 term of the ( [ valmuterm ] ) right magnitude is generated .
since the minimal susy @xmath2 does not involve right handed neutrino states the neutrinos are massless . to give them mass , we also include , together with the yukawa couplings ( [ yuksu5 ] ) , the lepton number violating bi - linear coupling ( [ bilins ] ) , which ( as discussed in sect .
4 ) induces the neutrino mass ( [ nubilins ] ) . for @xmath350 , @xmath351
according to ( [ bilinops ] ) we obtain @xmath246 ev , a desirable value to explain the atmospheric anomaly . with the couplings in ( [ yuksu5 ] ) , ( [ supmumssm ] ) , ( [ sups ] ) , ( [ bilins ] ) and taking into account the unified @xmath2 multiplets on 5d level we have @xmath352 @xmath353 ( l)=-k ( s)-(h_u)+ , ( q)=-k ( s)-3(h_u)+2 , = , where @xmath354 , @xmath255 are undetermined .
this still allows the lepton number violating couplings ( [ trilins ] ) , giving for @xmath350 , @xmath355 a radiative neutrino mass @xmath356 ev [ see eq .
( [ numastrilin ] ) , ( [ trilins1 ] ) ] .
this is the relevant scale for explaining the solar neutrino puzzle through the large angle vacuum oscillations of the @xmath357 state into the @xmath358 .
as we will see in the next subsection , the scale of unification is close to @xmath359 gev . assuming that the cutoff scale @xmath37 has the same magnitude for @xmath360 in ( [ sols ] ) we need @xmath361 . if in the eqs .
( [ zphases ] ) one takes @xmath350 , @xmath361 , @xmath362 , @xmath363 the discrete symmetry would be @xmath364 and one can verify that baryon number violating @xmath29 operators @xmath365 , @xmath366 are forbidden .
also , all other @xmath63-parity and baryon number violating couplings are absent in this scenario .
below the compactification scale @xmath62 , we have precisely the mssm field content with the @xmath273 factors given in ( [ bmssm ] ) , while above @xmath62 , due to the @xmath126 parities of the states given in ( [ z2z2ofv24 ] ) , ( [ z2z2higgs ] ) and for @xmath274 generations of ( [ z2z2fer ] ) in the bulk , we have @xmath367 _1 , _ 2 , _
3= - , -6 , -2 + 4(1 , 1 , 1 ) . from this and ( [ als])-([alg ] ) , and taking into account ( [ dekkzz1 ] ) , we get _ 3 ^ -1=_2 ^ -1-_1 ^ -1- ( s_1-s_2 ) , = ( _ 1 ^ -1-_2 ^ -1)- ( s_1-s_2 ) , _ g^-1=_2 ^ -1-+ s_1+s_2 .
( here we do not have an intermediate scale and we take @xmath280 . ) we see that contributions to ( [ susysu5als ] ) , ( [ susysu5scale ] ) from the power law functions @xmath182 , @xmath183 [ defined in ( [ skk ] ) ] are canceled out in the limit @xmath368 .
this is understandable , since in this limit the @xmath2 symmetry is restored above the @xmath62 scale and there are only contributions from complete @xmath2 multiplets [ according to ( [ bkksu5 ] ) @xmath369const . ] . to have a reasonable value for @xmath198
one needs @xmath370 , which means @xmath371 leading to @xmath372 .
because of this , from ( [ susysu5scale ] ) we get @xmath26 gev .
the value of @xmath193 in ( [ susysu5alg ] ) remains perturbative for @xmath373 .
thus in contrast to the 5d susy @xmath17 scenario , it is impossible to get low scale unification within the orbifold @xmath2 scenario .
in the previous section we have seen that within orbifold @xmath2 gut a power law unification does not take place .
although above the @xmath61 scale each coupling of @xmath17 has power law running , the renormalization of their relative slope ( e.g. running of @xmath374 ) is still logarithmic because above @xmath62 the full @xmath2 multiplets participate in renormalization .
because of this , one does not get low scale unification
. this result would be the same for any higher dimensional orbifold gut scenario with semisimple gauge group [ such as @xmath18 , @xmath375 , @xmath19 , @xmath376 , @xmath377 ] if the compactification of all extra dimensions occurs at a single mass scale .
then representations of all gauge groups listed above again can be decomposed to complete @xmath2 multiplets .
low scale unification within gut models with orbifold extra dimensions is however possible if we allow for compactifications of various extra dimensions at different mass scales .
suppose , we have a gut model with gauge group @xmath59 and with two extra spacial dimensions , and assume compactification in two steps with scales @xmath378 with a symmetry breaking chain g _ 1 .
for having power law unification it is crucial that @xmath379 must be different from @xmath2 and also must not contain it as a subgroup
. then the field content of @xmath379 , relevant between @xmath380 and @xmath381 , will not constitute full @xmath2 multiplets .
this would give us the possibility of low scale unification .
the bottom - up picture of such a scenario looks as follows : at an energy scale @xmath382 the gauge group @xmath383 is restored to @xmath379 and above @xmath62 kk states in incomplete multiplets give power law differential running ( of @xmath374 ) .
the gap @xmath378 must be big enough to reduce the intermediate scale @xmath62 [ see eq .
( [ scale ] ) , which gives a low intermediate scale @xmath162 in case of large @xmath284 ( or @xmath182 , @xmath183 ) and a negative coefficient ; a similar expression can be derived for scale @xmath62 in case of several compactification mass scales ] .
scale @xmath380 is close to @xmath22 : note , that the case with @xmath384 will not work because above @xmath380 the unification group @xmath59 is restored and its full multiplets would be alive. in order to realize this idea , the gauge group @xmath59 _ a ) _ must be higher ( in rank ) than @xmath2 ; _ b ) _ should have subgroups different from @xmath2 and _ c ) _ its subgroups must give a realistic phenomenology , e.g. they should contain the @xmath17 gauge group and mssm states .
one of the groups , which has these properties , is @xmath18 .
its maximal subgroups are @xmath4 and flipped @xmath5 .
it is straightforward to study compactification breaking of these groups and to see what is going on above the scale @xmath382 .
since for power law unification the region between @xmath381 and @xmath385 is relevant , we can consider five dimensional @xmath8 and @xmath9 orbifold models .
as we will see in the following sections this type of bottom - up approach turns out to be quite transparent and convenient for studying various phenomenological issues together with gauge coupling unification .
in the following we consider a supersymmetric @xmath65 model in five dimensions ( 5d ) . @xmath387 the adjoint of @xmath388 , in @xmath389 terms reads v(15_c)=v_s(1)_0+v_c(8)_0+v_t(3)_4+v_|t(|3)_-4 , where subscripts denote @xmath390 charges in @xmath391 units ( @xmath388 normalization ) : y_u(1)=(1 , 1 , 1 , -3 ) .
the decomposition of @xmath392 is identical .
the decomposition of the @xmath93 s adjoints through the @xmath393 channel has the form v(3_r)=v_r(1)_0+v_p(1)_2+v_|p(1)_-2 , and the same for @xmath394 .
here subscripts denote @xmath348 charges in @xmath395 units : y_u(1)_r=(1 , -1 ) .
[ [ matter - sector . ] ] matter sector .
+ + + + + + + + + + + + + + we introduce @xmath274 generations of @xmath71 chiral supermultiplets _ n=2=(f , ) , * f^c*_n=2=(f^c , ^c ) , where under @xmath8 f~(4 , 2 , 1 ) , f^c~(|4 , 1 , 2 ) , f=(q , l ) , f^c=(u^c , d^c , ^c , e^c ) , and also @xmath274 generations of copies _ n=2=(f , ) , * f^c*_n=2=(f^c , ^c ) . with
precisely the same content and transformation properties as @xmath396 and @xmath397 resp .
the remaining @xmath344 generations at the fixed point have the same massless field content as the @xmath274 bulk generations . note that the introduction of copies is crucial if one wants a @xmath181 orbifold invariant 5d action , with matter both in the bulk and on a fixed point , which reduces at low energies to the chiral content of the mssm , extended only by right handed neutrino states .
[ [ scalar - sector . ] ] scalar sector .
+ + + + + + + + + + + + + + we need two sets of scalars .
first we introduce hypermultiplets , which will contain the two higgs doublet superfields of the mssm .
_ n=2= ( , ) , * * _ n=2=( , ) , where under @xmath8 ~(1 , 2 , 2 ) , ~(1 , |2 , |2 ) . @xmath78 and
@xmath77 have the field content = ( h_u , h_d ) , = ( h_u , h_d ) . @xmath35 and
@xmath36 have the same quantum numbers as the mssm higgses responsible for the generation of up and down quark masses resp .
@xmath398 is a copy of @xmath399 . and its introduction is crucial for having vectorlike higgs content , keeping the theory anomaly free . to break the @xmath8 symmetry we use orbifold compactification
since this breaking does not reduce the rank of the group we should introduce some additional scalars to provide for the needed rank breaking via higgs mechanism .
the model with minimal field content thus possesses also the following two @xmath71 supermultiplets _ n=2=(h^c , h^c ) , * h*_n=2=(h^c , ^c ) , where h^c~(|4 , 1 , 2 ) , ^c~(4 , 1 , |2 ) , h^c=(u^c_h , d^c_h , ^c_h , e^c_h ) , ^c=(^c_h , ^c_h , ^c_h , ^c_h ) , and @xmath400 is a copy of @xmath401 . in this subsection
we will show how the @xmath8 symmetry can be broken down to the standard model gauge group @xmath17 . by a special selection of boundary conditions on the @xmath54 orbifold , at a first stage @xmath8
symmetry can be broken down to the @xmath404 symmetry ( where @xmath390 and @xmath348 come from @xmath388 and @xmath93 resp . ) . with the @xmath181 parities presented in table [ t:422gut ]
, the @xmath8 symmetry reduces to @xmath405 and @xmath71 susy reduces to @xmath3 susy . on the fixed point @xmath117
we have only states with @xmath181 parities @xmath406 . we therefore have three generations of @xmath407 , and two mssm higgs doublets @xmath35 , @xmath331 .
in addition , there are extra scalar supermultiplets @xmath408 , @xmath409 , @xmath410 , @xmath411 .
to break @xmath412 down to the standard @xmath158 we use the states @xmath408 , @xmath409 , which have zero @xmath158 hypercharge . by developing non zero vevs along their scalar components
they induce the desired breaking . with @xmath413 , @xmath414
we have unbroken y =- y_u(1)+y_u(1)_r , where @xmath202 is given in @xmath415 in the standard @xmath2 normalization .
the superposition orthogonal to ( [ suppos ] ) corresponds to the hypercharge of the broken rank .
@xmath416 below we use a mechanism similar to the one we used in the mssm and @xmath2 cases , to generate the @xmath413 and @xmath417 vevs . here , to solve the various phenomenological issues
, we do not need to introduce the singlets @xmath148 , @xmath149 since their role will be played by @xmath408 and @xmath418 .
we introduce a @xmath11 symmetry under which the combination @xmath419 ( @xmath405 invariant ) has the same transformation as the @xmath349 in ( [ ssym ] ) .
then the relevant soft breaking terms and consequently the whole potential is precisely the same as in ( [ sups ] ) , ( [ softspot ] ) , ( [ pots ] ) , but now with @xmath148 and @xmath149 replaced by @xmath408 and @xmath418 . also , as in ( [ sols ] ) , the solutions @xmath420 , @xmath417 with parameterization ^c_h~v , will be v~m^ . in this way the @xmath412 symmetry breaking scale @xmath421 is expressed by the interplay of the cutoff scale @xmath37 and the susy mass scale @xmath24 , and the magnitude of @xmath421
is controlled by the discrete @xmath11 symmetry , i.e. by @xmath152 .
depending on the scenario one considers , one can select @xmath152 in such a way as to obtain a reasonable ratio @xmath422 .
for instance for @xmath423 , @xmath424 tev , @xmath425 gev we get @xmath426 , which is indeed a desirable value ( see first row of table [ t : iunif422 ] which corresponds to model * i*-susy422 ) . for low values of @xmath37 ,
there are no large mass gaps , and there is no need for large @xmath152 s : e.g. for @xmath427 gev , @xmath428 tev , and @xmath235 we have @xmath429 .
also this case can be realized with successful unification of gauge couplings ( see rows 1 - 3 of table [ t : iiiunif422 ] , corresponding to the model * iii*-susy422 ) . to avoid the 4d superpotential coupling @xmath430 we postulate the transformation property ( [ mutermmssm ] ) , where @xmath36 is to be replaced by @xmath331 .
the coupling responsible for the @xmath0 term generation then is w_=m^n-1h_uh_d , and after substituting the vevs of the @xmath408 , @xmath418 states , taking into account ( [ nusol ] ) , we get ~m^2n-2~m . for the time
being the introduced discrete symmetry is acting as @xmath214 on the field combinations @xmath419 , @xmath333 while the transformation properties of the single fields were not specified .
the phases of the @xmath11 transformations for the introduced states are given in table [ t : z422 ] .
the 4d yukawa superpotential generating charged fermion masses reads @xmath431 w_y = qu^ch_u+qd^ch_d+e^clh_d , where family indices are suppressed . for the dirac and majorana couplings of neutrinos
we have w_=_^k ^clh_u+ ( ^c_h^c)^2 , where @xmath266 is some integer , the value of which is dictated by the model considered , depending on what suppressions for neutrino masses are needed . after substituting the appropriate vevs and integrating out the @xmath432 states ( with masses @xmath433 ) , we obtain neutrino masses m__^2 ^4k-4 . for @xmath434 , @xmath435 , @xmath232 gev
this gives @xmath436 ev . in the case
@xmath350 , @xmath437 it is possible to reduce the cut off scale even down to the multi tev : for @xmath234 tev and @xmath438 we get @xmath439 ev .
we will see below that both cases @xmath440 can be realized and give successful pictures of unification ( see models * i*-susy422 and * iii*-susy422 in sect .
in order to accommodate recent atmospheric data with neutrino mass scales @xmath441 ev , one might assume neutrino species with degenerate masses @xcite , e.g. @xmath442 ev so that neutrinos are candidates for dark matter .
an alternative solution of the atmospheric and solar neutrino puzzles could be provided by a hierarchical structure of masses , which requires @xmath443 ev . this scale can be obtained having @xmath444 in ( [ numass422 ] ) . a stronger suppression for the first two neutrino generations can be achieved by introducing some flavor symmetries in the spirit of @xcite .
matter @xmath63-parity violating operators _ h h_u l , ^c_h q d^c l , ^c_h u^c d^c d^c , ^c_h
e^c l l , which are invariant under @xmath405 , are forbidden by the @xmath364 symmetry with the charge selections given in table [ t : z422 ] , for @xmath435 or @xmath445 .
the @xmath29 baryon number violating operators of the type presented in ( [ d5mssm ] ) and also @xmath446 ( this operator violates also @xmath63 parity and leads to a @xmath29 baryon number violating coupling after the substitution of vev @xmath417 ) are forbidden in this scenario for both choices @xmath435 and @xmath350 .
there are also @xmath27 ( [ d6mssm ] ) type operators allowed by @xmath405 symmetry and in addition the @xmath447 coupling .
it is easy to check that these are absent due to the @xmath364 symmetry . in order that the 5d lagrangian terms of ( [ lagv ] ) , ( [ lagf ] ) , allowed by the @xmath126 orbifold parities , are invariant under the introduced @xmath364 discrete symmetry
, we must assure that the other fields transform properly .
this is the case if @xmath448 , @xmath449 , @xmath450 , @xmath451 , @xmath452 , @xmath453 , @xmath454 , @xmath455 ( and all mirrors with opposite phases ) .
since the states @xmath410 and @xmath411 transform as @xmath408 and @xmath456 resp .
, their mass term is generated through an operator @xmath457 and one gets @xmath458 .
first of all we must make sure that these triplet states do not cause nucleon decay .
the allowed couplings of @xmath410 , @xmath411 with matter are @xmath459 and @xmath460 for @xmath435 , while for @xmath350 the operators @xmath461 and @xmath462 are permitted .
however , the couplings @xmath463 and @xmath464 are forbidden ( see table [ t : z422 ] ) and the baryon number violating @xmath29 operators @xmath365 , @xmath465 do not emerge .
the issue of gauge coupling unification in this model , which we call * i*-susy422 , will be studied below .
as it turns out , successful unification can be obtained for various scales as presented in table [ t : iunif422 ] . for the case shown in the first row
we obtain @xmath426 ( obtained for @xmath423 according to ( [ nusol ] ) ) .
this mass gap is crucial for @xmath0-term generation with the correct magnitude . for this case
we have @xmath466 tev .
the existence of colored triplet states with this mass can have interesting phenomenological implications @xcite .
there might be a leptoquark like signature @xcite , similar to what is expected within some @xmath63-parity violating models .
a different scenario , with heavy @xmath410 , @xmath411 states , can be constructed introducing additional two @xmath71 supermultiplets @xmath467 ( @xmath301 ) of @xmath388 , where 6^(i)=t^(i)(3_-2)+^(i)(|3_2 ) , parity assignment @xmath468 ( t^(1 ) , ^(2))~(+ , - ) , ( t^(2 ) , ^(1))~(- , - ) , the triplet - antitriplet pair @xmath469 , @xmath470 will have zero modes and can therefore couple with the @xmath410 , @xmath411 giving them large masses . with @xmath364 phases @xmath471 ,
@xmath472 the relevant 4d superpotential couplings will thus be w_t=^c_hd^c_ht^(1)+ ^c_h^c_h^(2 ) . after substituting the vevs of @xmath408 , @xmath418 ,
the triplet states acquire masses @xmath473
. the allowed couplings of the @xmath469 , @xmath470 states with matter , are @xmath474 , @xmath475 .
however , the couplings @xmath476 , @xmath477 are forbidden by @xmath364 symmetry and baryon number is still conserved .
we refer to this model as * ii*-susy422 .
also in this case successful unification of gauge couplings occurs if @xmath478 ( see sect . 7.1.1 ) .
however , as we will see , with a specific extension it is possible to get unification near the multi tev region ( see table [ t : iiiunif422 ] for the model * iii*-susy422 , which presents mass scales for which unification holds ) . for this case , since triplets get masses @xmath479 through the couplings ( [ tripsup ] ) , their masses are a few tev , making this scenario testable in future collider experiments .
we conclude this section by noting that , together with a natural @xmath480 breaking pattern and @xmath0-term generation , the @xmath364 symmetry provides automatic @xmath63-parity and baryon number conservation within the 5d susy orbifold @xmath8 model . here
we will study the issue of gauge coupling unification for susy @xmath8 model with compactification breaking to @xmath405 . throughout this analysis
we will use the expressions obtained in section 3 .
* model i - susy422 * the field content of this scenario is as follows .
we have the scalar superfields of ( [ bidoublets ] ) , ( [ brscalars ] ) , which are necessary to obtain the pair of mssm higgs doublets and to realize the wanted @xmath405 breaking to @xmath17 .
we also have @xmath274 generations of @xmath86 , @xmath482 presented in ( [ matter ] ) , and @xmath274 copies , if @xmath274 generations of matter have kk excitations .
we then identify the scale of @xmath483 symmetry breaking @xmath484 with the intermediate scale @xmath162 in ( [ rge321 ] ) , ( [ match ] ) . below @xmath162 ,
the gauge group is @xmath17 and the field content is that of the mssm with the @xmath273-factors ( [ bmssm ] ) , plus the states @xmath410 , @xmath411 with a mass @xmath485 in the range @xmath486 gev-@xmath441 tev , which have @xmath273-factors b_i^d^c_h = , 0 , 1 . above the scale @xmath162
we have b_u(1)_r , b_u(1) , b_2 , b_3^m_i = 9 , 7 , 1 , -2 . with the @xmath181 parities shown in table [ t:422gut ] ,
the corresponding @xmath487 and @xmath488-factors of ( [ dekkzz1 ] ) , will be @xmath489 _u(1)_r , _
u(1) , _ 2 , _ 3= 2 , -6 , 2 , 0 + 41 , 1 , 1 , 1 . due to ( [ suppos ] ) , which determines the pattern of @xmath158 embedding in @xmath483 , the group - theoretical factor @xmath167 in ( [ match ] ) , ( [ bsupg1g2 ] ) , ( [ kksupg1g2 ] )
will be @xmath490 if @xmath491 and @xmath492 . taken this into account , using ( [ match])-([delog ] ) , ( [ dekkzz1 ] ) , ( [ kksupg1g2 ] ) , it is not difficult to derive from ( [ als])-([alg ] ) the following equations _ 3 ^ -1=_2 ^ -1-_1 ^ -1 + + + s_1- s_2 , = ( _ 1 ^ -1-_2 ^ -1)- - - s_1+s_2 , _
g^-1=_2 ^ -1-+ s_1-s_2 . without the four
last terms in ( [ modelals ] ) the one loop value of @xmath198 would be @xmath493 , which is close to the experimental value of the strong coupling @xmath23 . therefore , the contribution of the remaining terms should not be large . since @xmath182 and @xmath183 have nearly the same values , the sum of the last two terms in ( [ modelals ] ) will be negative and this negative number must be compensated by the third and fourth term by a proper choice of the mass scales . in ( [ modelscale ] )
the last four terms give a negative contribution .
this gives the possibility to have a relatively low scale @xmath162 .
this is realized for @xmath494 .
the latter is crucial for gauge constant s perturbativity . within all models with large gap between @xmath62 and @xmath22
scales we will take @xmath319 , i.e. matter is located at the fixed point ( in agreement with string models with intersecting branes @xcite ) .
mass scales , which give successful unification , are presented in table [ t : iunif422 ] .
@xmath186 , @xmath187 are the maximal numbers of even and odd kk states resp . , which lie below @xmath22 and are determined from the inequalities ( [ maxns ] ) .
the picture of unification for the @xmath495 case of table [ t : iunif422 ] is presented in fig . 1 .
@xmath496 & \log_{10}[\mu_0 ] & \log_{10}[m_g ] & \log_{10}[m_i / m_{d^c_h } ] \\
\hline \hline 30 & ( 14 , ~14 ) & 10.93 & 11.15 & 12.63 & 6.92 \\
\hline 28 & ( 13 , ~13 ) & 11.5 & 11.5 & 12.94 & 6.75 \\ \hline 26&(12 , ~13 ) & 11.64 & 11.64 & 13.08 & 6.5 \\ \hline \end{array}\ ] ] * model ii - susy422 * in this scenario we introduce two @xmath497 supermultiplets with the components shown in ( [ 6pl ] ) and with @xmath181 parities shown in ( [ trippar ] ) . as we will see , with only this extension , it is impossible to get unification with a mass gap between @xmath162 , @xmath62 and @xmath22 , as well as relatively low @xmath22 .
but this can be achieved with a simple additional extension ( model * iii*-susy422 ) . due to the couplings ( [ tripsup ] ) , all triplet zero mode states decouple on the @xmath498 scale and consequently , below @xmath162 , the gauge coupling runnings will be precisely the same as in the mssm , with @xmath273-factors ( [ bmssm ] ) . due to the presence of the @xmath469 , @xmath470 states ,
the @xmath273-factors will be modified above @xmath162 and ( [ b3211 ] ) will be changed by _ * 6 * b_u(1)_r , b_u(1) , b_2 , b_3^m_i = 0 , 1 , 0 , 1 .
the @xmath487 and @xmath488 factors in ( [ bkk ] ) are modified by _ * 6 * ^ _ u(1)_r , ^ _ u(1) , ^ _ 2,_3= _ * 6 * _u(1)_r , _
u(1) , _ 2 , _ 3=0 , 2 , 0 , 2 . (
the subscript *6* in @xmath499 indicates that the changes are due to states coming from the two @xmath500 supermultiplets . )
taking all this into account , we have _ 3 ^ -1=_2 ^ -1-_1 ^ -1 + + s_1-s_2 , = ( _ 1 ^ -1-_2 ^ -1)- -s_1+s_2 . in ( [ iisusy422als ] )
we see that the contribution of the last three terms is always positive and for reasonable @xmath198 the only possibility is to have @xmath501 . due to this fact , from ( [ iisusy422scale ] )
one can see that @xmath502 gev .
* model iii - susy422 : low scale unification * here we present an extension which gives low scale unification .
in addition to the field content of model * ii*-susy422 we introduce two @xmath503 ( @xmath504 ) , which are supermultiplets of @xmath159 . with @xmath181 parities ( d^(1 ) , ^(2))~(+ , + ) , ( d^(2 ) , ^(1))~(- , - ) , only the states @xmath505 , @xmath506 have zero modes .
the contributions to the @xmath507 , @xmath487 , and @xmath488-factors , due to @xmath508 supermultiplets , are @xmath509 _ * 2*_u(1)_r , _
u(1) , _ 2 , _ 3 = ( 0 , 0 , 2 , 0 ) , _ * 2*_i=0 , and consequently we obtain _ 3 ^ -1=_2 ^ -1-_1 ^ -1 + - + ( s_1-s_2 ) , = ( _ 1 ^ -1-_2 ^ -1)- + -s_1+s_2 , _
g^-1=_2 ^ -1 - - -s_1-s_2 , where @xmath510 is the mass of the zero mode of the doublet states @xmath505 , @xmath511 , which arises from the 4d superpotential coupling @xmath512 . from ( [ lowiisusy422als ] ) we see that for @xmath368 the contribution from kk states cancels out . since @xmath182 , @xmath183 differ slightly , the cancellation is partial and for a desirable value of @xmath23 , appropriate contributions from the logarithmic terms are needed . now
, the contribution from kk states in ( [ lowiisusy422scale ] ) is always negative ( there is no possible cancellation ) .
thus it is possible to get low scale unification in this scenario .
the values of the mass scales , which give a successful picture of unification , are presented in table [ t : iiiunif422 ] .
as one can see , for @xmath513 the couplings unify at the scale @xmath514tev ; the unified coupling constant @xmath193(@xmath515 ) is perturbative .
the picture of unification for case @xmath516 of table [ t : iiiunif422 ] is presented in fig .
2 . note , that in these cases the mass of the doublet pair @xmath505 , @xmath511 is @xmath517gev .
this makes the model testable in future collider experiments .
@xmath518 & \log_{10}[\mu_0 ] & \log_{10}[m_g ] & \log_{10}[m_g / m_d]&\al_3(m_z ) \\ \hline \hline 110 & ( 54 ,
~54 ) & 3.26 & 3.26 & 5.30 & 3.21 & 0.1184 \\ \hline 108 & ( 53 , ~53 ) & 3.48 & 3.48 & 5.52 & 3.21 & 0.1185 \\ \hline 107 & ( 52 , ~53 ) & 3.6 & 3.6 & 5.62 & 3.21 & 0.1186 \\ \hline 106 & ( 52 , ~52 ) & 3.71 & 3.71 & 5.73 & 3.21 & 0.1186 \\
\hline \end{array}\ ] ] in this subsection , similar to 7.1 , we will consider the breaking of @xmath8 by orbifold compactification down to @xmath520 .
the breaking of the latter to @xmath17 again will occur on the 4d level through the non - vanishing vevs of certain fields .
the decomposition of @xmath521 ( @xmath392 ) under @xmath522 is given in ( [ su4dec ] ) .
the gauge group @xmath523 is , in this case , not broken by the compactification .
the matter sector is the same as in ( [ matter ] ) with copies ( [ mattercopy ] ) needed if chiral states are introduced in the bulk .
the content of @xmath86 , @xmath482 from the viewpoint of @xmath524 is @xmath525 where @xmath526 for the copies we have similar expressions . to obtain one pair of mssm higgs doublets , it is enough in this case to have only one @xmath71 supermultiplet @xmath527 ( see ( [ bidoublets ] ) ) with field content as in ( [ decdoubl ] ) , ( [ phicont ] ) . for further @xmath528 breaking to @xmath158
we need the states ( [ brscalars ] ) with components ( [ decbrscal ] ) , ( [ brscalcont ] ) where now under @xmath524 @xmath529 decomposes as h^c = q^c_h(|3 , 1 , 2)_-1+l^c_h(1 , 1 , 2)_3 and similarly for @xmath328 , @xmath530 , @xmath531 .
the @xmath126 parities and @xmath390 charges of the appropriate fragments are given in table [ t : lr3 ] .
with these parity assignments , @xmath8 is broken to @xmath524 and the zero mode matter and scalar superfields are @xmath203 , @xmath532 , @xmath533 , @xmath534 , @xmath78 , @xmath535 , @xmath536 .
@xmath537 the @xmath538 breaking occurs through @xmath535 , @xmath536 states , after their neutral scalar components @xmath408 , @xmath418 have developed non - zero vev s . as for the models considered above , here
we also introduce a discrete @xmath11 symmetry .
the combination @xmath539 transforms similarly to ( [ ssym ] ) and the relevant couplings will be ( [ sups])-([pots ] ) but with @xmath148 , @xmath149 replaced by @xmath535 , @xmath536 .
thus , @xmath408 and @xmath418 which are components of @xmath535 and @xmath536 resp . will have non - vanishing vev s ( [ nuvevdef ] ) , ( [ nusol ] ) .
this provides the symmetry breaking down to @xmath17 at an intermediate mass scale between the susy scale @xmath24 and the unification scale @xmath37 . as a first example
suppose that @xmath540gev in which case @xmath541gev . with @xmath542
we get @xmath543 , the value needed to have successful high scale unification in the * i*-susy422 model ( see below ) . on the other hand
if the unification scale is as small as @xmath544gev , with @xmath235 we obtain @xmath545gev as needed in low scale unification scenarios . with @xmath78 transforming under @xmath546 as @xmath547 the coupling responsible for the @xmath548-term generation
is @xmath549 and we obtain the same @xmath0 term as in ( [ muterm ] ) .
the 4d yukawa superpotential , responsible for generation of up - down quark and charged lepton masses , is @xmath550 from this coupling the dirac type coupling for the neutrinos is also generated and @xmath551 .
there is therefore no additional suppression of the dirac neutrino masses . for a low fundamental scale
this turns out to be a problem . to overcome this difficulty we can introduce an additional singlet state @xmath552 .
the relevant couplings are @xmath553 due to the last term in ( [ nu4221model ] ) the state @xmath432 decouples at the scale @xmath554 and the first term , which violates lepton number , is suppressed by appropriate powers of @xmath555 .
this can lead to a neutrino mass generation of the needed magnitude .
it is not difficult to verify that together with the @xmath556 coupling , the operators in ( [ nu4221model ] ) give m_~^2k .
note that we consider @xmath557 and therefore for @xmath558 one can obtain the needed suppression . for instance , with @xmath559 , @xmath560 and @xmath561 gev , one can have @xmath439 ev . here
we address the question of gauge coupling unification for @xmath8 with compactification breaking to @xmath524 .
* model i-susy422 * the field content of this scenario is minimal : @xmath71 gauge supermultiplets , scalar superfields , and @xmath274 generation of matter in the bulk .
all these states , their @xmath390 charges , and @xmath126 parities are presented in table [ t : lr3 ] .
the corresponding @xmath507 , @xmath52 , and @xmath563 factors are @xmath564 @xmath565 @xmath566 on the scale @xmath162 we have matching conditions ( [ match ] ) , ( [ bsupg1g2 ] ) , where @xmath567 , @xmath492 and @xmath568 .
taking all this into account , from ( [ als])-([alg ] ) we obtain _
3 ^ -1= _ 2 ^ -1-_1 ^ -1- -(2s_1 - 3s_2 ) , = ( _ 1 ^ -1-_2 ^ -1)-- ( 4s_1+s_2 ) , _
g^-1=_2 ^ -1-+ s_1-s_2 . in table
[ t : lr5 ] we present several solutions of the above equations .
the experimental lower bound @xmath569gev for the @xmath93 symmetry breaking scale @xcite , @xcite puts a bound also on the unification scale @xmath570gev .
the picture of unification for the case in the last row of table [ t : lr5 ] is presented in fig .
3 . @xmath571 } & \log_{10}{[\mu_0 ] } & \log_{10}{[m_g ] } \\
\hline \hline 50 & ( 24,~24 ) & 5.40 & 10.32 & 12.02 \\ \hline 51 & ( 24,~25 ) & 5.18 & 10.20 & 11.91 \\ \hline 55 & ( 26,~27 ) & 4.31 & 9.75 & 11.49 \\ \hline 56 & ( 27,~27 ) & 4.10 & 9.64 & 11.39 \\ \hline \end{array}\ ] ] * model ii-susy422 : low scale unification * in order to have low scale unification , as for the model * iii*-susy422 , we introduce two @xmath572 plets of @xmath388 ( i=1 , 2 ) with decomposition ( [ 6pl ] ) , and two @xmath573 of @xmath159 ( @xmath504 ) .
in addition we introduce one bi - doublet @xmath398 state of ( [ bidoublets ] ) .
we will take @xmath126 parities ( [ trippar ] ) and ( [ 2par ] ) for fragments from @xmath572 and @xmath508 , resp . , while for the fragments of @xmath398 we take the parities @xmath574 , @xmath575 .
the contributions of these states to the @xmath273 , @xmath576 , and @xmath577 factors are @xmath578 @xmath579 @xmath580 @xmath581 with these changes we obtain @xmath582 @xmath583 @xmath584 in ( [ al3 ] ) the last term vanishes in the @xmath368 limit , while the last term of ( [ scalemi ] ) is negative .
this shows that low scale unification is possible . in table
[ t : lr6 ] we present several cases of successful unification in this scenario . with the experimental bound @xmath585gev we have @xmath586gev .
the picture of unification for the @xmath516 case of table [ t : lr6 ] is presented in fig .
4 . @xmath587 } & \log_{10}{[m_g / m_d ] } & \log_{10}{[m_i ] } & \log_{10}{[\mu_0 ] } & \log_{10}{[m_g ] } \\
\hline \hline 110 & ( 54,~54 ) & 3.44 & 2.52 & 3.85 & 3.86 & 5.90 \\ \hline 109 & ( 53,~54 ) & 3.62 & 2.65 & 3.97 & 3.98 & 6.00 \\ \hline 108 & ( 53,~53 ) & 3.80 & 2.80 & 4.10 & 4.10 & 6.15 \\ \hline 107 & ( 52,~53 ) & 3.78 & 2.78 & 4.21 & 4.21 & 6.23 \\ \hline 90 & ( 44,~44 ) & 5.26 & 1.96 & 4.71 & 6.67 & 8.63 \\ \hline 89 & ( 43,~44 ) & 5.42 & 1.95 & 4.73 & 6.82 & 8.77 \\ \hline 80 & ( 39,~39 ) & 7.90 & 1.90 & 4.45 & 8.46 & 10.36 \\ \hline \end{array}\ ] ] * model iii-susy422 : low scale unification * a different model which also gives low scale unification is a scenario extended with @xmath398 and four @xmath588 ( @xmath589 ) , where @xmath63 is a doublet of @xmath93 . with @xmath126 parities @xmath590 their contributions to the @xmath273-factors ( above the common mass @xmath591 ) and @xmath52 , @xmath563 factors are @xmath592 @xmath593 rges in this case are @xmath594 @xmath595 @xmath596 from the above equations we see that this scenario also allows low scale unification . in table
[ t : lr7 ] we present mass scales giving successful unification . in this case
the experimental bound @xmath597 gev gives @xmath598gev .
the picture of unification for the @xmath599 case of table [ t : lr7 ] is presented in fig .
@xmath600 } & \log_{10}{[m_i ] } & \log_{10}{[\mu_0]}&\log_{10}{[m_g ] } \\
\hline \hline 57 & ( 27,~28 ) & 2.89 & 3.41 & 3.41 & 5.17 \\ \hline 55 & ( 26,~27 ) & 2.85 & 3.85 & 3.85 & 5.59 \\ \hline 54 & ( 26,~26 ) & 2.84 & 4.08 & 4.08 & 5.81 \\ \hline 53 & ( 25,~26 ) & 2.82 & 4.30 & 4.30 & 6.02 \\ \hline 37 & ( 17,~18 ) & 7.11 & 5.55 & 7.85 & 9.42 \\ \hline 36 & ( 17,~17 ) & 7.29 & 5.67 & 8.07 & 9.63 \\ \hline 18 & ( 8,~8 ) & 11.08 & 7.54 & 12.14 & 13.39 \\ \hline 17 & ( 7,~8 ) & 11.24 & 7.66 & 12.36 & 13.59 \\ \hline \end{array}\ ] ]
finally we consider flipped @xmath66 gut in 5d .
the decomposition of @xmath2 s adjoint @xmath322 in terms of @xmath602 reads v(24)=v_c(8 , 1)_0+v_su(2)_l(1 , 3)_0+v_s(1 , 1)_0+v_x(3 , |2)_5 + v_y(|3 , 2)_-5 , where subscripts denote @xmath390 charges in @xmath306 units . since the @xmath390 comes from @xmath2 , it has the usual @xmath306 normalization .
@xmath603 in ( [ dec24of51 ] ) corresponds to the @xmath390 gauge superfield .
the decomposition of @xmath323 has an identical form .
* matter sector * the matter sector contains @xmath274 generations of @xmath71 supermultiplets _ n=2=(x , ) , * v*_n=2=(v , ) , * i*_n=2=(i , ) , where = * 10*_-1=(^c , q , d^c)_-1 , v=*|5_3=(l , u^c)_3 , i=*1*_-5=e^c_-5 are the usual chiral multiplets of flipped @xmath601 gut , and @xmath604 , @xmath605 , @xmath606 in ( [ matter51 ] ) are their mirrors .
subscripts in ( [ matterdec51 ] ) denote @xmath164 charges , defined up to some normalization factor . assuming that @xmath9 comes from @xmath18 ,
the normalization factor will be @xmath607 .
therefore , in @xmath18 normalization y_u(1)[x]=- , y_u(1)[v]= , y_u(1)[i]=- , in fact , the @xmath18 spinor @xmath608 in terms of @xmath9 reads _
so(10)=*10*_-1+*|5*_3+*1*_-5 .
* we also introduce @xmath274 copies @xmath609 , @xmath610 , @xmath611 .
@xmath344 generations are introduced at the brane . * scalar sector * to have mssm doublets we introduce the following set of supermultiplets : _
n=2=h(5_2 ) , ( |5_-2 ) , * h*_n=2=h(5_-2 ) , |h(|5_-2 ) , where subscripts denote @xmath164 charges in @xmath607 units . in ( [ n2sc51doubl ] )
one has h(5_2)=(h_d , ^c_h)_2 , ( |5_-2)=(h_u , d^c_|h)_-2 , and the same for states with primes . for reducing the rank of the group via higgs breaking
we need additional states .
thus we introduce _
n=2=h , , * h*_n=2=h , , where h=10_-1^h = ^c_h , d^c_h , q^ _ h _
-1 , = ^h_1=^c_h , ^c_h , ^ _ h_1 , and similarly for @xmath328 and @xmath612 .
the first stage of breakdown of the @xmath9 gauge group again occurs through orbifolding : prescribing to various fragments of @xmath9 multiplets certain @xmath181 parities some states are projected out , and at a fixed point we remain with a reduced gauge group .
the transformation properties of the fragments of 5d susy @xmath9 are given in table [ t:51gut ] .
@xmath614 one can easily see that at the @xmath117 fixed point we have a @xmath615 gauge group . at this fixed point together with three generations of quark - lepton and right handed neutrino superfields @xmath203 , @xmath345 , @xmath346 , @xmath204 , @xmath616 , @xmath432 and an mssm pair of higgs doublets @xmath35 , @xmath331 , we also have the states @xmath410 , @xmath411 , @xmath408 , @xmath418 .
the latter two states are responsible for an @xmath617 breaking down to @xmath158 .
if they develop vevs on the scale @xmath162 the second stage of symmetry breaking occurs and we have the unbroken @xmath158 generator y = y_u(1)-y_u(1) , where @xmath618 and @xmath619 charges of the states are given in table [ t:51gut ] .
[ for @xmath619 again we have used @xmath18 normalization ( [ so10norm ] ) , since flipped @xmath67 is one of its maximal subgroups @xcite , @xcite ] . the generation of @xmath413
, @xmath417 vevs can happen in the same way as for the @xmath153 scenario presented in sect .
7.1 . introducing a discrete @xmath11 symmetry and a transformation for @xmath419 as in ( [ ssym ] ) , the relevant superpotential , soft susy breaking , and all the potential couplings will have the form of ( [ sups ] ) , ( [ softspot ] ) , ( [ pots ] ) resp . if the states @xmath148 , @xmath149 are replaced by @xmath408 , @xmath418 .
from all this one can ensure non zero vev solutions ( [ nuvevdef ] ) , ( [ nusol ] ) . by a proper choice of @xmath152 one
can obtain a gap between the mass scales @xmath620 and @xmath37 .
also @xmath0 term generation can happen similarly .
with transformation ( [ mutermmssm ] ) for the @xmath333 combination and coupling ( [ musup ] ) the @xmath0 term ( [ muterm ] ) is derived . also for the susy @xmath9 model we use a @xmath11 symmetry to obtain a realistic phenomenology .
the lagrangian ( [ lagv ] ) , ( [ lagf ] ) with @xmath621 orbifold parities should be invariant under this @xmath11 ; therefore the fragments of matter and scalar superfields from the same @xmath601 multiplets should have identical @xmath11 phases , while the phases of mirrors must be opposite .
together with quark - lepton superfields and right handed neutrinos at the @xmath117 fixed point we have zero mode @xmath408 , @xmath418 , @xmath410 , @xmath411 states . with couplings ( [ sups ] ) ( with @xmath349 replaced by @xmath622 ) , ( [ musup ] ) we thus have ( ^c_h)+(^c_h)= ( h_d)+(h_u)= , and ( ^c)=(d^c ) , ( ^c_h)=(d^c_h ) , ( ^c_h)=(^c_h ) .
therefore , at this stage together with the phases of @xmath203 , @xmath623 , @xmath204 , @xmath616 , we have 8 independent phases .
writing yukawa couplings which generate masses for up - down quarks and charged leptons w_y
= qu^ch_u+qd^ch_d+le^ch_d , we will remain with 4 independent phases .
it is possible to fix two of them from the neutrino sector , writing dirac and majorana type couplings .
for the question of the mass scales needed for successful unification the couplings of type ( [ dirmaj422 ] ) are irrelevant , while terms @xmath624 give @xmath625 .
the latter leads to the reasonable value @xmath439 ev for @xmath626 in table [ t:51susy ] .
however , through these couplings the phases of appropriate states are selected in such a way , that some unacceptably large matter parity ( lepton number ) and baryon number violating couplings are allowed .
to avoid this , we will modify to a model , which gives a value for the neutrino mass accommodating atmospheric data and does not lead to any unacceptable process . including a matter parity violating ( [ bilins ] ) type coupling ^n-1^c_hlh_u , similar to the treatment in the mssm and @xmath2 cases , one obtains an operator @xmath627 with @xmath628 inducing a neutrino mass ( [ nubilins ] ) , where in this case @xmath629 ( still without assuming any alignment between superpotential and soft susy breaking couplings ) . for a suppressed neutrino mass we need @xmath630 .
this case indeed is realized with successful unification [ see case @xmath631 in table [ t:51susy ] ] : for @xmath632 gev and @xmath633 gev , three gauge couplings are unified and we have @xmath634 ev . couplings ( [ yuk51 ] ) , ( [ bilin ] ) and conditions ( [ phasecond1 ] ) , ( [ phasecond2 ] ) determine the phases @xmath635 @xmath636 ( ^c_h)=(^c_h)=(l)+(h_u ) , = , but @xmath637 , @xmath638 , @xmath254 are still free . with eqs .
( [ phases51 ] ) the lepton number violating operators qd^cl , e^cll , are allowed , and for @xmath639 they lead to a radiative neutrino mass @xmath640 ev [ see eq .
( [ numastrilin ] ) , ( [ trilins1 ] ) ] , with the relevant scale for the solar neutrino anomaly .
chosing @xmath637 , @xmath638 and @xmath254 as ( q)=(l)=/3 , ( h_u)= , the discrete @xmath11 symmetry will be @xmath641 . with assignment ( [ phases51 ] ) , ( [ phaseqlh ] )
the coupling @xmath642 is allowed . however , the coupling @xmath464 is forbidden and therefore the @xmath29 operator @xmath365 does not emerge .
one can verify easily that all other matter parity and baryon number violating operators are absent in this setting .
the coupling @xmath642 would induce decays of a light triplet @xmath410 ( mass @xmath643 few tev ) @xmath644 with leptoquark signature , observable in future collider experiments @xcite . concluding this subsection we note that in order to give masses to the states @xmath432 one can introduce @xmath645 singlet states with @xmath641 phase @xmath646 . through couplings
@xmath647 right handed states would decouple at the scale @xmath162 .
below @xmath162 we have the mssm field content with the b - factors ( [ bmssm ] ) , plus states @xmath410 , @xmath411 with mass @xmath485 in the tev range and b - factors as in ( [ bdc ] ) . above the @xmath162 scale
we obtain b_u(1 ) , b_u(1) , b_2 , b_3^m_i = , , 1 , -2 , and for the @xmath487 and @xmath488 factors @xmath648 _u(1 ) , _
u(1) , _ 2 , _
3= , - , 0 , 2 + 4(1 , 1 , 1 , 1 ) , if @xmath274 families have kk excitations . according to ( [ sup51charges ] ) ,
we get @xmath649 for the @xmath167 in ( [ match ] ) if @xmath650 and @xmath492 . taking into account this and also ( [ rge321])-([alg ] ) one obtains _ 3 ^ -1=_2 ^ -1-_1 ^ -1 + + + s_1 + s_2 , = ( _ 1 ^ -1-_2 ^
-1)- - - s_1-s_2 , _ g^-1=_2 ^ -1-+ s_1-s_2 . the last four terms in ( [ 51modelals ] ) are positive and in order to get a reasonable value for @xmath198 one should take @xmath651 . then from ( [ 51modelscale ] ) one obtains @xmath652 gev ( @xmath653 ) . on the other hand
we know that this scenario leads to light @xmath410 , @xmath411 states , inconsistent with unification . to resolve this problem
we can introduce additional @xmath71 susy states , which contain zero mode @xmath159 doublets , which being light ( @xmath643tev ) will compensate contributions from colored triplets : states @xmath654 ( @xmath301 ) , with @xmath601 representations @xmath655 , @xmath656 ; they decompose into doublets and triplets @xmath657 , @xmath658 . with @xmath181 orbifold parities @xmath659 ( t^(1 ) , ^(2)))~(- , + ) , ( t^(2 ) , ^(1))~(+ , - ) , only @xmath660 , @xmath661 states will have zero modes and contribute in the b - factors .
below @xmath162 _ b_1 , b_2 , b_3= , 1 , 0 , while above the @xmath162 scale _ b_u(1 ) , b_u(1) , b_2 , b_3^m_i= , , 1 , 0 .
contributions to the @xmath487 and @xmath488 factors from the fragments of @xmath662 states are @xmath663 _
_u(1 ) , _ u(1) , _ 2 , _ 3= , , 0 , 2 . taking into account ( [ 51b3211 ] ) , ( [ bkk51 ] ) , ( [ lowbpsi])-([gadepsi ] )
, we finally have _ 3 ^ -1=_2 ^ -1-_1 ^ -1- + - - s_1 + s_2 , = ( _ 1 ^ -1-_2 ^ -1)- - + -s_1-s_2 , _ g^-1=_2 ^ -1 - - - s_1-s_2 , where @xmath510 is the mass of @xmath660 , @xmath661 state s
zero modes .
the mass scales , for which successful unification takes place in this model , are presented in table [ t:51susy ] . as we see , the masses of the doublets are in a range @xmath664 tev@xmath665 gev .
the picture of unification for the case in the last row of table [ t:51susy ] is presented in fig .
@xmath666 & \log_{10}[\mu_0 ] & \log_{10}[m_g ] & \log_{10}[\fr{m_i}{m_{d^c_h } } ] & \log_{10}[\fr{m_i}{m_d } ] & \al_3(m_z ) \\ \hline \hline 1 & ( 0 , ~0 ) & 15.37 & 16.37 & 16.37 & 12 & 12.15 & 0.119 \\
\hline 3 & ( 0 , ~1 ) & 14.59 & 15.61 & 16.09 & 11 & 12.2 & 0.1184 \\ \hline 4 & ( 1 , ~1 ) & 12.79 & 15.69 & 16.29 & 8 & 10.2 & 0.119\\ \hline 5&(1 , ~2 ) & 10.77 & 15.58 & 16.27 & 6.7 & 8.6 & 0.117 \\ \hline 6 & ( 2 , ~2 ) & 5.53 & 15.75 & 16.53 & 2 & 3.3 & 0.1176\\ \hline 6 & ( 2 , ~2 ) & 6.45 & 15.67 & 16.45 & 3 & 4.3 & 0.116\\ \hline \end{array}\ ] ] concluding this subsection , we note that in the flipped @xmath67 gut it is impossible to get low scale ( near few or multi tev ) unification .
the reason is the following : introducing some additional states , one should cancel the positive power law contribution in ( [ 51modelals ] ) in order to get a reasonable @xmath667 .
the contribution from additional states will have the form @xmath668 , where @xmath669 is a contribution to the renormalization of @xmath670 .
since fragments from non trivial @xmath2 representations give the same contribution to factors of @xmath159 and @xmath160 we have @xmath671 .
therefore the final contribution to @xmath294 is @xmath672 .
the latter should cancel the last two terms in ( [ 51modelals ] ) , which in the @xmath673 limit are equal to @xmath674 .
thus , from the cancellation condition we have @xmath675 ; and thus the contribution in ( [ 51modelscale ] ) is @xmath676 .
this value precisely cancels the last two terms of ( [ 51modelscale ] ) in the @xmath677 limit .
this means that @xmath22 can not be lowered down to multi tev . in fig .
5 all couplings unify at one point .
however , it would be quite natural in spirit of a two step unification that at a first step only the three couplings of @xmath2 unify and the @xmath164 coupling joins at a higher scale .
this can be achieved either by a change of the intermediate scale or by a different choice of the extra state s masses .
we have considered 5d orbifold susy models and within them we have addressed numerous phenomenological issues .
orbifold constructions give an attractive resolution of several outstanding problems of guts , but some extensions are still needed to have full control of difficulties which even appear outside guts .
in fact , problems such as baryon number violation , @xmath0 problem and neutrino oscillations ( tied with lepton number violation ) are not cured by extra dimensions , and some care is needed to deal with them . in our approach
we have considered extensions with a discrete @xmath11 symmetry , which gives a natural and simultaneous understanding of these problems . in the essential part of the paper
we have addressed the question of gauge coupling unification , which in the presence of kk states gets new facets . since the orbifold approach to the celebrated @xmath2 gut does not allow for low or intermediate scale unification , we studied extended ( in rank ) guts . as we have seen
this opens up new and interesting possibilities from the viewpoints of unification and phenomenological implications .
for extended guts , symmetry breaking can occur by a step by step compactification and if the intermediate gauge group differs from semisimple @xmath2 , power law unification can take place . as an example we have considered the two maximal subgroups of @xmath18 - the pati - salam @xmath8 and flipped @xmath9 guts . within @xmath8 ,
low scale unification can take place , while @xmath9 only allows for unification at scales @xmath10 gev .
the latter scales are also possible for @xmath8 . within both scenarios extensions with a discrete @xmath11 symmetry
were pursued ; thus key phenomenological problems were resolved in an elegant way . the @xmath9 model and
also in some cases the @xmath8 one predict colored triplet states in the few tev range .
@xmath8 models with low scale unification lead to additional relatively light states . because of these the models have an interesting phenomenology , in particular if kk states appear near the tev scale .
future high energy collider experiments will test the relevance of such models . in our studies
we have used one loop rge analysis .
two loop power law contributions to the @xmath51-function would contribute significantly .
this would make estimates unstable .
however , due to higher dimensional supersymmetries , in higher loops full sets of @xmath71 supermultiplets can contribute @xcite , @xcite and in this case effects of higher loops will be logarithmic .
the latter contributions would not change the unification picture and might only imply a slight modification of mass scales .
the same argument could be applied for threshold corrections coming from different sources .
our studies of the @xmath8 and @xmath9 groups were performed in a bottom - up approach and therefore the low energy sector was more relevant for our considerations . having in mind that both @xmath8 and @xmath9 are maximal subgroups of @xmath18 , it would be interesting to construct a higher ( at least two extra ) dimensional @xmath18 model where extra dimensions do not compactify at a single mass scale ( in contrast to the models of @xcite ) but with intermediate groups @xmath8 or @xmath9 at a scale between @xmath378 . only this kind of step by step compactification breaking of an initial semisimple gauge group can give power law unification near the tev scale
. it would be also interesting to consider other extended gauge groups in higher dimensions and to study the breaking pattern and the phenomenological implications .
we wish to study these and other relevant issues in our future publications .
is supported by funda~ ao de ci^ encia e tecnologia ( grant sfrh / bd/4973/2001 ) .
in this appendix we investigate the influence of some brane operators on gauge coupling renormalization .
we will show that the 4d mass term of bulk vectorlike states does not change the expressions for the power law functions @xmath284 and @xmath182 , @xmath183 of ( [ dekkz ] ) and ( [ skk ] ) resp .. also , it turns out that some brane bi - linear derivative couplings , involving states with negative parities , do not change this picture for @xmath678 .
the latter condition is usually satisfied in models with power law unification .
is satisfied . ]
our discussion will be model independent .
we will consider here the case of @xmath1 orbifold , but generalization to @xmath7 scenarios is straightforward .
consider two @xmath71 supermultiplets @xmath679 and @xmath680 transforming non trivially under a certain gauge group @xmath59 . if @xmath681 , @xmath682 belong to some representation @xmath79 of @xmath59 , then @xmath683 , @xmath684 transform as @xmath80 .
let us assign the orbifold parities as ( e , )~*+ * , ( e , ) ~*- * .
thus , only @xmath681 , @xmath684 states have zero modes . writing the brane coupling @xmath685
we obtain , after performing the integration over @xmath123 , the 4d superpotential mass terms w^kk_e=_0_n=1^ne^(n)^(n)+ e^(n)^(n) . for our calculations
only a certain number of kk states is relevant , which contribute to the gauge coupling runnings .
so , we truncate the tower of kk states on the level @xmath180 . taking this into account and combining ( [ supa ] ) , ( [ masskka ] ) ,
the mass terms can be written in compact form as @xmath689 where @xmath690 @xmath691 ,
\quad { \mathcal m}_e = m_e \left [ \begin{array}{cccc } \sqrt{2 } & 2 & \ldots & 2\\ \vdots & & \ddots & \vdots\\ \sqrt{2 } & 2 & \ldots & 2 \\ 1 & \sqrt{2 } & \ldots & \sqrt{2 } \end{array } \right]~ , \la{mata}\ ] ] and @xmath692 , \qquad { \mathcal m}_0= \mu_0 \left [ \begin{array}{ccccc } 0 & 1 & \ldots & 0 & 0 \\ \vdots & & \ddots & & \vdots \\ 0 & 0 & & n_0 - 1 & 0 \\ 0 & 0 & \ldots & 0 & n_0 \end{array } \right]~. \la{defmata}\ ] ] to estimate the kk and zero mode contributions to the running , one should diagonalize the matrix @xmath693 and find the eigenvalues @xmath694 , @xmath695 , @xmath377 , @xmath696 .
if all these masses lie below @xmath22 , then their contribution to the renormalization of @xmath697 up to the gut scale will be @xmath698 from ( [ mata ] ) , ( [ defmata ] ) it is easy to see that = m_e_n=1^n_0m_n^2= m_e_n=1^n_0n_0 ^ 2 . using ( [ detma ] ) in ( [ dea ] ) we obtain ^e_i=-- -_n=1^n_0 , ( where we have set @xmath699 ) .
as one can see , the first term in ( [ de1a ] ) is just the naive logarithmic running of zero mode states , while the second term has precisely the form of the unaffected kk power law running ( [ dekkz ] ) . from this
we conclude that due to the brane coupling ( [ bra ] ) , the rges will in addition include a logarithmic term corresponding to the zero modes @xmath700 , @xmath701 above the scale @xmath313 , while the effect of the kk states must be estimated by the power law function @xmath284 .
these conclusions are valid for any value of @xmath702 .
higher order brane couplings , preserving orbifold symmetries , imply @xmath703 derivatives and must be cut off at the fundamental scale @xmath704 .
for example , the couplings @xmath705 after integration along the fifth dimension , change the masses by the amount @xmath706 .
the latter is negligible for @xmath707 and ( [ de1a ] ) is still valid .
h. georgi , s.l .
glashow , phys.rev.lett .
32 ( 1974 ) 438 ; h. fritzsch , p. minkowski , ann .
93 ( 1975 ) 193 .
h. georgi , h. quinn , s. weinberg , phys .
33 ( 1974 ) 451 ; w. marciano , a. sirlin , phys .
46 ( 1981 ) 163 .
r. arnowitt , p. nath , phys .
b 287 ( 1992 ) 89 ; phys
70 ( 1993 ) 3696 ; j. hisano , h. murayama , t. yanagida , nucl .
b 402 ( 1993 ) 46 ; k.s . babu , s.m .
barr , phys .
b 381 ( 1996 ) 137 ; z. berezhiani , z. tavartkiladze , m. vysotsky , hep - ph/9809301 ; r. dermisek , a. mafi , s. rabi , phys . rev .
d 63 ( 2001 ) 035001 ; h. murayama , a. pierce , hep - ph/0108104 .
i. antoniadis , j. ellis , s. kelley , d. nanopoulos , phys.lett .
b 272 ( 1991 ) 31 .
p. lamgacker , n. polonsky , phys .
d 47 ( 1993 ) 4028 ; phys .
d 52 ( 1995 ) 52 ; n. polonsky , hep - ph/0108236 ; see also references terein .
y. kawamura , prog .
phys . 105 ( 2001 ) 999 ; _ ibid .
_ 105 ( 2001 ) 691 .
g. altarelli , f. feruglio , hep - ph/0102301 ; a. kobakhidze , hep - ph/0102323 ; m. kakizaki , m. yamaguchi , hep - ph/0104103 ; n. maru , hep - ph/0108002 .
y. nomura , d. smith , n. weiner , hep - ph/0104041 ; t. kawamoto , y. kawamura , hep - ph/0106163 ; r. barbieri , l. hall , y. nomura , hep - th/0107004 ; a. hebecker , j. march - russell , hep - ph/0107039 ; c. csaki , g. kribs , j. terning , hep - ph/0107266 .
t. asaka , w. buchmller , l. covi , hep - ph/0108021 ; l. hall , y. nomura , t. okui , d. smith , hep - ph/0108071 ; r. dermisek , a. mafi , hep - ph/0108139 ; n. haba , t. kondo , y. shimizu , hep - ph/0112132 , hep - ph/0202191 .
r. blumenhagen , l. goerlich , b. krs , d. lst , hep - th/0010198 ; g. aldazabal , s. franco , l. ibanez , r. rabadan , a. uranga , hep - th/0011073 ; hep - ph/0011132 ; l.e .
ibanez , f. marchesano , r. rabadan , hep - th/0105155 ; r. blumenhagen , b. krs , d. lst , t. ott , hep - th/0107138 .
i. antoniadis , phys .
b 246 ( 1990 ) 377 ; i. antoniadis , c. munoz , m. quiros , nucl . phys .
b 397 ( 1993 ) 515 ; j. lykken , phys . rev .
d 54 ( 1996 ) 3693 ; i. antoniadis , s. dimopoulos , a. giveon , hep - th/0103033 .
j. price , proc .
int . symp . on experimental gravitational physics ,
p. michelson , guangzhou , china ( world sci . , singapore 1988 ) ; j. price et al .
, nsf proposal 1996 ; a. kapitulnik , t. kenny , nsf proposal , 1997 ; j. long et al .
, hep - ph/9805217 .
k. dines , e. dudas , t. gherghetta , phys .
b 436 ( 1998 ) 55 ; nucl .
b 537 ( 1999 ) 47 .
z. kakushadze , nucl .
b 548 ( 1999 ) 205 ; _ ibid .
_ b 552 ( 1999 ) 3 . c. carone , hep - ph/9902407 ; a. delgado , m. quiros , hep - ph/9903400 ; p. frampton , a. rasin , hep - ph/9903479 ; a. perez - lorenzana , r.n .
mohapatra , hep - ph/9904504 ; z. tavartkiladze , hep - ph/0105281 .
i. antoniadis , j. hagelin , d. nanopoulos , phys . lett .
b 194 ( 1987 ) 231 ; g. lazarides , c. panagiotakopoulos , q. shafi , phys . lett .
b 315 ( 1993 ) 325 ; g. dvali , q. shafi , phys . lett .
b 403 ( 1997 ) 65 ; q. shafi , z. tavartkiladze , hep - ph/9811282 ; hep - ph/9807502 .
g. dvali , q. shafi , phys .
b 326 ( 1994 ) 258 ; b 339 ( 1994 ) 241 .
kim , h.p .
nilles , mod .
lett . a 38 ( 1994 ) 3575 .
k. choi , e.j .
chun , h. kim , phys .
d 55 ( 1997 ) 7010 .
g. dvali , q. shafi , phys.lett .
b 339 ( 1994 ) 241 ; z. berezhiani , c. csaki , l. randall , hep - ph/9501336 ; see also q. shafi , z. tavartkiladze , in ref . @xcite .
t. banks , y. grossman , e. nardi , y. nir , phys .
d 52 ( 1995 ) 5319 .
h. nilles , n. polonsky , nucl .
b 484 ( 1997 ) 33 ; e. nardi , phys .
d 55 ( 1997 ) 5772 ; g. eyal , y. nir , hep - ph/9904473 ; j. chkareuli , i. gogoladze , a. kobakhidze , m. green , d. hutchcroft , hep - ph/9908451 ; q. shafi , z. tavartkiladze , hep - ph/9909238 ; j. feng , y. nir , y. shadmi , hep - ph/9911370 ; see also other references therein .
d. caldwell , r. mohapatra , phys .
d 48 ( 1993 ) 3259 ; c. carone , m. sher , hep - ph/9711259 ; f. vissani , hep - ph/9708483 ; h. georgi , s. glashow , hep - ph/9808293 ; y.l .
wu , hep - ph/9810491 ; c. wetterich , hep - ph/9812426 ; j. ellis , s. lola , hep - ph/9904279 ; see also references therein .
j. elwood , n. irges , p. ramond , hep - ph/9807228 ; f. vissani , hep - ph/9810435 ; z. berezhiani , a. rossi , hep - ph/9811447 ; r. barbieri , l. hall , g. kane , g. ross , hep - ph/9901228 ; q. shafi , z. tavartkiladze , hep - ph/9901243 ; hep - ph/0002150 ; g. altarelli , f. feruglio , i. masina , hep - ph/9907532 ; s. barr , i. dorsner , hep - ph/0003058 ; j. chkareuli , c. froggatt , h. nielsen , hep - ph/0109156 ; i. gogoladze , a. perez - lorenzana , hep - ph/0112034 ; f. feruglio , a. strumia , f. vissani , hep - ph/0201291 ; see also references therein .
g. dvali , m. shifman , phys .
b 399 ( 1997 ) 60 . g. abbiendi et al .
[ opal collaboration ] , eur .
j. c 12 ( 2000 ) 1 ; m. carena , g. giudice , s. lola , c. wagner , hep - ph/9612334 ; b. mukhopadhyaya , s. roy , hep - ph/9903418 . | we study gauge coupling unification and various phenomenological issues , such as baryon number conservation , the @xmath0 problem and neutrino anomalies , within susy 5d orbifold models .
the 5d mssm on an @xmath1 orbifold with minimal field content does not lead to low scale unification , while some of its extensions can give unification near the multi tev scale . within the orbifold @xmath2 gut , low scale unification
can not be realized due to full @xmath2 multiplets participating in the renormalization above the compactification scale . as alternative examples ,
we construct 5d @xmath3 susy pati - salam @xmath4 and flipped @xmath5 guts [ both maximal subgroups of @xmath6 on an @xmath7 orbifold .
new examples of low scale unifications within @xmath8 are presented . for @xmath9
the unification scale is shown to be necessarily close to @xmath10 gev .
the possible influence of brane couplings on the gauge coupling unification is also outlined . for the resolution of the various phenomenological problems extensions with a discrete @xmath11 symmetry turn out to be very effective . * gauge coupling unification and phenomenology + of selected orbifold 5d @xmath3 susy models * filipe paccetti correia@xmath12 , michael g. schmidt@xmath12 , zurab tavartkiladze@xmath13 @xmath14_institut fr theoretische physik , universitt heidelberg , philosophenweg 16 , + d-69120 heidelberg , germany + @xmath15 institute of physics , georgian academy of sciences , tbilisi 380077 , georgia + _ |
epsilon ursae majoris ( @xmath0 uma , hd 112185 , hr 4905 ) is the brightest ( @xmath9 = 1.77 mag ) chemically peculiar ( cp ) star and has been extensively studied during the last century .
@xcite established a period of 5.@xmath100887 d from variations in the intensity of the ca ii k line and also noticed a periodic splitting of some lines .
@xcite subsequently reported doubling of lines of cr ii , fe ii , v ii and other elements at certain phases .
since the overall widths of the double lines are the same and not all lines double , they ruled out the orbital motion and instead suggested that the phenomenon is related to rotation of the star .
@xcite measured a double wave light variation with the same 5.@xmath100887 period .
the star is brightest when the ca ii k line intensity is near its minimum and other elements are near their maximum strength .
perhaps the most interesting and controversial aspect of @xmath0 uma is its the radial velocity variations .
@xcite carried out the radial velocity analysis of the 62 brightest northern cp stars for spectroscopic binaries .
they were used special observing and measuring techniques of the hydrogen lines and did not find the radial velocity variations for @xmath0 uma .
not surprisingly , they could not to measure the radial velocity variations because the mean internal error per spectrum of 1.2 km @xmath11 is near the limit at which the photographic technique is useful . on the other hand , @xcite detected the duplicity for the system @xmath0 uma with a separation of 0.053 arcsec , using speckle interferometry technique .
although , the authors noted that the observed separation corresponds to the diffraction limit of the 2.5-m isaac newton telescope .
@xcite measured the radial velocity variations of many lines in the star and found sinusoidal variations in the radial velocities of fe , cr and ti lines with amplitudes of about 20 km @xmath11 and attributed this to the existence of a spots of enhanced abundance of these elements .
several surface abundance doppler images of @xmath0 uma have been produced @xcite .
most of the published doppler images relate to fe , o , ca , or cr abundance .
recently , @xcite determined for @xmath0 uma for the first time the abundance distributions of mn , ti , sr , and mg .
attempt to determine orbital motion using these spectral lines would be impossible .
however , the hydrogen lines do not show significant rotational effect in their radial velocity .
they would be preferable for the measurement of binary motion if the radial velocity of the hydrogen lines could be measured with precision . in this letter , we present the radial velocity measurement of the hydrogen lines in the spectrum of cp star @xmath0 uma .
the elodie observations of this star and data reduction are described in sect . 2 .
the orbital solution derived for the satellite candidate are presented in sect .
section 4 reports discussion of our results for @xmath0 uma and conclusions are presented in sect .
[ cols="^,>,^,^ , > , > , > , > , > , > " , ] @llcc@ parameter & & value & error ( fixed ) & ( d ) & 5.0887 & + _ t _ ( periastron ) & ( jd-2450000 ) & 1743.932 & 0.161 + _ e _ & & 0.503 & 0.063 + @xmath12 & ( km @xmath11 ) & -8.920 & 0.030 + @xmath13 & ( @xmath14 ) & 260.53 & 12.91 + _ a _
@xmath8(i ) & ( @xmath15 au ) & 0.256 & 0.022 + _ k _ & ( km @xmath11 ) & 0.634 & 0.067 + _
f_(m ) & ( @xmath16 @xmath17 ) & 0.868 & 0.227 + the spectra of @xmath0 uma were retrieved from the elodie archive @xcite .
this archive contains the complete collection of high - resolution echelle spectra using the elodie fiber - fed echelle spectrograph @xcite mounted on the 1.93-m telescope at the haute - provence observatory ( france ) .
the spectra have a resolution ( @xmath18/@xmath19@xmath18 ) of about 42000 .
the archived signal - to - noise ratio was between @xmath3200 and @xmath3400 in the spectral region near @xmath18 5550 .
in addition , eight spectra obtained 15 and 16 june 2000 at the same telescope were retrieved from the hypercat fits archive ( http://leda.univ-lyon1.fr/11/spectrophotometry.html ) . in table 1 , each spectrum is presented by its file name , exposure time , julian date of the observations and its corresponding phase .
note that the phases were computed with the ephemeris from table 2 ( see below ) .
special technique of radial velocity measurement of the hydrogen lines was used and should be explained .
well known , that in slowly rotating late b - type stars the hydrogen lines have broad wings and sharp cores . for @xmath0
uma the value of @xmath20 sin @xmath21 is equal 35 km @xmath11 @xcite . unfortunately , the wings of the hydrogen lines can be affected by different spectral lines .
for example , in spectrum of @xmath0 uma the wing of @xmath22 line affected by many cr lines @xcite . on the other hand ,
the sharp cores of the hydrogen lines are not affected by spectral lines and the flux is formed in the upper layers of the atmosphere . thus our technique was to use only the sharp cores of the hydrogen lines in spectra of @xmath0 uma .
the spectra were processed using the spectral reduction software spe developed by s. sergeev at crimean astrophysical observatory ( crao ) .
the program allows detecting the variations of the centre gravity of the sharp cores of the hydrogen lines .
after processing , all the spectra were corrected for the motion of the earth around the sun .
for example , the demonstration of the positional variability of the core @xmath23 line is presented in fig . 1 .
in table 1 , each spectrum is presented by its radial velocities computed from the hydrogen lines , the mean radial velocity and errors of the mean radial velocity .
line obtained on may 21 , 2000 ( phase = 0.93 ) and on june 6 , 2000 ( phase = 0.07 ) marked by the solid and dashed lines , respectively .
the velocity scale is given with respect to the @xmath18 = 6562.797 .,width=264 ]
orbital elements have been determined by a non - linear least - squares fitting of the mean radial velocities from table 1 using the program binary writing d.h .
gudehus from georgia state university ( http://www.chara.gsu.edu/@xmath3gudehus/binary.html ) .
the solution for a single - lined binary is modelled by up to six parameters : 1 .
_ p _ period 2 .
_ t _ time of periastron passage 3 .
_ e _ eccentricity 4 .
@xmath12 system radial velocity 5 .
@xmath13 longitude of periastron 6 . _ a _
@xmath8(i ) sine of inclination times semi - major axis the expected radial velocities are @xmath24\ ] ] where @xmath25 is the angular position of the star measured from the centre of mass at a given instant .
the program also calculate the projected semi - amplitude variation of the radial velocity : @xmath26 though this is never used as a parameter in the solution and the mass function : @xmath27 the program binary gives the estimated standard deviations of the orbital parameters as well .
the orbital solution of table 2 was obtained by fitting a keplerian orbit to the 29 elodie radial velocity measurements .
note that most of observations were obtained between jd=2451568 ( january 2000 ) and jd=2451714 ( june 2000 ) ( see table 1 ) .
experience shows that the best keplerian fit to the data with the fixed period _ p _
= 5.0887 d , the eccentricity _ e _ = 0.503 and the semi - amplitude _ k _ = 0.634 km @xmath11 .
the parameters of the best keplerian orbital solution for @xmath0 uma are presented in table 2 . in close binary system with bp - ap stars
, there is evidence for a tendency toward synchronization between the rotational and orbital motions .
this effect is thought to be produced by the tidal forces @xcite .
the radial velocity curve is displayed in fig . 2 with the residuals around solution .
a linear trend is not observed in the residuals around the orbital solution that can be explained by the absent of a second companion in a longer - period orbit .
although , the weighted r.m.s . around the best keplerian solution ( @xmath28(o - c ) ) is equal to 0.131 km @xmath11 .
this value is a bit large compared to the typical radial velocity measurements from the elodie cross - correlation function @xcite .
phase diagram of the radial velocity measurements and keplerian orbital solution for @xmath0 uma .
@xmath29 residuals around the solution.,width=264 ]
the presence of the spots on the stellar surface of @xmath0 uma can change the observed spectral line profiles and induces a periodic radial velocity signal similar to the one expected from the presence of a satellite .
the hydrogen lines analysis is one of the best tools to discriminate between radial velocity variations due to changes in the spectral line shapes and variations due to the real orbital motion of the star .
it is obviously of interest to compare the phase diagram of the radial velocity of @xmath0 uma derived above with the phase diagram of the radial velocity computed from metallic lines . in this way , we selected the spectral region of the very prominent unblended cr ii line at @xmath18 4558 .
4558 obtained on june 15 , 2000 ( phase = 0.84 ) and on january 26 , 2000 ( phase = 0.19 ) marked by the solid and dashed lines , respectively .
the velocity scale is given with respect to the @xmath18 = 4558.65 .,width=264 ] figure 3 shows the strong photosphere lines in the spectral region of the cr ii line at @xmath18 4558 obtained at phases before and after the epoch of periastron passage .
the radial velocity shift of the cr ii @xmath30 4554 , 4558 and 4565 lines and the fe ii @xmath18 4555 line at different phases is clearly seen .
the ti ii line at @xmath18 4563 shows the line doubling that appears at phase 0.84 and disappears at phase 0.19 .
the maximum splitting of this line is at phase 0.04 .
the splitting of the lines was first observed by @xcite in the spectrum of @xmath0 uma .
however , the authors concluded that the doubling is not caused by orbital motion but may be due to a combination of the physical effects with doppler effect in a rotating star .
@xcite estimated the masses for 10 magnetic cp stars using the position of stars in the log@xmath31
log@xmath32 plane .
position of @xmath0 uma in the log@xmath31
log@xmath32 plane seems to be among the most evolved cp stars and gives the value of @xmath33 = [email protected]@xmath35 . according to @xcite
the radius of @xmath0 uma is equal to [email protected]@xmath31 corresponding to their choice of @xmath36 = 35 km @xmath6 and @xmath21 = @xmath37 using the trigonometric parallax measured by _
hipparcos _ , @xmath38 = 40.30 mas ( eso , 1997 ) and an angular diameter of 1.561 mas .
the effective temperature of @xmath0 uma was taken from the paper by @xcite and is equal to 9340@xmath34530 k. on the hypothesis that the rotational axis of @xmath0 uma is perpendicular to the orbital plane we can estimate the mass of the secondary star
. for the value of @xmath39 = 3.0@xmath35 and @xmath21 = @xmath37 eq.(3 ) gives @xmath40 = 0.014@xmath35 .
this result gives a value @xmath314.7@xmath4 , strongly suggesting that the companion is in the typical brown - dwarf regime .
if we know the value @xmath39/@xmath40 then it is possible to estimate the projected semi - amplitude variation of the radial velocity for the companion according to formula : @xmath41 where @xmath42 is the projected semi - amplitude variation of the radial velocity for @xmath0 uma taken from table 2 . for the value of @xmath43 = 0.634 km @xmath6 eq.(4 )
gives @xmath5 = 135.9 km @xmath6 .
thus , we can estimate the sine of inclination times semi - major axis using eq.(2 ) .
computation gives the value of @xmath7@xmath8(i ) = 0.055 au .
these estimates show that the proposed brown - dwarf is quite close to the surface of @xmath0 uma at periastron .
but , such close orbits are not new .
for example , the subgiant star hd 118203 have the planet with eccentric orbit ( @xmath2 = 0.31 ) , the period of @xmath1 = 6.1335 d , and is close to its parent star ( @xmath44 = 0.06 au ) @xcite .
another way to interpret the radial velocity variations is the radial pulsation of @xmath0 uma .
@xcite are analysed observations of @xmath0 uma obtained with the star tracker on the _ wide field infrared explorer _ satellite .
the authors observed that a light curve has about 2 per cent amplitude of photometric variability .
on the other hand , @xcite has presented ultraviolet light curves for @xmath0 uma from the oao-2 satellite which indicate that the photometric variations of this star are due to variable ultraviolet absorption effects which redistribute flux into the visible region ( see his fig .
note that no colour index changes @xcite .
certainly the radial pulsation appears unlikely given that the rotational period is synchronized with the orbital period of @xmath0 uma .
the archival elodie high - resolution echelle spectra of @xmath0 uma permit us to analyse the radial velocity variations of the sharp cores of the hydrogen lines .
this allowed determining the orbital elements of binary system for the cp star @xmath0 uma .
the best keplerian fit to the data shown that the rotational period is synchronized with the orbital period .
we are estimated the mass of the secondary star which is equal @xmath314.7@xmath4 .
this result indicate that the companion is the brown - dwarf with the projected semi - amplitude variation of the radial velocity @xmath5 = 135.9 km @xmath6 and the sine of inclination times semi - major axis @xmath7@xmath8(i ) = 0.055 au .
the author would like to thank the referee dr . j.b .
rice of this letter for his extremely helpful comments .
99 abt h.a . ,
snowden m.s . , 1973 ,
apjss , 25 , 137 baranne a. , queloz d. , mayor m. , adrianzyk g. , knispel g. , kohler d. , lacroix d. , meunier j .-
, et al , 1996 , a&as , 119 , 373 da silva r. , udry s. , bouchy f. , mayor m. , moutou c. , pont f. , queloz d. , santos n.c . , et al , 2005 , ( astro - ph/0510048 ) gerbaldi m. , floquet m. , hauck b. , 1985 , a&a , 146 , 341 guthnick p. , 1934
berlin , 30 , 506 hatzes a.p . , 1991 , mnras , 253 , 89 holmgren d.e . , rice j.b . , 2000 ,
a&a , 364 , 660 lueftinger t. , kuschnig r. , piskunov n.e . , weiss w.w . , 2003 , a&a , 406 , 1033 molnar m.r . , 1975 ,
aj , 80 , 137 morgan b.l . , beddoes d.r . ,
scaddan r.j .
, dainty j.c . , 1978 , mnras , 183 , 701 moultaka j. , ilovaisky s.a . , prugniel p. , soubiran c. , 2004 , pasp , 116 , 693 naef d. , mayor m. , beuzit j.l . , perrier c. , queloz d. , sivan j.p .
, udry s. , 2003 , ( astro - ph/0310261 ) provin s.s . , 1953 ,
apj , 118 , 489 retter a. , bedding t , r . ,
buzasi d.l . , kjeldsen h. , kiss l.l . , 2004 , apj , 601 , l95 rice j.b . , wehlau w.h . , 1990 , a&a , 233 , 503 rice j.b . , wehlau w.h
, holmgren d.e . , 1997 ,
a&a , 326 , 988 sokolov n.a . , 1998 , a&as , 130 , 215 struve o. , hiltner w.a , 1943 , apj , 98 , 225 stibbs d.w.n . , 1950 , mnras , 110,395 wade g.a . , 1997 , a&a , 325,1063 wehlau w. , rice j. , piskunov n. , khokhlova v. , 1982 , pisma astron .
, 8 , 30 woszczyk a. , jasiski m. , 1980 , acta astron . , 30 , 331 iovsk j. , zverko j. , 1995 , contrib
skalnat pleso , 25 , 39 | in this letter , the radial velocity variability of the chemically peculiar star @xmath0 ursae majoris ( @xmath0 uma ) from the sharp cores of the hydrogen lines is investigated .
this study is based on the elodie archival data obtained at different phases of the rotational cycle .
the star exhibits low - amplitude radial velocity variations with a period of @xmath1 = 5.0887 d. the best keplerian solution yields an eccentricity @xmath2 = 0.503 and a minimum mass @xmath314.7@xmath4 on the hypothesis that the rotational axis of @xmath0 uma is perpendicular to the orbital plane .
this result indicate that the companion is the brown - dwarf with the projected semi - amplitude variation of the radial velocity @xmath5 = 135.9 km @xmath6 and the sine of inclination times semi - major axis @xmath7@xmath8(i ) = 0.055 au .
[ firstpage ] binaries : spectroscopic stars : chemically peculiar stars : individual : @xmath0 ursae majoris . |
calabi - yau manifolds of dimensions @xmath2 and @xmath3 , i.e. elliptic curves and k3-surfaces , have a long and successful tradition in geometry and number theory . in the 1980 s , in connection with developments in string theory , calabi - yau manifolds of dimension @xmath4 moved to the forefront . emphasis has been on their complex and symplectic geometry , and in particular on the variation of the hodge structure for calabi - yau threefolds near the so - called large complex structure limit .
this is the first of a series of papers in which we want to describe certain aspects of the arithmetic geometry of families of ordinary calabi - yau threefolds and analogies with complex calabi - yau threefolds near the large complex structure limit .
more specifically , this paper discusses the associated crystals , i.e. modules with an integrable connection , like the gauss - manin connection on the cohomology of a family of calabi - yau threefolds .
crystals which originate from geometry come with a hodge filtration , which is not preserved by the connection but instead satisfies griffiths transversality .
crystals which originate from geometry in characteristic @xmath1 additionally carry an action of frobenius operators behaving in an appropriate way with respect to the connection and the hodge filtration : they are divisible hodge @xmath5-crystals . for so - called divisible _
hodge @xmath5-crystals the space of flat sections decomposes into eigenspaces for the frobenius operators .
the position of the hodge filtration with respect to this eigenspace decomposition is the source for the canonical coordinates and the yukawa coupling : see theorems [ thmstructure ] , [ cy3structure ] , [ integralities ] .
the ordinariness requirement implies , exactly as the requirement of maximally unipotent local monodromy in the traditional analysis of complex calabi - yau threefolds near the large complex structure limit @xcite , that there is a _ filtration which is opposite to the hodge filtration and is invariant under the gauss - manin connection and whose associated graded is of hodge - tate type _
@xcite prop.1.3.6 , @xcite 6 ) .
the theory of ordinary hodge @xmath5-crystals has its origins in dwork s work on the variation of the zeta function in a family of varieties over a finite field @xcite and in the work of serre and tate on the formal moduli of deformations of an ordinary elliptic curve over a field of positive characteristic . in @xcite deligne and illusie
present a general theory of ordinary hodge @xmath5-crystals and apply it to investigate the formal moduli of ordinary abelian varieties and ordinary k3-surfaces .
the general theory works equally well for ordinary cy3 crystals , i.e. crystals of the kind that arises as the crystalline cohomology of ordinary calabi - yau threefolds in characteristic @xmath0 .
thus , exactly as in @xcite , there are canonical coordinates on the formal moduli space . in section [ section 1 ]
we briefly recall the general theory of @xcite . in section 2
we take the same algebraic path as bryant , griffiths , morrison and deligne @xcite in their analysis of the variation of the hodge structure of complex calabi - yau threefolds near a maximally degenerate ( @xmath6 maximally unipotent local monodromy @xmath6 large complex structure ) boundary point of the moduli space . in particular
we find for ordinary cy3 crystals an exact analog of the cubic form of bryant - griffiths ( @xmath6 yukawa coupling ) .
interestingly , besides the standard characterization ( [ eq : yukawa ] ) of the prepotential of this cubic form we find a characterization ( [ eq : yukinnerproduct ] ) by means of the action of the canonical lift of frobenius .
the latter characterization remains valid if there are no parameters , i.e. if @xmath7 in ( [ eq : a ] ) , while in that case the former description is vacuous .
also for the canonical coordinates the formalisms in the ordinary and the large complex structure situations match very well , but the match is not perfect : in the ordinary case the canonical coordinates take the value @xmath2 at the origin of the deformation space ( cf .
( [ eq : canorig ] ) ) , whereas the canonical coordinates at the large complex structure limit point vanish . the standard algorithms ( cf .
@xcite ) for computing canonical coordinates and yukawa coupling for complex calabi - yau threefolds near the `` maximally unipotent local monodromy boundary point '' use special solutions of the picard - fuchs equations associated with the family of cy3 s and a nowhere vanishing global @xmath4-form . in remark [ picard - fuchs ]
we point out that in the ordinary case one has basically the same relation between canonical coordinates , yukawa coupling and special solutions of the picard - fuchs equations ( although explicit computations seem here out of reach ) .
the present paper is a survey of structures none which except maybe ( [ eq : yukinnerproduct ] ) and ( [ eq : yukinteger ] ) is new by itself , but which appear in the literature in different contexts .
so we hope that there is an inspiring and cross - fertilizing effect from putting them into one ( con)text . for instance , being alerted that for the deformation theory of ordinary calabi - yau threefolds the cubic form of bryant - griffiths ( yukawa coupling ) should be relevant , one may wonder how that shows up in connection with the hodge - tate decomposition of @xmath1-adic tale cohomology as in @xcite p. 109 .
the original motivation for our work was to find a general method of proof for the integrality conjectures in mirror symmetry by reducing them to known results or easy to check conditions on the crystalline cohomology for families of calabi - yau threefolds in positive characteristics : it has been observed in many examples in the literature on mirror symmetry for calabi - yau threefolds , that the coefficients in appropriate expansions of the canonical coordinates and of the prepotential of the yukawa coupling are integers . in the famous example of the quintics in @xmath8 ( see @xcite ) one has for instance @xmath9 where @xmath10 is the parameter and @xmath11 is the canonical coordinate of the family @xmath12 the yukawa coupling @xmath13 is given by @xmath14 @xmath15 the prepotential @xmath16 is defined ( up to terms of order @xmath17 in @xmath18 ) by @xmath19 and has an expansion @xmath20 where @xmath21 is the trilogarithm function .
according to the mirror symmetry conjecture @xcite the number @xmath22 is the number of rational curves of degree @xmath23 on a generic quintic hypersurface in @xmath8 .
for the integrality conjecture one is interested in the less astonishing , but still highly non - trivial fact that these numbers are integers .
now one may note that the numbers @xmath22 are integers if and only if for every prime number @xmath1 @xmath24\ ] ] where @xmath25 is the ring of @xmath1-adic integers . in this paper we show that @xmath1-adic analogues of the integrality conjectures , in particular an analogue of ( [ eq : exa.quintic ] ) , hold true for ordinary cy3 crystals ; see formulas ( [ eq : qprops ] ) and ( [ eq : yukinteger ] ) .. ] throughout this paper the prime @xmath1 is fixed and the constructions involve choices which may depend on @xmath1 . for the integrality conjectures for families of calabi - yau threefolds , like the above quintics
, one must however deal with all primes .
thus one is confronted with the challenge to ` glue ' or to ` synchronize ' those choices for the various primes @xmath1 .
in @xcite we will describe the beginnings of such a synchronized set - up for `` ordinariness '' .
the present paper starts at the fairly abstract heights of @xmath5-crystals and wants to find out from there whether the path of ordinariness leads to interesting vistas .
in @xcite we will start at a less abstract more geometric level .
we recall the theory of ordinary hodge @xmath5-crystals ; the general reference is @xcite .
we work over a perfect field @xmath26 of characteristic @xmath0 and explicitly keep track of which results require passing to the algebraic closure @xmath27 of @xmath26 .
let @xmath28 be the ring of witt vectors of @xmath26 and @xmath29 the field of fractions of @xmath28 .
so @xmath28 is a local ring with residue field @xmath26 ; its field of fractions @xmath29 has characterisitic @xmath30 ; its maximal ideal is @xmath31 and @xmath28 is complete and separated for the @xmath1-adic topology : @xmath32 .
moreover we need the rings of formal power series @xmath33\:,\quad a = w[[t_1,\ldots , t_n ] ] .\ ] ] a _ crystal over @xmath34 _ is , by definition , a finitely generated free @xmath35-module @xmath36 together with a connection @xmath37 which is integrable and @xmath1-adically topologically nilpotent ; this means that if we set @xmath38 with the derivation @xmath39 viewed as a linear map @xmath40 , then @xmath41 and @xmath42 in the @xmath1-adic topology on @xmath43 . on the characteristic @xmath1 rings @xmath26 and @xmath34 one has the _ frobenius endomorphism _ @xmath44 raising elements to their @xmath1-th power : @xmath45 . on the perfect field @xmath26
this is an automorphism and it lifts canonically to an automorphism on @xmath28 , also denoted by @xmath44 .
there are many different _ lifts of frobenius on @xmath35
_ , i.e. ring endomorphisms @xmath46 of @xmath35 which restrict to @xmath44 on @xmath28 and reduce modulo @xmath1 to @xmath44 on @xmath34 . if @xmath36 is an @xmath35-module and
@xmath47 is a lift of frobenius we write @xmath48 where @xmath49 is @xmath35 viewed as an @xmath35-@xmath35-bimodule with left structure via the identity map @xmath50 and right structure via @xmath47 .
the natural map @xmath51 is @xmath46-linear , i.e. @xmath52 .
now let @xmath53 be a crystal as above and let @xmath54 be two lifts of frobenius .
since @xmath46 and @xmath55 are equal modulo @xmath1 the connection @xmath56 provides , via a kind of taylor expansion , a canonical isomorphism of @xmath35-modules @xmath57 more precisely , the map @xmath58 is given by the formula @xmath59 note that for @xmath60 the rational number @xmath61 has @xmath1-adic valuation @xmath62 .
[ fcrystal ] one says that a crystal @xmath53 is an _
@xmath5-crystal over @xmath34 _
if for every lift of frobenius @xmath47 there is given a homomorphism of @xmath35-modules @xmath63 which is horizontal for the connection @xmath56 , i.e. the square @xmath64 \mbox{\scriptsize{$f(\phi)\phi^*$}}\hspace{-1em}&\downarrow&&\downarrow & \hspace{-2.5em } \mbox{\scriptsize{$\phi\otimes f(\phi)\phi^*$ } } \\[.6ex ] & h&\stackrel{\nabla}{\longrightarrow}&{\omega}_{a / w}^1\otimes h & \end{array}\ ] ] is commutative , and such that for every pair of lifts of frobenius @xmath54 @xmath65 moreover @xmath66 should be an isomorphism . + if for one , and hence every , lift of frobenius @xmath47 the homomorphism @xmath67 in ( [ eq : frobcrys ] ) is an isomorphism one says that @xmath36 is a _
unit @xmath5-crystal_. combining formula ( [ eq : fvar ] ) with the taylor expansion formula for @xmath68 one gets @xmath69 the horizontality relation for @xmath67 and @xmath56 ( [ eq : horizontal ] ) makes it possible to solve the problem of finding @xmath56-flat sections ( i.e. essentially solving differential equations ) by finding @xmath70-fixed sections ( i.e. essentially solving polynomial equations ) .
indeed , one has the following result .
[ eq : solve ] suppose @xmath71 is such that @xmath72 for some lift of frobenius @xmath47 .
then @xmath73 and @xmath74 holds for every lift of frobenius @xmath55 . by definition
there is for every @xmath75 an element @xmath76 such that @xmath77 . from this one
gets by induction for every @xmath75 and @xmath60 elements @xmath78 such that @xmath79 and hence @xmath80 for every @xmath2-form @xmath81 .
+ if @xmath72 , then ( [ eq : horizontal ] ) yields for every @xmath60 @xmath82 taking @xmath83 we see @xmath84 . the equality @xmath74 for any lift of frobenius @xmath55 now follows from ( [ eq : ftaylor ] ) .
let @xmath36 be an @xmath5-crystal and the maps @xmath67 .
] and @xmath85 . for a lift of frobenius @xmath47 and an integer @xmath86
we set @xmath87 these are @xmath34-submodules of @xmath88
. one can show that they are independent of the chosen lift of frobenius @xmath46 ; see @xcite 1.3 .
trivially , @xmath89 for @xmath90 and @xmath91 for @xmath92 .
moreover , it can be shown that there is an integer @xmath93 such that @xmath94 and @xmath95 .
one then says that _
@xmath36 is of level @xmath96_. thus , one finds the filtrations @xmath97 ( [ eq : filcon2 ] ) is called the _ conjugate filtration _ and ( [ eq : filhodge2 ] ) is called the _
hodge filtration_. the connection @xmath56 on @xmath36 induces a connection @xmath98 on @xmath88 .
the conjugate filtration is horizontal for @xmath98 and the hodge filtration satisfies the _ griffiths transversality condition _ ,
i.e. for all @xmath86 : @xmath99 let @xmath100 and @xmath101 denote the graded modules associated with the conjugate and hodge filtrations respectively .
if @xmath100 is a free @xmath34-module , then so is @xmath101 and for every @xmath86 the modules @xmath102 and @xmath103 are canonically isomorphic .
the rank of the free @xmath34-module @xmath104 is called _ the @xmath86-th hodge number of @xmath36_. [ defhodgecrystal ] a _
hodge @xmath5-crystal _ over @xmath34 is an @xmath5-crystal @xmath36 over @xmath34 equipped with a filtration by free @xmath35-submodules @xmath105 ( called the hodge filtration on @xmath36 ) which lifts the hodge filtration from @xmath88 and satisfies griffiths transversality , i.e. for every @xmath86 @xmath106 [ defordinary ] an @xmath5-crystal @xmath36 over @xmath34 is said to be _ ordinary _ if @xmath100 is a free @xmath34-module and the conjugate filtration and hodge filtration on @xmath88 are opposite , i.e. @xmath107 [ propordinary ] let @xmath36 be an @xmath5-crystal over @xmath34 such that @xmath100 is a free @xmath34-module .
then @xmath36 is ordinary if and only if there is a filtration by sub-@xmath5-crystals @xmath108 such that for every @xmath86 @xmath109 where @xmath110 is a unit @xmath5-crystal and @xmath111 is tate twist , i.e. @xmath112 is the same @xmath35-module with connection as @xmath110 , but for every lift of frobenius @xmath46 the map @xmath70 on @xmath112 is @xmath113 on @xmath110 .
the filtration ( [ eq : ufilt ] ) is unique .
[ prophodgedeco ] for an ordinary hodge @xmath5-crystal @xmath36 the filtrations @xmath114 and @xmath115 are opposite , i.e. for every @xmath86 @xmath116 as a consequence one has a decomposition
@xmath117 the above proposition gives a first result on the structure of ordinary hodge @xmath5-crystals .
stronger results ( theorem [ thmstructure ] below ) can be obtained over the algebraic closure @xmath27 of the base field @xmath26 .
let @xmath118 the ring of witt vectors of @xmath27 and @xmath119\,,\qquad{{\overline{a}}}={{\overline{w}}}[[t_1,\ldots , t_n ] ] \,,\qquad { { k^\mathrm{u}}}={{\overline{w}}}[\textstyle{\frac{1}{p}}].\ ] ] one can base change an @xmath5-crystal @xmath36 over @xmath34 to an @xmath5-crystal @xmath120 over @xmath121 as follows . by tensoring with @xmath122
one gets the free @xmath122-module with connection @xmath123 if @xmath124 is a lift of frobenius such that @xmath125 , then the @xmath35-linear map @xmath126 gives the @xmath122-linear map @xmath127 . if @xmath128 is an arbitrary lift of frobenius , the connection @xmath56 provides an isomorphism @xmath129 and we can define @xmath130 . [ defdivisible ] let @xmath120 be a hodge @xmath5-crystal over @xmath121 with hodge filtration @xmath131 .
one says that @xmath120 is _ divisible _ if for some lift of frobenius
@xmath46 @xmath132 if @xmath133 and ( [ eq : divisible ] ) holds for one lift of frobenius , then it holds for every lift of frobenius ( see @xcite 5.0 ) .
thorme 1.4.2 of @xcite describes the structure of ordinary hodge @xmath5-crystals of level @xmath134 .
the following theorem generalizes that result to higher levels .
[ thmstructure ] let @xmath120 be a divisible ordinary hodge @xmath5-crystal over @xmath121 of level @xmath135 with filtrations @xmath114 and @xmath115 as in ( [ eq : ufilt ] ) and ( [ eq : hodgelift ] ) respectively .
then @xmath136 and there is a basis @xmath137 of @xmath120 with @xmath138 for all @xmath139 and there is a matrix @xmath140 with entries in @xmath141 $ ] , such that * @xmath142 with @xmath143 an @xmath144-matrix , @xmath145 for @xmath146 , + @xmath147identity - matrix , the matrix @xmath148 has entries in @xmath149 .
* the matrix of the connection @xmath56 with respect to the basis @xmath150 is @xmath151 * for every lift of frobenius @xmath55 the matrix of the map @xmath152 with respect to the basis @xmath150 is @xmath153 where @xmath154 is the diagonal matrix @xmath155 with the entry @xmath156 repeated @xmath157-times . for @xmath158
there is an isomorphism of @xmath122-modules @xmath159 with @xmath160 a unit @xmath5-crystal . according to @xcite prop.1.2.2
there is an @xmath122-basis @xmath161 of @xmath160 such that @xmath162 for every lift of frobenius @xmath163 .
lift the basis @xmath161 of @xmath160 to a basis @xmath164 of @xmath165 .
this gives the basis @xmath137 for @xmath120 .
equations ( [ eq : unitbase ] ) and griffiths transversality show that the connection matrix with respect to this basis has the following shape @xmath166 with @xmath167 a matrix of size @xmath168 with entries in @xmath169 .
integrability of the connection implies for all @xmath86 @xmath170 the entries of @xmath167 are closed @xmath2-forms and hence exact , by the poincar lemma .
so there is a matrix @xmath171 of size @xmath168 with entries in @xmath141 $ ] such that @xmath172 using the equations in ( [ eq : integrability ] ) and induction on @xmath173 one sees that for @xmath174 there exist matrices @xmath143 of size @xmath175 with entries in @xmath141 $ ] such that @xmath176 for @xmath146 we set @xmath177-zero - matrix and for @xmath178 we set @xmath179-identity - matrix .
we collect the matrices @xmath143 into one ( block structured ) matrix @xmath180 then ( [ eq : etatau1 ] ) and ( [ eq : etatau2 ] ) are equivalent with @xmath181 we fix the constants of integration by taking the constant term @xmath182 of @xmath140 as in ( [ eq : t0 ] ) below .
take the lift of frobenius @xmath183 given by @xmath184 with @xmath44 the standard frobenius map on @xmath118 .
equations ( [ eq : unitbase ] ) and the assumption on divisibility show @xmath185 therefore the matrix @xmath186 of the @xmath46-linear map @xmath187 has the shape @xmath188 with @xmath154 as in ( [ eq : p ] ) and @xmath189 an uppertriangular matrix with entries in @xmath122 and @xmath2 s along the diagonal .
let @xmath190 denote the constant term of @xmath189 .
after these preparations we define @xmath140 to be the unique matrix which satisfies the differential equation ( [ eq : dt ] ) and the initial condition @xmath191 this product ( with the factors ordered from right to left for increasing @xmath192 ) converges since the matrix @xmath193 has entries in @xmath194 .
the fact that @xmath67 is horizontal for the connection @xmath56 ( see diagram ( [ eq : horizontal ] ) ) is expressed by the matrix equation @xmath195 multiplying this matrix equation by @xmath140 and using equation ( [ eq : dt ] ) gives @xmath196 and hence @xmath197 is constant .
since @xmath198 by ( [ eq : t0 ] ) we conclude : @xmath199 this shows that ( [ eq : frobmat ] ) holds for the particular lift of frobenius @xmath46 . in order to check it for an arbitrary lift of frobenius
we first define the new basis @xmath200 for @xmath201 $ ] so that @xmath202 is the matrix whose columns give the coordinates of the new basis vectors with respect to the old basis @xmath137 : @xmath203 the triangular shape of the matrix @xmath140 implies @xmath204 \qquad\textrm{for all}\quad i , m.\ ] ] the connection matrix with respect to the new basis is @xmath205 .
so @xmath206 the matrix of @xmath70 with respect to the new basis is @xmath207 .
so @xmath208 for all @xmath139 .
this holds for the special lift of frobenius @xmath46 ( [ eq : lift ] ) .
if @xmath55 is any lift of frobenius ( [ eq : ftaylor ] ) and ( [ eq : dnewbase ] ) show @xmath209 .
thus for any lift of frobenius @xmath55 we have @xmath210 the matrix @xmath211 of @xmath152 with respect to the basis @xmath150 is therefore @xmath212 as claimed in ( [ eq : frobmat ] ) .
to finish the proof of theorem [ thmstructure ] we note that ( [ eq : t0 ] ) implies that the matrix @xmath148 has entries in @xmath149 for @xmath213 .
we now specialize the general result of theorem [ cy3structure ] to crystals of the type that appears as the third crystalline cohomology of an ordinary calabi - yau threefold over the algebraically closed field @xmath27 .
more precisely , we want to see what happens , if we combine the hypotheses in theorem [ cy3structure ] with the hypotheses for the variation of hodge structure of calabi - yau threefolds , which are explicitly stated in @xcite 2 .
so _ from now on @xmath214 and @xmath120 is a divisible ordinary hodge @xmath5-crystal over @xmath121 of level @xmath4 _ with filtrations @xmath215 such that in the decomposition @xmath216 , @xmath217 @xmath165 would be denoted as @xmath218 in the hodge theoretic setting of @xcite .
+ moreover we assume that there is given a non - degenerate alternating bilinear form @xmath219 such that for all @xmath220 and for every lift of frobenius @xmath183 @xmath221 \langle f(\phi)\phi^ * ( x),f(\phi)\phi^*(y)\rangle & = & p^3\phi(\langle x , y\rangle )
. \end{array}\ ] ] we also require the riemann bilinear relations : @xmath222 [ def cy3crys ] an _ ordinary cy3 crystal _ over @xmath121 is a divisible ordinary hodge @xmath5-crystal of level @xmath4 over @xmath121 equipped with a non - degenerate alternating bilinear form @xmath223 such that ( [ eq : hodgenumbers ] ) , ( [ eq : formconditions ] ) , ( [ eq : riemann ] ) are satisfied . in view of the griffiths tranversality condition ( [ eq : transversality ] ) the connection @xmath56 induces an @xmath122-linear map @xmath224 where @xmath225 .
[ cy3structure ] let @xmath214 and let @xmath120 be an ordinary cy3 crystal over @xmath121 .
then there is a basis @xmath226 , such that @xmath227 and such that the gramm matrix of the form @xmath223 with respect to this basis is @xmath228 with @xmath229-identity - matrix .
moreover there is a matrix @xmath140 such that the matrix of the connection @xmath56 with respect to the basis @xmath230 is @xmath231 and such that for every lift of frobenius @xmath55 the matrix of the map @xmath152 with respect to the basis @xmath230 is @xmath232 with @xmath154 the ( block ) diagonal matrix @xmath233 .
this matrix @xmath140 factors as means matrix transpose ] @xmath234 where @xmath235 are matrices with entries in @xmath141 $ ] of respective sizes @xmath236 , @xmath237 , @xmath237 , @xmath238 .
assume in addition to the preceding assumptions that the map in ( [ eq : conngrhodge ] ) is an isomorphism , and hence @xmath239 .
let @xmath240 $ ] denote the components of the @xmath237-matrix @xmath241 : @xmath242 then the matrix @xmath243 is invertible over @xmath244 $ ] and @xmath245 take any @xmath122-basis of @xmath120 and matrix for which the statements in theorem [ thmstructure ] hold and denote these as @xmath246 and @xmath247 respectively .
so @xmath248 also take the basis @xmath249 as in ( [ eq : filnewbase ] ) , related to the previous basis by the transformation with the matrix @xmath247 ; see ( [ eq : newbasetrafo ] ) .
then @xmath250 and hence @xmath251 for all @xmath252 .
this implies @xmath253 for all @xmath252 .
moreover we have @xmath254 with @xmath255 smallest integer @xmath256 , and @xmath257 .
this implies @xmath258 and hence @xmath259 if @xmath260 , whereas @xmath261 if @xmath262
. now take the ( block structured ) matrix with entries in @xmath25 : @xmath263 this matrix is invertible because the form @xmath223 is non - degenerate .
we use it for the basis transformations @xmath264 for the new basis @xmath265 , the ( block structured ) gramm matrix is @xmath266 we set @xmath267 so that we have the basis relation @xmath268 the block structure of the matrix @xmath140 is @xmath269 the statements contained in formulas ( [ eq : dnewbase ] ) , ( [ eq : fnewbase ] ) , ( [ eq : conmat ] ) , ( [ eq : frobmat ] ) are still valid with respect to the new bases .
the gramm matrix of the form @xmath270 with respect to the basis @xmath230 is @xmath271 the riemann bilinear relations ( [ eq : riemann ] ) boil down to @xmath272 in view of ( [ eq : gramme ] ) this means firstly that the gramm matrix is indeed as in ( [ eq : egramm ] ) and secondly that @xmath273 the latter equalities are equivalent with the factorization of @xmath140 as in ( [ eq : tfac ] ) , with @xmath274 from equations ( [ eq : etatau1 ] ) , ( [ eq : etatau2 ] ) , ( [ eq : taurel ] ) one gets immediately @xmath275 and with some straightforward extra calculation @xmath276 now assume that the map in ( [ eq : conngrhodge ] ) is an isomorphism .
the partial derivations @xmath277 ( @xmath278 ) constitute a natural basis for @xmath279 . a natural basis for @xmath280
consists of the maps @xmath281 ( @xmath282 ) .
with respect to these bases the map ( [ eq : conngrhodge ] ) is given by the matrix @xmath283 this matrix is therefore invertible .
equation ( [ eq : tau13b ] ) is an obvious restatement of ( [ eq : tau13 ] ) , while equation ( [ eq : tau12b ] ) follows from ( [ eq : tau12 ] ) and ( [ eq : tau13b ] ) . in the preceding proof
we have essentially followed the same algebraic path as bryant and griffiths in their analysis of the variation of the hodge structure of ( complex ) calabi - yau threefolds @xcite . formulas ( [ eq : yuk ] ) , ( [ eq : tau13b ] ) , ( [ eq : tau12b ] ) are familiar from the theory of bryant and griffiths .
the following computation shows that the equally familiar formula for the _ cubic form of bryant and griffiths _ ( also known as _
yukawa coupling _ ) holds in the present situation as well : [ coryukawa ] in the situation of theorem [ cy3structure ] one has @xmath284 here @xmath285 is shorthand for @xmath286 . @xmath287 all hypotheses and notations of theorem [ cy3structure ] remain in force . the divisibility hypothesis ( [ eq : divisible ] ) and formula ( [ eq : frobmat2 ] )
imply that for every lift of frobenius @xmath55 the matrix @xmath288 has entries in @xmath122 .
this means in particular @xmath289 from theorem [ thmstructure ] we also know that @xmath290 for @xmath282 . by a lemma of dwork this implies ( see @xcite cor.1.4.5 ) that , if we set @xmath291 then @xmath292 is well defined and @xmath293 @xcite cor.1.4.7 now yields that @xmath294 is a regular system of parameters for the ring @xmath122 i.e. @xmath295\:=\:{{\overline{w}}}[[q_1 - 1,\ldots , q_h-1]].\ ] ] the elements @xmath296 are called the _
canonical coordinates_. there is an associated _ canonical lift of frobenius _
@xmath297 for every lift of frobenius @xmath55 the matrix of the map @xmath152 with respect to the basis @xmath230 is given by formula ( [ eq : frobmat2 ] ) .
a straightforward computation makes the entries of this matrix explicit .
one then sees that exactly for the canonical lift of frobenius @xmath298 this matrix takes the following elegant form : + _ for the canonical lift of frobenius @xmath298 the matrix of @xmath299 with respect to the basis @xmath230 is _
@xmath300 this formula generalizes @xcite ( 1.4.7.1 ) to the cy3 case .
we summarize the preceding discussion in [ integralities ] all hypotheses and notations of theorem [ cy3structure ] remain in force . define @xmath301
then @xmath302 and @xmath294 is a regular system of parameters for the ring @xmath122
i.e. @xmath303\:=\:{{\overline{w}}}[[q_1 - 1,\ldots , q_h-1]].\ ] ] the elements @xmath296 are called the _
canonical coordinates_. there is an associated _ canonical lift of frobenius _
@xmath304 moreover @xmath305 and @xmath306.\ ] ] ( [ eq : yukinnerproduct ] ) and ( [ eq : yukinteger ] ) follow directly from ( [ eq : matcanfrob ] ) and the divisibility hypothesis ( [ eq : divisible ] ) .
having a good hold on the structures when the base field is algebraically closed we now turn to a base field which is just a _ perfect field _
so : + _ let @xmath214 and let @xmath36 be a divisible ordinary hodge @xmath5-crystal of level @xmath4 over @xmath34 equipped with a non - degenerate alternating bilinear form @xmath223 such that ( [ eq : hodgenumbers ] ) , ( [ eq : formconditions ] ) , ( [ eq : riemann ] ) hold and such that ( [ eq : conngrhodge ] ) is an isomorphism _ , everything with @xmath36 in place of @xmath120 .
let @xmath120 be the hodge @xmath5-crystal over @xmath121 obtained from @xmath36 by base change from @xmath26 to @xmath27 .
then all results of section [ sectioncy3 ] hold for @xmath120 .
we use the notations of section [ sectioncy3 ] for referring to matters of @xmath120 .
let @xmath315 and @xmath316 denote the respective images of @xmath307 and @xmath310 under the isomorphisms @xmath317 take the lift of frobenius @xmath47 such that @xmath318 for @xmath282 . then in the unit @xmath5-crystal @xmath319 @xmath320 with @xmath321 , @xmath322 , @xmath323 .
the product @xmath324 converges in @xmath35 , because @xmath325 .
moreover there is an element @xmath326 such that @xmath327 .
thus @xmath328 on the other hand , ( [ eq : frobmat2 ] ) implies @xmath329 and hence in the ( untwisted ) unit @xmath5-crystal @xmath330 we have @xmath331 since @xmath332 we see that @xmath333 and hence @xmath334 .
thus @xmath335 with @xmath336 .
we set for @xmath278 @xmath337 where @xmath338 .
lemma [ lemmaf ] shows @xmath339 and hence @xmath340 .
+ on the other hand , @xmath341 implies @xmath342 from ( [ eq : kodaira - spencer ] ) we know that @xmath343 is precisely the matrix of the isomorphism @xmath344 .
since @xmath345 is an @xmath122-basis of @xmath346 , we now see that @xmath347 is an @xmath35-basis for @xmath348 .
next we set for @xmath282 @xmath349 then @xmath350 the form @xmath270 induces a duality between @xmath351 and @xmath348 .
this fact together with ( [ eq : omj4 ] ) shows that @xmath352 which a priori are in @xmath353 , lie in fact in @xmath351 and constitute the basis of @xmath351 dual to the basis @xmath354 of @xmath348 .
[ propomega ] starting from @xmath361 define @xmath311 by ( [ eq : omegaf ] ) , define @xmath362 by ( [ eq : omj1 ] ) and define @xmath363 and @xmath364 by ( [ eq : omj5 ] ) resp .
( [ eq : omj3 ] ) .
write @xmath365 then @xmath366 is a basis for @xmath36 such that @xmath367 the matrix @xmath368 in the basis transformation @xmath369 ( i.e. the @xmath370-th column gives the coordinates of @xmath371 with respect to @xmath230 ) is @xmath372 recall that in the proof of theorem [ cy3structure ] we constructed a basis @xmath373 from the basis transformation @xmath374 the relation with the basis @xmath366 is therefore @xmath375 the virtues of @xmath265 are @xmath376 and for every lift of frobenius @xmath47 @xmath377 thus one finds the following analogue of formulas ( [ eq : conmat2 ] ) and ( [ eq : frobmat2 ] ) let the hypotheses and notations be as in proposition [ propomega ] . then the matrix of the connection @xmath56 with respect to the basis @xmath366 is @xmath378 and for every lift of frobenius @xmath47 the matrix of the map @xmath70 with respect to the basis @xmath366 is @xmath379 with matrices @xmath140 and @xmath154 as in theorem [ cy3structure ] and @xmath368 as in proposition [ propomega ] .
[ picard - fuchs ] up to now we have discussed all structures in terms of a connection , to be thought of as the gauss - manin connection of a family of calabi - yau threefolds . in the literature descriptions of canonical coordinates and yukawa coupling for complex calabi - yau threefolds near
the large complex structure limit usually start from the picard - fuchs equations and their solutions .
we want to point out how such a description can also be seen in our discusssion of ordinary cy3 crystals .
the hypotheses and notations are as before in this section .
we start with a non zero element @xmath307 in @xmath380 . in the geometric situation
@xmath380 would be @xmath381 and thus @xmath307 is the analogue of a nowhere vanishing global @xmath4-form .
the connection induces an action of differential operators ( in @xmath382 ) on @xmath36 .
the differential operators which annihilate @xmath307 are called _ the picard - fuchs operators_. we expand @xmath307 with respect to the basis @xmath373 : @xmath383 since @xmath384 for all @xmath86 , all @xmath385 are annihilated by the picard - fuchs operators , i.e. are solutions of the picard - fuchs equations .
this set of solutions is split into subsets @xmath386 according to the action of frobenius on the basis vectors @xmath265 . in the complex setting near the large complex structure point such a splitting of a basis of the solution space of the picard - fuchs equations
is made according to the monodromy action , i.e. according to the degree of the logarithmic terms in the solution .
since @xmath387 and @xmath388 the column vector @xmath389 is the last column of the matrix @xmath390 . thus @xmath391 and @xmath392 is the last column of the matrix @xmath140 .
this is analogous to the step of dividing the solutions of the picard - fuchs equations by the unique solution which is holomorphic and has value @xmath2 at the large complex structure point .
99 bloch , s. , k. kato , _
@xmath1-adic tale cohomology _ ,
publ . math .
ihes * 63 * , 1986 , 107152 .
bryant , r. , and p. griffiths , _ some observations on the infinitesimal period relations for regular threefolds with trivial canonical bundle _ , arithmetic and geometry vol .
ii , m. artin and j. tate ( eds . ) , progress in math .
36 , birkhuser , boston , 1983 , pp .
candelas , p. , x. de la ossa , p. green , l. parkes , _ a pair of calabi - yau manifolds as an exactly soluble superconformal theory _ , essays on mirror manifolds , s - t .
yau ( ed.),international press , hong kong , 1992 , pp .
deligne , p. , _ cristaux ordinaires et coordonnes canoniques _ , surfaces algbriques , j. giraud , l. illusie and m. raynaud ( eds . ) , lecture notes in math .
868 , springer - verlag , berlin , 1981 , pp .
deligne , p. , _ local behavior of hodge structures at infinity _ , mirror symmetry ii , b. greene and s - t .
yau ( eds . ) , ams / ip studies in advanced math .
1 , amer . math . soc . and international press , 1997 , pp . 683699 .
friedman , r. , _ on threefolds with trivial canonical bundle _ , complex geometry and lie theory , j. carlson , h. clemens and d. morrison ( eds . ) , proc .
pure math .
vol.53 , amer .
soc . , providence , ri , 1991 , pp.103134 .
katz , n. , _ travaux de dwork _ , sminaire bourbaki exp.409 , lecture notes in math . 317 , springer - verlag , berlin , 1973 , pp.167200 .
morrison , d. , _ mirror symmetry and rational curves on quintic threefolds : a guide for mathematicians _ , j. of the amer .
* 6 * ( 1993 ) , 223247 ; alg - geom/9202004 lian , b. , and s - t .
mirror maps , modular relations and hypergeometric series i _ , hep - th/9507151 stienstra , j. , _ the ordinary limit for varieties over @xmath393 $ ] _ , these proceedings . | we show that crystals with the properties of crystalline cohomology of ordinary calabi - yau threefolds in characteristic @xmath0 , exhibit a remarkable similarity with the well known structure on the cohomology of complex calabi - yau threefolds near a boundary point of the moduli space with maximal unipotent local monodromy
. in particular , there are canonical coordinates and an analogue of the prepotential of the yukawa coupling .
moreover in formulas ( [ eq : qprops ] ) and ( [ eq : yukinteger ] ) we show @xmath1-adic analogues of the integrality properties for the canonical coordinates and the prepotential of the yukawa coupling , which have been observed in the examples of mirror symmetry . |
atmospheric hazes , present in a range of solar system and extrasolar planetary atmospheres , play an important role in physical and chemical processes occurring in the atmosphere and for terrestrial planets , on the surface .
haze particles affect the radiative balance of an atmosphere , may serve as condensation nuclei for clouds and rain , play a role in fluvial and aeolian processes @xcite , and affect the elemental budget of an atmosphere and surface .
the effect on haze particles on the temperature structure of an atmosphere has implications for the habitability of a planet and an organic haze , such as seen in the atmosphere of saturn s moon titan , may potentially serve as the building blocks of life @xcite .
while haze formation in n@xmath0/ch@xmath1 atmospheres has been extensively studied for decades in the laboratory through the production of titan aerosol analogues or `` tholins '' ( see @xcite ) , the effect of other atmospheric constituents on the formation of haze in planetary atmospheres has not been well studied .
carbon monoxide ( co ) is particularly interesting because it might serve as a source of oxygen for incorporation into photochemical aerosols affecting both their radiative and chemical properties and it is found in numerous atmospheres throughout the universe .
it is present in the hazy , reducing n@xmath0/ch@xmath1 atmospheres of titan @xcite , pluto @xcite , and triton @xcite and in the h@xmath0 dominated atmospheres of the giant planets @xcite .
recently co and ch@xmath1 , in addition to handful of other molecules , have also been detected in the atmospheres of extrasolar planets ( see e.g. @xcite ) and haze layers have also been invoked to explain relatively featureless spectra of a number of exoplanets ( see e.g. @xcite ) .
it is therefore important to understand the effect of co on the formation of planetary atmospheric hazes .
a few previous titan atmosphere simulation experiments have included co in their initial gas mixtures . @xcite and @xcite focused on the effects of co on the production of gas phase products .
@xcite focused on both gas and solid phase composition and observed the formation of ketones and carbonyls in their solid phase products .
@xcite reported the detection of amino acids and nucleotide bases in the solid products .
however , none of these investigations reported the effect of co on the size and number of haze particles produced , which are important parameters for the radiative effects of haze particles and for the total organic inventory found in haze particles .
we present here an experimental investigation of the effect of co on the formation of planetary atmospheric hazes including measurements of the size and number density of haze particles produced using a spark discharge source or uv photons to irradiate a range of mixtures of ch@xmath1 , co , and n@xmath0 .
figure [ fig : experiment ] shows a schematic of our experimental setup .
previous uv and spark experiments were performed using a similar setup by @xcite and @xcite , respectively .
we introduced co ( 99.999% airgas ) in volume mixing ratios ranging from 50 ppm ( the abundance in titan s atmosphere @xcite ) to 5% and ch@xmath1 ( 99.99% airgas ) in volume mixing ratios of 0.1% and 2% ( see table 1 ) into a stainless steel mixing chamber , then filled the mixing chamber to 600 psi with n@xmath0 ( 99.999% airgas ) .
we allow the gases to mix for a minimum of 8 hours before running the experiment .
the reactant gases continuously flowed through a cold trap , to remove trace impurities in the gases , before flowing through a glass reaction cell .
we maintain a flow rate of 100 standard cubic centimeters per minute ( sccm ) using a mass flow controller ( mykrolis fc-2900 ) .
the glass cold trap consists of two lines ; one line is immersed in a slurry of 200 proof ethanol and liquid nitrogen , while the other line bypasses the cold trap entirely . for this work ,
the bypass line remained closed .
the slurry remained at a temperature of @xmath2 - 115 @xmath3c .
the temperature of the gas line into the production cell was also monitored and was found to be unaffected by the use of the cold bath .
we maintain the pressure in the reaction cell between 620 and 640 torr ( atmospheric pressure in boulder , co ) at room temperature .
we expose the reactant gases to one of two energy sources , spark discharge from a tesla coil or fuv photons , which initiate chemistry leading to particle formation .
the experimental setup is the same for both energy sources until the gases reach the reaction cells .
a tesla coil ( electro technic products ) is connected to the spark reaction cell , while the uv reaction cell is connected to a deuterium lamp with a mgf@xmath0 window ( hamamatsu l1835 ) .
aerosol particle formation results from gas phase chemistry initiated by energy from the tesla coil or the deuterium lamp .
photons play a dominant role in the dissociation and ionization of chemical species that eventually lead to the formation of aerosols in titan s atmosphere @xcite .
it is therefore of paramount importance to investigate aerosol formation from photochemistry .
the deuterium lamp we used for these experiments is a continuum source that produces photons from 115 - 400 nm ( with major peaks near 121 and 160 nm ) .
although these photons are not sufficiently energetic to directly dissociate n@xmath0 and co , @xcite and @xcite demonstrated that nitrogen is participating in the chemistry in our reaction cell .
work is ongoing in our laboratory to understand the mechanism(s ) responsible for the observed nitrogen incorporation .
it seems likely that co is dissociated through an analogous mechanism due to the similarity of their bonds .
we use the electrical discharge because it is known to dissociate the triple bonds of co and n@xmath0 and is therefore an analog of the relatively energetic environment of upper atmospheres .
however , we acknowledge that the resulting energy density is higher than the energy available in most planetary atmospheres to initiate chemistry .
we use a tesla coil that can operate at a range of voltages .
as described in @xcite , we set the tesla coil to minimize the energy density while still producing sufficient aerosol using 2% ch@xmath1 in n@xmath0 for our analytical techniques and used that setting for every experiment .
the flow exits the reaction cell and flows into a scanning mobility particle sizer ( smps ) , which measures the distribution of particle sizes .
the smps has three parts : an electrostatic classifier ( tsi 3080 ) , a differential mobility analyzer ( dma , tsi 3081 ) , and a condensation particle counter ( cpc , tsi 3775 ) .
the polydisperse aerosol first enters the dma , where an electric field is applied to the flow of particles , which are then size selected based on their electrical mobility against the drag force provided by the sheath flow .
sheath flows of either 3 l / min or 10 l / min were used depending on the range of particle sizes produced in the experiment ( covering @xmath4 ranging from 14.5 to 673 nm or 7.4 to 289 nm , respectively ) . once size - selected , the particles enter the cpc where the number of particles is measured by light scattering . in this manner
, we measure the number of particles as a function of their mobility diameter ( @xmath4 ) .
our standard flow rate of 100 sccm is determined by requirements of other instruments ( see e.g. @xcite ) and is used here for consistency ; however , the smps requires a higher flow rate .
we therefore add an additional flow of n@xmath0 after the particles exit the reaction chamber bringing the total flow rate to 260 sccm .
the dilution caused by the additional flow of n@xmath0 is accounted for during data analysis .
our _ in situ _ analysis technique prevents the particles from being exposed to earth s atmosphere and does not require sample collection , which could alter the observed particle sizes and number densities .
however , real time analysis requires higher production rates .
additionally , the smps requires pressures at or near atmospheric pressure for operation .
for those two reasons , we ran the experiments presented here at 620 - 640 torr ( boulder , co , atmospheric pressure , altitude @xmath21600 m ) . this pressure is higher than the surface pressure on pluto and triton and in titan s thermosphere where the chemical processes that result in formation of aerosol begin .
here we are interested only in comparing differences resulting from the addition of co and from the choice of energy source at our standard experimental pressure .
our previous titan simulation experiments have used 0.1% ch@xmath1 for uv experiments @xcite and 2% ch@xmath1 for spark experiments @xcite .
while 2% ch@xmath1 is analogous to titan s atmosphere , the 0.1% ch@xmath1 is determined by experimental production constraints . aerosol formation in our setup from the fuv lamp peaks near 0.1% ch@xmath1 due to optical depth in the reaction cell @xcite and
previous work has determined that the aerosol composition does not vary strongly with ch@xmath1 concentration @xcite . both for comparison purposes and to extend the range of planetary atmospheres where our results may shed light on aerosol formation , experiments were run for a range of co concentrations using 0.1% and 2% ch@xmath1 . in titan
s atmosphere the ch@xmath1 concentration has almost certainly varied over time , and since the co abundance is tied both to ch@xmath1 chemistry and the plumes of enceladus @xcite , the co abundance has almost certainly varied over time as well .
measurement of absolute mixing ratios of co and ch@xmath1 in exoplanet atmospheres is still quite difficult ( see e.g. the discussion in @xcite ) and further emphasizes the need to explore a range of co and ch@xmath1 concentrations .
the smps measurements of particle size as a function of initial co concentration are shown in panel a of figure [ fig : diameter ] . for the spark experiments ,
the addition of 50 ppm of co results in a decrease in particle diameter compared to experiments performed with no co , while even the addition of 50 ppm co results in an increase in particle size for the uv experiments . in general , the particle diameter increases as a function of increasing co starting at 50 ppm for both spark and uv energy sources regardless of methane concentration .
remarkably , the addition of 5% co results in the formation of particles with diameters 2 - 3 times larger than the experiments that did not include co ; particle distributions emphasizing this point are shown in figure [ fig : dist ] . while the strong influence of co on particle diameter could potentially be attributed to the increase of carbon in the system ,
this behavior is not observed as a function of increasing ch@xmath1 concentration in n@xmath0/ch@xmath1 in the same experiment @xcite . for uv experiments ,
particle size decreases as a function of increasing ch@xmath1 from 0.01% to 10% ch@xmath1 , while for the spark experiments the particle diameter increased until a peak at 2% ch@xmath1 and then decreased .
the number density of particles , shown in panel b of figure [ fig : diameter ] , exhibits behavior very similar to that of particle diameter ; the number density increases as a function of increasing co concentration for both energy sources at both ch@xmath1 concentrations from 50 ppm co to 5% co. as we observed with the n@xmath0/ch@xmath5 only experiments @xcite , the uv experiments always produce more particles than the spark discharge experiments . for number density , the addition of 50 ppm co results in decreases compared to no co for both spark experiments ( 0.1% ch@xmath1 and 2% ch@xmath1 ) and a slight decrease for the 0.1% ch@xmath1 uv experiment . however , the addition of 50 ppm co to the 2% ch@xmath1 uv experiment results in an increase in number density .
this indicates that the presence of co in titan s atmosphere can not be ignored in titan atmosphere simulation experiments .
the aerosol mass loading calculations are performed assuming that the density of the particle is 1 g/@xmath6 .
an extensive discussion of the assumptions made in analysis of smps measurements , as well experimental determinations of tholin particle density for n@xmath0/ch@xmath1 experiments can be found in @xcite .
a variation of particle density is observed in both spark and uv experiments in the absence of co. density calculations require additional measurements not obtained for this work .
however , the aerosol mass loading calculations are presented here , despite the assumption of density , so that the results may be compared to other works . since both particle size and number density increase , aerosol mass loading also increases as a function of increasing co concentration .
the increase is most pronounced for the spark and 2% ch@xmath1 uv experiments .
for the 2% ch@xmath1 uv experiment , an increase of more than 2 orders of magnitude is observed in the mass loading with a concentration of 5% co compared to the case where no co is used .
taken together , the measurements of particle diameter and number density demonstrate that the addition of co to n@xmath0/ch@xmath1 experiments results in more , larger particles .
co is therefore affecting both the formation and growth of aerosol particles in our experiments , even at relatively low concentrations .
further work , particularly on the composition of these particles , is necessary to fully understand the chemical mechanisms by which co is affecting particle formation and growth .
however , we have a few possible explanations for the observed behavior . first
, previous tholin works have suggested that the presence of h@xmath0 and h can decrease particle production .
this explanation is often invoked to explain results of multiple plasma experiments which demonstrate that aerosol production first increases with increasing ch@xmath1 concentration , reaches a peak , and then decreases .
the decrease is attributed to an increase in the production of h@xmath0 and h at high ch@xmath1 concentrations @xcite .
the addition of h@xmath0 has also been shown to reduce aerosol formation in early earth simulation experiments that photolyzed ch@xmath1 , co@xmath0 , and n@xmath0 @xcite .
perhaps oxygen , produced from co in our experiments , is reacting with h@xmath0 and h and removing them from the system .
this could then lead to an aerosol formation by reducing one of the limiting factors .
second , as mentioned earlier , the presence of co in the gas mixture does increase the total amount of carbon atoms present in the system .
based on our work looking at aerosol production as a function of ch@xmath1 concentration , we know that simply increasing the amount of carbon in the system does not result in the formation of more aerosol for spark or uv experiments . however , the molecule that carries the carbon atom play a significant role . in the uv experiments ,
the decrease in aerosol production as a function of increasing ch@xmath1 in the gas mixture has been attributed to the increase in optical depth at the wavelengths produced by our fuv lamp @xcite .
however , co does not absorb these wavelengths and therefore may serve as source of carbon in the experiment without increasing the optical depth in the cell ; thus more photons are available to drive aerosol chemistry . for spark experiments , increasing co in the system
increases the amount of carbon available without increasing the amount of hydrogen present in the system , which may also result in an increase of aerosol formation .
third , if the degree of oxygen participation in the chemistry is increasing as a function of increasing co then the vapor pressures of the molecules produced may be lower , on average , than the molecules produced from n@xmath0/ch@xmath1 mixtures .
this oxygen incorporation may shift the partitioning of gas and solid phase species toward the solid phase , which would result in the formation of more aerosol .
the actual chemistry occurring in the reaction cell may be a combination of these three ideas or some other possibility .
future measurement of the aerosol and gas phase composition will provide insight into the partitioning of gas and solid phase species and allow us to assess the degree that oxygen is participating in the chemistry occurring in our experiment and the possible effect of co on hydrogen chemistry .
isotopic labeling experiments will help determine the degree to which carbon in the aerosol originates from co or ch@xmath1 .
we obtained _ in situ _ particle size and number density measurements for tholins produced using co concentrations from 50 ppm to 5% and ch@xmath1 concentrations of 0.1% and 2% and two different energy sources , spark discharge and uv as summarized in table 1 .
for both energy sources and both ch@xmath1 concentrations investigated , the particle size , number density and aerosol mass loading all increase as a function of increasing co concentration above 50 ppm co. the inclusion of co has a dramatic effect on aerosol production , increasing the aerosol mass loading by orders of magnitude over the range of co mixing ratios investigated . the fact that both the particle size and number density increase indicates that inclusion of co increases both particle formation and growth . intriguingly , the behavior as a function of co mixing ratio is quite similar for both energy sources , in contrast to the behavior observed with only n@xmath0/ch@xmath1 gas mixtures where the production rate trends differ greatly with ch@xmath1 concentrations based on energy source @xcite .
products of co destruction may be decreasing the presence of h@xmath0 and h in the reaction cell , species which are believed to inhibit aerosol formation , thus resulting in an increase in particle formation and growth .
co may effect aerosol formation by serving as an additional source of carbon , without affecting the optical depth at fuv wavelengths or introducing more hydrogen into the system .
the increase in aerosol formation may also result from a shift in the partitioning between gas and solid phase species due to changes in vapor pressures of the molecules produced .
however , further work is necessary to understand the effects of co on the gas phase and particle phase composition before the role of co in aerosol formation can be fully understood . , r. , irwin , p. g. j. , teanby , n. a. , lellouch , e. , bzard , b. , vinatier , s. , nixon , c. a. , fletcher , l. , howett , c. , calcutt , s. b. , bowles , n. e. , flasar , f. m. , & taylor , f. w. 2007 , icarus , 186 , 354 , l. a. , tomasko , m. g. , archinal , b. a. , becker , t. l. , bushroe , m. w. , cook , d. a. , doose , l. r. , galuszka , d. m. , hare , t. m. , howington - kraus , e. , karkoschka , e. , kirk , r. l. , lunine , j. i. , mcfarlane , e. a. , redding , b. l. , rizk , b. , rosiek , m. r. , see , c. , & smith , p. h. 2007 , planet .
space sci .
, 55 , 2015 | organic haze plays a key role in many planetary processes ranging from influencing the radiation budget of an atmosphere to serving as a source of prebiotic molecules on the surface .
numerous experiments have investigated the aerosols produced by exposing mixtures of n@xmath0/ch@xmath1 to a variety of energy sources .
however , many n@xmath0/ch@xmath1 atmospheres in both our solar system and extrasolar planetary systems also contain co. we have conducted a series of atmosphere simulation experiments to investigate the effect of co on formation and particle size of planetary haze analogues for a range of co mixing ratios using two different energy sources , spark discharge and uv .
we find that co strongly affects both number density and particle size of the aerosols produced in our experiments and indicates that co may play an important , previously unexplored , role in aerosol chemistry in planetary atmospheres . |
solitons are solutions of non - linear wave equations that maintain their shape during time evolution .
they appear in different phenomena investigated in the fields of non - linear optics , ultra - cold atomic gases and in many other physical systems .
both in the optics and atomic gases bright and dark solitons have been observed experimentally @xcite .
the former corresponds to a localized pulse propagating without a change of shape through an optical self - focusing kerr medium or to a localized matter - wave packet formed by attractively interacting atoms @xcite .
dark solitons are related to dark holes on continuous - wave background that can be created in a self - defocusing kerr medium or in repulsively interacting massive particles @xcite . in the areas of non - linear optics and ultra - cold atomic gases , the non - linear wave equations are results of the approximation that assumes occupation of single modes by a large number of photons or atoms
these equations provide a very accurate description of the soliton experiments @xcite .
evidently there are observables related to higher order correlations which can not be obtained within a mean - field description @xcite .
they are however very difficult for experimental detection .
rapid development of experimental techniques in the area of ultra - cold gases opens possibility for realization and precise detection of systems with a relatively small particle number .
in such a case , many - body effects that go beyond the mean - field approximation become important and experimentally relevant @xcite .
there are examples of quantum many - body systems where solutions of the schrdinger equation can be obtained analytically . in one - dimensional ( 1d ) space ,
the bethe ansatz approach turns out to be invaluable in description of bosonic or fermionic atoms and various mixtures of atomic species interacting via contact @xmath0-potentials @xcite .
the celebrated lieb - liniger model describes bosonic particles with contact interactions in a 1d space @xcite . in the case of periodic boundary conditions , all energy eigenstates of the system
are also eigenstates of the unitary operator that translates all particles by the same distance .
thus , it is not clear which solutions of the many - body schrdinger equation can be identified with the mean - field solitons because the latter break the translational symmetry .
the identification is particularly difficult in the case of repulsive particle interactions where a dark soliton does not correspond to the lowest energy solution of the non - linear mean - field equation .
nearly 40 years ago it was conjectured that eigenstates belonging to the so - called type ii excitation spectrum of the lieb - liniger model were related to the mean - field dark soliton solutions due to the coincidence of the spectrum with the soliton dispersion relation @xcite .
there are many publications that confirm this conjecture @xcite .
especially it is shown that a proper superposition of the type ii eigenstates allows one to prepare states where dark soliton signatures are visible in the reduced one - particle density and in the behavior of the matrix element of the bosonic field operator @xcite .
demonstration that a single type ii eigenstate reveals dark soliton character has been given recently only @xcite .
the single eigenstate possesses the translational symmetry and consequently the corresponding reduced one - particle density is uniform with no soliton signature . however , one - particle density is related to density of atoms averaged over many realizations of the same experiment . for many - body systems ,
already a single experiment allows one to plot a particle density and such a single outcome can be very different from the average result as demonstrated in different many - body problems @xcite .
it turns out that dark solitons can emerge in the course of measurement of particle positions @xcite .
that is , after the measurement of a single particle , the probability density for the measurement of the position of the second particle is not longer uniform because the translational symmetry is broken in the first measurement .
if we continue the particle measurements , the shape of probability density for measurements of consecutive particles approaches the dark soliton profile .
when @xmath1 particles is detected and only the last particle is left its wave - function matches perfectly a soliton solution of the mean - field equation @xcite . in ref .
@xcite we have demonstrated the emergence of single and double solitons in the measurement of positions of particles prepared in type ii eigenstates in the weak interaction limit .
we have also analyzed results of the measurement in the case of moderate interactions . here , we focus on dynamics of many - body states that result from translational symmetry breaking of initial symmetric eigenstates and try to answer the question if signatures of the mean - field evolution of dark solitons can be observed .
moreover , we analyze the bose system prepared in type ii eigenstates in the strong interaction regime .
the paper is organized as follows . in sec .
[ liebliniger ] we describe shortly the lieb - liniger model and present the key elements of the numerical methods that we use . in sec .
[ weak ] dark solitons in the weak interaction regime are considered .
section [ strong ] presents analysis of the system for strong interactions .
both single and double soliton solutions are considered .
we conclude in sec .
[ conclc ] .
we consider @xmath2 bosons with repulsive contact interactions in the 1d space with periodic boundary conditions @xcite .
the hamiltonian of the system in the second quantization formalism reads , h=_0^ldx , [ h ] where @xmath3 is the bosonic field operator and @xmath4 is the system size ( the length of the 1d ring ) .
we use units where @xmath5 with @xmath6 the particle mass .
the strength of the interactions is characterized by = , where @xmath7 is the parameter of the interaction potential and @xmath8 denotes the average density of particles .
for @xmath9 the interactions between particles are weak , for @xmath10 the lieb - liniger model describes strongly interacting impenetrable bosons @xcite .
the eigenvalue problem for the hamiltonian ( [ h ] ) has analytical solutions in terms of the bethe ansatz @xcite .
there are two branches of the excitation spectrum of the system identified by lieb @xcite .
the first , called type i , corresponds to the bogoliubov dispersion relation in the weak interaction limit .
the interpretation of the other branch ( type ii ) was not clear initially . later on it
was found out that the type ii branch followed the dark soliton dispersion relation in the weak interaction regime @xcite .
eigenstates of the hamiltonian ( [ h ] ) are determined by a set of @xmath2 integers ( half integers ) @xmath11 for odd ( even ) total particle number @xmath2 which are parameters of the bethe equations k_jl+2_mj^n()=2i_j . for a given collection @xmath12 , solution of the bethe equations results in a set of @xmath2 real numbers @xmath13
which are called quasi - momenta and which determine total momentum and energy of the system , i.e. p(\{k}_n)&=&_j=1^nk_j=_j=1^ni_j , + e(\{k}_n)&=&_j=1^nk_j^2 , respectively @xcite . the ground state of the system corresponds to the ordered set \{i}_n=\{-,-+1, ,}. [ gss ] numbers @xmath11 ( integers or half - integers ) are distributed symmetrically around zero , are equally spaced by a unit distance , and belong to the set @xmath14 for even @xmath2 , or @xmath15 if @xmath2 is odd .
the eigenstates of the system in the lieb - liniger model read @xcite @xmath16 _ n < m .
[ ba ] it is convenient to bring here the analogy between the studied interacting bose system and the ideal fermi gas in 1d .
this analogy does not have formal character , but it provides a familiar platform to think about a structure of many - body eigenstates of the system .
the following discussion and analogies to the fermi gas can not be understood literally . first note , that the numbers @xmath12 are closely related to the quasi - momenta .
there is one to one correspondence between @xmath11 and @xmath13 , a function relating them is monotonic ( preserves ordering ) , i.e. if @xmath17 then @xmath18 .
therefore , to simplify our discussion we will sometimes identify these two quantities addressing numbers @xmath11 as ` momenta ' .
if the ground state were of a form of the slater determinant , it could be thought of as a fully occupied ` fermi sea ' .
all states with @xmath19 are occupied , while the others are empty .
the ` fermi momentum ' here is @xmath20 .
the ` fermi surface ' is composed of two states @xmath21 of opposite momenta .
the ` lowest energy ' state has ` momentum ' @xmath22 for odd @xmath2 , while is doubly ` degenerate ' for even @xmath2 .
there are two states of minimal opposite ` momenta ' @xmath23 then .
the type i elementary excitations correspond to the bogoliubov spectrum , with a phonon branch for low excitation momenta .
these are particle - like excitations : the ` particle from the fermi level ' is excited to the ` higher momentum state ' .
more precisely the ` momentum ' @xmath24 ( or @xmath25 ) is missing from the set @xmath12 and is substituted for @xmath26 , where @xmath27 . the type ii excitation spectrum is obtained by creating a hole below the fermi surface , substituting @xmath28 , for the state next to the fermi surface , @xmath29 @xcite .
there are two accessible ` momenta ' next to the fermi surface : @xmath30 , and the opposite one @xmath31 . despite the simple form of the bethe ansatz solution ( [ ba ] ) ,
it not easy to draw information from @xmath32 because number of terms in ( [ ba ] ) grows dramatically with @xmath2 . in ref .
@xcite we have performed numerical simulations of measurements of particle positions for @xmath33 employing a sequential procedure .
that is , particles have been measured one by one and at each time the conditional probability for the measurement of a next particle has been calculated @xcite .
the calculation of the conditional probability is possible because matrix elements @xmath34 , so - called form factors of the field operator , can be easily obtained by means of the determinant formulas @xcite .
it turns out that the simulations of the particle measurements can be significantly simplified with the help of monte carlo algorithm of metropolis et al .
@xcite , based on the markovian walk in the configuration space . having @xmath35 , we can employ the metropolis procedure which allows us to obtain a sequence of random sets of particles position @xmath36 from the @xmath2-dimension probability distribution .
this procedure is equivalent to direct sampling of the @xmath2-paricle probability density @xmath37 and also equivalent to the sequential choice of the particle positions applied in ref .
@xcite . in the latter case
, simulations become difficult in the strong interaction regime ( @xmath38 ) because the number of the form factors of the field operator needed to obtain converged results , increases significantly .
the monte carlo method does not suffer from this problem because the form of the eigenstates ( [ ba ] ) is the same regardless of the value of the interaction strength . in this work
we want to show that the lieb - liniger type ii excitations occur to be solitons not only at a given instant of time , while observed , but , what is also very important , they propagate like genuine solitons should do , including double - soliton solutions and their collisions . to this end
the two - point time correlations have to be studied . to observe a time evolution of the type ii soliton
, first the translational symmetry of the many body state has to be broken during measurement of positions of @xmath39 out of @xmath2 particles .
the soliton gets localized , the larger @xmath39 , the better .
then , the system of remaining @xmath40 particles evolves in time till the second measurement of positions of all remaining particles .
the question is if the second measurement reveals the soliton - like shape , shifted in the laboratory frame with respect to the initial position of solitons in agreement with the corresponding mean - field soliton dynamics . to answer this question
we shall probe the conditional probability density distribution of finding @xmath41 particles at time @xmath42 and at positions @xmath43 , provided that previously at @xmath44 another @xmath39 particles were detected in the initial state of @xmath2 particles prepared in a type ii exited lieb - linger state .
we implement the above scheme as follows .
first we choose positions @xmath45 at @xmath44 as any @xmath39 values belonging to the set @xmath46 maximizing the @xmath2-body probability distribution @xmath47 of the given state .
alternatively , we could select them randomly .
the values of these @xmath39 positions are assumed to be set in the course of the first measurement .
the detection leads to a reduction of the system wave function .
the state of remaining @xmath40 particles is given by the wave function [ phi0 ] ( \{x}_n_r ) _
\{k}_n(\{x_0}_n_i,\{x}_n_r ) , which determines a conditional probability density @xmath48 of remaining @xmath41 particles positions at the initial time . to get the time correlations between particles distribution at the initial time @xmath44 , and at the final time @xmath42
, we have to evolve the initial reduced @xmath41-body state , @xmath49 .
let us note , that it is not the eigenstate of the lieb - linger hamiltonian anymore .
expansion of @xmath49 in the basis of the eigenstates of the @xmath50-particle system allows us to get time evolution ( \{x}_n_r , t)= _ \{k}_n_re^-it e(\{k}_n_r )
c_\{k}_n_r _ \{k}_n_r(\{x}_n_r ) , [ expan ] as well as the two - time conditional probability distribution @xmath51 .
this distribution is then sampled with the help of monte carlo algorithm .
every sample corresponds to a measurement of positions of all @xmath41 particles at time @xmath42 provided that @xmath39 particles were previously found at positions @xmath45 at the initial moment .
we expect that in the limit @xmath52 , @xmath53 , and @xmath54 the first measurements of @xmath39 positions @xmath45 disturb only infinitesimally the initial type ii eigenstate .
they simultaneously perfectly localize the soliton .
the reduced state obtained after the first measurement @xmath49 should be infinitesimally close to the type ii eigenstate of the lieb - liniger model , a perfect soliton again .
here we are going to study rather small systems of several particles .
as we are going to show , effects of a small system will lead to some ` imperfections ' as compared to mean - field solutions . the expansion ( [ expan ] ) requires calculations of @xmath55 that is not easy because it involves multi - dimensional integration .
however , it turns out that the expansion can be obtained in a simpler way . choosing many different sets of the positions @xmath43 , i.e. sampling the @xmath50-dimensional space , and calculating values of @xmath49 and values of eigenstates @xmath56 which are chosen to perform the expansion
, one can obtain @xmath57 by fitting the rhs of ( [ expan ] ) to the sets of values of @xmath49 .
such a linear regression procedure turns out to be very efficient .
in the weak interaction limit ( @xmath58 ) , the ground state of the lieb - liniger hamiltonian and collectively excited states can be described by the mean - field approach where all bosons are assumed to occupy the same single particle state @xmath59 which is a solution of the gross - pitaevskii equation ( gpe ) @xcite . in the limit of the infinite system ( @xmath60 )
, the dark soliton solution of the gpe reads @xmath61 where @xmath62 is the position of the dark soliton notch and @xmath63 is the healing length that describes the width of the notch @xcite . in the case of the finite system with periodic boundary conditions ,
the dark soliton solution has to be modified because the hyperbolic tangent function alone is not able to fulfill the boundary conditions .
if @xmath64 , the dark soliton solution can be approximated by @xmath65 which describes density profile that drops to zero at the position of the soliton but the density is not stationary because the notch moves periodically along the 1d ring .
dark solitons are stationary in a frame moving with a certain velocity @xmath66 with respect to the laboratory frame . that is , there are dark soliton solutions of the stationary gpe , ( x)=(x ) , [ gpe ] with periodic boundary conditions , given in terms of jacobi functions @xcite where @xmath67 . for chosen values of the interaction strength @xmath68 and average particle density @xmath69 , one can find intervals of @xmath66 where single- , double- or multiple - soliton solutions exist @xcite .
they are characterized by the depth of the dark soliton notch d = minl , and by the winding number @xmath70 which determines a complex phase @xmath71 the wave - function @xmath72 acquires with respect to @xmath59 when the position on the ring changes by @xmath4 @xcite . ) in the measurement of positions of @xmath2 particles prepared initially in the one - hole excited eigenstate with the total momentum per particle @xmath73 ; ( a ) corresponds to the probability density and ( b ) to the phase of the wave - function . in each measurement realization , the soliton notch is localized at different position but the wave - function always matches the properly shifted mean - field solution that corresponds to the single dark soliton .
the latter solution is not plotted because it would not be distinguishable from the last particle wave - function .
panels ( c ) and ( d ) show the probability density and phase , respectively , of two examples of the last particle wave - function obtained in the measurements of positions of @xmath2 particles prepared in the one - hole excited eigenstate with @xmath74 . in this case
there are two types of the last particle wave - functions observed in the measurement realizations .
the first is related to mean - field solutions with the winding number @xmath75 [ black solid lines in ( c ) and ( d ) ] .
the corresponding distributions of their kinetic energies and average momenta are presented in ( e ) and ( f ) , respectively , see black histograms .
the other wave - functions are related to @xmath76 and they are represented by red dash lines and red filled histograms in the panels ( e)-(f ) . for each last particle wave - function obtained in the simulation , we can find a mean - field solution that follows it very well .
such mean - field solution is chosen so that its average momentum and @xmath70 are the same as those of a given last particle wave - function .
the size of the 1d space @xmath77 , @xmath78 and @xmath33 . ] particles at time @xmath42 proceeded by a previous detection of @xmath79 particles at @xmath44 .
( a ) short time evolution .
a value of the time @xmath42 is indicated in each individual panel .
initially , the @xmath2-particle system is prepared in the one - hole excited eigenstate with the total momentum per particle @xmath73 .
such a many - body eigenstate is related to the mean - field dark soliton that propagates in the laboratory frame with the velocity @xmath80 without changing its shape .
short time evolution of the histogram follows the mean - field dynamics very well .
however , in the long time dynamics , presented in ( b ) , many - body effects emerge and lead to smearing of the density notch in the histogram .
each realization of the measurement results in a wave - function of the last particle [ examples , plotted with red dash lines , are shown in ( b ) ] that reveals dark soliton profile nearly identical to the mean - field prediction but localized at different positions . in the short time
evolution the uncertainty of soliton localization around the position predicted by the mean - field description is small .
however , in the course of the time evolution the uncertainty increases that is responsible for blurring of the histogram . at @xmath81
the histogram is almost flat without clear signature of the solitons . at even longer time , we observe revival of the initial shape of the histogram .
evolution of the uncertainty is quantitatively illustrated in ( c ) where the standard deviation of the notch position of the last particle wave - functions versus time is presented .
the size of the 1d space @xmath77 , @xmath78 and @xmath33.,title="fig : " ] particles at time @xmath42 proceeded by a previous detection of @xmath79 particles at @xmath44 .
( a ) short time evolution . a value of the time @xmath42 is indicated in each individual panel .
initially , the @xmath2-particle system is prepared in the one - hole excited eigenstate with the total momentum per particle @xmath73 .
such a many - body eigenstate is related to the mean - field dark soliton that propagates in the laboratory frame with the velocity @xmath80 without changing its shape .
short time evolution of the histogram follows the mean - field dynamics very well .
however , in the long time dynamics , presented in ( b ) , many - body effects emerge and lead to smearing of the density notch in the histogram .
each realization of the measurement results in a wave - function of the last particle [ examples , plotted with red dash lines , are shown in ( b ) ] that reveals dark soliton profile nearly identical to the mean - field prediction but localized at different positions . in the short time
evolution the uncertainty of soliton localization around the position predicted by the mean - field description is small .
however , in the course of the time evolution the uncertainty increases that is responsible for blurring of the histogram . at @xmath81
the histogram is almost flat without clear signature of the solitons . at even longer time , we observe revival of the initial shape of the histogram .
evolution of the uncertainty is quantitatively illustrated in ( c ) where the standard deviation of the notch position of the last particle wave - functions versus time is presented .
the size of the 1d space @xmath77 , @xmath78 and @xmath33.,title="fig : " ] particles at time @xmath42 proceeded by a previous detection of @xmath79 particles at @xmath44 .
( a ) short time evolution .
a value of the time @xmath42 is indicated in each individual panel .
initially , the @xmath2-particle system is prepared in the one - hole excited eigenstate with the total momentum per particle @xmath73 .
such a many - body eigenstate is related to the mean - field dark soliton that propagates in the laboratory frame with the velocity @xmath80 without changing its shape .
short time evolution of the histogram follows the mean - field dynamics very well .
however , in the long time dynamics , presented in ( b ) , many - body effects emerge and lead to smearing of the density notch in the histogram .
each realization of the measurement results in a wave - function of the last particle [ examples , plotted with red dash lines , are shown in ( b ) ] that reveals dark soliton profile nearly identical to the mean - field prediction but localized at different positions . in the short time
evolution the uncertainty of soliton localization around the position predicted by the mean - field description is small .
however , in the course of the time evolution the uncertainty increases that is responsible for blurring of the histogram . at @xmath81
the histogram is almost flat without clear signature of the solitons . at even longer time , we observe revival of the initial shape of the histogram .
evolution of the uncertainty is quantitatively illustrated in ( c ) where the standard deviation of the notch position of the last particle wave - functions versus time is presented .
the size of the 1d space @xmath77 , @xmath78 and @xmath33.,title="fig : " ] let us start our comparison between the mean - field and many - body solutions with an analysis of a single - soliton which is completely dark , i.e. @xmath82 . for @xmath80 there is a single - soliton solution of the gpe ( [ gpe ] ) corresponding to the average momentum @xmath83 , for any value of the interaction strength @xmath68 @xcite .
the phase of the mean - field solution jumps by @xmath84 at the soliton position .
however , because the probability density drops to zero at the soliton notch , the jump of the phase can be also considered as @xmath84 plus any multiple of @xmath85 .
thus , the winding number is meaningless in the present case . in the many - body lieb - liniger model
one can find a one - hole excited eigenstate ( type ii eigenstate ) that is characterized by the total momentum per particle @xmath73 and the energy per particle nearly the same as the mean - field value .
the many - body eigenstate possesses the translational symmetry which is broken if we perform measurements of particle positions .
the intriguing question is if the system collapses to the mean - field solution during the measurement process .
it turns out that when we simulate many realizations of the measurement of positions of @xmath1 particles , the wave function for the last @xmath2-th particle matches the mean - field solution very well , see fig .
[ fweak1 ] .
the last particle wave - function is defined as follows ( x)_\{k}_n(\{x_0}_n-1,x ) , [ lastp ] where @xmath32 is a chosen one - hole excited eigenstate and @xmath86 are fixed positions of @xmath1 particles obtained in the measurement process .
note that in every realization of the measurement the localization of the soliton notch is random .
that is , before the measurement is performed we do not know where the soliton is going to appear because the system preserves the translational symmetry @xcite .
let us now consider a solution of the gpe that describes a gray single - soliton ( @xmath87 ) related to the average momentum @xmath88 and to the winding number @xmath75 . in the many body picture simulation of the measurement processes performed on the corresponding one - hole excited eigenstate of ( [ h ] ) , i.e. the state with @xmath74 , results in a bunch of different wave - functions for the last particle .
they correspond both to @xmath75 ( with the probability 0.79 ) and to @xmath76 ( with the probability 0.21 ) . in fig .
[ fweak1 ] we show examples of the last particle wave - functions and distributions of kinetic energies and average momenta of the last particles that we have obtained collecting results of many realizations of the measurement process .
all observed wave - functions match certain mean - field single - soliton solutions .
the choice of @xmath89 that is substituted for @xmath30 in ( [ gss ] ) in a one - hole excitation determines properties of wave - functions of the last particle obtained in simulation of the measurement process . for even @xmath2 and for @xmath90
we obtain the wave - functions that , in the mean - field approach , are related to a completely dark soliton moving with the velocity @xmath80 , cf .
[ fweak1 ] . when the chosen @xmath11 increases and approaches @xmath91 , the last particle wave - functions reveal gray solitons that reduce to the zero - momentum solution of the gpe ( [ gpe ] ) for @xmath92 .
on the other hand if @xmath11 decreases and reaches @xmath25 , the last particle wave - functions correspond to gray solitons that reduce to the plane wave solution of the gpe with the momentum @xmath93 if @xmath92 .
this is the scenario how the mean - field solitons allow us to pass smoothly between the plane wave solutions with @xmath94 and @xmath93 @xcite .
signatures of similar passing can be also observed in the many - body description .
if @xmath11 is substituted for @xmath25 ( instead of @xmath24 ) we obtain analogous relation between the results of the measurements and the mean - field solutions but with @xmath95 . in order to investigate dynamics of solitons
we follow the procedure described in the previous section . for a given one - hole eigenstate we choose positions of @xmath96 particles corresponding to the configuration of maximal probability . in this way we break the translational symmetry of an initial eigenstate and localize position , with a small uncertainty , of a soliton formed by the remaining particles . then , we evolve the remaining @xmath97 particle state for a certain period of time to get the conditional two - time probability distribution . finally we sample this distribution with the help of monte carlo algorithm . in fig .
[ fweak_dark_dyn ] we present histograms resulting from the many - body evolution of the previously described dark soliton that in the mean - field limit moves in the laboratory frame with the velocity @xmath80 .
short time dynamics shows that the profile of the histogram does not change and the notch , visible in the histogram , moves with the velocity predicted by the mean - field approach .
however , in the long time evolution we see that the depth of the notch decreases and when @xmath42 is of the order of @xmath98 the notch nearly disappears .
the time scale @xmath98 is much larger than the quantum speed limit time , i.e. a typical lifetime of a generic quantum state @xcite .
the notch reappears again if we continue the time evolution .
the smearing of the notch is a many - body effect which is related to an uncertainty of a soliton position @xcite .
the initial choice of positions of @xmath39 particles leads to localization of the soliton with a small uncertainty .
that is , in different realizations of the measurement of the remaining @xmath97 particles , the last particle wave - functions reproduce the mean - field dark soliton profile but different profiles are localized at slightly different positions [ the initial standard deviation of the soliton position is about 0.1 in fig . [ fweak_dark_dyn](c ) ] .
in the course of time evolution of the remaining @xmath97 particles , the uncertainty of the soliton position increases [ see fig .
[ fweak_dark_dyn](c ) ] and causes smearing of the soliton notch visible in the histograms in fig . [ fweak_dark_dyn](b ) .
we observe also some contributions to the histogram of particle positions coming from the last particle wave - functions corresponding to gray solitons where @xmath87 .
the average @xmath99 , however , is never greater than 0.1 and these contributions can not explain the smearing of the histograms . for a longer time evolution the profile of the histograms returns to its initial shape .
such a quantum revival phenomenon can be expected in a few - body quantum system .
we are not able to perform simulations for large @xmath2 but one may expect that for @xmath92 but with @xmath100constant , the many - body evolution follows the mean - field dynamics for time that increases linearly with @xmath2 .
( b)-[fweak_dark_dyn](c ) .
( a ) shows long time dynamics of the histogram of positions of @xmath97 particles obtained in many measurement realizations .
initially , the @xmath2-particle system is prepared in the two - hole excited eigenstate with the total momentum @xmath101 , i.e. , @xmath102 in ( [ gss ] ) .
each realization of the measurement results in a wave - function of the last particle ( examples , plotted with red dash lines , are shown in the panels ) that in most of the cases matches the mean - field double dark soliton solution that is stationary in the laboratory frame ( @xmath103 ) . in each realization , the two soliton notches are always situated at the distance @xmath104 apart but they localize at different positions . at @xmath81 the histogram is almost flat large uncertainty of the position of the solitons is responsible for this effect .
evolution of the uncertainty is illustrated in ( b ) where the standard deviation of the position of one of the notches versus time is presented .
the size of the 1d space @xmath77 , @xmath78 and @xmath33.,title="fig : " ] ( b)-[fweak_dark_dyn](c ) .
( a ) shows long time dynamics of the histogram of positions of @xmath97 particles obtained in many measurement realizations .
initially , the @xmath2-particle system is prepared in the two - hole excited eigenstate with the total momentum @xmath101 , i.e. , @xmath102 in ( [ gss ] ) .
each realization of the measurement results in a wave - function of the last particle ( examples , plotted with red dash lines , are shown in the panels ) that in most of the cases matches the mean - field double dark soliton solution that is stationary in the laboratory frame ( @xmath103 ) . in each realization , the two soliton notches are always situated at the distance @xmath104 apart but they localize at different positions . at @xmath81
the histogram is almost flat large uncertainty of the position of the solitons is responsible for this effect .
evolution of the uncertainty is illustrated in ( b ) where the standard deviation of the position of one of the notches versus time is presented .
the size of the 1d space @xmath77 , @xmath78 and @xmath33.,title="fig : " ] but the initial @xmath2-body state corresponds to the two - hole excited eigenstate for @xmath33 where @xmath105 in ( [ gss ] ) that results in the total momentum per particle @xmath106 . the mean - field double soliton corresponding to the average momentum @xmath107 propagates with the velocity @xmath108 .
the notches in the histogram moves with the velocity very close to @xmath109 .
the size of the 1d space @xmath77 and @xmath78.,title="fig : " ] but the initial @xmath2-body state corresponds to the two - hole excited eigenstate for @xmath33 where @xmath105 in ( [ gss ] ) that results in the total momentum per particle @xmath106 . the mean - field double soliton corresponding to the average momentum @xmath107 propagates with the velocity @xmath108 .
the notches in the histogram moves with the velocity very close to @xmath109 .
the size of the 1d space @xmath77 and @xmath78.,title="fig : " ] but the initial @xmath2-body state corresponds to the two - hole excited eigenstate for @xmath33 where @xmath105 in ( [ gss ] ) that results in the total momentum per particle @xmath106 . the mean - field double soliton corresponding to the average momentum @xmath107 propagates with the velocity @xmath108 .
the notches in the histogram moves with the velocity very close to @xmath109 .
the size of the 1d space @xmath77 and @xmath78.,title="fig : " ] ( a ) and fig . [ fweak2b](a ) but
the initial @xmath2-body state corresponds to the two - hole excited eigenstate where @xmath110 in ( [ gss ] ) .
the subsequent panels show moments of time where two different minima : propagate towards each other , even their depth and exchange their positions .
( b ) : distance @xmath111 between the minima , visible in ( a ) , versus time .
the size of the 1d space @xmath77 , @xmath78 and @xmath33.,title="fig : " ] ( a ) and fig .
[ fweak2b](a ) but the initial @xmath2-body state corresponds to the two - hole excited eigenstate where @xmath110 in ( [ gss ] ) .
the subsequent panels show moments of time where two different minima : propagate towards each other , even their depth and exchange their positions .
( b ) : distance @xmath111 between the minima , visible in ( a ) , versus time .
the size of the 1d space @xmath77 , @xmath78 and @xmath33.,title="fig : " ] double - soliton solutions of the gpe are related to two - hole excited eigenstates of the lieb - liniger model @xcite .
that is , when we remove two values of @xmath112 and @xmath113 in the ground state sequence ( [ gss ] ) and add new ones next to the fermi surface : @xmath114 and @xmath115 or @xmath115 and @xmath116 instead , we observe that in the measurement processes the wave - functions of the last particle match mean - field double - soliton states .
there is a double - soliton solution of the gpe that is characterized by the average momentum @xmath94 .
this solution is stationary in the laboratory frame ( @xmath103 ) and describes two completely dark solitons ( @xmath82 ) . in the many - body picture
the corresponding two - hole excited eigenstate , related to @xmath117 , can be obtained by the exchange @xmath118 and @xmath119 for @xmath114 and @xmath115 in the sequence ( [ gss ] ) for even @xmath2 .
it is also possible to obtain a mean - field solution where two completely dark solitons move together in the laboratory frame with the velocity @xmath120 .
then , the related two - hole excited eigenstate corresponds to the substitution @xmath121 in ( [ gss ] ) for even @xmath2 .
when we carry out the measurements on the many - body system prepared in one of these two two - hole excited eigenstates , wave - functions of the last particle reproduce very well the corresponding mean - field solutions , obviously , in the view of the previous discussion , localized at different positions in different realizations of the measurement process .
two - hole excitations can be chosen in many different ways .
we can find two - hole excited eigenstates that reveal signatures of solutions of the gpe ( [ gpe ] ) if we perform the measurement of particle positions .
equation ( [ gpe ] ) describes two ( or more ) identical solitons that either move together with the same velocity in the laboratory frame or do not move at all .
however , one may also expect situations where solitons move with different velocities and collide .
such a process can not be described by the stationary gpe ( [ gpe ] ) .
it is interesting if signatures of two solitons moving with different velocities can be observed in the many - body system prepared initially in a two - hole excited eigenstate .
let us begin the analysis of soliton dynamics with the two - hole excited eigenstate corresponding to the completely dark mean - field solitons that are stationary in the laboratory frame ( @xmath103 ) .
figure [ fweak2a ] show how the histogram of positions of @xmath97 particles evolves in time if @xmath96 particles have been measured at @xmath44 . similarly to the single - soliton case , short time dynamics agrees with the mean - field predictions but in the long time evolution we observe that the notches are gradually smeared and disappear at @xmath81 .
the uncertainty of the soliton positions is responsible for such a many - body phenomenon as discussed in sec .
[ ssola ] . in the second example
we consider a two - hole eigenstate obtained by the exchange of @xmath118 and @xmath122 for @xmath115 and @xmath116 in the ground state sequence ( [ gss ] ) for @xmath33 which leads to @xmath106 .
the double - soliton solution of the gpe ( [ gpe ] ) corresponding to the the same value of the average momentum @xmath107 describes two gray solitons moving together with the velocity @xmath123 .
simulations show that the notches in the histogram of positions of the last @xmath97 particles , when the first @xmath96 particles are measured at @xmath44 , move with a very similar velocity , see fig .
[ fweak2b ] .
wave - functions of the last particle obtained in each final measurement of the 3-particle system are different but all correspond to mean - field solitons that propagate also with @xmath123 .
the smearing of the soliton notches is visible for @xmath42 of the order of @xmath98 but it is less apparent than in the cases corresponding to the completely dark mean - field solitons . in the third example we would like to consider a two - hole eigenstate that reveals signatures of two solitons moving towards each other with different velocities .
we begin the simulation with an eigenstate where @xmath124 in the sequence ( [ gss ] ) for @xmath33 , then we measure positions of @xmath96 particles ( @xmath44 ) , evolve the system of the remaining @xmath97 particles in time and at different @xmath42 perform simulations of measurements of positions of 3 particles .
analysis of the wave - functions of the last particle indicates that they may be very different .
there are wave - functions that show dark and gray solitons that are clearly separated one from each other but there are also profiles where clear separation of the two solitons is not observed .
the latter cases can be related to a collision of the two solitons .
the average behavior of the three - particle system is represented by the time evolution of the histogram of the detected positions of three particles in fig .
[ fweak2c ] .
we can see that two density notches ( one nearly dark and the other gray ) approach each other , even their depth and exchange their positions as if two sloitons collide , pass each other and propagate further . to sum up
, two - hole eigenstates we have investigated can be divided into three classes .
the first class is related to completely dark solitons in the mean - field approach .
that is , when the measurement of particle positions is performed , we always observe that the last particle wave - functions match quite well dark soliton solutions with average momentum @xmath125 equal to the initial total momentum per particle @xmath126 of the many - body system .
there are two such eigenstates .
the first , obtained by the exchange @xmath127 for even @xmath2 , corresponds to @xmath117 and reveals the double - soliton mean - field solution that is stationary in the laboratory frame .
the other eigenstate , related to @xmath128 or to @xmath129 , where @xmath130 , reveals double - solitons that move with @xmath131 in the mean - field approach .
the second class consists of eigenstates which are obtained by an exchange @xmath132 with @xmath133 .
the measurements reveal different wave - functions of the last particle that match gray double - soliton solutions of the gpe ( [ gpe ] ) but corresponding to different average momentum @xmath125 in different measurement realizations .
that is , @xmath125 is not necessarly equal to @xmath126 of the initially chosen many - body eigenstate .
the third class corresponds to two - hole excitations where @xmath134 or @xmath135 but @xmath136 .
then , the last particle wave - functions reveal double - solitons with two notches that are not identical , appear at different distances one from each other and show signatures of a collision of two solitons .
such a soliton dynamics can not be described the stationary gpe ( [ gpe ] ) .
left panels are related to the one - hole excited eigenstate obtained by the exchange @xmath137 in ( [ gss ] ) , i.e. to @xmath73 , while right panels to the eigenstate where @xmath138 where @xmath74 . in ( a)-(b )
particle interactions correspond to @xmath139 ( black solid lines ) and @xmath140 ( red dash lines ) the average phases change very little with a change of @xmath141 , therefore in ( c ) and ( d ) , we show the results for @xmath140 only . in ( d )
red dash line is related to the average over the wave - functions with the winding number @xmath75 and red dotted line to the average over the wave - functions with @xmath76 both kind of the wave - functions appear in the simulations similarly to the weak interaction case , cf .
[ fweak1](d ) .
the size of the 1d space @xmath77 and @xmath33 . ] in ( [ gss ] ) .
the wave - functions reveal two phase flips . before averaging the wave - functions are shifted so that position of one of the flips coincides with @xmath142 .
similarly as in fig .
[ fstrong1](a)-[fstrong1](b ) : @xmath139 ( black solid lines ) and @xmath140 ( red dash lines ) .
panels ( c)-(d ) show histograms of distances @xmath111 between the positions of the phase flips for @xmath139 ( black solid lines ) and @xmath140 ( red dash lines ) .
left pannels are related to @xmath143 , right panels to @xmath33 .
the size of the 1d space @xmath77 . ] particles obtained in many measurement realizations for different moments of time ( indicated in individual panels ) after the initial measurement of @xmath79 particles for @xmath139 ( a ) and @xmath144 ( b ) .
initially , @xmath2-particle system is prepared in the one - hole excited eigenstate with the total momentum per particle @xmath73 . in the weak interaction limit
such a many - body eigenstate is related to the dark soliton that propagates with the velocity @xmath80 , cf .
[ fweak_dark_dyn ] . for @xmath139 , the smearing of the density notch , visible initially in the histogram at @xmath145
[ see the first pannel in ( a ) ] , occurs on the time scale @xmath146 that is similar to the period of the dark soliton motion in the mean - field description . for @xmath144 ,
five minima are clearly visible in the histogram at @xmath44 [ see the first panel in ( b ) ] .
these minima correspond to space points where the first @xmath147 particles have been measured . at these points
it is not possible to observe the remaining 3 particles if repulsion between particles is strong .
analysis of the last particle wave - functions shows that the phase flip by @xmath148 is still present and localizes at different positions ( usually at @xmath149 ) in different measurement realizations . in the course of time evolution
, the five minima are blurred very rapidly but the phase flip of the last particle wave - functions is preserved ( see fig .
[ fstrong4 ] ) even at @xmath150 where @xmath151 is a typical lifetime of a generic quantum state , see the text .
the size of the 1d space @xmath77 and @xmath33.,title="fig : " ] particles obtained in many measurement realizations for different moments of time ( indicated in individual panels ) after the initial measurement of @xmath79 particles for @xmath139 ( a ) and @xmath144 ( b ) .
initially , @xmath2-particle system is prepared in the one - hole excited eigenstate with the total momentum per particle @xmath73 . in the weak interaction limit
such a many - body eigenstate is related to the dark soliton that propagates with the velocity @xmath80 , cf .
[ fweak_dark_dyn ] . for @xmath139 , the smearing of the density notch , visible initially in the histogram at @xmath145
[ see the first pannel in ( a ) ] , occurs on the time scale @xmath146 that is similar to the period of the dark soliton motion in the mean - field description . for @xmath144 ,
five minima are clearly visible in the histogram at @xmath44 [ see the first panel in ( b ) ] .
these minima correspond to space points where the first @xmath147 particles have been measured . at these points
it is not possible to observe the remaining 3 particles if repulsion between particles is strong .
analysis of the last particle wave - functions shows that the phase flip by @xmath148 is still present and localizes at different positions ( usually at @xmath149 ) in different measurement realizations . in the course of time evolution
, the five minima are blurred very rapidly but the phase flip of the last particle wave - functions is preserved ( see fig .
[ fstrong4 ] ) even at @xmath150 where @xmath151 is a typical lifetime of a generic quantum state , see the text .
the size of the 1d space @xmath77 and @xmath33.,title="fig : " ] ( black solid lines ) and @xmath76 ( red dash lines ) are averaged independently .
the results correspond to one - hole excited eigenstates analyzed in fig .
[ fstrong3 ] .
panel ( a ) is related to @xmath139 and @xmath152 [ cf .
[ fstrong3](a ) ] while ( b ) to @xmath144 and @xmath153 [ cf . fig .
[ fstrong3](b ) ] .
the size of the 1d space @xmath77 and @xmath33 . ] it is not easy to answer the question if signatures of soliton - like behavior can be observed in the strong interaction limit because the mean - field approach is not valid . in ref .
@xcite it is shown that the phase imprinting , which is a standard method for dark soliton excitation in experiments , is not able to create dark solitons in ultra - cold atoms in a periodic lattice potential if the system is in the mott insulator regime .
this is due to the lack of phase coherence . on the other hand , analysis of the lieb - liniger model
indicates that the type ii excitation branch is present regardless how strong the particle interactions are @xcite .
moreover , superpositions of the one - hole excited eigenstates allows one to create states that reveal density notches in the reduced single particle density and to observe the phase behavior of the matrix elements of the field operator similar as in the weak interaction case @xcite . in the following
we attempt to answer a question if signatures of soliton - like character can emerge in the course of the particle measurements in the system prepared in a one- and two - hole excited eigenstate for @xmath10 .
let us begin with one - hole excitations . in fig .
[ fstrong1 ] the average probability density and average phase of the last particle wave - functions , obtained in simulation of the measurement process , are presented for two different one - hole excited eigenstates and for different @xmath141 .
the first eigenstate is defined by the exchange @xmath137 in ( [ gss ] ) and corresponds to the dark soliton in the weak interaction limit shown in fig .
[ fweak1](a)-1(b ) . the other , where @xmath138 , is similar to the gray soliton analyzed in the weak interaction case , cf .
[ fweak1](c)-1(f ) . in each realization
we identify a position of the phase flip and we shift all the wave - functions so that the positions always coincide with @xmath104 .
the average phase depends very weakly on @xmath141 and reproduces the behavior observed in the weak interaction limit .
the average probability density changes initially with an increase of @xmath141 but it freezes when @xmath141 becomes of the order of 100 . for strong repulsive interactions ,
particles behave like impenetrable bosons and tend to localize with separations @xmath154 . the width of the density notches and the period of the density oscillations that are visible in fig .
[ fstrong1](a)-(b ) for @xmath140 , coincide with such a length scale .
the oscillations dyes out with an increasing distance from a notch . for @xmath155 ,
the eigenstates ( [ ba ] ) of the lieb - liniger model can be approximated by _
\{k}_n(x_1, ,x_n)_n < msgn(x_m - x_n)det , [ bac ] where @xmath156 for odd @xmath2 @xcite .
the corresponding probability densities are identical to the probability densities for eigenstates of non - interacting fermions , i.e. @xmath157\right|^2 $ ] .
such a fermionisation is observed in the probability density in the configuration space .
note that the wave - functions ( [ bac ] ) remember still the bose statistics , i.e. they are symmetric under an exchange of any pair of particles .
results related to the two - hole excitation where @xmath158 in ( [ gss ] ) , are presented in fig .
[ fstrong2 ] for two different even values of @xmath2 .
we shift the last particle wave - functions so that the position of one of the two observed phase flips coincides with @xmath142 . in different realizations of the particle measurements ,
the two flips appear at different distances one from each other in contrast to the corresponding weak interaction case .
histograms of relative distances between the phase flips are depicted fig .
[ fstrong2 ] .
if @xmath155 , the size of the notches is of the order of @xmath154 and the smallest clear separation between the phase flips takes place when only one particle localizes between them .
such cases correspond to the first maximum in the histograms located at @xmath111 slightly greater than @xmath154 .
the next maxima appear at @xmath159 . for @xmath139 ,
the size of the notches is greater than @xmath154 and we can see the remnant of the second notch in the plots of the average densities .
analysis of the dynamics is performed in a way similar to the weak interaction case .
for @xmath33 we start with the one - hole excited eigenstate corresponding to @xmath73 [ cf .
[ fstrong1](a ) and fig .
[ fstrong1](c ) ] , measure @xmath160 particles , evolve the remaining @xmath97 particles reduced state in time and sample the final probability distribution of positions .
the obtained histograms , for different moments of time and for @xmath139 are shown in fig .
[ fstrong3](a ) .
the density notch , initially visible in the histogram , disappears rapidly , i.e. when @xmath42 is of the order of @xmath146 .
the situation is more complicated when @xmath10 . for @xmath144 , time evolution of a wave - function @xmath161 of three bosons
can be easily obtained by observing that ( x_1,x_2,x_3)_j > s=1
^ 3 sgn(x_j - x_s ) , evolves like a wavefunction of three non - interacting fermions , see ( [ bac ] ) .
the latter evolution can be performed by switching to the momentum space with the help of the fourier transform . in fig .
[ fstrong3](b ) we show the results for @xmath144 . even at @xmath44 ,
the histogram does not show a clear signature of a soliton - like notch .
there are 5 deep minima that correspond to the positions in space where the first @xmath160 particles have been detected at these points it is not possible to find remaining 3 particles because of the strong particles repulsion .
the phase flip of the last particle wave - functions [ very similar like in fig .
[ fstrong1](c ) ] localizes randomly at different positions ( usually at @xmath149 ) . in the course of time evolution the histograms change very quickly , however , the phase flip of the last particle wave - functions can be clearly observed even for @xmath162 , see fig .
[ fstrong4 ] . the typical lifetime of a generic quantum state @xmath163 , where @xmath164 is the averaged energy of the @xmath97 particle system with respect to its ground state energy , is related to the so - called speed limit time @xcite . in fig .
[ fstrong4](b ) we show average phases related to the winding number @xmath75 and @xmath76 .
at this moment of time there are also contributions ( @xmath165 ) related to different winding numbers for which position of the phase flip is difficult to identfy .
the presence of the phase flip is a remnant of the one - hole excitation and a signature that we do not deal with the bose system in the ground state .
we have considered relation of one- and two - hole excited eigenstates ( the so - called type ii eigenstates ) of the lieb - liniger model to dark soliton - like behavior .
the eigenstates possess translational symmetry and are not able to reveal solitons .
however , the symmetry can be broken in a measurement of positions of particles and the solitons can emerge . in the weak interaction limit , successive measurements of positions of particles result in a formation of solitons by the remaining particles .
it is clearly visible in wave - functions of the last particle , i.e. the wave - functions of the system when all but one particles have been measured .
we have also analyzed dynamics of the system prepared initially in a one- or two - hole excited eigenstate .
it turns out that when the translational symmetry is broken in a measurement of @xmath166 particles , the remaining three particles follow the corresponding mean - field soliton dynamics for time @xmath42 smaller than @xmath98 . at @xmath81 many - body effects , that are beyond the mean - field description ,
emerge and lead to smearing soliton notches due to a large uncertainty of soliton positions . in the strong interaction regime ,
i.e. when @xmath155 , the bose system undergoes fermionisation .
that is , the particle probability densities become identical to the probabilities of non - interacting fermions .
then , bosons tend to localize with separation @xmath154 .
flips of the phase of the last particle wave - functions occur also on the length scale of the order of @xmath154 .
these phase flips are preserved for a short period of time despite substantial distortion of the probability densities . in the case of two - hole
excited eigenstates , the distance between two phase flips is different in different realizations of the measurement process . even in the weak interaction limit
, the initial many - body eigenstates do not correspond to a bose - einstein condensate .
however , measurement of positions of particles drives the remaining particles to a state which is the closer to a bec the more particles were measured .
it means that the state of a remaining subsystem approaches a product state with each measurement of particle position .
we have considered the cases where in the course of the measurements , the system approaches a bec with a condensate wave - function corresponding to a single- or multiple - soliton solution of the gross - pitaevskii equation .
it turns out that such a phenomenon can be observed for a number of particles as small as @xmath33 .
numerical simulations , that we perform , are not attainable for a large number of particles .
however , we expect the same behavior also for large @xmath2 .
especially , if @xmath92 but @xmath167 , the many - body dynamics is expected to follow the mean - field prediction for longer and longer time .
that is , the time period when many - body effects turn up should increase like @xmath98 .
dealing with a small @xmath2 could be an advantage if one wants to observe many - body effects in soliton dynamics because the time scale needed for these effects to emerge is shorter .
we have considered the emergence of dark solitons from translationally symmetric eigenstates that exist due to the assumed periodic boundary conditions . such a system can be realized if atoms are loaded to a toroidal trap and prepare in proper excited eigenstates .
this is not an easy experiment and its realization requires further research .
we are gratefull to krzysztof giergiel for the help in numerical programming .
the work was performed within the focus action of faculty of physics , astronomy and applied computer science of jagiellonian university .
support of eu horizon-2020 quic 641122 fet program is also acknowledged .
moreover , a.s . acknowledges support in the form of a special scholarship from the marian smoluchowski scientific consortium `` matter - energy - future '' , from the so - called know funding .
j. denschlag , j. e. simsarian , d. l. feder , charles w. clark , l. a. collins , j. cubizolles , l. deng , e. w. hagley , k. helmerson , w. p. reinhardt , s. l. rolston , b. i. schneider and w. d. phillips , science * 287 * , 97 ( 2000 ) . | eigenstates of bose particles with repulsive contact interactions in one - dimensional space with periodic boundary conditions can be found with the help of the bethe ansatz .
the type ii excitation spectrum identified by e. h. lieb , reproduces the dispersion relation of dark solitons in the mean - field approach .
the corresponding eigenstates possess translational symmetry which can be broken in measurements of positions of particles .
we analyze emergence of single and double solitons in the course of the measurements and investigate dynamics of the system . in the weak interaction limit , the system follows the mean - field prediction for a short period of time .
long time evolution reveals many - body effects that are related to an increasing uncertainty of soliton positions . in the strong interaction regime particles
behave like impenetrable bosons .
then , the probability densities in the configuration space become identical to the probabilities of non - interacting fermions but the wave - functions themselves remember the original bose statistics .
especially , the phase flips that are key signatures of the solitons in the weak interaction limit , can be observed in the time evolution of the strongly interacting bosons . |
the recent market turbulence caused by the credit crunch has exposed in a drastic way the consequences of overconfidence in financial modelling assumptions .
typically , a financial model , such as the famous black - scholes model , will assume that the price of an asset follows a given stochastic process whose parameters need to be calibrated to market prices .
if a model becomes an accepted standard and most market participants adopt it , problems can occur when assumptions that hold under normal market conditions are also expected to hold under abnormal ones .
an example is the stock market crash of @xmath2 , where the volatilities used for pricing at - the - money options were also used for pricing far out - of - the - money put options . as the market headed downwards , it turned out that the true hedging cost for somebody who had sold such puts was far greater than the received premium .
another good example is described in the recent paper @xcite , where the authors demonstrate for cdos and cdo@xmath3s what can happen to asset prices when model parameters that are hard to observe or estimate with sufficient accuracy are put to a true stress test .
however , they write : `` the good news is that this mistake can be fixed .
for example , a bayesian approach that explicitly acknowledges that parameters are uncertain would go a long way towards solving this problem . ''
@xcite another well - established way to obtain estimates for such parameters from observable data , which we will follow here , is via maximum entropy methods ( @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite ) .
such an estimate `` is the least biased estimate possible on the given information , i.e. , it is maximally noncommittal with regard to missing information . ''
@xcite for example , the probability distribution over the interval @xmath4 $ ] which maximises entropy is the uniform distribution .
there is no entropy maximiser for distributions over @xmath5 .
however , when the mean and variance are specified , the gauss distribution with these parameters maximises entropy .
we concentrate on the distribution of an asset price at a given time in the future , for which there are some option data .
we develop a highly robust technique to find a maximum entropy distribution ( med ) for the asset in case we have call and digital option prices .
the density is obtained by partitioning the range of possible stock prices into buckets , i.e. the intervals between adjacent strikes given by the option data , but , in contrast to the black - scholes model , making no a priori assumption about the asset s distribution .
instead , we maximise the boltzmann - shannon entropy to obtain a distribution that respects only the given option prices and is otherwise unbiased .
the density can in turn be used to interpolate implied volatilities and , by repeating this operation for a range of maturities , obtain a volatility surface .
the results agree surprisingly well with observed volatility surfaces from the markets .
buchen and kelly ( @xcite ) have proposed a similar entropy maximisation method to infer the probability distribution for an asset from call prices .
this maximisation problem corresponds to finding a set of lagrange multipliers . in @xcite
the authors write : `` there is a problem with this type of calculation , '' meaning that the formal lagrange multipliers approach is not mathematically rigorous . using convex programming arguments they legitimise those calculations .
like @xcite we legitimise the results found in @xcite .
however , we follow a simpler approach by applying a result of csiszr s @xcite . both @xcite and @xcite present numerical methods to find the lagrange multipliers by solving an @xmath6-dimensional non - linear problem ( where @xmath6 is the number of constraints given by call prices ) . as mentioned in @xcite , in the case of close strikes
the problem can be poorly conditioned . in the present work ,
we add @xmath6 digital prices to our constraints and the resulting numerical problem is highly simplified : instead of an @xmath6-dimensional equation , we need to solve a one - dimensional problem @xmath7 for @xmath6 different values of @xmath8 ( the same @xmath9 though ) , allowing for easy parallelisation and avoiding any ill - conditioned problem . additionally , @xmath9 is a strictly monotonic function whose derivative is known analytically . therefore , the newton - raphson method can be used for excellent speed of convergence and stability . alternatively
to iterative methods , one may try to find an analytical approximation for @xmath10 . in a nutshell ,
the advantage we obtain is the localisation of the maximum entropy density into asset price buckets , in which the functional form is a simple exponential function .
but of course there is a price to pay for this localisation technique , and the price here is the necessity of an additional constraint for each call option used .
this extra constraint is the price of a digital option at the same strike .
the density in our case differs slightly from the one given by the method in @xcite .
we therefore investigate the differences between them . in both approaches
one can also use information from a so - called prior density , if available , leading to the concept of relative entropy ( also called @xmath1-divergence and kullback - leibler information number ) , and we compare the densities obtained by this method . after finding the med
we give the expressions for the cumulative distribution function and its inverse .
these formulas involve only arithmetic operations and exponential- and logarithm - functions .
they are therefore very easy to implement and fast to compute .
this is a highly useful feature for fast monte carlo simulations .
furthermore , we obtain an analytical formula for the price of a call at a given strike . by calculating several such prices at different strikes , one can recover the implied volatility smile .
we also include a section in which we calibrate to real market data .
digital options on the s&p 500 index ( spx ) and the cboe volatility index ( vix ) are traded on the chicago board option exchange ( cboe ) , where they are called _
binary options_. they are specified such that `` expiration dates and settlement values are the same as for traditional options '' @xcite , which is just what we need in our setup . we show results for two cases : the first , in which we have cboe quotes for call _ and _ digital options on the spx and calibrate to them , and the second , for a different maturity , in which we only have quotes for call options and therefore have to estimate digital prices from call spreads .
the method we propose here is found to work very well in both cases , and we compare our results to those obtained by calibrating only to call prices as in @xcite .
we are given a fixed maturity @xmath11 , strictly increasing strikes @xmath12 @xmath13 , and undiscounted call and digital prices @xmath14 at these strikes .
the payoffs of the call and digital options are given in equations and .
@xmath15 denotes the discount factor . throughout
we make the convention @xmath16 assuming risk neutral pricing , we will determine a density @xmath17 for the underlying asset price @xmath18 which maximises entropy @xmath19 under the constraints @xmath20 } = \tilde{c}_i , \quad i.e. \quad \int_{k_i}^\infty ( x - k_i ) g(x ) dx = \tilde{c}_i\ ] ] and @xmath21 } = \tilde{d}_i , \quad i.e. \quad \int_{k_i}^\infty g(x ) dx = \tilde{d}_i\ ] ] for all @xmath22 .
in particular , these two constraints for @xmath23 mean that @xmath17 is a density , since @xmath24 , and that the martingale condition @xmath25 } = \int_0^\infty x g(x ) dx = \tilde{c}_0\ ] ] is satisfied , since @xmath26 is the forward price of @xmath27 for time @xmath11 . from the second constraint
it immediately follows that @xmath28 looking at a call spread with strikes @xmath29 raised to level @xmath30 , i.e. a derivative that pays @xmath18 if @xmath31 and zero otherwise , we obtain the condition @xmath32 we now calculate the density @xmath17 under the constraints given above . the purpose of theorem [ localglobalmaximiser ] is to show that the local constraints and are equivalent to the global constraints and .
moreover , @xmath33 and , thus , we only need to maximise @xmath34 subject to and over each bucket .
let @xmath35 be the set of positive borel - measurable functions defined on @xmath36 define @xmath37 and , for all @xmath22 , @xmath38 [ intersectionx ] @xmath39 it is straightforward to show this using , , and .
@xmath40 for @xmath41 we define @xmath42 [ localglobalmaximiser ] if @xmath17 is a maximiser of @xmath43 on @xmath44 , then @xmath17 is a maximiser of @xmath45 on @xmath46 .
conversely , if @xmath17 is a maximiser of @xmath45 on @xmath46 for all @xmath47 then @xmath17 is a maximiser of @xmath43 on @xmath44 .
let @xmath17 be a maximiser of @xmath43 on @xmath44 , and let @xmath48 .
define @xmath49 since @xmath50 on @xmath51 , we have @xmath52 .
moreover , for @xmath53 , we have @xmath54 on @xmath55 , and thus @xmath56 .
it follows from proposition [ intersectionx ] that @xmath57 .
hence , from the maximality of @xmath58 , we get @xmath59 .
a simple computation gives @xmath60 , and therefore @xmath61 .
it follows that @xmath17 maximises @xmath45 on @xmath46 .
conversely , suppose that @xmath17 is a maximiser of @xmath45 on @xmath46 for all @xmath62 let @xmath63 .
we have @xmath64 which means that @xmath17 is a maximiser of @xmath43 on @xmath44 .
@xmath40 we now give a heuristic way of determining the entropy maximiser , but in the next subsection we also give a rigorous proof that this is indeed the correct result . formally applying the lagrange multipliers theorem , we conclude that the maximiser has the form @xmath65 to see this , define the functionals @xmath66 and solve the equation @xmath67 for the frchet derivatives .
it follows that @xmath68 therefore , on the interval @xmath51 , we must have @xmath69 and , introducing @xmath70 and @xmath71 , we obtain .
using the explicit form of @xmath17 just found in and gives @xmath72 for all @xmath22 .
for @xmath73 , solving for @xmath74 and then for @xmath75 , using integration by parts , gives @xmath76 define @xmath77 it follows that @xmath78 and @xmath79 ( here @xmath80 means derivative with respect to @xmath81 ) . ]
figure [ fig:1 ] shows the graphs of @xmath82 and @xmath83 for @xmath84 and @xmath85 .
it suggests that equation has a unique solution if the quantity on the right hand side is in @xmath86k_i , k_{i+1}[$ ] .
this turns out to be the case , as we show with the following proposition .
let @xmath87 .
if there is no arbitrage opportunity implied by @xmath88 , then there is a unique solution @xmath89 for equations and .
define @xmath90 we first show that we must have @xmath91 .
this can be seen by comparing the prices of three derivatives : they pay , respectively , @xmath30 , @xmath18 and @xmath92 if @xmath31 and zero otherwise . under the assumption that there is no arbitrage opportunity
, it follows immediately that the second derivative is more expensive than the first one and cheaper that the third one .
it is also clear that they can be replicated by portfolios of calls and digitals and their prices , in increasing order , are @xmath93 from the definition of @xmath94 the middle quantity above is @xmath95 and the result follows .
next we show that if @xmath96 , then there is a unique solution @xmath89 for equations and .
we begin with the case @xmath73 . as we have just seen , and are then equivalent to and . without loss of generality , we may assume @xmath97 and @xmath98 .
indeed , it is straightforward to see that the change of variables @xmath99 transforms the equation @xmath100 into @xmath101 , with @xmath1020 , 1[$ ] . using lhpital s rule we obtain that the function @xmath103 given by @xmath104 is a continuous extension of @xmath105 .
it is easy to see that @xmath106 and @xmath107 .
hence the equation @xmath108 has a solution . to prove that the solution is unique
, we shall now show that @xmath9 is strictly increasing .
again by lhpital s rule , we obtain that @xmath9 is differentiable at @xmath109 and @xmath110 ( this is particularly useful because @xmath109 is an ideal starting point for the newton - raphson method ) . for @xmath111
we have @xmath112 recall that @xmath113 for all @xmath114 .
hence @xmath115 we conclude that @xmath116 for all @xmath117
. therefore @xmath9 is strictly increasing .
finally , we consider the case @xmath118 .
equations and then become @xmath119 the first equation implies that @xmath120 . solving it for @xmath121 and the second equation for @xmath122
gives @xmath123 @xmath40 note that we have shown that @xmath9 is itself a continuously differentiable probability distribution function .
for such a function there might already exist an inversion - algorithm . like others
, we have formally derived the expression for the entropy maximiser using the lagrange multipliers method .
however , as pointed out in @xcite `` there is a problem with this type of calculation . ''
recall that the lagrange multipliers theorem requires continuous differentiability for objective and constraint functionals in a neighbourhood of the maximiser .
however , the boltzmann - shannon entropy functional is finite only for densities in @xmath124 which has empty interior on @xmath125 . therefore a maximiser is not an interior point of @xmath126 .
even worse , the entropy is far from being continuously differentiable since it is nowhere continuous . in @xcite ,
convex programming arguments are considered to circumvent this problem . here
we present a new approach based on a result by csiszr @xcite .
when no prior density is given we are interested in the ( non - relative ) entropy of @xmath17 @xmath127 where @xmath128 is an interval . however
, csiszr s results deal with relative entropy of a probability density @xmath17 with respect to a probability measure @xmath129 on @xmath1 @xmath130 roughly speaking , we are interested in the `` relative entropy '' with respect to the lebesgue measure which , in general , is not a probability measure . for @xmath131 , @xmath132 is bounded . in that case
, the problem can easily fit in csiszr s framework by considering the normalised lebesgue probability measure @xmath133 .
however , it is impossible to use this trick for the global problem , @xmath134 , and for the last bucket , @xmath135 , since there is no normalisation constant which turns the lebesgue measure into a probability measure on these intervals .
nevertheless , it is possible to turn the two problems over unbounded intervals into equivalent ones that do fit in csiszr s framework .
this is the subject of the next proposition .
moreover , the same arguments also apply to bounded intervals . therefore ,
in contrast to @xcite , we do not need to make any distinction between bounded and unbounded intervals . for the sake of simplicity
, the statement in the following proposition considers only two _ main _ constraints , namely , the total mass and the mean .
this includes the bucket problems and excludes the global problem ( where additional constraints are given ) .
however , the proof works even for an infinite number of constraints provided the two main ones are among them .
[ bijection ] let @xmath136 be an interval .
define @xmath137 for all @xmath138 , where @xmath139 is a normalisation constant such that @xmath140 is a probability measure on @xmath1 .
let @xmath141 .
then the mapping @xmath142 is a bijection from @xmath143 onto @xmath144 moreover , @xmath17 is a maximiser of @xmath145 on @xmath146 if and only if @xmath147 is a maximiser of @xmath148 on @xmath149 .
define @xmath150 by @xmath151 .
since @xmath152 is strictly positive , it follows immediately that @xmath153 is a well defined bijection .
we shall show that @xmath153 preserves some linear functionals .
let @xmath154 and @xmath155 .
then we have @xmath156 in particular , applying this result to @xmath157 and to @xmath158 , it follows immediately that @xmath153 maps @xmath146 onto @xmath149 . to complete the proof it suffices to show that if @xmath159 , then @xmath160 in fact , we shall show a stronger result , namely , that the two differences above are equal
this is equivalent to showing that @xmath161 does not depend on @xmath162 .
we have @xmath163 @xmath40 later we will restate and apply a partial version of a theorem by cziszr .
but before we do so , let us say a few words about it
. it is very natural to apply the lagrange multipliers theorem for maximisation problems under constraints .
however , there are many cases where other techniques are used - for instance in the proof of the existence of projection on a convex set of a hilbert space . in that case ,
geometric arguments , including the parallelogram identity , are used .
many texts suggest thinking of the relative entropy of one probability measure with respect to another as a quantity measuring how much they differ .
moreover , they present some similarities between relative entropy and a metric . unfortunately , they say , this analogy does not go too far .
csiszr s paper pushes these similarities a bit further , showing a relation analogous to the parallelogram identity .
furthermore , he proves the existence of an entropy minimiser under convex constraints by similar arguments that show the existence of projection on convex subsets of hilbert spaces .
we restate here a partial version of his theorem 3.1 sufficient for our purposes .
[ csiszr ] let @xmath129 be a probability on a measurable space @xmath164 .
let @xmath165 be an arbitrary set of real - valued @xmath166-measurable functions on @xmath167 and @xmath168 be real constants .
let @xmath169 be the set of all those probabilities @xmath170 on @xmath171 for which the integrals @xmath172 exist and equal @xmath173 @xmath174 . then
, if there exists @xmath175 such that @xmath176 and its radon - nikodym derivative has the form @xmath177 where @xmath178 and @xmath179 belongs to the linear space spanned by the @xmath180 s , then @xmath181 for all @xmath182 such that @xmath183 .
now we prove that @xmath17 , given by , and , is indeed an entropy maximiser .
let @xmath184 , @xmath132 .
let @xmath74 and @xmath75 be defined by equations and .
then @xmath185 given by @xmath186 maximises @xmath45 on @xmath46 .
set @xmath187 and @xmath188 .
let @xmath152 , @xmath129 , @xmath146 and @xmath149 be as in proposition [ bijection ] .
note that for this choice of @xmath1 , @xmath189 and @xmath190 , we have @xmath191 and @xmath192 .
let @xmath193 , @xmath194 be the @xmath195-algebra of lebesgue measurable subsets of @xmath1 , @xmath196 , @xmath197 ( @xmath198 ) and @xmath169 as in csiszr s theorem .
given @xmath199 , define the measure @xmath200 by @xmath201 . from the definition of @xmath149
, it follows that @xmath202 .
conversely , if @xmath182 and @xmath183 , then @xmath203 . then a simple computation yields @xmath204 by definition of @xmath74 and @xmath75 we have @xmath205 .
proposition [ bijection ] yields @xmath206 .
moreover , @xmath207 for all @xmath138 .
let @xmath208 .
it follows that @xmath175 , and its radon - nikodym derivative with respect to @xmath129 ( which is @xmath209 ) has the form with @xmath210 and @xmath211 .
therefore , csiszr s theorem gives @xmath181 for all @xmath182 such that @xmath183 .
in particular , for all @xmath199 , from we obtain @xmath212 we conclude that @xmath213 maximises @xmath148 on @xmath149 and , again by proposition [ bijection ] , that @xmath17 is a maximiser of @xmath45 on @xmath46 . @xmath40 we have the explicit form of the density given by equation .
this allows us to give formulas in several important cases . to do this ,
we first state two useful results for the following proofs . for @xmath214 , we have @xmath215 \nonumber \\ & = & \frac{\alpha_i}{\beta_i}(ke^{\beta_ik } - k_ie^{\beta_ik_i } ) - \frac{\alpha_i}{\beta_i^2}(e^{\beta_ik } - e^{\beta_ik_i } ) .
\label{k_i - k - mean}\end{aligned}\ ] ] it is straightforward to integrate the density @xmath17 and obtain an explicit form of the probability distribution @xmath216 its inverse can also be expressed analytically , which is a useful feature for monte carlo simulations .
the results are stated in the following proposition .
[ distributionfunction ] suppose @xmath217 .
then @xmath218 given @xmath219 , find @xmath220 such that @xmath221\tilde d_{i+1 } , \tilde d_i]$ ] .
then @xmath222 we treat only the case @xmath223 .
the simpler case @xmath224 is left to the reader .
first , notice that @xmath225
. then , using , we get @xmath226 since @xmath227 , we have @xmath228 . therefore solving for @xmath229
concludes the proof . @xmath40
it is also straightforward to express the prices of call and digital options analytically .
[ prices ] given a strike @xmath230 , find @xmath231 such that @xmath214 .
if @xmath223 , then @xmath232 if @xmath224 , then @xmath233 again we prove only the case @xmath223 . from we obtain @xmath234 for the ( undiscounted ) call price we have @xmath235 now putting and into leads to the stated result .
@xmath40 finally , using euler s relationship for homogeneous functions , we can also give an explicit formula for spot - delta .
given a strike @xmath230 , find @xmath231 such that @xmath214 .
let @xmath27 be today s underlying spot price and @xmath236 be the spot - delta of a call with strike @xmath237 maturing at @xmath11 . if @xmath223 , then @xmath238 if @xmath224 , then @xmath239 again we consider only the case @xmath223 and leave the simpler case @xmath224 for the reader .
let @xmath240 and @xmath241 be , respectively , the discounted prices of call and digital options with strike @xmath237 maturing at @xmath11 , i.e. @xmath242 and @xmath243 , where @xmath244 and @xmath245 are as in proposition [ prices ] .
since @xmath240 is a positively homogeneous function of degree @xmath246 in @xmath247 , from euler s theorem we have @xmath248 recalling that @xmath249 , we can rewrite the last relation as @xmath250 now using and gives the result .
@xmath40 note that the analogous statement for the forward - delta can be obtained by replacing the spot - price with the forward - price in the corollary and proof above .
if we hold a prior belief about the distribution , we can maximise relative entropy instead in order to stay as `` close '' as possible to the prior distribution .
suppose @xmath251 is a probability density for this prior distribution . for @xmath252
, define relative entropy @xmath253 ( the kullback - leibler information number or @xmath1-divergence is given by @xmath254 .
this can be thought of as a measure of distance between two distributions .
for example , @xmath255 , and @xmath256 if and only if @xmath257 .
) we have @xmath258 and essentially the same argument as the one given above shows that the maximum relative entropy density ( mred ) @xmath259 is given by @xmath260 therefore the resulting density @xmath261 is now given by the product of a piecewise exponential density and the prior density . even in the simple case where the prior density @xmath262 is just log - normal
, we no longer have explicit formulas for call and digital prices .
since we can not separate the two constraints @xmath263 for each @xmath47 as in equations and , we must solve them simultaneously using numerical integration and a two - dimensional root - finder .
however , if the prior density @xmath262 is already given by an med , then @xmath264 and we can solve everything analytically as before .
we also recover explicit formulas for call and digital prices .
buchen and kelly @xcite propose a similar method to find an entropy - maximising density @xmath265 under constraints given by european payoffs .
the case of most interest is where these are the payoff - functions of call options at different strikes @xmath266 and the actual constraints are given by ( undiscounted ) call option prices @xmath267 such that @xmath268 } = \tilde{c}_i\ ] ] must hold for all @xmath269 .
the density @xmath265 must therefore satisfy the conditions @xmath270 and @xmath271 to find @xmath265 , they construct the functional @xmath272 where @xmath273 are the lagrange multipliers , and then solve the equation @xmath274 the solution is given by @xmath275 where @xmath276 is a normalising constant .
buchen and kelly show that numerically , finding the parameters @xmath277 is an m - dimensional root - finding problem that can be tackled with the multi - dimensional newton algorithm .
they show how to compute the jacobian , and that it is invertible , by expressing it as a covariance matrix .
if a call option with strike @xmath278 , i.e. the forward , is among the input data , the mean of the distribution is given .
since the total mass , @xmath246 , is also known , we have the two main constraints needed to apply the arguments from subsection [ rigorousproof ] and can therefore also rigorously find the entropy maximiser when only call options are given as input .
of course , the forward should be known in most situations , so that this is certainly the most important case . similarly ,
if a prior distribution @xmath262 is given , the distribution maximising relative entropy under the same constraints is given by @xmath279 where @xmath280 is again the normalising constant .
in this section we give some numerical examples for the entropy maximisers described so far .
we suppose that the market data is given by @xmath281 we assume a flat volatility and make no skew correction when calculating the digital prices in this scenario .
we calculate three densities using strikes * @xmath282 * @xmath283 * @xmath284 lllllllllll + strike & 0.00 & 20.00 & 40.00 & 60.00 & 80.00 & 100.00 & 120.00 & 140.00 & 160.00 & 180.00 + call & 100.0000 & 80.0000 & 60.0005 & 40.1454 & 22.2656 & 9.9477 & 3.7059 & 1.2139 & 0.3659 & 0.1049 + digital & 1.0000 & 1.0000 & 0.9998 & 0.9725 & 0.7786 & 0.4503 & 0.1965 & 0.0707 & 0.0225 & 0.0066 + + entropy & 4.6714 + @xmath285 & 1.3582e-04 & n / a & n / a & n / a & n / a & 1.8835 & n / a & n / a & n / a & n / a + @xmath81 & 0.0539 & n / a & n / a & n / a & n / a & -0.0453 & n / a & n / a & n / a & n / a + call & 100.0000 & 80.0402 & 60.2562 & 40.9886 & 23.2384 & 9.9477 & 4.0232 & 1.6271 & 0.6581 & 0.2661 + implied vol . &
n / a & 62.13% & 46.26% & 36.17% & 28.88% & 25.00% & 25.95% & 27.04% & 27.84% & 28.41% + digital & 1.0000 & 0.9951 & 0.9808 & 0.9386 & 0.8146 & 0.4503 & 0.1821 & 0.0736 & 0.0298 & 0.0120 + + entropy & 4.6143 + @xmath285 & 6.0682e-08 & n / a & n / a & 0.0016 & n / a & 0.5397 & n / a & 14.2333 & n / a & n / a + @xmath81 & 0.1894 & n / a & n / a & 0.0255 & n / a & -0.0343 & n / a & -0.0582 & n / a & n / a + call & 100.0000 & 80.0001 & 60.0033 & 40.1454 & 22.4905 & 9.9477 & 3.7539 & 1.2139 & 0.3790 & 0.1183 + implied vol . & n / a & 38.76% & 28.60% & 25.00% & 25.93% & 25.00% & 25.14% & 25.00% & 25.15% & 25.38% + digital & 1.0000 & 1.0000 & 0.9994 & 0.9725 & 0.7765 & 0.4503 & 0.1978 & 0.0707 & 0.0221 & 0.0069 + + entropy & 4.6076 + @xmath285 & 6.0682e-08 & n / a & n / a & 1.5393e-04 & 0.0129 & 0.2389 & 1.6987 & 14.2333 &
n / a & n / a + @xmath81 & 0.1894 & n / a & n / a & 0.0584 & 0.0027 & -0.0268 & -0.0433 & -0.0582 & n / a & n / a + call & 100.0000 & 80.0001 & 60.0033 & 40.1454 & 22.2656 & 9.9477 & 3.7059 & 1.2139 & 0.3790 & 0.1183 + implied vol . &
n / a & 38.76% & 28.60% & 25.00% & 25.00% & 25.00% & 25.00% & 25.00% & 25.15% & 25.38% + digital & 1.0000 & 1.0000 & 0.9994 & 0.9725 & 0.7765 & 0.4503 & 0.1978 & 0.0707 & 0.0221 & 0.0069 + table [ tab:1 ] gives the ( undiscounted ) option prices we used and the parameters describing the density . and
three maximum entropy densities obtained by calibrating to @xmath246 , @xmath286 and @xmath287 strikes . ]
figure [ fig:2 ] shows the three densities and the actual log - normal density .
it can be seen that already with @xmath287 strikes and the forward , the fit of the piecewise - exponential distribution to the log - normal distribution is very good . in practice , however , implied volatilities are not flat as in the example above , i.e. they are not the same for different strikes and maturities .
this is discussed in detail in gatheral s book @xcite , in particular in the section `` the spx implied volatility surface '' in chapter 3 . we show that our method has by its nature a tendency to give good fits to observed volatility surfaces . to do this
, we will assume that we now only have at - the - money ( atm ) option prices , and that already with this minimal amount of market data our method generates a very realistic looking volatility surface .
we show in figure [ fig:3 ] the implied volatility surface obtained by using just the atm strike .
more precisely , if we assume @xmath288 and @xmath289 again , then the atm strike is @xmath290 .
moreover , we consider a constant atm volatility @xmath291 ( it could of course be time dependent ) . for each maturity @xmath292
, we compute the black - scholes prices of the atm call @xmath293 and atm digital @xmath294 . then applying the maximum entropy approach to just one strike @xmath295 , we compute @xmath74 and @xmath75 for @xmath296 . in other terms
, we recover the maximum entropy density of @xmath18 compatible with @xmath297 , @xmath293 and @xmath294 .
we emphasise that no other strike , call or digital is used in this calibration .
according to the med , the price of a call with strike @xmath237 and maturity @xmath11 is given by @xmath298 , where @xmath299 is given in proposition [ prices ] .
from @xmath300 we recover the implied volatility @xmath301 with a bisection root - finder from the black - scholes formula .
readers interested in a more robust method can consult the one proposed in @xcite . as expected , as a consequence of calibration , @xmath302 .
what is surprising is the fact that the curve @xmath303 has a profile very similar to smile curves typically seen in equity markets .
now , by varying @xmath11 one constructs a volatility surface which , again , is qualitatively very similar to those observed in equity markets .
the maximum entropy method seems to be able to transform just one volatility number from a flat black - scholes ( atm ) world into a very realistic looking volatility surface , with important features such as a strongly pronounced smile at the short end that decays as the maturity increases .
different strikes and different numbers of option prices can of course be used at different maturities , so that any arbitrage - free option data can easily be converted into an implied volatility surface .
the density @xmath17 is usually discontinuous at the @xmath30 s .
the distribution function is of course continuous .
many monte carlo models work by drawing a random uniform variable and inverting the distribution . in black - scholes type models , for example , a normal distribution has to be inverted at some stage . in our case , only one logarithm needs to be taken , a circumstance which accelerates a simulation .
let the prior distribution be a log - normal distribution with fixed volatility parameter @xmath195 @xmath304 we still have an explicit form of the density , namely @xmath261 , where @xmath17 is a piecewise exponential density , although the parameters @xmath305 are of course different from the parameters @xmath306 used for the med of the previous subsection . since we are now unable to express call prices analytically , we calculate them via numerical integration .
lllllll strike & 0.00 & 60.00 & 80.00 & 100.00 & 120.00 & 140.00 + + @xmath307 & 12.2600 & n / a & n / a & 0.0833 & n / a & n / a + @xmath308 & -0.0298 & n / a & n / a & 0.0206 & n / a & n / a + + @xmath307 & 11.2900 & 7.2379 &
n / a & 0.2930 & n / a & 0.3267 + @xmath308 & -0.0194 & -0.0237 & n / a & 0.0098 & n / a & 0.0116 + + @xmath307 & 11.2900 & 5.9910 & 2.0430 & 0.5970 & 0.5815 & 0.3267 + @xmath308 & -0.0194 & -0.0210 & -0.0097 & 0.0031 & 0.0047 & 0.0116 + table [ tab:2 ] gives the parameters describing the density .
of course , should a prior density already meet the constraints , we will have @xmath309 and @xmath310 for all @xmath311 . , actual log - normal density with @xmath312 and three maximum relative entropy densities obtained by calibrating to @xmath246 , @xmath286 and @xmath287 strikes . ]
figure [ fig:4 ] shows the three maximum relative entropy densities and the prior log - normal density ( @xmath313 ) .
the density for the forward and one call is already much closer to the actual one than in the previous case , so that convergence is not as pronounced as before when the number of strikes is increased .
we see that @xmath17 has the effect of pushing the prior density downwards and widening it as to be closer to the actual density .
the explicit form of the density given by equation allows one to obtain analytic expressions for call and digital prices like those in proposition [ prices ] . as an example , using just the forward and an at - the - money call , i.e. @xmath314 , we obtained @xmath315 on our computer .
this leads to a very similar volatility smile as the one given at @xmath316 in subsection [ med - cd ] above .
we refer to @xcite for graphs and numerical data regarding this distribution . as in subsection [ relmed - cd ] ,
in general there will be no analytic expressions for call or digital prices .
if the chosen prior distribution is continuous , then the resulting relative entropy maximiser will also be continuous .
again , we advise the reader to look at @xcite for graphs and numerical data regarding this distribution .
digital options are traded on the chicago board option exchange ( cboe ) .
they are called _ binary options _ there .
we quote the following paragraph from the `` binaries '' product description @xcite : `` cboe offers binary options on the s&p 500 index ( spx ) and the cboe volatility index ( vix ) .
the ticker symbols for these binary contracts is bsz and bvz respectively .
expiration dates and settlement values are the same as for traditional options . ''
the specification that digital option strikes and maturities are the same as those of call options is exactly what we need for our setup .
we calibrate to cboe option prices from 10 april 2010 for two different maturities .
the first maturity is 18 september 2010 .
we have digital option bid and ask quotes for strikes @xmath237 from @xmath317 to @xmath318 usd , usually in steps of @xmath319 usd .
we also have call option bid and ask quotes for these same strikes .
we calibrate to the `` mid '' prices , i.e. the average of the bid and ask quotes , at the ten strikes from @xmath317 to @xmath318 in steps of @xmath320 usd .
rrrrrrr & & & + strike & digital & call & digital & call & digital & call + 950 & 0.9400 & 246.30 & 0.9259 & 246.30 & 0.9400 & 246.30 + 975 & 0.9150 & 223.20 & 0.9171 & 223.25 & 0.9153 & 223.12 + 1000 & 0.8950 & 200.50 & 0.9014 & 200.50 & 0.8950 & 200.50 + 1025 & 0.8750 & 178.15 & 0.8787 & 178.24 & 0.8795 & 178.30 + 1050 & 0.8550 & 156.60 & 0.8516 & 156.60 & 0.8550 & 156.60 + 1075 & 0.8150 & 135.70 & 0.8191 & 135.70 & 0.8195 & 135.65 + 1100 & 0.7750 & 115.70 & 0.7802 & 115.70 & 0.7750 & 115.70 + 1125 & 0.7250 & 96.75 & 0.7336 & 96.76 & 0.7367 & 96.76 + 1150 & 0.6700 & 79.10 & 0.6776 & 79.10 & 0.6700 & 79.10 + 1175 & 0.6050 & 63.00 & 0.6117 & 62.97 & 0.6137 & 63.01 + 1200 & 0.5350 & 48.60 & 0.5357 & 48.60 & 0.5350 & 48.60 + 1225 & 0.4550 & 36.25 & 0.4541 & 36.23 & 0.4585 & 36.13 + 1250 & 0.3550 & 25.90 & 0.3720 & 25.90 & 0.3550 & 25.90 + 1300 & 0.1850 & 11.35 & 0.2112 & 11.35 & 0.1850 &
11.35 + 1350 & 0.0700 & 4.10 & 0.0896 & 4.10 & 0.0700 & 4.10 + 1400 & 0.0450 & 1.33 & 0.0307 & 1.33 & 0.0450 & 1.33 + table [ tab:3 ] shows cboe prices for call and digital options on the spx from 10 april 2010 in columns 2 and 3 .
columns 4 and 5 show option prices obtained by calibrating an med to call prices at strikes @xmath317 , @xmath321 , @xmath322 , @xmath323 , @xmath318 .
columns 6 and 7 show option prices obtained by calibrating an med to call and digital prices at the same strikes .
note that the second med matches market call _ and _ digital prices at the strikes calibrated to exactly , whereas the first med matches only the call prices .
the second maturity is 31 december 2010 .
we have call option bid and ask quotes for strikes @xmath237 from @xmath324 to @xmath325 usd in steps of @xmath320 usd .
we do not have any digital option quotes for this maturity . as a substitute ,
we calculate symmetric call spread prices @xmath326 using mid call prices .
we calibrate to this data at the three strikes @xmath327 . in the previous example
, we showed that our method can be used to calibrate to quotes at many ( @xmath328 ) strikes , and that this leads to a very good fit . in this example , we calibrate to only a small number ( @xmath286 ) of strikes in order to show that our method has an excellent natural tendency to fit a market smile .
the call spread prices @xmath329 needed for are obtained by using the call quotes at @xmath330 . [ tab:4 ] rrrrrrrrr & & & & + & & & & + strike & call & implied vol . & call & implied vol . & call & implied vol . & call & implied vol .
+ 500 & 681.15 & 62.90% & 678.87 & 59.45% & 677.05 & 56.22% & 679.66 & 60.70% + 550 & 631.75 & 57.19% & 630.27 & 55.25% & 628.62 & 52.84% & 630.92 & 56.12% + 600 & 582.35 & 51.94% & 581.72 & 51.22% & 580.41 & 49.63% & 582.18 & 51.75% + 650 & 533.45 & 47.55% & 533.22 & 47.32% & 532.44 & 46.53% & 533.45 & 47.55% + 700 & 484.75 & 43.52% & 484.75 & 43.52% & 484.75 & 43.52% & 484.75 & 43.52% + 750 & 436.55 & 39.99% & 436.54 & 39.98% & 437.36 & 40.59% & 436.55 & 39.99% + 800 & 388.85 & 36.77% & 388.90 & 36.81% & 390.39 & 37.75% & 388.97 & 36.85% + 850 & 341.85 & 33.85% & 342.02 & 33.95% & 343.97 & 35.01% & 342.11 & 34.00% + 900 & 295.95 & 31.28% & 296.11 & 31.36% & 298.32 & 32.39% & 296.20 & 31.40% + 950 & 251.25 & 28.89% & 251.49 & 28.99% & 253.72 & 29.88% & 251.55 & 29.01% + 1000 & 208.25 & 26.70% & 208.54 & 26.80% & 210.56 & 27.50% & 208.54 & 26.80% + 1050 & 167.50 & 24.68% & 167.76 & 24.76% & 169.36 & 25.25% & 167.72 & 24.75% + 1100 & 129.70 & 22.82% & 129.84 & 22.86% & 130.87 & 23.15% & 129.77 & 22.84% + 1150 & 95.60 & 21.09% & 95.64 & 21.10% & 96.06 & 21.21% & 95.60 & 21.09% + 1200 & 66.30 & 19.52% & 66.30 & 19.52% & 66.30 & 19.52% & 66.30 & 19.52% + 1250 & 42.70 & 18.13% & 43.09 & 18.24% & 42.93 & 18.20% & 42.70 & 18.13% + 1300 & 25.10 & 16.91% & 25.83 & 17.13% &
25.61 & 17.06% & 25.19 & 16.93% + 1350 & 13.35 & 15.87% & 13.79 & 16.05% & 13.63 & 15.99% & 13.35 & 15.87% + 1400 & 6.35 & 15.01% & 6.35 & 15.01% & 6.35 & 15.01% & 6.35 & 15.01% + 1450 & 2.68 & 14.27% & 2.74 & 14.34% & 2.83 & 14.43% & 2.68 & 14.27% + 1500 & 1.13 & 13.91% & 1.18 & 14.02% & 1.26 & 14.16% & 1.08 & 13.82% + 1550 & 0.43 & 13.57% & 0.51 & 13.88% & 0.56 & 14.05% & 0.43 & 13.59% + 1600 & 0.20 & 13.69% & 0.22 & 13.83% & 0.25 & 14.03% & 0.17 & 13.49% + table [ tab:4 ] shows cboe prices for call options on the spx from 10 april 2010 and their implied volatilities in columns 2 and 3 .
columns 4 and 5 show call option prices and implied volatilities obtained by calibrating an med to call prices at strikes @xmath331 , @xmath332 , @xmath318 .
columns 6 and 7 show call option prices and implied volatilities obtained by calibrating an med to call and digital prices at the same strikes , using call spread prices at @xmath333 , @xmath334 , @xmath335 as substitutes for the digital prices .
columns 8 and 9 show call option prices and implied volatilities obtained by calibrating an med to call and digital prices at all nine strikes .
of course the med obtained using calls and digitals at three strikes uses more `` information '' than the med obtained using just calls at three strikes .
it is therefore not surprising that it leads to a better fit .
however , to show that this fit is already very good , we also report the med obtained from call prices at all nine strikes used in this example .
figure [ fig:5 ] illustrates graphically that these last two med s are indeed very close to each other . .
in most situations the information observed in the market regarding an asset consists of option prices at a discrete set of strikes . using this to extrapolate the second derivative of a function everywhere , as suggested by breeden and litzenberger @xcite or the volatility approach @xcite , @xcite , relies on additional assumptions about the distribution of returns , the sde the asset follows and/or the choice of an interpolation method .
even when there are strong reasons for such assumptions , we believe that it is important to know the shape of the distribution function given by the principle of maximum entropy ( pme ) in case these assumptions turn out to be flawed . in the local volatility model call prices for all strikes in @xmath336 are needed , and , additionally , it assumes that the smile volatility is twice continuously differentiable .
hence this approach requires an infinity of non - quoted prices together with a strong regularity .
`` since the market provides call prices at only a small number of strike prices , the second derivative must be estimated by interpolation .
this method is not very robust as the results are very sensitive to the interpolation scheme used . ''
@xcite for the method proposed here , if there are no observable digital quotes in the market , the `` artificial '' data required consists only of digital prices for a finite , usually small , set of strikes @xmath337 . if one assumes ( and our approach does not ) that the volatility smile is differentiable with respect to the strike at the points of @xmath338 , then prescribing digital prices there is indeed equivalent to prescribing the value of the smile derivative at those points .
this is still a much weaker requirement than that of the local volatility model .
moreover , the example in section [ spx - calibration ] shows that centered call spread prices are very good estimators for digital prices .
entropy has been one of the main concepts in information theory @xcite , and since market participants react to information when taking their positions , we believe entropy is a very natural tool for use in finance .
entropy can be seen as a measure of how unbiased a probability distribution is .
hence , by maximising entropy , what we propose is to find the most unbiased probability distribution which agrees with information provided from the market .
we then show how this hypothesis leads to a piecewise exponential density .
the method we propose can be used reliably and efficiently in practice . on the one hand
, we have seen that it produces a remarkably realistic volatility surface from just one volatility number as in the original black - scholes model , with a steep skew for short maturities that decays with increasing maturity . on the other hand ,
if the actual distribution is known , then with option prices given at five or more strikes , the fit to it is very close . in particular
, it can be used as a robust interpolation method for volatility curves .
if additionally there is knowledge of a prior distribution , the principle of maximum relative entropy can be applied to find a density that takes this into account and also meets the new constraints .
we give an example of such a scenario with two log - normal distributions , and show that the convergence to the actual distribution is particularly quick .
buchen and kelly have proposed a similar method of finding a probability density that maximises entropy when the market data consists only of call options .
the density they obtain is continuous .
however , to find its parameters they must solve a multi - dimensional root - finding problem with the newton - raphson algorithm .
one criticism often raised in this application of the pme is that the method of finding the form of the density uses lagrange multipliers and is not rigorous .
indeed , this technique works well in practice and leads to the correct form , but we also give a complete mathematical proof that avoids them .
relative entropy has often been compared to a metric for probability distributions .
our proof uses results by csiszr that give additional insights into `` distances '' between distributions and establish remarkable `` geometric '' results .
since we have an explicit form of the density , we are able to give analytical formulas for the distribution , inverse distribution , and call and digital option prices . using euler s relation for homogeneous functions , we give formulas for spot- and forward - deltas .
we also include two examples in which we calibrate to real market data from the cboe .
we show that our method performs very well in both cases and compare our results to those obtained by using buchen and kelly s approach .
cassio neri currently works as quantitative analyst for lloyds banking group in london .
lorenz schneider is assistant professor in quantitative finance at emlyon business school in lyon .
they have ph.d.s in mathematics from universities paris ix and vi , respectively . | we obtain the maximum entropy distribution for an asset from call and digital option prices . a rigorous mathematical proof of its existence and exponential form is given , which can also be applied to legitimise a formal derivation by buchen and kelly @xcite .
we give a simple and robust algorithm for our method and compare our results to theirs .
we present numerical results which show that our approach implies very realistic volatility surfaces even when calibrating only to at - the - money options .
finally , we apply our approach to options on the s&p 500 index .
* keywords : * entropy @xmath0 information theory @xmath0 @xmath1-divergence @xmath0 asset distribution @xmath0 option pricing @xmath0 volatility smile * mathematics subject classification ( 2000 ) : * 91b24 @xmath0 91b28 @xmath0 91b70 @xmath0 94a17 * jel : * c16 @xmath0 c63 @xmath0 g13 |
extending the standard model ( sm ) higgs sector to two higgs doublets , @xmath0 gives the two - higgs - doublet model ( thdm ) .
there , the potential may contain many more terms than in the sm ; see e.g. @xcite .
the most general thdm higgs potential can be written @xcite @xmath1 + [ \lambda_6 ( \varphi_1^\dagger \varphi_2 ) + \lambda_6^ * ( \varphi_2^\dagger \varphi_1 ) ] ( \varphi_1^\dagger \varphi_1 ) + [ \lambda_7 ( \varphi_1^\dagger \varphi_2 ) + \lambda_7^ * ( \varphi_2^\dagger \varphi_1 ) ] ( \varphi_2^\dagger \varphi_2)\;,\end{gathered}\ ] ] with @xmath2 , @xmath3 , @xmath4 real and @xmath5 , @xmath6 complex .
many properties of thdms turn out to have a simple geometric meaning if we introduce gauge invariant bilinears @xcite , @xmath7 in terms of these bilinears @xmath8 , @xmath9 , the higgs potential reads @xmath10 with parameters @xmath11 , @xmath12 , 3-component vectors @xmath13 , @xmath14 and a @xmath15 matrix @xmath16 , all real .
the standard cp transformation of the higgs - doublet fields is defined by @xmath17 in terms of the bilinears , this standard cp transformation is @xcite @xmath18 where @xmath19 , corresponding in @xmath20 space to a reflection on the 13 plane . generalised cp transformations ( gcps ) are by @xcite @xmath21 with @xmath22 an arbitrary unitary @xmath23 matrix . in terms of the bilinears
this reads @xcite @xmath24 with an improper rotation matrix @xmath25 . requiring @xmath26 leads to two types of gcps . in @xmath20
space : @xmath27 while the cp transformations of type @xmath28 are equivalent to the standard cp transformation , respectively , the point reflection transformation of type @xmath29 is quite different and turns out to have very interesting properties . motivated by this geometric picture of generalised cp transformations , the most general thdm invariant under the point reflection ( @xmath30 ) has been studied in @xcite .
the corresponding potential has to obey the conditions @xmath31 , @xmath32 this model is , besides the point reflection symmetry of type @xmath29 , invariant under three gcps of type @xmath28 .
we call this model therefore maximally cp symmetric model , mcpm . requiring also maximally cp symmetric yukawa couplings
we find that at least two fermion families are necessary in order to have non - vanishing fermion masses .
that is , we find a reason for family replication in the mcpm .
furthermore , requiring absence of large flavor changing neutral currents it was shown that the yukawa couplings are completely fixed .
for instance for the lepton sector we get the yukawa couplings @xmath33 in the unitary gauge electroweak symmetry breaking gives @xmath34 with the standard vacuum - expectation value .
the physical higgs - boson fields are @xmath35 , @xmath36 , @xmath37 , and @xmath38 .
let us briefly summarize the essential properties of the mcpm : * there are 5 physical higgs particles , two cp even ones @xmath35 , @xmath36 , one cp odd one @xmath37 , and a charged higgs - boson pair @xmath38 . * the @xmath35 boson couples exclusively to the third @xmath39 family , @xmath35 behaves like the sm higgs boson . * the higgs bosons @xmath36 , @xmath37 , @xmath38 couple exclusively to the second @xmath40 family with strengths proportional to the masses of the third generation fermions .
* the first @xmath41 family is uncoupled to the higgs bosons . for further details we refer to @xcite .
since the yukawa couplings of the @xmath36 , @xmath37 , @xmath38 higgs bosons to the second fermion family are proportional to the third - fermion - family masses we have large cross sections for drell yan type higgs - boson production . for the same reason we have large decay rates of these higgs bosons to the second generation fermions . in figure [ fig - diag ]
we show the diagrams for these production and decay reactions in @xmath42 collisions . in @xcite
the cross sections were computed for drell - yan higgs - boson production at the tevatron and the lhc for center - of - mass energies of 1.96 tev and 14 tev , respectively . in @xcite radiative effects were considered . here
we add the cross sections for a center - of - mass energy of 7 tev at lhc , which is currently available .
the corresponding total cross sections for the drell yan production of the @xmath36 , @xmath37 , @xmath38 bosons are shown in figure [ figs ] . in this figure
we also recall the branching ratios of the @xmath37 boson decays .
left : total cross section of drell yan type higgs boson production at tevatron and lhc .
right : branching ratios of the cp odd @xmath37 higgs boson , where a mass of @xmath44 gev is assumed.,title="fig:",scaledwidth=45.0% ] left : total cross section of drell yan type higgs boson production at tevatron and lhc .
right : branching ratios of the cp odd @xmath37 higgs boson , where a mass of @xmath44 gev is assumed.,title="fig:",scaledwidth=45.0% ] as an example consider higgs - boson masses @xmath36 , @xmath37 , @xmath38 of 200 gev where we get very large total production cross sections , around 850 pb , for lhc7 . these higgs bosons decay mainly into light @xmath45 and @xmath46 quarks .
however , tagging of @xmath45 and @xmath46-quarks in the detectors is at least challenging .
channels involving muons should be more easily accessible experimentally . with the branching ratio of @xmath47 into @xmath48-pairs , we predict about 25 @xmath48 events from a 200 gev @xmath36 ( @xmath37 ) at lhc7 for 1 fb@xmath49 integrated luminosity . for further details of the calculations we refer to @xcite .
j. f. gunion , h. e. haber , g. l. kane and s. dawson , `` errata for the higgs hunter s guide '' , [ arxiv : hep - ph/9302272 ] .
h. e. haber and r. hempfling , phys .
d * 48 * 4280 ( 1993 ) [ arxiv : hep - ph/9307201 ] .
m. maniatis , a. von manteuffel , o. nachtmann and f. nagel , c. c. nishi , phys .
d * 74 * 036003 ( 2006 ) .
m. maniatis , a. von manteuffel and o. nachtmann , t. d. lee and g. c. wick , phys .
rev . * 148 * , 1385 ( 1966 ) .
g. ecker , w. grimus and w. konetschny , nucl .
b * 191 * ( 1981 ) 465 , g. ecker , w. grimus and h. neufeld , nucl .
b * 247 * ( 1984 ) 70 , m. maniatis , a. von manteuffel and o. nachtmann , m. maniatis and o. nachtmann , m. maniatis and o. nachtmann , | predictions for lhc physics are given for a two - higgs - doublet model having four generalized cp symmetries . in this _ maximally - cp - symmetric model _
( mcpm ) the first fermion family is , at tree level , uncoupled to the higgs fields and thus massless . the second and third fermion families have a very symmetric coupling to the higgs fields .
but through the electroweak symmetry breaking a large mass hierarchy is generated between these fermion families .
thus , the fermion mass spectrum of the model presents a rough approximation to what is observed in nature . in the mcpm the couplings of the higgs bosons to the fermions are completely fixed .
this allows us to present clear predictions for the production at the lhc and for the decays of the physical higgs bosons . as salient feature
we find rather large cross sections for higgs - boson production via drell yan type processes . in this paper
we present a short outline of the model and extend a former study by the predictions at lhc for a center - of - mass energy of 7 tev . |
due to brownian motion , it is impossible to keep a molecule within the detection volume for an extended period of time .
the study of nanofluidics became popular to circumvent this problem with wide scientific applications @xcite . with the advancement of nanotechnologies , nanofluidic devices comprising arrays of nanochannels with diameters
less than hundred nanometres have become significant for bioanalytical diagnostics applications , such as dna optical mapping @xcite , single virus and nanoparticle detection @xcite .
nanochannels are useful tools allowing systematic studies of entities from single molecule to viruses over long periods of time .
it can induce 1d fluidic confinement by suppressing the thermal motions in two directions .
the primary problem is to establish a simple and efficient way of nanofabricating 1d nanofluidic channels .
different methods have been published , which described the fabrication processes of such nanofluidic devices @xcite .
however , majority of them are technically challenging , costly , and not easily applicable for high - throughput production .
hence , we present a cost - effective and simple nanofabrication technique based on electron beam lithography ( ebl ) and shadow - angle - electron - beam - deposition ( saebd ) . the nanochannels were used to detect a flow of photoluminescent nanomaterials , such as small dna molecules labelled with single organic fluorophores , carbon nanodots ( cnd ) @xcite , and pure organic fluorophores . using two - focus fluorescence correlation spectroscopy ( 2ffcs ) @xcite , we recorded the transits of single molecules or nanodots through the nanochannels , and quantitatively analysed their flow velocities . here
, the detection regions were two diffraction limited focal spots with a lateral displacement of 400 nm . to ensure 1d tranport between the two foci , the diameter of the nanochannels should be smaller than the focal diameters .
the nanochannels presented here have diameters ranging from 30 nm to 100 nm , which are smaller than the focal diameter .
the process steps to create an enclosed nanochannels involve nanofabrication of open nanochannels using ebl and rie and enclosing them using saebd . the ballistic path of the electron beam ( e - beam ) assisted evaporation @xcite is the principle behind saebd .
when a straight beam of depositing material hits onto an open nano - trench at shallow incident angles , no deposition occurs in the shadowed region @xcite
. such deposition can enclose a large number of parallel nanochannels ( depending on the e - beam diameter ) leaving the shadowed region as the fluidic path of interest .
the process is unaffected by nanometre sized leftover residues of the e - beam resists and does not require an atomically clean surface , unlike any wafer bonding process @xcite . for optical reasons ,
the final nanochannels were prepared using pure silicon dioxide .
being an insulator silica contributes to surface charging under e - beam which affects ebl and sem .
ebl and sem induced surface charges were compensated using a thin - film of gold . here
we show the saebd process using silicon to obtain high resolution electron microscopy images of the intermediate steps .
figure [ fig:1]*a * shows a schematic flow - chart of creating nanochannels on [ 100 ] silicon wafers .
the width of the nano - trenches ( i.e. the final width of the enclosed nanochannels ) can be optimised by the e - beam exposure of the ebl to the positive e - beam resist .
different widths of nano - trenches were created ranging from 30 nm to 100 nm . in figure
[ fig:1]*b * we observe nano - trenches of such widths , 65 nm and 100 nm .
the nanolithographed e - beam resist acted as a mask for rie to etch the final nano - trenches on silicon .
the depth of the nano - trenches was investigated with afm as 35 nm to 40 nm . .
* d. * shadows of electron beam arrows represent angular e - beam evaporation . * e. * schematic of high angle deposition and * f. * sem of deposition at @xmath1 . * g. * schematic of low angle deposition and * h. * sem of deposition at @xmath2 .
* i. * fib cross - section of the enclosed nanochannels .
two layers of platinum were used to protect the nanochannel form high energy ions . *
j. * magnified view of the enclosed nanochannel , where si-1 is the silicon substrate on which nano - trenches were fabricated , ti-2 is the 50 nm thick titanium layer which was used in saebd , pt-3 and pt-4 were platinum deposited inside fib .
* k. * further magnified view of the nanochannel where ` 1 ' represent silicon ( si-1 ) and ` 2 ' represent titanium ( ti-2 ) .
[ fig:1 ] ] in the next step , saebd was used to enclose the nano - trenches to create enclosed nanochannels .
figure [ fig:1]*c * schematically explains the concept of the saebd process .
a high - energy e - beam ( bent with magnet ) sublimates the material which deposits onto the substrate .
figure [ fig:1]*d * schematically explains the role of deposition angle ( @xmath3 ) .
the shadowed region is created due to the angular depositions .
this region is unexposed to the depositing material .
the deposited region and shadowed region are colour coded red and yellow , respectively . in the time evolution
schematic , the saebd leads to enclosing the nano - trenches leaving apart a void . by decreasing @xmath3 , one can increase the hypotenuse . at an acute angle close to @xmath4 ,
the hypotenuse becomes nearly parallel to the base as shown in figure [ fig:1]*e*-*h*. figure [ fig:1]*e * schematically represents a high angle deposition which was realised at @xmath1 in figure [ fig:1]*f*. here , an array of 5 mm long nano - trenches were cross - sectioned using a wafer sawing instrument to observe the intermediate steps while performing saebd .
figure [ fig:1]*g * schematically represents a low angle deposition which was realised at @xmath2 in figure [ fig:1]*f*. designing the deposition stage is difficult for an acute angle close to @xmath4 .
nevertheless , satisfactory results were obtained using @xmath5 as shown in figure [ fig:1]*i*-*k*. 60 nm titanium was deposited on the open nano - trenches at an angle of @xmath2 with a deposition rate of 1 / s at a pressure of @xmath6 mbar .
we predict that at an angle @xmath7 one can prepare high - quality flat edge .
we used fib to cross - section the produced nanochannels for characterisation . to avoid ion beam induced damage ,
the top part of the region to be milled was protected with metallic thin - film layers .
we deposited two thin - films on the top surface of the enclosed nanochannels .
figure [ fig:1]*i * - * k * show sem images of the milled regions from low to high magnification . in figure
[ fig:1]*j * the first layer ( si-1 ) is the silicon substrate on which nano - trenches were fabricated .
the second layer is the @xmath2 saebd - ed titanium ( ti-2 ) layer .
the third and fourth layers
pt-3 and pt-4 are platinum layers of 100 nm and 450 nm , respectively acted as protective layers to avoid fib induced damage .
pt-3 and pt-4 were deposited using fib , which have no relation to the saebd process . in figure
[ fig:1]*k * , we observe the magnified cross - section of an enclosed nanochannel . here , ` 1 ' and ` 2 ' refer to si-1 and ti-2 layers , respectively . as speculated ,
the non - conformal saebd growth of titanium thin - film produces a well defined flat encloser of the nanochannel .
a 60 nm of titanium saebd at @xmath2 angle should produce a vertical thickness of 51 nm . in figure [ fig:1]*j * ,
the vertical thickness of the titanium film is 47.8 nm .
considering the uncertainty of few nanometres deposition thickness , this agrees well with our expectation .
after this proof of principle experiment with titanium , fused silica based nanofluidic devices were prepared . here , 5 nm of gold thin - film was sputter coated prior to spin - coating e - beam resist on the silica wafer .
monte - carlo simulations were performed to optimise gold thin - film thickness in order to reduce charging effect of silica under ebl . fused silica based nano - trenches were enclosed with saebd using silicon dioxide at @xmath2 .
the design of the silica based nanofluidic device to perform single molecule experiments is schematically illustrated in figure [ fig:2]*a*. it has two reservoirs with gradual decrease of length - scale from milli - scale to microscale and finally , the nanometre - sized channels .
two milli - scale reservoirs were used as inlet and outlet .
these two inlets were sand blasted on silica wafers using 70 @xmath8 m silica particles prior to nanofabrication process .
the microscale reservoirs with the same thickness of nanochannels and micrometre width were etched with rie . in the figure [ fig:2]*a * ,
the white regions and blue regions are milli - scale and microscale reservoirs , respectively ( supporting information figure s1 and s2 ) .
the red stripes correspond to nanochannels .
they are connected to the microscale reservoirs .
an sem image of these silica nanochannels is also shown in the right inset of figure [ fig:2]*a*. we filled these nanochannels with a solution of fluorescent probes .
they were unidirectionally transported with electro - osmotic flow @xcite as shown in figure [ fig:2]*b*. after filling one reservoir with fluorescent molecules ( diluted in pbs buffer - see methods - electro - osmotic flow ) capillary force transported the fluid to the other reservoir .
a relaxation time of 30 s was given to avoid development of trapped air bubbles , and then only , the second reservoir was filled .
two 100 @xmath8 m thick platinum electrodes were immersed into the pbs filled reservoirs ( supporting information figure s1 ) , and an electric field was applied along the nanochannel ( supporting information figure s1 and s4 electrode integration with the nanofluidic device ) . in figure [ fig:2]*c * , single alexa-647 fluorophore molecules ( purchased from thermo fisher , massachusetts , usa ) are horizontally lined up in all the parallel nanochannels .
the pixel to photon counts profile plot shows an average snr of 90 .
this image was captured with a wide - field optical microscope by exciting the molecules with a 640 nm continuous wave laser ( coherent laser systems gmbh , gttingen , germany ) . beside the evidence of fib and sem
, this also infers that the nanochannels are properly enclosed .
no cross - talking of single molecules is observed among nanochannels @xcite .
fluid flow inside a nanochannel is 1d when @xmath9 , where @xmath10 and @xmath11 are the diameters of the detecting region and the nanochannels , respectively ( figure [ fig:2]*a * ) .
the degree of confinement is acceptable until the molecule is forced within the detection volume .
this condition is satisfied because in our case @xmath10 is 300 nm to 500 nm ( exciation wavelenght dependent ) and @xmath11 is 30 nm to 100 nm .
the in - time and out - time of flowing single molecules within detection volumes were correlated using two apds of 2ffcs ( figure [ fig:2]*b * ) @xcite .
the diffraction limited two foci of 2ffcs were 40 mhz pulsed interleaved excitation .
the two foci were partially overlapping with each other .
they were generated using a nomarski prism , a 100@xmath12 1.49 na of oil immersion objective , and two linearly polarised lasers perpendicular to each other .
an _ x - y - z _ piezo - scanner was used to acquire scan images and point measurements of photon counts . to restrict unwanted surface adsorption of single molecule , we used fluorescent molecules carrying the same charge as the nanochannels wall .
pure silica is negatively charged above its isoelectric point ( ph(i ) = 2 ) @xcite .
a buffer of ph 8.5 was used to increase the negative charges on the walls of silica nanochannels .
the condition was suitable to flow negatively charged molecules without being adsorbed .
we chose carbon - nanodots ( cnd ) as the first probe of interest .
they are @xmath132 nm in size @xcite and negatively charged ( see supporting information figure s5 ) .
a confocal scanned image was recorded along the _ x - y _ plane when cnds were flowing inside nanochannel . in the schematic top - view of the nanochannels in figure [ fig:4]*a * dark lines
are the nanochannels .
they were filled with carbon - dots ( figure [ fig:4]*b * ) .
the pixel size ( @xmath14 ) of the image is 320 nm with dwell time ( @xmath15 ) of 5 ms . to record
this focused _ z _ plane in figure [ fig:4]*b * , we performed a _
y - z _ scanned image and went to the position , where the count - rate was maximum .
the scheme of _ y - z _ piezo - scan is shown in figure [ fig:4]*c*. the height - wise cross - section of several periodic point spread functions ( psf ) were recorded from periodic nanochannels .
when single cnds are flowing in unidirectionally , jittery ( or fluctuating ) fluorescent signals are recored as observed in the periodic psfs figure [ fig:4]*d * , * e * , and * f*. an ascending concentrations of aqueous cnd soluions were used . in figure
[ fig:4]*d * to * g * , the volume percentage of cnds in water relative to its stock solution are @xmath16 , and @xmath17 , respectively . here
, @xmath14 is 100 nm and @xmath15 is 2 ms .
the applied electric field to induce electro - osmotic flow during the measurement was 15 v / mm .
as the concentration of cnds increased , the fluctuations of the photon signal decreased due to increasing rate of cnds arrival in the focal volume ( figure [ fig:4]*d * - * g * ) . .
the data was fitted with 1d fokker - planck equation . from the fitting , we found diffussion coefficient of @xmath18 @xmath19 .
* b. * plot of velocity versus applied voltage along the nanochannel .
the data was fitted with linear fit . * c. * burst size distribution of single molecule transits fitted with two poissoinian distributions .
total time trace binned with 5 ms is shown in the inset .
the inset showing total time trace the dashed bordered grey region is shown in * d*. * d. * a part of the time trace plots of single photon bursts due single dna molecules transits through focus .
[ fig:5 ] ] time - correlated analysis of photon counts was carried out to investigate flow velocities of single molecules .
these measurements were carried out at the highest photon counted points of _ y - z _ scanned psfs .
prior to performing every 2ffcs measurement , a _
y - z _ scan was taken as shown in figure [ fig:5 ] . to substantiate that our method is not restricted to a particular fluorescent probe , we performed 2ffcs of atto 488 dye ( see supporting information for a control measurement of diffusion figure s6 ) and 48 base - pair of dna labelled with single alexa-647 fluorophore ( purchased from iba gmbh , gttingen , germany ) .
the electrolytic aqueous buffer solution used for the electrosmostic flow is detailed in the method section .
we measured the flow velocities of single dna molecules inside 1d nanochannel within a range of applied electric field from 27 v / mm to 300 v / mm ( see supporting information figure s7 ) .
these measurements were performed inside a nanochannel with @xmath11 of 30 nm .
the correlated photon counts provide two autocorrelations for two foci , and one forward and one backward cross - correlation between them @xcite .
a difference should be observed in two cross - correlations due to a unidirectional flow profile as shown earlier by arbour and enderlein @xcite .
the correlated data points from 2ffcs are fitted with the fokker - planck equation @xcite considering the 1d electro - osmotic flow .
an exemplary correlated photon counts of 1d flow measurement performed at 220 v / mm is shown in figure [ fig:5]*a*. here , sky - blue and red fitted curves are two cross - correlations , and other two curves are autocorrelations .
the velocity and diffusion coefficient from the curve fitting are @xmath20 @xmath21 and -207 @xmath22 m/s , respectively . the negative value of the flow velocity infers the direction of the flow , which can be altered by changing the applied polarity of the electric field .
the linear relationship of electro - osmotic flows at different applied electric fields is plotted in figure [ fig:5]*b*. the linear fit has an @xmath23 value of @xmath24 . to confirm whether the photon used in 2ffcs were from aggregates of single molecules , photon burst sizes of molecular transits were analysed @xcite . burst size distribution ( bsd ) of the complete time trace for a flow measurement
is shown in figure [ fig:5]*c*. the complete time trace used in the bsd analysis is shown in the inset .
we fit the bsd with poisson distributions .
the first peak is due to the single molecule bursts , and the second is due to multiple - molecule - events as observed by enderlein et al .
multiple - molecule - events occur when more than one molecule transits through focus in close succession that the detectors can not resolve the temporal discretisations between two or more molecules .
figure [ fig:5]*d * shows a magnified region of the inset 5 ms binned time - trace of figure [ fig:5]*c * where 0 to 600 ms is highlighted with a dashed box ( grey inside ) .
we clearly observe single molecule bursts in figure [ fig:5]*d*. in summary , this letter presented an efficient method ( saebd ) of nanofabricating nanochannels and their usage in single molecule nanofluidics to confine the thermal fluctuations .
the nanofabrication process demonstrated here is simple and high - throughput for large scale producion of nanofluidic chips .
the process is not material restricted , unlike the oxidation and bonding based techniques .
single fluorophore molecules , cnds , and dna molecules were studied inside nanochannels .
1d flows of them inside nanochannel were achieved using electro - osmosis .
wide - field and confocal microscopy were used for qualitative analysis of the 1d flow . using 2ffcs method
the 1d flows of the mentioned species were quantitatively analysed .
a broad range of velocities was observed by varying the applied electric field along the nanochannels .
the bsd analysis confirm that the observed transits were mainly due to single emitters .
all the experiments were performed inside nanochannels of 30 nm @xmath12 35 nm in cross - section .
this simple and high - throughput approach of fabricating nanochannels paves the way towards detecting early onset of any disease at single molecule level . in future , trapping nanoscale objects of less than 2 nm in size for large residence time should be also feasible using these nanochannels .
biomolecular interactions such as dynamics of dna , protein aggregation , and structural biology of molecules in physiological condition can be also studied at single molecule level using the saebd based nanofluidic devices .
* nanofabrication : * a rigorous cleaning of the surface of interest is required to eliminate unwanted organic or dust - particle contaminations .
the silicon wafers were ultrasonicated with acetone prior to piranha cleaning at 120@xmath25 c for 15 min .
the piranha cleaned wafers were then rinsed with de - ionised water .
the wafers were heated at 180@xmath25 c for 10 min to make sure all the water molecules were evaporated .
the cleaning protocol was carried out inside a class 100 cleanroom .
prior to spin - coating , the wafers with e - beam resist , an upright optical microscope with 100@xmath12 objective lens was used to investigate the surface of the silicon .
fused silica wafers were cleaned with rca1 reagents instead of piranha .
the pristine wafers were spin - coated with the e - beam resist pmma - a2 ( polymethylene methacrylate from microchem corp . ,
newton , united states ) at 2000 rpm for 60 s. the spin - coated wafers were then baked at 180@xmath25 c for 90 s. a uniform 100 nm thick pmma film was produced on the wafers .
ebl was carried out on the pmma coated wafers using fixed - e - beam - moving - stage condition @xcite the first experiment where silicon wafers were used
nanochannel dimensions were 2 mm long and 30 nm - 100 nm wide .
the final fused silica based nanofludic device contained 100 150 @xmath0 m long nanochannels . to avoid overlapping periodic psfs , the distance between every two nanochannels was 2 3 @xmath0 m .
the ebl exposed pmma was monomerised and removed by immersing the wafers into mibk : ipa ( methyl isobutyl ketone dissolved in isopropanol purchased from microchem corp .
, newton , united states ) developer at -20@xmath25 c for 30 s. after 30 s , the wafers were immersed into isopropanol to stop over - development and nitrogen blown dried .
nanometre scale openings were formed revealing the silicon or silica where pmma was removed with the developer .
the remaining undeveloped pmma regions acted as a mask where no etching was performed for rie .
the reactive plasma was created at high vacuum .
* saebd : * saebd was performed using a leybold univex 350 e - beam evaporation system purchased from oerlikon leybold vacuum gmbh , cologne , germany .
the vacuum pressure inside the chamber while evaporation was @xmath26 mbar .
the dicing system used to cross - section the silicon wafers was disco dad 320 purchased from disco corporation , tokyo , japan . the diamond blade used in
the system was zh05-sd2000-n1 - 70 ( disco corporation , tokyo , japan ) .
* fib : * fib used here is nova nanolab 600 dualbeam purchased from fei company , hillsboro , usa . the protection layer of platinum was deposited in two separate growth rates . at the interface of ti-2 we used 1 pa current for the deposition to achieve an uninterrupted and atomic interface .
the top most surface ( pt-4 ) was grown at fast deposition rate because this surface is the first interacting surface with fib dissipating majority of the damaging ionic bombardments . to avoid any damage at the region of interest ,
ion beam was created with 1 pa current at system vacuum of @xmath27 mbar . *
2ffcs setup : * we used the commercial confocal system microtime 200 ( picoquant , berlin , germany ) for scanning our samples and performing 2ffcs .
the details of the setup are included in the supporting information .
the supporting information contains the description of integrating the nanochannels to electrodes , design of the nanofludic device , carbon nanodots cation - pi interaction behaviour , 2ffcs measured 1d diffusion of single atto 488 fluorophores and electro - osmotic flow of single dna molecules at different electric fields .
thanks the international max planck research school for physics of biological and complex systems and the ministry of science and culture ( lower saxony ) for awarding an excellence stipend / mwk ph.d . scholarship .
the work presented here was carried out at the research group of prof .
jrg enderlein .
the authors thank the internal funding from the university of gttingen associated to the third institute of physics for this project .
lastly , the authors are extremely grateful to prof .
enderlein for providing enormous amount of critical scientific advices .
s.g . has written the manuscript , conceptualised the idea of nanofabrication , prepared the nanofluidic device , performed all the measurements and numerical fitting presented in this paper .
and n.k . have performed intial measurements together ( not presented here ) .
have modified the fitting routine provided by prof .
jrg enderlein and i.g .
has supervised the project and provided all necessary scientific advices .
all the authors have read and made adequate correction in the manuscripts .
miao wang , nan jing , chin b. su , jun kameoka , chao - kai chou , mien - chie hung , and kuang - an chang .
_ electrospinning of silica nanochannels for single molecule detection .
_ applied physics letters 88 , no .
3 ( 2006 ) : 033106 .
chao wang , robert l. bruce , elizabeth a. duch , jyotica v. patel , joshua t. smith , yann astier , benjamin h. wunsch et al . _ hydrodynamics of diamond - shaped gradient nanopillar arrays for effective dna translocation into nanochannels .
_ acs nano 9 , no . 2 ( 2015 )
: 1206 - 1218 .
mark n. hamblin , jie xuan , daniel maynes , h. dennis tolley , david m. belnap , adam t. woolley , milton l. lee , and aaron r. hawkins .
_ selective trapping and concentration of nanoparticles and viruses in dual - height nanofluidic channels . _ lab on a chip 10 , no . 2 ( 2010 )
: 173 - 178 .
jean - luc fraikin , , tambet teesalu , christopher m. mckenney , erkki ruoslahti , and andrew n. cleland .
_ a high - throughput label - free nanoparticle analyser .
_ nature nanotechnology 6 , no . 5 ( 2011 ) : 308 - 313 .
fredrik persson , l h thamdrup , m bl mikkelsen , s e jaarlgar , p skafte - pedersen , h bruus , and a kristensen , _
double thermal oxidation scheme for the fabrication of sio@xmath28 nanochannels_. nanotechnology 18 ( 2007 ) : 245301 - 245305 .
so hyun kim , yidan cui , min jung lee , seong - won nam , doori oh , seong ho kang , youn sang kim and sungsu park , _ simple fabrication of hydrophilic nanochannels sing the chemical bonding between activatied ultrathin pdms layer and cover glass by oxygen plasma_. lab chip 11 ( 2011 ) : 348 - 353 .
siddharth ghosh , anna m. chizhik , narain karedla , mariia o. dekaliuk , ingo gregor , henning schuhmann , michael seibt et al .
_ photoluminescence of carbon nanodots : dipole emission centers and electron - phonon coupling . _ nano letters 14 , no . 10
( 2014 ) : 5656 - 5661 .
michael brinkmeier , klaus drre , jens stephan , and manfred eigen .
_ two - beam cross - correlation : a method to characterize transport phenomena in micrometer - sized structures .
_ analytical chemistry 71 , no . 3 ( 1999 ) : 609 - 616 .
thomas dertinger , victor pacheco , iris von der hocht , rudolf hartmann , ingo gregor , and jrg enderlein .
_ two - focus fluorescence correlation spectroscopy : a new tool for accurate and absolute diffusion measurements . _ chemphyschem 8 , no . 3 ( 2007 )
: 433 - 443 .
thomas dertinger , anastasia loman , benjamin ewers , claus b. moeller , benedikt kraemer , and jrg enderlein . _ the optics and performance of dual - focus fluorescence correlation spectroscopy .
_ optics express 16 , no .
19 ( 2008 ) : 14353 - 14368 .
j. j. merkel , t. sontheimer , b. rech and c. becker .
_ directional growth and crystallization of silicon thin films prepared by electron - beam evaporation on oblique and textured surfaces .
_ journal of crystal growth 367 ( 2013 ) : 126 - 130 .
stefan egger , adelina ilie , shinichi machida , and tomonobu nakayama .
_ integration of individual nanoscale structures into devices using dynamic nanostenciling .
_ nano letters 7 , no . 11 ( 2007 )
: 3399 - 3404 .
siddharth ghosh , manoj kumbhakar , mitja platen , ingo gregor , and jrg enderlein .
_ single - molecule fluorescence inside solid - state nanochannels . _ in spie bios , pp .
895008 - 895008 .
international society for optics and photonics , 2014 .
xiaohua huang , manuel j. gordon , and richard n. zare . _
current - monitoring method for measuring the electroosmotic flow rate in capillary zone electrophoresis . _ analytical chemistry 60 , no .
17 ( 1988 ) : 1837 - 1838 .
manish keswani , srini raghavan , pierre deymier , and steven verhaverbeke .
_ megasonic cleaning of wafers in electrolyte solutions : possible role of electro - acoustic and cavitation effects .
_ microelectronic engineering 86 , no . 2 ( 2009 ) : 132 - 139 .
jrg enderlein , david l. robbins , w. patrick ambrose , and richard a. keller . _ _ molecular shot noise , burst size distribution , and single - molecule detection in fluid flow : effects of multiple occupancy.__the journal of physical chemistry a 102 , no .
30 ( 1998 ) : 6089 - 6094 .
michael wahl , hans - juergen rahn , tino roehlicke , gerald kell , daniel nettels , frank hillger , ben schuler , and rainer erdmann .
_ _ scalable time - correlated photon counting system with multiple independent input channels.__review of scientific instruments 79 , no . 12 ( 2008 ) | ensemble averaging experiments may conceal many fundamental molecular interactions . to overcome that
, a high - throughput detection of single molecules or colloidal nanodots is crucial for biomedical , nanoelectronic , and solid - state applications .
one - dimensional ( 1d ) discrete flow of nanoscale objects is an efficient approach in this direction .
the development of simple and cost - effective nanofluidic devices is a critical step to realise 1d flow .
this letter presents a nanofabrication technique using shadow - angle - electron - beam - deposition for a high - throughput preparation of parallel nanofluidic channels . these were used to flow and detect small fragments of dna , carbon - nanodots , and organic fluorophores .
the 1d molecular mass transport was performed using electro - osmotic flow .
the 1d flow behaviour was analysed using two - focus fluorescence correlation spectroscopy ( 2ffcs ) .
a range of flow velocities of single molecules was achieved from 15 @xmath0 m/s to 290 @xmath0 m/s .
the transitions of single molecules or nanodots through the two foci were quantitatively analysed using confocal scanning fluorescence microscopy , correlative photon detection , and burst size distribution analysis .
the results suggest an efficient nanofabrication technique is developed to prepare nanofluidic devices .
this first demonstration of high - throughput nanochannel fabrication process and using 2ffcs - based single molecule flow detection should have potential impact on ultra - sensitive biomedical diagnostics and studying biomolecular interactions as well as nanomaterials . |
direct @xcite and indirect @xcite evidence of the present acceleration of the universe is accumulating .
nonetheless , the source of the accelerated expansion is as elusive as ever . in the context of general relativity
, one needs to hypothesize a dark energy source with negative pressure to accommodate the acceleration .
alternative ideas include various modifications to gravity .
little is known about the dark energy except that its pressure is negative , and that it accounts for about @xmath0 of the critical density of the universe . whether the dark energy is the cosmological constant or a dynamically evolving source of energy is a fundamental unanswered question .
while the cosmological constant has a fixed ratio of pressure to energy density , @xmath1 , dynamical dark energy ( dde ) will in general have a varying equation of state ( eos ) , @xmath2 . observing a deviation from @xmath3 or
a time - evolution in the eos will be decisive evidence in favor of the existence of dde .
however , there are known degeneracies @xcite which make this task extremely difficult , unless the deviation from a cosmological constant is strong .
the current observational limit on the eos of dark energy is roughly @xmath4 at the @xmath5 level @xcite , which is consistent with a cosmological constant .
future experiments hold out the possibility of pinning down this limit by maybe a factor of @xmath6 . for a recent review ,
see @xcite .
the effect of the cosmological constant and dde on the expansion rate can be identical , and there is a need for probes that go beyond the background to distinguish between the two .
an example of such a probe is structure formation .
there are numerous works exploring the formation of structure in the presence of homogeneous dark energy @xcite .
an exciting and somewhat controversial possible difference between the cosmological constant and dde is their clustering behavior .
while the cosmological constant is exactly homogeneous on all scales , dde is expected to be not perfectly homogeneous @xcite , and the implications of this on the cmb are well known @xcite .
however , it is usually assumed that the clustering of dde is negligible on scales less than @xmath7 .
whether small perturbations in dde can be neglected is debatable , and a deeper understanding of the dde inhomogeneous dynamics is clearly needed .
several recent works have explored the consequences of dde clustering on scales shorter than @xmath7 . some have adopted a phenomenological approach , parameterizing the clustering degree of dde @xcite .
these works point out potential observables of dde clustering , justifying further investigation .
works which attempt a more fundamental treatment are mostly in the context of coupled dark energy @xcite , or other non - trivial models of dde @xcite , as clustering is most probable in such theories .
however , less attention has been given to the clustering in simpler models of dde .
in the present work , we aim to further explore the inhomogeneous behavior of dde .
our approach is straightforward : starting with a gravitational action which includes matter and dde , we numerically follow the linear evolution of spherical perturbations of matter and the dde response to these perturbations . for the sake of simplicity , our model for the dde is a light scalar field , which is not explicitly coupled to the matter density . as the only coupling between the dde and the matter is gravitational , our results are conservative in the sense that any model more complicated will exhibit stronger dde perturbations than shown here .
the striking feature that emerges from our calculation is that in the vicinity of collapsing matter , the dde develops a spatial profile and tends to form voids .
the mechanism that allows the void to form is that although initially the field s evolution is friction dominated due to the cosmic expansion , the collapse of matter slows down the local expansion .
this allows the field to locally roll down and lose energy , creating the void .
the presence of the matter perturbation is necessary to trigger this mechanism .
the plan of the paper is as follows : in section [ the model ] we describe our model in detail . in section [ results ]
we present our results .
we discuss our results and state our conclusions in sections [ discussion ] and [ conclusions ] respectively .
we are interested in spherical perturbations around a flat frw universe .
the most general line element in comoving coordinates is then @xmath8 where @xmath9 and @xmath10 are general functions @xcite .
we take a cosmic mix of non - relativistic matter and a dde component as the energy source .
the matter component is described by a perfect and pressureless fluid , with an energy - momentum tensor given by @xmath11 where @xmath12 is the energy density of matter .
we model the dde with a classical scalar field @xmath13 with a lagrangian @xmath14 given by @xmath15 and an energy - momentum tensor given by @xmath16 the eos of the dde @xmath17 is defined as @xmath18 with the energy density @xmath19 and the pressure @xmath20 are read off the energy momentum tensor , @xmath21 and @xmath22 respectively .
+ it is convenient to rewrite einstein s equations in the following way , @xmath23 where @xmath24 is the reimann tensor , and @xmath25 .
as there is no explicit interaction between the matter and the dde , energy conservation applies to each separately , @xmath26 the time - evolution of the system is given by the following equations ( where dots denote time - derivatives and primes denote derivatives with respect to the radial coordinate , except for @xmath27 ) : @xmath28 these are subject to the following constraint equations : @xmath29 we now proceed to separate our variables to a homogeneous background and a time and space - dependent perturbation , which we will then linearize . working in the synchronous gauge @xcite , we redefine the metric functions @xmath30 and @xmath31 as follows : @xmath32 here @xmath33 is the scale factor of the spatially homogenous and flat background , and @xmath34 and @xmath35 are the deviations .
we introduce a perturbation around a homogeneous background also in the matter and the dde , @xmath36 the zeroth order of equations - gives @xmath37&=&0 \\
\label{h0 } \dot{h}+3h^2-k\left(\frac{1}{2}\rho+ v\right)&=&0 \\
\label{rho0 } \dot{\rho}+3h\rho&=&0 \\
\label{phi0 } \ddot{\phi}+3h\dot{\phi}+v'&=&0 ~,\end{aligned}\ ] ] where @xmath38 is the hubble function .
+ to linear order , the evolution equations - give @xmath39 and the constraint equations - reduce to @xmath40 combining equations , and gives @xmath41 the only combination which is relevant to the equations of motion and is @xmath42 . comparing and
it is clear that @xmath43 can be thought of as @xmath44 , and therefore characterizes the spatial profile of the hubble function . at the cost of losing some information about the metric
, we can reduce the number of our variables and equations from @xmath45 to @xmath46 by solving for @xmath43 instead of for @xmath47 and @xmath48 .
equations , and yield @xmath49 finally , by fourier - transforming @xmath50 , @xmath51 and @xmath52 into @xmath53 , @xmath54 and @xmath55 respectively , equations - can be written as a set of ordinary differential equations : @xmath56 where we have used the fact that to linear order , @xmath57 and @xmath58 .
observationally distinguishing between various potentials of dde is a formidable task @xcite , and a careful analysis of the growth of structure in various potentials might prove a useful tool .
our present goal though is to trace generic properties of dde .
accordingly , we choose to work with a simple mass potential , @xmath59 we take the mass scale comparable to the present hubble scale , @xmath60 .
the light mass assures a slow roll behavior , which will provide accelerated cosmic expansion . unless noted otherwise
, the figures presented here refer to this mass potential . in order to verify the generality of our results
, we repeated the analysis for a more realistic potential - the double exponential @xcite , @xmath61 with @xmath62 and @xmath63 .
as we later show , the resulting behavior for the two potentials is qualitatively the same .
we want to study how the dde reacts to the clustering of matter .
thus our initial conditions are of perturbed matter and homogeneous dde .
the matter perturbation is taken as a spherical gaussian , @xmath64 and the dde is taken to be initially homogeneous , @xmath65 . a non - homogeneous evolution for the dark energy is nonetheless allowed .
to ensure that the matter perturbation has no peculiar velocity , we choose the initial condition for the metric variable as @xmath66 . from equation , we find that this amounts to the statement that initially , @xmath67 , _ i.e.
_ the matter particles making up the perturbation are being simply carried along with the hubble expansion ( or in other words , they are at rest in a comoving frame ) .
however , we have verified that choosing different initial values of @xmath43 does not qualitatively affect our results .
the numerical calculation begins at a redshift of @xmath68 with a perturbation amplitude of @xmath69 , and is run to the present , @xmath70 .
we focus on relatively short - scale perturbations , @xmath71 . with this choice of redshift span , we aim to explore the full linear range of the matter perturbation , while also allowing for a period of dde dominance .
the initial values of the background variables @xmath72 , @xmath73 and @xmath74 are chosen such that their present values ( denoted with a @xmath75 subscript ) converge to @xmath76 , @xmath77 , and a normalized hubble value , @xmath78 .
this is done with the help of a root - finding algorithm .
we take @xmath79 , which means that the initial state of the scalar field is homogeneous and with equation of state @xmath80 , similar to the cosmological constant .
we now have the layout to numerically evolve equations - .
the solutions for @xmath81 , @xmath82 and @xmath83 are then fourier transformed back to real space .
the results shown in this section are for the mass potential , equation , unless specifically noted otherwise .
the numerical run begins at a redshift of @xmath68 , when the dde is subdominant .
the dde remains subdominant through most of the growth time of the perturbation as well , and accordingly , we expect the matter density contrast @xmath84 to grow as the scale factor @xmath85 .
this is indeed confirmed by our results .
the most striking result that emerges from our calculation is that in the vicinity of collapsing matter , the dde tends to form voids , or growing regions of under - density .
the anti - correlation between the perturbations of the matter and the dde was noted in @xcite and in @xcite for particular cases .
figure [ void_demo1 ] plots the dde density contrast @xmath86 at the center of the matter perturbation , @xmath87 , against the redshift @xmath88 .
we find that the amplitude of the perturbation grows sharply at late times .
figure [ void_demo2 ] shows the growth of the absolute value of the dde perturbation with redshift on a logarithmic scale .
this plot reveals another interesting effect : the initial response of the dde to the gravitationally collapsing matter is a very weak tendency to collapse .
the collapsing phase , however , is extremely short - lived ( an @xmath89 $ ] fraction of the total time of the run ) .
the kink in the figure is the cross - over from a positive to negative perturbation .
we next look at the spatial profiles of the perturbations .
figure [ profiles ] shows the profiles of both the @xmath90 and @xmath91 at an early stage of the run ( @xmath92 ) , at an intermediate redshift ( @xmath93 ) and at the final stage ( @xmath70 ) .
the x - axis shows the physical scale as a fraction of the horizon size .
the amplitude of the perturbations is shown on a logarithmic scale , and it is worth noting the change in gap between the two scales .
the growth rate of @xmath91 is significantly faster than that of @xmath90 , so that the two amplitudes are almost comparable at late times .
this suggests that a calculation of the non - linear regime might reveal interesting behavior of the dde .
figures [ sigma_effect1 ] and [ sigma_effect2 ] show how the growth of the dde perturbation is affected by the initial width of the matter perturbation .
the figure confirms the sensitivity of dde perturbations to the scale , and as expected , shorter scales exhibit a suppressed behavior .
nonetheless , a possibly significant amplitude can be found on relevant scales .
for instance , our local supercluster has diameter of order @xmath94 and an overdensity of about @xmath95 .
our runs of @xmath96 end with an overdensity of @xmath97 , but roughly they can be taken as a measure of what we should expect on these scales . also , one must remember that this analysis only captures the physics of the linear regime .
given the sharp increase in the void amplitude at later times , it is not unreasonable to expect interesting effects in the strongly non - linear regime .
figures [ mass_effect1 ] and [ mass_effect2 ] examine the sensitivity to the field s mass .
as expected , increasing the mass of the scalar field causes @xmath91 to grow stronger and at increasingly earlier redshifts .
one can also see that if the mass is heavy enough ( an order of magnitude larger than the hubble mass ) , the field will have had enough time to enter a period of rapid oscillations , and will effectively behave as regular matter .
we would now like to focus on the local behavior of the equation of state ( eos ) of the scalar field .
let us define @xmath98 as the background homogeneous eos , @xmath99 as the leading order correction to @xmath98 and let the first - order corrected @xmath17 be @xmath100 : @xmath101 where to first order @xmath102 and @xmath103 , and we have suppressed the @xmath13 subscripts .
figure [ total_w ] shows how @xmath104 increases with time and becomes non - homogenous .
the eos at the perturbation is less negative than the background , but for this choice of mass it is still negative enough to behave as dark energy .
+ to quantify the extent of the inhomogeneity in @xmath17 we define @xmath105 to characterize the % deviation of the local @xmath17 from the background value , @xmath106 the evolution of the @xmath105 spatial profile is shown in figure [ delta_w ] .
we now aim to intuitively explain why the dde perturbation initially increases and then sharply drops in the presence of the matter perturbation .
initially , both the scalar field and the hubble function are homogenous .
the matter perturbation introduces an inhomogeneity in the hubble function , as regions with matter overdensity expand slower due to the increased gravitational pull . as a result
, @xmath107 acquires a spatial profile which evolves in time along with the matter overdensity , due to the @xmath108 term in equation .
this is illustrated in figure [ dh_vs_r ] , which plots the spatial profile of the % deviation of @xmath107 from the homogeneous background value , at three different redshifts .
regions which have a lower local value of @xmath107 offer less hubble damping to the scalar field , as the @xmath109 contribution in equation is negative .
therefore , in these regions the scalar field accelerates down its potential slightly faster , the local @xmath110 has a bigger absolute value than the acceleration of the background , @xmath111 .
thus the matter perturbation imparts a local `` downhill kick '' to the field , and the strength of the kick depends on the magnitude of the matter perturbation at that point .
the presence of matter is essential to trigger this mechanism .
a similar conclusion was reached in @xcite .
since the matter perturbation is gaussian , we can expect the profile of @xmath112 to be gaussian as well , which is confirmed in figure [ de_accn_vs_r ] .
the acceleration of the field perturbation @xmath113 is zero initially ( @xmath68 ) , but it quickly takes on a gaussian profile , which initially grows , and then shrinks at later times . the important point to note in figure [ de_accn_vs_r ]
is that the acceleration weakens , but does nt change its sign throughout the evolution .
the acceleration profile leads to spatial variations in the energy density of the dde .
for the mass potential , the linear order of the density perturbation is @xmath114 , where the first term represents the local variation in kinetic energy ( ke ) and the second term represents the local variation in potential energy ( pe ) . whether the initial value of the scalar field was shifted to the right or left of its minimum , @xmath115 and @xmath116 will be of the same sign , assuring a positive correction to the ke . on the other hand , @xmath13 and @xmath117
will be of opposite signs , assuring the pe correction will be negative .
the total correction to the energy density then will be positive when the ke dominates , creating an overdensity , and negative when the pe correction dominates , creating a void .
the fact that @xmath113 weakens but does nt change its sign assures that @xmath116 approaches a constant value , but @xmath117 keeps growing .
the pe contribution @xmath118 soon becomes dominant over the ke correction , @xmath119 .
as soon this happens , a void is created .
+ similar reasoning should apply to other slow - roll potentials .
the results we have presented so far were for the mass potential , equation .
this is not a very attractive model from the theoretical point of view , as the choice of the mass scale and initial conditions are fine tuned .
while the general problem of fine tuning has not yet been resolved , the issue of the initial conditions is alleviated in tracking potentials such as equation . to verify that our results are not unique to the mass potential , we have repeated the analysis for the double exponential potential .
we present here the equivalent of figures [ profiles ] and [ total_w ] to show that the results are essentially similar : as a reaction to the perturbation in the matter fluid , the dde quickly forms a void . starting very low ,
the rate of growth of its amplitude is significantly faster than that of the matter perturbation , motivating a further investigation into the non linear regime .
in this work we have investigated the clustering properties of dde .
we modeled the dde as a scalar field with a light mass , and have shown that in the vicinity of gravitationally collapsing matter , the dde develops inhomogeneities and forms voids .
our results show a high sensitivity to the mass scale of the field . for a mass much larger or smaller than the hubble scale ,
the field imitates the behavior of dust or the cosmological constant , respectively .
the interesting dynamics is most prominent in the window where the mass and hubble scales are comparable .
this window is within the relevant mass range for dark energy models . as the hubble function was larger in the past , heavier fields would have had comparable mass and hubble scales at some point in the past .
our results should apply to any model of dde which achieves the present acceleration through a slow roll phase , as the slow roll assures that only a small patch of the potential is probed .
we have shown this explicitly for the double exponential potential .
whether our results apply to an even wider class of models which achieve accelerated expansion not through slow roll should be further investigated .
one thing which is clear from our results , is that dde has potentially non - trivial behavior during the growth of inhomogeneities , though full non - linear analysis is needed to confirm whether the amplitude of the dde inhomogeneities is relevant to observations . as
inhomogeneities of dark energy are a clear signature differentiating between the cosmological constant and dde , such possibilities should be fully explored and exhausted .
a full treatment of the observational consequences of our results is beyond the scope of this work , but we would like to mention a few possibilities . some obvious places to look for dde inhomogeneities include lensing , the isw effect in the cmb , number counts , and mass functions . some of these directions are being pursued @xcite .
our results show that both the energy density and the eos of dde develop a spatial dependence .
thus any observation constraining either of the above which can separately be measured locally and globally is valuable . +
another possibility which we would like to point out , is that it would be useful to quantify the effect of a statistical distribution of dde voids on the cmb , or , following reference @xcite , on the directional distribution of supernovae .
for example , detecting an angular inhomogeneity in @xmath107 which is not in accordance with the matter distribution might suggest the presence of a dde void .
o. lahav , p. b. lilje , j. r. primack and m. j. rees , mon . not .
soc . * 251 * , 128 ( 1991 ) ; l. m. wang and p. j. steinhardt , astrophys .
j. * 508 * , 483 ( 1998 ) [ arxiv : astro - ph/9804015 ] ; i. t. iliev and p. r. shapiro , mon . not .
soc . * 325 * , 468 ( 2001 ) [ arxiv : astro - ph/0101067 ] ; n. n. weinberg and m. kamionkowski , mon . not .
soc . * 341 * , 251 ( 2003 ) [ arxiv : astro - ph/0210134 ] ; r. a. battye and j. weller , phys .
d * 68 * , 083506 ( 2003 ) [ arxiv : astro - ph/0305568 ] ; c. horellou and j. berge , mon . not .
360 * , 1393 ( 2005 ) [ arxiv : astro - ph/0504465 ] ; p. wang , astrophys .
j. * 640 * , 18 ( 2006 ) [ arxiv : astro - ph/0507195 ] ; m. bartelmann , m. doran and c. wetterich , arxiv : astro - ph/0507257 ; w. j. percival , arxiv : astro - ph/0508156 .
d. f. mota and c. van de bruck , astron .
astrophys . * 421 * , 71 ( 2004 ) [ arxiv : astro - ph/0401504 ] ; i. maor and o. lahav , jcap * 0507 * , 003 ( 2005 ) [ arxiv : astro - ph/0505308 ] ; n. j. nunes , a. c. da silva and n. aghanim , arxiv : astro - ph/0506043 ; t. koivisto and d. f. mota , phys .
d * 73 * , 083502 ( 2006 ) [ arxiv : astro - ph/0512135 ] ; m. takada , phys .
d * 74 * , 043505 ( 2006 ) [ arxiv : astro - ph/0606533 ] . c. wetterich , astron .
astrophys . * 301 * , 321 ( 1995 ) [ arxiv : hep - th/9408025 ] ; l. amendola , phys .
d * 62 * , 043511 ( 2000 ) [ arxiv : astro - ph/9908023 ] ; m. manera and d. f. mota , mon . not .
soc . * 371 * , 1373 ( 2006 ) [ arxiv : astro - ph/0504519 ] ; r. mainini and s. bonometto , phys .
d * 74 * , 043504 ( 2006 ) [ arxiv : astro - ph/0605621 ] .
j. garriga and v. f. mukhanov , phys .
b * 458 * , 219 ( 1999 ) [ arxiv : hep - th/9904176 ] ; f. perrotta and c. baccigalupi , phys .
d * 65 * , 123505 ( 2002 ) [ arxiv : astro - ph/0201335 ] ; m. kunz and d. sapone , arxiv : astro - ph/0609040 ; t. koivisto and d. f. mota , arxiv : hep - th/0609155 ; k. a. malik , arxiv : astro - ph/0610864 ; r. bean , d. bernat , l. pogosian , a. silvestri and m. trodden , arxiv : astro - ph/0611321 .
i. maor and r. brustein , phys .
d * 67 * , 103508 ( 2003 ) [ arxiv : hep - ph/0209203 ] ; m. sahlen , a. r. liddle and d. parkinson , phys . rev .
d * 72 * , 083511 ( 2005 ) [ arxiv : astro - ph/0506696 ] ; m. sahlen , a. r. liddle and d. parkinson , arxiv : astro - ph/0610812 . | we investigate the clustering properties of a dynamical dark energy component . in a cosmic mix of a pressureless fluid and a light scalar field ,
we follow the linear evolution of spherical matter perturbations .
we find that the scalar field tends to form underdensities in response to the gravitationally collapsing matter .
we thoroughly investigate these voids for a variety of initial conditions , explain the physics behind their formation and consider possible observational implications .
detection of dark energy voids will clearly rule out the cosmological constant as the main source of the present acceleration . |
in the photonics and information science of recent years , much attention is paid to the spin - based , rather than charge - based , electronic systems .
this area of research , nowadays referred to as spintronics , treats spin as a carrier of quantum information and is considered as a highly promising pathway to new information technologies @xcite .
this tendency has arisen a great interest to dynamic properties of spin - systems both in equilibrium and under different kind of perturbations capable of controlling spin state of the system .
the new technique - spin noise spectroscopy , developed in the last decade , refers to the experimental methods intended for studying spin - systems under conditions of thermal equilibrium .
the term spin noise spectroscopy ( sns ) signifies experimental technique that implies spectral investigations of spontaneous fluctuations of spin - system magnetization ( spin polarization ) . until recently ,
this spectroscopy was not considered as a practical instrument of experimental physics .
a few experiments @xcite , in which magnetic resonance was observed experimentally in the spin noise spectrum , were primarily aimed at demonstration of practical feasibility of such an approach .
of course , the fact that spin - system in a magnetic field should display an excess noise at the frequency of magnetic resonance justified by the fluctuation - dissipation theorem ( see , e.g. , @xcite ) was beyond question .
still , fluctuations of this kind , fundamentally small for macroscopic systems , could be either hardly detectable , or of no practical sense .
a certain interest to this technique was first revealed by researchers dealing with fundamental problems of quantum nondemolition measurements ( see , e.g. , @xcite ) .
for the last several years , however , the situation has drastically changed , mainly , due to remarkable advances in the digital systems of data acquisition @xcite . at present , the faraday - rotation - based sns , which is considered to be the most efficient approach to spin noise detection , has found its niche among other methods of the magnetic resonance spectroscopy and gradually turns into a standard experimental technique with a wide and , in many respects , unique range of applications . in this paper
, we briefly outline the main stages in development of this new branch of spectroscopy starting with the first experimental observation of magnetic resonance in the faraday - rotation ( fr ) noise spectrum to the most recent achievements that revealed unique potentialities of this technique both in the rf and optical spectroscopy of paramagnets .
we will consider basic ideas underlying this experimental technique , will discuss technical problems associated with its implementation , and will describe specific properties of the fr - based sns that determine its rich informative potential .
we will also consider recent technical progress in this method of research that has radically changed its place in the contemporary experimental physics .
this paper is intended not to review the experimental measurements on sns , but rather to review properties of the noise and spin - noise spectroscopy that make these measurements possible .
the fr - based sns combines in itself two essentially different methods of research . on the one hand , it can be considered as a branch of the light intensity noise ( lin ) spectroscopy and , on the other , as a specific modification of the optical method of magnetization detection . to make clear the essence of the fr - based sns
, it makes sense to consider these two basic methods in more detail .
the pioneering experimental work on the lin spectroscopy that has arisen a great interest was performed by a. forrester _ et al . _ in 1955 @xcite ( fig .
1 ) . in that work
, the cell with hg vapor placed into a magnetic field was excited by an electrodless microwave discharge , and two zeeman components of the emission line 546.1 nm were detected photoelectrically by means of a specially designed photodetector coupled to a microwave cavity .
the detected spectrum of the light intensity noise was found to contain a peak at the frequency of magnetic splitting of these two components .
the excess noise of the light intensity related to this splitting was less than @xmath0 of its shot - noise level .
the complexity of the experiment was additionally aggravated by high frequency of the detected signal ( @xmath1 hz ) , needed to provide sufficiently high q - value of the noise peak .
still , the signal has been detected , in this experiment , with the signal - to - noise ratio of around 2 , for the accumulation time of 250 s. it should be emphasized that , in spite of the fact that the frequency of the observed peak in the intensity noise spectrum exactly corresponded here to magnetic splitting of the emission line , this experiment did not have anything to do with optical detection of magnetic resonance ( see @xcite and references therein ) .
suffice it to say that the width of this peak was determined by the doppler broadening of these zeeman components which should not be revealed in any way in the magnetic resonance spectrum .
the result of fundamental importance obtained in this paper was that , in the process of photoelectric conversion of the light field , the emission probability for electrons is proportional to the square of the total electric field amplitude arising due to interference between its fourier components . in other words ,
this paper has shown experimentally the possibility of interference between the light fields originating from different incoherent sources .
this conclusion , at that time , was far from trivial , and , as the authors of @xcite wrote , `` many physicists found it contradictory to their ideas about the nature of interference '' .
the discussion raised by publication @xcite was further heated by another important experiment in the field of noise spectroscopy .
the experiment in point was performed and interpreted by hunbury - brown and twiss in 1956 @xcite .
now , the measurements were focused on detecting spatial correlations in the intensity noise of a remote thermal source and were aimed at application of this effect to evaluating angular dimensions of stars .
the authors took advantage of the fact that emission of a remote thermal source of finite sixe is characterized by a certain spatial scale of the light coherence controlled by angular size of the source . indeed , if we consider intensity relief created by such a source with the diameter @xmath2 located at a distance @xmath3 from the plane of observation ( fig .
2 ) , we will see that it comprises a multitude of bright and dark spots .
the brightness created at each particular point depends on whether the resultant interference of the waves from all emitters of the source is constructive or destructive .
these spots are very pronounced in monochromatic laser fields and are called _
speckles_. the size of a speckle is determined by the distance at which the total phase difference between the rays providing constructive or destructive interference changes by @xmath4 radians . in our case , the average size of the spot is given by @xmath5 , where @xmath6 is the light wavelength .
thus , we see that the angular size of the light source in this geometry can be directly estimated from the average size of speckles .
hunbury - brown and twiss performed their experiment with a star as a light source and evidently could not examine instantaneous relief of brightness over a large area of the illuminated surface .
so , the authors made use of the intrinsic intensity noise of the light emitted by the incoherent source .
these temporal fluctuations had to be correlated or not , in two spatial points of the observation plane , depending on whether these points are closer to each other than the size of the spot or not .
thus , by measuring correlation between photocurrents of two detectors placed at a distance of @xmath7 from each other as a function of @xmath7 one can estimate the average size of the speckles ( of the _ coherence area _ ) and to find angular size of the source .
the stellar interferometer created on the basis of this effect ( the _ intensity interferometer _ ) was successfully used in astrophysical measurements .
the experiments @xcite and @xcite were the earliest and the most famous that have shown that noise in optics may be useful and informative .
note that these measurements were aimed at studying properties of the light , rather than properties of a medium . at the same time , in 60s-70s , there have been performed a considerable number of experiments in which the intensity - noise - based technique ( yet without lasers ) was used for studying dynamic properties of atomic systems .
highly important results of these studies were related to possibility of manifestation of the excited state dynamics and fine energy structure in the noise spectrum of spontaneous emission of the system .
the noise of the light transmitted by an absorbing atomic medium was employed for measuring diffusion parameters of atoms in gaseous medium .
it should be noted that all these experiments , in the pre - laser era , were highly labor - consuming . in more detail , this story is considered in the review @xcite .
the situation has changed dramatically , however , with the advent of lasers . due to extremely high luminosity of laser emission
, the spectral power density of the light intensity modulation in these sources at the beat frequencies could exceed the background shot - noise level by many orders of magnitude , and the lin spectroscopy has gained a much deeper sense .
it became clear that the lin spectroscopy could be used not only for studying spectral and correlation characteristics of the light source proper ( like , e.g. , the laser output mode structure ) , but also for investigating dynamic properties of the media interacting with the light .
specifically , the lin spectroscopy , referred to as dynamic light scattering ( dls ) or photon correlation spectroscopy ( pcs ) , has found application for studying morphological and dynamic properties of suspensions , solutions of macromolecules and polymers , liquid crystals , biological solutions and microorganisms .
this technique may be efficiently applied to ensembles of particles with a wide range of dimensions ( from 0.001 @xmath8 m to several @xmath8 m ) , inaccessible for other methods ( see , e.g. , @xcite ) .
it is interesting to compare the spectroscopy of lin with the conventional optical spectroscopy ( which also inevitably deals with fluctuating optical fields ) .
both methods are aimed at getting information about spectral characteristics of the optical field @xmath9 oscillating at about @xmath10 hz by measuring its _ intensity _
@xmath11 ( in reality , by measuring photocurrent or photocharge ) , since we can not directly measure the field amplitude @xmath9 . in the standard optical spectroscopy , which implies measuring the spectrum of the field at optical frequencies and , therefore ,
may be considered as the spectroscopy of optical field , the procedure of spectral decomposition of the light field precedes measurement of the light intensity . as a result , the intensity of each detected spectral component retains information about correlation properties of the optical field which is connected with its spectrum , according to the wiener - khinchin theorem , through the relationship @xmath12 in other words , optical spectroscopy allows one , in this way , to get information about correlation properties of the field in the range of @xmath13 s without photodetectors with so high temporal resolution . in the lin spectroscopy , the experimental arrangement is inverted : the light field is first converted into a photocurrent ( thus losing its carrier frequency ) and then spectrally analyzed in the electronic channel of the detection system . in this case , the power spectrum of the signal under study is determined by correlation characteristics of the _ light intensity _ @xmath14 ( rather than light field @xmath9 ) @xmath15 and does not contain , in an explicit form , information about correlation properties of the field at optical frequencies . in the lin spectroscopy , the scale of the frequencies of interest is limited by the bandwidth of the detection system , and the lin spectrum , therefore , can contain information only about relatively fine features of the optical spectrum .
this fundamental distinction between the two spectroscopic techniques is schematically illustrated by fig .
3 . the sns , as a sort of lin spectroscopy
, evidently implies the second type of the measurements , with the photodetection preceding spectral decomposition of the signal .
at the same time , combination of the two types of spectroscopy , with preliminary spectral decomposition of the light field , as we will see below , may provide valuable additional information about the system . in some cases ,
the _ light intensity spectrum _ may provide information about the light field inaccessible for conventional optical spectroscopy .
the matter is that the light intensity variations , as was already mentioned , are caused , in fact , by the beats between different spectral components of the optical field and , therefore , are able to reveal the phase correlation between the components not seen or not resolved in the optical spectrum .
the simplest example may be given by a broadband ( or `` white '' ) light with harmonically modulated intensity .
this light reveals , in its intensity spectrum , a sharp peak at the modulation frequency , while its optical spectrum remains practically unperturbed by the modulation .
this informative capability of the light - intensity spectroscopy constitutes the basis for spectroscopy of superhigh - resolution .
this spectroscopy , developed in the 70s , made it possible , in particular , to realize sub - doppler resolution in optical spectroscopy of atomic systems and to study fundamental phenomena of interference of quantum states hidden in the inhomogeneously broadened optical transitions ( @xcite ) .
the fr - based spin noise spectroscopy ( sns ) , we will talk about , can be regarded as a polarization version of the lin spectroscopy , with the light field fluctuations provided by spontaneous noise of the spin - system magnetization .
possibility of conversion of the magnetization fluctuations to those of the light polarization is determined by direct relationship between the faraday rotation and magnetization of the spin - system , constituting the basis of the optical method of magnetic measurements ( see , e.g. , @xcite ) .
optical methods of detecting spin - system magnetization ( spin polarization ) employ the fact that magnetization of a paramagnet affects its optical properties .
the so - called `` paramagnetic '' fr that directly reflects magnetization of the spin - system is formed in the following way ( fig.4 ) .
the difference between populations of the ground - state magnetic sublevels of the paramagnet ( of the spin - system energy levels ) creates , for the light propagating along the field , corresponding difference in the optical absorption for the two circular polarized transitions from these unequally populated states ( _ magnetic circular dichroism _ , mcd ) .
this dichroism is , evidently , observed only in the region of optical absorption and manifests itself in the form of ellipticity acquired by the linearly polarized light transmitted through the medium .
the same kind of optical anisotropy , in conformity with the kramers - kronig relations , is observed in difference of the refractive indices for the two circular polarizations ( _ magnetic circular birefringence _ ) and is revealed as rotation of the polarization plane of the linearly polarized light traveling through the medium ( _ the faraday effect _ ) . and
@xmath16 ) and refractive indices ( @xmath17 and @xmath18 ) for two circular polarizations become different .
these differences ( @xmath19 and @xmath20 ) give rise to the magnetic circular dichroism and faraday rotation , respectively .
magnetic splitting of the transition energies is supposed negligible compared to the line width . , width=377 ] in this description , we ignored magnetic splitting of the transitions into circularly polarized components of different handedness .
this contribution to the magneto - optical activity of the medium ( usually referred to as `` diamagnetic '' ) is not related to spin - system magnetization and is irrelevant to our consideration .
thus , magnetization of a paramagnet can be measured by detecting either circular dichroism or faraday rotation .
an essential difference between these two methods is that the method of dichroism implies inevitable optical excitation of the sample and , therefore , can not be nonperturbative , whereas detection of the faraday effect may be performed in the region of transparency without producing any real optical transitions .
in addition , the mcd - based measurements of spin - system magnetization may be performed in a pure photometric way be measuring intensity of the transmitted circularly polarized light , while the faraday rotation in the transparency region does not imply ant changes in the light intensity and requires the use of polarimetric technique .
it is also worth to note that the faraday rotation , in the context of our treatment , is not necessarily to be observed in the longitudinal magnetic field , as implies its classical definition .
it is supposed that the fr detects spin polarization along the light propagation direction regardless of the direction ( and even of the presence ) of the external magnetic field . in particular , the fr can be used to detect oscillation of transverse magnetization of the spin - system under conditions of its coherent precession and , thus , to optically detect its magnetic resonance .
this method was first proposed and realized in @xcite .
application of magnetooptics to studying spin dynamics is known since 50s of the last century .
the fr and mcd effects were used , in particular , for detecting magnetic resonance in atomic systems polarized by optical pumping . in those experiments ,
the magnetic resonance was revealed either as a suppression of longitudinal magnetization ( mcd or fr ) for the light propagating along the field ) or as oscillations of the transverse magnetization ( mcd or fr for the light propagating across the field ) @xcite .
this magneto - optical technique was also used for studying dynamic properties and energy structure of transparent paramagnets by measuring their magnetic susceptibility at subresonant frequencies @xcite .
it is important that , in all these experiments , the detected optical signal was a result from _ coherent response _ of the spin - system to a modulated ( or , at least , time dependent ) perturbation ( microwave pumping or external magnetic field ) .
the idea of the spin noise spectroscopy was to detect intrinsic fluctuations of the spin - system magnetization under conditions of thermal equilibrium without any regular external perturbation .
since any real spin - system consists of a finite number of spins which perform incessant random motion ( maintaining thermal equilibrium of the system ) , its magnetization should inevitably exhibit spontaneous fluctuations due to deviations of instantaneous values of the magnetization from its mean value ( fig.5 ) . of a spin - system in a static external magnetic field @xmath21 , due to permanent motion of individual apins ,
exhibits random fluctuations both in its magnitude ( along the magnetizatio @xmath22 ) and in its direction ( across the mean magnetization).,width=302 ] spectrum of these random fluctuations and their correlation times should be directly connected with dynamic parameters of this motion ( characteristic frequencies and correlation times ) .
these dynamic characteristics , in principle , can be extracted from results of the noise measurements .
of course , the notion of thermodynamic fluctuations had been known long before the late 70 s , when the experiment @xcite was planned , and the problem was more of practical , rather than fundamental nature : it was clear that these fluctuations , in a macroscopic spin - system , should be extremely small , and the question was whether it was possible to detect them experimentally for a reasonable accumulation time with a sufficiently high signal - to - noise ratio and , after all , whether it had any practical sense .
conventional magneto - optical measurements imply two main geometries of the measurements , with the light beam propagating , respectively , along and across the magnetic field .
these geometries are commonly referred to as the faraday ( longitudinal ) and voigt ( transverse ) configurations .
standard magneto - optical effects in a static magnetic field , observed in these configurations , in conformity with the symmetry of the problem , correspond to the cases of magnetic - field - induced circular and linear anisotropy .
this is not the case , however , for fluctuations of the magneto - optical anisotropy , which may break the symmetry of the problem . the simplest experimental geometry for measuring the fr noise , which may seem the most natural , implies detection of _ longitudinal _ fluctuations of the magnetization , with the probe laser beam aligned along the external magnetic field and along the equilibrium magnetization of the spin - system ( faraday configuration , fig . 6 , left panel ) .
in this case , the detected fluctuations of the magnetization do not break axial symmetry of the system .
dynamics of these fluctuations is controlled by the only relevant characteristic time - longitudinal spin relaxation time @xmath23 , which determines the correlation and spectral characteristics of this random process . in the time domain , the longitudinal fr noise ( fig .
6,b , left panel ) looks as a `` white '' noise with the removed higher frequencies ( exceeding the relaxation rate @xmath24 ) .
the autocorrelation function , correspondingly , acquires exponential shape ( rather than @xmath25-wise , as for the `` white '' noise ) with the characteristic time @xmath23 ( fig .
6,c , left panel ) , while the power spectrum of the process , given by fourier transform of the autocorrelation function , is a lorentzian centered at zero frequency with the width determined by the relaxation rate @xmath24 ( fig . 6,d , left panel ) .
thus , by measuring the fr noise spectrum in the faraday geometry , we can obtain information about time @xmath23 , its dependence on magnetic field , temperature , and other external parameters , i.e. , about properties of the system usually obtained from the data of epr spectroscopy or from nonresonant magnetic measurements in oscillating magnetic fields .
it is worth to note that this experimental approach has much in common with the paramagnetic relaxation technique which has been developed by the dutch physicist c. gorter @xcite for measuring magnetic susceptibility of paramagnets in parallel fields and was later transferred to optical basis with the aid of laser polarimetry @xcite . in gorter
s technique , dynamics of the spin - system is studied by measuring its linear response to a rf magnetic field applied along the dc field .
the essential difference between these two approaches is that in the case of noise spectroscopy , the system is not supposed to be perturbed by the external oscillating field . in terms of the fluctuation - dissipation theorem ,
the longitudinal magnetization noise ( or the fr moise ) is the counterpart of the linear magnetic ( or magneto - optic)response observed in parallel fields . in this longitudinal geometry
, however , the noise spectrum is located in the vicinity of zero frequencies where the measurements are often hampered by the universal @xmath26 ( flicker ) noise .
in addition , spectrum of the longitudinal fluctuations of magnetization , bringing much data about relaxation characteristics of the system , does not contain any information about its magnetic splitting , which is highly important for characterization of the spin - system .
the above drawbacks of the faraday configuration can be easily overcome by passing to the voigt geometry ( fig .
6 , right panel ) . in this cobfifuration
, the light beam traveling through the paramagnet across the applied magnetic field will detect transverse magnetization of the spin - system , which , by symmetry considerations , should vanish in this geometry and may arise only due to spontaneous violation of the symmetry by fluctuations .
in this configuration , any random transverse fluctuation of magnetization will precess around the magnetic field direction during the transverse relaxation time @xmath27 ( or what is usually called @xmath28 , when the spin - system is inhomogeneous ) and then will be replaced by another realization of the transverse fluctuating magnetization with another magnitude and another phase of precession . as a result , the probe light beam will exhibit randomly oscillating fr at larmor frequency ( fig.6,b , right panel ) .
correlation function of this random process will now have the shape of oscillatory decay with the characteristic time @xmath29 ( fig .
6,c , right panel ) .
the peak of the fr noise spectrum will now be shifted away from zero frequency , and the width of the peak will be determined by the dephasing time @xmath27 ( or @xmath28 ) of the spin - system ( fig .
6,d , right panel ) . in other words ,
the fr noise spectrum will present , in this geometry , the _ magnetic resonance spectrum _ of the spin - system .
the idea of detecting magnetic resonance in the noise of the spin - system magnetization was first mentioned by f. bloch @xcite in 1946 .
an essential contribution to understanding of optical manifestations of spin - system magnetization , spin precession , and spin dynamics and , implicitly , to optical detection of magnetization noise , was made by a. kastler in his works on optical pumping @xcite .
the fr noise spectroscopy is closely related to the time resolved faraday rotation @xcite and may be considered as its incoherent version .
a specific feature of the sns is that it is , in fact , not a _ spectroscopy of response _ , and , therefore , the magnitude of the detected signal ( _ noise signal _ ) can not be controlled by varying the strength of the perturbation .
in addition , magnitude of the magnetization ( and fr ) noise , for a macroscopic sample , as was already mentioned , should be extremely small as compared with the values of `` coherent '' magnetization ( or magnetization of saturation ) induced , by the external perturbation acting in the same way upon all the spins of the ensemble .
thus , it is clear that the problem of polarimetric sensitivity may have a critical importance for the spin noise spectroscopy .
the sensitivity of polarimetric measurements in optics is known to be fundamentally limited by the so - called photon noise or shot noise of the detector photocurrent in accordance with the relationship ( see , e.g. , @xcite ) @xmath30 ( @xmath31 is the angle of the polarization plane rotation detected with the signal - to - noise ratio equal to unity , @xmath32 is the bandwidth of the detection channel , @xmath33 is the light intensity measured in the number of photons per second , is supposed to be given in photons per second . ] and @xmath34 is the quantum yield of the photodetector ) .
for instance , for the light power 20 mw , wavelength 550 nm , and bandwidth 1 hz , this quantity lies in the range of @xmath35 rad .
such a sensitivity , as may be shown by appropriate estimates , is high enough to solve , in many cases , the problem of spin noise detection .
the most frequent reason why the shot - noise - limited sensitivity can not be achieved with the lasers ( which are considered , in this context , the only suitable light sources ) is related to their excess noise , which may exceed the shot - noise level by a few orders of magnitude .
the excess noise can be suppressed using different expedients @xcite , the simplest and most efficient of them being _ balanced detector _ ( fig.7,a ) .
( right side ) .
( a ) standard @xmath36-geometry . at @xmath37 ,
the photocurrents become equal , and the current flowing through the resistor @xmath38 vanishes for any light intensity .
( b ) the scheme with variable polarization extinction .
the beam is split by a non - polarizing beamsplitter ( npbs ) , retaining polarization of the light beam , and the level of polarization extinction is set in each arm independently ( using polarizers @xmath39 and @xmath40 ) to equalize photocurrents . at high extinction
( at small angles of detuning @xmath25 from the crossed position ) , the steepness of the dependence @xmath41 may become much higher .
pd1 and pd2 - photodetectors , p1 and p2 - polarizers .
, width=453 ] the linearly polarized light beam whose polarization behavior is examined passes through a polarization beamsplitter with its two outputs coupled to two photodetectors . the photodetectors are included into a differential circuit so that their photocurrents are subtracted at the load resistor @xmath38 .
when the polarization plane of the incident light makes an angle of @xmath36 with polarizing directions of the beamsplitter , photocurrents of the two detectors cancel at the resistor ( regardless of the light intensity and its variations ) . at the same time , the changes of the photocurrent related to variations of the light beam polarization plane are always anticorrelated ( have opposite signs ) and , as a result , are summed up . in this way
, it is possible , in practice , to suppress the excess intensity noise by approximately three orders of magnitude and to achieve the shot - noise limit of polarimetric sensitivity with noisy laser sources .
for the first time , as far as we know , the shot - noise - limited polarimetric sensitivity has been realized ( with the use of the balanced detector ) , in @xcite .
nowadays , the balanced detectors are produced commercially and , in combination with polarization beamsplitters , are commonly used in the high - sensitive polarimeters .
there are some other interesting methods of the excess noise suppression based on the fact that the shot - noise - limited polarimetric sensitivity does not substantially change when the angle between the analyzer and polarization plane varies from @xmath42 to total extinction .
indeed , when moving towards the total extinction ( let it be @xmath43 ) , the polarimetric response ( @xmath44 ) varies approximately linearly with @xmath45 ( @xmath46 ) .
at the same time , the transmitted light intensity at small @xmath45 varies as @xmath47 ( @xmath48 ) providing the shot noise ( varying as @xmath49 ) also proportional to @xmath45 .
thus , the signal - to - noise ratio remains in this range of angles @xmath45 practically the same ( to within the factor @xmath50 ) . in reality , however , this conclusion is violated at small @xmath45 either because of nonideality of the polarization system ( the extinction ratio of the polarizers usually can not be smaller than 10@xmath51 10@xmath52 , or when the detected light intensity noise sinks in the noise of electronics . in the absence of excess fluctuations ,
capable of reducing polarimetric sensitivity , the high - extinction polarization geometries can still be used for other purpose , namely , to raise the probe beam intensity ( and thus to increase the shot - noise - limited polarimetric sensitivity ) leaving the input light power on the detectors at low level .
this is especially important for the sns , which employs broadband photodetectors with small photosensitive area unable to endure high input power .
one of the methods that allows one to realize the high - extinction polarization geometry and , at the same time , to retain opportunity to further suppress the excess intensity noise is based on the use of polarization pile , which serves , in this case , as a dichroic medium with no birefringence .
as has been shown in @xcite , the polarization pile makes it possible to considerably magnify the polarization plane rotation angle at the expense of reduction of the beam intensity .
this gadget can also be used for choosing favorable ratio between intensity of the light passing through the sample and the input light power of the photodetector with no loss in sensitivity . among other high - extinction polarization geometries
can be mentioned the scheme with polarization - insensitive beamsplitter and independent adjustments of polarization extinction in two channels ( fig.7,b ) .
this method has all the merits of the polarization pile , but can be easier realized experimentally . as applied to measurements of the spin - system magnetization noise ,
there are certain additional requirements that should be met .
in particular , the wavelength of the probe laser beam should provide the most efficient conversion of the magnetization noise to that of the fr ( this wavelength does not necessarily coincide with the region of the greatest verdet constant of the system , see below ) .
the other requirement is related to maximizing fr per a single spin , which can be achieved by reducing cross section of the probed volume and increasing its length .
indeed , let we probe a paramagnet with the length @xmath53 and spin density @xmath54 by the light beam with the cross section @xmath55 ( all quantum systems of the ensemble are supposed to be identical ) .
then , the number of spins confined in the beam will be @xmath56 .
fluctuations of the faraday rotation at the exit of the paramagnet will be proportional to fluctuations of this quantity ( @xmath57 ) and to the total , independent of @xmath55 , faraday rotation ( @xmath58 ) . as a result , for the fr noise , we have @xmath59 so , for a given paramagnetic sample , the fr noise varies in inverse proportion with the radius of the light spot @xcite .
this is a highly important feature of the sns , which implies that the spin noise ( with all other factors being the same ) _ increases _ with decreasing number of spins .
this is why , in the sns , the probe laser beam passing through the sample should be preferably tightly focused .
note that , using the light beam for detecting spin fluctuations , we probe a fluctuating medium ( spin - system ) by a fluctuating agent ( photon flux ) and measure intensity fluctuations of the light on the background of its own shot noise @xcite .
it may seem , at first sight , that the spin noise can be detected only when the noise introduced by the spin - system exceeds ( or is comparable with ) the shot noise of the probe beam . in reality , however , this is not the case .
as can be shown , the mean - square level of poissonian noise transmitted through a lorentzian filter with the bandwidth @xmath60 for the accumulation time @xmath61 can be measured with a relative accuracy of @xmath62^{1/2}$ ] ( @xmath63 is the photocurrent and @xmath64 is the electron charge ) , which turns into the known factor @xmath65 at sufficiently large photocurrents ( @xmath66 ) @xcite .
so , one can easily estimate that , under real experimental conditions and for sufficiently large accumulation times , the factor @xmath67 may reach many orders of magnitude . as a result , the shot - noise power appears to be defined with a fairly high accuracy and , in the form of a stable spectral background , does not preclude measuring intensity fluctuations whose power lies essentially below the shot noise level .
the first experiment on detection of magnetic resonance in the fr noise spectrum @xcite was performed on sodium atoms in the atmosphere of buffer gas ( neon , @xmath68 10 torr ) .
schematic of the experimental setup is shown in fig.8 . as a light source
, we used a cw dye laser tuned to close vicinity of the d1 or d2 absorption line of na .
the cell with sodium vapor was placed into a transverse magnetic field created by a pair of helmholtz coils .
the field was slightly modulated at a frequency of @xmath69 hz .
the differential signal of the balanced photodetector was first filtered by a resonant circuit at a frequency of 1.3 mhz , then selectively amplified , quadratically detected ( rectified ) , lock - in amplified at the field modulation frequency , and recorded as a function of the applied magnetic field . in other words , the magnetic resonance frequency was swept with respect to resonance frequency of the photodetection system ( much like the resonance frequency is swept with respect to frequency of the rf or microwave oscillator in conventional epr spectroscopy ) .
figure 9 shows two typical plots of the detected spin - noise resonance .
the probe beam was tuned to the edge of d1 line or to the midpoint between the two lines .
polarimetric signal of the fr noise , in this experiment , substantially exceeded the shot noise level , and the signal - to - noise ratio ( or , better to say , the ratio of useful noise to harmful noise ) for the accumulation time 2 s , was close to 100:1 .
this was the first experimental demonstration of feasibility of the spin noise spectroscopy . in 1985
, there has been performed the experiment @xcite on nuclear spin noise , which did not have anything to do with the fr technique , but ideologically was identical to the experiment @xcite . as a spin - system , in that experiment , was used an ensemble of nuclei @xmath70cl in the naclo@xmath71 crystal .
the experiment was performed at 4 k , and the magnetization noise was detected in a straightforward way using a squid sensor .
this work has considerably contributed to the sns as a first observation of the resonant spin noise in a nuclear spin - system predicted by f. bloch @xcite . here , the number of spins contributed to the signal was much greater than in the experiment with electron spins @xcite , and the noise signal was expected to be extremely small .
indeed , the peak of spin noise at the frequency of nuclear quadrupole resonance of @xmath70cl was detected with the signal - to - noise ratio of a few units for the accumulation time of 7 hrs .
still , further technical advances in this area made it possible to observe the nuclear spin noise at room temperature @xcite and even to develop , on its basis , an efficient method of nuclear spin - noise imaging @xcite .
this technique , though less sensitive than the conventional method of nmr imaging , may be indispensable , in certain cases , as an entirely noninvasive tomography that does not use any external rf or x - ray fields .
the effect of magnetic resonance in the fr noise spectrum remained practically unnoticed , as a possible experimental tool of magnetic - resonance spectroscopy , until the verge of our century , when , with some modifications , it was reproduced on other alkaline atoms and then applied to semiconductor systems .
below , we will briefly outline the main steps in the development of the experimental sns . .
in 2000 , an interesting experiment was performed by t. mitsui @xcite , who independently observed optically detected spontaneous spin noise of rubidium atoms at the larmor frequency .
the idea of the experiment was the same , but realized differently : the magnetization noise was detected in the optical absorption , rather than in the faraday rotation , of the atomic vapor .
a circularly polarized beam of a diode laser , tuned in resonance with the d1 absorption linea of @xmath72rb ( 5 @xmath73s@xmath74 - @xmath73p@xmath74 ) was transmitted through the cell with atomic vapor at 80@xmath75c placed into a transverse magnetic field . to simplify qualitative analysis of the results ,
the transition was strongly broadened ( up to 6 ghz ) by a buffer gas ( nitrogen ) at 200 torr .
fluctuations of the spin - system magnetization , in this experiment , were observed in the intensity ( rather than fr ) noise spectrum of the transmitted light .
a specific feature of this approach was that the excess intensity noise of the light source , in this case , should have been eliminated before the sample , rather than suppressed after it , because , otherwise , it could induce nonlinear effects undesirable in this experiment ( they were studied independently ) .
the laser source used for these measurements was characterized by extremely low excess intensity fluctuations , and , therefore , there was no need to employ balanced detector for their suppression . in that experiment ,
fluctuations of the spin - system magnetization were , in fact , detected by the noise of the mcd ( rather than fr ) of the paramagnet . from the viewpoint of linear optics , these two quantities ( mcd and fr ) ,
being connected through the kramers - kronig relations , are identical and should give the same results unless the optical excitation affects spin dynamics of the system .
at the same time , in the mcd - based measurements , each act of absorption interrupts spin precession of the atom , and it may seem that the later absorption events can not be correlated , in any way , with the previous .
the author has shown , however , that this is not the case for a thermodynamically fluctuating spin - system , and the mcd - based method of the spin noise detection , which looks more perturbative than the one based on the fr , is also applicable to macroscopic systems .
it is worth to note here that the question about possibility of detecting superposition states of a quantum system ( as , e.g. , a precessing spin ) in the noise spectrum of electromagnetic field is not trivial .
in particular , it has been shown in @xcite that the spontaneous emission noise spectrum of an atomic medium excited randomly , in a poissonian way , does not reveal any features related to its decay kinetics .
it means that the intensity fluctuations of spontaneous emission excited stationary in this way can not be used for detecting spin noise in the excited state of the atoms .
a rigorous quantum - mechanical description of the intensity - fluctuation spectrum of spontaneous emission is given in @xcite . in 2004 ,
s. crooker et al .
@xcite examined in more detail the spin - noise spectra of rb and @xmath76k atoms using the fr - based method and provided a deeper insight into capabilities of this technique .
it was shown , in particular , that magnetic resonance in the fr noise spectrum can be also observed in the faraday configuration ( in longitudinal magnetic field ) .
this is possible , in conformity with the fluctuation - dissipation theorem , for the esr transitions allowed in the ac magnetic field oscillating along the external dc field .
an illustration of such a situation was presented in @xcite , where the resonances of this type were originated from transitions with @xmath77 ( @xmath78 is the projection of the total angular momentum of the state ) in @xmath76k atoms .
one of the questions that arose after successful detection of spin noise in atomic gases , with strongly off - resonant probing of the system , was whether this technique could be considered as nonperturbative or not . at first glance
, it looked fairly passive or nonperturbative because it utilized faraday rotation in the transparency region , the probe beam did not induce any real electronic transitions and , therefore , did not disturb the spin - system under study . on the other hand , however , it could not be nonperturbative for evident reasons : the initially monochromatic probe light , after passing through the medium , became modulated and thus acquired sidebands with shifted frequencies .
it means that some photons of the beam experienced inelastic interaction with the medium inevitably accompanied by some energy exchange between the light and the spin - system .
it is also known that the effect of transparent medium on polarization of the light beam is accompanied by the back action of the light upon the medium even in static magnetic field @xcite .
this contradiction was resolved by gorbovitskii and perel @xcite who have shown that the noise of the fr at the frequency of larmor precession can be considered as a result of _ coherent forward raman scattering _ of the probe laser light by the ensemble of spins randomly precessing in the transverse magnetic field .
the oscillating polarization in the probe beam arises dues to interference of the forward - scattered light with the light directly transmitted by the medium .
this effect , according to @xcite , is analogous to the effect of optical mixing used in @xcite for studying brilloin scattering .
so , this process can be regarded as nonperturbative , because the probe light interacts with macroscopic fluctuations of the system , and does not select any particular spin to flip .
such a conclusion is supported by the possibility to observe magnetic resonance in the mcd noise , when each absorption event explicitly destroys coherent superposition of the ground - state sublevels .
the technique evidently becomes more and more perturbative as we pass from macroscopic ensembles to microscopic ones .
note that in semiconductors , which are more complicated than pure atoms and can not be treated in the framework of so simple models , a small residual absorption may dramatically affect properties of the spin - system under study even in the regions of nominal optical transparency @xcite .
the atomic systems , mainly alkali atoms , as the most convenient model objects of sns , have been widely used in fundamental research related to non - demolition measurements and interaction between squeezed states of light and matter .
it was shown , in particular , that the spin - noise measurements performed in the region of optical transparency make it possible to surpass the standard quantum limit of phase measurements and thus to produce the squeezed atomic spin states ( see , e.g. , @xcite ) .
this interesting topic and associated experimental results , however , lie outside the scope of this review .
it is worth to mention here another experimental approach to the optical detection of incoherent spin precession , usually also regarded as a sort of noise spectroscopy @xcite .
this approach is based on transformation of modulation spectrum of the light transmitted through a paramagnet in the voigt geometry . in this case
, the light beam serves simultaneously as a probe and as a pump , and its modulation ( either in intensity or in polarization ) is supposed to be equivalent to modulation of the effective magnetic field applied along the light propagation .
this can be achieved either by intensity modulation of a circularly polarized beam or by modulating the degree of its circular polarization .
experimental setup for these measurements may look as shown in fig .
10 .
it is well known that the coherent larmor precession of a spin - system can be induced optically by the resonant pump light modulated at the appropriate frequency .
the first experimental observations of the optically driven spin precession were announced in @xcite .
the effect was detected in the vapors of alkali metals ( cs and rb ) and in metastable helium .
the cell was resonantly pumped by circularly polarized light propagating across the applied magnetic field and modulated in intensity at the frequency close to that of larmor precession of the spin - system .
the resonance was observed as a peak in the transmitted light intensity at the point where the modulation frequency coincided with that of the ground - state zeeman splitting .
this effect of optically driven spin precession , often referred to as ` optical orientation in the rotating coordinate frame ' or as a ` resonance of coherence ' , is very close , in its physical content , to the effects of coherent population trapping @xcite and electromagnetically induced transparency @xcite .
all these effects imply excitation of coherent superposition of two states ( in our case - of two low - lying states ) by superposition of two optical fields ( or by a single modulated field , which is practically the same ) .
the `` active '' spin noise spectroscopy exploits the same ideal , but uses , for this purpose , the pump modulated by more or less `` white '' noise ( rather than harmonically ) . under these conditions ,
the system is offered , so to say , to choose by its own the frequency component capable of inducing its larmor precession .
this approach utilizes the effects of nonlinear optics and is evidently essentially perturbative .
regularities of the `` active '' sns strongly differ from those of the fr - based spectroscopy of spontaneous spin noise . at the same time , in certain respects , it can be more convenient in practice and may provide additional information about the system not related directly to its magnetic properties . from the viewpoint of experimentalist , this type of noise spectroscopy is , in many respects , the exact opposite of the fr - based sns : it does not imply high polarimetric sensitivity and the laser intensity noise should be well pronounced or even purposely increased , rather than suppressed . the laser beam is not supposed to be tightly focused on the sample to reduce the number of spins participating in formation of the signal . at the same time , the light power density on the sample , in these measurements , should be high enough to provide the required optical nonlinearity . as has been shown in @xcite ,
under sufficiently high power densities , when the rabi frequency becomes comparable with relaxation rates of the excited state , the intensity noise spectrum of the transmitted light becomes much more complicated , with its peaks and singularities not connected in a straightforward way with the zeeman and rabi frequencies of the system .
this technique of noise spectroscopy , as far as we know , was applied so far only to atomic systems . as the noisy light sources , in those experiments ,
were used diode lasers with the frequency - modulated output emission which was converted into the intensity - modulated light in the process of its interaction with narrow spectral features of the sample under study . meanwhile
, the `` active '' noise spectroscopy , in our opinion , is a promising method of research that may find application in experimental studies of solid - state ( including semiconductor ) systems highly important for the up - to - date photonics and information science .
the experimental setups , with the laser sources intentionally modulated in intensity or in polarization in a broad frequency range may be useful in the cases when the nonperturbative nature of the sns is not of primary importance .
it should be noted that this method of nonlinear optics may provide information inaccessible to the conventional sns and lying far beyond the bounds of the field of magnetic resonance and spin dynamics .
atomic gases seemed to be highly favorable objects of the sns due to their intense and narrow optical resonances providing strong peaks of fr in the vicinity of the lines ( the so - called macaluso - corbino effect ) .
semiconductors and other solid - state systems with their broader optical spectra did not look so promising .
still , in 2005 , oestreich et al .
@xcite have managed to successfully apply this technique to a solid semiconductor system .
the measurements were performed on a thick @xmath79-doped gaas wafer at 10 k with the wavelength of the probe laser @xmath80 nm below the gaas bandgap .
the level of si - doping ( @xmath81 @xmath82 ) was chosen corresponding to the longest spin dephasing time measured using the femtosecond time - resolved faraday rotation technique @xcite .
the fr noise of the sample placed into a transverse magnetic field was detected with a broadband balanced photoreceiver and conventional sweeping spectrum analyzer in the frequency range of 200 - 400 mhz .
behavior of the discovered peaks well correlated with the known magnetic - resonance and relaxation properties of the donor - bound electrons in @xmath79-gaas .
this work was , at that moment , more of fundamental than practical significance since the accumulation time needed to reliably detect the noise signal lied in the range of several hours .
still , in these experiments , the authors have successfully used , for the first time , the sns technique for measuring the electron lande @xmath83-factor and electron spin relaxation time in a semiconductor ( @xmath79-gaas ) and have demonstrated applicability of sns to semiconductor systems , thus laying the foundation for the semiconductor sns ( see the review @xcite ) .
the `` academic '' period of the sns was completed when it became possible to perform these measurements for much shorter times in a wider frequency range on system of greater practical importance .
a real breakthrough in this field of research was made when the sweeping spectrum analyzer , in the system of data acquisition , has been replaced by the one with the real - time fast fourier transform ( fft ) processing system with a wide frequency range @xcite . the conventional sweeping spectrum analyzer , as is known , measures the signal only in a narrow frequency interval at a time and , therefore , constantly disregards most information containing in the time - dependent signal .
the fft spectrum analyzer , on the contrary , is capable of using the whole bulk of incoming information .
it digitizes the signal in the whole bandwidth of the system ( with a sampling rate of around @xmath84 s@xmath85 ) , performs fft of the signal _ in real time _ , and accumulates the spectrum thus obtained . in other words
, it accumulates signals in all frequency channels simultaneously , rather than only in a single channel . as a result
, the accumulation time needed to achieve the same signal - to - noise ratio has decreased by a few orders of magnitude . the process of accumulation of the signal is schematically illustrated by fig . 11 .
this technical advancement has turned the sns into a real practical tool of magnetic spectroscopy and made it suitable , in particular , for studying spin dynamics of low - dimensional semiconductor systems ( quantum wells , quantum dots , quantum wires ) highly important for the present - day applications in photonics and optoelectronics .
another interesting idea was proposed by mller et al .
@xcite to overcome bandwidth limitations of the optical detectors . for that purpose
, it was proposed to use a pulsed laser with a high repetition rate ( e.g. , a mode - locked laser ) instead of a cw laser , as a source of the probe light . in this case , the intensity spectrum of the light comprises a comb of discrete lines spaced by the pulse repetition rate @xmath86 , and it becomes possible to observe , in the spectrum of the detected fr noise signal , not only the peak at the frequency of magnetic resonance @xmath87 , but also the peaks at the frequencies shifted from @xmath87 towards lower frequency by multiples of the repetition rate @xmath86 ( provided that @xmath88 ) . as a result , by mixing the resonance signal with the nearest peak of the probe beam intensity spectrum , the frequency of the spin noise resonance can be transferred to the frequency range @xmath89 $ ] .
this experimental approach allowed the authors to detect spin noise at larmor frequencies up to 16 ghz . under these conditions , the total bandwidth of the detection system ( including spectrum analyzer )
may not exceed the repetition rate @xmath86 . at the same time , the measured spin dephasing rates are limited , in this technique , by approximately half the repetition rate . as noted in @xcite , the sensitivity of this technique is not reduced as compared to the conventional sns . .
( b)- dependence of the noise signal on the time delay .
vectorial diagrams show summation of contributions of the two pulses to the fr signal for different ratios of the time delay @xmath90 and the larmor precession period @xmath91 .
( c ) the resulted dependence of the noise signal on the time delay @xmath90 , corresponding to resonant transient of the system or its autocorrelation function , and the spin resonance spectrum obtained by fourier transform of the autocorrelation function . , width=377 ] one more remarkable approach to the problem of bandwidth of the fr - based sns was proposed in @xcite .
the idea of the approach was to take advantage of an extremely broad spectrum of the femtosecond ( or picosecond ) light pulses emitted by the mode - locked ti - sapphire laser and to measure correlation characteristics of the fr noise with a very high temporal resolution inaccessible to conventional photodetectors .
for that purpose , the sample under study was suggested to be probed by a train of pairs of ultrashort pulses with the time interval between them variable within the range comparable with the period of larmor precession of the system ( fig.12,a ) . in spite of the fact that the two closely spaced pulses are not resolved by the detection system , the detected fr noise will depend on the time interval between them ( provided that there is some distinguished oscillation frequency ( and , therefore , a distinguished characteristic time ) in the fr noise , as shown in fig .
in particular , when the time interval between the pulses is equal to integer number of the oscillation periods , then contributions of this oscillating process to both pulses will have the same sign , and total fr of this pair of pulses will fluctuate . if , however , the time interval between the pulses equals odd number of half - periods of the oscillation , then contributions of these oscillations to the two pulses will compensate for each other , so that total contribution of this process to fr of the pair of pulses will vanish .
thus , by scanning the time delay between the two pulses of the probe beam and detecting the fr noise power in the transmitted light , we will observe an oscillatory transient of the process directly corresponding to its correlation function , whose fourier transform provides the autocorresponding spin - noise spectrum ( fig .
it is important that the two pulses are not supposed to be resolved by the detection system .
the efficiency of this ultrafast sns has been recently successfully demonstrated experimentally on a heavily @xmath79-doped bulk gaas @xcite .
the train of picosecond pulses was produced , in that experiment , by two synchronized lasers with a repetition rate of 80 mhz .
this technique , that implies measuring autocorrelation function instead of its fourier image , should allow one to directly detect decay of the transverse magnetization in time domain and to extend the accessible frequency range up to several thz .
the above proposals demonstrating possibility of considerable extension of the sns bandwidth give promise that this technique will find application in the epr spectroscopy of standard microwave ranges for nonperturbative investigations of transparent paramagnets .
in spite of remarkable advances in the data acquisition technique , the problem of polarimetric sensitivity retains its significance in the fr - based sns , and all suggestions that can help to improve this characteristic remain to be of great interest .
it is well known that faraday rotation can be strongly enhanced with the aid of a fabry - perot cavities @xcite , due to multiple passes of the light through the magneto - optical medium .
such an approach is especially popular nowadays for studying dynamics of spin states of low - dimensional semiconductor structures in high - finesse microcavities , where the observed fr can be increased by a few orders of magnitude . in these studies ,
it is usually tacitly assumed that the measured times are much longer than intrinsic times of the cavity ( the cavity photon lifetime and intermode beat period ) .
this is really the case for microcavities , when the photon round - trip time over the cavity is equal or just a few times longer than the oscillation period of the light wave .
when , however , this is not the case , and the fr oscillation frequency becomes comparable with or higher than that of the intermode beats , the response of the cavity becomes more complicated . in @xcite , it was shown that polarization of a monochromatic light resonant to a longitudinal mode of a fabry - perot cavity , may be highly sensitive to modulation of the intracavity anisotropy at frequencies multiple of spacing between its longitudinal modes .
this effect can be qualitatively understood in terms of the light traveling over the cavity back and forth and has much in common with the known effect of mode - locking in lasers @xcite .
indeed , intuitively it seems evident that enhancement of the cavity s response to a weak oscillating optical anisotropy of the intracavity element will be observed when the light moving back and forth inside the cavity finds this element with the same phase of the oscillation , so that the new contribution to the light polarization state is added to those already accumulated .
this reasoning , though formally incorrect ( being inapplicable to a monochromatic wave ) , proves to be useful for qualitative understanding of the resonant fr enhancement effect .
this effect can be also interpreted in terms of spectral transformation of the light passing through a polarization modulator : the resonant enhancement of the fr occurs when the modulation frequency coincides with intermode spacing of the fabry - perot cavity , and the sidebands of the modulated laser light hit adjacent modes of the resonator .
figure 13 shows frequency dependence of the polarization modulation gain factor ( @xmath92 ) for two different positions of the anisotropic element inside the cavity ( for more detail , see @xcite ) . for the sample placed in the middle of the cavity ( a )
, the light hits the sample twice per a round - trip , and , correspondingly , the effect of enhancement can be observed at frequencies that are by a factor of two lower than those for the sample placed at the edge of the cavity ( b ) . ) in a fabry - perot cavity for the sample placed at the edge of the cavity ( a ) and in the middle of it ( b ) .
the frequency @xmath93 is given in the units of intermode spacing @xmath94 .
figure ( c ) illustrates synchronism of the light traveling over the cavity with oscillations of anisotropy of the intracavity sample ( for @xmath95 ) .
the layers in the center of the cavity and near its edge ( colored green ) depict the medium with oscillating anisotropy , black sinusoid is the time variation of the anisotropy , and red lines show propagation of the light beam.,width=302 ] amplification , in this scheme , is realized in the optical ( rather than electronic ) channel , and the corresponding gain factor proves to be approximately equal to the cavity q - factor ( @xmath96 ) . in other words ,
the fabry - perot cavity can be used as a _ selective optical amplifier _ of a polarization signal produced by intracavity element at the frequency of intermode spacing ( @xmath94 , @xmath3 is the cavity length and @xmath97 is the speed of light ) or its multiples ( @xmath98 , @xmath79 is the integer ) .
this effect can evidently be helpful in sns for amplification of the polarization modulation arising in the probe light beam due to the random spin precession of the spin - system .
it is noteworthy that the increase in the fr signal in this method is achieved , exactly as in the case of high - extinction polarimetric measurements , at the expense of strongly increasing power density on the sample , which should be taken into account in the experiments . .
one more curious possibility of application of this effect is related to prospects of creating an all - optical spin - noise spectrometer @xcite .
the idea of this proposal combines the effect of amplification of the polarization signal with its optical quadratic detection .
let the studied transparent paramagnetic sample be placed inside an optical cavity ( fig.14 ) .
a monochromatic linearly polarized laser beam incident upon the cavity coincides in frequency with one of its longitudinal modes and thus passes through the cavity with no loss . at the exit of the cavity , we place a linear polarizer in a crossed position , so that no light passes through the system .
now , we start sweeping the transverse magnetic field applied to the sample . at the moment when the larmor precession frequency of the spins becomes equal to ( or multiple of ) the double intermode frequency of the cavity ( @xmath99 ) , oscillations of the fr appear to be strongly enhanced by the cavity , and the light intensity at the exit ( after the output polarizer ) becomes nonzero .
a unique property of this system is that it allows one to detect magnetic resonance in the fr noise spectrum at any frequency without any broadband electronics and broadband photodetectors - all the needed information is contained in the dc optical signal .
what is observed here is just a result of resonant coupling of the spin - system with the fabry - perot cavity .
it is clear that such a design of the all - optical spin spectrometer imposes heavy demands both on q - value of the cavity and on extinction ratio of the polarization system .
still , it seems feasible and may be useful , in particular , for applied purposes , as a basic system for magnetometers of new generation .
a considerable interest has been currently attracted to particularities of spin - noise spectroscopy under conditions of resonant or near - resonant optical probing of the spin - system ( @xcite ) .
it is clear that , generally speaking , the sns , in this case , looses its important property of being perturbation - free , when the probe laser beam induces real optical transitions . at the same time , the degree of perturbation of the spin - system by the probe beam depends on how large is the optical excitation rate compared with the dephasing rate of the spin - system , and , in most cases , the light power density can be decreased to the level when optical perturbation of the system may be neglected ( see , e.g. , @xcite ) . under these conditions ,
the sns acquires additional remarkable properties @xcite .
interesting possibilities of sns stems from the question about mutual correlations of the fr noise at different wavelengths of the probe beam . in @xcite , it was pointed out that these correlations depend on whether the appropriate fluctuations are contributed by the same spin ensemble ( ensemble of identical quantum systems ) or not .
it is important that this fact can be revealed not only in a straightforward way in the cross - correlation spectral characteristics of the fr noise ( which may be thought of ) , but also , much easier , in optical spectra of the fr noise power , which , generally , appear to be related to the conventional fr spectra in a nontrivial way .
the use of the probe beam wavelength as a tunable parameter of the standard spin noise spectroscopy allows one to look at this technique as at a sort of _ optical spectroscopy_. such an approach makes it possible , in certain cases , to detect the structure of optical transitions hidden in the linear optical or magneto - optical spectra and , thus , opens new possibilities of the sns technique .
let us consider dependence of the _ integrated fr - noise _ ( spin noise ) power on the probe beam frequency . as was already mentioned , the fr - based sns exploits `` paramagnetic '' part of the faraday rotation to monitor magnetization of the spin - system .
it seems evident that , to make conversion of the spin - system magnetization to the fr most efficient , one has to select the wavelength of the probe light in the region of the greatest fr ( greatest verdet constant ) , and , vice versa , the fr noise can not be detected at the wavelengths where the fr proper turns into zero .
it was shown , however , that this is not the case .
magnetization of a macroscopic spin - system is created by an ensemble of individual spins whose optical spectra may be either identical or different . as a result
, the magnetization noise of the spin - system may be transformed into that of the fr in different ways .
if optical spectrum of a paramagnet comprises several spectral features , then the fr angle @xmath45 at each frequency @xmath100 will be given by the sum of partial contributions of all these features : @xmath101 when all the spins of the system are identical , and optical spectrum of the whole ensemble just reproduces the one corresponding to an individual spin - system , then fluctuations of the fr will be evidently _ coherent _ or _ correlated _ over the whole spectrum , and spectral dependence of the mean - square fluctuation of the fr @xmath102 will be proportional to the square of the total fr ( square of sum of partial contributions ) : @xmath103 when , however , the paramagnet under study contains several spin subsystems with different optical spectra , their contributions of these subsystems to the fr noise will fluctuate _ independently _ , in uncorrelated way , and optical spectrum of the fr noise will be described by the _ sum of squares _ of the individual contributions rather than by their sum squared : @xmath104 as a result , optical spectra of spin noise in these two cases , due to different interference of partial contributions in the region of their overlap , may be essentially different .
figure 15 illustrates this difference for two closely spaced absorption lines associated either with the same spin - system or with different ( independent ) ones characterized by different optical spectra .
spectral dependence of the fr is taken here in the form of a dispersion - like curve characteristic of the `` paramagnetic '' part of the fr for the band with the width substantially exceeding magnetic splitting of the optical transition .
one can see that the distinction between these two spectra ( fig .
15 , c , d ) is most pronounced in the region between the lines , where contributions of two optical transitions either compensate for each other ( for a single spin - system , fig .
15 , c ) or sum up ( in statistical sense , for two different spin - systems , fig . 15 , d ) . one more manifestation of correlation characteristics of spin noise in the optical spectrum , as noted in @xcite , is that the optical spectroscopy of spin noise may show , in some cases , spectral resolution higher than the conventional optical or magneto - optical spectroscopy and , thus , may be helpful in resolving hidden structure of optical spectra .
this effect is revealed in a highly spectacular form in optical spectra with inhomogeneous broadening @xcite . the absorption and fr spectra in the vicinity of an isolated absorption band ( fig .
16 , a , b ) are known to be the same regardless of whether the band is broadened homogeneously or inhomogeneously .
linear magneto - optics ( as well as linear optics in general ) can not distinguish between these two cases .
the results of calculations presented in @xcite show that the optical spectra of spin noise in these two cases are drastically different .
16 , c , d ) for the homogeneously broadened band , the fr noise , which is proportional to the verdet constant squared , vanishes at the center of the band ( where the faraday rotation proper turns into zero ) , whereas for the band with a strong inhomogeneous broadening , this central dip disappears , and the fr noise proves to be the ) .,width=340 ] greatest at the band center .
qualitatively , it can be easily understood .
the faraday rotation at the center of the inhomogeneously broadened band vanishes because positive and negative contributions of the higher- and lower - lying spectral components compensate for each other , whereas their fluctuations , being uncorrelated , are summed up statistically and attain the greatest value at the point where the density of spectral components is the greatest , i.e. , at the center of the band .
when the inhomogeneous broadening is comparable with homogeneous , the dip in the optical spectrum of the fr noise power acquires some intermediate depth , which can be used to estimate the ratio of these two contributions to the line broadening .
it was also established in @xcite that magnitude of the spin noise considerably increasers with increasing ratio of the inhomogeneous bandwidth to homogeneous ( approximately in direct proportion with it ) .
it results from the fact that , in the case of strongly inhomogeneously broadened band , the main contribution to the fr noise of the probe beam is made by the spectral components lying nearby the laser wavelength ( within the range of homogeneous width ) .
this relatively small number of narrow spectral components with relatively large partial contribution of each of them may provide strong enhancement of the spin noise power for inhomogeneously broadened bands .
this fact , on the one hand , makes easier sns experiments with inhomogeneously broadened systems ( like quantum dots ) and , on the other , can be used for measuring homogeneous linewidth of optical transitions .
the above properties of the optical spin - noise ( osn ) spectroscopy have been confirmed experimentally in @xcite .
the osn spectra of the homogeneously broadened d1 line of potassium atoms and inhomogeneously broadened band of the ingaas / gaas quantum dots were strongly different in accordance with the above results .
the effect of enhancement of the spin noise power in inhomogeneously broadened systems was also confirmed experimentally .
thus , the optical spectroscopy of the fr noise may be considered as the other side of the fr - based sns that may provide interesting additional information about spin - system under study .
the sns , as a method of the esr spectroscopy , was primarily intended for providing standard information about g - factors and relaxation rates of the spin - system . as an optical technique
, the fr - based sns has much in common with the conventional optical methods of the esr detection @xcite , with an essential difference that the detected spin precession , in the sns , is spontaneous ( stochastic ) rather than coherently excited by an external field , and the spin - system is supposed to remain in the state of thermal equilibrium .
what is highly significant , in our opinion , is that the use of magnetization _ noise _ in the capacity of signal , in combination with the optical ( laser - assisted ) technique , allows one to get additional information fundamentally inaccessible for conventional methods of spectroscopy and makes the sns unique in many respects .
one of the most important features of the fr - based sns , as was already mentioned , is related to its nonperturbative character : the light interacting with the paramagnet in the region of its transparency produces virtually no real excitation of the system .
this property may be highly important , e.g. , in studies of ultracold atoms @xcite or semiconductor systems , when appearance even of small amount of photo - induced charge carriers may substantially distort dynamics of the spin - system .
another property of the sns that essentially distinguishes it from other ( fundamentally perturbative ) methods of the magnetic - resonance spectroscopy is that it does not require population difference between magnetic sublevels of the spin - system to detect the resonance .
this technique remains equally efficient at high temperatures and low magnetic fields ( including zero field ) , when the population difference becomes negligibly small ( see , e. g. , @xcite . by detecting spin noise as described above , we , in fact , monitor random microscopic motion of a macroscopic object confined spatially by the probe laser beam .
this allows us to get information hardly accessible for the conventional radio- or linear optical spectroscopy .
in particular , sns was used to identify fermi - dirac statistics of a degenerate electron gas in heavily doped n - gaas , to distinguish between localized and delocalized conduction - band electrons , and to detect the effect of brownian motion of the electrons in the conduction band on the spin noise line width .
essential information can be also extracted from absolute value of the spin noise power which is known to be directly related to total number of spins probed by the beam @xcite .
a widely - known method of nonlinear optics is the so - called `` @xmath105-scan technique '' , which allows one to identify optical nonlinearity and to measure , in a simple way , coefficients of nonlinear absorption or refraction of the medium @xcite . in all modifications of this -scan arrangement ( a ) and characteristic dependence of the transmitted light intensity on position of the sample @xmath105 for the sample with optical nonlinearity.,width=340 ]
technique , the sample is drawn through the waist of a tightly focused laser beam ( along the @xmath105-axis , fig .
17 ) , and intensity of the transmitted light is measured as a function of @xmath105-coordinate .
when the sample is optically linear and its optical properties do not depend on the light power density , the transmitted light intensity does not show any dependence on @xmath105 .
the presence of such a dependence with a peculiarity at @xmath106 usually serves as an indicator of nonlinearity of the medium .
if we apply this @xmath105-scan technique to the measurements of spin noise , we will evidently obtain a @xmath105-dependent noise signal ( as if the medium were nonlinear ) , because , when moving the sample through the waist of the beam , the number of probed spins , controlling magnitude of the spin noise power , changes . in other words , the tightly focused laser beam traveling through a bulk paramagnet probes the spin - system mainly by a small spatial region in the vicinity of its waist .
this situation has evidently much in common with the case of nonlinear medium when the main response is provided by the region of the beam with the greatest light power density .
another curious illustration of this effect may be provided by a simple two - beam experiment , which can be considered as a sort of pump - probe spectroscopy of the light intensity noise ( fig.18 ) .
let a light beam pass through a layer of transparent medium and let us ask the question ; can we find the spot illuminated by this beam on the layer with the aid of another light beam ?
the usual reply is : yes , we can do this provided that the first beam ( usually called
pump ) changes in some way optical properties of the layer in the illuminated spot . in other words , this is possible ( and is a usual story ) in _ nonlinear _ optics . in the light intensity noise ,
however , this is also possible , in spite of the fact that the probe beam is not supposed to modify , in any way , properties of the layer . indeed , if we detect total intensity noise of the two beams ( pump and probe ) transmitted by the layer ( fig .
18 ) , then the measured signal will depend on whether they pass through the same spot of the layer or not . in the first case , their fluctuations will be correlated , and total noise power will be given by their sum squared , while , in the second case , they will be uncorrelated , and the measured noise power will be given by the sum of their squares . as a result ,
the spot illuminated by the pump can be easily detected by the
probe. we can say , that the illuminate spot is coded in a unique way , and the hey to its code is provided by the noise of the pump ` ' . it is appropriate to mention here the idea of two - beam spin noise spectroscopy , which was put foeward in @xcite and may be promising for studying spatial characteristics of spin systems .
these features of the sns have been used in @xcite to realize a three - dimensional sns - based tomography .
efficiency of this technique was demonstrated with a pair of @xmath79-doped gaas wafers , around 350 @xmath8 m thick each , probed , as in the conventional z - scan technique , by a focused laser beam .
the wavelength of the probe beam was chosen well below the bandgap of gaas ( 849 nm ) .
the measurements were performed with no external magnetic field , so that the spin noise spectrum was centered at zero frequency . due to different doping concentrations of the two plates , the corresponding spin relaxation times ( and spin - noise spectra ) were different , which made it possible to distinguish them in the sns experiment . in this proof - of - principle experiment ,
the spatial in - depth resolution of 50 @xmath8 m was demonstrated .
the method may allow , in authors opinion , to realize 3d doping measurements with submicrometer spatial resolution even at low doping concentrations .
this is one of the features of sns that , along with its ability to penetrate inside hidden structure of optical transitions , mentioned above , bring is close to the methods of nonlinear optics .
we have briefly outlined development of the sns for the last several years and described its main achievements and potentialities as applied to scientific research . it is curious that the idea of detecting magnetic resonance in the faraday - rotation noise spectrum , that looked initially more like an academic trick useful primarily for tutorial purposes , gave birth to a highly efficient and , in many respects , unique experimental tool .
this technique keeps certain properties of conventional experimental methods like magnetic resonance spectroscopy , optical and raman spectroscopy , spectroscopy of double rf - optical resonance , and gorter s paramagnetic relaxation method , but essentially differs from any of them and acquires thereby qualitatively new features , some of which were considered in this paper .
nowadays , as we believe , the fr - based spin noise spectroscopy , as a new experimental tool , is only at the beginning of its life in experimental research , and its potentialities are far from being exhausted . especially promising looks its application to studies of micro - samples and nanostructures , bearing in mind that smallness of the optically probed volume of the sample ( smallness of the number of spins ) may be considered , under certain conditions , as a favorable factor from the viewpoint of relative magnitude of the signal .
an important degree of freedom of the sns , which may considerably widen its informative capacity , is provided by the wavelength of the probe light .
nontrivial shape of optical spectra of the spin noise power may contain information about hidden structure of optical transitions and thus may reveal apparent features of nonlinear optics .
interesting possibilities may be provided by different modifications of the intentionally perturbative ( resonant ) methods of the sns .
these methods , generally , do not have much to do with polarization measurements and do not need high polarimetric sensitivity . their main advantage is similar to advantage of the fft spectrum analyzer compared to the sweeping one : the broadband ( rather than monochromatic ) intensity modulation of the acting light substantially improves sensitivity of the technique .
these methods of the `` active '' noise spectroscopy , however , pertaining more to nonlinear optics , may provide the information related , to a greater extent , to the optical ( rather than spin ) dynamics of the system . at present , we have every reason to believe that the novel technique of sns will soon find a wide use in studies of spin - systems and will turn into a standard method of experimental research .
this is a unique case when a pure noise , usually considered as a nuisance factor , turns into a basic source of information in a field of science .
the author is thankful to gleb g. kozlov , aleksey kavokin , evgenii aleksandrov , and dmitri yakovlev for numerous useful discussions . the financial support from the russian ministry of education and science ( contract no .
11.g34.31.0067 with spbsu and leading scientist a. v. kavokin ) is acknowledged .
99 s. a. wolf , d. d. awschalom , r. a. buhrman , j. m. daughton , s. von molnar , m. l. roukes , a. y. chtchelkanova , and d. m. treger , _ science _ , * 294 * , 1488 ( 2001 ) i. i. @xmath107uti@xmath108 , j. fabian , s. das sarma , _ rev .
_ , * 76 * , 323 ( 2004 ) j. m. elzerman , r. hanson , l. h. willems van beveren , b. witkamp , l. m. k. vandersypen , and l. p. kouwenhoven _ nature _ ( london )
* 430 * , 431 ( 2004 ) e. b. aleksandrov , and v. s. zapasskii , _ jetp _ * 54 * , 64 ( 1981 ) sleator , t. , hahn , e. l. , hilbert , c. and clarke , j. _ phys.rev .
lett . _ * 55 * , 1742 ( 1985 ) t. mitsui , _ phys.rev .
lett . _ * 84 * , 5292 ( 2000 ) l. d. landau , and e. m. lifshitz , _ statistical physics _ 1980 ( pergamon ) kuzmich , a. , molmer , k. and polzik , e. s. _ phys.rev
. lett . _ * 79 * , 4782 ( 1997 ) sorensen , j. l. , hald , j. and polzik , e. s. _ phys.rev . lett_.
* 80 * , 3487 ( 1998 ) rmer , m. , hubner , j. h. and oestreich , m. _ rev .
_ * 78 * , 103903 ( 2007 ) mller , g. , d. schuh , j. hubner , and m. oestreich , _ phys . rev .
b _ 81 , 075216 ( 2010 ) crooker , s. a. , brandt , j. , sandfort , s. , greilich , a. , yakovlev , d. r. , reuter , d. , wieck , a. d. and bayer , m. _ phys .
lett . _ * 104 * , 036601 ( 2010 ) forrester , a. t. gudmundsen , r. a. and johnson , p. o. _ phys .
rev . _ * 99 * , 1691 ( 1955 ) zapasskii , v. s. and feofilov , p. p. , _ sov .
* 18 * , 323 ( 1975 ) hunbury - brown , r. and twiss , r. q. _ nature _ * 177 * , 27 ( 1956 ) aleksandrov , e. b. , golubev , y. m. , lomakin , a. v. and noskin , v. a. _ sov .
* 26 * , 643 ( 1983 ) b. j. berne and r. pecora , _ dynamic light scattering with applications to chemistry , biology , and physics _ , wiley interscience : new york , 1976 e.b.aleksandrov , phys .
, v. * 107 * , 595 - 622 ( 1972 ) aleksandrov , e. b. , kaliteevskii , n. i. and chaika , m. p. _ sov .
_ * 22 * , 760 ( 1979 ) zapasskii , v. s. , in _ spectroscopy of solids contain- ing rare - earth ions
a. a. kaplyanskii and m. f. macfarlane , 1987 ( elsevier science publishers b.v . ) p. 674 a. d. buckingham and p. j. stephens , ann .
phys . chem .
* 17 * , 399 ( 1966 ) bitter , f. , _ phys .
, _ * 76 , * 833 ( 1949 ) h. g. dehmelt , phys . rev . *
105 * , 1924 ( 1957 ) w. e. bell and a. l. bloom , phys . rev . * 107 * , 1559 ( 1957 ) zapasskii , v. s. , kozlov , g. g. and malyshev v. a. _ proc .
spie _ , * 1274 * , 193 ( 1990 ) gorter , c. j. , _ paramagnetic relaxation _ , 1947 ( elsevier ) aleksandrov , e. b. and zapasskii , v. s. _ sov .
phys . - solid state _
_ 20 _ , 679 ( 1978 ) bloch , f. _ phys .
rev . _ * 70 * , 460 ( 1946 ) kastler , a. , _ j. opt
. soc . am . , _ * 53 * , 902 ( 1963 ) kastler , a. , _ science _ ,
* 158 * , 214 ( 1967 ) happer , w. , _ rev .
phys . _ * 44 * , 169 ( 1972 ) crooker , s. a. , awschalom , d. d. , baumberg , j. j. , flack .
f. , and samarth , n. _ phys . rev .
b _ * 56 * , 7574 ( 1997 ) zapasskii , v. s. , _ j. appl
. spectrosc . _ * 37 * , 857 ( 1982 ) aleksandrov , e. b. and zapasskii , v. s. _ opt . spectrosc . _ * 41 * , 502 ( 1976 ) jones , r. v. _ proc .
_ 349 , 423 ( 1976 ) v. s. zapasskii , _ opt .
_ , 47 , 450 - 451 ( 1979 ) crooker , s. a. , rickel , d. g. , balatsky , a. v. and smith , d. _ nature _ * 431 * , 49 ( 2004 ) mller , g. m. , oestreich , m. , rmer , m. and hubner , j. _ physica e _ 43 , 569 ( 2010 ) kharkevich , a. a. _ spectra and analysis _ , 1995 ( springer ) mccoy , m. a. and ernst r. r. _ chem .
lett . _ * 159 * , 587 ( 1989 ) gueron , m. and leroy , j. l. _ j. magn .
res . _ * 85 * , 209 ( 1989 ) mller , n. and jerschow , a. _ proc .
usa _ * 103 * , 6790 ( 2006 ) aleksandrov , e. b. , kozlov , v. p. , and kulyasov , v. n. , _ sov .
jetp _ , * 39 * , 620 ( 1974 ) aleksandrov , e. b. and kulyasov , v. n. , _ opt .
spectrosc , ( ussr ) _ , * 40 * , 449 ( 1976 ) smirnov , d. f. , sokolov , i. v. , _ sov .
jetp _ , * 43 * , 1095 ( 1976 ) mihaila , b. , crooker , s. a. , rickel , d. g. , blagoev , k. b. , littlewood , p. b. and smith , d. l. _ phys .
a _ * 74 * , 043819 ( 2006 ) happer , w. and b.s.mathur , _ phys .
lett . _ * 18 * , 577 ( 1967 ) gorbovitskii , b. m. and perel , v. i. _ opt
. spectrosc .
_ * 54 * , 229 ( 1983 ) eden , d. and swinney , h. , _ opt . commun . , _ * 10 , * 191 ( 1974 ) crooker , s. a. , chemg , l. and smith , d. l. _ phys . rev .
b _ * 79 * , 035208 ( 2009 ) a. kuzmich , n. p. bigelow , and l. mandel , _ europhys
_ , * 42 * , 481 ( 1998 ) a. kuzmich , l. mandel , and n. p. bigelow , _ phys
_ , * 85 * , 1594 ( 2000 ) yabuzaki , t. , mitsui , t. and tanaka , u. _ phys .
rev . lett_. * 67 * , 2453 ( 2001 ) mcintyre , d. h. , fairchild , c. e. , cooper , j. and walser , f . _
lett . _ * 18 * , 1816 ( 1993 ) ito , t. , shimomura , n. , and yabuzaki t. _ j. phys .
japan , _ * 72 * , 962 ( 2003 ) martinelly , m. , valente , p. , failache , h. , felinto , d. , cruz , l. s. , nussenzveig , p. and lezama , a. _ phys .
a _ * 69 * , 043809 ( 2004 ) valente , p. , failache , h. and lezama , a. _ eur .
phys . j_. * 50 * , 133 ( 2008 ) w. e. bell and a. l. bloom , _ phys .
lett . _ * 6 * , 280 ( 1961 ) g. alzetta , e. arimondo , c. ascoli , _ nuovo cimento b _ , * 54 * , 107 ( 1968 ) s. harris , _ physics today _
, * 50 * , 36 ( 1997 ) r. walser and p. zoller , _ phys .
a _ , * 49 * , 5067 ( 1994 ) g. g. kozlov and v. s. zapasskii , arxiv : 1206.1921v1 [ physics.optics ] 9 june 2012 oestreich , m. , rmer , m. , haug , r. g. and hagele , d. _ phys .
* 95 * , 216603 ( 2005 ) kikkawa , j. m. and awschalom , d. d. , _ phys .
_ , * 80 * , 4313 ( 1998 ) starosielec , s. and hogele , d. _ appl . phys . lett_. * 93 * , 051116 ( 2008 ) berski , f. , kuhn , h. , lonnemann , j.g . , hubner , j. , and oestreich , m. , arxiv:1207.0081 [ cond-mat.mes-hall ] r. rosenberg , c. b. rubinstein , and d. r. herriott _ applies optics _ , * 3 * , 1079 ( 1964 ) ling , h. y. , _ j. opt .
_ * 11 * , 754 ( 1994 ) kavokin , a. v. , vladimirova , m. r. , kaliteevski , m. a. , lyngnes , o. berger , j. d. gibbs , h. m. and khitrova , g. _ phys.rev .
b _ * 56 * , 1087 ( 1997 ) salis , g and moser , m. _ phys.rev .
b _ * 72 * , 115325 ( 1997 ) li , y. q. , steuerman , d. w. , berezovsky , j. , seferos , d. s. , bazan , g. and awschalom , d. d. _ appl . phys
* 88 * , 193126 ( 2006 ) dong , l. , jiang , y. , chen , h. , and shi , y. _ j. appl .
phys . _ * 107 * , 093101 ( 2010 ) giri , r. , cronenberger , s. , vladimirova , m. , and scalbert , d. _ phys.rev .
b _ * 85 * , 195313 ( 2012 ) zapasskii , v. s. and przhibelskii , s. g. _ opt . spectrosc_. * 110 * , 917 ( 2011 ) lamb , w. e. , jr . , _ phys
* 134 * ( 6a ) , a1429 ( 1964 ) chalupczak , w. and godun , r. m. _ phys .
a _ * 83 * , 032512 ( 2011 ) horn , h. , mller , g. m. , rasel , e. m. , santos , l. , hubner , j. and oestreich , m. _ phys .
a _ * 84 * , 043851 ( 2011 ) q. huang and d. s. steel , phys . rev . *
b 83 * , 155204 ( 2011 ) zapasskii , v. s. , greilich , a. , crooker , s. a. , li , y. , kozlov , g. g. , yakovlev , d. r. , reuter , d. , wieck , a. d. and bayer , m. arxiv:1206.4168v1 [ cond-mat.mes-hall ] ( 2012 ) brossel , j. and bitter , f. _ phys .
_ , * 86 , * 308 ( 1952 ) aleksandrov , e. b. and zapasskii , v. s. _ j. russian laser research _ * 6 * , 291 ( 1985 ) mihaila , b. , crooker , s. a. , blagoev , k. b. , rickel , d. g. , littlewood , p. b. and smith , d. l. _ phys .
a _ * 74 * , 63608 ( 2006a ) mller , g. m. , rmer , m. , schuh , d. , wegscheider , w. , hubner , j. and oestreich , m. _ phys . rev .
, 206601 ( 2008 ) sheik - bahae , m. , said , a. a. and stryland , e. w. v. _ opt . lett . _
* 14 * , 955 ( 1090 ) patterson , w. m. , sheik - bahae , m. , epstein , r. i. and hehlen , m. p. _ ieee j. quant .
electron _ * qe-26 * , 760 ( 1990 ) pershin , yu .
v. , slipko , v. a. , roy , d. , and sinitsyn , n. a. , arxiv:1211.6764v1 [ condmat.mes-hall ] 28 nov 2012 rmer , m. , hubner , j. and oestreich , m. _ appl .
lett . _ * 94 * , 112105 ( 2009 ) | more than 30 years ago , the feasibility of detecting magnetic resonance in the faraday - rotation noise spectrum of transmitted light has been demonstrated experimentally .
however , practical applications of this experimental approach have emerged only recently thanks , in particular , to a number of crucial technical ements .
this method has now become a popular and efficient tool for studying magnetic resonance and spin dynamics in atomic and solid - state paramagnets . in this paper
, we present a review of research in the field of spin noise spectroscopy including its physical basis , its evolution since its first experimental demonstration , and its recent experimental advances .
main attention is paid to the specific capabilities of this technique that render it unique compared to other methods of magnetic and optical spectroscopy . the paper is primarily intended for the experimentalists who may wish to use this novel optical technique . |
the axial vector anomaly plays a fundamental role in the structure of the standard model and describes many physical processes , including the classic decay @xmath4 @xcite .
one can summarize the traditional current algebra manipulations used to treat anomalous processes by an effective action .
this is a functional , @xmath5 , of a chiral field of the pseudoscalar mesons , @xmath6 , and background gauge fields @xmath7 , @xmath8 , coupled to left- and right - handed chiral currents .
it generates ( consistent ) anomalies under local gauge transformations and is known as the wess - zumino - witten ( wzw ) term @xcite .
the wzw term has been developed into a phenomenologically useful form by kaymakcalan , rajeev and schechter ( krs ) @xcite following witten s pioneering work .
it can be understood as arising from a chern - simons term built of yang - mills gauge fields in @xmath9 , suitably compactified such that @xmath10 zero modes emerge as the mesons @xcite .
it applies to any effective theory of pseudo - nambu - goldstone bosons ( pngb s ) coupled to gauge fields , , as in the case of little higgs theories @xcite .
the wzw term arises naturally in connection with topological physics in extra dimensions , and occurs in both top - down @xcite and bottom - up @xcite approaches to holographic qcd in order to correctly match the flavor anomalies of qcd .
the wzw term for spontaneously broken @xmath11 flavor symmetry describes anomalous processes involving the pseudoscalar pngb s and fundamental gauge fields such as the photon .
for example , we have @xmath12 with @xmath13 and @xmath14
. however , the wzw term also summarizes processes containing effective vector mesons in pole approximation , such as @xmath15 , @xmath16 , where @xmath17 ; and @xmath18 , @xmath19 , . , where @xmath20 , with @xmath21 and @xmath22 @xcite .
to leading order in an expansion in @xmath23 the wzw term is seen to contain `` pseudo - chern - simons '' terms ( pcs ) , such as @xmath24 .
recently it has been proposed that terms involving @xmath25 may be of phenomenological interest @xcite , which has stimulated the present work .
presently we note that the standard model implies a pcs term in the lagrangian involving the photon , the z - boson , and the isoscalar @xmath1 vector meson of the form : [ one ] _ ^z^f^ .
the derivation of this result requires a careful accounting of anomalies . in the absence of @xmath1 , gauge invariance
is maintained by the combination of the wzw term and anomaly cancelling contributions from the lepton sector . in the presence of @xmath1 , @xmath26 gauge invariance
must be enforced by including appropriate counterterms , @xmath27 , .
this uniquely specifies the coefficient of eq .
, apart from coupling constant normalizations .
@xmath28 in eq .
should be thought of as a gauge - invariant stueckelberg field , , we are in unitary gauge for the broken @xmath29 generators . for simplicity , we restrict attention to two light flavors and the @xmath1 is coupled to baryon number @xmath30 , inducing the coupling to the nucleon isodoublet @xmath31 , as @xmath32 in this way , the wzw term contains a term @xmath33 , where @xmath34 is the properly normalized goldstone - wilczek skyrmion baryon current when @xmath35 . ] .
we also note that the one - loop diagram responsible for the pcs term is closely related to the electroweak baryon number anomaly .
the full details will be presented elsewhere @xcite .
physically interesting effects arise when we integrate out the @xmath1 and the @xmath0 , replacing them with the baryon current and neutrino current respectively .
the new anomalous interaction is then : s_int = _ f d^4x ^ j _
|^f_l_^f_l f _ , where @xmath36 is the isosinglet baryon current and we sum over all left - handed neutrino flavors .
the normalization factor is : = , _
one could include similar terms for the @xmath37 and @xmath38 mesons , but @xmath1 is expected to dominate .
interesting effects will arise in processes such as : + n & & + + n ( or n ) + + n & & + | + n ( or n ) .
coherent enhancements to these basic processes can occur in situations described by an ambient background baryon density .
we presently turn to two such classes of phenomena : neutrino scattering on nucleons ; and the cooling of a young neutron star .
we briefly mention that an analogous axion - photon - baryon current interaction arises with similarly important implications .
we consider an incident neutrino of any flavor with 4-momentum @xmath39 colliding with a stationary nucleon , producing an outgoing neutrino of momentum @xmath40 , and a photon of momentum @xmath41 .
we assume @xmath42 , where @xmath43 is the nucleon mass and recoil is treated as negligible .
the action then becomes : [ result2 ] s_int d^4x _
0ijk|_l ^i _ l f^jk ^3 ( ) .
the resulting total cross - section is : [ result ] & = & + & = & 2.610 ^ -41 ( e_/gev)^6 ( g_/10.0 ) ^4 ^ 2 .
we remark that there is considerable uncertainty in the value of @xmath44 , generally extracted from nuclear potential models @xcite , and the value we use here is `` conservative . '' to assess the possible experimental sensitivity to this effect we note that miniboone observes an excess of @xmath45 events with electromagnetic energy and with reconstructed mean neutrino energies of order @xmath46 mev @xcite .
these events look like @xmath47 mev @xmath48 charged current events , but the electron could be faked by the hard photon in our process .
the miniboone @xmath49 beam spectrum is fairly flat from @xmath50 mev to @xmath51 , peaking at @xmath52 mev .
we focus on @xmath53 mev @xmath49 s and assume that these produce photons with @xmath54 mev via our process
. in this energy range miniboone accepts @xmath55 charged current quasi - elastic ( ccqe ) @xmath56 events with cross - section , @xmath57 @xcite .
thus we expect to produce @xmath58 events .
the extra factor of @xmath59 arises from the fact that our process involves both @xmath60 and @xmath61 while ccqe involves only @xmath60 in a carbon target .
our cross - section has a distinctly flat angular distribution , @xmath62 where @xmath63 .
we note , however , that the photon will be pulled forward as form - factor effects , which we have ignored , set in .
while this estimate is extremely naive , it is encouraging that existing experiments may already have sensitivity to our effect . for a more refined analysis
we require an improved @xmath64 on nuclear ( carbon ) targets , including coherence and form - factor effects , and convolution of @xmath64 with the beam spectrum .
additional anomaly mediated interactions can also be studied systematically starting from the wzw term .
for instance , a term @xmath65 involves pion exchange with the nucleus in place of @xmath1 exchange .
however , this process contains @xmath66 , and is not the dominant effect . also , in contrast to ( [ one ] ) , the amplitude for this process does not add coherently over individual nucleons at low energy .
however , at the energies of interest , the assumption of coherence on the entire nucleus leads to only small effects confined to the forward direction . ] .
we thus conjecture that the anomaly mediated process may be relevant to the miniboone excess .
moreover , it could lead to multiple enhanced prospects for observing quasi - elastic neutrino scattering on nuclei .
modern neutrino experiments could ultimately provide a normalization of the uncertain input parameter , @xmath44 . while the anomaly mediated
neutrino process is higher order , it contains a hard photon .
even for relatively low energy reactor neutrinos , @xmath67 , the enormous flux availability suggests that the anomaly process may be observable .
the hard photon also provides a possible observable for a quartz - cerenkov detector , or liquid halogen bubble detectors , .
the anomaly mediated process is competitive with the conventional processes for neutron star cooling . for illustration
, we consider a particular coherent subprocess in a superconducting neutron star core , but the anomaly mediated process will have a more general applicability .
neutron star cooling is reviewed in @xcite .
the full set of processes contributing to neutron star cooling is complex . in the ultra - high density inner core
the energy loss is dominated by the direct urca process , @xmath68 throughout most of star the nucleons are at lower densities , @xmath69 , and are fermi degenerate . in this case , the direct urca process is highly suppressed .
energy is then typically lost by the modified urca ( murca ) process where a bystander nucleon is included to conserve energy and momentum .
the murca process is affected by the superfluid phase , which appears below a critical temperature @xmath70 .
this reduces the murca rate .
however , new cooling mechanisms associated with cooper pairing of nucleons may turn on at @xmath71 , compensating the murca suppression .
we presently consider typical neutron star densities of @xmath72 ( twice nuclear density @xmath73 ) and assume that the regime @xmath74 contains the standard murca processes .
we summarize these effects by ( see , table 2 of @xcite ) : @xmath75 at temperature @xmath76 .
this describes a collection of various processes , ignoring superfluidic suppression , and permits a naive comparison to our process .
we now want to estimate the cooling rate due to the anomaly mediated process .
our basic assumptions are : 1 .
the degenerate protons pair to make a superconductor .
this spontaneously breaks @xmath77 and gives a mass to the photon , also known as the inverse penetration length of the superconductor .
the mass depends sensitively on strong interaction models , we take @xmath78 as a typical value .
the ( massive ) photons are in thermal equilibrium with the neutrons , protons and electrons in the superconducting interior of the neutron star and we work in the limit @xmath79 .
the photons are therefore characterized by the phase - space distribution function and number density , @xmath80^{-1 } \
, , \quad n_\gamma = g \int \frac{d^3p_\gamma } { ( 2 \pi)^3 } f(p_\gamma ) \,,\ ] ] with @xmath81 for a massive spin - one particle . for @xmath82 , we can approximate @xmath83 .
the emissivity of neutrinos due to @xmath84 is @xmath85 thus requiring the decay rate of the photon , @xmath86 . in the star rest frame the baryon number current is @xmath87 , and the photon 4-momentum is @xmath88 .
however , it is convenient to work in the photon rest - frame , where @xmath89 . in this frame
the nucleon current is slightly boosted : @xmath90 , where @xmath91 and @xmath92 . in the nonrelativistic limit
we thus have : _ p _ & = & + & & |m|^2 ( 2)^4 ^ 4(p_^- p_1-p_2 ) , and the invariant final - state spin - summed and initial state spin - averaged matrix element is : |m|^2 = ^0 _ ^0 _ _ .
this results in the decay rate : _ p _ & = & |_|^2 , and emissivity : [ lum ] q_^anom & = & ce^-m_/t(m_t)^5/2 , + c & = & = 0.012 - 0.96 .
the range of @xmath93 corresponds to a range of @xmath94 to @xmath95 .
we note that both @xmath44 and @xmath96 run with nuclear density , , refs .
@xcite obtain @xmath96 reduced by a factor @xmath97 at @xmath98 .
holding @xmath96 fixed , larger values of @xmath44 account for this effect , as well as other effects , , higher resonance contributions . for standard density @xmath99 ,
this yields an emissivity of q_^anom & = & 2.31 10 ^ 22 erg^-1 ^ -3 + & & m^9/2(g_/10)^4 e^(-11.6m / t_9)(t_9)^5/2 , where @xmath100 .
[ cool ] we see in fig.(2 ) that this is competitive with the cooling rate from murca processes , by a factor of @xmath101 . ] .
we note that in the early phase of neutron star formation with @xmath102 , our process may actually dominate .
this will be developed and reported elsewhere @xcite .
anomaly mediated interactions between photons and neutrinos at finite baryon density may play an important role in laboratory neutrino experiments and astrophysical processes .
the new interaction ( [ one ] ) may be relevant in accounting for the miniboone low energy excess .
it may also play a significant role in neutron star cooling and early stage evolution .
there are many potentially important applications in various other physical regimes
. we will present a more detailed analysis and discussion elsewhere , including the detailed derivation of pcs and axion interactions from the wzw term @xcite@xcite .
we further remark that the axion will have a similar induced coupling to the photon and the @xmath1 , leading to an interaction of the form : c_axion _
f^|n ^n , where @xmath103 is calculable from a given axion model .
an important application is to consider axion emission and the resulting bounds on axion couplings from supernovae ( sn1987a ) .
0.2 in * acknowledgements * 0.1 in we thank w. bardeen , j. conrad , and p. cooper for helpful discussions . research supported by the u.s .
department of energy grant de - ac02 - 76cho3000 and by nsf grants phy-00506630 and 0529954 .
this work was supported in part by the joint theory institute funded by argonne national laboratory and the university of chicago .
jh acknowledges the hospitality of the aspen center for physics during part of this work .
j. wess and b. zumino , phys .
b * 37 * , 95 ( 1971 ) .
e. witten , nucl .
b * 223 * , 422 ( 1983 ) .
o. kaymakcalan , s. rajeev and j. schechter , phys .
d * 30 * , 594 ( 1984 ) . c. t. hill and c. k. zachos , phys .
d * 71 * , 046002 ( 2005 ) ; c. t. hill , phys . rev .
d * 73 * , 126009 ( 2006 ) ; phys .
d * 73 * , 085001 ( 2006 ) c. t. hill and r. j. hill , 0705.0697 [ hep - ph ] .
d * 75 * , 115009 ( 2007 ) t. sakai and s. sugimoto , prog .
. phys . * 113 * , 843 ( 2005 ) j. erlich , e. katz , d. t. son and m. a. stephanov , phys .
lett . *
95 * , 261602 ( 2005 ) l. da rold and a. pomarol , nucl .
b * 721 * , 79 ( 2005 ) a standard reference for nucleon - meson couplings is : r. machleidt , k. holinde and c. elster , phys .
* 149 * , 1 ( 1987 ) .
j. c. caillon and j. labarsouque , j. phys .
g * 21 * , 905 ( 1995 ) .
h. kouno , y. horinouchi and k. tuchitani , prog .
. phys . * 112 * , 831 ( 2004 ) a. a. aguilar - arevalo _ et al . _ [ miniboone collaboration ] , phys .
lett . * 98 * , 231801 ( 2007 ) and `` measurement of muon neutrino quasi - elastic scattering on carbon , '' 0706.0926 [ hep - ex ] .
j. monroe [ miniboone collaboration ] , nucl .
suppl . * 139 * , 59 ( 2005 ) [ hep - ex/0408019 ] .
`` black holes , white dwarfs , and neutron stars : the physics of compact objects , '' s. l. shapiro and s. a. teukolsky ( john wiley , new york , 1983 ) .
d. g. yakovlev , a. d. kaminker , o. y. gnedin and p. haensel , phys .
rept . * 354 * , 1 ( 2001 ) d. g. yakovlev and c. j. pethick , ann .
astrophys .
* 42 * , 169 ( 2004 ) e. flowers , m. ruderman and p. sutherland , astrophys .
j. * 205 * , 541 ( 1976 ) .
some implications of @xmath104 anomalies at constant baryon density have been studied in : d. t. son and a. r. zhitnitsky , phys .
d * 70 * , 074018 ( 2004 ) [ hep - ph/0405216 ] . | we propose new physical processes based on the axial vector anomaly and described by the wess - zumino - witten term that couples the photon , @xmath0-boson , and the @xmath1-meson .
the interaction takes the form of a pseudo - chern - simons term , @xmath2 .
this term induces neutrino - photon interactions at finite baryon density via the coupling of the @xmath0-boson to neutrinos .
these interactions may be detectable in various laboratory and astrophysical arenas .
the new interactions may account for the miniboone excess .
they also produce a competitive contribution to neutron star cooling at temperatures @xmath3 . these processes and related axion photon interactions at finite baryon density appear to be relevant in many astrophysical regimes . |
in the continuum euclidean path - integral , chiral anomaly comes from jacobian of the fermion measure @xcite .
the point is that @xmath0 of the transformed fermion measure gives nonzero contribution because of the infinite degrees of freedom . on the other hand ,
lattice field theory is a framework to simulate the continuum theory with a finite number of lattice sites .
chiral anomaly can not be reproduced on the finite lattice in the same way as the continuum theory .
we should definitely distinguish finite and infinite lattice formulations and concentrate on reproducing chiral anomaly based on the finite lattice for practical numerical studies .
the species doubling problem of the lattice fermion @xcite is closely related with chiral anomaly @xcite . according to the nielsen - ninomiya theorem
, a single weyl fermion can not exist on the lattice @xcite .
when formulating lattice dirac fermion , one has almost no choice but to break chiral symmetry explicitly using wilson terms @xcite .
otherwise , one has to give up one of the other assumptions of the theorem as seen in the literature .
lscher s implementation of lattice chiral symmetry based on the ginsparg - wilson relation @xcite was a breakthrough @xcite .
the most significant feature to be stressed is that chiral anomaly can be devised even with a finite lattice . in the lscher s formulation ,
new chiral transformation is introduced and chiral anomaly is obtained from jacobian of the fermion measure in a different way from the continuum theory .
the lattice index theorem holds for arbitrary lattice spacing @xcite and chiral anomaly agrees with the continuum result in the continuum limit @xcite .
the lesson to be learned is that one can modify the axial current in order to reproduce the correct chiral anomaly on the finite lattice . in the electroweak theory , left- and
right - handed fermions couple to gauge fields in different ways .
this type of formulation is called chiral gauge theory . for consistent construction of chiral gauge theories
, gauge symmetry needs to be maintained at the quantum level ( see ref .
@xcite , for example ) .
introduction of wilson terms has trouble because the mixing of left- and right - handed fermions complicates discussion of gauge anomaly cancelation @xcite .
slac fermion partially simplifies the problem because it does not use wilson terms @xcite .
however , it has not been successful due to breakdown of locality and lorentz invariance associated with axial currents in the continuum limit @xcite
. there have been discussions that deny these defects and the non - conservation of the axial current @xcite .
the defects originate in the definition of axial currents with a derivative different from the slac derivative .
careful and consistent treatment is necessary when discussing species doubler and chiral anomaly with slac fermion . in this paper , we give a method to save the slac derivative . on a finite lattice ,
a derivative is defined as a discrete fourier transform of momentum in a similar way to the conventional slac derivative . to remove doubler modes , antiperiodic boundary conditions
are chosen for the derivative in real space .
the doubler modes at the momentum boundary are lifted and the dispersion relation of the continuum theory is reproduced discretely .
although long - range hopping interactions appear , the pathology of slac fermion is avoided and the correct continuum limit is guaranteed .
however , long - range interactions are not useful in practical numerical calculations because sparser dirac operator is better for numerical efficiency . by using the lanczos factor technique demonstrated in ref .
@xcite , we effectively truncate long - range hopping interactions and improve the fermion propagator . since the proposed lattice derivative does not use wilson terms , left- and right - handed fermions are completely independent . as a result ,
the dirac operator constructed with the derivative has exact chiral symmetry .
this means that the dirac fermion does not reproduce chiral anomaly because the fermion measure of path - integral is trivially invariant under the chiral transformation on the finite lattice . to simulate the correct chiral anomaly
, we modify chiral transformation using the neuberger s solution to the ginsparg - wilson relation @xcite . as a result
, we obtain an anomalous ward identity for the modified chiral transformation .
the zero mode of the axial current divergence is generally non - zero and gives the index theorem .
the absence of chiral anomaly under the usual chiral transformation implies the existence of a single anomaly - free weyl fermion .
chiral gauge theories can be constructed using the single weyl fermion as a building block .
this work is a tricky extension of slac fermion . in the conventional slac fermion ,
periodic boundary conditions and an infinite lattice are assumed , where the doubler remains as a singularity at the momentum boundary . on the other hand ,
our formulation is based on antiperiodic boundary conditions and a finite lattice and therefore free from species doubler .
in addition , the axial current is defined with the modified chiral transformation and evaluated nonperturbatively .
the nielsen - ninomiya theorem does not apply to our implementation of chiral anomaly because the modified chiral transformation is used to define a symmetry .
this paper is organized as follows . in sec .
[ lattice_derivative ] , the lattice derivative is defined based on the finite lattice formulation .
locality of the derivative is discussed .
long - range hopping interactions are truncated using the lanczos factor . in sec .
[ chiral_anomaly ] , a modified chiral transformation is introduced to reproduce the correct chiral anomaly . a way of constructing gauge - invariant chiral gauge theories is given in a nonperturbative way .
[ sumamry_and_discussions ] is devoted to summary and discussions .
we define lattice derivative as a discrete fourier transform of momentum @xmath1 where @xmath2 represents lattice sites and takes integer values between @xmath3 and @xmath4 .
the lattice size @xmath5 is a finite even number and @xmath6 corresponds to momentum .
@xmath7 antiperiodic boundary conditions have been chosen in real space , @xmath8 . in eq .
( [ nabla ] ) , the summation can be carried out easily .
@xmath9,\ ] ] where @xmath10 . the lattice derivative ( [ nabla ] ) is doubler - free and reproduces discretely the dispersion relation of the continuum theory for arbitrary lattice spacing . in the large @xmath5 limit , eq .
( [ nabla ] ) becomes an integral @xmath11 where @xmath12 is a lattice spacing and @xmath13 is a space coordinate . in the continuum limit @xmath14
, we obtain the first order derivative of the continuum theory . @xmath15 the lattice derivative ( [ nabla ] )
is local in the continuum limit .
figure [ fig1 ] plots the lattice derivative @xmath16 with @xmath10 in a range @xmath17 for two lattice sizes @xmath18 and @xmath19 .
the absolute value of @xmath20 is large around @xmath21 and @xmath22 and small around @xmath23 .
the points @xmath21 and @xmath22 are equivalent because of antiperiodicity .
therefore , @xmath20 around @xmath24 does not mean severe nonlocality .
the points @xmath23 give the most long - range hopping . as expected from eq .
( [ nabla3 ] ) , locality of the derivative @xmath20 is quite good when @xmath25 . on the other hand , decay of the derivative @xmath20
is slow when the lattice size is small .
we are interested in constructing a theory with better locality for a practical purpose . in order for eq .
( [ nabla ] ) to be a useful lattice derivative , long - range hopping interactions need to be truncated . however , truncation of hopping interactions may cause errors .
we need to find a systematic way to reproduce spectra effectively with only short - range hopping interactions . for simplicity ,
let us consider a classical action for a free massless fermion in one - dimensional space .
@xmath26 fourier transforms of the one - component fermions are @xmath27 where antiperiodicity is assumed @xmath28 then we have @xmath29 where @xmath30 with the inverse transform @xmath6 of eq .
( [ nabla ] ) @xmath31 to obtain this , antiperiodicity of @xmath32 , @xmath33 , and @xmath34 has been used .
antiperiodicity in real space gives rise to periodicity in momentum space , @xmath35 , @xmath36 , and @xmath37 .
some explanations would be necessary for the derivation of eq .
( [ pl0 ] ) .
consider the matrix contained in eq .
( [ pll1 ] ) @xmath38 which has periodicity , @xmath39 . in fig .
[ fig2 ] , the matrix @xmath40 is shown schematically ( see the upper diagram ) . some examples for the indices @xmath41
are given .
the matrix elements on the dotted lines are not contained in eq .
( [ pll1 ] ) . in the upper diagram , the vertices @xmath42 and @xmath43 correspond to the points with @xmath24 in fig .
[ fig2 ] . using the periodicity of the matrix ,
the triangles 1 and 2 can be moved to form the parallelogram ( see the lower diagram ) . as a result ,
( [ pll1 ] ) can be evaluated by calculating the summation for each @xmath44 , which draws a line segment parallel to the oblique sides of the parallelogram .
the last term of eq .
( [ pl0 ] ) is a contribution of the term with @xmath45 , which corresponds to the left oblique side of the parallelogram . in eq .
( [ pl0 ] ) , we truncate long - range terms with a parameter @xmath46 , which represents the largest distance of fermion hopping . @xmath47 . \label{pl1}\ ] ] the truncation can be implemented to eq .
( [ pll1 ] ) .
@xmath48 the indices @xmath49 and @xmath2 run from @xmath3 to @xmath4 .
the prime symbol means that the summation is taken over @xmath50 , @xmath51 , and @xmath52 .
fermion hopping has been restricted to a finite range .
figure [ fig3 ] plots @xmath6 of eq .
( [ pl1 ] ) as a function of @xmath53 for @xmath54 , @xmath55 , and @xmath56 with a lattice size @xmath18 .
@xmath57 gives the exact result with no truncation , which satisfies the dispersion relation of the continuum theory .
if one does not mind inclusion of long - range hopping , doubler - free formulation of a single weyl fermion is possible maintaining the correct dispersion relation .
when @xmath58 is small , there is also no doubler modes because of antiperiodic boundary conditions .
although some modes around the momentum boundary deviate from the correct dispersion , those are not so harmful because there is no genuine degeneracy with low - lying modes .
however , @xmath6 oscillates around the exact result .
the oscillation tends to be large as @xmath58 goes to small .
the small oscillation around the correct dispersion comes from the truncated terms having large @xmath2 s in eq .
( [ pl1 ] ) . as shown in ref .
@xcite , such oscillation can be removed by introducing the lanczos factor , which is used in fourier analysis to cancel the gibbs phenomenon .
@xcite we modify eq .
( [ pl1 ] ) as follows : @xmath59 .
\label{pl2}\ ] ] where @xmath60 is the lanczos factor . as a result ,
( [ pll2 ] ) becomes @xmath61 the final form of the action is given as @xmath62 figure [ fig4 ] plots @xmath6 of eq .
( [ pl2 ] ) improved with the lanczos factor as a function of @xmath53 for @xmath54 , @xmath55 , and @xmath56 with @xmath18 , which are compared with the exact result ( @xmath57 ) shown in fig .
[ fig3 ] . as before , there is no doubler for every @xmath58 .
in addition to this , the oscillation has been removed with the lanczos factor .
as @xmath58 goes to large , the deviation around the momentum boundary tends to be small . in this way
, we can construct a doubler - free ultralocal derivative .
if the lanczos factor is introduced , ultralocal formulation of a single weyl fermion is possible maintaining almost correct dispersion relation .
on the finite four - dimensional euclidean lattice , consider an effective action @xmath63 $ ] @xmath64}=\int { \cal d}\psi { \cal d}\bar{\psi } e^{-s[u ] } , \label{ea}\ ] ] where @xmath65 is a classical action for a massless dirac fermion coupled to gauge .
@xmath66 is a dirac operator and the euclidean dirac matrices satisfy @xmath67 and @xmath68 .
chirality is defined with @xmath69 .
the fermion variables @xmath70 and @xmath33 are grassmann valued .
the indices @xmath49 and @xmath2 are four - component numbers to indicate lattice sites and each component runs from @xmath3 to @xmath4 as before .
the lattice covariant derivative @xmath71 is diagonal with respect to the space - time indices @xmath49 and @xmath2 except for the @xmath72-th ones .
the dirac operator @xmath73 is antihermitian and satisfies @xmath74 .
the classical action is invariant under the usual chiral transformation .
the derivative @xmath32 can be replaced with the truncated one in the same way as eq .
( [ action1 ] ) , if ultralocal construction is preferred . the gauge variable @xmath75 is a product of all link variables that compose a line segment between the two sites @xmath49 and @xmath2 parallel to the @xmath72-th direction .
when connecting the two sites with link variables along the @xmath72-th direction , there are two ways because of periodicity of the action .
the most natural choice is the shorter path .
one of the two ways is chosen depending on the distance between two sites . when @xmath76 , @xmath77 which corresponds to the hexagon sandwiched between the triangles 1 and 2 in fig .
@xmath78 is a unit vector in the @xmath72-th direction .
when @xmath79 , @xmath80 which corresponds to the triangles 1 and 2 in fig .
[ fig2 ] and therefore intersects the boundary .
the link variables @xmath81 are elements of a gauge group and satisfy @xmath82 and @xmath83 . in the continuum limit ,
the lattice covariant derivative becomes @xmath84 where @xmath85 and @xmath86 , and @xmath87 with a parameterization @xmath88 .
( [ lcd ] ) reduces to the covariant derivative of the continuum theory in the continuum limit . with our dirac operator
, the usual chiral transformation does not reproduce chiral anomaly because the classical action and the fermion measure is invariant . to simulate the correct chiral anomaly on the finite lattice ,
we introduce the modified chiral transformation with @xmath89 @xmath90 where @xmath91 the operator @xmath92 is the neuberger s solution @xcite to the ginsparg - wilson relation @xmath93 and has nothing to do with the dirac operator ( [ lcd ] ) .
the modified chiral transformation is a symmetry of the effective action @xmath94 for arbitrary lattice spacing independent of whether the transformation is local or global .
when the transformation is global @xmath95 , it is also a symmetry of the classical action @xmath96 in the continuum limit @xmath14 because the variation of the lagrangian density induced by the transformation is proportional to lattice spacing . @xmath97
although the variation is classically zero in the continuum limit , the vacuum expectation value of the variation gives the index theorem for arbitrary lattice spacing .
the axial current divergence is defined as a variation of the classical action under the local chiral transformation with @xmath98 .
@xmath99.\end{aligned}\ ] ] the axial current is obtained by inverting the derivative @xmath100,\end{aligned}\ ] ] which is gauge invariant . in the free theory ,
the currents are local in the continuum limit because the leibniz rule holds in the limit .
the fermion measure transforms as @xmath101 chiral anomaly @xmath102 is a gauge - invariant quantity . as shown in ref .
@xcite , the index theorem holds for arbitrary lattice spacing .
@xmath103 ( see also ref . @xcite . ) since the effective action @xmath94 is invariant under transformation of the integration variables , the modified axial current @xmath104 does not conserve . @xmath105
the ward identity for the modified chiral transformation holds also for the zero mode , which gives the index theorem and corresponds to the variation under the global transformation ( [ var ] ) . ) is necessary for the existence of the non - vanishing zero mode of the axial current divergence . ] in the continuum limit , the chiral anomaly agrees with the continuum result @xcite @xmath106 where @xmath107 and @xmath108 $ ] .
the effect of chiral anomaly can be implemented to physical quantities via the modified anomalous axial current .
construction of anomaly - free chiral gauge theory is easy if our lattice derivative is used . in our formulation ,
weyl fermions are defined with the ordinary @xmath109 ( not @xmath98 ) . as a result ,
the fermion measure of the path integral do not depend on gauge variables .
the weyl fermions are free from gauge anomaly beforehand . consider a classical action for a dirac fermion @xmath110 where only the right - handed fermion is coupled to gauge .
the left - handed fermion is redundant and does not contribute to the effective action .
once the absence of gauge anomaly is confirmed , the left - handed fermion can be integrated out .
the definition of the dirac operator @xmath111 is same as eq .
( [ lcd ] ) except that gauge is defined as a product of link variables @xmath112 where @xmath113 and @xmath114 .
@xmath111 is no longer antihermitian . under local gauge transformation ,
the classical action ( [ action2 ] ) is invariant . on the finite lattice ,
infinitesimal local gauge transformation @xmath115 does not change the fermion measure @xmath116 { \cal d}\psi { \cal d}\bar{\psi } = { \cal d}\psi { \cal d}\bar{\psi}\ ] ] and hence the effective action .
a single weyl fermion can exist on the lattice without violating gauge symmetry .
we have constructed a doubler - free covariant derivative and an anomalous ward identity for the modified chiral symmetry on the lattice .
the index theorem holds for arbitrary lattice spacing and the dependence of chiral anomaly on gauge fields agrees with the continuum result in the continuum limit .
the zero mode of the anomalous ward identity gives the index theorem . in this formulation ,
a single weyl fermion can exist on the lattice maintaining gauge symmetry .
chiral gauge theories can be constructed nonperturbatively using the single anomaly - free weyl fermion as a building block .
the proposed lattice derivative is a simple matrix that gives the correct dispersion .
we have shown that introduction of the lanczos factor enables us to construct lattice derivatives with good locality . in this case
, almost the correct dispersion is reproduced except for deviation at high momentum .
it depends on fermion mass how large the truncation parameter @xmath58 should be .
the correct chiral anomaly has been obtained without violating locality and lorentz invariance in the continuum limit in spite of the existence of long - range hopping interactions .
this is a result of a nonperturbative formulation based on the modified chiral transformation ( or equivalently the modified axial current ) . to reproduce chiral anomaly , wilson terms have to be introduced somewhere .
we have introduced them only in the modified chiral transformation to maintain complete independence of left- and right - handed fermions in the classical action .
when calculating physical quantities dependent on chiral anomaly such as the @xmath117 meson mass , @xmath109 needs to be replaced with @xmath98 in the concerned vertex operators to include the effect of chiral anomaly .
it depends on the construction of the operator @xmath92 how precisely the effect of chiral anomaly is implemented with finite lattice spacing . in lattice qcd , the modification of the axial current does not affect physical quantities independent of chiral anomaly because the axial current does not couple to gauge directly .
this research was supported in part by riken
. 0 k. fujikawa , phys .
lett . * 42 * , 1195 ( 1979 ) ; phys .
d * 21 * , 2848 ( 1980 ) ; phys . rev .
d * 22 * , 1499 ( 1980 ) .
j. kogut and l. susskind , phys .
d * 11 * , 395 ( 1975 ) ; s. d. drell , m. weinstein , and s. yankielowicz , phys . rev .
d * 14 * , 487 ( 1976 ) . k. g. wilson , in _ new phenomena in subnuclear physics _ , erice , 1975 , edited by a. zichichi ( plenum , new york , 1977 ) .
d. b. kaplan , phys .
b * 288 * , 342 ( 1992 ) ; m. f. golterman , k. jansen and d. b. kaplan , phys .
b * 301 * , 219 ( 1993 ) .
y. shamir , nucl .
b * 406 * , 90 ( 1993 ) ; v. furman and y. shamir , nucl .
b * 439 * , 54 ( 1995 ) .
h. neuberger , phys .
b * 417 * , 141 ( 1998 ) ; _ ibid _ * 427 * , 353 ( 1998 ) .
h. b. nielsen and m. ninomiya , phys .
b * 105 * , 219 ( 1981 ) ; nucl .
phys . b 185 , 20 ( 1981)[e : b 195 , 541 ( 1982 ) ] ; b 193 , 173 ( 1981 ) . l. h. karsten and j. smit , nucl .
b * 144 * , 536 ( 1978 ) ; phys .
b * 85 * , 100 ( 1979 ) .
l. h. karsten and j. smit , nucl .
b * 183 * , 103 ( 1981 ) .
e. seiler and i. o. stamatescu , phys .
d * 25 * , 2177 ( 1982 ) [ erratum - ibid .
d * 26 * , 534 ( 1982 ) ] .
p. h. ginsparg and k. g. wilson , phys .
d * 25 * , 2649 ( 1982 ) .
m. lscher , phys .
b * 428 * , 342 ( 1998 ) .
p. hasenfratz , v. laliena and f. niedermayer , phys .
b * 427 * , 125 ( 1998 ) m. lscher , nucl . phys .
b * 549 * , 295 ( 1999 ) ; * 568 * , 162 ( 2000 ) .
y. kikukawa and a. yamada , phys .
b * 448 * , 265 ( 1999 ) .
k. fujikawa , nucl .
b * 546 * , 480 ( 1999 ) .
h. suzuki , prog .
* 102 * , 141 ( 1999 ) .
d. h. adams , annals phys . * 296 * , 131 ( 2002 ) .
r. a. bertlmann , _ anomalies in quantum field theory _
( oxford university press , oxford , 1996 ) .
m. lscher , arxiv : hep - th/0102028 .
k. melnikov and m. weinstein , phys . rev .
d * 62 * , 094504 ( 2000 ) j. m. rabin , phys . rev .
d * 24 * , 3218 ( 1981 ) .
m. ninomiya and c. i. tan , phys .
lett . * 53 * , 1611 ( 1984 ) .
t. sugihara , phys .
d * 68 * , 034502 ( 2003 ) .
g. b. arfken , _ mathematical methods for physicists _
( academic press , new york , 1970 ) .
t. w. chiu , phys .
d * 58 * , 074511 ( 1998 ) k. fujikawa , phys .
d * 60 * , 074505 ( 1999 ) e. witten , phys .
b * 117 * , 324 ( 1982 ) . | we combine a pair of independent weyl fermions to compose a dirac fermion on the four - dimensional euclidean lattice .
the obtained dirac operator is antihermitian and does not reproduce anomaly under the usual chiral transformation . to simulate the correct chiral anomaly ,
we modify the chiral transformation .
we also show that chiral gauge theories can be constructed nonperturbatively with exact gauge invariance .
the formulation is based on a doubler - free lattice derivative , which is a simple matrix defined as a discrete fourier transform of momentum with antiperiodic boundary conditions .
long - range fermion hopping interactions are truncated using the lanczos factor . #
10= 0=0 1= 1=1 0>1 # 1 / |
sobolev spaces have been playing a major role in the study of nonlinear partial differential equations and calculus of variations for more than seven decades now .
variational approach to mappings between manifolds leads in a natural way to the notion of sobolev mappings between manifolds @xmath0 . here
the symbol @xmath1 stands for the sobolev space of @xmath2 integrable mappings with @xmath2 integrable first order derivatives .
this class of mappings was considered for the first time as early as in 1948 by morrey @xcite in his celebrated paper on regularity of minimizing harmonic mappings from two dimensional surfaces into manifolds . while it is a well known theorem of meyers and serrin that smooth functions are dense in the sobolev space @xmath3 , where @xmath4 is a riemannian manifold , similar question , raised by eells and lemaire , @xcite , about density of smooth mappings in the class of sobolev mappings between manifolds
is no longer obvious .
actually schoen and uhlenbeck @xcite , @xcite were the first to observe that in general the answer to this question is in the negative .
namely they demonstrated that the radial projection mapping @xmath5 , @xmath6 , can not be approximated by @xmath7 mappings when @xmath8 . in the same paper they proved that @xmath9 is a dense subset of @xmath0 when @xmath10 and @xmath4 , @xmath11 are compact smooth manifolds , @xmath12 .
later , mainly due to the work of bethuel , @xcite , it turned out that the condition for the density of smooth mappings can be formulated in terms of algebraic topology describing the topological structure of the manifolds @xmath4 and @xmath11 .
finally a necessary and sufficient condition for the density of smooth mappings in the space of sobolev mappings @xmath0 was discovered by hang and lin @xcite , @xcite .
this result corrects an earlier statement of bethuel @xcite about a necessary and sufficient condition for the density .
other papers on this topic include @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite .
the theory of sobolev mappings between manifolds has been extended to the case of sobolev mappings with values into metric spaces .
the first papers on this subject include the work of ambrosio , @xcite , on limits of classical variational problems and the work of gromov and schoen , @xcite , on sobolev mappings into the bruhat - tits buildings , with applications to rigidity questions for discrete groups .
later the theory of sobolev mappings with values into metric spaces was developed in a more elaborated form by korevaar and schoen @xcite in their approach to the theory of harmonic mappings into alexandrov spaces of non - positive curvature .
other papers on sobolev mappings from a manifold into a metric space include @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite .
finally analysis on metric spaces , the theory of carnot
carathodory spaces and the theory of quasiconformal mappings between metric spaces lead to the theory of sobolev mappings between metric spaces , @xcite , @xcite , @xcite . in the case of sobolev mappings into metric spaces heinonen , koskela , shanumgalingam and tyson , ( * ? ? ?
* remark 6.9 ) , ask for the conditions that would guarantee the density of lipschitz mappings .
our paper is devoted to study this problem . from the point of view of geometric analysis one
is willing to identify spaces that are bi - lipschitz equivalent .
the main result of the paper , theorem [ main ] , shows that if we change the metric in the target space to a bi - lipschitz equivalent metric , than the property of the density of lipschitz mappings may be lost .
that shows a major difficulty in attempts to develop the theory in the case of metric space valued maps .
other main results in the paper are theorems [ one ] , [ dens ] , [ subaru ] , [ smooth ] .
all the main results but theorem [ dens ] are counterexamples .
let @xmath4 and @xmath11 be two smooth compact manifolds , @xmath11 without boundary .
assume that @xmath11 is smoothly embedded into the euclidean space @xmath13 .
then for @xmath14 we define the class of sobolev mappings between manifolds @xmath4 and @xmath11 as @xmath15 the space @xmath0 is equipped with the metric of the sobolev norm @xmath1 i.e. @xmath16 for @xmath17 . by the theorem of meyers and serrin
every sobolev mapping @xmath18 can be approximated by smooth mappings @xmath19 , but it is not always possible to choose the approximating sequence to have values in @xmath11 .
clearly the space @xmath0 depends on the embedding of @xmath11 into the euclidean space .
however , if @xmath20 and @xmath21 are two different smooth embeddings then the mapping @xmath22 is a homeomorphism of the spaces @xmath23 and @xmath24 .
in particular the answer to the question about density of smooth mappings in @xmath0 does not depend on the smooth embedding of @xmath11 .
there are various possibilities to define metric space valued sobolev mappings , see e.g. @xcite , @xcite , @xcite .
as was proved by heinonen , koskela , shanmugalingam and tyson @xcite , in a reasonable generality different approaches are equivalent .
the results in @xcite concern sobolev mappings between metric spaces , but in our paper we will consider the case of mappings from a closed manifold to metric spaces @xmath25 only .
one of the definitions of such mappings mimics the definition ( [ betweenman ] ) .
first one defines sobolev mappings from @xmath4 into banach spaces .
then one considers an isometric embedding of @xmath26 into a banach space @xmath27 ( any metric space @xmath26 admits an isometric embedding into @xmath28 ) and defines @xmath25 as the set of all mappings in @xmath29 such that values of the mapping belong to @xmath30 . for a precise definition and proof of the equivalence with other approaches ,
see @xcite .
the space @xmath25 inherits the metric from the banach structure of the space @xmath29 .
then under reasonable assumptions every sobolev mapping @xmath31 can be approximated by lipschitz mappings @xmath32 , ( * ? ? ?
* theorem 6.7 and remark 6.6 ) .
the authors ask : _ it is an interesting problem to determine when one can choose the lipschitz approximation to have values in the target @xmath26 . [ ] for instance , one can ask to what extent bethuel s results have analogs for general spaces . _ in the case of sobolev mappings between riemannian manifolds @xmath0 it is often very important to have a riemannian metric on @xmath11 fixed as changing the metric could result in turning a given variational problem to a different one . in the case of mappings into metric spaces @xmath25 it is often that there is no canonical way to define the metric in @xmath26 and there are several possibilities which are in some sense equally good
. one of the requirements is , however , that the metrics be bi - lipschitz equivalent . actually , from the point of view of analysis on metric spaces , spaces which are bi - lipschitz homeomorphic are equivalent .
a fundamental example in this context is the heisenberg group @xmath33 .
there are two canonical metrics on the heisenberg group .
the carnot
carathodory metric @xmath34 and the gauge metric @xmath35 .
they are bi - lipschitz equivalent i.e. there is a constant @xmath36 such that for all @xmath37 @xmath38 capogna and lin , @xcite study sobolev mappings from a domain @xmath39 in the euclidean space into the heisenberg group .
they write : _ because of ( [ equiv ] ) both the definitions of the lebesgue space @xmath40 , and the sobolev space @xmath41 do not depend on the choice between the carnot
carathodory distance and the one coming from the gauge norm . to simplify the computations
we will deal with the metric @xmath42 .
_ they consider the space of sobolev mappings with values into the heisenerg group as a set and they _ do not _ equip this space with any metric .
they are perfectly right when they say that the set does not depend on the choice of the metric in the target as long as the metrics that are considered are bi - lipschitz equivalent , like those in ( [ equiv ] ) . however , if one equips the space of the sobolev mappings @xmath25 from a manifold into a metric space with a metric that comes from the isometric embedding of @xmath26 into a banach space ( as was done in @xcite ) it is not clear how changing the metric in @xmath26 to a bi - lipschitz equivalent one would affect the metric structure of the space @xmath25 .
our paper is devoted to study this problem .
easy examples show that different isometric embeddings of @xmath26 into banach spaces may give different metrics in @xmath25 equipped with the norm @xmath43 is a banach space and @xmath44 , @xmath45 if @xmath46 , @xmath47 if @xmath48 is an isometric embedding .
it is easy to see that this embedding equips the space @xmath49,\bbbr)$ ] with a metric which is different than the standard one . ] and it is not even clear what is the answer to the following question .
* question 1 . *
_ does the answer to the question about the density of lipschitz mappings in @xmath25 depend on the isometric embedding of @xmath26 into a banach space ?
_ it is easy to see that in the case of isometric embeddings of @xmath26 into hilbert spaces the answer to question 1 is no .
since we will be concerned with metric spaces @xmath26 that admit isometric embeddings to euclidean spaces , we will be concerned with question 2 ( formulated below ) rather than with question 1 which is obvious in this case .
as we assume @xmath26 to have an isometric embedding to the euclidean space , we can actually identify it with a subset of @xmath13 and define @xmath50 the space @xmath25 is equipped with the metric of the sobolev norm @xmath1 .
two closed manifolds can be identified if they are diffeomorphic . in the case of metric spaces
a natural counterpart of a diffeomorphism is a bi - lipschitz homeomorphism .
this is the class of the `` most smooth '' homeomorphisms that can be considered for arbitrary metric spaces .
many of the metric spaces @xmath26 admit a bi - lipschitz embedding into the euclidean space @xmath13 , rather than an isometric one .
in such a case one could try to use this embedding to define the space @xmath25 by ( [ 719 ] ) .
how does the space @xmath25 depend on the bi - lipschitz embedding of @xmath26 ?
* question 2 .
* _ assume that @xmath26 and @xmath51 are compact subsets of @xmath13 that are bi - lipschitz homeomorphic .
assume that @xmath4 is a smooth closed @xmath52-dimensional manifold and that lipschitz mappings @xmath53 are dense in @xmath25 for some @xmath14 .
are the lipschitz mappings @xmath54 dense in @xmath55 ?
_ we will examine the case of sobolev mappings @xmath56 from an @xmath52-dimensional closed manifold to a compact set @xmath57 .
we will prove that if @xmath26 is a lipschitz neighborhood retract then lipschitz mappings @xmath53 are dense in @xmath56 , see theorem [ dens ] .
now if @xmath51 is bi - lipschtz homeomorphic to @xmath26 , then @xmath54 is dense in @xmath58 because the class of lipschitz neighborhood retracts is closed under bi - lipschtz homeomorphisms ( lemma [ lnr ] ) and hence @xmath51 is a lipschitz neighborhood retract as well .
on the other hand we will show in theorem [ main ] that there are two bi - lipschitz homeomorphic compact subsets @xmath26 and @xmath51 of @xmath59 such that for any @xmath52-dimensional closed manifold @xmath4 lipschitz mappings @xmath53 are dense in @xmath56 , but lipschitz mappings @xmath54 are not dense in @xmath58 .
obviously @xmath26 can not be a lipschitz neighborhood retract .
other new results are theorems [ one ] , [ subaru ] and [ smooth ] .
in all theorems but theorem [ one ] we deal with the sobolev space @xmath56 , where @xmath60 .
one can ask similar questions for spaces @xmath25 with @xmath61 .
i believe that our counterexamples have counterparts for other exponents as well
. it would be , however , more interesting if one could prove positive results for some @xmath61 .
* statement of main results .
* consider the class of sobolev mappings @xmath62 and a bounded lipschitz function @xmath63 . here
@xmath64 is an open set with finite measure , but we could also consider a closed manifold @xmath4 instead of @xmath65 .
it is well known and easy to prove that the nonlinear operator @xmath66 is bounded i.e. it maps bounded sets to bounded sets .
more precisely the inequality @xmath67 is satisfied .
the same holds in the case in which @xmath65 is replaced by @xmath4 .
* question 3 .
* _ is the operator @xmath68 continuous ? _
this is a crucial problem for us . indeed
, continuity would immediately imply that @xmath69 is a homeomorphism whenever @xmath70 is a bi - lipschitz homeomorphism and @xmath71 .
marcus and mizel @xcite proved continuity of @xmath68 given by ( [ tuptus ] ) in the case @xmath72 .
we are , however , interested mainly in the case in which @xmath73 .
let us start with a positive result that will be needed later .
[ jeden ] let @xmath4 be a smooth closed manifold and @xmath74 a bounded lipschitz function on @xmath13 that is of the class @xmath75 away from one point @xmath76 .
then the operator @xmath77 is continuous for all @xmath78 .
unfortunately , in general , the answer to question 3 is in the negative when @xmath73 .
this is shown by the following result .
i believe the result is known , but i was unable to locate it in the literature . in any case
the proof is very easy and the phenomenon is so important for our paper that we provide a short proof .
[ one ] there is a lipschitz function @xmath79 with compact support such that the bounded operator @xmath80,\bbbr^2)\to w^{1,p}([0,1])$ ] defined as composition @xmath71 is not continuous for any @xmath81 .
similar claim follows from theorem [ main ] below .
indeed , the mapping @xmath82 induced by the bi - lipschitz homeomorphism @xmath70 from theorem [ main ] can not be continuous .
however , the proof of theorem [ one ] is much simpler than that of theorem [ main ] . even if we know
that @xmath83 is not continuous for a bi - lipschitz homeomorphism @xmath70 it does not immediately imply that the spaces @xmath25 and @xmath55 must have different properties .
in fact it easily follows that if @xmath70 is a bi - lipschitz homeomorphism then the corresponding mapping @xmath68 gives a one - to - one correspondence between the spaces @xmath25 and @xmath55 and one - to - one correspondence between the spaces @xmath53 and @xmath54 .
hence one could expect positive answer to question 2 .
the following result answers question 2 in the positive in the case in which @xmath84 and @xmath26 is a lipschitz neighborhood retract .
[ dens ] let @xmath4 be a smooth closed @xmath52-dimensional manifold and @xmath57 a compact lipschitz neighborhood retract
. then lipschitz mappings @xmath53 are dense in @xmath56 .
recall that @xmath57 is a _ lipschitz neighborhood retract _ if there is an open set @xmath85 and a lipschitz mapping ( called _ lipschitz retraction _ ) @xmath86 such that @xmath87 for all @xmath88 .
the following result is well known .
[ lnr ] if @xmath57 is a lipschitz neighborhood retract and @xmath89 is bi - lipschiz homeomorphic to @xmath26 , then @xmath51 is a lipschitz neighborhood retract as well . _
_ if @xmath86 is a lipschitz retraction , @xmath90 a bi - lipschitz homeomorphism and @xmath91 is a lipschitz extension ( mcshane ) of @xmath92 , then @xmath93 is a lipschitz retraction .
we say that @xmath57 is an @xmath95-dimensional _ lipschitz submanifold _ if every point @xmath88 has a neighborhood in @xmath26 which is bi - lipschtz homeomorphic to @xmath96 .
theorem [ dens ] applies to lipschitz submanifolds because of the following result .
( ( * ? ? ?
* theorem 5.13 ) ) lipschitz submanifolds of @xmath13 are lipschitz neighborhood retracts . _
_ we sketch the proof which is based on a localization of the argument of almgren @xcite ( cf .
* proposition 2.13 ) ) .
let @xmath57 be a lipschitz submanifold .
we decompose @xmath97 into whitney cubes and define the lipschitz retraction first on the vertices of the cubes as the projection to a closest point in @xmath26 .
we take only those whitney cubes which are in a close neighborhood of @xmath26 , as we require that all the vertices of each cube are mapped into a domain in @xmath26 which is bi - lischitz homeomorphic to a ball .
next we extend the lipschitz projection to higher dimensional edges of the cubes by induction .
@xmath94 schoen and uhlenbeck @xcite , @xcite proved theorem [ dens ] in the case in which @xmath26 is a compact smooth manifold .
they proved that the convolution approximation of @xmath98 takes values into a tubular neighborhood of @xmath26 in @xmath13 and hence one can obtain desired approximation by composing the convolution approximation with the smooth nearest point projection .
this argument can not be directly applied in the case of theorem [ dens ] , even when @xmath26 is a lipschitz manifold , because now the retraction of a neighborhood of @xmath26 onto @xmath26 is lipschitz only and according to theorem [ one ] the composition with a lipschitz function need not be continuous in the sobolev norm .
proof of theorem [ dens ] will require a much more delicate approximation method than the one coming from convolution .
a nice and nontrivial class of examples of lipschitz manifolds was provided by toro , @xcite ( see also @xcite ) .
she proved that surfaces in @xmath99 that are locally graphs of @xmath100 functions are lipschitz submanifolds .
if we replace the class of @xmath100 surfaces by continuous @xmath101 surfaces , then we can construct a counterexample to the density .
such a counterexample is provided by the case @xmath102 of the following result .
[ subaru ] there is a subset @xmath103 homeomorphic to @xmath104 and @xmath105 such that @xmath106 is a smooth submanifold of @xmath107 and @xmath11 is a graph of a continuous @xmath108 function near @xmath76 .
that means @xmath11 is a regular @xmath52-dimensional submanifold of @xmath107 with one singular point .
this construction can be done in such a way that for every closed @xmath52-dimensional manifold @xmath4 , lipschitz mappings @xmath109 are not dense in @xmath110 .
one should compare theorem [ subaru ] with theorem [ dens ] . as we will see
theorem [ subaru ] is a consequence of the proof of theorem [ smooth ] . namely we will see that if @xmath111 , then the singular manifold @xmath112 constructed in the proof of theorem [ smooth ] has properties required in theorem [ subaru ]
. if @xmath57 is a subset , then by smooth mappings @xmath113 we will mean the class of all mappings @xmath114 such that @xmath115 .
[ main ] fix an integer @xmath116 .
there is a compact and connected set @xmath117 and a global bi - lipschitz homeomorphism @xmath118 with the property that for any closed @xmath52-dimensional manifold @xmath4 smooth mappings @xmath113 are dense in @xmath56 , but lipschitz mappings @xmath54 are not dense in @xmath58 , where @xmath119 . according to theorem [ dens ] and lemma [ lnr ] , @xmath26 can not be a lipschitz neighborhood retract .
in particular there is no such example as in theorem [ main ] in the case in which @xmath26 is a closed manifold .
however , if we allow a global homeomorphism @xmath68 of the euclidean space to be locally bi - lipschitz everywhere but in one point , then one can find counterexamples even in the class of closed manifolds .
[ smooth ] let @xmath120 be a closed @xmath52-dimensional manifold .
then there is a homeomorphism @xmath121 which is a diffeomorphism in @xmath122 , which is identity outside a sufficiently large ball and has the property that lipschitz mappings @xmath123 are not dense in @xmath124 , where @xmath125 . the example constructed in the proof of theorem [ smooth ] will be employed in the proof of theorem [ main ]
. * notation .
* notation employed in the paper is rather standard .
the @xmath126-dimensional hausdorff measure will be denoted by @xmath127 and the average value by @xmath128 symbol @xmath129 will be used to denote a ball . by @xmath130 ,
where @xmath131 we will denote a ball concentric with @xmath129 and with the radius @xmath132 times that of @xmath129 .
we say that @xmath4 is a closed manifold if it is smooth , compact and without boundary .
the @xmath2 norm of a function @xmath133 will be denoted by @xmath134 .
the sobolev space @xmath3 consists of all @xmath135 such that @xmath136 .
the space is equipped with the norm @xmath137 .
by @xmath138 we will denote a general constant which can change its value even in the same string of estimates .
we will write @xmath139 to designate any function @xmath140 such that @xmath141 .
coordinates in @xmath142 will often be denoted by @xmath143 , where @xmath144
. * acknowledgement . *
i wish to express my deepest gratitude to the referee for the incredibly careful work .
referee s comments helped me a great deal to improve theorem [ dens ] and other parts of the paper as well .
for @xmath145 in @xmath146 and a.e . we will show that @xmath147 in @xmath1 .
we split @xmath4 into three pairwise disjoint sets @xmath148 , @xmath149 note that the measure of @xmath150 goes to @xmath151 as @xmath152 and hence for every @xmath153 there is @xmath126 such that @xmath154 because the family @xmath155 is equiintegrable as @xmath156 in @xmath2 .
in particular @xmath157 for all @xmath158 .
the chain rule applies to @xmath159 and to @xmath160 on @xmath161 when @xmath158 and hence @xmath162 in @xmath2 on @xmath161 .
sine @xmath92 is constant on @xmath163 , @xmath164 a.e . on @xmath163 and @xmath165 a.e . on @xmath163 .
hence @xmath166 in @xmath2 on @xmath163 .
thus @xmath167 in @xmath2 on @xmath163 .
this , however , implies that @xmath168 in @xmath2 on @xmath163 .
all the facts put together easily imply that @xmath162 in @xmath169 .
the proof is complete .
we will construct a bounded lipschitz function @xmath170 with compact support @xmath171\times [ 0,1]$ ] such that the bounded operator @xmath172,\bbbr^2 ) \to w^{1,p}([0,1],\bbbr ) , \qquad \phi(u)=\vi\circ u\ ] ] is not continuous for any choice of @xmath173 in the interval @xmath14 . to this end
we first define a compact set @xmath174\times [ 0,1]$ ] as union of an infinite family of segments defined below .
let @xmath175 for @xmath176 we define @xmath177 \quad \mbox{and } \quad i_\infty = \ { 2\ } \times [ 0,1].\ ] ] this is a family of vertical segments .
next for each @xmath176 we define a finite family of horizontal segments @xmath178 that join @xmath179 with @xmath180 namely we set @xmath181\times \ { k/2^i\}.\ ] ] now we define the set @xmath182 as follows @xmath183 the set @xmath182 is the union of boundaries an infinite family of squares .
now we define the function @xmath74 by the formula @xmath184\times [ 0,1]$,}\\ 0 & \mbox{if $ x\in\bbbr^2\setminus [ 0,2]\times [ 0,1]$. } \end{array } \right.\ ] ] the graph of @xmath74 over each of the squares has the shape of a pyramid , so the graph of @xmath74 is an infinite collection of pyramids .
the lipschitz constant of @xmath74 is @xmath185 .
fix @xmath14 .
we will show that the operator @xmath68 is not continuous in @xmath1 .
let @xmath186\to\bbbr^2 $ ] , @xmath187 .
the image of @xmath188 is the vertical segment that is in the middle between @xmath179 and @xmath180 .
obviously @xmath189 in @xmath1 as @xmath190 , where @xmath191\to\bbbr^2 $ ] , @xmath192 .
however , @xmath193 does not converge to @xmath194 in @xmath1 because @xmath195 a.e .
the proof is complete .
the approximation technique employed here is similar to that used in @xcite , @xcite . fix a riemannian tensor on the manifold @xmath4 .
denote the induced distance and the gradient by @xmath196 and @xmath197 respectively .
@xmath198 will be used to denote the hardy - littlewood maximal function
. there exists @xmath199 depending on the riemannian structure of @xmath4 such that the following inequalities @xmath200 and @xmath201 hold true whenever the radius @xmath202 of @xmath129 is less than @xmath199 and @xmath203 , @xmath14 .
the first inequality is known as the poincar inequality .
the second inequality follows e.g. from lemma 1.50 and theorem 1.32(i ) with @xmath204 in @xcite . as a corollary
one can easily prove the following version of a well known pointwise inequality , see e.g. @xcite , @xcite , ( * ? ? ?
* theorems 3.2 and 3.3 ) and references therein .
[ pointwise ] let @xmath4 be a closed manifold and @xmath205 , where @xmath14 .
then @xmath206 let @xmath207 .
our aim is to construct a family of lipschitz mappings @xmath208 such that 1 .
the lipschitz constant of @xmath209 is bounded by @xmath210 .
2 . @xmath211 as @xmath212 .
3 . @xmath213 as @xmath212 .
before we construct a family @xmath214 we will show how to use the properties ( a ) , ( b ) and ( c ) to complete the proof of the theorem .
let @xmath86 be a lipschitz neighborhood retraction . by ( c ) , @xmath215 for all sufficiently large @xmath216 and
hence @xmath217 is a family of lipschitz mappings well defined for large @xmath216 .
we will show now that @xmath218 in @xmath108 when @xmath212 .
indeed , @xmath218 a.e . by ( b ) and
hence @xmath218 in @xmath219 ( mappings are bounded ) .
finally ( a ) and ( b ) imply @xmath220 now we proceed to the construction of the family @xmath214 .
let @xmath221 .
the set @xmath222 is open .
there is a whitney decomposition of @xmath222 into balls and subordinated lipschitz partition of unity .
this is to say there is a constant @xmath223 depending on the riemannian structure of @xmath4 only and a sequence @xmath224 of points in @xmath222 such that with @xmath225 we have 1 .
@xmath226 ; 2 .
@xmath227 for all @xmath228 ; 3 . for every @xmath228 and
all @xmath229 we have @xmath230 ; 4 .
no point of @xmath222 belongs to more than @xmath138 balls @xmath231 ; 5 .
there is a family of lipschitz continuous functions @xmath232 such that @xmath233 , @xmath234 , @xmath235 and the lipschitz constant of @xmath236 is bounded by @xmath237 .
the proof of the existence of a whitney decomposition of a domain in @xmath142 into cubes with pairwise disjoint interiors can be found in @xcite .
standard modification of the method gives the above decomposition into balls .
actually the construction is so general that it can be carried out on arbitrary doubling metric spaces , see @xcite , ( * ? ? ?
* lemma 2.9 ) . in what follows
we will use notation @xmath238 and @xmath239 .
we define @xmath240 note that by lemma [ pointwise ] the function @xmath241 is lipschitz continuous with the lipschitz constant @xmath210 . by the hardy littlewood theorem , @xcite , @xmath242 , and hence @xmath243 as @xmath212 .
this immediately implies the property ( b ) of @xmath209 .
now we prove the property ( a ) .
given @xmath244 , let @xmath245 be such that @xmath246 . note that whenever @xmath247 is such that @xmath248 , then @xmath249 .
observe that radius of @xmath250 is comparable to @xmath251 whenever @xmath252 .
choose @xmath216 large enough to make the measure @xmath253 so small that radii of all of the balls @xmath250 are less than @xmath199 .
for @xmath244 we have @xmath254 where @xmath255 ( observe that @xmath248 for @xmath256 ) .
we will prove now that @xmath257 this together with inequality ( [ last ] ) , the observation that @xmath258 for @xmath256 and the fact that the number of elements in @xmath259 is bounded by a constant that does not depend on @xmath260 will give @xmath261 to prove ( [ asia ] ) note that @xmath262 ) yields @xmath263 because @xmath264 . to estimate the second term we apply the poincar inequality @xmath265 the above inequalities
complete the proof of ( [ asia ] ) and hence that for ( [ trzy ] ) .
we proved inequality ( [ trzy ] ) for @xmath244 .
the inequality holds , however , for all @xmath266 because @xmath267 whenever @xmath268 .
now we are ready to estimate the lipschitz constant of @xmath209 .
if @xmath269 , then @xmath270 we employed here the fact that @xmath271 is lipschitz continuous with the lipschitz constant @xmath210 . thus we can assume that @xmath272 .
since @xmath273 we have @xmath274 in the inequality second to last we employed the fact that @xmath275 for all @xmath276 .
the proof of this fact is very similar to the proof of inequality ( [ asia ] ) .
the proof of the second claim is complete .
now we will prove the property ( c ) of @xmath209 .
for @xmath268 we have @xmath277 and hence we can assume that @xmath278 .
given @xmath153 there is @xmath199 such that for every ball @xmath129 in @xmath4 with radius @xmath279 we have @xmath280 hence if we choose @xmath216 large enough we can guarantee that ( [ ast ] ) holds for each of the balls @xmath281 . since @xmath282 , the poincar inequality yields @xmath283 this means for sufficiently large @xmath216 all averages @xmath250 are very close to @xmath26 .
since @xmath284 it suffices to prove that @xmath285 .
we have @xmath286 ) and the fact that the number of indices in @xmath259 is bounded by a constant independent of @xmath260 .
this completes the proof of the last claim and hence that of the theorem .
let us start with a brief description of the main steps in the proof .
let @xmath287 be a closed @xmath52-dimensional manifold .
in the first step we define a continuous @xmath108 function @xmath288 on @xmath142 with compact support contained in the unit ball @xmath289 . in the second step we use the fact that @xmath4 contains a subset diffeomorphic to @xmath290 to replace a part of @xmath4 by the graph of @xmath288 .
the resulting space is denoted by @xmath291 .
it is still a subset of @xmath13 .
the function @xmath288 is constructed in such a way that it is smooth everywhere but in one point and hence @xmath291 is a smooth manifold with one nonsmooth point on it . obviously @xmath291 is homeomorphic to @xmath4 , actually a careful construction of @xmath291 can make it @xmath292 homeomorphic in the sense that there is a homeomorphism @xmath293 such that @xmath294 . moreover @xmath68 can be chosen to be a diffeomorphism everywhere but in one point and identity outside a sufficiently large ball . while @xmath295 is a smooth homeomorphism , it is shown that there is no smooth homeomorphism from @xmath4 onto @xmath291 . actually due to high oscillations of @xmath288 near the nonsmooth point
, it is proved that there is no lipschitz mapping @xmath296 . in the last step
we show that lipschitz mappings @xmath123 are not dense in @xmath124 .
the reason being the following : @xmath297 and hence there is a homeomorphism @xmath298 of the class @xmath108 .
it follows from the homotopy properties of @xmath108 mappings that every lipschitz mapping @xmath299 sufficiently close to @xmath74 in the @xmath108 norm is homotopic to @xmath74 and hence @xmath300 .
this contradicts , however , the lack of lipschitz mappings @xmath296 .
actually the steps of the proof will be organized in a slightly different order due to the idea of collecting technical constructions at the end of the proof . * construction of @xmath288 . * the construction of @xmath288 presented below is closely related to the fact that the @xmath52-capacity of a point is zero .
those who know what it means can avoid the part of the construction involving truncation and convolution , but just invoke basic definitions .
it is well known and easy to prove that @xmath301 .
define the truncation of @xmath302 between levels @xmath303 and @xmath216 , @xmath304 by @xmath305 fix an arbitrary @xmath306 . for every @xmath153
there is a sufficiently large @xmath303 such that @xmath307 is a lipschitz function on @xmath142 with the properties : @xmath308 @xmath309 @xmath310 the function @xmath311 is not smooth because it is defined as a truncation , however , mollifying @xmath311 gives a smooth function , denoted by @xmath312 , with the same properties as those of @xmath311 listed above .
let @xmath143 denotes the coordinates in @xmath142 , where @xmath313 .
let @xmath314 , @xmath315 , and let @xmath316 .
the balls @xmath150 are pairwise disjoint and all contained in @xmath317 . observe that we can find @xmath318 pairwise disjoint balls of radius @xmath319 inside @xmath150 .
select such a family for each @xmath150 and denote the balls in the family by @xmath320 , @xmath321 .
denote the centers of @xmath320 by @xmath322 and define @xmath323 where @xmath324 .
each of the functions @xmath325 is supported in @xmath326 and hence the supports of the functions in the double sum ( [ myszka ] ) are pairwise disjoint .
this immediately implies that @xmath327 , @xmath288 is continuous on @xmath142 , @xmath328 and @xmath329 .
moreover @xmath330 and hence @xmath331 .
denote the graph of @xmath288 over @xmath332 by @xmath333 i.e. @xmath334 we equip @xmath333 with the metric induced from @xmath107 .
the following lemma is a consequence of high oscillations of @xmath288 near the origin .
[ niema ] let @xmath11 be a closed @xmath52-dimensional riemannian manifold .
then there is no lipschitz mapping @xmath335 .
_ the claim of the lemma remains the same ( with the same proof ) if we replace @xmath11 by any metric space @xmath26 equipped with a finite borel measure @xmath336 satisfying the lower bound for the growth @xmath337 for all @xmath88 and all @xmath338
. clearly @xmath11 equipped with the hausdorff measure @xmath339 is an example .
_ proof of lemma [ niema ] .
_ we argue by contradiction .
suppose @xmath335 is an @xmath340-lipschitz mapping .
fix a positive integer @xmath126 .
the graph of @xmath325 over @xmath326 is of height @xmath341 and hence we can find @xmath342 points on that graph whose mutual distances are at least @xmath319 .
choose such points and denote them by @xmath343 , @xmath344 .
since distances between the balls @xmath345 are no less than @xmath319 we conclude that mutual distances of the points in the family @xmath346 are at least @xmath319 .
the family @xmath346 consists of @xmath347 points .
choose @xmath348 arbitrarily . since @xmath349 is @xmath340-lipschitz
the distances between points @xmath350 are at least @xmath351 .
hence the balls @xmath352 are pairwise disjoint .
this in turn gives the lower estimate for the volume of @xmath11 .
@xmath353 as @xmath152 .
this clearly contradicts the finiteness of the volume of @xmath11 .
the proof of the lemma is complete .
@xmath94 * the construction of @xmath291 .
* take a subset of @xmath4 that is diffeomorphic to @xmath332 and replace it by the graph @xmath333 .
denote the resulting space by @xmath354 .
the construction is quite obvious and does not require further explanations . in any case , we will present all details in the next subsection because it will be needed for the construction of @xmath68 .
the following properties of @xmath291 are obvious .
the property ( c ) below is a consequence of the fact that the mapping @xmath355 is lipschitz continuous , and property ( d ) is a consequence of lemma [ niema ] . 1 .
there is @xmath356 such that @xmath357 is a smooth submanifold of @xmath13 .
2 . there is a homeomorphism @xmath298 in the class @xmath358 with the property that @xmath359 is a @xmath292 diffeomorphism of manifolds , where @xmath360 .
@xmath74 can be choosen in such a way that @xmath361 is lipschitz continuous .
4 . there is no lipschitz mapping @xmath296 .
the above properties do not immediately guarantee the existence of a homeomorphism @xmath293 being diffeomorphism everywhere but in one point and such that @xmath294 .
one could try to construct @xmath68 as an extension of @xmath361 , however , high oscillations of @xmath291 near @xmath362 seem to produce a singularity of @xmath363 at @xmath362 , and this is just one of the technical problems that we will have to face .
we postpone the technically complicated construction of @xmath68 for a while and show that lipschitz mappings @xmath123 are not dense in @xmath124 .
we will actually show that the homeomorphism @xmath358 can not be approximated by lipschitz mappings @xmath123 in the @xmath108 norm .
suppose by contrary that @xmath364 is a sequence of lipschitz mappings such that @xmath365 in @xmath108 .
then @xmath366 in @xmath367 .
indeed , by lemma [ jeden ] the composition with @xmath368 is continuous in the @xmath108 norm .
admits an extension to @xmath121 .
hence @xmath368 is smooth everywhere and there is no problem with the continuity of the operator of composition . ] .
we will also need the following result of white @xcite . [
5.2 ] let @xmath4 and @xmath11 be closed manifolds and let @xmath60
. then there exists @xmath153 such that any two lipschitz mappings @xmath369 satisfying @xmath370 are homotopic .
the degree modulo @xmath371 of the identity mapping @xmath372 equals @xmath185 ( we do not assume @xmath4 be orientable , so the brouwer degree needs to be replaced by degree modulo @xmath371 , see @xcite ) .
lemma [ 5.2 ] implies that for sufficiently large @xmath126 , @xmath373 is homotopic to the identity map .
hence by the homotopy lemma @xcite , the degree modulo @xmath371 of @xmath374 equals @xmath185 as well , and the definition of the degree modulo @xmath371 yields @xmath375 .
this , in turn , implies @xmath376 which contradicts property ( d ) . * the construction of @xmath68 . * to complete the proof of the theorem it remains to show that there is a @xmath292 homeomorphism @xmath377 which is diffeomorphism everywhere but in one point , identity outside a sufficiently large ball , and satisfies @xmath294 . to this end
we will repeat the construction of @xmath291 providing more details . in
what follows @xmath378 , @xmath379 , will be identified with a subset of @xmath13 being the ball in the subspace generated by the first @xmath126 coordinates .
we use @xmath380 to designate coordinates in @xmath381 . with no loss of generality
we may assume that @xmath382 let @xmath288 be as before .
define @xmath291 by replacing @xmath383 with the graph of @xmath288 @xmath384 now we will extend this graph to a graph of a function defined in @xmath385 .
let @xmath386 be such that @xmath387 and @xmath388 define @xmath389 by @xmath390 it is easily seen that @xmath391 , @xmath392 is continuous on @xmath393 and equals zero outside the ball @xmath394 .
now define the graph of @xmath392 over @xmath395 by @xmath396 since @xmath397 it follows that @xmath398 .
clearly @xmath399 is an extension of @xmath333 in the sense that the intersection of @xmath399 with the linear subspace @xmath400 coincides with @xmath333 .
observe that @xmath401 is a homeomorphism @xmath402 , @xmath403 , @xmath404 and @xmath405 is a smooth diffeomorphism off the line @xmath406 .
however , @xmath405 is _ not _ differentiable in points of the line @xmath407 .
moreover @xmath405 is not identity outside any finite ball .
hence we have to modify @xmath405 in order to improve its properties .
actually we will prove .
[ king ] there is a @xmath292 homeomorphism @xmath408 which is diffeomorphism in @xmath122 , @xmath409 , @xmath410 and @xmath68 is identity in @xmath411 .
it is obvious now that lemma [ king ] implies theorem [ smooth ] , so we are left with the proof of the lemma . since a version of lemma [ king ] holds for a much larger class of graphs , and not just for the graph of the function @xmath392 constructed above , it makes sense to formulate the lemma in a more general form and deduce lemma [ king ] as a special case .
lemma [ king ] is a direct consequence of lemma [ michalek ] and lemma [ natusia ] below .
notation @xmath380 for the coordinates in @xmath13 will no longer be convenient for us , so in the remaining part of section [ four ] we will denote the coordinates in @xmath13 by @xmath412 , where @xmath413 .
[ michalek ] assume that a nonnegative function @xmath392 defined on @xmath393 is continuous , @xmath414 , @xmath415 , and @xmath416 .
denote the graph of @xmath392 over @xmath395 by @xmath399 i.e. , @xmath417 assume that the graph of @xmath392 over @xmath418 is contained in the ball @xmath419 .
then there is a homeomorphism @xmath420 of the class @xmath421 with the following properties 1 .
@xmath422 is of the form @xmath423 , 2 .
@xmath424 , 3 .
@xmath422 is a diffeomorphism in @xmath122 , 4 .
@xmath425 , 5 .
@xmath422 is identity in @xmath411 .
the mapping @xmath422 constructed in the above lemma has all the properties needed in lemma [ king ] except for the smoothness at the origin .
the following lemma shows how to modify the mapping @xmath422 in a way that it will have all the properties needed in lemma [ king ] .
[ natusia ] let @xmath426 be a continuous mapping such that @xmath427 and @xmath428 .
then there is a homeomorphism @xmath429 such that 1 .
@xmath430 , 2 .
@xmath431 is a diffeomorphism in @xmath432 , 3 .
@xmath433 and @xmath434 for all @xmath435 , 4 .
@xmath436 is of the form @xmath437 for some homeomorphism @xmath438and hence @xmath439 for @xmath440 .
5 . @xmath441 .
if we set @xmath442 , where @xmath422 is defined for @xmath392 given by ( [ gwiazdka ] ) , then @xmath443 has the properties required in lemma [ king ] .
indeed , it follows from the property ( 1 ) of @xmath422 that @xmath444 and hence @xmath445 by the property ( d ) of @xmath431 .
_ proof of lemma [ michalek ] .
_ let @xmath446 .
then @xmath447 is strictly positive in @xmath448 , @xmath449 and @xmath450 in @xmath451 .
fix a function @xmath452 such that @xmath453 for @xmath454 , @xmath455 $ ] and @xmath456 .
for @xmath457 we define @xmath458 then @xmath459 , the support of the function @xmath460 is contained in the interval @xmath461 $ ] , the function is nonpositive in the interval , but @xmath462 .
for @xmath457 we also define @xmath463 then the support of the function @xmath464 is contained in the interval @xmath465 $ ] , the function is nonnegative in the interval and @xmath466 .
now we define @xmath467 the integrand is greater than or equal to @xmath468 and hence the function @xmath469 is strictly increasing in @xmath470 with @xmath471 .
moreover @xmath472 for @xmath473 and @xmath474 for @xmath475 .
clearly the mapping @xmath423 is a homeomorphism of @xmath476 onto @xmath476 satisfying claims ( 1 ) , ( 2 ) and ( 4 ) .
property ( 5 ) follows from the formula for @xmath469 because @xmath477 whenever @xmath478 , and the fact that support of @xmath392 is concentrated near @xmath151 .
we still have to prove that @xmath422 is smooth in @xmath122 and that the jacobian of @xmath422 is positive in @xmath122 .
smoothness of @xmath422 at @xmath479 , @xmath480 follows from the definition of @xmath469 and smoothness of @xmath422 at @xmath481 , @xmath482 follows from the observation that @xmath422 is the identity in a neighborhood of @xmath481 .
the mapping @xmath422 is a diffeomorphism in @xmath483 because direct computation shows that @xmath484 .
this completes the proof of all the properties of @xmath422 .
@xmath94 _ proof of lemma [ natusia ] . _ in the proof we will need the following result .
[ elem ] let @xmath485 be an arbitrary increasing sequence of positive numbers .
then there is a homeomorphism @xmath486 such that @xmath487 , @xmath488 for @xmath489 , @xmath490 for @xmath491 and @xmath492 as @xmath493 .
_ there is a nonnegative function @xmath494 with @xmath495 $ ] and such that @xmath496 for all @xmath489 .
indeed , take an arbitrary function @xmath497 such that @xmath498 $ ] , and @xmath499 in @xmath500
. then @xmath501 , where @xmath502 satisfies ( [ 76 ] ) .
this follows from the observation that @xmath503 is a strictly positive @xmath292 function for all @xmath489 and @xmath504 for all integers @xmath52 .
let @xmath505 be a sequence of positive numbers less than @xmath185 such that @xmath506 as @xmath493 .
let also @xmath507 for all integers @xmath508 .
the function @xmath509 is smooth and strictly positive .
moreover @xmath510 as @xmath511
. define @xmath512 obviously @xmath513 is a homeomorphism , @xmath514 and @xmath488 for @xmath489 .
moreover @xmath515 for @xmath491 .
however , @xmath516 for @xmath491 and hence @xmath517 for all @xmath491 and some constant @xmath518 .
let @xmath519 be an arbitrary nondecreasing function such that @xmath520 for @xmath521 and @xmath522 for @xmath523 .
now the function @xmath524 has all the properties required in the lemma except perhaps for ( [ square ] ) .
we will show that an appropriate choice of @xmath525 will guarantee ( [ square ] ) as well .
since @xmath520 for @xmath521 we have that @xmath526 for small @xmath216 and hence it suffices to prove ( [ square ] ) for @xmath527 .
let @xmath528 we will show that the function @xmath527 constructed for the sequence @xmath529 satisfies ( [ square ] ) . since for every @xmath489 no more than two of the functions @xmath530 are different from zero we conclude that for any @xmath531 there is @xmath532 such that for any @xmath533 @xmath534 hence @xmath535 and thus @xmath536 as @xmath493 .
it remains to show that @xmath537 as @xmath493 .
this is easy and we leave it to the reader .
the proof of the lemma is complete .
@xmath94 now we can complete the proof of lemma [ natusia ] .
we will construct a homeomorphism @xmath431 of the form @xmath437 where @xmath538 is as in lemma [ elem ] for an appropriate sequence @xmath539 .
the sequence will be chosen later .
obviously @xmath540 is a homeomorphism such that @xmath433 , @xmath434 for @xmath435 and @xmath431 is a diffeomorphism of @xmath541 of the class @xmath292 .
observe that @xmath542 where @xmath543 denotes the vector whose components are all partial derivatives of order @xmath95 . hence if @xmath485 is a sequence such that @xmath544 , then @xmath545 for each positive integer @xmath95 .
this , in turn , implies that @xmath431 has partial derivatives at @xmath151 of an arbitrary order and all the derivatives are equal to @xmath151 .
hence @xmath546 .
obviously @xmath547 .
now it remains to prove that @xmath441 for some increasing sequence @xmath485 , such that @xmath544 . to this end
it suffices to show that for some sequence @xmath485 we have @xmath548 for each nonnegative integer @xmath95 .
a multiple application of the leibniz rule gives the following variant of the fa di bruno formula @xmath549 where the last sum is taken over all nonnegative integers @xmath550 , @xmath551 ,
, @xmath552 such that @xmath553 define @xmath554 where the supremum is taken over all @xmath260 satisfying @xmath555 and @xmath556 . for such @xmath260 we havelatexmath:[\[\begin{aligned }
\lefteqn { c m_n \sum_{i=1}^{m}|(\nabla^{i}\theta)(\lambda(x))| } \\ & \leq & c m_n \sum_{i=1}^{m } \left ( \sum_{j=0}^{i } |\lambda(x)|^{-j } c m_n 2^{(n+1)m } m^2 \sup_{\ell\leq m } @xmath138 depending on @xmath95 for @xmath558 we have latexmath:[\[|\nabla^{m}(\theta\circ\lambda)(x)|/|x| \leq c m_n 2^{nm } \sup_{0<t\leq 2^{-n } } \sup_{\ell\leq n } @xmath485 , such that @xmath560 .
this completes the proof of lemma [ natusia ] and hence that for lemma [ king ] .
thus the proof of theorem [ smooth ] is complete .
let @xmath111 and let @xmath112 be the singular submanifold of @xmath107 constructed in the proof of theorem [ smooth ] .
let @xmath561 be the @xmath108 homeomorphism constructed in the proof of theorem [ smooth ] .
let @xmath562 be a closed @xmath52-dimensional manifold and let @xmath563 be a subset diffeomorphic to @xmath289 .
then there is a smooth mapping @xmath564 which is one - to - one on @xmath129 and which maps all of @xmath565 into one point .
clearly such a mapping has degree modulo @xmath371 equal @xmath185 .
now the same topological argument based on lemma [ niema ] , as in the proof of theorem [ smooth ] shows that @xmath566 can not be approximated by lipschitz mappings @xmath567 . @xmath94
* construction of @xmath26 . *
let @xmath11 be a submanifold of @xmath107 diffeomorphic to @xmath104 having @xmath332 as a subset .
let @xmath325 be defined as in the proof of theorem [ smooth ] and define @xmath568 thus the sequence @xmath569 is a smooth approximation of @xmath288 defined by formula ( [ myszka ] ) in the @xmath108 norm .
let @xmath570 be the graph of @xmath571 over the unit ball @xmath332 . denote by @xmath572 the manifold obtained from @xmath11 by substituting @xmath332 with the graph @xmath570 .
since each of the functions @xmath571 is smooth , @xmath572 is a smooth manifold diffeomorphic to @xmath104 .
the manifolds @xmath572 converge in some sense to @xmath573 , the space obtained from @xmath11 by substituting @xmath332 with @xmath333 , the graph of @xmath288 defined by formula ( [ myszka ] ) .
let @xmath150 denote balls defined as in the definition of the function @xmath288 .
denote the coordinates in @xmath107 by @xmath574 , @xmath575 , @xmath576 .
there is a natural projection @xmath577 of the graph @xmath333 onto @xmath570 , namely @xmath578 the projection extends in a natural way to a homeomorphism @xmath579 .
denote the orthogonal projection from @xmath107 onto @xmath142 by @xmath582 .
observe that @xmath583 .
hence @xmath577 restricted to @xmath584 is smooth with the constant norm of the derivative that does not depend on @xmath95 .
since @xmath577 is identity outside @xmath580 , @xmath577 extends to a smooth mapping of a neighbourhood of @xmath573 in @xmath107 onto @xmath572 . _ proof . _
since @xmath577 extends to a smooth mapping we conclude that @xmath590 .
now @xmath591 on the set @xmath592 .
this is a decreasing sequence of subsets of @xmath4 with empty intersection and hence @xmath593 as @xmath589 .
this , in turn , easily implies that @xmath587 in @xmath108 as @xmath589 because @xmath594 as @xmath589 .
we employed here the fact that @xmath595 for all @xmath596 .
the proof of the lemma is complete .
@xmath94 @xmath572 is a subset of @xmath107 and hence it is a subset of @xmath597 .
let @xmath598 be the translation of @xmath572 by the vector @xmath599 in @xmath59 .
now we define @xmath600 we equip @xmath601 with the metric induced from @xmath59 .
@xmath601 is a compact set , but it is not connected .
we define @xmath26 by connecting all parts of @xmath601 by a curve whose each part is non - rectifiable .
it is an easy consequence of the absolute continuity of @xmath92 on lines ( * ? ? ?
* theorem 2.1.4 ) that images of almost all lines under @xmath92 are rectifiable .
hence sobolev mappings from @xmath4 into a space that contains no rectifiable curves ( except constant ones ) must be constant . accordingly ,
if @xmath602 is not constant , then @xmath603 . in the next two steps we will prove the density of smooth mappings @xmath604 in @xmath56 and we will construct a global bi - lipschitz homeomorphism @xmath605 such that lipschitz mappings @xmath606 are not dense in @xmath607 .
_ the lemma is a fairly easy consequence of the absolute continuity of @xmath92 on almost all lines . indeed ,
if we would find two subsets @xmath613 of positive measure such that @xmath614 and @xmath615 , @xmath616 , then we would find a curve joining @xmath163 and @xmath129 such that @xmath92 is continuous along that curve .
hence the image of @xmath92 would contain a curve joining @xmath617 and @xmath618 inside @xmath601 .
this is , however , impossible .
@xmath94 we know that @xmath610 for some @xmath126 finite or @xmath612 .
if @xmath126 is finite , then there is no problem with the approximation because @xmath619 is a closed manifold and the smooth mappings @xmath620 are dense in the class @xmath621 according to schoen and uhlenbeck s theorem ( cf . theorem [ dens ] ) .
thus we are left with the case in which @xmath622 .
let @xmath623 i.e. first we compose @xmath92 with the projection @xmath577 from @xmath573 onto @xmath572 and then we translate the image by the vector @xmath599 , so that the resulting mapping @xmath624 maps @xmath4 to @xmath598 .
it follows from lemma [ dwanascie ] that @xmath625 in @xmath108 as @xmath589 .
now it suffices to observe that @xmath626 can be approximated by smooth mappings @xmath627 .
this completes the proof of density of @xmath608 in @xmath609 .
* construction of @xmath68 .
* we can assume that @xmath628^n$ ] is a subset of @xmath11 .
since @xmath572 was obtained from @xmath11 by modifying it on @xmath289 only , @xmath628^n$ ] is a subset of @xmath572 as well .
thus @xmath629^n\times \ { 0\ } \times \ { 2^{-m}\}\ ] ] is a subset of @xmath598 .
moreover the intersection of some @xmath630-dimensional box containing @xmath631 with @xmath601 equals @xmath631 .
more precisely we can assume that @xmath632 where @xmath633^n\times [ -2^{-(m+2)},2^{-(m+2)}]\times [ 2^{-m}-2^{-(m+2)},2^{-m}+2^{-(m+2)}].\ ] ] now for each @xmath95 we will construct an @xmath340-bi - lipschitz homeomorphism @xmath634 which is identity on the boundary @xmath635 , and with the constant @xmath340 that does not depend on @xmath95 .
it follows that the mapping @xmath636 is bi - lipschitz .
indeed , it is enough to observe that the norm of the derivative is bounded from below by a positive constant and bounded from above .
the set @xmath637 will be a subset of @xmath638 with `` wrinkles '' .
it will be a graph of a lipschitz function @xmath639 constructed later . thus the mapping @xmath68 transforms each @xmath598 into a manifold with `` wrinkles '' added to its surface .
note that @xmath640 .
let @xmath4 be a closed @xmath52-dimensional manifold . according to theorem [ subaru ] there is a mapping @xmath585 which can not be approximated by lipschtz mappings @xmath641 .
since @xmath640 , we have @xmath642 .
our aim is to construct a bi - lipschitz mapping @xmath605 of the form described above which has the property that there is @xmath153 such that every lipschitz mapping @xmath643 satisfies @xmath644 .
this will readily imply that the mapping @xmath642 can not be approximated by lipschitz mappings @xmath606 .
note that @xmath92 can be approximated by lipschitz mappings into @xmath598 ( see ( [ asterix ] ) ) , however , adding `` wrinkles '' to @xmath598 by applying @xmath68 will have the effect that mappings into @xmath645 stay away from @xmath92 .
let @xmath647 be an @xmath648-dimensional box contained in @xmath638 and let @xmath649 for @xmath650 .
the set @xmath651 is an @xmath52-dimensional subset of @xmath652 .
let @xmath653 be the function defined by the formula @xmath654 using notation @xmath655 , @xmath575 , for the coordinates in @xmath59 , we define the graph of @xmath639 by @xmath656 @xmath657 is a compact subset of @xmath638 .
the lipschitz constant of @xmath639 is @xmath185 as it is the distance function .
[ joasia ] there is a constant @xmath658 depending on @xmath52 only such that for every integer @xmath659 there is an @xmath340-bi - lipschitz homeomorphism @xmath634 which is identity on the boundary and which satisfies @xmath660 .
_ let @xmath661 . with this notation
@xmath662\subset e_m\times\bbbr$ ] .
the mapping @xmath663 given by @xmath664 is @xmath340-bi - lipschitz for some constant @xmath340 that does not depend on @xmath95 .
indeed , the inverse mapping is @xmath665 and both @xmath666 and @xmath667 have partial derivatives bounded by @xmath185 , because the lipschitz constant of @xmath639 is @xmath185 .
the mapping @xmath666 is identity on the part of the boundary of @xmath638 ( because @xmath639 is zero on @xmath668 ) , however , it is not identity on the entire boundary of @xmath638 .
fortunately , one can easily use an argument similar to that employed in the proof of lemma [ michalek ] and construct a mapping @xmath669 of the form @xmath670 being a modification of @xmath666 and satisfying the claim of lemma [ joasia ] .
indeed , one can write the fomula for @xmath671 explicitly making it linear as a function of @xmath672 on each of the segments @xmath673 $ ] and @xmath674 $ ] .
we leave easy details to the reader .
@xmath94 * lack of density .
* now we can complete the proof of theorem [ main ] by showing that @xmath642 can not be approximated by mappings @xmath606 .
as we have already observed it suffices to prove that there is no sequence @xmath675 , such that @xmath676 as @xmath152 .
let us briefly recall the construction of the mapping @xmath92 ( cf .
the proof of theorem [ subaru ] ; in that proof @xmath562 plays the role of @xmath4 ) .
@xmath573 is homeomorphic to @xmath104 and @xmath677 is diffeomorphic to @xmath104 with one point removed .
we choose a subset @xmath678 of @xmath4 diffeomorphic to a closed euclidean ball .
the mapping @xmath92 maps @xmath679 into one point @xmath680 and @xmath681 is a homeomorphism .
moreover @xmath682 is a diffeomorphism .
we can assume that @xmath628^n\subset n^\infty\setminus \ { x^*,0\}$ ] as the point @xmath680 can be chosen arbitrarily .
that means , applying a suitable change of variables in @xmath4 , we can assume that @xmath628^n\subset m$ ] and that @xmath683 satisfies @xmath684 for @xmath685^n$ ] .
suppose that @xmath675 , @xmath676 as @xmath152 .
let us denote the coordinates in @xmath628^n$ ] by @xmath686^{n-1}\times [ 1,2]=[1,2]^n.\ ] ] fubini s theorem implies that for a suitable subsequence of @xmath687 ( also denoted by @xmath687 ) and for almost all @xmath688^{n-1}$ ] , the sequence of functions of one variable @xmath689\ni x^n\mapsto f_{m_k}(\hat{x},x^n)\ ] ] converges to the function on one variable @xmath690\ni x^n\mapsto f(\hat{x},x^n)\ ] ] in @xmath691)$ ] .
in particular it implies that the functions converge uniformly and that the lengths of the curves ( [ b421 ] ) converge to the length of the curve ( [ b422 ] ) as @xmath152 .
the length of the curve ( [ b422 ] ) equals @xmath185 , because @xmath92 is identity on @xmath628^n$ ] .
however , for @xmath126 sufficiently large , the length of the curve ( [ b421 ] ) is at least @xmath692 , because it goes across the `` wrinkles '' @xmath693 .
indeed , the projection of the curve ( [ b421 ] ) onto the coordinate plane generated by coordinates @xmath694 contains the projection of the set @xmath693 on that plane .
this projection is , however , isometric to the curve @xmath628\ni t\mapsto ( t,\dist ( t , s))$ ] , where @xmath695 and the length of that curve equals @xmath692 . since the orthogonal projection does not increase the length , we conclude that the length of the curve ( [ b421 ] ) is at least @xmath692 .
this is a contradiction with the convergence of the length .
the proof of theorem [ main ] is complete .
@xmath94 _ remark .
_ the homeomorphisms @xmath634 are not smooth .
however , one can easily modify the construction in a way that the they are diffeomorphism .
then the mapping @xmath605 is a diffeomorphism in @xmath696^{n}$ ] i.e. it is a diffeomorphism outside a square of dimension @xmath52 .
888 a approximation lemma for @xmath1 functions .
_ in : _ proc . of the symposium on material instabilities and continuum mech
. j. m. ball , pp .
15 , oxford science publications , 1988 .
the homotopy groups of the integral cycle groups .
_ topology _ 1 ( 1962 ) , 257299 .
metric space valued functions of bounded variation .
_ ann . scuola norm .
pisa cl .
sci._17 ( 1990 ) , 439478 .
a characterisation of maps in @xmath697 , which can be approximated by smooth maps , _ ann .
i. h. p. analyse non linaire _ 7 ( 1990 ) , 269286 . ,
the approximation problem for sobolev maps between two manifolds .
_ acta math .
_ 167 ( 1991 ) , 153206 . a cohomological criterion for density of smooth maps in sobolev spaces between two manifolds . _ in : _ nematics ( orsay , 1990 ) , pp .
1523 , nato adv .
, 332 , kluwer acad . publ . ,
dordrecht , 1991 .
density of smooth functions between two manifolds in sobolev spaces .
_ j. functional anal .
_ , 80 ( 1988 ) , 6075
. improved @xmath698 approximation of higher order sobolev functions in norm and capacity _ indiana univ .
j. _ 51 ( 2002 ) , 507540 .. topology and sobolev spaces .
_ j. funct .
anal . _ 183 ( 2001 ) , 321369 .
degree and sobolev spaces . _
methods nonlinear anal . _ 13 ( 1999 ) , 181190 . degree theory and bmo .
i. compact manifolds without boundaries .
_ selecta math .
_ 1 ( 1995 ) , 197263 . degree theory and bmo .
ii . compact manifolds with boundaries . with an appendix by the authors and petru mironescu . _ selecta math .
_ 2 ( 1996 ) , 309368 .
legendrian energy minimizers
. i. heisenberg group target .
var . partial differential equations _ 12 ( 2001 ) , 145171 . ,
lecture notes in math.242 , springer 1971 . minimizing @xmath173-harmonic maps into spheres . _
j. reine angew .
_ 401 ( 1989 ) , 82100 .
une caractrisation des applications de @xmath699 qui peuvent etre approches par des fonctions rgulires .
_ c. r. acad.sci .
paris _ 310 , srie i , ( 1990 ) , 553557 . harmonic maps between riemannian polyhedra .
cambridge tracts in mathematics , 142 .
cambridge university press , cambridge , 2001 .
a report on harmonic maps . _ bull .
london math .
_ 10 ( 1978 ) , 168 .
some remarks on the density of regular mappings in sobolev classes of @xmath700-valued functions .
_ rev . mat .
complut .
madrid _ 1 ( 1988 ) , 127144 .
degree formulas for maps with nonintegrable jacobian . _
methods nonlinear anal . _
6 ( 1995 ) , 8195 .
carnot - carathodory spaces seen from within . in : _ sub - riemannian geometry .
_ , 79323 , progr .
, 144 , birkhuser , basel , 1996 . harmonic maps into singular spaces and @xmath173-adic superrigidity for lattices in groups of rank one .
hautes tudes sci .
_ 76 ( 1992 ) , 165246 . , equivalent statement of the poincar conjecture , _ annali .
pura appl .
_ 167 ( 1994 ) , 2531 . ,
approximation of sobolev mappings , _ nonlinear analysis _ 22 ( 1994 ) , 1579 - 1591 .
sobolev spaces on an arbitrary metric space .
_ potential analysis _ 5 ( 1996 ) , 403415 .
hlder quasicontinuity of sobolev functions on metric spaces _ rev
. mat .
iberoamericana _ 14 ( 1998 ) , 601622 .
sobolev met poincar . _
memoirs amer .
_ 688 ( 2000 ) , 1101 .
weakly differentiable mappings between manifolds .
_ memoirs amer .
( to appear ) topology of sobolev mappings .
_ 8 ( 2001 ) , 321330 .
topology of sobolev mappings ii .
_ acta math .
_ 191 ( 2003 ) , 55107
university of jyvskyl .
department of mathematics and statistics , 100 .
university of jyvskyl , jyvskyl , 2005 .
quasiconformal maps in metric spaces with controlled geometry .
_ acta math .
_ 181 ( 1998 ) , 161 .
sobolev classes of banach space - valued functions and quasiconformal mappings .
_ j. anal .
_ 85 ( 2001 ) , 87139 .
approximation of sobolev maps between an open set and euclidean sphere , boundary data , and singularities . _ math .
_ , 285 ( 1989 ) , 125140 . characterization of the strong closure of @xmath701 in @xmath702 ( @xmath703 ) .
_ j. math .
_ 190 ( 1995 ) , 361372 . equilibrium maps between metric spaces .
_ 2 ( 1994 ) , 173205 .
generalized dirichlet forms and harmonic maps .
_ 5 ( 1997 ) , 119 .
nonlinear dirichlet forms . in : _ new directions in dirichlet forms .
_ , 147 , ams / ip stud .
, 8 , amer .
soc . , providence , ri , 1998 . harmonic maps into bruhat tits buildings and factorizations of @xmath173-adically unbounded representations of @xmath704 of algebraic varieties .
j. algebraic geom .
_ 9 ( 2000 ) , 142 .
sobolev spaces and harmonic maps for metric space targets .
_ 1 ( 1993),561659 . some function spaces on spaces of homogeneous type .
_ manuscripta math .
_ 106 ( 2001 ) , 219248 . elements of lipschitz topology .
ser . a i math . _ 3 ( 1977 ) , 85122 .
lipschitz functions on spaces of homogeneous type .
_ 33 ( 1979 ) , 257270 . mathematical surveys and monographs , 51 .
american mathematical society , providence , ri , 1997 .
every superposition operator mapping one sobolev space into another is continuous . _
j. funct .
_ 33 ( 1979 ) , 217229 .
topology from the differentiable viewpoint . the university press of virginia , charlottesville , va .
1965 the problem of plateau on a riemannian manifold .
_ ann . of math .
_ 49 ( 1948 ) , 807851 . on surfaces of finite total curvature
. _ j. diff .
_ 42 ( 1995 ) , 229258 .
the imbedding problem for riemannian manifolds .
_ ann . of math .
_ 63 ( 1956 ) , 2063
. weak density of smooth maps for the dirichlet energy between manifolds .
anal . _ 13 ( 2003 ) , 223257 .
sobolev classes of functions with values in a metric space .
( russian ) _ sibirsk . mat .
38 ( 1997 ) , 657675 , translation in _ siberian math . j. _ 38 ( 1997 ) , 567583 .
the dirichlet problem for harmonic maps . _ j. differential geometry _ , 18 ( 1983 ) , 153268 .
approximation theorems for sobolev mappings , ( preprint ) .
boundary regularity of harmonic maps to nonpositively curved metric spaces .
_ 2 ( 1994 ) , 139153 .
princeton mathematical series , no .
30 princeton university press , princeton , n.j .
1970 surfaces with generalized second fundamental form in @xmath705 are lipschitz manifolds . _
j. diff .
_ 39 ( 1994 ) , 65101 .
approximately lipschitz mappings and sobolev mappings between metric spaces . in : _ proceedings on analysis and geometry _ ( russian ) .
( novosibirsk akademgorodok , 1999 ) , 585594 , izdat . ross .
nauk sib .
inst . mat .
, novosibirsk , 2000 .
infima of energy functionals in homotopy classes of mappings . _ j. differential geometry _ , 23 ( 1986 ) 127142 .
homotopy classes in sobolev spaces and the existence of energy minimizing maps .
_ acta math .
_ 160 ( 1988 ) , 117 . , graduate texts in mathematics 120 ( springer 1989 ) . | we study a question of density of lipschitz mappings in the sobolev class of mappings from a closed manifold into a singular space .
the main result of the paper , theorem [ main ] , shows that if we change the metric in the target space to a bi - lipschitz equivalent one , than the property of the density of lipschitz mappings may be lost .
other main results in the paper are theorems [ one ] , [ dens ] , [ subaru ] , [ smooth ] . |
transiting exoplanets provide invaluable information on planetary physics , their formation , and evolution .
the search for planets in the habitable zone has so far focused on solar - type stars and m dwarfs .
however , transiting planets in the habitable zone around white dwarfs ( wds ) may be common and they may provide our best chance to detect biomarkers on an exoplanet in the near future @xcite .
wds are as common as sun - like stars , and they provide an energy source for planets for billions of years .
typical wds have @xmath2 , and @xmath3 , so a planet must orbit at @xmath40.01 au to be at a temperature for liquid water to exist on its surface .
the habitable zone around wds extends from 0.005 au ( p = 4
h ) to 0.02 au ( p = 30 h , * ? ? ?
* ; * ? ? ?
* ) for wds older than about 1 gyr .
figure [ fig : agol ] shows the evolution of the habitable zone around wds ( wdhz ) with time .
a planet enters at the bottom of figure [ fig : agol ] and moves vertically up the figure as its wd host ages , so it starts off too hot for liquid water , passes through the wdhz , and then becomes too cold .
the duration a planet spends within the wdhz has a maximum of 8 gyr at 0.01 au ; wds have long - lived habitable zones .
we expect the planets within 1 au of solar type stars to be destroyed in the giant phase ( * ? ? ?
* but see silvotti et al .
hence , planets in the habitable zone around wds must arrive there after this phase .
@xcite , @xcite , @xcite , and @xcite describe several ways to form such planets near the wd or bring them closer .
planets can form out of gas near the wd , via the interaction or merger of binary stars @xcite , or by capture or migration from larger distances .
planets have been detected around evolved , post - main - sequence stars , including millisecond pulsars and pulsar + wd systems @xcite .
@xcite have a dynamical instability model that can produce short period planets on gyr timescales ; the same mechanism is likely at work after the giant phases and could populate the habitable zone around wds .
the discovery of close - in planets around the post - main - sequence stars kic 05807616 and v391 pegasi demonstrate that such planets should also exist around wds @xcite . @xcite and
@xcite have emphasized that the tidal heating of a planet , until it achieves full circularization , in the wdhz would lead to the loss of water .
however , the young earth was also a hot and dry place , but volatiles and water were then delivered to it by a large number of comets .
it is likely that the process that populates the wd habitable zone could also lead to a heavy bombardment phase that could deliver water and volatiles to these planets @xcite .
about 30% of the wds in the solar neighborhood have metal - polluted atmospheres @xcite .
the source of metals is not accretion from the interstellar medium @xcite .
in addition , @xmath54.3% of wds host debris disks
@xcite , which are the remnants of tidally destroyed asteroids and planets ( e.g. , * ? ? ?
hence , there is direct evidence from these debris disks that the interactions between giant planets can send asteroids , moons , or small ( earth - like ) planets closer to the wd @xcite . therefore , short - period planets around wds likely exist ; but we have never looked at enough wds to find them .
wds are about the same size as earth .
hence , earth - size ( and even smaller ) planets can easily be detected through photometric observations @xcite .
figure [ fig : eclipse ] shows model light - curves of a 0.6 @xmath6 wd as it is eclipsed by a planet with @xmath7 and another with @xmath8 orbiting in the habitable zone of the wd . in the case
of the 1 @xmath9 planet , the eclipse produces a reduction in flux of 50% , whereas a 10 @xmath9 super - earth would effectively block out the wd completely .
it is this high contrast ratio between the planet and the host wd that makes it possible to obtain an atmospheric transmission spectrum of planets in the wdhz with future telescopes , like the james webb space telescope ( jwst , * ? ? ?
* ; * ? ? ?
the most prominent biomarker on earth , molecular oxygen ( o@xmath10 ) , can be detected in the transmission spectrum of an earth - like planet orbiting wds ( see fig . 1 in * ? ? ?
* ) . to detect these 2-min long transits
, one must monitor at high cadence a large sample of wds for the duration of the orbital period at 0.02 au ( @xmath11 h ) . assuming random inclinations ,
the transit probability is roughly 1% for earth - size planets . to measure @xmath12 ,
the frequency of earth - like planets , to an accuracy of 33% requires detecting 9 planets , so one must survey @xmath13 wds .
similarly , @xcite argue that a survey of at least 500 wds is required to detect transiting planets around them . for values of @xmath1410% ,
more stars must be observed to detect planets , so the best strategy is to observe multiple wds simultaneously with a wide - field imager .
given the unknown frequency of exoplanets in the wdhz , a survey of @xmath15 wds would be essential in constraining the frequency and types of planets in the wdhz . a ground - based telescope with a modest field of view ( less than a couple sq .
deg . ) would have to observe the stars one at a time .
so @xmath1 wds would require 2 - 4@xmath16 days ( including overhead ) , or about 100 years . for comparison
, kepler can perform the same survey in 200 days .
there are no planned or ongoing surveys that will accomplish the survey requirements ; 30 h long observations of @xmath15 wds with 1 - 2 min cadence .
the transiting exoplanet survey satellite ( tess ) is not sensitive enough to perform a transit survey for a large number of wds .
there are also no ground - based telescopes with sufficient aperture size and wide field - of - view to perform a survey of @xmath1 wds and to obtain continuous coverage over 2 days .
for example , the large synoptic survey telescope ( lsst ) and gaia @xcite surveys will obtain 50 to 1000 epochs of observations over a decade , possibly detecting one epoch in eclipse for a few per cent of the wds with transiting planets .
these surveys will be biased toward detecting shorter period and larger planets that have yet to enter the wdhz .
in addition , confirmation of the planets around many of the faint lsst wds will be challenging . on the other hand ,
the transits detected in our proposed kepler survey can be easily followed - up with ground - based telescopes and , most importantly , with the jwst .
kepler in the 2-wheels mode provides an unprecedented opportunity to perform a photometric survey that is capable of finding the first exoplanets in the wdhz .
thanks to its sky coverage and depth , for the first time , kepler can image a large number of wds at the same time and search for transiting planets and other substellar or stellar companions .
the majority of the known wds are in the sdss fields .
hence , kepler imaging of the known wds in the sdss fields provides the best opportunity for such a transit survey .
there are @xmath17 spectroscopically confirmed wds in the sdss dr7 area @xcite .
we have identified an additional @xmath18 wds in the same area through proper motion measurements using the sdss + usno - b astrometry @xcite . even though many of these targets lack spectroscopy data , @xcite
demonstrate that this sample is almost pure , with a contamination rate from subdwarfs of 1% .
overall , there are @xmath19 unique wds currently known in 11,663 sq .
deg . of imaging in the sdss dr7 area .
figure [ fig : sdss ] shows the magnitude and color distribution of these targets .
there are @xmath13 and @xmath1 wds with @xmath20 mag and @xmath21 mag , respectively .
the latter include 2253 cool wds with @xmath22 k. we propose a kepler imaging survey of the @xmath1 sdss wds brighter than @xmath21 mag to discover stellar , substellar , and planetary ( including sub - earth - and - up companions ) around all wds .
this survey will measure the planetary frequency and survival around both young and old systems and discover planets in the wdhz .
thanks to the sdss astrometry , accurate positions are known for each target and target apertures can be defined easily . the limiting factor for kepler observations of faint sources
is set by source confusion , rather than the photometric accuracy .
all of our targets have sdss imaging down to @xmath23 mag , and the usno - b+sdss proper motion catalog @xcite avoids sources with nearby background sources within @xmath24 .
hence , source confusion is not a problem for our targets .
the proposed survey will image each field for 2 days , with 1 min cadence .
this survey will be sensitive to planets within 0.03 au of the stars .
imaging the dr7 area requires about 100 pointings .
hence , the entire survey can be done over 200 days .
the nominal kepler mission provided @xmath450 ppm photometry for a 12 mag g2v star integrated over 30 minutes , enabling discoveries of earth - like planets around main - sequence stars . given the pointing problems in the 2-wheels mode , the image will be spread out over more pixels , and the increased read noise and larger responsivity variations will reduce the photometric precision to @xmath40.3 - 1% .
given the eclipse signature of earth - size and larger planets around wds ( see figure [ fig : eclipse ] ) , the systematic errors due to the pointing problems is not the limiting factor for wdhz observations . instead , the brightness of the target wds limits the delivered photometric precision .
kepler can obtain 1% photometry for targets brighter than about 16.5 mag , and 10% photometry for targets brighter than 18.7 mag .
hence , the 50% eclipse signature of an earth - like planet can be detected at @xmath25 for targets as faint as 18.7 mag .
it may be possible to obtain photometry of the fainter targets in the same fields by using a smaller number of pixels for computing the light curve .
figure [ fig : sdss ] shows that the number of objects rises dramatically between @xmath26 and 20 mag .
hence , the number of targets can easily be doubled by a @xmath40.5 mag deeper survey ; operational testing of the 2-wheels mode is necessary to decide on the limiting magnitude for such observations .
kepler can effectively search for transiting earth - size planets around @xmath1 wds in the sdss . with its unprecedented photometric sensitivity
, kepler could also detect smaller occulting objects down to the scale of the earth s moon , albeit with lower significance .
in addition to planetary companions , kepler will also be sensitive to short period substellar and stellar companions
. this sample will be essentially complete to @xmath27 d. however , eclipsing systems with longer periods will also be detected .
about 22% of field wds have late - type main - sequence star companions @xcite , while 3.4% of these systems are eclipsing with @xmath28 day orbits @xcite .
hence , the frequency of eclipsing wd + m dwarf binaries is likely around 0.75% . in a sample of @xmath1 wds , we are likely to find 75 eclipsing wd + m dwarf binaries .
earth - density planets will not survive at orbital periods shorter than 4 h. hence , eclipses with @xmath29 h are almost certainly from stellar or brown dwarf companions .
given the differences in size , the eclipse signatures of wd + earth - size planets and wd + stellar binaries would be significantly different and easy to distinguish based on the light - curves .
in addition , follow - up near - infrared photometry and optical radial velocity observations of the eclipsing systems can easily distinguish between planetary companions and brown dwarf or stellar companions .
stellar binaries will show primary and secondary eclipses of different depths in different filters , offset secondary eclipses due to light travel time , gravitational lensing @xcite , and doppler beaming @xcite .
these systems would provide stringent constraints on the mass - radius relations for wds and late - type m dwarfs , and also age - activity relation for the latter .
it is widely accepted that sne ia are caused by the thermonuclear explosion of wds @xcite , but the nature of the progenitor binary systems is not settled . until we find the progenitors , it is impossible to understand the systematic uncertainties and to optimize sne ia for precision cosmology .
the sne ia progenitor problem is therefore a key problem in astronomy today @xcite .
recent theoretical studies suggest that sub - chandrasekhar mass wd systems may also form sne ia ( e.g. , * ? ? ?
* ; * ? ? ?
major contenders for sn ia progenitors are mergers of binary wds .
discovering and quantifying the population properties of close double wds , including co+co and co+he binary wd systems , can provide key insights into the sne ia progenitor problem @xcite .
there are currently four eclipsing double wd systems known with orbital periods ranging from 12.75 min @xcite to 5.9 h @xcite .
the lack of eclipsing systems at longer orbital periods is a selection effect due to the difficulty of observing a target for longer than 6 h a night .
the proposed kepler imaging survey will identify all of the eclipsing double wd merger systems among our targets , thanks to the 2 day long observations .
these eclipses last 2 - 3 min , and include total eclipses for similar mass wds . based on a population synthesis model for binary stars ,
@xcite predict a frequency of 2.5% for double wds with @xmath30 h. our proposed survey will provide the first constraints on the frequency of eclipsing double wds and their merger rate .
eclipsing double wds provide the most accurate constraints on the masses and radii of each target .
this will in turn provide the mass distribution and merger times for each system and constrain the contribution of the double degenerate channel to sne ia .
this project may finally provide direct evidence for or against the double degenerate channel for sne ia .
the shortest period binary wds are also excellent gravitational wave sources .
the gravitational wave radiation and the orbital decay in the shortest period systems may be detected directly by space based missions like the laser interferometer space antenna ( lisa ) and indirectly by ground based observations ( see * ? ? ?
@xcite lists 12 ultra - compact systems that are guaranteed lisa sources , but predicts that lisa should detect at least several hundred systems .
double wds in the galaxy outnumber all other known types of gravitational wave sources and they form a gravitational wave foreground
. some of the eclipsing systems found in the proposed kepler survey will be amongst the strongest gravitational wave sources known . with accurate positions , masses , and distance estimates
, these will be verification sources for gravitational wave missions in the milli - hertz range ( e.g. , elisa , * ? ? ?
wds go through the dov , dbv , and dav instability strips as they evolve .
the zz ceti ( dav ) stars are the most common type of pulsators , with @xmath4150 currently known .
davs are found in a very well defined region in the @xmath31 plane .
they exhibit pulsation periods in the range 100 - 1400 s , corresponding to low - degree gravity - mode oscillations .
the detected pulsation modes have amplitudes ranging from a few millimagnitudes to a fraction of a magnitude ( @xmath410% ) .
the cooler pulsators tend to show longer periods and larger amplitudes . also , for a given effective temperature
, the observed periods tend to be longer for lower gravity objects @xcite .
high speed photometric studies of the pulsating wds enable us to constrain the structure and evolutionary timescales of wds .
for example , the pulsation modes and amplitudes are sensitive to the thickness of the surface h and he layers , the depth of the convection zone , and the degree of crystallization ( e.g. * ? ? ?
there are @xmath41500 wds with @xmath3212,000 k in our proposed kepler survey .
these wds are in the right temperature range to pulsate as a dav .
kepler will be sensitive to large amplitude pulsations , and as a bonus , it will provide a magnitude - limited sample of pulsators . with 2 d coverage , the majority of the modes can be identified , and the most interesting targets can be followed up from the ground for further characterization and theoretical modeling of the structure of each star .
kepler will significantly increase the number of known pulsating wds and change this field dramatically .
before the discovery of hot jupiters around main - sequence stars , our view of planet formation was very different . the discovery of hot jupiters was unexpected and unintentional .
here we outlined an alternate mission for kepler in the 2-wheels mode targeting a large sample of wds to find the first planets in the habitable zone . if the history of exoplanet science has taught us anything , it is that planets are ubiquitous and they exist in the most unusual places , including very close to their host stars and even around pulsars @xcite
. currently there are no known planets around wds , but we have never looked at a sufficient number of wds at high cadence to find them through transit observations .
it is essentially impossible to find earth - jupiter size planets around wds by any other method @xcite .
if habitable planets exist around wds , the proposed kepler imaging survey will find them .
biomarkers , including o@xmath10 , on such planets can be detected with the jwst .
hence , even though this is a completely unexplored search area for transiting planets , the scientific yield of the proposed survey will be enormous .
in addition to the planetary studies , the proposed survey will also dramatically change the field of short period binary wds , including eclipsing double wds , wd + m dwarfs , and wd + brown dwarfs .
this survey will constrain the merger rate and mass distribution of double wds , which is important for binary population synthesis studies , understanding sne ia explosions , and identifying new verification binaries and the galactic foreground in gravitational waves .
short cadence targets are limited to no more than 512 across the entire focal plane .
our proposed survey has a density of @xmath33 short cadence targets per pointing .
therefore , this survey can be combined with other projects targeting the sdss fields .
for example , similar high ( or low ) cadence observations of m dwarfs ( e.g. , the mearth project , * ? ? ? * ) or other types of ( variable ) stars in the sdss fields can be obtained simultenously .
nearby m dwarfs can be easily selected based on their high proper motion from the sdss + usno - b positions @xcite .
alternatively , the sdss photometry can be used to select a nearly complete and contamination free sample of variable stars like rr lyrae over the entire sdss footprint .
the survey can also be extended to an all - sky survey through the selection of wds from other proper motion surveys like the usno - b , supercosmos @xcite , and lspm @xcite .
if this survey is not selected as the primary mission for kepler in the 2-wheels mode , it should at least be considered as a filler project for other surveys .
agol , e. 2002 , , 579 , 430 agol , e. 2011 , , 731 , l31 badenes , c. , & maoz , d. 2012 , , 749 , l11 barber , s. d. , patterson , a. j. , kilic , m. , et al .
2012 , , 760 , 26 barnes , r. , & heller , r. 2013 , astrobiology , 13 , 279 berta , z. k. , irwin , j. , & charbonneau , d. 2013 , arxiv:1307.3178 beuermann , k. , dreizler , s. , hessman , f. v. , & deller , j. 2012 , , 543 , a138 brown , w. r. , kilic , m. , hermes , j. j. , et al . 2011 , , 737 , l23 charpinet , s. , fontaine , g. , brassard , p. , et al .
2011 , , 480 , 496 debes , j. h. , & sigurdsson , s. 2002 , , 572 , 556 di stefano , r. , howell , s. b. , & kawaler , s. d. 2010 , , 712 , 142 drake , a. j. , beshore , e. , catelan , m. , et al . 2010 ,
arxiv:1009.3048 faedi , f. , west , r. g. , burleigh , m. r. , goad , m. r. , & hebb , l. 2011 , , 410 , 899 farihi , j. , becklin , e. e. , & zuckerman , b. 2005 , , 161 , 394 gianninas , a. , bergeron , p. , & fontaine , g. 2006 , , 132 , 831 gould , a. , & kilic , m. 2008 , , 673 , l75 hermes , j. j. , kilic , m. , brown , w. r. , et al .
2012 , , 757 , l21 hermes , j. j. , montgomery , m. h. , winget , d. e. , et al .
2013 , , 765 , 102 howell , d. a. 2011 , nature communications , 2 , 350 iben , i. , jr . , &
tutukov , a. v. 1984 , , 54 , 335 jura , m. 2003 , , 584 , l91 kilic , m. , & redfield , s. 2007 , , 660 , 641 kilic , m. , leggett , s. k. , tremblay , p .- e . , et al . 2010 , , 190 , 77 klein , b. , jura , m. , koester , d. , & zuckerman , b. 2011 , , 741 , 64 kleinman , s. j. , kepler , s. o. , koester , d. , et al .
2013 , , 204 , 5 kratter , k. m. , & perets , h. b. 2012 , , 753 , 91 lpine , s. , & shara , m. m. 2005 , , 129 , 1483 livio , m. , pringle , j. e. , & wood , k. 2005 , , 632 , l37 loeb , a. , & gaudi , b. s. 2003 , , 588 , l117 loeb , a. , & maoz , d. 2013 , , 432 , l11 monteiro , h. 2010 , bulletin of the astronomical society of brazil , 29 , 22 montgomery , m. h. , provencal , j. l. , kanaan , a. , et al .
2010 , , 716 , 84 munn , j. a. , monet , d. g. , levine , s. e. , et al .
2004 , , 127 , 3034 nelemans , g. 2009 , classical and quantum gravity , 26 , 094030 nordhaus , j. , & spiegel , d. s. 2013 , , 432 , 500 pakmor , r. , kromer , m. , taubenberger , s. , & springel , v. 2013 , , 770 , l8 parsons , s. g. , gnsicke , b. t. , marsh , t. r. , et al .
2013 , , 429 , 256 passy , j .- c .
, mac low , m .- m .
, & de marco , o. 2012 , , 759 , l30 perryman , m. a. c. , de boer , k. s. , gilmore , g. , et al . 2001 , , 369 , 339 rowell , n. , & hambly , n. c. 2011 , , 417 , 93 seoane , p. a. , et al . 2013 ,
arxiv:1305.5720 sigurdsson , s. , richer , h. b. , hansen , b. m. , stairs , i. h. , & thorsett , s. e. 2003 , science , 301 , 193 silvotti , r. , schuh , s. , janulis , r. , et al .
2007 , , 449 , 189 sim , s. a. , fink , m. , kromer , m. , et al .
2012 , , 420 , 3003 vennes , s. , thorstensen , j. r. , kawka , a. , et al .
2011 , , 737 , l16 veras , d. , mustill , a. j. , bonsor , a. , & wyatt , m. c. 2013 , , 431 , 1686 villaver , e. , & livio , m. 2007 , , 661 , 1192 webbink , r. f. 1984 , , 277 , 355 wolszczan , a. , & frail , d. a. 1992 , , 355 , 145 wu , y. , & lithwick , y. 2011 , , 735 , 109 zucker , s. , mazeh , t. , & alexander , t. 2007 , , 670 , 1326 zuckerman , b. , melis , c. , klein , b. , koester , d. , & jura , m. 2010 , , 722 , 725 | a large fraction of white dwarfs ( wds ) may host planets in their habitable zones .
these planets may provide our best chance to detect bio - markers on a transiting exoplanet , thanks to the diminished contrast ratio between the earth - sized wd and its earth - sized planets .
the james webb space telescope is capable of obtaining the first spectroscopic measurements of such planets , yet there are no known planets around wds . here
we propose to take advantage of the unique capability of the kepler spacecraft in the 2-wheels mode to perform a transit survey that is capable of identifying the first planets in the habitable zone ( @xmath0 4 - 30 h ) of a wd .
we propose to obtain kepler time - series photometry of @xmath1 wds in the sloan digital sky survey imaging area to search for planets in the habitable zone .
thanks to the large field of view of kepler , for the first time in history , a large number of wds can be observed at the same time , which is essential for discovering transits .
our proposed survey requires a total of 200 days of observing time , and will find up to 100 planets in the wd habitable zone .
this survey will maintain kepler s spirit of searching for habitable earths , but near new hosts .
with few - day observations and minute - cadences per field , it will also open up a completely unexplored discovery space .
in addition to planets , this survey is sensitive to pulsating wds , as well as eclipsing short period stellar and substellar companions .
these have important implications for constraining the double wd merger rate and their contribution to type ia supernovae and the gravitational wave foreground . given the relatively low number density of our targets ,
this program can be combined with other projects that would benefit from high cadence and ` many - fields ' observations with kepler , e.g. a transit survey of a magnitude - limited , complete sample of nearby m dwarfs or asteroseismology of variable stars ( e.g. rr lyrae ) in the same fields . |
nowadays the electron bernstein waves ( ebws ) having no density cut - offs and effectively damped even at high electron cyclotron harmonics are considered as the most promising candidate to provide an auxiliary heating and current drive in a dense plasma of a spherical tokamaks and stellarators @xcite .
the ebw could be excited via direct conversion of x mode to bernstein mode in a vicinity of the upper hybrid resonance ( uhr ) or via so - called o - x - b scheme .
the efficiency of last scheme , as it was demonstrated theoretically @xcite in the frame of 1d slab model , is determined by the efficiency of o to x mode conversion , which can reach 100 percent value at the certain parallel refractive index being constant in slab geometry . due to , in a real spherical tokamak s configurations , where the poloidal inhomogeneity of the magnetic field is important , both the parallel refractive index is no longer constant and the components of the dielectric tensor are functions of two co - ordinates , an analysis of the full - wave equations in the frame of 2d model is important .
the first attempt to consider 2d model of ox conversion has been undertaken a couple of years ago @xcite . the main conclusion provided by author of @xcite concerning the absence of the o mode reflection from the o - mode cut - off surface seems to be quite doubtful and lacks support from the last two papers devoted to this topic @xcite .
unfortunately , the ox mode conversion in @xcite has been considered in the frame of the oversimplified model ignoring as it does both the poloidal magnetic field and , as result , varying of the parallel refractive index on the magnetic surface .
these effects were taken into account in @xcite , where the the explicit expressions for the ox and xo conversion coefficients have been obtained simultaneously with @xcite but for @xmath0 ( @xmath1 and @xmath2 are the poloidal and toroidal components of the magnetic field and @xmath3 is an angle between the toroidal direction and the magnetic field ) . unlike the assumption used in @xcite , @xmath4 , in spherical tokamak s plasma @xmath3 is no longer small as it is demonstrated in figure 1 for a typical mast tokamak discharge . therefore ,
straightforward using the results of @xcite , without mentioning paper @xcite , to describe ox conversion in spherical tokamaks seems to be overhasty . because of
the intensive efforts are paid to provide an auxiliary heating in spherical tokamaks using oxb scheme , calculation of the ox conversion coefficient in the realistic 2d model is of great interest . in this paper a reduced set of the partial differential wave equations valid in a vicinity of the o - mode cutoff and accounting for the magnetic field 2d inhomogeneity with no restriction to an angle between the toroidal direction and the magnetic field are derived . a solution to the reduced set of equations
has been obtained and it properties are considered in details .
there are three effects that remain beyond the scope of the present paper .
we neglect , first , the curvature of the magnetic field line at the magnetic surface because of its local radius @xmath5 is considerably greater than the beam radius @xmath6 , second , the curvature of the magnetic flux surfaces assuming high localization of the conversion region , third , the magnetic field shear which is not important for ox conversion @xcite . we restrict ourselves to the case of not extremely strong plasma density inhomogeneity @xmath7 when geometrical optics can be applied except possibly near cut - off layer or possibly mode conversion layer .
one introduces two cartesian co - ordinate systems @xmath8 and @xmath9 with distances scaled in the units @xmath10 and their origin located at the o - mode cut - off surface ( figure 2 ) .
the co - ordinates @xmath11 , @xmath12 and @xmath13 imitate the flux surface label , the poloidal and the toroidal co - ordinates , respectively , and axes @xmath14 and @xmath15 are along the magnetic field and perpendicular to it on the o - mode cut - off surface , respectively ( figure 3 ) . the transformation from @xmath12 , @xmath13 components to @xmath16 ,
@xmath17 components convenient to represent the maxwellian equations is @xmath18 one introduces the set of the wave equations for a monochromatic wave @xmath19 as @xmath20 where @xmath21 , @xmath22 @xmath23 @xmath24 @xmath25 , @xmath26 and @xmath27 are electron langmuir and cyclotron frequencies , respectively .
since a domain in a vicinity of o - mode cut off surface beyond the tokamak mid - plane is of interest , one can expand the plasma parameters , namely density @xmath28 and magnetic field modulus @xmath29 , into the taylor serious at @xmath30 @xmath31 where @xmath32 are parameters being the first order quantity ( @xmath33 ) and @xmath34 is one being the second order quantity ( @xmath35 ) . in order to study the properties of the waves in the mode conversion region we keep in mind that the component @xmath36 is small compared to two others
@xmath37 there . omitting in ( 2 ) the terms being higher order quantity than the first one obtains @xmath38 due to tokamak symmetry we may assume that the wave fields vary as @xmath39 , where @xmath40 being constant is large .
we would like to construct a solution @xmath41 of the system ( 6 ) .
to this end we look to develop the integral representation for required solutions of laplace integral type @xmath42 where @xmath43 is assumed analytic in some domain and the path of integration in the @xmath44 plane is such that the integrand vanishes rapidly at the ends of the contour @xmath45 or at infinity . next @xmath46 where @xmath47 and @xmath48 we can not easily deal with the system ( 8) as it stands after transformation , but recalling the conclusions of 1d theory @xcite that the effective conversion is possible if both @xmath49 @xmath33 and @xmath50 @xmath51 we may reduce it to a simpler form .
expanding @xmath52 around @xmath53 and @xmath54 around @xmath55 we have found that @xmath56 and @xmath57 . keeping in ( 8) terms being the first order quantity one has @xmath58 where the explicit expressions for @xmath59 and @xmath60 could be found by varying ( 9 ) . making the backward transformation with respect to @xmath61 yields @xmath62 where the electric field s components @xmath63 obey the system of equations @xmath64 introducing new notations @xmath65
@xmath66 we read ( 13 ) in the compact form @xmath67 in order to make the first step in analysis of the electric field components behavior in the conversion region , we consider in the next section wkb solution to the system ( 14 ) , namely we focus on the quality analysis of the ray trajectory along which the energy of the incident beam of the ordinary ( or extraordinary ) waves is carried over .
although in the mode conversion region the wkb approximation breaks down and a local full wave equations ( 14 ) to be solved , the ray trajectory analysis could be useful .
the ray representation of the wave field in four dimensional ray phase space @xmath68 @xmath69 is governed by hamilton s equations : @xmath70 where @xmath71 is the local dispersion relation and @xmath72 denotes the orbit parameter .
since the general picture of the ray behavior in the two - dimensional subspace @xmath73 is of interest , let us re - parameterize @xmath74 , where @xmath75 . substituting ( 16 ) into ( 15 ) we obtain @xmath76 where @xmath77 .
the equations constituting ( 17 ) are coupled with @xmath78 being the coupled coefficient .
one can introduce the normal co - ordinates @xmath79 reducing the system of equations ( 17 ) to two independent ones .
the set of linear independent solutions for either of the two is @xmath80 the expressions ( 18 ) , which show the projections of hamiltonian rays on the ( @xmath81 ) plane , deserves a few comments .
first , note the oscillatory behavior along the @xmath82 direction , while the @xmath83-motion displays the influence of a retarding force .
second , due to @xmath84 oscillations along the @xmath82 direction are not to be clear - cut .
third , we can anticipate the full wave solution to the set of equations ( 14 ) could be represented as the superposition of the eigenmodes intrinsic to the confining potential ( @xmath85 ) in the @xmath82 direction leading to oscillatory @xmath82 motion and the superposition of the linear independent parabolic cylinder functions along @xmath83 direction , which proper for the effective potential @xmath86 .
the full - wave solution to the set of equations ( 14 ) is studied in the next section .
we would like to obtain the appropriate particular solution of the system ( 14 ) matching to the wkb solution outside the mode conversion region @xmath87 where * e * is the polarization vector , @xmath88 is a set of eigen functions to be found , @xmath89 and @xmath90 correspond to incident o mode ( ox conversion ) and incident x mode ( xo conversion ) .
with that end in view we introduce two possible functional substitutions @xcite providing the identity of the equations constituting ( 14 ) @xmath91 with @xmath92 and @xmath93 being required functions that satisfy the equations @xmath94 let us use the transformation @xmath95 where @xmath96 @xmath97 and @xmath98 , that makes the potential in the brackets in ( 22 ) symmetrical @xmath99 if @xmath100 ( see figure 4 ) then @xmath101 .
seeking the solutions to ( 24 ) in the form @xmath102 where the definition for @xmath103 is given by @xmath104 and @xmath105 are hermitian polynomials , we obtain the equation for @xmath106 and @xmath107 @xmath108 where @xmath109 are important for subsequent analysis we list theirs below expressing its argument @xmath110 explicitly @xmath111 the solutions @xmath106 and @xmath107 to ( 27 ) matching to the wkb solutions ( 19 ) for the incident ordinary and extraordinary waves ( see appendix a ) are @xmath112 where @xmath113 are an arbitrary constants , which we choose so that ( 20 ) and ( 21 ) fit the incident wkb wave outside the conversion layer .
as it is demonstrated in appendix a @xmath114 and @xmath115 correspond to the incident ordinary and extraordinary waves , respectively .
the parameter @xmath116 has meaning the length of an evanescence layer ( see the equation ( a.3 ) ) .
its value depends on the mode s number @xmath117 and combination of the parameter s @xmath110 sign and the direction of the process ( see ( 28 ) ) .
that is , first , either of the mode tunnels through an evanescence layer with the efficiency inherent itself , second , the conversion efficiency being in 1d model the invariant under the conversion s direction reversal and the poloidal position of the incident beam changing has no longer the same property in 2d inhomogeneous plasma .
being mentioned for the first time in @xcite for the oversimplified plasma model ignoring the poloidal magnetic field and confirmed in @xcite for the reasonable plasma model accounting for the poloidal magnetic field @xmath118 this property depends entirely on the essential two - dimensional character of the waves . inserting ( 29 ) and ( 30 ) into ( 20 ) , ( 21 ) and ( 11 ) after simple algebra we obtain an integral representation of solution to ( 6 ) @xmath119 where upper signs at @xmath120 correspond to the incident ordinary wave and lower signs correspond to the incident extraordinary wave .
the explicit expressions for @xmath121 and @xmath122 are @xmath123 and @xmath124 while outside the mode conversion region at @xmath125 @xmath126 and @xmath127 match to the geometrical optics solutions that correspond to the wkb incident and conversed waves ( compare the first terms in the r.h.s .
of ( a.1 ) and ( a.2 ) , ( a.6 ) and ( a.7 ) ) , the wkb asymptotics of @xmath128 and @xmath129 correspond to the wkb wave reflected from the evanescence layer ( see ( a.5 ) ) .
it is possible to obtain asymptotic expansion of the solution for @xmath130 . for the incident ordinary wave at @xmath131 we have @xmath132
last expressions recover the expressions for the electric field components derived in @xcite for plasma with @xmath118 .
for usual spherical tokamak s configuration o mode cut - off layer and the x mode cut - off layer cross at the small angle , i.e. @xmath133 . keeping only terms in ( 23 ) being zero and first order quantity with respect to the parameter @xmath78 we read it as @xmath134 then we simplify @xmath135 where @xmath136 and @xmath12 is scaled in the units @xmath137 . comparing the asymptotic representation of @xmath138 for the argument @xmath139 and asymptotic representation of @xmath138 for the argument @xmath140 ( see appendix a ) we obtain an explicit expressions for the conversion coefficient , @xmath141 , ( in energy ) and reflection coefficient , @xmath142 , where the subscribes @xmath143 and @xmath144 correspond to the direction of the process @xmath145 and @xmath146 the coefficient @xmath147 is given by the expression @xmath148 where @xmath149 means the complex conjugation , @xmath150
is given field distribution in the incident beam in the wkb region , @xmath11 and @xmath13 are scaled in @xmath151 and @xmath152 is a normalizing coefficient proportional to the beam intensity .
the expressions ( 38 ) and ( 39 ) derived similar to ones obtained in @xcite excepting the formula for @xmath153 given by ( 40 ) and deserve one comment .
noting @xmath154 at the fixed @xmath78 , we could read the reciprocal relation for the conversion coefficients in 2d inhomogeneous plasma as @xmath155 where the argument @xmath156 is presented explicitly .
for the first time the property ( 41 ) was mentioned in @xcite . in
@xcite it was proved for plasma with poloidal magnetic field @xmath157 . for illustration , in figure 5 the conversion coefficient , @xmath158 , of an incident beam with transversal distribution @xmath159 calculated with the use of 1d model @xcite ( dashed curve ) and the formula ( 38 ) ( solid curves ) for @xmath160 and different @xmath3 versus the dimensionless radius , @xmath161 , is shown .
calculations are executed under conditions of ebw heating experiment in mast @xmath162 , @xmath163 , @xmath164 , @xmath165 .
rather strong variation of dependence presented in figure 5 with variation of @xmath3 is observed .
the mode conversion of o mode to x mode have been examined in a spherical tokamak geometry for plasma with a cold plasma dielectic tensor .
we have assumed the wavelengths are much shorter than the equilibrium plasma gradient length , @xmath7 .
thus , we expect wkb approximation to apply except possibly near o - mode cut - off surface or mode conversion layer .
expanding wave equations in the region near the intersection of o and x mode cut - off surfaces , one finds reduced wave equations appropriate for this region which depends on two coordinates , @xmath11 and the poloidal angle @xmath12 .
we seek the required solution to the set of equations so that it matches to the wkb solution outside the mode conversion region .
after functional substitution we have found the integral representations of the required solution . to the case typical of a spherical tokamak s configuration for which the o mode cut - off layer and the x mode cut - off layer cross at the small angle we have been simplified the obtained integral representations of solution and found the wave fields and the conversion coefficients explicitly .
we have shown that at the fixed poloidal position , i.e. fixed @xmath78 , the conversion coefficients obey the reciprocial relation ( 41 ) as it was shown earlier in the frame of the model neglecting the poloidal magnetic field @xcite and of the model in which the poloidal magnetic field is assumed to be small , @xmath118
. the importance of 2d effects has been demonstrated by the example under the mast conditions .
the work was supported by rfbr grants 04 - 02 - 16404 , 06 - 02 - 17212
we pay our attention to the functions of the parabolic cylinder with different arguments and indices @xmath166 and @xmath167 .
let us begin from asymptotic expressions for the first one at two limits @xmath125 which describe the corresponding wkb solutions beyond the conversion layer .
when the argument @xmath83 tends to @xmath168 @xcite @xmath169 while at @xmath170 @xmath171 the second term in ( a.2 ) is much smaller ( with factor @xmath172 ) compared to the first one .
therefore the function @xmath166 in the wkb sense describes the wave approaching an evanescence layer from @xmath173 and conversed wave propagating from an evanescence layer to @xmath174 . comparing the coefficients at @xmath175 in ( a.1 ) and
( a.2 ) one obtains the conversion coefficient ( in energy ) for the wave described by the parabolic cylinder equation @xmath176 farther we obtain the derivative of the parabolic cylinder function @xmath177 the first term in the r.h.s .
of ( a.4 ) , which has the asymptotics @xcite @xmath178 describes the reflected wave . comparing the asymptotic expansions for the parabolic cylinder function @xmath167 @xcite @xmath179 and @xmath180 we note that it describes the wave which incident on an evanescent layer from @xmath181 ( see the first term in ( a.5 ) ) and propagates from an evanescent layer to the plasma boundary ( @xmath170 ) ( see ( a.6 ) ) .
10 laqua h p , erckmann v , hartfuss h j and w - a ecrh group 1997 _ phys .
lett . _ * 78 * 3467 efthimion p c , hosea j c , kaita r , majeski r and taylor g 1999 _ rev .
instrum . _ * 70 * 1018 shevchenko v , baranov y , obrien and saveliev a 2002 _ phys .
lett . _ * 89 * 265005 prienhalter j and kopecky v 1973 _ j. plasma physics _ * 10 * 1 mjolhus j j 1984 _ plasma phys .
_ 31 7 tokman m d 1985 _ sov . j. plasma phys .
_ 11 689 ( 1985 ) weitzner h 2004 _ phys .
plasmas _ * 11 * 866 gospodchikov e d , shalashov a g , suvorov e v 2006 _ ppcf _ * 48 * 869 popov a yu , piliya a d 2006 _ plasma phys
. reports _ ( to be pub . )
cairns r a , lashmore - davies c n 2000 _ phys
. plasmas _ * 7 * 4126 zharov a a 1984 _ sov
. j. plasma phys .
_ 10 642 _ high transcendental functions _ ( harry bateman ) ( mc graw - hill book company , inc , 1953 ) | in this paper a reduced set of the partial differential wave equations valid in the conversion layer close to o - mode cutoff surface and accounting for the magnetic field 2d inhomogeneity with no restriction to an angle between the toroidal direction and the magnetic field direction is derived .
an integral representation of a solution to the derived set of equation is given .
for the particular case of small angle between o mode cut - off surface and x mode cut - off surface an explicit expressions for both the electric field components and the conversion coefficients are obtained and its properties are considered in details . |
in electroencephalography ( eeg ) electrical currents are recorded at different positions on the scalp to measure the brain activity .
the correlations between the time series of these currents strongly depend on the overall state of the brain . during an epileptic seizure ,
for example , the correlations are much stronger than in normal periods @xcite . this time dependence of the correlations is the kind of non stationarity that we wish to address .
non stationarities are also seen when wave packets travel through disordered systems . even if the disorder is static , the correlations between the wave intensities measured at different positions versus time will change , when the direction or the composition of the wave packet is altered @xcite .
finance provides another important example for this type of non stationarity .
the correlations between stock price time series change in time , just as the business relations between the firms and the traders market expectations @xcite .
similar non stationarities exist in many complex systems , including velocity fluctuations in turbulent flows , heartbeat dynamics , series of waiting times , etc . @xcite .
a system showing non stationary correlations may be interpreted as being out of equilibrium , implying that some of the key tools in statistical physics are not applicable . yet
, the challenges are similar to the one faced for equilibrium systems : is there generic or universal behavior ? how can we identify it ?
can we set up statistical models for these non stationarities ? in the context of finance , we recently put forward a random matrix approach to tackle these issues @xcite .
we also succesfully applied it in a study of credit risk and its impact on systemic stability @xcite .
inspite of the conceptual differences , random matrix theory @xcite formally has much in common with statistical mechanics .
observables are averaged over an ensemble ; in statistical mechanics , it usually is the microcanonical , canonical or macrocanonical one , in random matrix models , it is an ensemble of those matrices which describe or characterize the system . in the context of the present discussion
, random matrix models can be divided into two classes : 1 .
the ensemble is fictitious .
it comes into play via an ergodicity argument only .
the ensemble really exists and can be identified in the system .
the issue of ergodicity does not arise . the vast majority of random matrix models in , _
e.g. _ , quantum chaos falls into class 1 , for a review see ref .
one is interested in the spectral statistics of one individual system .
its hamiltonian is viewed as a random matrix , whose dimension is eventually sent to infinity .
ergodicity holds in this limit , meaning that a smoothing energy average of an observable over one individual spectrum equals the average over an ensemble of random matrices . a noticeable exception
are random matrix applications to quantum chromodynamics @xcite . in lattice gauge theory , the quarks first propagate in frozen configurations of the gauge fields , before an average over the gauge fields , modeled by random matrices , is carried out as second step .
this clearly belongs in class 2 .
the fluctuating gauge fields truly exist , the partition function involves an integral over them .
ergodicity reasoning is not evoked .
there are numerous applications of random matrix theory in finance @xcite which address statistical properties of correlation matrices .
many of them also deal with non gaussian ensembles . to the best of our knowledge ,
all of these applications fall into class 1 , because one is interested in the statistics of one individual correlation matrix , measured at one particular instant in time . in our study
@xcite , we put forward a first application of random matrices in finance that belongs in class 2 .
stationarity makes the covariances fluctuate and thereby creates an ensemble of covariance matrices which we approximated by a gaussian wishart ensemble of random matrices @xcite .
we derived how the multivariate distribution of dimensionless price changes , referred to as returns , acquires heavy tails due to the non stationarity .
hence , we showed that the non stationarities indeed have universal features . here
, we have three goals : first , we present in sec .
[ sec2 ] a statistically significant way to construct a proper and analytically tractable random matrix ensemble from the data .
we emphasize that this is an important issue for random matrix models in the context of correlations .
in contrast to quantum chaos , where universality holds on the scale of the mean level spacing , there is not such a local scale when studying statistical properties of correlation matrices .
thus , a gaussian assumption is not always justified and it does matter what the ensemble looks like in reality . in particular , realistic ensembles considerably help to understand and model large events .
our construction is general and not tied to any specific system .
its merit lies in the fact that once the enemble is known , it can be used to work out generic statistical properties of any observable depending on the correlation matrices , see ref .
@xcite for an example .
second , we apply our approach to financial data in sec .
we identify an algebraic ensemble , which is quite relevant for risk estimation .
third , we discuss two issues arising in our general construction in secs .
[ sec4 ] and [ sec5 ] , namely a certain conceptual caveat and yet a further extension , respectively .
conclusions are given in sec .
after setting up the general problem in sec .
[ sec21 ] , we introduce the deformed wishart ensemble and derive the correponding amplitude distribution in sec . [ sec22 ] .
the determination of the deformation functions which characterize the ensemble and the amplitude distribution is discussed in sec .
[ sec23 ] . here
, we derive the approach for the general case , for sake of illustration , the reader is referred to ref .
@xcite and sec .
[ sec3 ] suppose we have measured in a system with randomness @xmath0 amplitudes as time series @xmath1 over a long interval @xmath2 of time @xmath3 . for examples ,
these amplitudes can be electric or magnetic fields at @xmath0 different points in a disordered system , positions of @xmath0 randomly moving particles or financial returns , _
i.e. _ dimensionless price changes for @xmath0 stocks .
importantly , we assume that there are correlations between the time series . in complex systems ,
one often encounters the situation that crucial system parameters , in particular the covariances or correlations , are seemingly random functions of time @xcite . to be more precise ,
we consider a time window of length @xmath4 that is much shorter than the total interval , @xmath5 .
we now want to average over the subinterval @xmath6 $ ] of length @xmath4 whose position in the total interval is determined by the time @xmath7 .
sample averages of a function @xmath8 in this subinterval are then written as @xmath9 we are particularly interested in the covariances @xmath10 where we introduced the amplitudes normalized to zero mean value , @xmath11 we keep in mind that the resulting @xmath12 covariance matrix @xmath13 is calculated from time series of length @xmath4 .
we now move this time window of length @xmath4 through the data , the resulting covariances @xmath14 fluctuate .
this non stationarity has an important impact on other statistical observables . in the present study ,
we focus on the distribution of the amplitudes .
we now consider a time interval @xmath4 as short as possible such that the covariance matrix @xmath15 in this time interval is in good approximation constant .
we begin with addressing the case in which the distribution of the amplitudes is , for a given time @xmath7 , well approximated by a multivariate gaussian @xmath16 with the @xmath0 component vector @xmath17 and the @xmath12 covariance matrix @xmath15 .
we suppress the argument @xmath7 of @xmath18 and use @xmath19 to indicate the transpose .
we refer to @xmath20 as _ static amplitude distribution_. due to the correlations , a gaussian assumption for the static distribution is not as restrictive as it may seem . in the eigenbasis of @xmath15 ,
the amplitudes only appear in linear combinations .
thus , for large @xmath0 , the mechanisms that lead to the central limit theorem start working and drive the distributions towards gaussians . later on in sec .
[ sec5 ] we will nevertheless relax the gaussian assumption for the static amplitude distribution and look at more general functional forms .
how does the non stationarity affect the amplitude distribution when data from the total interval @xmath2 are analyzed ? as in ref .
@xcite , we model this by random matrices . as the covariance matrix is different at each time
@xmath7 where it is analyzed , we replace the covariance matrix in the distribution ( [ gaussian ] ) by the expression @xmath21 where @xmath22 is a real rectangular @xmath23 random matrix without any symmetries . the right hand side of eq .
( [ repl ] ) has to have the form given to ensure that it can model a properly defined covariance matrix .
this follows directly from the definition ( [ covar ] ) .
although @xmath0 , the first dimension is fixed , the second one , @xmath24 , is for the time being a free model parameter . it can be viewed as the length of the model time series .
further clarifications will follow . to obtain the amplitude distribution for the total interval , we average over the random matrices @xmath25 \overline{w}(a|\sigma , n ) g\left(r\left|\frac{1}{n}aa^\dagger\right.\right ) \
, \label{average}\ ] ] where @xmath26 $ ] is the volume element , _
i.e. _ , the product of all independent variables in @xmath22 .
following wishart @xcite , the gaussian distribution @xmath27 was assumed for the random matrices in ref .
it describes the gaussian fluctuations of the model covariance matrices @xmath28 about the given empirical covariance matrix @xmath29 , which is evaluated over the total time interval of length @xmath2 .
it should not be confused with the empirical covariance matrix @xmath13 calculated in the subintervals @xmath6 $ ] .
the crucial difference compared to ref .
@xcite is a generalization of the ensemble ( [ uswis ] ) .
we introduce the deformed wishart ensemble @xmath30 which is defined by the _
ensemble deformation function _
@xmath31 with the properties @xmath32 for later convenience , @xmath29 on the right hand side of eq .
( [ defwis ] ) is rescaled with @xmath24 .
the flutuations of the model covariance matrices @xmath28 deviate from gaussian , but always about the empirical covariance matrix @xmath29 .
the meaning of the model parameter @xmath24 now becomes clearer .
it sets the variance for these fluctuations .
the above rescaling only changes the functional dependencies , but not the rle of @xmath24 .
we emphasize once more that @xmath29 is evaluated over the total time interval .
similar deformations of random matrix ensembles but in a hamiltonian , not wishart setting were apparently first put forward in refs .
@xcite . after inserting the ansatz ( [ defwis ] ) into eq .
( [ average ] ) , we may use the result @xcite @xmath33 = \int\limits_0^\infty \chi^2_n(z ) g\left(r\left|\frac{z}{n}\sigma\right.\right ) dz \ , \label{gaver}\ ] ] which reformulates the whole random matrix average as a univariate average over the @xmath34 distribution @xmath35 of @xmath24 degrees of freedom . on the mathematical side , there are connections between formula ( [ gaver ] ) and the calculation of certain distributions in scattering theory @xcite . using the result ( [ gaver ] )
, the amplitude distribution reduces to the double integral @xmath36 again , we point out the rescaling of @xmath29 with @xmath24 , _ cf . _
( [ gaver ] ) .
it is useful to rewrite that as a single integral @xmath37 by introducing the variable @xmath38 and its distribution @xmath39 we refer to it as _ amplitude distribution deformation function_. one easily obtains @xmath40 which establishes the relation between the two deformation functions .
we notice that the ansatz ( [ defwis ] ) restricts the form of the deformed distribution @xmath41 to functions of @xmath42 only .
even though the inclusion of further terms such as @xmath43 is likely to improve the quality of the data fits , we stick to the ansatz ( [ defwis ] ) .
its considerable advantage is the guaranteed but otherwise questionable analytical tractability as will be shown in the sequel .
moreover , further terms will also increase the number of deformation functions which will hamper their unambiguous determination .
apart from the deformation functions , the distributions @xmath44 and @xmath45 depend on the usual covariance matrix @xmath29 analyzed by sampling over the total interval .
we notice that the corresponding covariance matrix @xmath46 in the deformed ensemble slightly differs from that . by definition
we have @xmath47 \ . \label{sigmad}\end{aligned}\ ] ] inserting eq .
( [ defwis ] ) , we can do the ensemble average in the gaussian case which yields the covariance matrix @xmath48 .
thus , only the @xmath49 integral remains and we have @xmath50 implying that the two covariance matrices differ by the average of @xmath51 .
alternatively , one can calculate @xmath46 from the amplitude distribution , @xmath52 \ , \label{sigmada}\ ] ] which yields @xmath53 here , the two covariance matrices differ by the first moment of @xmath54 . with the help of eq .
( [ aampldef ] ) , the results ( [ sigmadd],[sigmadad ] ) are easily seen to coincide .
having extracted the covariance matrix for the total time interval from the data , we can proceed with the determination of the deformation functions .
the exponential function in the integrand of eq .
( [ pdis ] ) allow us to interpret it as a laplace transform , @xmath55 where we introduced @xmath56 to avoid inconvenient factors of two .
thus , the ensemble deformation function is the inverse laplace transform @xmath57 of the amplitude distribution deformation function divided by a power of @xmath54 .
this makes it possible to determine @xmath31 by extracting @xmath58 from the amplitude time series and carrying out the inverse laplace transform .
in contrast , extracting @xmath31 directly from the data is cumbersome and burdened by limited statistics , as the following discussion shows .
the rows of @xmath22 are the model time series of length @xmath24 and can not easily be identified with the amplitude time series @xmath59 of lenght @xmath4 . however , the matrices @xmath28 form the ensemble of model covariance matrices and can be compared with the empirical ones . as a certain sample length
is required for meaningful results , it is out of question to compare the matrices directly , _
i.e. _ their individual matrix elements
. a better observable is the distribution @xmath60 \label{trace}\ ] ] of the traces , which can easily be written as a single integral involving the ensemble deformation function @xmath31 .
the distribution ( [ trace ] ) is empirically obtained by moving a time window through the amplitude time series and calculating the empirical covariance matrices and their traces .
this then gives @xmath31 .
the problem with the above procedure is its still limited statistical significance .
instead , extracting the amplitude distribution deformation function @xmath58 from the data gives much more meaningful results .
as we discuss in the sequel , the number of data points is larger by a factor of @xmath0 .
the amplitudes @xmath61 appear in eq .
( [ averagesingle ] ) only via the bilinear form @xmath62 .
we rotate the amplitude vector @xmath18 into the eigenbasis of the empirically obtained covariance matrix @xmath29 . by definition ,
the eigenvalues of @xmath29 are positive and larger than zero , provided the length of the sampling interval is larger than @xmath0 .
we divide each component of the rotated amplitude vector by the square root of the corresponding eigenvalue and denote the resulting vector by @xmath63 . within our model , all components of @xmath63 should be equally distributed .
we integrate out all but one , @xmath64 , and arrive at the distribution @xmath65 thus , @xmath58 may be identified with the distribution of the variances @xmath54 of the gaussian distributed random variables @xmath64 . conceptually , this is our main result .
it provides a simple and statistically significant method to obtain the amplitude distribution deformation function @xmath58 which then yields upon inverse laplace transform the ensemble distribution function @xmath31 .
as we have @xmath0 time series @xmath59 , we gain a factor of @xmath0 by aggregation .
we now apply our method to stock market data .
this is of particular interest as heavy tails are ubiquituous in finance .
a better modeling for multivariate distributions is urgently called for to improve risk estimation .
we present the data in sec .
[ sec31 ] , extract the deformation functions in sec .
[ sec32 ] and calculate the ensemble and return distributions in sec .
[ sec33 ] .
we analyze the @xmath66 continuously traded stocks with prices @xmath67 in the s&p 500^^ index between 1992 and 2012 @xcite , which we previously analyzed with a purely gaussian , _
i.e. _ , non deformed wishart ensemble @xcite .
the amplitudes are here the dimensionless price changes @xmath68 which are referred to as returns .
they depend on the chosen return horizon @xmath69 . according to eq .
( [ returns ] ) , we calculate the returns @xmath59 normalized to zero mean . to make our presentation self contained ,
we show once more how strongly the whole @xmath12 correlation matrix @xmath70 for this data set changes in time . in fig .
[ fig1 ] , it is displayed for subsequent three months time companies in s&p 500^^ for the fourth quarter of 2005 and the first quarter of 2006
. the darker , the stronger the correlation . the companies are sorted according to industrial sectors . taken from ref .
@xcite.,title="fig:",scaledwidth=23.5% ] companies in s&p 500^^ for the fourth quarter of 2005 and the first quarter of 2006 .
the darker , the stronger the correlation .
the companies are sorted according to industrial sectors .
taken from ref .
@xcite.,title="fig:",scaledwidth=23.5% ] windows . in most random matrix approaches ,
the ensemble is fictitious and enters only by means of an ergodicity argument .
this is not so here , as fig .
[ fig1 ] illustrates .
our ensemble exists in reality , it is the whole set of matrices analyzed by moving a sample time window through the data . in fig .
[ fig1 ] , one also sees rather stable stripes in these correlation matrices which are due to the different industrial sectors , see , _
e.g. _ , ref .
obviously , basis invariance is not present in this data set , and probably neither in any other real data set .
hence , a direct extraction of the ensemble deformation function @xmath31 which preserves the basis invariance of the random matrix ensemble is problematic . yet
, there is still another reason : market states were identified which reveal a fine structure of the ensemble @xcite . as every random matrix ensemble has an effective character , one is advised to analyze quantities which already reflect this . in the present case , such quantities are the amplitude , in the present case return , distribution and the corresponding deformation function @xmath58 .
we use daily data , _
i.e. _ , @xmath71 trading day .
rotation of the return vector @xmath18 into the eigenbasis of the empirical covariance matrix @xmath29 , normalization to the square roots of the eigenvalues and aggregation on a five day window yield the empirical distributions of variances shown in fig .
aggregation on a distribution as dashed dotted line.,scaledwidth=40.0% ] distribution as dashed dotted line.,scaledwidth=40.0% ] ten day window gives similar results ; hence , the estimation noise does not have a major impact on the distribution .
a variety of functions is capable of describing the data . in finance
, one often employs the log normal distributions to model volatilities , see _
e.g. _ ref . @xcite . in finance ,
the standard deviations are referred to as volatilities
. however , the log normal distribution fails to capture the empirically found tail behavior .
more suitable is the beta prime distribution @xmath72 with two positive parameters @xmath24 and @xmath73 .
anticipating the following discussion , we choose their combination in the expression ( [ betaprime ] ) in such a way that @xmath24 coincides with the parameter @xmath24 introduced in sec . [ sec2 ] .
the fit is depicted in fig .
[ fig2 ] , the agreement with the data is much better than for a @xmath34 distribution corresponding to the ensemble of ref .
@xcite which is formally obtained by setting @xmath74 or @xmath75 , respectively , depending on the rescaling with @xmath24 .
we carry out fits for different return horizons @xmath69 .
the results for @xmath24 and @xmath73 are shown in fig .
[ fig3 ] . while @xmath73 stays constant around two , @xmath24 increases ( upper , ascending dots ) and @xmath73 ( lower dots ) for the beta prime distribution versus the return horizon @xmath69 .
triangles and diamonds for the fits with constraint to integer values , suares and circles without constraint.,scaledwidth=40.0% ] from about seven for daily data to about 23 for @xmath76 trading days .
we postpone the interpretation up to the evaluation of the ensemble and return distribution .
having extracted the amplitude , _
i.e. _ , return distribution deformation function @xmath77 , we calculate the inverse laplace transform ( [ invlaplace ] ) , @xmath78 and find @xcite with @xmath79 for the ensemble deformation function @xmath80 this is a @xmath81 distribution with @xmath82 degrees of freedom . as required , @xmath83 is a positive and normalized function . inserting eq .
( [ betaprime2 ] ) into eq .
( [ defwis ] ) yields after a straightforward calculation @xmath84 for the distribution of the random matrices @xmath22 .
thus , we arrive at an ensemble characterized by an algebraic distribution . for a similar ensemble , but in the special case of @xmath85 , spectral correlation functions were studied in refs .
@xcite . here , however , we derived our ensemble from data , and the dependence on a non
trivial @xmath29 is in the present essential . anticipating the result ( [ wisal ] ) , we rescaled @xmath29 with @xmath24 as compared to ref .
thereby , @xmath24 and @xmath73 appear on equal footing in the formulae . to obtain the ensemble averaged return distribution , we plug eq .
( [ betaprime ] ) into eq .
( [ averagesingle ] ) and find @xmath86 with the confluent hypergeometric function @xcite @xmath87 for positive real parts of @xmath88 and @xmath89 . from eq .
( [ sigmadd ] ) or ( [ sigmadad ] ) the covariance matrix @xmath90 for the deformed ensemble follows . to compare with the empirical return distribution ,
we compute the integral ( [ rotscale ] ) , @xmath91 a comment on the permissible values for the parameter @xmath24 is in order . in the return distributions ( [ wisret ] ) and ( [ wisretmarginal ] )
, @xmath24 can take any positive real value . in the ensemble distribution ( [ wisal ] ) , however , @xmath24 is the length of the model time series or , equivalently , one of the dimensions of the @xmath23 matrices @xmath22 and thus restricted to integer values .
it is thus a matter of interpretation whether one wants to impose the constraint that @xmath24 be integer .
there is no such restriction for the parameter @xmath73 . in any case
, we also carried out fits with the integer constraint .
the results shown in fig .
[ fig3 ] do not indicate a strong influence of this constraint .
the results of the data comparison are displayed in fig .
[ fig4 ] for daily returns , @xmath71 trading day .
the fitted parameter values trading day .
empirical results as dots , fit to the distribution ( [ wisretmarginal ] ) as solid line .
the corresponding result of ref .
@xcite as dashed line .
center of the distribution on a linear scale ( top ) , whole distribution on a logarithmic scale ( bottom).,scaledwidth=40.0% ] trading day .
empirical results as dots , fit to the distribution ( [ wisretmarginal ] ) as solid line .
the corresponding result of ref .
@xcite as dashed line .
center of the distribution on a linear scale ( top ) , whole distribution on a logarithmic scale ( bottom).,scaledwidth=40.0% ] are @xmath92 and @xmath93 .
the center of the empirical distribution is slightly better described by employing the deformed wishart ensemble instead of the non deformed one in ref .
the heavy tails clearly reveal that the deformed wishart ensemble yields overall a better description , since the result of ref .
@xcite consistently underestimates the large events . in fig .
[ fig5 ] we present the same analysis for returns with @xmath94 trading days , the fit ( bottom ) for returns with @xmath94 trading days.,scaledwidth=40.0% ] gives @xmath95 and @xmath96 .
here , the tails are still strong , but less pronounced than for daily data . for the interpretation of these results ,
we recall the well established fact that _ univariate _ distributions of returns for one stock acquire heavy tails as the return horizon @xmath69 becomes smaller , see _
e.g. _ ref .
@xcite . here
, however , we analyze the _ multivariate _ distribution of @xmath0 correlated stocks .
thus , there are two competing effects .
first , as discussed in general in connection with eq .
( [ gaussian ] ) and for the financial data in ref .
@xcite , the superposition of the amplitudes , in the present case the returns , drives the multivariate distribution towards a gaussian , provided that the covariances are sufficiently constant .
second , as observed in ref .
@xcite and extended here , the fluctuations of the non stationary covariances lift the tails of the distributions evaluated over long time intervals and make them heavier .
not surprisingly , the heavier the tails of the _ univariate _ distributions , the heavier are also those of the ensemble averaged multivariate ones shown above .
this is nicely reflected in the nearly linear increase of the parameter @xmath24 on the return horizon @xmath69 , see fig .
the smaller @xmath24 , the heavier are the tails in the ensemble distributions ( [ wisal ] ) and in the enesmble averaged return distribution ( [ wisret ] ) . as @xmath24 grows and @xmath73 is held fixed , the distribution ( [ wisal ] ) comes closer to a gaussian , _
i.e. _ , to the non deformed ensemble .
it is quite remarkable that the fit of @xmath73 always yields values close to two . according to eq .
( [ covard ] ) , this implies @xmath97 .
put differently , the heavy tails in the cases considered alter the measured covariances only slightly as compared to a gaussian assumption .
when looking at financial risk , however , the tails are very important .
finally , we use the opportunity to discuss an issue of general interest when presenting and fitting a multivariate distribution that depends on the statistical variables only via a bilinear form such as @xmath98 . instead of the above procedure which involves rotation of @xmath18 into the eigenbasis of @xmath29 and aggregation , one might also view the bilinear form as a generalized radius @xmath99 in the @xmath0 dimensional space and study its distribution @xmath100 \ . \label{rho2}\ ] ] a rather straightforward calculation yields @xmath101 the jacobian @xmath102 , typical for such a radial distribution , appears and , because of @xmath66 , dominates the functional form of the distribution for small @xmath103 , as can be seen in fig .
[ fig6 ] . trading day .
empirical results as dots , fit to the distribution ( [ disrad ] ) as solid line .
the corresponding result of ref .
@xcite as dashed line , on a linear ( top ) and a logarithmic scale ( bottom).,scaledwidth=40.0% ] trading day .
empirical results as dots , fit to the distribution ( [ disrad ] ) as solid line .
the corresponding result of ref .
@xcite as dashed line , on a linear ( top ) and a logarithmic scale ( bottom).,scaledwidth=40.0% ] the theoretical result ( [ disrad ] ) describes the data much better than the corresponding result of ref .
nevertheless , the dominance of @xmath102 gives a somewhat misleading picture and we infer that using the distributions ( [ wisret ] ) and ( [ wisretmarginal ] ) is more appropriate .
when extracting the return distribution deformation function @xmath58 from the data , we encountered a puzzling problem that we wish to report here .
the log logistic distribution @xmath104 with @xmath105 yields a very good description of the data , even slightly better than the beta prime distribution .
the @xmath106 values are around one , @xmath107 for @xmath71 trading day and increasing for larger @xmath69
. however , the resulting ensemble deformation function can take positive and negative values .
for example , for @xmath107 , we find @xmath108 hence , it can not be interpreted as a distribution . indeed
, we are confronted with a problem of interpretation . while the discussion in sec .
[ sec23 ] clearly revealed that @xmath58 is a well defined distribution of a random variable , namely of a variance , it is not obvious that @xmath31 also represents a well defined distribution .
the corresponding random variables , the matrices @xmath22 , do not have a direct data interpretation .
thus , one might simply view @xmath31 as a continuous coefficient function for the expansion of the distribution ( [ defwis ] ) in terms of gaussians
. nevertheless , even if one does not want to enforce an interpretation of @xmath31 as a distribution , the resulting ensemble distribution ( [ defwis ] ) must be positive definite .
we test this by calculating the distribution of the traces @xmath109 \ . \label{tracesig}\ ] ] as we only wish to test the positivity , it is convenient to choose it different from the above distribution ( [ trace ] ) by including the covariance matrix . after some algebra
, we can express it as a high order derivative involving the return distribution deformation function , @xmath110 this in principle general result gives for the log logistic distribution ( [ logp ] ) @xmath111 restricting ourselves to even @xmath24 , we may employ the theory of complex functions to calculate the pole expansion @xmath112 with the poles @xmath113 the derivatives in eq .
( [ u2 ] ) can now easily be evaluated and we arrive at @xmath114 inspite of the complex poles , this is by construction a real function . yet , it takes positive and negative values which outrules an interpretation of @xmath115 and thus also of @xmath116 as distributions . by means of this example we face the somewhat surprising result that a well defined distribution @xmath58 does not necessarily yield a well defined ensemble .
each case has to be investigated individually .
we argued in sec .
[ sec21 ] that the gaussian assumption ( [ gaussian ] ) for the static amplitude distribution is not as restrictive as it might appear at first sight .
nevertheless , we now extend our construction by assuming more general functional forms . at present
, we do not have data at our disposal in which the static amplitude distribution is non gaussian , but we nevertheless now extend our construction , as it might be useful for future data analyses .
moreover , we will also come across some interesting observations . instead of eq .
( [ gaussian ] ) , we now assume that the static amplitude distribution can be expressed as an average over the gaussian ( [ gaussian ] ) , @xmath117 with a new deformation function @xmath118 that fulfils @xmath119 we proceed as in sec .
[ sec22 ] . instead of the ensemble average ( [ average ] ) , we now have @xmath120 \overline{w}(a|\sigma , n ) \overline{g}\left(r\left|\frac{1}{n}aa^\dagger\right.\right ) \ . \label{averageg}\ ] ] this is readily cast into the form @xmath121 which differs from eq .
( [ averagesingle ] ) only by the definition of the amplitude distribution deformation function .
it is now given by @xmath122 for fixed @xmath123 , we introduce the new variable @xmath124 and find @xmath125 which coincides with eq .
( [ aampldef ] ) , but now involving the new ensemble deformation function @xmath126 this integral is reminiscent of a convolution .
thus , the case of a deformed , non gaussian static amplitude distribution is formally traced back to the gaussian case .
the difference can be fully absorbed into the ensemble deformation function .
importantly , this means that all other results of sec .
[ sec2 ] continue to hold , in particular the laplace transform ( [ laplace ] ) and its inversion ( [ invlaplace ] ) .
nevertheless , the following problem remains .
we can extract @xmath118 and @xmath58 from the data by using the methods outlined in sec .
[ sec23 ] for very short time intervals and for the whole , long time interval , respectively . from the inverse laplace
transform ( [ invlaplace ] ) , we obtain @xmath127 , but to determine @xmath31 , we are left with the task to invert eq .
( [ aampldefg3 ] ) .
although that is definitely possible for some special cases , a general inversion formula is lacking . in practical applications , however , the extension sketched above is more likely to be needed for consistency tests .
for example , if some of the available data for the same system permit the gaussian assumption for the static amplitude distribution and others do not , one can first determine @xmath31 as described in sec .
[ sec2 ] and then turn to the data which require an additional deformation function @xmath118 . once both of these deformation functions are known , one can evaluate @xmath127 and check if it is consistent with the inverse ( [ invlaplace ] ) of @xmath58 which is independently extracted from the data .
as an example , we consider the case that both , @xmath31 and @xmath118 , are @xmath34 distributions @xmath128 of @xmath82 and @xmath129 degrees of freedom , respectively . the choice for @xmath31 coincides with the result of sec . [ sec32 ] . with eq
( [ aampldefg3 ] ) , we obtain @xmath130 where @xmath131 is the modified bessel function of the second kind of order @xmath132 .
this function already appeared in the ensemble averaged return distribution of ref .
@xcite . according to eq .
( [ aampldefg2 ] ) , the distribution @xmath58 is an integral involving the modified bessel function and the return distribution averaged over the deformed ensemble is an integral over a product of modified bessel functions , but we do not give the formulae here .
non stationarity is an often encountered feature in complex systems . here , we addressed non stationarity of correlations .
we presented a method to determine their distribution from the amplitude distribution .
put differently , we showed how to extract the proper ensemble of random covariance matrices from amplitude data . carrying out our analysis for the case of financial data
, we found an algebraic distribution of covariance matrices reflecting the heavy tails in the amplitude , _
i.e. _ , return distributions .
a conceptually important comment is in order .
consider any two empirical covariance matrices @xmath133 and @xmath134 .
they are certainly not independent , because , first , the correlation structure due to , _
e.g. _ , the industrial sectors in the financial markets only changes on a large time scale and , second , one would expect that they are the more dependent the smaller the time difference @xmath135 .
the first cause is a built in feature of our model , as the random model covariance matrices fluctuate around the empirical @xmath29 .
the second cause is effectively accounted for as the length of our model time series @xmath24 , _
i.e. _ , the second dimension of the random matrices @xmath22 , is different from the length @xmath4 of the subintervals in which the empirical covariance matrices @xmath13 are evaluated .
the values for @xmath24 resulting from the data analysis are very small @xmath136 while @xmath137 when measuring @xmath13 .
our random matrix ansatz does not aim at modeling the ensemble of the empirical covariance matrices @xmath13 in a one to one fashion .
this is never the goal of a statistical approach .
our model has an effective character which is obvious , for example , in the relation @xmath138 .
model time series much shorter than the empirical ones suffice to properly grasp the statistical effects induced by the fluctuations around the average covariance matrix .
this also reflects the mutual dependence of the empirical covariance matrices .
we demonstrated by our analysis of financial data how useful our model is .
furthermore , observables such as the amplitude distribution do not resolve autocorrelations in time of any kind . even reshuffeling
the order of the empirical covariance matrices @xmath13 in time does not change the amplitude distribution for the total time interval .
a similar situation is encountered when analyzing statistical properties of quantum chromodynamics .
the gauge fields may be viewed as a really existing ensemble over which an average is carried out this is the very definition of the partition function .
although these gauge fields are not independent either , an effect of this autocorrelation is only seen if corresponding observables are used .
densities are not affected .
random matrices as models for the highly non trivial gauge fields are applied with great success .
yet another aspect is worth mentioning .
in contrast to random matrix applications for hamiltonian systems , there is not a local scale that can enforce universal statistical behavior .
thus , it is important how the covariances are actually distributed .
gaussian assumptions are only acceptable if really justified by data analysis .
the present study extends a previous one in which we employed such a gaussian assumption in finance . here
, we reconsidered the same data set and clearly demonstrated that the gaussian assumption underestimates the tails .
the algebraic distributions that we found here are relevant for risk estimation as they help to better understand large events .
importantly , once the ensemble is properly extracted , meaningful averages can be computed for all observables that depend on non stationary covariances .
when developing our construction , we came across a puzzling feature which calls for a caveat .
the deformation function extracted from the amplitude distribution determines , on the one hand , uniquely the ensemble , but , on the other hand , this ensemble is not necessarily well defined .
each case has to be studied individually .
we do not expect this to cause severe problems in applications , but conceptually it is an interesting aspect .
we also extended our construction by including deformed static amplitude distributions . the additional freedom accompanying
this extension might offer a possibility to circumvent the above mentioned puzzling problem . from a more general viewpoint
, we have to emphasize that our construction only includes functional forms of the ensemble that depend on the trace over the product of the random covariance matrix and the mean covariance matrix .
although this is quite natural , as it guarantees a certain amount of basis invariance which all random matrix models need , more general functional forms pose an interesting and potentially important challenge .
hitherto , we only applied our method to finance .
we plan applications to other complex systems , too .
this may be rewarding , as large events and risk estimations are not only important in finance .
we thank desislava chetalova , tobias nitschke , thilo schmitt and yurij stepanov for fruitful discussions . | in complex systems , crucial parameters are often subject to unpredictable changes in time .
climate , biological evolution and networks provide numerous examples for such non stationarities . in many cases , improved statistical models are urgently called for . in a general setting , we study systems of correlated quantities to which we refer as amplitudes .
we are interested in the case of non stationarity , _
i.e. _ , seemingly random covariances .
we present a general method to derive the distribution of the covariances from the distribution of the amplitudes . to ensure analytical tractability
, we construct a properly deformed wishart ensemble of random matrices .
we apply our method to financial returns where the wealth of data allows us to carry out statistically significant tests .
the ensemble that we find is characterized by an algebraic distribution which improves the understanding of large events . |
since many years considerable effort has been made to investigate the problem of nucleation and growth of a new phase upon martensitic transformation . despite these attempts , there is no clear understanding so far of the atomic scale mechanism of the bcc - hcp reconstructive transformation occurring even in non - magnetic materials . in the case of
fe - considered here - there are convincing arguments , both from theoretical and experimental side , that magnetism plays a crucial role for the stability of bcc structure , making the problem more complicated @xcite .
the bcc - hcp transformation observed in non - magnetic materials has been extensively discussed in the literature @xcite .
the corresponding mechanism suggested by burgers @xcite consists in two types of simultaneous distortions ( see appendix , fig .
[ fig_burgers ] ):
( i ) opposite displacement of adjacent ( 110 ) planes along the @xmath3_{bcc}$ ] direction , described by the parameter @xmath4 , associated with the transverse @xmath0-point @xmath1 phonon mode with @xmath2 $ ] polarization ; ( ii ) shear deformation along the [ 001 ] direction , characterized by the angle @xmath5 between the diagonals in ( 110 ) plane , which should change from @xmath6 in the case of bcc structure to @xmath7 in the case of hcp structure .
the shear modulus is determined by the slope of the @xmath1 phonon branch @xmath8 $ ] with @xmath9 $ ] polarization .
note that even for non - magnetic materials a phenomenological description of this type of martensitic transformation is not straightforward : as pointed out in the literature @xcite , the transition is discontinuous , having large critical displacements and no group - subgroup relationship between the symmetries of the initial and final phases .
this causes problems for a landau free energy expansion with respect to an order parameter . to deal with first - order transitions
the phenomenological landau theory was extended @xcite using two order parameters representing shuffle and shear deformations and have been applied rather successfully to the bcc - hcp phase transitions in ti @xcite and zr @xcite , associated with the softening of the @xmath0 point @xmath1 phonon mode . according to these theoretical findings , already small phonon softening ( as it takes place in the case of ti and zr )
can be sufficient for a first order transformation @xcite .
the theoretical approach used for non - magnetic systems has been applied to fe , showing the important role of magnetism .
in contrast to the non - magnetic bcc metals mentioned above , no softening of the @xmath0 point phonon modes have been observed under pressure neither experimentally up to @xmath10 gpa @xcite nor theoretically @xcite .
however , some dft based theoretical investigations report about the key role of the shear stress in fe under pressure for the bcc - hcp transition @xcite . their role has been investigated by sanati et al .
@xcite in application to ti and zr showing that the bcc structure in these materials is completely stable with respect to shear deformation and only the @xmath0 point phonon mode is responsible for bcc - hcp transformation .
ab - initio investigations by ekman et al .
@xcite clearly showed that the stability of the bcc phase of fe is due to magnetic ordering .
they have shown that the burgers type of lattice distortion results in the transition to the paramagnetic state at a certain amplitude of the atomic displacements and that way to the instability of the bcc structure .
liu and johnson @xcite have analyzed the potential - energy surface and minimum - energy pathway obtained within ab - initio calculations and have also found that the magnetization collapse during the shuffle - shear ( burgers type ) deformation leads to the instability of the bcc phase of fe under pressure . to investigate experimentally the role of magnetism in the bcc - hcp transformation , mathon et al .
@xcite have performed measurements of near edge x - ray absorption ( xanes ) including a determination of the x - ray magnetic circular dichroism ( xmcd ) for fe under pressure .
the high sensitivity of xmcd allows very precise measurements of the ordered magnetic moment on the absorber at the magnetic phase transition and to observe its correlation with the local geometrical structure monitored via xanes .
the pressure dependence of the xmcd and xanes spectra around the transition pressure suggests that the magnetic transition slightly precedes the structural one .
this finding allowed the authors to ascribe a leading role in the bcc - hcp transformation to the magnetic order in the system .
nevertheless , this interrelation requires further clarification , because , in general , the vanishing of magnetism should lead immediately to the instability of the bcc state and therefore , no difference in the transition pressure deduced from xanes and xmcd spectra should be expected .
although these results give information on the origin of the pressure induced instability of bcc fe and showing the minimal - energy pathway for the transition , there is still the question how the instability condition which needs a certain phonon softening under pressure occurs while no softening could be found .
one possible way has been suggested by vul and harmon @xcite in their fluctuation - less mechanism for martensitic transformations triggered by the defects presented in the crystal . in the present work
we analyze the conditions which can result in the instability in perfect bcc fe via lattice fluctuations , in particular the effect of temperature induced spin fluctuations for the bcc - hcp transition .
these investigations in particular give an answer to the question , why the decrease of the xmcd signal slightly precedes that of the xanes signal corresponding to bcc structure upon a pressure increase .
spin - polarized electronic structure calculations have been performed using the spin - polarized kkr ( korringa - kohn - rostoker ) green s function method @xcite in the fully relativistic approach .
the generalized gradient approximation ( gga ) for density functional theory was used with the exchange - correlation potential due to perdew , burke , ernzerhof ( pbe ) @xcite .
the potential was treated within the full potential ( fp ) scheme . for
the angular momentum expansion of the green s function a cutoff of @xmath11 was applied . to treat spin disorder in the system , the self - consistent coherent potential approximation ( cpa )
method was employed . for the calculation of the x - ray absorption coefficient @xmath12
the following expression has been used @xmath13 where @xmath14 , @xmath15 , and @xmath16 stand for the wave vector , frequency , and polarization of the radiation , and @xmath17 is the electron - photon interaction operator @xcite .
the finite temperature magnetic properties have been investigated via monte carlo ( mc ) simulations based on the classical heisenberg model , using a standard metropolis algorithm @xcite .
the exchange coupling parameters @xmath18 for these calculations are obtained within the approach described by lichtenstein @xcite . fig .
[ fig_etot ] represents the results of total energy calculations for bcc fe in ferromagnetic ( fm ) and non - magnetic ( nm ) states as well as for hcp structure in the nm state .
( below we will use the notation nm and pm to denote the non - magnetic state with with zero magnetic moments and the paramagnetic state implying a random distribution of the localized magnetic moments , respectively . ) the calculations for the hcp structure have been performed at a fixed @xmath19 ratio equal to @xmath20 . according to the results shown in fig .
[ fig_etot ] , the minimum of the total energy corresponds to the equilibrium lattice parameter @xmath21 a.u . in the case of the fm state of bcc - fe , and in the case of hcp - fe - to @xmath22 a.u .
the critical pressure evaluated from the equivalence of the enthalpy of the bcc and hcp phases is equal to 8 gpa in reasonable agreement with experiment .
( a ) total energy as a function of volume for ferromagnetic ( fm ) and non - magnetic ( nm ) states of bcc - fe as well as for nm state of hcp - fe ; ( b ) pressure as a function of volume for bcc - fe.,title="fig:",scaledwidth=33.0%](a ) ( a ) total energy as a function of volume for ferromagnetic ( fm ) and non - magnetic ( nm ) states of bcc - fe as well as for nm state of hcp - fe ; ( b ) pressure as a function of volume for bcc - fe.,title="fig:",scaledwidth=33.0%](b ) the fm to nm transition leading to the instability of the bcc structure occurs at the lattice parameter @xmath23 a.u .
( @xmath24 ( a.u.)@xmath25 , fig .
[ fig_etot ] ) that corresponds to a pressure of about 200 gpa .
this value is much too high when compared directly to experiment .
therefore this type of instability does not seem to play a role for the real system . the energy of burgers type lattice distortion in bcc fe have been calculated first assuming fm order in the system .
in contrast to the works of ekman @xcite and vul and harmon @xcite , where the authors used two parameters @xmath5 and @xmath4 to describe the shear deformation within the ( 110 ) plane and displacements of the neighbor ( 110 ) planes with respect to each other ( see introduction ) , in the present investigations we consider the path from bcc to hcp transformation as suggested by friak and sob @xcite , which includes both deformations using only one parameter @xmath26 ( see appendix ) .
this approach avoids the high - energy configurations occurring upon independent variation of the parameters @xmath4 and @xmath5 , accounting only those being close to the minimum total energy path @xcite .
the parameter @xmath27 corresponds to the bcc structure , while @xmath28 corresponds to the hcp structure with @xmath29 .
as soon as additional calculations ( not presented here ) exhibit only a weak dependence on the @xmath19 ratio for the position of the minimum of total energy as a function of lattice parameter , the following discussion concerns the results for @xmath29 for the sake of convenience .
[ fig_e_path ] represents the total energy as a function of the parameter @xmath26 , @xmath30 , for different lattice parameters @xmath31 corresponding to different pressure values .
the @xmath30 curves have two minima corresponding to the fm bcc structure ( @xmath32 ) and nm hcp structure ( @xmath33 ) , which is a quasi - equilibrium state . at low pressure
the fm bcc structure of fe is more stable , while as pressure increases , the energy of the nm hcp structure becomes deeper leading to the stability of this state .
note that the @xmath30 curves ( see fig .
[ fig_e_path ] ) calculated for three different pressure values ( or , equivalently , lattice parameters @xmath31 equal to @xmath34 , @xmath35 , and @xmath36 a.u . ) for the fm state of fe , have nearly the same dispersion .
this is in line with previous phonon calculations exhibiting no softening of the @xmath0-point phonon modes under pressure as discussed above . as follows from fig .
[ fig_etot ] , the nm state of bcc fe is higher in energy than the fm state .
@xmath37 increases while @xmath38 decreases monotonously with @xmath26 varying from 0 to 1 ( open symbols in fig .
[ fig_e_path ] ) , indicating the instability of nm bcc fe with respect to this type of distortions , in line with the results on phonon calculations @xcite .
the critical distortion values , @xmath39 , correspond to the cross - points of the total energy curves as functions of the lattice distortion @xmath26 , calculated for the fm ( @xmath37 ) and nm ( @xmath38 ) states .
i.e. , at the critical values of distortions @xmath40 .
the magnitude of the energy cusps in fig .
[ fig_e_path ] , defined by @xmath41 for the various lattice parameters @xmath42 , @xmath43 , @xmath44 , @xmath45 and corresponding amplitude of the critical distortions ( @xmath46 , @xmath47 , and @xmath48 ) are getting smaller when the pressure increases . energy as a function of the parameter @xmath26 .
full symbols correspond to the fm state , open symbols - to the nm state .
the vertical arrow indicates the energy difference @xmath49,scaledwidth=45.0% ] total dos per atom as a functions of the parameter @xmath26 for pauli paramagnetic ( non - magnetic ) ( a ) and paramagnetic ( via dlm calculations ) ( b ) states of bcc fe .
dashed ( dashed - dotted ) lines represent spin - up ( spin - down ) states.,title="fig:",scaledwidth=37.0% ] ( a ) total dos per atom as a functions of the parameter @xmath26 for pauli paramagnetic ( non - magnetic ) ( a ) and paramagnetic ( via dlm calculations ) ( b ) states of bcc fe . dashed ( dashed - dotted ) lines represent spin - up ( spin - down ) states.,title="fig:",scaledwidth=37.0% ] ( b ) fig .
[ fig_dos](a ) shows the dos , @xmath50 , corresponding to the nm state of distorted bcc fe . for small distortions
, @xmath50 has a maximum at the fermi level @xmath51 indicating the instability of the nm state . this dos maximum
is formed by the double - degenerated @xmath52 electronic energy bands
. therefore there are two scenarios to remove the instability by breaking the symmetry of the system : either due to structure distortion ( e.g. , burgers distortion ) leading to the splitting of doubly degenerated @xmath52 states at @xmath51 or due to spontaneous magnetization leading to the exchange splitting of the electronic states having opposite spin directions . at ambient pressure
the second scenario is more favorable leading to the stabilization of the bcc structure with fm order . under high pressure , however , due to a reduced tendency towards spontaneous magnetization via the stoner mechanism the first scenario ( structure distortion ) could be more preferable , as discussed by ekman @xcite .
the same effect can be achieved even at lower pressure , when the amplitude of the distortion is big enough , as demonstrated in fig .
[ fig_e_path ] by the critical points @xmath53 and @xmath54 .
however , in this case the energy barrier @xmath55 is much too high to be overcome assuming the fluctuation mechanism of phase transition , even at the pressure @xmath56 gpa ( @xmath57 a.u . ) , when the hcp structure becomes energetically more preferable .
so far we have discussed the stability of the system with respect to lattice distortions assuming no deviations from perfect fm order . to discuss in more details the interrelation between the structural and magnetic degrees of freedom , the effect of spin moment fluctuations
should also be investigated . moreover ,
such investigations are required to account for the conditions of the experimental measurements performed at room temperature . to take the effect of temperature induced spin fluctuations at @xmath58 into account , the paramagnetic state with non - zero local moments of fe was simulated using the so - called disordered local moment ( dlm ) scheme describing a random distribution of the local magnetic moments over the lattice sites .
accordingly , discussing the results of calculations for the pm state we will use the term dlm state implying the approximation used in these calculations .
the dlm calculations are done by means of the cpa alloy theory applied to the effective alloy , fe@xmath59fe@xmath60 , with equal amount of alloy component ( @xmath61 ) having the spin magnetic moment along the @xmath62 direction ( fe@xmath63 ) and in the opposite direction ( fe@xmath64 ) @xcite .
the @xmath65 dependence on @xmath26 for the pm state obtained ( via dlm ) at the pressure corresponding to @xmath66 a.u .
is shown in fig .
[ fig_etot_dlm](a ) by solid circles . for @xmath67 ( i.e. ideal bcc structure ) the difference
@xmath68 is about two times smaller than @xmath69 .
@xmath70 decreases slowly with @xmath26 varying from 0 to 1 , indicating an instability of the dlm state of bcc fe at this pressure value .
note that the @xmath30 dependence for dlm state , @xmath70 , is weaker than that corresponding to the nm state , @xmath38 .
this behavior can be understood using the dos plots shown in fig .
[ fig_dos ] . in the case of dlm state
, one can clearly see a local dos minimum ( fig .
[ fig_dos](b ) ) at the fermi energy , created by the exchange - split majority and minority @xmath71-states of fe .
the dependence on @xmath26 of the dos obtained for dlm state is rather weak , in contrast to the dos obtained for the nm state ( fig .
[ fig_dos](a ) ) , having a pronounced maximum at @xmath51 strongly modified due to burgers type of distortions . however , instability of the dlm state of bcc fe with respect to the lattice distortion @xmath26 occurs only at the pressure exceeding a certain critical value ( closed circles and squares in fig . [ fig_etot_dlm](b ) ) . in the vicinity of ambient pressure
, the @xmath70 curve has a minimum at @xmath72 ( closed triangles in fig . [
fig_etot_dlm](b ) ) , that means stability of this state with respect to burgers distortions . to simulate the magnetic disorder corresponding to the temperature below the critical one , @xmath73 , so - called non - compensated dlm ( ndlm ) calculations
have been performed with the ndlm state simulated by an effective alloy fe@xmath74fe@xmath75 with @xmath76 $ ] . in this case
the normalized magnetic moment @xmath77 at each lattice site is equal to @xmath78 , assuming @xmath79 to be a saturated local spin magnetic moment of the fe atoms .
open circles in fig .
[ fig_etot_dlm](a ) correspond to the ndlm state of bcc fe ( @xmath80 a.u . ) with @xmath81 . in this case
the curve @xmath30 has a minimum at @xmath67 ( open circles in fig . [
fig_etot_dlm](b ) ) and exhibits a slow increase with @xmath26 increasing up to the critical displacement @xmath82 .
the total energy @xmath83 decreases further , when @xmath84 changes up to @xmath85 , that is associated with the temperature decrease and increase of fm order . as an example , open squares in fig .
[ fig_etot_dlm](a ) represent the @xmath86 dependence for @xmath87 ( i.e. , @xmath88 ) .
thus , the energy of burgers type lattice distortions , @xmath89 , decreases in the presence of temperature induced magnetic disorder and close to the pm state ( dlm state ) becomes much smaller than in the case of perfect fm order .
as a consequence , this leads to an increase of atomic displacements , caused directly by magnetic disorder in the system .
additional calculations have been performed to check explicitly the influence of magnetic disorder on the energy of the @xmath0 point @xmath1 phonon mode ( opposite displacement of adjacent ( 110 ) planes along the @xmath3_{bcc}$ ] direction ) , i.e. , without accounting for shear deformation along the [ 001 ] direction present in the case of the burgers type deformation .
the results are shown in fig . [ fig_etot_dlm](c ) .
the energy dependence @xmath90 on the atomic displacements @xmath4 is rather similar to that obtained for the burgers deformations , shown in fig .
[ fig_etot_dlm](a ) .
thus , these results show an explicit evidence of the crucial role of magnetic disorder for the softening of the @xmath0 point @xmath1 phonon mode making it responsible as the driving mechanism for the bcc to hcp transition .
total energy as a function of the lattice distortion parameter @xmath26 for bcc fe with @xmath66 a.u . : ( a ) comparison of the results for fm ( triangles ) , pm ( dlm , closed circles ) , partially disordered fm ( ndlm ) with non - compensated magnetic moment @xmath81 ( opened circles ) and @xmath87 ( opened squares ) ; ( b ) comparison of the dlm and ndlm ( @xmath81 ) results for different lattice parameters ; ( c ) total energy as a function of the amplitude of the @xmath0 point ta@xmath91 phonon mode for fm state ( closed triangles ) , nm state ( opened triangles ) and dlm state ( closed circles ) .
, title="fig:",scaledwidth=41.0% ] ( a ) total energy as a function of the lattice distortion parameter @xmath26 for bcc fe with @xmath66 a.u . : ( a ) comparison of the results for fm ( triangles ) , pm ( dlm , closed circles ) , partially disordered fm ( ndlm ) with non - compensated magnetic moment @xmath81 ( opened circles ) and @xmath87 ( opened squares ) ; ( b ) comparison of the dlm and ndlm ( @xmath81 ) results for different lattice parameters ; ( c ) total energy as a function of the amplitude of the @xmath0 point ta@xmath91 phonon mode for fm state ( closed triangles ) , nm state ( opened triangles ) and dlm state ( closed circles ) . , title="fig:",scaledwidth=40.0% ] ( b ) total energy as a function of the lattice distortion parameter @xmath26 for bcc fe with @xmath66 a.u . : ( a ) comparison of the results for fm ( triangles ) , pm ( dlm , closed circles ) , partially disordered fm ( ndlm ) with non - compensated magnetic moment @xmath81 ( opened circles ) and @xmath87 ( opened squares ) ; ( b ) comparison of the dlm and ndlm ( @xmath81 ) results for different lattice parameters ; ( c ) total energy as a function of the amplitude of the @xmath0 point ta@xmath91 phonon mode for fm state ( closed triangles ) , nm state ( opened triangles ) and dlm state ( closed circles ) .
, title="fig:",scaledwidth=40.0% ] ( c ) additional investigations have been performed to show an effect of lattice displacements on the magnetic order in the system . considering atomic fluctuations corresponding to the burgers type of lattice deformation @xcite , two series of the exchange coupling parameters , @xmath18 ,
have been calculated for the dlm state : ( 1 ) for different pressure values ( i.e. different lattice parameters ) for the perfect bcc structure and ( 2 ) for the distorted bcc lattice with @xmath66 a.u . with different distortion parameter @xmath92 .
[ fig_jxc_tvib ] shows the results for the case ( 1 ) , for which @xmath18 values are presented together with corresponding curie temperatures , determined by means of mc simulation , and exhibiting their decrease when the pressure increases .
the corresponding exchange coupling parameters @xmath18 for distorted bcc fe ( case ( 2 ) ) are shown in fig .
[ fig_jij_delta ] where one can easily see that lattice burgers distortions are accompanied by pronounced variations of the exchange coupling parameters .
in particular , we can point out a decrease of the fm and increase of the afm exchange interactions when the @xmath26 parameter increases .
assuming that the biggest amplitude of atomic displacements is related to this type of distortions ( due to their softening ) , the corresponding critical temperature have been calculated using the simulations for different values for the @xmath26 parameter ( fig . [ fig_mc](a ) ) .
this simplified approach was used to demonstrate the effect of atomic displacements on the finite - temperature magnetic properties of fe under fixed pressure ( corresponding to @xmath66 a.u . ) . at small distortions
the system has fm order with the curie temperature decreasing upon @xmath26 increase . at @xmath93 the system exhibit afm properties ( non - collinear or collinear , depending on @xmath26 ) .
the average magnetic moment calculated for different @xmath26 values at the temperature @xmath94 k is presented in fig .
[ fig_mc](b ) .
it drops down at rather small lattice distortions , @xmath95 , first , due to transition to the state with non - collinear magnetic structure and then , for @xmath96 , to the pm state . in spite the simplifications used for our analysis
, the results presented above directly demonstrate the strong mutual influence of the lattice burgers distortion and spin moment fluctuations resulting in a pronounced pressure and temperature dependence of the geometric and magnetic structure in the system .
both effects are the counterparts of the mechanism leading to a softening of the corresponding phonon modes responsible for the instability of the bcc state of fe and leading to the bcc - hcp transition .
( a ) exchange coupling parameters for bcc fe at different lattice parameter via dlm calculations for case ( 1 ) ( see text ) . , scaledwidth=35.0% ] dependence of the exchange coupling parameters on the amplitude of lattice distortion for bcc fe under pressure corresponding to lattice parameter @xmath66 a.u . for case ( 2 ) ( see text),scaledwidth=45.0% ] results of mc simulations : ( a ) critical temperature as a function of lattice distortion @xmath26 for bcc fe with @xmath66 a.u . ;
( b ) the normalized average magnetic moment corresponding to the temperature @xmath94 k.,title="fig:",scaledwidth=35.0% ] ( a ) results of mc simulations : ( a ) critical temperature as a function of lattice distortion @xmath26 for bcc fe with @xmath66 a.u . ; ( b ) the normalized average magnetic moment corresponding to the temperature @xmath94 k.,title="fig:",scaledwidth=35.0% ] ( b ) experimental investigations on bcc fe under pressure have been performed using x - ray absorption spectroscopy at the k - edge of fe , with the xanes and xmcd spectra measured simultaneously @xcite .
this allowed to check experimentally the role of magnetic order for the stability of bcc - fe , as discussed in the literature ( see , e.g. @xcite ) .
in particular , a synchronous decrease of the structural and magnetic xas signals related to the ferromagnetic bcc phase of fe would be expected at the critical pressure , where this loss of stability is associated with a transition to the pauli paramagnetic ( i.e. , nm ) state .
k - edge xanes for bcc fe at different pressures as well as hcp fe at a pressure above the critical one : ( a ) xanes and ( b ) xmcd spectra .
solid and dashed lines correspond to theoretical and experimental results , respectively.,title="fig:",scaledwidth=40.0% ] ( a ) k - edge xanes for bcc fe at different pressures as well as hcp fe at a pressure above the critical one : ( a ) xanes and ( b ) xmcd spectra .
solid and dashed lines correspond to theoretical and experimental results , respectively.,title="fig:",scaledwidth=40.0% ] ( b ) theoretical calculations of the x - ray absorption and xmcd spectra at the k - edge of fe have been performed for different lattice parameters corresponding to the pressure values below the critical one .
these results are compared in fig .
[ fig_xas ] with experimental spectra measured at different pressures below the critical value . in this case
the spectra are associated to the bcc structure with a slow variation upon pressure increase caused by the pressure induced variation of the lattice parameter . as one can see , the theoretical calculations reproduce the experimental xanes results quite well .
the same applies for the xmcd spectra . at the pressure above the critical value ,
the xanes spectrum reflects the hcp structure and is again in good agreement with experiment ( see fig .
[ fig_xas ] ) . in this case
the xmcd signal is very weak and corresponds to the remnant bcc phase at this pressure , which disappears completely if the pressure in further increased . to explain the pressure dependence of the experimental xmcd spectra
, we refer to the theoretical results discussed above .
the xanes experiment performed at the @xmath97 edge of fe implies that the induced orbital polarization of the @xmath98-electrons is probed . in ordered fm systems
this implies that the k - edge xmcd signal should be roughly proportional to the spin magnetic moment of the 3d electrons ( see , e.g. , @xcite and references therein ) . at finite temperature , however , a decrease of the dichroic signal can occur due to an increase of the magnetic disorder in the system , even for weak changes of the local magnetic moment . in the present case of bcc fe at fixed ( room ) temperature ,
the increase of the magnetic disorder is governed by the increasing pressure . at the same time , the xanes signal does not change because the bcc structure remains stable until the critical value of the magnetic disorder ( associated with a critical pressure ) is reached .
this is indeed observed in the experimental xmcd spectra
. a further pressure increase should result in a quick drop down for the average magnetic moment leading to the instability of the bcc structure and to a transition to the hcp phase , as it is observed in the xanes and xmcd spectra .
in summary , our investigations led to the following results : ( i ) assuming no structural nor spin distortions for bcc fe , a transition to the hcp structure can occur only at very high pressures , corresponding to a lattice parameter @xmath99 a.u .
, at which the system becomes paramagnetic ; ( ii ) burgers - type lattice distortions in collinear fm bcc fe can lead to a transition to the nm state and as a result to the bcc - hcp transition at the pressure close to the one observed experimentally , implying fluctuation mechanism of transition . however ,
the energy of lattice fluctuations required for the fm - pm transition is too high ; ( iii ) dlm calculations show an instability of spin disordered bcc fe upon a pressure increase .
this requires the transition to the pm state . to get this condition at the temperature of measurements , @xmath100
, the curie temperature should be low enough , i.e. @xmath101 .
however , this is not the case . on the other hand
, even a partial spin disorder can result in the decrease of the energy of lattice fluctuations required for transition to the pm state .
this implies that both effects , spin and lattice fluctuations , are the counterparts of the mechanism leading to a softening of the corresponding phonon modes responsible for the instability of the bcc state of fe and leading to bcc - hcp transition .
financial support within the framework of deutsches elektronen - synchrotron desy ( bmbf ) 05k13wma program and the priority schwerpunktprogramm spp 1415 is gratefully acknowledged .
according to friak and sob @xcite , transformations using only one parameter @xmath26 includes both transformations represented by the parameters @xmath4 or @xmath5 ( see fig .
[ fig_burgers ] ) , avoiding the high - energy configurations occurring upon their independent variation .
variation of @xmath26 parameter keeps the volume @xmath102 unchanged and leads to the hcp structure with the ratio @xmath103 , following the path being close to the minimum energy path @xcite .
the lattice parameters for the distorted orthorombic structure with 4 atoms / u.c . varies as follows : @xmath104 \\ c & = & a[\delta(2\sqrt{2 } - 3)/3 + 1 ] \\
v / v_0 & = & \sqrt{2}[\delta(2\sqrt{3 } - 3\sqrt{2})/6 + 2\sqrt{2}][\delta(2\sqrt{3 } - 3)/3 + 1 ] \end{aligned}\ ] ] the atomic positions in the unit cell with distortion are @xmath105 . transformation from bcc to hcp structure according to burgers scheme : ( a ) ( 110 ) planes of the bcc structure ; ( b ) opposite displacement of adjacent ( 110 ) planes along the @xmath3_{bcc}$ ] direction ; ( c ) hcp structure - after shear deformation along [ 001 ] direction with @xmath5 changing from @xmath106 in the case of bcc structure to @xmath107 . ,
title="fig:",scaledwidth=11.0%](a ) transformation from bcc to hcp structure according to burgers scheme : ( a ) ( 110 ) planes of the bcc structure ; ( b ) opposite displacement of adjacent ( 110 ) planes along the @xmath3_{bcc}$ ] direction ; ( c ) hcp structure - after shear deformation along [ 001 ] direction with @xmath5 changing from @xmath106 in the case of bcc structure to @xmath107 .
, title="fig:",scaledwidth=11.0%](b ) transformation from bcc to hcp structure according to burgers scheme : ( a ) ( 110 ) planes of the bcc structure ; ( b ) opposite displacement of adjacent ( 110 ) planes along the @xmath3_{bcc}$ ] direction ; ( c ) hcp structure - after shear deformation along [ 001 ] direction with @xmath5 changing from @xmath106 in the case of bcc structure to @xmath107 .
, title="fig:",scaledwidth=11.0%](c ) | the pressure induced bcc to hcp transition in fe has been investigated via ab - initio electronic structure calculations .
it is found by the disordered local moment ( dlm ) calculations that the temperature induced spin fluctuations result in the decrease of the energy of burgers type lattice distortions and softening of the transverse @xmath0-point @xmath1 phonon mode with @xmath2 $ ] polarization . as a consequence ,
spin disorder in an system leads to the increase of the amplitude of atomic displacements .
on the other hand , the exchange coupling parameters obtained in our calculations strongly decrease at large amplitude of lattice distortions .
this results in a mutual interrelation of structural and magnetic degrees of freedom leading to the instability of the bcc structure under pressure at finite temperature . |
we are concerned with quantum compression , which is best understood in the context of quantum information theory .
informally , the schumacher coding theorem @xcite states that a qubit source can be compressed to its von neumann entropy with high fidelity .
this statement is strongly analogous to the shannon coding theorem for classical information theory which states that a bit source can be compressed to its shannon entropy with high probability .
quantum compression is therefore concerned with algorithms for achieving compression near to the von neumann entropy of a qubit source in a reasonable amount of time and space .
the main result of this paper is a compilation algorithm which takes as input any family of classical block compression ( and decompression ) algorithm as well as a description of the quantum source .
the output is a quantum block compression and decompresion algorithm with the following properties : 1 . if the classical compression algorithm asymptotically approaches shannon entropy , then the quantum compression algorithm asymptotically approaches the von neumann entropy .
2 . if the classical compression and decompresion algorithm each take space @xmath0 and time @xmath1 , then the quantum compression algorithm uses space @xmath2 and time @xmath3 ( for any @xmath4 ) .
several hand constructed algorithms for block quantum data compression have been proposed .
these include quantum arithmetic compression @xcite which runs in @xmath5 time and a combinatoric index compression algorithm @xcite running in @xmath5 time and @xmath6 space . in further work ,
the combinatoric index compression was parallelized to a circuit of only polylogarithmic depth @xcite .
the remainder of this paper will discuss some information theory preliminaries , present the compiler , compare the performance of this compiler with hand crafted approaches , and then close with discussion .
suppose that we have a qubit source producing qubits in the state @xmath7 with probability @xmath8 .
then the density matrix of this source will be given by the average outer product : @xmath9 where @xmath10 are the eigenvectors of the density matrix and @xmath11 is the eigenvalue of @xmath10 .
we wish to compress the tensor product of @xmath12 qubits each drawn independently from this source so as to achieve a high `` fidelity '' .
fidelity is defined as the probability that a measurement can distinguish between the input and output states .
let @xmath13 be the input state chosen with random bits @xmath14 , and @xmath15 be the output of a compression / decompression cycle .
then , the fidelity is defined by : @xmath16 where @xmath17 is an expectation with respect to the quantum source as well as with respect to the outcome of the quantum measurement .
the schumacher coding theorem states that the minimum coding rate of any compression scheme with a high expected fidelity is given by the von neumann entropy : @xmath18 thus , we are particularly motivated by compression schemes which can ( asymptotically ) achieve the von neumann entropy .
a generic ( classical ) block compression algorithm , @xmath19 , maps @xmath12 bits to @xmath20 bits with @xmath12 larger than @xmath20 . the corresponding decompression algorithm , @xmath21 ,
will then ( hopefully ) invert the compression to give the original state .
we can use the general result that any classical function can be implemented @xcite ( reasonably efficiently ) on a quantum computer .
in particular , if we let @xmath22 be the bits of the input string and @xmath23 be the bits of the compressed string then we can compile the compression circuit into the reversible circuit which calculates : @xmath24 where @xmath25 are auxiliary bits .
the same trick can be done with the decompression circuit to get : @xmath26 pictorially , our circuits are the following : these elements will be our two fundamental building blocks in the construction of generic quantum block compression .
two challenges exist : 1 . @xmath27 and @xmath28 both have outputs which are entangled with their input .
this is unacceptable in any quantum compression scheme .
2 . there are some quantum sources with a low von neumann entropy but a large shannon entropy .
we will need to make a circuit which achieves compression near the shannon entropy to achieve compression near the von neumann entropy .
these two challenges are met in the next subsections .
we are not finished with the compression because the output of @xmath27 leaves the compressed bits entangled with the input bits .
there is a simple construction which removes this entanglement : compose the decompressor with the compressor to get the following circuit : here we have set the auxiliary bits to @xmath29 .
the output of the first compression block is the compressed bits along with the uncompressed bits .
the uncompressed bits are then fed into the decompression block instead of @xmath29 s , resulting in an output consisting of the compressed bits plus things of the form @xmath30 .
this term is @xmath29 precisely when the compression is lossless . by assumption ,
the compression is lossless with high probability implying that we will essentially always measure the @xmath29 vector .
at this point , we can measure all of the `` @xmath29 '' qubits . if any of our measurements are nonzero , we have detected a low probability event .
the process of this measurement will project the wavefunction onto the typical subspace , letting us achieve the projection done explicitly in @xcite for `` free '' as a byproduct of our compression process .
this construction is also particularly elegant because decompression ( with entaglement removal ) can be accomplished by simply reading the circuit backwards . in summary
, we now have a technique for constructing a quantum compression circuit which outputs the compressed string _ without _ extra entanglement .
the above approach only does _ classical _ compression of qubits and so it can only achieve the shannon entropy .
in particular , it will fail to compress some highly compressible quantum state distributions such as the state which is @xmath31 with probability @xmath32 .
this state is highly compressible because it has a trivial distribution over possibilities and yet no classical algorithm will compress the qubits well . in order to accomplish compression of this state , we must add an extra rotation to the input and output of the compression and decompression .
for example , the operation @xmath33 ( hadamard applied to each bit ) , would convert the input to @xmath34 which is highly compressible .
this extra rotation involves @xmath35 operations since it can be done on each symbol from the quantum source individually giving us the following circuit : the interpretation of this circuit is simple : we first rotate every qubit to align the eigenbasis and the computational basis . then
, we apply the compression circuit using the decompression circuit to remove entanglement with the input values . mathematically
, we compute @xmath36 where @xmath37 is the rotation into the eigenbasis .
once again , decompression is done using the circuit backwards .
this solution works for _ any _ quantum source .
given a description of any quantum source , the eigenvalue decomposition theorem says that we can construct a matrix @xmath37 which satisfies : @xmath38 where @xmath39 is a diagonal matrix of eigenvalues and @xmath37 is a matrix consisting of eigenvectors . the @xmath37 matrix will rotate the input source so that the computational basis and the eigenbasis coincide .
this rotation does not affect the von neumann entropy of the source .
now , given that the computational basis and eigenbasis coincide , the von neumann entropy will be exactly the same as the shannon entropy .
the compiled circuit will therefore achieve the von neumann limit .
to summarize , the compiler is the following : _ make_quantum_compression_circuit(@xmath40 , @xmath41 , @xmath42 ) _ 1 . compile the quantum reversible circuits @xmath28 and @xmath27 from the classical circuits @xmath42 and @xmath41
2 . compose @xmath28 and @xmath27 as in section [ unentangle ] .
3 . calculate the eigenvalue matrix , @xmath37 , given @xmath40 .
4 . rotate all input and output qubits of the quantum circuit by @xmath37 as in section [ rotate ] .
how efficiently can we compile compression algorithms made for classical ( irreversible ) computers into reversible quantum mechanical programs ?
there are a very few results here which are well summarized in @xcite .
there are a variety of tradeoffs all of which are typically parameterized by the space @xmath0 and time @xmath1 of the original irreversible algorithm .
one of the more concrete results is that reversible space is @xmath43 and the reversible time is ( for any @xmath4 ) @xmath44 . applying this result to the common bzip2 block compression algorithm which runs in space @xmath45 and time @xmath46 will require @xmath47 space and @xmath48 time ( for any @xmath4 ) .
if we choose @xmath49 , this is a smaller running time than @xmath5 as used by previous hand crafted quantum block compression algorithms .
it is easy to imagine that hand compilation of irreversible compression / decompression algorithms can yield further functional improvements .
the technique for automatically compiling classical compression algorithms into quantum compression algorithms gives us a baseline and general structure for comparison with hand crafted quantum compression algorithms .
the automatic approach is reasonably efficient and can give results better than current hand crafted approaches in some cases .
hand crafted algorithms which are functionally more efficient than this baseline are a nontrivial improvement .
the technique of section [ unentangle ] applies to _ any _ pair of classical algorithms with an inversion property : @xmath50 .
this includes encryption and decryption or encoding and decoding . for algorithms with this property
, we can construct quantum analogs with a special property : the `` inputs '' are disentangled from the `` outputs '' .
this disentanglement comes at a relatively small price : a factor of @xmath51 increase in the circuit size .
in addition to hand crafting efficient compression algorithms , one other significant problem in quantum data compression exists : removing the requirement for full information about the quantum source .
we know this is ( information theoretically ) possible @xcite although no explicit quantum algorithm yet exists that achieves asymptotically efficient compression without knowledge of the source .
intuitively , this is possible because the precision of the specification of @xmath37 improves exponentially with increasing ( qu)bits while the necessary precision required in specifying @xmath37 scales linearly with @xmath12 , the number of input qubits .
consequently , the portion of the bits which must be allocated to specifying @xmath37 is asymptotically @xmath29 .
an explicit algorithm for doing universal quantum compression would yield a significant improvement in our understanding of quantum compression .
i would like to thank andris ambainis , klejda bega , david relyea , peter shor , luis von ahn , and ke yang for advice and comments in preparing this paper
. 1 b. schumacher , `` quantum coding '' physical review a , vol . 51 pp .
2738 - 2747 , 1995 .
i. l. chuang and d. s. modha , `` reversible arithmetic coding for quantum data compression '' ieee transactions on information theory , vol .
1104 - 1116 .
r. cleve and d. p. divincenzo , `` shumacher s quantum data compression as a quantum computation '' , physical review a , vol .
54 , 4707 , 1996 . s. l. braunstein , c. a. fuchs , d. gottesman , and h. lo , `` a quantum analog of huffman coding '' , 1998 ieee symposium on information theory , p. 353 .
e. bernstein and u. vazirani , `` quantum complexity theory '' , special issue on quantum computation of the siam journal of computing , oct .
1997 h. buhrman , j. tromp , p. vitanyi , `` time and space bounds for reversible simulation '' , proc .
international conference on automata , languages and programming , lecture notes in computer science , springer - verlag , berlin , 2001 .
r. josza , m. horodecki , p. horodecki , r horodecki , `` universal quantum information compression '' , phys .
letters 81 , 1714 ( 1998 ) . | a generic approach for compiling any classical block compression algorithm into a quantum block compression algorithm is presented . using this technique , compression asymptoticaly approaching the von neumann entropy of a qubit source
can be achieved .
the automatically compiled algorithms are competitive ( in time and space complexity ) with hand constructed quantum block compression algorithms . |
a generalized concept in present theoretical physics is that description of all physical phenomena should be derivable from first principles in a unified way .
however , the bridge towards concrete applications requires elaborated procedures and judicious arguments .
this is the case of quantum chromodynamics ( qcd ) , which is the accepted theoretical model for the strong interactions . despite the fact that it is perturbative in the high energy regime , the fundamental state of matter corresponds to the opposite limit , where confinement and the breakdown of symmetries make it mathematically intractable .
different effective models , such as nambu- jona lasinio , skyrme , bag - like , and chiral perturbation theory attempts to translate the main features of qcd into the hadronic phase . on the other hand , lattice simulations and qcd sum - rules share this aim , although using different methods .
+ qcd sum - rules is a ingenious procedure to reveal the foundations of certain hadronic properties .
the method is based on the evaluation of correlation functions in terms of quarks and gluons degrees , then applying the operator product expansion it is possible to express it as combinations of perturbative contributions and condensates ( non - perturbative ) .
finally this expansion is connected with the hadronic counterpart by means of a borel transformation .
the method was developed to study meson @xcite as well as baryon @xcite properties in vacuum , it was subsequently generalized to study finite density systems @xcite .
these calculations provide a useful guide for some static properties of hadrons immersed in a dense medium , as concerning qcd symmetries and phenomenology .
however , they are not able to take into account the dynamical aspects of hadronic matter , which may be achieved by inserting coherently these results into a theoretical model of hadronic interactions .
+ a similar situation was found in the study of nuclear structure , as the standard combination of _ microscopic _ potentials and the relativistic dirac - brueckner approximation for nuclear matter gives rise to reliable , density dependent , nucleon self - energies .
notwithstanding , this procedure was inadequate to treat finite nuclei due to unsurmountable mathematical difficulties . a feasible solution to this dilemma
was proposed in @xcite , by defining density dependent meson - nucleon vertices in terms of the self - energies obtained in dirac - brueckner calculations .
the enlarged hadronic model keeps the mathematical versatility of the quantum hadrodynamic models @xcite , but is equipped with couplings reflecting the properties of the nuclear environment .
the structure of spherical nuclei , in the medium to heavy range , was studied within this framework using the hartree approximation .
+ further improvements , developed in @xcite , replace the density dependence of the meson - nucleon vertices by an expansion in terms of in - medium nucleon condensates .
this conceptual replacement restores the covariance and the thermodynamical consistency of the model , giving rise to the density dependent hadronic field theory ( ddhft ) .
it was recently appended by introducing momentum dependent vertices @xcite , and a expansion of the vertices in terms of meson mean field values @xcite .
+ the validity of the method is only justified _ a posteriori _ , and relies in the flexibility of hadronic field models to accommodate pieces of information provided by another field of research . the enlarged model yields a simpler and more intuitive description , instead of making involved calculations based on first principles interactions .
a significative exemplification of this standpoint was given by the brown - rho scaling of hadronic masses .
taking into account the chiral and scale symmetries of qcd solely , an approximate scaling law for the in - medium hadronic masses was derived in @xcite .
this hypothesis was applied to describe heavy ion collision , reaching an excellent agreement with the experimental results for the low mass dilepton production rate @xcite .
+ given a set of physically meaningful self - energies as input for ddhft , there is room for a full family of hadronic models , according to the field parameterization assumed for the vertices . in principle , this may lead to different predictions for nuclear observables .
this point has not been investigated yet , and it is the main purpose of the present work .
+ in a previous paper @xcite , the author attempted to make use of the scheme outlined above to relate nuclear observables with qcd inspired results . for this purpose we took as input the nucleon self - energies in symmetric nuclear matter obtained in @xcite by using qcd sum rules .
however , ref .
@xcite does not exhaust the physical description given there , as it covers isospin asymmetric nuclear matter also .
so , we aim to complete our theoretical model including isospin degrees of freedom , based on the results presented in @xcite .
we have organized this work presenting in the following section the general features of ddhft , and particularly the meson parameterization of the hadronic vertices . in section [ seciii ]
we present and discuss the results for symmetric nuclear matter .
a final summary is given in section [ seciv ] .
as mentioned above , ddhft was designed to incorporate into the hadronic field formalism pieces of information produced by other theoretical frames . strictly speaking
, ddhft takes as input the nucleon self - energies , which can be decomposed into iso - scalar and iso - vector components , each one containing lorentz scalar and vector contributions : _
i. e. _ @xmath0 , @xmath1 , @xmath2 , @xmath3 , where the superscript @xmath4 distinguishes iso - vector quantities , and @xmath5 stands for proton and neutron respectively . for homogeneous nuclear matter in steady state , they can be parameterized as functions of continuous parameters , like baryonic current density @xmath6 and temperature , characterizing the macroscopic state of hadronic matter . in particular
, there exists a reference frame where the spatial part of the baryonic current vanishes in the mean . in its former version @xmath7
was taken from relativistic brueckner - hartree - fock calculations with one boson exchange potentials @xcite .
on the other hand , a lagrangian density is proposed in terms of meson fields @xmath8 , @xmath9 in the iso - scalar sector , and @xmath10 , @xmath11 in the iso - vector one .
the indices @xmath12 stand for the meson isospin projection @xmath13 where @xmath14 , @xmath15 has been used , and @xmath16 represents a isospinor .
we have adopted the convention that repeated isospin indices must be summed .
the isovector vertices @xmath17 take values over the space generated by the pauli matrices @xmath18 and the identity .
the masses have been fixed at the phenomenological values @xmath19 mev , @xmath20 mev , @xmath21 mev , @xmath22 mev , and @xmath23 mev .
the iso - scalar vertices are assumed to depend on the scalar combinations : @xmath24 , @xmath25 , @xmath26 , and @xmath27 , whereas the isovector @xmath17 could depend linearly on @xmath28 , or @xmath10 , and on the scalars @xmath29 , @xmath30 , @xmath31 , and @xmath32 .
the baryonic and isospin current have been written as @xmath33 and @xmath34 , respectively .
this is not the most general dependence , but it keeps a clear separation between isospin scalar and vector degrees of freedom . in ref .
@xcite the dependence on @xmath35 or @xmath36 was presented and the case with only @xmath36 was explicitly studied ; in @xcite the meson dependence of the vertices was introduced within ddhft , for symmetric nuclear matter .
the field equations obtained for this lagrangian density are @xmath37 \psi_b \nonumber\\ & & \;\;\;+ \bar{\psi } \frac{\partial}{\partial \bar{\psi_a } } \left ( \gamma_s \sigma - \gamma_w\not \ ! \omega+ \gamma_{f}^c \,\phi_c+\gamma_{r}^c \ , \not \ !
\rho_c \right ) \psi=0 \label{nucleoneq}\\ & & \;\;\;\left ( \square+m_s^2 \right)\sigma=\bar{\psi}\left [ \gamma_s+ \left(\frac{\partial \gamma_s}{\partial m_2 } \sigma-\frac{\partial \gamma_w}{\partial m_2 } \not\ !
\omega\right ) \right ] \psi \label{sigmaeq } \\ & & \;\;\;\partial_\mu w^{\mu \nu}+m_w^2 \omega^\nu=\bar{\psi}\left [ \gamma_w \gamma^\nu-\frac{\partial m_1}{\partial \omega_\nu } \left(\frac{\partial \gamma_s}{\partial m_1 } \sigma-\frac{\partial \gamma_w}{\partial m_1 } \not\ !
\omega\right ) \right]\psi .
\label{omegaeq } \\ & & \;\;\;\left ( \square+m_f^2 \right)\phi^c=\bar{\psi}\left [ \gamma_f^c+\frac{\partial m_4}{\partial \phi_c } \left ( \frac{\partial \gamma_f^a}{\partial m_4 } \phi_a+\frac{\partial \gamma_r^a}{\partial m_4 } \not\ !
\rho_a\right ) \right ] \psi \label{phieq } \\ & & \;\;\;\partial_\mu r^{\mu \nu}_c+m_r^2 \rho^\nu_c=-\bar{\psi}\left [ \gamma_{r\,c } \gamma^\nu+\frac{\partial m_3}{\partial \rho_\nu^c } \left(\frac{\partial\gamma_f^a}{\partial m_3 } \phi_a+\frac{\partial \gamma_r^a}{\partial m_3 } \not\ !
\rho_a\right ) \right]\psi , \label{rhoeq}\end{aligned}\ ] ] where the isospin indices have been denoted as : @xmath38 for nucleons , and @xmath39 for mesons . in the first equation , the vertex derivatives must be interpreted as @xmath40 the mean field approximation ( mfa ) is suited to describe homogeneous matter . within this scheme ,
meson fields are replaced by their uniform mean values and bilinear combinations of spinors are replaced by their expectation values . for homogeneous , isotropic nuclear matter in steady state , some simplifications arise in the meson mean values .
for instance , their coordinate dependence can be neglected ; the spatial components of the vectors @xmath9 and @xmath41 become null , and only the third component of the isovector mesons are non - zero , due to isospin conservation . + therefore within mfa , eqs .
( [ nucleoneq]-[rhoeq ] ) reduce to @xmath42 \psi_b , & & \ ; \label{nucleoneq2}\\ m_s^2 \tilde{\sigma}=<\bar{\psi}\,\gamma_s\psi > + < \bar{\psi}\frac{\partial \gamma_s}{\partial m_2 } \psi \,\sigma > & -&<\bar{\psi}\frac{\partial \gamma_w}{\partial m_2}\gamma_\mu \psi \omega^\mu > ; \label{sigmaeq2 } \\
g^\nu_0\;m_w^2 \tilde{\omega}=<\bar{\psi}\,\gamma_w \gamma^\nu\psi>-<\bar{\psi}\,\frac{\partial \gamma_s } { \partial m_1}\psi\,\frac{\sigma\omega^\nu}{m_1 } > & + & < \bar{\psi}\frac{\partial \gamma_w}{\partial m_1}\gamma_\mu\psi \,\frac{\omega^\nu\omega^\mu}{m_1 } > ; \label{omegaeq2}\\ \delta_{3c}\;m_f^2 \tilde{\phi}=<\bar{\psi}\gamma_f^c \psi>+<\bar{\psi}\frac{\partial \gamma_f^a}{\partial m_4}\psi\,\frac{\phi_c\phi_a}{m_4}>&+&<\bar{\psi}\frac{\partial \gamma_r^a}{\partial m_4}\gamma_\mu\psi \frac{\phi_c}{m_4 } \rho_a^\mu > ; \label{phieq2 } \\
-\delta_{3c}g^\nu_0 m_r^2 \tilde{\rho}=<\bar{\psi}\gamma_{r\,c } \gamma^\nu \psi>+<\bar{\psi } \frac{\partial\gamma_f^a}{\partial m_3}\psi \frac{\rho^\nu_c\phi_a}{m_3 } > & + & < \bar{\psi}\frac{\partial \gamma_r^a}{\partial m_3}\gamma_\mu\psi \frac{\rho^\nu_c \rho^\mu_a}{m_3}>;\label{rhoeq2}\end{aligned}\ ] ] where tildes over the meson symbols stand for their mean field values , and the two last terms within brackets in eq .
( [ nucleoneq2 ] ) can be regarded as the scalar and vector components of the nucleon self - energies : @xmath43 with @xmath44 . for the sake of concreteness ,
we have examined three possible parameterizations : 1 .
symmetric nuclear matter , with @xmath45 , @xmath46 ; 2 .
symmetric nuclear matter , with @xmath47 , @xmath46 ; 3 .
asymmetric nuclear matter , with @xmath48 , @xmath49 , @xmath50 , @xmath51 .
the first two cases are comparable to the instances presented in @xcite as vector and scalar dependencies .
the last case is a development of the preliminaries calculations of @xcite , the field dependence of the isospin vertices have been chosen so as to adjust the qcd sum rules calculations of @xcite .
+ these results are simplified within the assumptions ( a - c ) above , for the sake of completeness we consider separately each of these cases : @xmath52,\label{casea } \\ & & \tilde{\sigma}=\gamma_s n_s / m_s^2 , \
, \tilde{\omega}=\gamma_w n_b / m_v^2 , \nonumber \\ \mbox{b)}&&\ , \sigma_s=\tilde{\sigma}\left(\gamma_s+ \frac{d\gamma_s}{dn_s } n_s\right),\ , \sigma_\nu = g_{\nu 0}\,\tilde{\omega } \left ( \gamma_w+\frac{d\gamma_w}{dn_b}\,n_b\right ) , \label{caseb}\\ & & \,\tilde{\sigma } , \,\mbox{and } \;\tilde{\omega}\ , \mbox{have the same expressions as in ( a ) } , \nonumber\\ \mbox{c)}&&\ , \sigma_s^{ab}=\sigma_s^{isos}\delta_{ab}+\sigma_s^{isov\,ab } , \ ; \sigma_\nu^{ab}=g_{\nu 0}\left(\sigma_v^{isos}\delta_{ab}+\sigma_v^{isov\,ab}\right ) , \label{casec } \\ & & \sigma_s^{isos}= \gamma_s\tilde{\sigma},\;\sigma_s^{isov\,ab}=\gamma^{3\,ab}_f \tilde{\phi } , \;\sigma_v^{isos}=\gamma_w \tilde{\omega } , \;\sigma_v^{isov\,ab}=-\gamma_r^{3\,ab}\tilde{\rho},\nonumber\\ & & m_s^2 \tilde{\sigma}=\frac{d\sigma_s^{isos}}{d\tilde{\sigma}}n_s , \ ; m_w^2 \tilde{\omega}= \frac{d\sigma_v^{isos}}{d\tilde{\omega}}n_b,\nonumber\\ & & m_f^2\tilde{\phi}= < \bar{\psi}\frac{d\sigma_s^{isov}}{d\tilde{\phi}}\psi>-
< \bar{\psi}\frac{d\sigma_v^{isov}}{d\tilde{\phi}}\gamma_0\psi > , \ ; m_r^2\tilde{\rho}=<\bar{\psi}\frac{d\sigma_v^{isov}}{d\tilde{\rho}}\gamma_0\psi>.\nonumber\end{aligned}\ ] ] obtaining the equations listed above , we have used the wick theorem and the hartree approximation , therefore contractions of fields belonging to different full contracted terms ( both in lorentz and isospin indices ) have been neglected .
this approach allowed us to extract the vertices @xmath53 and its derivatives from the expectation values , and to express the mean value of products of more than two fermion fields as product of baryonic densities and currents .
finally , we have adopted the reference frame of static matter , therefore we have @xmath54 , @xmath55 , with @xmath56 the number of neutrons or protons per unit volume , for @xmath57 respectively , and @xmath58 the baryonic number density .
furthermore we have introduced the simplifying notation @xmath59 , @xmath60 , @xmath61 , @xmath62 , @xmath63 , @xmath64 , @xmath65 , @xmath66 , @xmath67 , @xmath68 , and @xmath69 .
cases ( a ) and ( b ) yield meson equations resembling those of the walecka model @xcite , and nucleon self - energies with the _ rearrangement _ contributions added . in the instance ( c ) the self - energies have the same structure as in qhd - i , but the source term of the meson equations are modified . with respect to the isovector meson field equations , it must be noted that @xmath70 for @xmath71 as isospin is a conserved charge . in order to give explicit expressions for the corresponding vertices ,
we follow the parameterization of @xcite : @xmath72 with @xmath73 , for protons ( neutrons ) , @xmath74 , @xmath75 , the remaining coefficients can be expressed in terms of the quantities used in ref .
@xcite as @xmath76 , @xmath77 , @xmath78 , @xmath79 , @xmath80 ) , @xmath81 , which were obtained in @xcite , by averaging over the borel mass . for the sake of completeness we give here the numerical values @xmath82 , @xmath83 , @xmath84 , @xmath85 , @xmath86 , @xmath87 gev ; @xmath88 gev ; @xmath89
; @xmath90 gev ; @xmath91 gev , @xmath92 , and @xmath93 for @xmath94 and @xmath95 respectively ; @xmath96 , @xmath97 , and @xmath98 for @xmath99 and @xmath100 respectively .
+ from eq .
( [ qcdself ] ) , we can extract isoscalar and isovector contributions , _
i.e. _ we consider the splitting @xmath101 , with @xmath102 with @xmath103 . + a choice of the isovector interaction in eq .
( [ lagrangian ] ) , which is coherent with these results is @xmath104 where @xmath105 is assumed to depend only on @xmath106 , whereas @xmath107 and @xmath108 are considered as functions of @xmath109 only .
the following vertices are deduced from it @xmath110 in the mfa , they give rise to the following contributions to the nucleon self - energies @xmath111 the isovector meson field equations can now be written as @xmath112 in ddhft @xcite , the isoscalar vertices @xmath113 and @xmath114 are defined by means of the relations @xmath115 where @xmath116 are the self - energies obtained with one boson exchange potentials in the dirac - brueckner - hartree - fock approach for symmetric nuclear matter . it must be noted that the right hand sides of these equations do not coincide in general with the dynamical self - energies
. therefore eqs .
( [ ddhft ] ) and ( [ casea])-([caseb ] ) do not coincide , unless the _ rearrangement _ terms were omitted . in @xcite ,
the author proposed that the vertices are solutions of someone of the differential equations ( [ casea])-([casec ] ) , instead .
they must be solved together with the self - consistent condition for the meson fields .
the explicit form of part of these solutions have been anticipated in @xcite , so we summarize them below and add new results to the case ( c ) concerning isospin asymmetric nuclear matter .
\mbox{b)}\,\gamma_s^2&=&2\ , \left(\frac{m_s}{n_s}\right)^2\int_0^{n_b } \ , dn'\frac{dn_s}{dn'}\,\sigma_s^{(in)},\;\;\gamma_w^2=2\ , \left(\frac{m_w}{n_b}\right)^2\int_0^{n_b } \
, dn'\,\sigma_v^{(in)};\nonumber \\
\mbox{c)}\,\gamma_s&=&\sigma_s^{isos \,(in)}/\tilde{\sigma},\;\gamma_w=\sigma_v^{isos\,(in)}/\tilde{\omega } , \;g=\frac{\alpha t n_b}{\tilde{\phi}(n_0+\lambda n_b)},\nonumber\\ g_f&=&\frac{\beta_s-\alpha'_s \alpha n_b}{n_0+\lambda n_b}\,\frac{t n_b}{\tilde{\phi } } , \ ,
g_r=-\frac{\beta_v-\alpha'_v \alpha n_b}{n_0+\lambda n_b}\,\frac{t n_b}{\tilde{\rho}},\nonumber \\
\tilde{\sigma}^2&=&\frac{2}{m_s^2}\int_0^{n_b } \ , dn'\,n_s\ , \frac{d\sigma_s}{dn'}^{isos\,(in)},\;\;\tilde{\omega}^2=\frac{2}{m_w^2}\int_0^{n_b } \ , dn'\,n ' \,\frac{d\sigma_v}{dn'}^{isos\,(in)},\nonumber\\ \tilde{\phi}^2&=&\frac{1}{m_f^2}\sum_{a , c=1}^2(1-i_3^c t)\int_0^{n_b } \ , dn'\left[n_s^a \frac{\partial\sigma_{s\,a}}{\partial n_c'}^{isov\,(in)}-\frac{1}{2}n'(1-i_3^a t ) \frac{\partial\sigma_{v\,a}}{\partial n_c'}^{isov\,(in)}\right],\nonumber \\
\tilde{\rho}^2&=&\frac{1}{2m_r^2}\sum_{a , c=1}^2(1-i_3^c t)(1-i_3^a t)\int_0^{n_b } \ , dn ' n ' \frac{\partial\sigma_{v\,a}}{\partial n_c'}^{isov\,(in)}.\nonumber\end{aligned}\ ] ] we have assumed that the input functions have been parameterized in terms of the partial nucleon densities @xmath118 , furthermore within the two last eqs . of case ( c ) , the isospin parameter @xmath119 is held constant in the integration .
+ thus we have obtained a set of relations , defining a hadronic field model suited to reproduce the nucleon self - energies provided by other theoretical framework . in particular ,
the meson dependent vertices have been extended to deal with isospin asymmetric matter .
the resulting vertices are given as functions of @xmath120 , and eventually @xmath121 , but they can be rewritten in terms of the meson fields or nucleon condensates . + in the next section , these results will be compared with the standard ddhft treatment , and the ability to adjust the nuclear matter phenomenology will be examined .
a well known feature of isospin symmetric nuclear matter is its energy per particle , having a minimum at a baryonic density about @xmath122 .
this property gives rise to bound states at zero temperature , and it is the least requirement that a model of nuclear matter should satisfy . starting with eq .
( [ lagrangian ] ) , we have evaluated the energy - momentum tensor @xmath123 by the canonical procedure , and the energy per unit volume in the mfa is obtained by taking the in - medium expectation value of @xmath124 . @xmath125 where we have introduced the fermi momentum @xmath126 , related to the baryonic density by @xmath127 , and the effective nucleon mass @xmath128 .
the mean value of the isovector meson fields becomes zero for symmetric nuclear matter .
+ finally , the binding energy is defined as @xmath129 , which should have a minimum value @xmath130 mev at the normal density @xmath131 , to satisfy the nuclear matter phenomenology . using the thermodynamical relation @xmath132 we have obtained the pressure for the nuclear matter , with the chemical potential given by @xmath133 , being @xmath134 .
+ in first place , we have examined the effects of introducing the effective vertices by means of the algebraic eqs .
( [ ddhft ] ) , or by solving a differential equation of the type listed in eqs .
( [ casea])-([casec ] ) .
it must be noted that the first option has the property of reproducing the energy density of the original dirac - brueckner calculations , whenever the parameterization of the vertices in terms of only the baryon number density is chosen .
otherwise , it is a definition without any special physical significance .
the alternative procedure proposed in @xcite , aims to make a connection between the hadronic vertices and another theoretical framework , by imposing the equality of the nucleon self - energies evaluated in both cases . + for the purpose stated , we have chosen as input the results of @xcite . in this work
, an ansatz for the vertices is proposed that fits several specific dirac - brueckner outcomes , but avoiding the unphysical behavior they exhibit in the zero density limit .
the ansatze for the couplings @xmath135 and @xmath136 are rational functions of the relative baryonic density @xmath137 .
the corresponding self - energies are obtained as @xmath138 , @xmath139 , with the meson mean field values given by equations similar to those of eq .
( [ casea ] ) . for details , see ref .
[ fig1 ] we show the binding energy and pressure evaluated within the standard ddhft treatment of @xcite . given the functions @xmath140 as above , we can also use them to deduce the vertices @xmath141 which reproduce these self - energies within the approach ( a ) ; _ i.e. _ solving eqs .
( [ casea ] ) .
the corresponding equation of state is shown in the same figure .
we have examined densities up to four times @xmath131 , a range that may be reached inside neutron stars .
it can be seen that similar results for the pressure are obtained for densities below @xmath142 , but the binding energy shows appreciable differences .
these results are qualitatively comparable , although for high densities the standard ddhft treatment gives a softer growth for both @xmath143 and @xmath144 .
these differences could be significative for those phenomena dominated by the regime of extreme densities , such as the structure of neutron stars . as the next subject we have examined the effect of multiple representations . indeed , given a set of functions @xmath145 depending on the macroscopic variables of nuclear matter , there exist an indefinite number of models like that of eq .
( [ lagrangian ] ) , capable of reproduce them within mfa .
some of these possibilities have been listed as cases ( a - c ) in the previous section .
the formal aspects have been summarized in eqs .
( [ casea])-([casec ] ) , now we focus on some thermodynamical observables of symmetric nuclear matter .
+ to avoid biased conclusions we have considered two inputs of diverse source . in addition to the results of @xcite , based on one - boson exchange potentials , we have included the qcd sum rules calculations of @xcite .
we have evaluated the binding energy and the pressure , for each of these inputs under the three assumptions ( a - c ) .
we have found that despite the different parameterizations used for a given input , the results are practically undistinguishable in both cases . in fig .
[ fig2 ] we show this feature for the calculations using the inputs obtained from @xcite , where it can be seen that the three curves are nearly coincident .
the range of densities has been restricted to the region of validity of the qcd sum rules computations .
the binding energy is monotonically decreasing , it seems to have a minimum near @xmath146 where the effective nucleon mass tends to zero .
this description does not adjust to the nuclear matter phenomenology , however we keep ref .
@xcite into consideration as it offers a physically meaningful input .
it must be noted that the result ( c ) does not coincide with the preliminary result given in ref .
@xcite , as the @xmath147 meson contribution to the energy was not properly taken into consideration there .
+ a similar conclusion is obtained by using the self - energies extracted from @xcite , since the curves corresponding to the instances ( b ) and ( c ) ( not shown in the figure ) follow closely the solid curves of fig .
[ fig1 ] , in the full range @xmath148 .
+ the different behavior between the inputs obtained from @xcite and @xcite , for the energy per particle , can be justified by examination of fig .
[ fig3 ] . in the upper panel
the self - energies as functions of the density , show a monotonous increase in both cases .
however the rate of growth of @xmath149 becomes lesser than the corresponding to @xmath150 for densities @xmath151 , in the parameterization of @xcite .
the qcd sum rules outputs instead , keeps a faster increase of the scalar as compared to the vector self - energy , over all the domain of densities .
this behavior is emphasized in the bottom panel , where the difference @xmath152 is monotonous increasing in one case , but exhibits a extremum in the other one .
furthermore , a constant quotient @xmath153 is obtained for the qcd sum rules input , but it increases slowly for the other instance .
anyhow , the rational density dependence of eq .
( [ qcdsrself ] ) for the iso - scalar self - energies could be retained in order to simulate the nuclear matter mechanism of saturation , for instance with @xmath154 and @xmath155 we have obtained a binding energy of @xmath156 mev at @xmath157 .
these values must be contrasted with the parameterization deduced from @xcite , @xmath158 and @xmath159 .
+ the density dependence of the vertices , obtained by using the input of ref .
@xcite under the three approaches ( a - c ) is presented in fig .
it must be noted that the expressions for @xmath160 become singular at zero density , nevertheless they have finite limits : @xmath161 , for all the cases under consideration . +
a common feature of the vertices is a strong decrease up to @xmath162 , which moderates above this value .
the cases ( b ) and ( c ) yield very similar results for both @xmath113 and @xmath114 , differences with the case ( a ) may lead to qualitatively distinct descriptions for subsequent applications .
all these results for the vertices @xmath163 can be summarized by an expression in terms of its natural variable @xmath164 , @xmath165 , with @xmath166 and the numerical coefficients are shown in table 1 .
.numerical values of the fitting coefficients for the density dependence of the vertices , obtained in the approaches ( a - c ) using the self - energies deduced from @xcite . [ cols="<,^,^,^,^,^,^,^,^,^,^",options="header " , ] another insight into the basic phenomenology of nuclear matter , can be given by the landau parameters .
they can be evaluated as the legendre projections of the second derivative of @xmath167 with respect to the baryonic density , for a deduction see @xcite .
these parameters are very useful in regard to collective phenomena of the dense nuclear environment , as for instance , phase transitions , the giant monopolar and quadrupolar modes , or the sound velocity @xcite . in order to keep track of the momentum dependence , we assume a non - zero spatial component of the self - energy @xmath168 , the baryonic current @xmath169 and the omega meson @xmath170 , taking the zero limit of these quantities at the end of the calculations .
so , for example , we write @xmath171 into the integral of eq .
( [ eos ] ) . denoting by @xmath172 the occupation number of nucleon states with a well defined momentum @xmath173
, then we define @xmath174 in the second equality , the limit of isotropic nuclear matter has been taken .
the derivative of the baryonic current can be evaluated in the mfa @xmath175 the remaining derivatives are @xmath176 whose explicit form depends on the approach used ( a - c ) . + the landau parameters @xmath177 are defined by means of the projections @xmath178 with @xmath179 the cosine of the angle between @xmath173 and @xmath180 , and @xmath181 the legendre polynomial of order @xmath182 . at the end of the calculations , the replacement @xmath183 is made . by this procedure
, we have obtained the landau parameters of zero and first orders @xmath184 .
\nonumber\end{aligned}\ ] ] it is costumary to normalize the landau parameters with the density of states at the fermi surface @xmath185 , so that @xmath186 .
+ for a given model of self - energies , the result for @xmath187 does not depend on the field parameterization of the vertices .
the parameter @xmath188 instead , depends on @xmath189 which certainly differs for each of the instances ( a - c ) . in fig .
[ fig5 ] the density dependence of both landau parameters is shown , significative departures for @xmath188 are found for densities above @xmath190 .
+ consequently all physical magnitudes depending solely on @xmath187 , such as the nuclear compressibility @xmath191 , do not distinguish among the field parameterization of the vertices .
the same feature is shared by the giant monopole mode energy @xmath192 , with @xmath193 , and the first sound velocity @xmath194 . on the other hand ,
the excitation energy of the quadrupole state is given by @xmath195 , with @xmath196 , therefore it is sensitive to the approach used .
we have obtained in our calculations @xmath197 mev , @xmath198 mev , and @xmath199 mev at the normal density , which must be compared with the empirical values 230 mev @xmath200 270 mev , @xmath201 mev , and @xmath202 mev .
+ another interesting phenomenological property that can be related to the landau parameters , are the zero sound modes .
these are longitudinal collective modes propagating in nuclear matter at zero temperature , whose dispersion relation can be found as the zeros of the longitudinal dielectric function @xmath203 with @xmath204 and @xmath205 the lindhard function .
collective modes have been studied in detail within relativistic field models , see for instance @xcite .
low lying collective modes can be classified as instability modes , in the low density regime , and zero sound modes with a characteristic linear dispersion relation @xmath206 . in fig .
[ fig6 ] , the dispersion relation in terms of the baryonic density is shown for the standard ddhft treatment and for the approaches ( a - c ) , within the self - energy parametrization of @xcite .
these results are appropriate for the regime @xmath207 , but keeping the linear dispersion relation .
the instability and zero sound modes are clearly distinguishable , as the first one is closely related to the equation of state , all the corresponding curves practically coincide .
the zero sound yields very different behaviors , the extremum values for its threshold are obtained within the approaches ( b ) and the standard ddhft treatment .
we have obtained @xmath208 and @xmath209 , respectively . in all the cases
we have obtained two - folded zero sound , and taking into account that @xmath210 has non zero imaginary part for @xmath211 , only the upper mode is undamped .
we have examined in this work , two definitions for the vertices of a hadronic field model in terms of the self - energies provided by another theoretical framework .
the first one is the standard ddhft treatment , which uses eqs .
( [ ddhft ] ) to define the vertices @xmath212 between nucleon and scalar and vector mesons , respectively .
there , @xmath213 denotes the nucleon self - energies , based on dirac - brueckner calculations with one - meson exchange potentials .
although vector and scalar dependencies have been stated @xcite , most practical applications have used only the first one , similar to our case ( a ) .
this particular choice has the property of reproducing the equation of state of the source calculations , but on the contrary the self - energies formally differ from the inputs @xmath213 .
an alternative definition was proposed in @xcite , it consisted in equating both , the input self - energy and that evaluated within the hadronic model with vertices depending on scalar combinations of hadronic fields .
this procedure ensures a coherent overlap of the effective hadronic field model and a more _ fundamental _ theoretical description , at least in the baryon sector .
three different parametrizations of the vertices , cases ( a - c ) , and two theoretical sources , namely the qcd sum rules of ref . @xcite and the dirac - brueckner parametrization of ref .
@xcite , have been examined within this scheme .
+ thermodynamical aspects of symmetric nuclear matter evaluated in mfa , are undistinguishable for the instances ( a - c ) within the approach of @xcite , see eqs.([casea]-[casec ] ) , but they differ noticeably from the standard ddhft results .
although the almost perfect coincidence within the ( a - c ) approaches for the energy and pressure , the vertices obtained present significative deviations in the medium to high densities regime .
the results for @xmath214 are similar in the ( b ) and ( c ) instances , as they use parametrizations in terms of variables strongly related in the mfa .
we have given fittings for all the vertices in terms of its corresponding variables .
+ in reference to the theoretical source for the inputs @xmath213 , the parameterization given in @xcite yields observables in agreement with the nuclear matter phenomenology , as expected .
the qcd sum rules instead , do not provide an acceptable description , at least in its present status .
taking the nucleon self - energies proposed in @xcite , we have presented formal expresions for the vertices of a hadronic field model for both isoscalar and isovector channels .
however , examination of the binding energy obtained for symmetric nuclear matter , show a qualitative mismatching .
this assertion is in opposition to the preliminary results of @xcite , where the vector meson contribution to the energy density was not taken into account correctly .
however we have shown that the functional dependence of the isoscalar self - energies given in @xcite is not completely reasonless , so as a redefinition of its numerical parameters produce a satisfactory equation of state .
+ we have also considered in detail the landau parameters for symmetric nuclear matter , under the four assumptions and the parametrization of @xcite .
the hadronic model yields two non - zero landau parameters @xmath215 and @xmath216 , the first one has a common behavior while the second one strongly depends on the vertices definition .
therefore , nuclear observables depending on @xmath215 only , such as isothermal compressibility , first sound velocity and the giant monopole energy , give almost coincident values . those quantities depending on @xmath216
may discriminate among the different approaches , for instance the zero sound modes exhibit very different threshold densities according to the choice made .
+ to sum up , we have compared several formal schemes of merging external physical information into hadronic field models . for this purpose
we have introduced interaction vertices @xmath113 and @xmath114 , which are functionals of the hadronic fields , and we have determined them by else the standard ddhft procedure or by requiring the reproduction of the self - energies used as input . either of them , yield comparable results for the binding energy and pressure of symmetric nuclear matter .
however , physical quantities can be found able to distinguish among the different approaches .
+ qcd sum rules calculations for the nucleon self - energy , do not meet the requirements of the procedure , however further refinements could improve its performance .
shifman , a.i .
vainshtein , and v.i .
zakharov , nucl .
b147 * , 385 ( 1979 ) .
ioffe , nucl . phys . *
b188 * , 317 ( 1981 ) .
e.g. drukarev and e.m .
levin , nucl .
a511 * , 679 ( 1990 ) ; * a532 * , 695 ( 1991 ) ; prog .
* 27 * , 77 ( 1991 ) ; e. g. drukarev , m.g .
ryskin , v.a .
sadovnikova , th .
gutsche , and a. faessler , phys .
c * 69 * , 065210 ( 2004 ) .
e. g. drukarev , m.g .
ryskin , and v.a .
sadovnikova , phys .
rev c * 70 * , 065206 ( 2004 ) .
r. brockmann and h. toki , phys .
lett . * 68 * , 3408 ( 1992 ) .
s. haddad and m. weigel , phys . rev .
c * 48 * , 2740 ( 1993 ) .
h. lenske and c. fuchs , phys . lett .
* b345 * , 355 ( 1995 ) .
s. typel and h.h .
wolter , nucl .
a656 * , 331 ( 1999 ) .
b.d . serot and j.d .
walecka , adv .
nucl . phys .
* 16 * , 1 ( 1986 ) ; b.d .
serot , rep .
phys . * 55 * , 1855 ( 1982 ) .
s. typel , phys .
c * 71 * , 064301 ( 2005 ) .
r. aguirre , phys . lett . *
b611 * , 248 ( 2005 ) .
brown and m. rho , phys .
lett . * 66 * , 2720 ( 1991 ) .
ko , and g.e .
brown , phys .
lett . * 75 * , 4007 ( 1995 ) ; nucl . phys . *
a606 * , 565 ( 1996 ) ; g.q .
brown , and h. sorge , nucl .
a611 * , 539 ( 1996 ) .
t. matsui , nucl .
phys . * a 370 * , 365 ( 1981 ) .
s. nishizaki , h. kurasawa , and t. suzuki , nucl .
a462 * , 687 ( 1987 ) .
k. lim , and c.j .
horowitz , nucl .
a501 * , 729 ( 1989 ) .
caillon , p. gabinski , and j. labarsouque , nucl . phys . *
a696 * , 623(2001 ) ; j. phys .
g * 29 * , 2291 ( 2003 ) . v. greco , m. colonna , m. di toro , and f. matera , phys .
c * 67 * , 015203 ( 2003 ) . | different representations of an effective , covariant theory of the hadronic interaction are examined . for this purpose
we have introduced nucleon - meson vertices parametrized in terms of scalar combinations of hadronic fields , extending the conceptual frame of the density dependent hadronic field theory .
nuclear matter properties at zero temperature are examined in the mean field approximation , including the equation of state , the landau parameters , and collective modes .
the treatment of isospin channels in terms of qcd sum rules inputs is outlined . |
there have been various techniques to evaluate performance of machine learning methods theoretically .
taking lasso @xcite as an example , lasso has been analyzed by nonparametric statistics @xcite , empirical process @xcite , statistical physics @xcite and so on . in general ,
most of these techniques require either asymptotic assumption ( sample number @xmath0 and/or feature number @xmath1 go to infinity ) or various technical assumptions like boundedness of features or moment conditions .
some of them are much restrictive for practical use . in this paper
, we try to develop another way for performance evaluation of machine learning methods with as few assumptions as possible .
an important candidate for this purpose is barron and cover s theory ( bc theory ) , which is one of the most famous results for the minimum description length ( mdl ) principle .
the mdl principle @xcite claims that the shortest description of a given set of data leads to the best hypotheses about the data source . a famous model selection criterion based on the mdl principle was proposed by rissanen @xcite .
this criterion corresponds to a codelength of a two - stage code in which one encodes a statistical model to encode data and then the data are encoded with the model . in this case ,
an mdl estimator is defined as the minimizer of the total codelength of this two - stage code .
bc theory @xcite guarantees that a risk of the mdl estimator in terms of the rnyi divergence @xcite is tightly bounded from above by redundancy of the corresponding two - stage code .
because this result means that the shortest description of the data by the two - stage code yields the smallest risk upper bound , this result gives a mathematical justification of the mdl principle .
furthermore , bc theory holds for finite @xmath0 without any complicated technical conditions .
however , bc theory has been applied to supervised learning only approximately or limitedly .
the original bc theory seems to be widely recognized that it can be applicable to both unsupervised and supervised learning .
though it is not false , bc theory actually can not be applied to supervised learning without a certain condition ( condition [ criticalcondition ] defined in section [ bctheory ] ) .
this condition is critical in a sense that lack of this condition breaks a key technique of bc theory .
the literature @xcite is the only example of application of bc theory to supervised learning to our knowledge .
his work assumed a specific setting , where condition [ criticalcondition ] can be satisfied .
however , the risk bound may not be sufficiently tight due to imposing condition [ criticalcondition ] forcedly , which will be explained in section [ bctheory ] .
another well - recognized disadvantage is the necessity of quantization of parameter space .
barron et al . proposed a way to avoid the quantization and derived a risk bound of lasso @xcite as an example .
however , their idea can not be applied to supervised learning in general .
the main difficulty stems from condition [ criticalcondition ] as explained later .
it is thus essentially difficult to solve .
actually , their risk bound of lasso was derived with fixed design only ( i.e. , essentially unsupervised setting ) .
the fixed design , however , is not satisfactory to evaluate generalization error of supervised learning . in this paper
, we propose an extension of bc theory to supervised learning without quantization in random design cases .
the derived risk bound inherits most of advantages of the original bc theory .
the main term of the risk bound has again a form of redundancy of two - stage code .
thus , our extension also gives a mathematical justification of the mdl principle in supervised learning .
it should be remarked that , however , an additional condition is required for an exact redundancy interpretation .
we also derive new risk and regret bounds of lasso with random design as its application under normality of features .
this application is not trivial at all and requires much more effort than both the above extension itself and the derivation in fixed design cases .
we will try to derive those bounds in a manner not specific to our setting but rather applicable to several other settings .
interestingly , the redundancy and regret interpretation for the above bounds are exactly justified without any additional condition in the case of lasso .
the most advantage of our theory is that it requires almost no assumptions : neither asymptotic assumption ( @xmath3 is also allowed ) , bounded assumptions , moment conditions nor other technical conditions .
especially , it is remarkable that our risk evaluation holds for finite @xmath0 without necessity of boundedness of features though the employed loss function ( the rnyi divergence ) is not bounded .
behavior of the regret bound will be investigated by numerical simulations .
it may be worth noting that , despite we tried several other approaches in order to extend bc theory to supervised learning , we can hardly derive a risk bound of lasso as tight as meaningful by using them .
we believe that our proposal is currently the unique choice that could give a meaningful risk bound .
this paper is organized as follows .
section [ setting ] introduces an mdl estimator in supervised learning .
we briefly review bc theory and its recent progress in section [ bctheory ] .
the extension of bc theory to supervised learning will appear in section [ extension ] .
we derive new risk and regret bounds of lasso in section [ lasso ] .
all proofs of our results are given in section [ proofs ] .
section [ simulation ] contains numerical simulations .
a conclusion will appear in section [ conclusion ] .
suppose that we have @xmath0 training data @xmath4 generated from @xmath5 , where @xmath6 is a domain of feature vector @xmath7 and @xmath8 could be @xmath9 ( regression ) or a finite set ( classification ) according to target problems . here ,
the sequence @xmath10 is not necessarily independently and identically distributed ( i.i.d . ) but can be a stochastic process in general .
we write the @xmath11th component of the @xmath12th sample as @xmath13 . to define an mdl estimator according to the notion of two - stage code @xcite
, we need to describe data itself and a statistical model used to describe the data too . letting @xmath14
be the codelength of the two - stage code to describe @xmath15 , @xmath14 can be decomposed as @xmath16 by the chain rule . since
a goal of supervised learning is to estimate @xmath17 , we need not estimate @xmath18 . in view of the mdl principle , this implies that @xmath19 ( the description length of @xmath20 ) can be ignored .
therefore , we only consider the encoding of @xmath21 given @xmath20 hereafter .
this corresponds to a description scheme in which an encoder and a decoder share the data @xmath20 . to describe @xmath21 given @xmath20
, we use a parametric model @xmath22 with parameter @xmath23 .
the parameter space @xmath24 is a certain continuous space or a union of continuous spaces .
note that , however , the continuous parameter can not be encoded .
thus , we need to quantize the parameter space @xmath24 as @xmath25 . according to the notion of the two - stage code
, we need to describe not only @xmath21 but also the model used to describe @xmath21 ( or equivalently the parameter @xmath26 ) given @xmath20 .
again by the chain rule , such a codelength can be decomposed as @xmath27 here , @xmath28 expresses a codelength to describe @xmath21 using @xmath29 , which is , needless to say , @xmath30 . on the other hand
, @xmath31 expresses a codelength to describe the model @xmath29 itself . note that @xmath32 must satisfy kraft s inequality @xmath33 the mdl estimator
is defined by the minimizer of the above codelength : @xmath34 let us write the minimum description length attained by the two - stage code as @xmath35 because @xmath36 also satisfies kraft s inequality with respect to @xmath21 for each @xmath20 , it is interpreted as a codelength of a prefix two - stage code .
therefore , @xmath37 is a conditional sub - probability distribution corresponding to the two - stage code .
we briefly review barron and cover s theory ( bc theory ) and its recent progress in view of supervised learning though they discussed basically unsupervised learning ( or supervised learning with fixed design ) . in bc theory ,
the rnyi divergence @xcite between @xmath38 and @xmath39 with order @xmath40 @xmath41 is used as a loss function .
the rnyi divergence converges to kullback - leibler ( kl ) divergence @xmath42 as @xmath43 , i.e. , @xmath44 for any @xmath45 .
we also note that the rnyi divergence at @xmath46 is equal to bhattacharyya divergence @xcite @xmath47 we drop @xmath0 of each divergence like @xmath48 if it is defined with a single random variable , i.e. , @xmath49 bc theory requires the model description length to satisfy a little bit stronger kraft s inequality defined as follows .
let @xmath50 be a real number in @xmath51 .
we say that a function @xmath52 satisfies @xmath50-stronger kraft s inequality if @xmath53 where the summation is taken over a range of @xmath54 in its context .
the following condition is indispensable for application of bc theory to supervised learning .
[ criticalcondition ] both the quantized space and the model description length are independent of @xmath20 , i.e. , @xmath55 under condition [ criticalcondition ] , bc theory @xcite gives the following two theorems for supervised learning .
though these theorems were shown only for the case of hellinger distance in the original literature @xcite , we state these theorems with the rnyi divergence .
[ bctheorem1 ] let @xmath50 be a real number in @xmath51 .
assume that @xmath56 satisfies @xmath50-stronger kraft s inequality .
under condition [ criticalcondition ] , @xmath57\label{origbound } \\ & = & e_{\bar{p}_*(x^n , y^n)}\log\frac{p_*(y^n|x^n)}{{\tilde{p}_{2}}(y^n|x^n)}\label{origbound2}\end{aligned}\ ] ] for any @xmath58 $ ] .
[ bctheorem2 ] let @xmath50 be a real number in @xmath51 .
assume that @xmath56 satisfies @xmath50-stronger kraft s inequality . under condition
[ criticalcondition ] , @xmath59 for any @xmath58 $ ] .
since the right side of ( [ origbound2 ] ) is just the redundancy of the prefix two - stage code , theorem [ bctheorem1 ] implies that we obtain the smallest upper bound of the risk by compressing the data most with the two - stage code .
that is , theorem [ bctheorem1 ] is a mathematical justification of the mdl principle .
we remark that , by interchanging the infimum and the expectation of ( [ origbound ] ) , the right side of ( [ origbound ] ) becomes a quantity called `` index of resolvability '' @xcite , which is an upper bound of redundancy .
it is remarkable that bc theory requires no assumption except condition [ criticalcondition ] and @xmath50-stronger kraft s inequality .
however , condition [ criticalcondition ] is a somewhat severe restriction .
both the quantization and the model description length can depend on @xmath20 in the definitions . in view of the mdl principle , this is favorable because the total description length can be minimized according to @xmath20 flexibly .
if we use the model description length that is uniform over @xmath60 in contrast , the total codelength must be longer in general .
hence , data - dependent model description length is more desirable .
actually , this observation suggests that the bound derived in @xcite may not be sufficiently tight .
in addition , the restriction by condition [ criticalcondition ] excludes a practically important case ` lasso with column normalization ' ( explained below ) from the scope of application .
however , it is essentially difficult to remove this restriction as noted in section [ introduction ] .
another concern is quantization .
the quantization for the encoding is natural in view of the mdl principle .
our target , however , is an application to usual estimators or machine learning algorithms themselves including lasso .
a trivial example of such an application is a penalized maximum likelihood estimator ( pmle ) @xmath61 where @xmath62 is a certain penalty .
similarly to the quantized case , let us define @xmath63 that is , @xmath64 note that , however , @xmath65 is not necessarily a sub - probability distribution in contrast to the quantized case , which will be discussed in detail in section [ extension ] .
pmle is a wide class of estimators including many useful methods like ridge regression @xcite , lasso , dantzig selector @xcite and any maximum - a - posteriori estimators of bayes estimation . if we can accept @xmath66 as an approximation of @xmath67 ( by taking @xmath68 ) , we have a risk bound by direct application of bc theory .
however , the quantization is unnatural in view of machine learning application . besides
, we can not use any data - dependent @xmath69 .
barron et al . proposed an important notion ` _ risk validity _ ' to remove the quantization @xcite .
[ riskvalidity ] let @xmath50 be a real number in @xmath51 and @xmath70 be a real number in @xmath71 $ ] . for fixed @xmath20
, we say that a penalty function @xmath72 is risk valid if there exist a quantized space @xmath73 and a model description length @xmath31 satisfying @xmath50-stronger kraft s inequality such that @xmath25 and @xmath31 satisfy @xmath74 where @xmath75 note that their original definition in @xcite was presented only for the case where @xmath76 . here , @xmath77 is the rnyi divergence for fixed design ( @xmath20 is fixed ) .
hence , @xmath78 does not depend on @xmath18 in contrast to the rnyi divergence for random design @xmath79 defined by ( [ renyidiv ] ) .
barron et al . proved that @xmath67 has bounds similar to theorems [ bctheorem1 ] and [ bctheorem2 ] for any risk valid penalty in the fixed design case .
their way is excellent because it does not require any additional condition other than the risk validity .
however , the risk evaluation only for a particular @xmath20 like @xmath80 $ ] is unsatisfactory for supervised learning . in order to evaluate the so - called ` generalization error ' of supervised learning , we need to evaluate the risk with random design , i.e. , @xmath81 $ ] .
however , it is essentially difficult to apply their idea to random design cases as it is .
let us explain this by using lasso as an example .
the readers unfamiliar to lasso can refer to the head of section [ lasso ] for its definition . by extending the definition of risk validity to random design straightforwardly ,
we obtain the following definition . [ riskvalidityinrd ]
let @xmath50 be a real number in @xmath51 and @xmath70 be a real number in @xmath71 $ ] .
we say that a penalty function @xmath72 is risk valid if there exist a quantized space @xmath82 and a model description length @xmath83 satisfying @xmath50-stronger kraft s inequality such that @xmath84 in contrast to the fixed design case , ( [ rv ] ) must hold not only for a fixed @xmath85 but also for all @xmath86 .
in addition , @xmath87 and @xmath83 must be independent of @xmath20 due to condition [ criticalcondition ] .
the form of rnyi divergence @xmath88 also differs from @xmath89 of the fixed design case in general .
let us rewrite ( [ rvrd ] ) equivalently as @xmath90 for short , we write the inside part of the minimum of the left side of ( [ rv2 ] ) as @xmath91 .
we need to evaluate @xmath92 in order to derive risk valid penalties
. however , it seems to be considerably difficult . to our knowledge , the technique used by chatterjee and barron @xcite is the best way to evaluate it , so that we also employ it in this paper .
a key premise of their idea is that taking @xmath93 close to @xmath94 is not a bad choice to evaluate @xmath95 .
regardless of whether it is true or not , this premise seems to be natural and meaningful in the following sense . if we quantize the parameter space finely enough , the quantized estimator @xmath66 is expected to behave almost similarly to @xmath67 with the same penalty and is expected to have a similar risk bound . if we take @xmath96 , then @xmath91 is equal to @xmath97 , which implies that @xmath97 is a risk valid penalty and has a risk bound similar to the quantized case .
note that , however , we can not match @xmath93 to @xmath94 exactly because @xmath93 must be on the fixed quantized space @xmath87 .
so , chatterjee and barron randomized @xmath93 on the grid points on @xmath87 around @xmath94 and evaluate the expectation with respect to it .
this is clearly justified because @xmath98 $ ] .
by using a carefully tuned randomization , they succeeded in removing the dependency of @xmath99 $ ] on @xmath21 .
let us write the resultant expectation as @xmath100 $ ] for convenience .
any upper bound @xmath72 of @xmath101 is a risk valid penalty . by this fact ,
risk valid penalties should basically depend on @xmath20 in general . if not ( @xmath102 ) , @xmath103 must bound @xmath104 , which makes @xmath105 much larger .
this is again unfavorable in view of the mdl principle . in particular
, @xmath101 includes an unbounded term in linear regression cases with regard to @xmath20 , which originates from the third term of the left side of ( [ rv2 ] ) .
this can be seen by checking section iii of @xcite .
though their setting is fixed design , this fact is also true for the random design .
hence , as long as we use their technique , derived risk valid penalties must depend on @xmath20 in linear regression cases .
however , the @xmath106 norm used in the usual lasso does not depend on @xmath20 .
hence , the risk validity seems to be useless for lasso .
however , the following weighted @xmath106 norm @xmath107 plays an important role here .
the lasso with this weighted @xmath106 norm is equivalent to an ordinary lasso with column normalization such that each column of the design matrix has the same norm .
the column normalization is theoretically and practically important .
hence , we try to find a risk valid penalty of the form @xmath108 , where @xmath109 and @xmath110 are real coefficients .
indeed , there seems to be no other useful penalty dependent on @xmath20 for the usual lasso .
in contrast to fixed design cases , however , there are severe difficulties to derive a meaningful risk bound with this penalty .
we explain this intuitively .
the main difficulty is caused by condition [ criticalcondition ] .
as described above , our strategy is to take @xmath93 close to @xmath94 .
suppose now that it is ideally almost realizable for any choice of @xmath111 .
this implies that @xmath91 is almost equal to @xmath112 . on the other hand , for each fixed @xmath94 , the weighted @xmath106 norm of @xmath94 can be arbitrarily small by making @xmath20 small accordingly .
therefore , the penalty @xmath113 is almost equal to @xmath110 in this case .
this implies that @xmath110 must bound @xmath114 , which is infinity in general .
if @xmath56 depended on @xmath20 , we could resolve this problem .
however , @xmath56 must be independent of @xmath20 .
this issue does not seem to be specific to lasso .
another major issue is the rnyi divergence @xmath88 . in the fixed design case ,
the rnyi divergence @xmath89 is a simple convex function in terms of @xmath94 , which makes its analysis easy .
in contrast , the rnyi divergence @xmath88 in case of random design is not convex and more complicated than that of fixed design cases , which makes it difficult to analyze .
we will describe why the non - convexity of loss function makes the analysis difficult in section [ someremarks ] .
the difficulties that we face when we use the techniques of @xcite in the random design case are not limited to them .
we do not explain them here because it requires the readers to understand their techniques in detail .
however , we only remark that these difficulties seem to make their techniques useless for supervised learning with random design .
we propose a remedy to solve these issues in a lump in the next section .
in this section , we propose a way to extend bc theory to supervised learning and derive a new risk bound of lasso .
there are several possible approaches to extend bc theory to supervised learning .
a major concern is how tight a resultant risk bound is .
below , we propose a way that gives a tight risk upper bound for at least lasso .
a key idea is to modify the risk validity condition by introducing a so - called typical set of @xmath20 .
we postulate that a probability distribution of stochastic process @xmath115 is a member of a certain class @xmath116 .
furthermore , we define @xmath117 by the set of marginal distribution of @xmath118 of all elements of @xmath116 .
we assume that we can define a typical set @xmath119 for each @xmath120 , i.e. , @xmath121 as @xmath122 .
this is possible if @xmath123 is stationary and ergodic for example .
see @xcite for detail . for short ,
@xmath124 is written as @xmath125 hereafter .
we modify the risk validity by using the typical set .
[ epsilonriskvalid ] let @xmath126 be real numbers in @xmath51 and @xmath70 be a real number in @xmath71 $ ] .
we say that @xmath72 is @xmath127-risk valid for @xmath128 if for any @xmath129 , there exist a quantized subset @xmath130 and a model description length @xmath131 satisfying @xmath50-stronger kraft s inequality such that @xmath132 note that both @xmath87 and @xmath56 can depend on the unknown distribution @xmath18 .
this is not problematic because the final penalty @xmath69 does not depend on the unknown @xmath18 .
a difference from ( [ rv2 ] ) is the restriction of the range of @xmath20 onto the typical set . from here to the next section , we will see how this small change solves the problems described in the previous section .
first , we show what can be proved for @xmath127-risk valid penalties .
[ mytheorem1 ] define @xmath133 as a conditional expectation with regard to @xmath134 given that @xmath135 .
let @xmath126 be arbitrary real numbers in @xmath51 .
for any @xmath58 $ ] , if @xmath72 is @xmath127-risk valid for @xmath136 , @xmath137 [ mytheorem2 ] let @xmath126 be arbitrary real numbers in @xmath51 . for any @xmath58 $ ] ,
if @xmath72 is @xmath127-risk valid for @xmath136 , @xmath138 a proof of theorem [ mytheorem1 ] is described in section [ mytheorem1proof ] , while a proof of theorem [ mytheorem2 ] is described in section [ mytheorem2proof ] .
note that both bounds become tightest when @xmath76 because the rnyi divergence @xmath79 is monotonically increasing in terms of @xmath70 ( see @xcite for example ) .
we call the quantity @xmath139 in theorem [ mytheorem2 ] ` regret ' of the two - stage code @xmath140 on the given data @xmath15 in this paper , though the ordinary regret is defined as the codelength difference from @xmath141 , where @xmath142 denotes the maximum likelihood estimator . compared to the usual bc theory
, there is an additional term @xmath143 in the risk bound ( [ finriskbound ] ) . due to the property of the typical set
, this term decreases to zero as @xmath144 .
therefore , the first term is the main term , which has a form of redundancy of two - stage code like the quantized case .
hence , this theorem gives a justification of the mdl principle in supervised learning .
note that , however , @xmath145 needs to satisfy kraft s inequality in order to interpret the main term as a conditional redundancy exactly . a sufficient conditions for this was introduced by @xcite and is called ` codelength validity ' .
[ codelengthvalid ] we say that @xmath72 is codelength valid if there exist a quantized subset @xmath73 and a model description length @xmath31 satisfying kraft s inequality such that @xmath146 for each @xmath20 . we note that both the quantization and the model description length on it depend on @xmath20 in contrast to the @xmath127-risk validity .
this is because the fixed design setting suffices to justify the redundancy interpretation .
let us see that @xmath147 can be exactly interpreted as a codelength if @xmath72 is codelength valid .
first , we assume that @xmath8 , the range of @xmath148 , is discrete . for each @xmath20
, we have @xmath149 hence , @xmath145 can be exactly interpreted as a codelength of a prefix code .
next , we consider the case where @xmath8 is a continuous space .
the above inequality trivially holds by replacing the sum with respect to @xmath21 with an integral .
thus , @xmath65 is guaranteed to be a sub - probability density function .
needless to say , @xmath145 can not be interpreted as a codelength as itself in continuous cases .
as is well known , however , a difference @xmath150 can be exactly interpreted as a codelength difference by way of quantization .
see section iii of @xcite for details .
this indicates that both the redundancy interpretation of the fist term of ( [ finriskbound ] ) and the regret interpretation of the ( negative ) second term in the left side of the inequality in the first line of ( [ regretbound ] ) are justified by the codelength validity .
note that , however , the @xmath127-risk validity does not imply the codelength validity and vice versa in general .
we discuss about the conditional expectation in the risk bound ( [ finriskbound ] ) .
this conditional expectation seems to be hard to be replaced with the usual ( unconditional ) expectation .
the main difficulty arises from the unboundedness of the loss function .
indeed , we can immediately show a similar risk bound with unconditional expectation for bounded loss functions .
as an example , let us consider a class of divergence , called @xmath151-divergence @xcite @xmath152 the @xmath151-divergence approaches kl divergence as @xmath153 @xcite .
more exactly , @xmath154 we also note that the @xmath151-divergence with @xmath155 is four times the squared hellinger distance @xmath156 which has been studied and used in statistics for a long time .
we focus here on the following two properties of @xmath151-divergence : * the @xmath151-divergence is always bounded : @xmath157 \label{boundedalpha}\ ] ] for any @xmath45 and @xmath158 . *
the @xmath151-divergence is bounded by the rnyi divergence as @xmath159 for any @xmath45 and @xmath158 .
see @xcite for its proof . as a corollary of theorem [ mytheorem1 ]
, we obtain the following risk bound .
[ coro1 ] let @xmath126 be arbitrary real numbers in @xmath51 . define a function @xmath160 . for any @xmath161 , if @xmath72 is @xmath127-risk valid for @xmath162 and @xmath65 is a sub - probability distribution , @xmath163&\le & \frac{1}{\lambda(\alpha ) } e_{\bar{p}_*}\left [ \log \frac{p_*(y^n|x^n)}{{p_{2}}(y^n|x^n ) } \right]\\ & & + \frac{{p_{\epsilon}^{n}}}{\lambda(\alpha)\beta}\log \frac{1}{{p_{\epsilon}^{n } } } + \frac{(1-{p_{\epsilon}^{n}})}{\lambda(\alpha)(\lambda(\alpha)+\alpha)},\end{aligned}\ ] ] in particular , taking @xmath164 yields the tightest bound @xmath163&\le & \frac{1}{\lambda(\alpha ) } e_{\bar{p}_*}\left [ \log \frac{p_*(y^n|x^n)}{{p_{2}}(y^n|x^n ) } \right]\nonumber \\ & & + \frac{{p_{\epsilon}^{n}}}{\lambda(\alpha)(\lambda(\alpha)+\alpha)}\log \frac{1}{{p_{\epsilon}^{n}}}\nonumber\\ & & + \frac{(1-{p_{\epsilon}^{n}})}{\lambda(\alpha)(\lambda(\alpha)+\alpha)}.\label{alphabound}\end{aligned}\ ] ] its proof will be described in section [ coro1proof ] .
though it is not so obvious when the condition `` @xmath65 is a sub - probability distribution '' is satisfied , we remark that the codelength validity of @xmath72 is its simple sufficient condition . the second and
the third terms of the right side vanish as @xmath144 due to the property of the typical set .
the boundedness of loss function is indispensable for the proof . on the other hand
, it seems to be impossible to bound the risk for unbounded loss functions .
our remedy for this issue is the risk evaluation based on the conditional expectation on the typical set .
because @xmath20 lies out of @xmath119 with small probability , the conditional expectation is likely to capture the expectation of almost all cases . in spite of this fact ,
if one wants to remove the unnatural conditional expectation , theorem [ mytheorem2 ] offers a more satisfactory bound .
note that the right side of ( [ regretbound ] ) also approaches to zero as @xmath144 .
we remark the relationship of our result with kl divergence @xmath165 .
because of ( [ renyikl ] ) or ( [ alphakl ] ) , it seems to be possible to obtain a risk bound with kl divergence .
however , it is impossible because taking @xmath166 in ( [ finriskbound ] ) or @xmath167 in ( [ alphabound ] ) makes the bounds diverge to the infinity .
that is , we can not derive a risk bound for the risk with kl divergence by bc theory , though we can do it for the rnyi divergence and the @xmath151-divergence .
it sounds somewhat strange because kl divergence seems to be related the most to the notion of the mdl principle because it has a clear information theoretical interpretation .
this issue originates from the original bc theory and has been casted as an open problem for a long time .
finally , we remark that the effectiveness of our proposal in real situations depends on whether we can show the risk validity of the target penalty and derive a sufficiently small bound for @xmath168 and @xmath169 .
actually , much effort is required to realize them for lasso . in this section ,
we apply the approach in the previous section to lasso and derive new risk and regret bounds . in a setting of lasso
, training data @xmath170 obey a usual regression model @xmath171 for @xmath172 , where @xmath173 is a true parameter and @xmath174 is a gaussian noise having zero mean and a known variance @xmath175 . by introducing @xmath176 , @xmath177 and an @xmath178 matrix @xmath179^t$ ]
, we have a vector / matrix expression of the regression model @xmath180 .
the parameter space @xmath24 is @xmath181 .
the dimension @xmath1 of parameter @xmath94 can be greater than @xmath0 .
the lasso estimator is defined by @xmath182 where @xmath109 is a positive real number ( regularization coefficient ) .
note that the weighted @xmath106 norm is used in ( [ lassoestimator ] ) , though the original lasso was defined with the usual @xmath106 norm in @xcite . as explained in section [ bctheory ]
, @xmath67 corresponds to the usual lasso with ` column normalization ' .
when @xmath20 is gaussian with zero mean , we can derive a risk valid weighted @xmath106 penalty by choosing an appropriate typical set .
[ erv4lasso ] for any @xmath183 , define @xmath184 where @xmath185 is a gaussian distribution with a mean vector @xmath186 and a covariance matrix @xmath187 .
here , @xmath188 denotes the @xmath11th diagonal element of @xmath187 and @xmath13 denotes the @xmath11th element of @xmath189 .
assume a linear regression setting : @xmath190 let @xmath50 be a real number in @xmath51 and @xmath70 be a real number in @xmath71 $ ] .
the weighted @xmath106 penalty @xmath191 is @xmath127-risk valid for @xmath136 if @xmath192
we describe its proof in section [ erv4lassoproof ] .
the derivation is much more complicated and requires more techniques , compared to the fixed design case in @xcite .
this is because the rnyi divergence is a usual mean square error ( mse ) in the fixed design case , while it is not in the random design case in general .
in addition , it is important for the risk bound derivation to choose an appropriate typical set in a sense that we can show that @xmath125 approaches to one sufficiently fast and we can also show the @xmath127-risk validity of the target penalty with the chosen typical set . in case of lasso with normal design ,
the typical set @xmath119 defined in ( [ tset ] ) satisfies such properties .
let us compare the coefficient of the risk valid weighted @xmath106 penalty with the fixed design case in @xcite .
they showed that the weighted @xmath106 norm satisfying @xmath193 is risk valid in the fixed design case .
the condition for @xmath110 is the same , while the condition for @xmath109 in ( [ rvcondition ] ) is more strict than that of the fixed design case .
we compare them by taking @xmath194 ( the tightest choice ) and @xmath195 in ( [ rvcondition ] ) because @xmath127 can be negligibly small for sufficiently large @xmath0 .
the minimum @xmath109 for the risk validity in the random design case is @xmath196 times that for the fixed design case .
hence , the smallest value of regularization coefficient @xmath109 for which the risk bound holds in the random design is always larger than that of the fixed design case for any @xmath40 but its extent is not so large unless @xmath70 is extremely close to @xmath197 ( see fig .
[ compfixed ] ) . against @xmath70.,width=259 ] next , we show that @xmath125 exponentially approaches to one as @xmath0 increases .
[ expbound ] suppose that @xmath198 independently . for any @xmath183
, @xmath199 see section [ expboundproof ] for its proof . in the lasso case
, it is often postulated that @xmath1 is much greater than @xmath0 .
due to lemma [ expbound ] , @xmath169 is @xmath200 , which also implies that the second term in ( [ finriskbound ] ) can be negligibly small even if @xmath2 . in this sense ,
the exponential bound is important for lasso . combining lemmas [ erv4lasso ] and [ expbound ] with theorems [ mytheorem1 ] and [ mytheorem2 ] , we obtain the following theorem .
[ theorem4lasso ] for any @xmath183 , define @xmath201 assume a linear regression setting : @xmath190 let @xmath50 be a real number in @xmath51 .
for any @xmath58 $ ] , if @xmath202 the lasso estimator @xmath203 in ( [ lassoestimator ] ) has a risk bound @xmath204\le } \nonumber\\ & & \hskip-7mm{e^{n}_{\epsilon}}\biggl[\inf_{\theta\in \theta}\bigg\{\frac{\left(\|y - x\theta\|_2 ^ 2-\|y - x\theta^*\|_2 ^ 2\right)}{2n\sigma^2 } + \mu_1\|\theta\|_{w,1}+\mu_2\bigg\}\biggr]\nonumber\\ & & \hskip-7mm-\frac{p\log\left(1\!-\!2\exp\left(-\frac{n}{2}(\epsilon-\log(1+\epsilon))\right)\right)}{n\beta},\label{riskboundlasso}\end{aligned}\ ] ] and a regret bound @xmath205 with probability at least @xmath206 which is bounded below by @xmath207 with @xmath208 . since @xmath20 and @xmath21 are i.i.d .
now , @xmath209 .
hence , we presented the risk bound as a single - sample version in ( [ riskboundlasso ] ) by dividing the both sides by @xmath0 . finally , we remark that the following interesting fact holds for the lasso case .
[ cv4lasso ] assume a linear regression setting : @xmath190 if @xmath109 and @xmath110 satisfy ( [ rvcondition ] ) , then the weighted @xmath106 norm @xmath210 is codelength valid .
that is , the weighted @xmath106 penalties derived in lemma [ erv4lasso ] are not only @xmath127-risk valid but also codelength valid .
its proof will be described in section [ proofcv4lasso ] . by this fact , the redundancy and regret interpretation of the main terms in ( [ riskboundlasso ] ) and ( [ regretboundlasso ] ) are justified .
it also indicates that we can obtain the unconditional risk bound with respect to @xmath151-divergence for those weighted @xmath106 penalties by corollary [ coro1 ] without any additional condition .
we give all proofs to the theorems , the lemmas and the corollary in the previous section . here , we prove our main theorem .
the proof proceeds along with the same line as @xcite though some modifications are necessary .
define @xmath211 by the @xmath127-risk validity , we obtain @xmath212}\nonumber\\ & \hspace{-9mm}\le & \hspace{-5mm}{e^{n}_{\epsilon}}\big[\exp\big(\beta\max_{{\tilde{\theta}}\in { \widetilde{\theta}}}\big\{f_{\lambda}^{{\tilde{\theta}}}(x^n , y^n)-{\tilde{l}}({\tilde{\theta}}|q_*)\big\}\big)\big]\nonumber\\ & \hspace{-9mm}\le & \hspace{-5mm}\sum_{{\tilde{\theta}}\in { \widetilde{\theta}}(q_*)}{e^{n}_{\epsilon}}\big[\exp\big(\beta\big(f_{\lambda}^{{\tilde{\theta}}}(x^n , y^n)-{\tilde{l}}({\tilde{\theta}}|q_*)\big)\big)\big]\nonumber\\ & \hspace{-9mm}= & \hspace{-5mm}\sum_{{\tilde{\theta}}\in { \widetilde{\theta}}(q_*)}\exp(-\beta{\tilde{l}}\big({\tilde{\theta}}|q_*)\big){e^{n}_{\epsilon}}\big[\exp\big(\beta f_{\lambda}^{{\tilde{\theta}}}(x^n , y^n)\big)\big].\label{intermediate4 } \end{aligned}\ ] ] the following fact is an extension of the key technique of bc theory : @xmath213}\\ & = & \exp\left(\beta d^n_{\lambda}(p_*,p_{{\tilde{\theta}}})\right ) { e^{n}_{\epsilon}}\left[\left(\frac{p_{{\tilde{\theta}}}(y^n|x^n)}{p_*(y^n|x^n)}\right)^{\beta}\right]\\ & \le&\frac{1}{{p_{\epsilon}^{n}}}\exp\left(\beta d^n_{\lambda}(p_*,p_{{\tilde{\theta}}})\right ) e_{\bar{p } _ * } \left[\left(\frac{p_{{\tilde{\theta}}}(y^n|x^n)}{p_*(y^n|x^n)}\right)^{\beta}\right]\\ & = & \frac{1}{{p_{\epsilon}^{n}}}\exp\left(\beta d^n_{\lambda}(p_*,p_{{\tilde{\theta}}})\right)\exp\left(-\beta d_{1-\beta}^n(p_*,p_{{\tilde{\theta}}})\right ) \\ & \le & \frac{1}{{p_{\epsilon}^{n}}}\exp\left(\beta d^n_{\lambda}(p_*,p_{{\tilde{\theta}}})\right)\exp\left(-\beta d^n_{\lambda}(p_*,p_{{\tilde{\theta}}})\right)=\frac{1}{{p_{\epsilon}^{n}}}. \end{aligned}\ ] ] the first inequality holds because @xmath214\ge { p_{\epsilon}^{n}}{e^{n}_{\epsilon}}\left[a\right]$ ] for any non - negative random variable @xmath215 .
the second inequality holds because of the monotonically increasing property of @xmath88 in terms of @xmath70 .
thus , the right side of ( [ intermediate4 ] ) is bounded as @xmath216\\ & & \le \frac{1}{{p_{\epsilon}^{n } } } \sum_{{\tilde{\theta}}\in { \widetilde{\theta}}(q_*)}\exp(-\beta{\tilde{l}}\big({\tilde{\theta}}|q_*)\big)\le \frac{1}{{p_{\epsilon}^{n}}}.\end{aligned}\ ] ] hence , we have an important inequality @xmath217.\label{coroused}\end{aligned}\ ] ] applying jensen s inequality to ( [ coroused ] ) , we have @xmath218\right)\nonumber \\ & \ge & \!\!\!\!\exp\left({e^{n}_{\epsilon}}\left[\beta\left(f_{\lambda}^{\hat{\theta}}(x^n , y^n)-{l}(\hat{\theta}|x^n)\right)\right]\right).\nonumber \end{aligned}\ ] ] thus , we have @xmath219.\end{aligned}\ ] ] rearranging the terms of this inequality , we have the statement .
it is not necessary to start from scratch .
we reuse the proof of theorem [ mytheorem1 ]
. we can start from ( [ coroused ] ) . for convenience ,
we define @xmath220 by markov s inequality and ( [ coroused ] ) , @xmath221 hence , we obtain @xmath222 the proof completes by noticing that @xmath223 for any @xmath20 and @xmath21 . the proof is obtained immediately from theorem [ mytheorem1 ] .
let again @xmath133 denote a conditional expectation with regard to @xmath134 given that @xmath224 .
let further @xmath225 be an indicator function of a set @xmath226 .
the unconditional risk is bounded as @xmath227}\\ & & \hskip-7 mm = e_{\bar{p}_*}[i_{{a_{\epsilon}^{n}}}(x^n){\mathrsfs{d}}^n_{\alpha}(p_*,p_{\hat{\theta } } ) ] \!+\!e_{\bar{p}_*}[(1-i_{{a_{\epsilon}^{n}}}(x^n)){\mathrsfs{d}}^n_{\alpha}(p_*,p_{\hat{\theta } } ) ] \\ & & \hskip-7mm\le { p_{\epsilon}^{n}}{e^{n}_{\epsilon}}[{\mathrsfs{d}}^n_{\alpha}(p_*,p_{\hat{\theta } } ) ] + ( 1-{p_{\epsilon}^{n}})\cdot \frac{4}{1-\alpha^2}\\ & & \hskip-7mm\le \frac{{p_{\epsilon}^{n}}}{\lambda(\alpha)}{e^{n}_{\epsilon}}[d^n_{\lambda(\alpha)}(p_*,p_{\hat{\theta } } ) ] + \frac{(1-{p_{\epsilon}^{n}})}{\lambda(\alpha)(\lambda(\alpha)+\alpha)}\\ & & \hskip-7mm\le \frac{{p_{\epsilon}^{n}}}{\lambda(\alpha)}\left({e^{n}_{\epsilon}}\log \frac{p_*(y^n|x^n)}{{p_{2}}(y^n|x^n ) } + \frac{1}{\beta } \log \frac{1}{{p_{\epsilon}^{n } } } \right)\\ & & + \frac{(1-{p_{\epsilon}^{n}})}{\lambda(\alpha)(\lambda(\alpha)+\alpha)}\\ & & \hskip-7mm= \frac{1}{\lambda(\alpha)}e_{\bar{p}_*}\left [ i_{{a_{\epsilon}^{n}}}(x^n ) \log \frac{p_*(y^n|x^n)}{{p_{2}}(y^n|x^n)}\right ] + \frac{{p_{\epsilon}^{n}}}{\lambda(\alpha)\beta}\log \frac{1}{{p_{\epsilon}^{n}}}\\ & & + \frac{(1-{p_{\epsilon}^{n}})}{\lambda(\alpha)(\lambda(\alpha)+\alpha)}\\ & & \hskip-7mm\le \frac{1}{\lambda(\alpha)}e_{\bar{p}_*}\left [ \log \frac{p_*(y^n|x^n)}{{p_{2}}(y^n|x^n)}\right ] + \frac{{p_{\epsilon}^{n}}}{\lambda(\alpha)\beta}\log \frac{1}{{p_{\epsilon}^{n}}}\\ & & + \frac{(1-{p_{\epsilon}^{n}})}{\lambda(\alpha)(\lambda(\alpha)+\alpha)}. \end{aligned}\ ] ] the first and second inequalities follow from the two properties of @xmath151-divergence in ( [ boundedalpha ] ) and ( [ alpharenyi ] ) respectively .
the third inequality follows from theorem [ mytheorem1 ] because @xmath228 by the assumption .
the last inequality holds because of the following reason . by the decomposition of expectation , we have @xmath229}\\ & & \hskip-7mm= e_{q_*(x^n)}\bigg [ i_{{a_{\epsilon}^{n}}}(x^n)e_{p_*(y^n|x^n)}\bigg [ \log \frac{p_*(y^n|x^n)}{{p_{2}}(y^n|x^n)}\bigg]\bigg ] . \end{aligned}\ ] ] since @xmath230 is a sub - probability distribution by the assumption , the conditional expectation part is non - negative
. therefore , removing the indicator function @xmath231 can not decrease this quantity .
the final part of the statement follows from the fact that taking @xmath76 makes the bound in ( [ finriskbound ] ) tightest because of the monotonically increasing property of rnyi divergence with regard to @xmath70 .
again , we remark that the sub - probability condition of @xmath65 can be replaced with a sufficient condition `` @xmath232 is codelength valid . '' in addition , the sub - probability condition can be relaxed to @xmath233 under which the bound increases by @xmath234 . in this section and the next section ,
we prove a series of lemmas , which will be used to derive risk valid penalties for lasso .
first , we show that the rnyi divergence can be understood by defining @xmath235 in lemma [ renyiview ] .
then , their explicit forms in the lasso setting are calculated in lemma [ bqlemma ] .
[ renyiview ] define a probability distribution @xmath235 by @xmath236 where @xmath237 is a normalization constant .
then , the rnyi divergence and its first and second derivatives are written as @xmath238,\label{firstderivative}\\ \frac{\partial^2d_{\lambda}(p_*,p_{\theta})}{\partial \theta\partial \theta^t}\!\!&=&\!\!-e_{\bar{p}^{\lambda}_{\theta}}\left [ g_{\theta}(x , y ) \right]\nonumber\\ & & -(1-\lambda){\rm var}_{\bar{p}^{\lambda}_{\theta}}\left(s_{\theta}(y|x)\right),\label{hessiangeneral}\end{aligned}\ ] ] where @xmath239 denotes a covariance matrix of @xmath215 with respect to @xmath1 and @xmath240 the normalizing constant is rewritten as @xmath241.\end{aligned}\ ] ] thus , the rnyi divergence is written as @xmath242 next , we calculate the partial derivative of @xmath243 as @xmath244\\ & = & \frac{1-\lambda}{z^{\lambda}_{\theta } } e_{\bar{p}_*}\left[\left(\frac{p_{\theta}(y|x)}{p_*(y|x)}\right)^{1-\lambda}\frac{\partial \log p_{\theta}(y|x)}{\partial \theta } \right]\\ & = & \frac{1-\lambda}{z^{\lambda}_{\theta } } \int q_*(x)p_*(y|x)^{\lambda}p_{\theta}(y|x)^{1-\lambda}s_{\theta}(y|x)dxdy\\ & = & ( 1-\lambda)e_{\bar{p}^{\lambda}_{\theta}}[s_{\theta}(y|x ) ] .
\end{aligned}\ ] ] therefore , the first derivative is @xmath245.\end{aligned}\ ] ] furthermore , we have @xmath246\\ & = & ( 1-\lambda ) \left(s_{\theta}(y|x)-e_{\bar{p}^{\lambda}_{\theta}}[s_{\theta}(y|x)]\right ) .
\end{aligned}\ ] ] hence , @xmath247\right)^t\\ & & + \frac{\partial ^2\log p_{\theta}(y|x)}{\partial \theta\partial\theta^t}\bigg]\\ & \hskip-4mm=&\hskip-3mm - e_{\bar{p}^{\lambda}_{\theta}}\left[\frac{\partial ^2\log p_{\theta}(y|x)}{\partial \theta\partial\theta^t}\right]-(1-\lambda)\\ & \hskip-4mm&\hskip-7mm\cdot e_{\bar{p}^{\lambda}_{\theta}}\bigg[\!\ ! \left(s_{\theta}(y|x)\!-\!e_{\bar{p}^{\lambda}_{\theta}}\left[s_{\theta}(y|x)\right]\right)\left(s_{\theta}(y|x)\!-\!e_{\bar{p}^{\lambda}_{\theta}}[s_{\theta}(y|x)]\right)^t\!\bigg]\\ & \hskip-4mm=&\hskip-3mm - e_{\bar{p}^{\lambda}_{\theta}}\left[\frac{\partial ^2\log p_{\theta}(y|x)}{\partial \theta\partial\theta^t}\right]\!\!-\!(1-\lambda)\mbox{var}_{\bar{p}^{\lambda}_{\theta}}\left(s_{\theta}(y|x)\right).\end{aligned}\ ] ] [ bqlemma ] let @xmath248 if we assume that @xmath249 ( i.e. , linear regression setting ) , @xmath250\bar{\theta},\nonumber\\ \frac{\partial^2d_{\lambda}(p_*,p_{\theta})}{\partial \theta\partial \theta^t } \!\!\!\!&=&\!\!\!\ ! \frac{\lambda}{\sigma^2}e_{q^{\lambda}_{\theta}}[xx^t]-\frac{\lambda}{\sigma^2c}{\rm var}_{q^{\lambda}_{\theta}}\!\left(xx^t\bar{\theta}\right).\label{genhessianlasso}\end{aligned}\ ] ] if we additionally assume that @xmath251 with a non - singular covariance matrix @xmath187 , @xmath252 where @xmath253 by completing squares , we can rewrite @xmath235 as @xmath254 hence , @xmath255 is @xmath256 .
integrating @xmath148 out , we also have @xmath257 when @xmath258 , @xmath259 since @xmath187 is strictly positive definite by the assumption , @xmath260 is non - singular . hence , by the inverse formula ( lemma [ inverseformula ] in appendix ) , @xmath261 therefore , @xmath262 .
the score function and hessian of @xmath263 are @xmath264 using ( [ firstderivative ] ) , the first derivative is obtained as @xmath265\\ & = & -e_{q^{\lambda}_{\theta}}\left[e_{p^{\lambda}_{\theta}}[s_{\theta}(y|x)]\right]\\ & = & -e_{q^{\lambda}_{\theta}}\left[e_{p^{\lambda}_{\theta}}\left[\frac{1}{\sigma^2}x(y - x^t\theta)\right]\right]\\ & = & -e_{q^{\lambda}_{\theta}}\left[\frac{1}{\sigma^2}xx^t(\theta(\lambda)-\theta)\right]\\ & = & \frac{\lambda}{\sigma^2}e_{q^{\lambda}_{\theta}}\left[xx^t\right]{\bar{\theta}}\end{aligned}\ ] ] because @xmath266 . when @xmath267 , @xmath268 from ( [ qsigma ] ) , we have @xmath269 which gives ( [ normalfirstderiv ] ) .
though ( [ normalhessianlasso ] ) can be obtained by differentiating ( [ normalfirstderiv ] ) , we derive it by way of ( [ hessiangeneral ] ) here . to calculate the covariance matrix of @xmath270 in terms of @xmath271
, we decompose @xmath270 as @xmath272 note that the covariance of @xmath273 and @xmath274 vanishes since @xmath275}\\ & = & e_{q^{\lambda}_{\theta}}\left[xx^t(x^t{\bar{\theta}})e_{p^{\lambda}_{\theta}}\left[(y - x^t\theta(\lambda))\right]\right ] = 0 .
\end{aligned}\ ] ] therefore , we have @xmath276 + \frac{\lambda^2}{\sigma^4}\mbox{var}_{q^{\lambda}_{\theta}}\left(xx^t{\bar{\theta}}\right)\\ & = & \frac{1}{\sigma^2}e_{q^{\lambda}_{\theta}}\left[xx^t\right ] + \frac{\lambda^2}{\sigma^4}\mbox{var}_{q^{\lambda}_{\theta}}\left(xx^t{\bar{\theta}}\right)\end{aligned}\ ] ] by ( [ hessiangeneral ] ) combined with ( [ secondderiv ] ) , the hessian of rnyi divergence is calculated as @xmath277-(1-\lambda)\left(\frac{1}{\sigma^2}e_{q^{\lambda}_{\theta}}[xx^t]+\frac{\lambda^2}{\sigma^4}\mbox{var}_{q^{\lambda}_{\theta}}\left(xx^t\bar{\theta}\right)\right)\\ & = & \!\!\!\!\frac{\lambda}{\sigma^2}e_{q^{\lambda}_{\theta}}[xx^t]-\frac{\lambda^2(1-\lambda)}{\sigma^4}\mbox{var}_{q^{\lambda}_{\theta}}\left(xx^t\bar{\theta}\right ) \\ & = & \!\!\!\!\frac{\lambda}{\sigma^2}e_{q^{\lambda}_{\theta}}[xx^t]-\frac{\lambda}{\sigma^2c}\mbox{var}_{q^{\lambda}_{\theta}}\left(xx^t\bar{\theta}\right ) .
\end{aligned}\ ] ] when @xmath258 , @xmath278 is calculated as follows .
note that @xmath279 -({\sigma^{\lambda}_{\theta}}{\bar{\theta}})({\sigma^{\lambda}_{\theta}}{\bar{\theta}})^t.\end{aligned}\ ] ] the @xmath280 element of @xmath281 $ ] is calculated as @xmath282 \!=\!\!\!\!\sum_{j_3,j_4=1}^p\!\!\!{\bar{\theta}}_{j_3}{\bar{\theta}}_{j_4}e_{q^{\lambda}_{\theta}}\left[x_{j_1}x_{j_2}x_{j_3}x_{j_4}\right],\end{aligned}\ ] ] where @xmath283 denotes the @xmath11th element of @xmath7 only here .
thus , we need all the fourth - moments of @xmath284 .
we rewrite @xmath285 as @xmath286 to reduce notation complexity hereafter . by the formula of moments of gaussian distribution , we have @xmath287=s_{j_1j_2}s_{j_3j_4}+s_{j_1j_3}s_{j_2j_4}+s_{j_2j_3}s_{j_1j_4}.\ ] ]
therefore , the above quantity is calculated as @xmath288}\\ & = & \!\!\!\!\sum_{j_3,j_4=1}^p{\bar{\theta}}_{j_3}{\bar{\theta}}_{j_4}(s_{j_1j_2}s_{j_3j_4}+s_{j_1j_3}s_{j_2j_4}+s_{j_2j_3}s_{j_1j_4})\\ & = & { \bar{\theta}}^ts{\bar{\theta}}s_{j_1j_2}+2(s{\bar{\theta}})_{j_1}(s{\bar{\theta}})_{j_2}.\end{aligned}\ ] ] summarizing these as a matrix form , we have @xmath289 & = & ( { \bar{\theta}}^ts{\bar{\theta}})s+2s{\bar{\theta}}(s{\bar{\theta}})^t.\end{aligned}\ ] ] as a result , @xmath290 is obtained as @xmath291 using ( [ sthetabar ] ) , the first and second terms of ( [ varst ] ) are calculated as @xmath292 combining these , @xmath293 using lemma [ bqlemma ] in section [ bqlemmasec ] , we show that the negative hessian of the rnyi divergence is bounded from above .
[ hessianlemma ] assume that @xmath251 and @xmath249 , where @xmath187 is non - singular . for any @xmath294 , @xmath295 where @xmath296 implies that @xmath297 is positive semi - definite . by lemma [ bqlemma ]
, we have @xmath298 for any nonzero vector @xmath299 , @xmath300 by cauchy - schwartz inequality .
hence , we have @xmath301 thus , @xmath302 define @xmath303 for @xmath304 . checking the properties of @xmath305 , we have @xmath306 therefore , @xmath307 . as a result , we obtain @xmath308 we are now ready to derive @xmath127-risk valid weighted @xmath106 penalties . similarly to the rewriting from ( [ rv ] ) to ( [ rv2 ] ) ,
we can rewrite the condition for @xmath127-risk validity as @xmath309 we again write the inside part of the minimum in ( [ erv2 ] ) as @xmath310 . as described in section [ bctheory ]
, the direct minimization of @xmath310 seems to be difficult . instead of evaluating the minimum explicitly
, we borrow a nice randomization technique introduced in @xcite with some modifications .
their key idea is to evaluate not @xmath95 directly but its expectation @xmath311 $ ] with respect to a dexterously randomized @xmath54 because the expectation is larger than the minimum .
let us define @xmath312 , where @xmath313 and @xmath314 .
we quantize @xmath24 as @xmath315 where @xmath316 is a quantization width and @xmath317 is the set of all integers . though @xmath87 depends on @xmath20 in fixed design cases @xcite , we must remove the dependency to satisfy the @xmath127-risk validity as above . for each @xmath94 , @xmath54 is randomized as @xmath318 where @xmath319 and each component of @xmath54 is statistically independent of each other .
its important properties are @xmath320=\theta,\quad \mbox{(unbiasedness)}\nonumber\\ & & e_{{\tilde{\theta}}}[|{\tilde{\theta}}|]=|\theta| , \label{absoluteunbiasedness}\\ & & e_{{\tilde{\theta}}}[({\tilde{\theta}}_j-\theta_j)({\tilde{\theta}}_{j'}-\theta_{j'})]\le i(j = j')\frac{\delta}{w_j^*}|\theta_j| ,
\nonumber \end{aligned}\ ] ] where @xmath321 denotes a vector whose @xmath11th component is the absolute value of @xmath322 and similarly for @xmath323 .
using these , we can bound @xmath311 $ ] as follows .
the loss variation part in ( [ erv2 ] ) is the main concern because it is more complicated than squared error of fixed design cases .
let us consider the following taylor expansion @xmath324 where @xmath325 is a vector between @xmath94 and @xmath54 . the first term in the right side of ( [ type1expansion ] )
vanishes after taking expectation with respect to @xmath54 because @xmath326=0 $ ] . as for the second term , we obtain @xmath327 by lemma [ hessianlemma ] .
thus , expectation of the loss variation part with respect to @xmath54 is bounded as @xmath328\le \frac{\delta n\lambda}{16\sigma^2}\|\theta\|_{w^*,1}. \label{elvp}\ ] ] the codelength validity part in ( [ erv2 ] ) have the same form as that for the fixed design case in its appearance . however , we need to evaluate it again in our setting because both @xmath87 and @xmath56 are different from those in @xcite . the likelihood term is calculated as @xmath329 taking expectation with respect to @xmath54 , we have @xmath330 \!\!\!\!\!&=&\!\!\!\!\ ! \frac{n}{2\sigma^2}e_{{\tilde{\theta}}}\!\left[{{\rm tr}}\!\left(w^2({\tilde{\theta}}-\theta)({\tilde{\theta}}-\theta)^t\right)\right]\nonumber\\ \!\!\!\!\!&\le&\!\!\!\!\!\frac{\delta n}{2\sigma^2}\sum_{j=1}^p\frac{w_j^2}{w_j^*}|\theta_j|,\nonumber\end{aligned}\ ] ] where @xmath331 .
we define a codelength function @xmath332 over @xmath333 .
note that @xmath334 satisfies kraft s inequality .
let us define a codelength function on @xmath335 as @xmath336 by this definition , @xmath56 satisfies @xmath50-stronger kraft s inequality and does not depend on @xmath20 but depends on @xmath337 through @xmath338 . by taking expectation with respect to @xmath54
, we have @xmath339=\frac{\log 4p}{\beta\delta}\|\theta\|_{w^*,1}+\frac{\log 2}{\beta}\end{aligned}\ ] ] because of ( [ absoluteunbiasedness ] ) .
thus , the codelength validity part is bounded above by @xmath340 combining with the loss variation part , we obtain an upper bound of @xmath311 $ ] as @xmath341 since @xmath135 , we have @xmath342 thus , we can bound @xmath311 $ ] by the data - dependent weighted @xmath106 norm @xmath343 as @xmath344}\nonumber\\ & \hskip-3mm\le & \hskip-2mm\frac{\delta n\lambda}{16\sigma^2 } \frac{\|\theta\|_{w,1}}{\sqrt{1-\epsilon } } + \frac{\delta n\sqrt{1+\epsilon}}{2\sigma^2}\sum_{j=1}^p\frac{w_j^{2}}{w_j}|\theta_j| + \frac{\log 4p}{\beta\delta}\frac{\|\theta\|_{w,1}}{\sqrt{1-\epsilon}}\nonumber\\ & & + \frac{\log 2}{\beta}\nonumber\\ & \hskip-3mm=&\hskip-2mm\left ( \frac{\delta n}{\sigma^2 } \left(\frac{\lambda}{16\sqrt{1-\epsilon}}\!+\!\frac{\sqrt{1+\epsilon}}{2}\right)\!+\ ! \frac{\log
4p}{\delta \beta\sqrt{1-\epsilon}}\right ) \|\theta\|_{w,1}\!+\!\frac{\log 2}{\beta}.\nonumber\end{aligned}\ ] ] because this holds for any @xmath316 , we can minimize the upper bound with respect to @xmath345 , which completes the proof .
the main difference of the proof from the fixed design case is in the loss variation part . in the fixed design case , the rnyi divergence @xmath346 is convex in terms of @xmath94 . when the rnyi divergence is convex , the negative hessian is negative semi - definite for all @xmath94
hence , the loss variation part is trivially bounded above by zero . on the other hand
, @xmath347 is not convex in terms of @xmath94 .
this can be intuitively seen by deriving the explicit form of @xmath347 instead of checking the positive semi - definiteness of its hessian . from ( [ qcalc ] ) , we have @xmath348 where @xmath349 is the identity matrix of dimension @xmath1 . prof .
a. r. barron suggested in a private discussion that @xmath237 can be simplified more as follows .
let @xmath350 $ ] be an orthogonal matrix such that @xmath351 . using this ,
we have @xmath352 hence , the resultant @xmath237 is obtained as @xmath353 thus , we have a simple expression of the rnyi divergence as @xmath354 from this form , we can easily know that the rnyi divergence is not convex .
when the rnyi divergence is non - convex , it is unclear in general whether and how the loss variation part is bounded above .
this is one of the main reasons why the derivation becomes more difficult than that of the fixed design case .
we also mention an alternative proof of lemma [ erv4lasso ] based on ( [ andrewsuggestion ] ) .
we provided lemma [ renyiview ] to calculate hessian of the rnyi divergence .
however , the above simple expression of the rnyi divergence is somewhat easier to differentiate , while the expression based on ( [ oldrenyi ] ) is somewhat hard to do it .
therefore , we can twice differentiate the above rnyi divergence directly in order to obtain hessian instead of lemma [ bqlemma ] in our gaussian setting . however , there is no guarantee that such a simplification is always possible in general setting . in our proof , we tried to give a somewhat systematic way which is easily applicable to other settings to some extent .
suppose now , for example , we are aim at deriving @xmath127-risk valid @xmath106 penalties for lasso when @xmath337 is subject to non - gaussian distribution . by ( [ genhessianlasso ] ) in lemma [ bqlemma ]
, it suffices only to bound @xmath290 in the sense of positive semi - definiteness because @xmath355 $ ] is negative semi - definite . in general
, it seemingly depends on a situation which is better , the direct differential or using ( [ genhessianlasso ] ) . in our gaussian setting , we imagine that the easiest way to calculate hessian for most readers is to calculate the first derivative by the formula ( [ firstderivative ] ) and then to differentiate it directly , though this depends on readers background knowledge . for other settings
, we believe that providing lemmas [ renyiview ] and [ bqlemma ] would be useful in some cases . here
, we show that @xmath20 distributes out of @xmath119 with exponentially small probability with respect to @xmath0 .
the typical set @xmath119 can be decomposed covariate - wise as @xmath356 where @xmath357 and the above @xmath358 denotes a direct product of sets . from its definition ,
@xmath359 is subject to a gamma distribution @xmath360 when @xmath361 .
we write @xmath359 as @xmath362 and @xmath363 as @xmath364 ( the index @xmath11 is dropped for legibility ) .
we rewrite the gamma distribution @xmath365 in the form of exponential family : @xmath366 where @xmath367 that is , @xmath368 is a natural parameter and @xmath362 is a sufficient statistic , so that the expectation parameter @xmath369 is @xmath370 $ ] .
the relationship between the variance parameter @xmath364 and natural / expectation parameters are summarized as @xmath371 for exponential families , there is a useful sanov - type inequality ( lemma [ exp2bound ] in appendix ) . using this lemma
, we can bound @xmath372 as follows . for this purpose
, it suffices to bound the probability of the event @xmath373 .
when @xmath374 and @xmath375 , @xmath376 where @xmath377 is the single data version of the kl - divergence defined by ( [ kldiv ] ) .
it is easy to see that @xmath378 for any @xmath379 . by lemma [ exp2bound ] , we obtain @xmath380 hence @xmath125 can be bounded below as @xmath381 the last inequality follows from @xmath382 for any @xmath383 $ ] and @xmath384 . to simplify the bound , we can do more .
the maximum positive real number @xmath385 such that , for any @xmath386 $ ] , @xmath387 is @xmath388 .
then , the maximum integer @xmath389 such that @xmath390 is @xmath391 , which gives the last inequality in the statement . we can prove this lemma by checking the proof of lemma [ erv4lasso ] .
let @xmath392 similarly to the rewriting from ( [ rvrd ] ) to ( [ rv2 ] ) , we can restate the codelength validity condition for @xmath393 as there exist a quantize subset @xmath25 and a model description length @xmath31 satisfying the usual kraft s inequality , such that @xmath394 recall that ( [ rvcondition ] ) is a sufficient condition for the @xmath127-risk validity of @xmath395 , in fact , it was derived as a sufficient condition for the proposition that @xmath393 bounds from above @xmath396= & \!\!\!\!\underbrace{e_{{\tilde{\theta}}}\left [ d^n_{\lambda}(p_*,p_{\theta})- d^n_{\lambda}(p_*,p_{{\tilde{\theta } } } ) \right ] } _ { \mbox{(i ) } } \nonumber\\ & & \!\!\!\!+ \underbrace { e_{{\tilde{\theta}}}\left [ \log \frac{p_{\theta}(y^n|v^n)}{p_{{\tilde{\theta}}}(y^n|v^n ) } + { \tilde{l}}({\tilde{\theta}}|q_*)\right]}_{\mbox{(ii ) } } \label{eh } \end{aligned}\ ] ] for any @xmath397 , @xmath398 , @xmath399 , @xmath23 , where @xmath54 was randomized on @xmath335 and @xmath400 were defined by ( [ quantizationrd ] ) and ( [ ctpen4lasso ] ) , in particular , @xmath131 satisfies @xmath50-stronger kraft s inequality .
recall that @xmath310 is the inside part of the minimum in ( [ erv2 ] ) . here , we used @xmath401 instead of @xmath20 so as to discriminate from the above fixed @xmath20 . to derive the sufficient condition
, we obtained upper bounds on the terms ( i ) and ( ii ) of ( [ eh ] ) respectively , and shown that @xmath402 with @xmath403 is not less than the sum of both upper bounds if ( [ rvcondition ] ) is satisfied .
a point is that the upper bound on the term ( i ) we derived is a non - negative function of @xmath94 ( see ( [ elvp ] ) ) .
hence , if @xmath403 and ( [ rvcondition ] ) hold , @xmath402 is an upper bound on the term ( ii ) , which is not less than @xmath404 now , assume ( [ rvcondition ] ) and let us take @xmath120 given @xmath20 , such that @xmath188 is equal to @xmath405 for all @xmath11 .
then we have @xmath406 , which implies @xmath407 since @xmath408 is determined by @xmath20 and @xmath409 satisfies kraft s inequality , the codelength validity condition holds for @xmath395 .
we investigate behavior of the regret bound ( [ regretboundlasso ] ) . in the regret bound ,
we take @xmath194 with which the regret bound becomes tightest .
furthermore , @xmath109 and @xmath110 are taken as their smallest values in ( [ rvcondition ] ) .
as described before , we can not obtain the exact bound for kl divergence which gives the most famous loss function , the mean square error ( mse ) , in this setting .
this is because the regret bound diverges to the infinity as @xmath43 unless @xmath0 is accordingly large enough .
that is , we can obtain only the approximate evaluation of the mse .
the precision of that approximation depends on the sample size @xmath0 .
we do not employ the mse here but another famous loss function , squared hellinger distance @xmath410 ( for a single data ) .
the hellinger distance was defined in ( [ hellinger ] ) as @xmath0 sample version ( i.e. , @xmath411 ) .
we can obtain a regret bound for @xmath412 by ( [ regretboundlasso ] ) because two times the squared hellinger distance @xmath413 is bounded by bhattacharyya divergence ( @xmath414 ) in ( [ bhattacharyya ] ) through the relationship ( [ alpharenyi ] ) .
we set @xmath415 , @xmath416 and @xmath417 to mimic a typical situation of sparse learning .
the lasso estimator is calculated by a proximal gradient method @xcite . to make the regret bound tight , we take @xmath418 that is close to zero compared to the main term ( regret ) . for this @xmath419 , fig .
[ figprob ] shows the plot of ( [ lowerboundp ] ) against @xmath127 .
we should choose the smallest @xmath127 as long as the regret bound holds with large probability .
our choice is @xmath420 at which the value of ( [ lowerboundp ] ) is @xmath421 .
we show the results of two cases in figs .
[ fig1]-[fig3 ] .
these plots express the value of @xmath414 , @xmath413 and the regret bound that were obtained in a hundred of repetitions with different signal - to - noise ratios ( snr ) @xmath422/\sigma^2 $ ] ( that is , different @xmath175 ) . from these figures and other experiments
, we observed that @xmath423 almost always equaled @xmath414 ( they were almost overlapped ) .
as the sn ratio got larger , then the regret bound became looser , for example , about six times larger than @xmath423 when snr is @xmath424 .
one of the reasons is that the @xmath127-risk validity condition is too strict to bound the loss function when snr is high .
hence , a possible way to improve the risk bound is to restrict the parameter space @xmath24 used in @xmath127-risk validity to a range of @xmath67 , which is expected to be considerably narrower than @xmath24 due to high snr .
in contrast , the regret bound is tight when snr is 0.5 in fig .
finally , we remark that the regret bound dominated the rnyi divergence over all trials , though the regret bound is probabilistic .
one of the reason is the looseness of the lower bound @xmath425 of the probability for the regret bound to hold .
this suggests that @xmath127 can be reduced more if we can derive its tighter bound . ) against @xmath183 when @xmath426 and @xmath418 .
the dotted vertical line indicates @xmath420.,width=259 ] ( bhattacharyya div . ) , @xmath423 ( hellinger dist . ) and the regret bound with @xmath418 in case that snr=1.5.,width=259 ] ( bhattacharyya div . ) , @xmath423 ( hellinger dist . ) and the regret bound with @xmath418 in case that snr=10.,width=259 ] ( bhattacharyya div . ) , @xmath423 ( hellinger dist . ) and the regret bound with @xmath418 in case that snr=0.5.,width=259 ]
we proposed a way to extend the original bc theory to supervised learning by using a typical set . similarly to the original bc theory , our extension also gives a mathematical justification of the mdl principle for supervised learning . as an application
, we derived a new risk and regret bounds of lasso .
the derived bounds still retains various advantages of the original bc theory .
in particular , it requires considerably few assumptions .
our next challenges are applying our proposal to non - normal cases for lasso and other machine learning methods .
[ sanovsection ] the following lemma is a special case of the result in @xcite .
below , we give a simpler proof . in the lemma
, we denote a random variable of one dimension by @xmath427 and denote its corresponding one dimensional variable by @xmath7 .
[ exp2bound ] let @xmath428 where @xmath7 and @xmath94 are of one dimension .
then , @xmath429 where @xmath430 is the expectation parameter corresponding to the natural parameter @xmath94 and similarly for @xmath431 .
the symbol @xmath377 denotes the single sample version of the kl - divergence defined by ( [ kldiv ] ) . in this setting
, the kl divergence is calculated as @xmath432=(\theta-\theta')\eta-\psi(\theta)+\psi(\theta').\ ] ] assume @xmath433 .
because of the monotonicity of natural parameter and expectation parameter of exponential family , @xmath434 by markov s inequality , we have @xmath435}{\exp\left((\theta'-\theta)\eta'\right)}\\ & = & \int \exp(\theta x-\psi(\theta))\exp((\theta'-\theta)x)dx\cdot \exp(-(\theta'-\theta)\eta')\\ & = & \int \exp(\theta ' x-\psi(\theta))dx\cdot \exp(-(\theta'-\theta)\eta')\\ & = & \exp(\psi(\theta'))\exp(-\psi(\theta))\cdot \exp(-(\theta'-\theta)\eta')\\ & = & \exp(-\left((\theta'-\theta)\eta'-\psi(\theta')+\psi(\theta)\right)).\end{aligned}\ ] ] the other inequality can also be proved in the same way .
[ inverseformula ] let @xmath215 be a non - singular @xmath436 matrix .
if @xmath437 and @xmath438 are both @xmath439 vectors and @xmath440 is non - singular , then @xmath441 see , for example , corollary 1.7.2 in @xcite for its proof .
we thank professor andrew barron for fruitful discussion .
the form of rnyi divergence ( [ andrewsuggestion ] ) is the result of simplification suggested by him .
furthermore , we learned the simple proof of lemma [ exp2bound ] from him .
we also thank mr .
yushin toyokihara for his support . | the minimum description length ( mdl ) principle in supervised learning is studied .
one of the most important theories for the mdl principle is barron and cover s theory ( bc theory ) , which gives a mathematical justification of the mdl principle .
the original bc theory , however , can be applied to supervised learning only approximately and limitedly . though barron et al .
recently succeeded in removing a similar approximation in case of unsupervised learning , their idea can not be essentially applied to supervised learning in general . to overcome this issue , an extension of bc theory to supervised learning
is proposed .
the derived risk bound has several advantages inherited from the original bc theory .
first , the risk bound holds for finite sample size .
second , it requires remarkably few assumptions .
third , the risk bound has a form of redundancy of the two - stage code for the mdl procedure .
hence , the proposed extension gives a mathematical justification of the mdl principle to supervised learning like the original bc theory .
as an important example of application , new risk and ( probabilistic ) regret bounds of lasso with random design are derived .
the derived risk bound holds for any finite sample size @xmath0 and feature number @xmath1 even if @xmath2 without boundedness of features in contrast to the past work .
behavior of the regret bound is investigated by numerical simulations .
we believe that this is the first extension of bc theory to general supervised learning with random design without approximation .
lasso , risk bound , random design , mdl principle |
the assumption that the universe is homogeneous on large scales is a fundamental pillar of our current concordance cosmology .
however , it is well known that very large - scale inhomogeneity exists in the form of sheets , voids , and filaments of matter . as structure formation proceeds ,
gravity acts to pile more matter onto the over - dense regions , and simultaneously evacuate the under - dense regions .
models have shown that this phenomenon implies that cosmological observables measured by an observer will vary significantly depending on where that observer is located with respect to these large - scale structures .
in two previous papers @xcite , we investigated large - scale inhomogeneity in the local universe .
these studies were inspired , in part , by the fact that a number of recent cosmological modeling efforts have shown that the observational phenomena typically used to infer an accelerating expansion , namely the dimming " of type ia supernovae , can be equally well fit by invoking a large local under - density instead of dark energy ( e.g. , @xcite ) . other models @xcite have shown that , even in the context of a universe dominated by dark energy , a large local under - density could explain the apparent discrepancy between local measurements of the hubble constant and those inferred by the planck team @xcite . in the study presented here , we investigated the @xmath0band luminosity density ( as a proxy for stellar mass density ) as a function of distance from our position in the local universe . to accomplish this we combined photometry from the ukirt infrared deep sky large area survey ( ukidss - las , @xcite ) and the two micron all sky survey extended source catalog ( 2mass - xsc , @xcite ) with redshifts from the the sloan digital sky survey ( sdss , @xcite ) , the two - degree field galaxy redshift survey ( 2dfgrs , @xcite ) , the galaxy and mass assembly survey ( gama , @xcite ) the two micron redshift survey ( 2mrs , @xcite ) , and the six degree field galaxy redshift survey ( 6dfgs , @xcite ) .
key results from our study @xcite are presented in the figures below . in figure 1a
, we show our measurement of the @xmath0band luminosity function using our compiled sample in three separate redshift ranges ( with residuals showing the normalization offset between these ) . in figure 1b , we show the relative contribution to the total light as a function of absolute magnitude to demonstrate that we are resolving @xmath8 of the total light at @xmath9 . in figure 2a , we show a comparison of our estimate of the total luminosity density in different directions on the sky . in figure 2b , we show our results compared to other studies from the literature and models .
we conclude that the local universe appears under - dense on a scale and amplitude sufficient to introduce significant biases into local measurements of the expansion rate , and , according to some models , to cause what looks like an accelerating expansion even when no dark energy is present .
( a ) the ukidss - las @xmath0band luminosity function split into three redshift ranges : @xmath10 ( red ) , @xmath11 ( green ) , and @xmath12 ( blue ) .
we separate the redshift range @xmath11 to demonstrate that the excess at @xmath13 is not due to the sloan great wall or the other over - densities we observe at higher declination in this redshift range . here
we fit @xmath14 in each redshift range with @xmath15 and @xmath16 fixed ( as described in @xcite ) .
we list the @xmath17 values for each redshift range in the figure , and note that relatively good fits can be obtained using fixed lf shape parameters and letting the normalization ( @xmath14 ) vary as a function of redshift .
residuals are shown relative to the low - redshift normalization . here
we also include our own analysis of the 2m++ catalog ( all sky , @xmath18 galaxies ) .
we have fit the normalization of the lf derived from the 2m++ catalog ( lf and residuals shown as light blue squares ) with the same shape parameters as for the ukidss sample . to extend the faint end of the @xmath19 lf
, we use the gama redshift sample ( @xmath20 , shown as orange squares ) .
( b ) the relative contribution to the total luminosity density ( % per magnitude ) as a function of absolute magnitude .
the blue shaded region shows the range of magnitudes covered by the ukidss sample at @xmath19 .
in orange we show the extra fraction of light resolved by extending the faint end of the lf with the gama sample .
this demonstrates that , at @xmath19 , we are making a robust measurement of the peak of the luminosity density distribution ( occurring at @xmath21 ) and resolving @xmath8 of the total light.,width=453 ] @xmath0band luminosity density as a function of comoving distance .
( a ) our measured @xmath0band luminosity density for the full sample ( red circles ) versus different directions on the sky ( green , blue , and orange circles corresponding to different subregions in the ukidss sample ) .
light blue squares indicate the @xmath0band luminosity density we measure in three different directions ( sdss , 6dfgrs , and 2mr regions ) using the 2m++ all sky catalog compiled by @xcite .
( b ) our measured @xmath0band luminosity density for the full sample ( red circles ) as a function of comoving distance compared with other studies from the literature .
our estimate of the @xmath22 luminosity density from the 2m++ catalog ( sdss and 6dfgrs regions only ) is shown as a light blue circle .
our estimates in three redshift bins for the gama survey only ( @xmath20 , same methods as for the ukidss sample ) are shown as black circles .
the density contrast , @xmath23 , is displayed on the right - hand vertical axis .
the scale of the right - hand axis was established by performing an error - weighted least - squares fit ( for the normalization only , not shape ) of the radial density profile of @xcite ( gray solid curve ) to all the luminosity density data in panel ( b ) .
the dashed curve shows the radial density profile of @xcite .
both @xcite and @xcite claimed these density profiles can provide for good fits to the snia data without dark energy .
the dash - dot curve shows the scale and amplitude of the hubble bubble " type perturbation that @xcite would require to explain the discrepancy between local measurements of the hubble constant and those inferred by planck.,width=453 ] | our recent estimates of galaxy counts and the luminosity density in the near - infrared @xcite indicated that the local universe may be under - dense on radial scales of several hundred megaparsecs .
such a large - scale local under - density could introduce significant biases in the measurement and interpretation of cosmological observables , such as the inferred effects of dark energy on the rate of expansion . in @xcite , we measured the @xmath0band luminosity density as a function of distance from us to test for such a local under - density .
we made this measurement over the redshift range @xmath1 ( radial distances @xmath2 @xmath3 mpc ) .
we found that the shape of the @xmath0band luminosity function is relatively constant as a function of distance and environment .
we derive a local ( @xmath4 @xmath3 mpc ) @xmath0band luminosity density that agrees well with previously published studies . at @xmath5 ,
we measure an increasing luminosity density that by @xmath6 rises to a value of @xmath7 times higher than that measured locally .
this implies that the stellar mass density follows a similar trend . assuming that the underlying dark matter distribution is traced by this luminous matter
, this suggests that the local mass density may be lower than the global mass density of the universe at an amplitude and on a scale that is sufficient to introduce significant biases into the measurement of basic cosmological observables .
at least one study has shown that an under - density of roughly this amplitude and scale could resolve the apparent tension between direct local measurements of the hubble constant and those inferred by planck team .
other theoretical studies have concluded that such an under - density could account for what looks like an accelerating expansion , even when no dark energy is present . |
in astrophysics , the accretion of fluids or matter onto compact objects such as neutron stars or black holes , is an interesting physical process as it describes a scenario which is most likely to explain the high energy output from the active galactic nuclei and quasars .
accretion is the process of capturing the matter by a gravitating object towards its center which leads to increase the mass and angular momentum of the accreting body .
the stars and planets are formed by the process of accretion in dust clouds .
the existence of supermassive black holes at the center of galaxies suggests that such black holes could have been gradually developed through the accretion process .
an accretion disk is developed when dust and gases rotate around a compact object and accumulate into a disk .
however , accretion does not always increase the mass of the accreting body but it could also decrease the mass such as accretion of phantom energy @xcite .
a unique feature of black hole is the presence of an event horizon which acts as a boundary through which infalling fluid disappears .
this may have various implications .
for instance , it provides the inner boundary condition which describes the motion of the fluid and further it helps to avoid the uncertainties regarding the correct boundary conditions . in @xmath1 bondi investigated the accretion process for a star within the newtonian framework @xcite .
after the evolution of the relativistic theory of gravitation given by einstein , it became possible for the astrophysicists to study the accretion process within the relativistic framework .
so , michel was the first one , who investigated the accretion process onto the schwarzschild black hole @xcite .
further , shapiro and teukolsky @xcite also contributed to michel s idea .
later on , babichev et al.@xcite discussed the effect of accretion of a dark energy onto schwarzschild black hole and found that accretion of phantom energy will decrease the mass of the black hole .
debnath @xcite further generalized the babichev s idea and presented a framework of a static accretion onto general static and spherically symmetric black holes .
moreover in the series of recent papers @xcite accretion of a spherically symmetric spacetime with is investigated .
the authors were mainly interested to determine that how the presence of the cosmological constant @xmath2 effects the accretion rate . during the phase transition ,
many topological defects are produced such as cosmic strings , domain walls , monopoles , etc .
monopoles are the three - dimensional topological defects that are formed when the spherical symmetry is broken during the phase transition .
these monopoles have goldstone fields with high energy density that decrease with the distance only as @xmath3 , so that the total energy density is divergent at the large distances @xcite .
this large density suggests that global monopoles can produce the strong gravitational fields .
another interesting point is that the spacetime around a cosmic string is conical globally but locally the spacetime is flat i.e. just a minkowski spacetime .
so we can think about this spacetime as minkowski spacetime but , in case of a global monopole the spacetime is not locally flat . however , globally the spacetime at the equatorial plane is conical having deficit angle as @xmath4 @xcite .
hence , it is evident that these monopoles exerts practically no gravitational force on non - relativistic matter but the space around it has a deficit solid angle and all the light rays are deflected by that angle , independent of the impact parameter @xcite .
it is believed that the charged black holes do not exist in nature , since initial charge will be neutralized quickly @xcite . however ,
these charged black holes may be produced during the gravitational collapse of massive stars particularly magnetized rotating stars surrounded by a magnetosphere having an equal and opposite charge .
the magnetosphere preserves the black hole from a neutralization due to the accretion of charge , so that the black hole remains stable in a typical astrophysical environment having low density .
barriola and vilenkin found the approximate solution of the einstein equations for the static spherically symmetric black hole with a global monopole @xcite .
several other authors have also studied the physical properties of the black holes with global monopoles @xcite . in the absence of the electric charge @xmath5 and the cosmological constant @xmath2 , letelier @xcite obtained the metric representing the black hole spacetime with a spherical mass @xmath6 centered at the origin of the system of coordinates , surrounded by a spherical cloud of strings . the accretion process onto the black hole with a string cloud parameter , which can be interpreted as a global monopole , is examined by ganguly et al.@xcite .
the circular geodesics and accretion disk of a black hole with global monopole is discussed by sun et al @xcite .
recently , the accretion process of a charged black hole in the anti - de - sitter spacetime is investigated by ficek @xcite .
the author calculated the analytical solutions for the isothermal and polytropic fluids and found a way to find sonic points as critical points of a hamiltonian system .
he has also proved the existence of closed trajectories and therefore homoclinic solutions in the phase space of subrelativistic isothermal flows . here , we extend his idea by considering a global monopole charge @xmath0 , which is different from the electric charge @xmath5 in the black hole . in this paper
we investigate the spherically symmetric , steady flow of the most general perfect fluid ( gas ) .
the basic motivation comes from the fact that some features of reissner - nordstrm black hole resemble with the kerr solution as they both share the similar structure of horizons @xcite .
the accretion dynamics for schwarzschild monopole de - sitter black hole with @xmath7 has already been discussed in @xcite .
the paper is organized as follows : in sec ii we derive the metric ( line element ) for the reissner - nordstrm anti - de - sitter ( rn - ads ) black hole with global monopole charge . in sec
iii a general formalism and conservation laws for the accretion process are derived .
the speed of sound at the sonic ( critical ) point is evaluated in sec iv .
next , we have assumed the isothermal equation of state and by choosing different values of state parameter the behaviour of flow for different cases such as ultra - stiff , ultra - relativistic , rotation and sub - relativistic fluids is analysed . in sec vi the general solutions for the polytropic equation of state are obtained as there exists no global ( bondi ) solutions for them .
in the subsequent section , we have calculated the matter accretion rate of the black hole for the isothermal flow and finally we conclude our discussion .
throughout we use the common relativistic notation while the greek indices run through @xmath8 .
the chosen metric signature is @xmath9 , and the geometric units as @xmath10 .
we adopted the metric formalism as given in @xcite .
we assume the general static spherically symmetric metric ansatz of the form @xmath11 the einstein s field equations with the cosmological constant @xmath2 are @xmath12 where @xmath13 and @xmath14 are the energy - momentum tensors for an electromagnetic field and the energy - momentum tensor for the nonminimally coupled global monopole ( gm ) field , respectively . the field eqs .
( [ 2 ] ) have non - vacuum solution when the total energy - momentum tensor @xmath15 does not vanish .
the energy momentum tensor for an electromagnetic field is @xmath16 where @xmath17 , @xmath18 , @xmath5 is the electric charge and the trace of @xmath13 is zero . the lagrangian which describes a global monopole is given by @xmath19 with @xmath20 and @xmath21 is the self - coupling term and @xmath22 is scale of gauge - symmetry breaking @xmath23 gev @xcite . here
@xmath24 is a triplet scalar field with @xmath25 given by @xmath26 with @xmath27 which yields @xmath28 .
hence , the energy - momentum tensor for the global monopole field is @xmath29 , \label{3 - 2}\end{aligned}\ ] ] where @xmath30 .
outside the monopole core it must be assumed that @xmath31 as @xmath32 for @xmath33 to vanish asymptotically .
thus far from the core the stress tensor of the system has the components as @xmath34 the equation of motion for @xmath24 is given by @xmath35 which upon simplification gives @xmath36-\lambda\eta^{2}h(h^{2}-1)&=&0,\end{aligned}\ ] ] where @xmath37 and @xmath38 are the function of @xmath39 due to spherical symmetry and is similar to the equation of motion which we obtain through the energy - momentum conservation i.e. @xmath40 @xcite . a global monopole could readily be added to the scalar field by a general prescription due to dadhich and patel @xcite for any spherically symmetric solution . using ( [ 3 ] ) and ( [ 3 - 2 ] ) together with ( [ 3 - 4 ] ) one can obtain from the einstein field equations ( [ 2 ] ) that @xmath41 where @xmath42 is the newtonian gravitational potential and should satisfy the following constraint equations @xmath43 clearly , eq . ( [ 5b ] ) is the cauchy - euler equation which gives a well known solution @xmath44 where both @xmath45 and @xmath6 are the constants of integration .
these constants are actually the global monopole charge and the mass of the black hole respectively . so using ( [ 6 ] ) the metric function @xmath46 in ( [ 4 ] )
gets the form @xmath47 for @xmath48 , eq .
( [ 7 ] ) reduces to the rn - ads black hole while @xmath49 gives schwarzschild black hole .
it is known that a horizon is the null - hypersurface .
the normal vector to a constant @xmath39 hypersurface is @xmath50 .
clearly @xmath51 therefore , we can say that @xmath52 is the null - hypersurface at @xmath53 and we can determine all the possible horizons from it .
for this , we have a polynomial of degree four by using the eq .
( [ 7 ] ) . when @xmath54 , we get the algebraic equation of the form @xmath55 where @xmath56 and @xmath57 .
this equation can be solved by making it factorizable such that @xmath58 which gives rise to the resolvent cubic equation .
then the quantities @xmath59 and @xmath60 in perfect square have the form given by @xmath61 if the variable @xmath62 is chosen such that @xmath63 i.e. the resolvent cubic with @xmath64 / \lambda^2 $ ] .
thus , we note that @xmath60 is linear and @xmath59 is quadratic in @xmath39 , so each term @xmath65 and @xmath66 is quadratic . therefore
, solving these quadratic formulas one can give all four solutions to the original quartic equation ( [ h - eq-2 ] ) .
the cubic equation ( [ h - eq-4 ] ) can be simplified by making the substitution @xmath67 . in terms of the new variable @xmath68 , eq.([h - eq-4 ] ) then becomes @xmath69 , where @xmath70 and @xmath71 .
now we can solve algebraically the above cubic equation defining the polynomial discriminant @xmath72 . if @xmath73 , one of the roots is real and the other two roots are complex conjugates . if @xmath74 , all roots are real and unequal . in this case , defining @xmath75 , then the real valued solutions of ( [ h - eq-4 ] ) are of the form @xmath76 where @xmath77 and @xmath78 which yields @xmath79 for @xmath80 ( de - sitter spacetime ) and @xmath81 for @xmath82 ( anti - de - sitter spacetime ) .
thus , the value of @xmath0 for @xmath83 has to satisfy the relation @xmath84 , i.e. @xmath85 for both @xmath80 and @xmath82 .
let @xmath86 be a real root of ( [ h - eq-4 ] ) , the four roots of the original quartic ( [ h - eq-2 ] ) are given by the roots of the quadratic equations @xmath87 which are @xmath88 where @xmath89 and @xmath90 .
the polynomial @xmath91 has at most three real and positive roots , which are representing the cauchy horizon , event horizon and cosmological horizon respectively . in order to find the generic case where these three horizons exist ,
the values of the parameters @xmath92 and @xmath0 must be constraint .
when all of those parameters are positive and @xmath93 then three horizons exist , but when @xmath94 and the remaining parameters positive then there exists only two horizons .
further , from curvature invariants given by eq .
( [ e1 ] ) , ( [ e2 ] ) and ( [ e3 ] ) in appendix we see that the curvature is finite everywhere outside the horizon and curvature invariants diverge at @xmath95 . therefore , to remove the singularity at horizon we introduce the eddington - finkelstein ( ef ) coordinate @xcite which is regular at the event horizon , such that @xmath96 this leads us to the following form of metric : @xmath97 dt^{\prime2 } -2\left [ 2 \left ( \beta-\frac{m}{r}\right ) + \frac{q^{2}}{r^{2 } } - \frac{\lambda r^{2}}{3 } \right ] dt^{\prime}dr \nonumber \\ & & + \left[1 - 2\left ( \beta-\frac{m}{r } \right ) -
\frac{q^{2}}{r^{2}}+\frac{\lambda r^{2}}{3}\right ] dr^{2 } + r^2(d\theta^2+\sin^2\theta d\phi^2).\end{aligned}\ ] ] the determinant of the metric defined in ( [ 11 ] ) is @xmath98 and @xmath99 .
now we derive the governing equations for spherical accretion . here , we consider the perfect fluid and analyse the accretion rate and flow of a perfect fluid near rn - ads black hole with a global monopole charge . for this
we define the two basic laws of accretion , i.e. particle conservation and energy conservation .
let @xmath100 be the number of particles and @xmath101 be the four velocity of the fluid , then the particle flux will be @xmath102 . from the law of particle number conservation
there will be no change in the number of particles , their number remain conserved .
in other words , we can say that for this system , the divergence of @xmath103-vector current density vanishes .
mathematically , it means that @xmath104 where @xmath105 is the covariant derivative . on the other hand , the energy momentum tensor for a perfect fluid
is given by @xmath106 with @xmath107 as the energy density and @xmath108 as the pressure and its conservation is given by @xmath109 the bondi - type accretion is steady state and spherically symmetric @xcite , so all the quantities must be function of the radial coordinate only .
furthermore , we are assuming that the fluid is flowing radially in the equatorial plane @xmath110 therefore , @xmath111 . by the normalization condition for @xmath103-velocity
@xmath112 we obtain , @xmath113 which yields @xmath114 at the equatorial plane the continuity equation ( [ 12 ] ) becomes @xmath115 which upon integration yields @xmath116 where @xmath117 is a constant of integration .
for inward flow @xmath118 and so @xmath119 .
similarly , we know that enthalpy is the ratio between density and the total internal energy of the system at constant pressure . therefore ,
by using the first law of thermodynamics , we can define enthalpy as @xmath120 . as
the flow is smooth therefore , eq .
( [ 13 ] ) leads us to @xmath121 further , we assume that the entropy of a fluid moving along a stream line is constant , so the flow must be isentropic @xcite .
hence , eq . ( [ 18 ] ) gives @xmath122 taking the zeroth component of eq .
( [ 20 ] ) we obtain @xmath123 which upon integration gives @xmath124 where @xmath125 is an another constant of integration .
now , these eqs .
( [ 17 ] ) and ( [ 24 ] ) are the main equations which will be used further to analyse the flow of a perfect fluid in the background of rn - ads with global monopole .
sonic point is the critical point where the four - velocity of the moving fluid becomes equal to the local speed of sound therefore , the flow passing through the sonic point has the maximum accretion rate .
if we take constant pressure into account , i.e. @xmath126 then fluid becomes barotropic and the equation of state for barotropic flow can be written as @xcite @xmath127 where @xmath128 is the local speed of sound .
[ 33 ] ) yields @xmath129 . from eqs .
( [ 17 ] ) , ( [ 24 ] ) and ( [ 33 ] ) , we obtain in this paper , the quantities referring to the critical point will be denoted with the subscripted letter
@xmath131 " . at critical point
both sides of eq .
( [ 34 ] ) must be equal to zero .
as @xmath132 so , the local speed of sound at the sonic point becomes @xmath133 where @xmath134 is the value of local speed of sound at sonic point , @xmath135 is the distance of fluid from the black hole at sonic point and @xmath136 is the velocity of the fluid at the sonic point .
then , the rhs of eq .
( [ 34 ] ) at the sonic point given by @xmath137 where @xmath138 , @xmath139 and @xmath140 .
putting eq .
( [ 35 ] ) into ( [ 36 ] ) we obtain the expression for the radial velocity at sonic point as @xmath141 using eqs .
( [ 15 ] ) , ( [ 36 ] ) and ( [ 37 ] ) we get @xmath142 , \label{38 - 2}\end{aligned}\ ] ] which leads to @xmath143 this equation allows us to determine @xmath135 once the speed of sound @xmath144 is known . solving eqs .
( [ 37 ] ) and ( [ 38 - 4 ] ) , we may find the values of @xmath135 and @xmath145 and so we get the critical point as @xmath146 .
isothermal fluids are those fluids which flow at a constant temperature so that the sound speed throughout the accretion process remains constant
. as the fluid is flowing at a very fast speed so it does not take the opportunity to exchange the heat with the surroundings , so it is more likely that our dynamical system is adiabatic .
the equation of state for such fluids is of the form @xmath147 , where @xmath148 is the state parameter such that @xmath149 and @xmath107 is the energy density @xcite . in general , the adiabatic sound speed is defined as @xmath150 . if we compare it to the equation of state , we find @xmath151 . since there is no change in entropy , so @xmath152 where @xmath153 denotes the entropy . by the first law of thermodynamics @xmath154 on integrating it from the sonic point to any point inside the fluid , we obtain @xmath155 for the isothermal equation of state @xmath156 , eq .
( [ 40 ] ) becomes @xmath157 comparing this to enthalpy , we get @xmath158 by using eq .
( [ 41 ] ) in eq .
( [ 24 ] ) we have @xmath159 where @xmath160 . comparing eqs .
( [ 17 ] ) and ( [ 42 ] ) , we get @xmath161 on the other hand , we can define the hamiltonian @xcite as @xmath162 where @xmath163 is the three - dimensional speed for the radial motion in equatorial plane and is defined as @xmath164 .
consequently , we have @xmath165 such that @xmath166 , @xmath167 , @xmath168 . as we are mainly concerned in finding the solutions at the sonic point , therefore eqs .
( [ 37 ] ) and ( [ 38 - 2 ] ) lead to @xmath169 it is worthwhile to generalize this discussion from a single point to the continuous flowing fluid .
so , we will now analyse the behaviour of fluid by choosing different values of state parameter . for instance
, we have @xmath170 ( ultra - stiff fluid ) , @xmath171 ( ultra - relativistic fluid ) , @xmath172 ( radiation fluid ) and @xmath173 ( sub - relativistic fluid ) . for the ultra - stiff fluids ,
energy density becomes equal to the pressure so that the equation of state is of the form @xmath174 , ( i.e. @xmath170 ) . from eqs .
( [ 44a ] ) and ( [ 44b ] ) one can find @xmath175 , i.e. the expression for @xmath135 is identical to the expression for the locations of event horizon , so the sonic point and the event horizon are located at the same place , i.e. @xmath176 . the hamiltonian ( [ 41b ] ) in this case acquires the form : @xmath177 the plot in figure [ f1 ] shows the two types of fluid motion . first is the supersonic accretion flow in the upper half , i.e. the region where @xmath178 and the other is the subsonic accretion flow in the lower region where @xmath179 . ) in phase space and the parameters taken as @xmath170 , @xmath180 , @xmath181 , @xmath182 and @xmath183 .
the black curves show the solution at critical point for which @xmath184 , the red curves show the solution at @xmath185 and the green for @xmath186.,width=302 ] for ultra - relativistic fluids the equation of state is of the form @xmath187 ( i.e. @xmath171 ) . in this case pressure
is always less than the energy density .
thus , from eqs .
( [ 44a ] ) and ( [ 44b ] ) we find @xmath188 which further reduces to the quartic equation @xmath189 where @xmath190 and @xmath191 . on further simplification
, we may find that the values of @xmath135 are same as given in ( [ h - eq-7 - 1 ] ) and ( [ h - eq-7 - 2 ] ) . putting this @xmath135 into ( [ 44a ] ) we get the value @xmath136 and then we solve these two equations to get the two critical points as @xmath146 .
for instance , the relation between @xmath39 and @xmath192 is obtained from eq .
( [ 43 ] ) as @xmath193 where @xmath194 .
the hamiltonian ( [ 41b ] ) reduces to @xmath195 the plot in figure [ f2 ] shows the different behaviours for the motion of a fluid .
the green and red curves are unphysical since they are double valued .
the black curves correspond to the transonic behaviour whereas , the blue and magenta curves show the supersonic behaviour of fluid in the region @xmath196 and subsonic behaviour for @xmath197 , where @xmath198 is the three - dimensional speed for the radial motion in equatorial plane at the sonic point , which is defined as @xmath199 by the eq .
( [ 41bb ] ) . ) in phase space where the parameters are @xmath171 , @xmath180 , @xmath181 , @xmath182 and @xmath183 .
the black curves show the solution for @xmath184 , the red curves show the solution for @xmath200 , the green curves show the solution for @xmath201 , the magenta curve is for @xmath202 and the blue curve is for @xmath203.,width=302 ] the fluid which obeys the equation of state @xmath205 ( i.e. @xmath172 ) is called radiation fluid .
so eqs . ( [ 44a ] ) and ( [ 44b ] )
lead to @xmath206 , i.e. @xmath207 which has the solutions @xmath208 where @xmath209 . using eq .
( [ 51 ] ) in ( [ 43 ] ) , we obtain @xmath210^{3}}{c_3 ^ 6 r_c^{4}}.\end{aligned}\ ] ] from here , we can find the value of @xmath145 , and so we get the critical points @xmath146 .
thus , we can express @xmath192 in terms of @xmath39 by eq .
( [ 44a ] ) .
the hamiltonian ( [ 41b ] ) in this case will be @xmath211 in figure [ f3 ] we see that fluid s motion is supersonic in the region @xmath212 where the curves with blue and magenta color shows the unphysical behaviour of the fluid . ) in phase space with the parameters @xmath172 , @xmath180 , @xmath181 , @xmath182 and @xmath183 .
the black curve shows the solution for @xmath184 , the red curve shows the solution for @xmath202 , the green curve shows the solution for @xmath203 , the magenta curve is for @xmath200 and the blue curve is for @xmath201.,width=302 ] when energy density exceeds the isotropic pressure , we get the sub - relativistic fluids and they obey the equation of state @xmath213 ( i.e. @xmath173 ) . eqs .
( [ 44a ] ) and ( [ 44b ] ) yield @xmath214 which is a quartic equation and has the following form @xmath215 where @xmath216 and @xmath217 . now eq . (
[ 53 ] ) has the same form as eq .
( [ h - eq-2 ] ) so its roots can be found by following the same procedure as given in eqs .
( [ h - eq-7 - 1 ] ) to ( [ h - eq-7 - 2 ] ) . using the value of @xmath135
, we obtain @xmath145 and similarly @xmath218 from ( [ 43 ] ) .
thus , we can express @xmath145 in terms of @xmath135 by eq .
( [ 43 ] ) as @xmath219^{2}}{c_3^{2 } r_c^{2}}.\end{aligned}\ ] ] by repeating the same procedure as in previous cases again we can find an explicit form of solution and critical points @xmath146 .
the hamiltonian ( [ 41b ] ) takes the form @xmath220 figure [ f4 ] shows the motion of the fluid in different regions .
the motion of the fluid is supersonic in the region where @xmath212 and subsonic where @xmath221 whereas , the region of vertical curves show unphysical behaviour of the flow . ) in phase space with parameters as @xmath173 , @xmath180 , @xmath181 , @xmath182 and @xmath183 .
the black curve shows the solution for @xmath184 , the red curve shows the solution for @xmath222 , the green curve shows the solution for @xmath223 , the magenta curve is for @xmath224 and the blue curve is for @xmath225.,width=302 ] in the above cases we have discussed the non - transonic solutions but we are also concerned with the flows which pass through the transonic point .
transonic flows are often considered in a context of spherical accretion which yields to the maximum accretion rate .
so , in figure [ f6 ] we have shown the transonic behaviour of the flow for the above mentioned fluids . since transonic solutions form a closed orbit therefore , the figure demonstrates that it may possible to get homoclinic orbits around a black hole in case of subsonic solutions . .
the other parameters are taken as @xmath180 , @xmath182 , @xmath181 and @xmath183.,width=377 ]
polytropes are self - gravitating gaseous spheres that are , very useful as crude approximation to relativistic fluid models .
the term polytropic " was originally used to describe a reversible process on any open or closed system of gas or any fluid which involves both heat and work transfer , such that a specified combination of properties were maintained constant throughout the process . in such a process , the expression relating the properties of the system throughout the process is called the polytropic path .
there are an infinite number of reversible polytropic paths between two given states ; the most commonly used polytropic path is @xmath226 , where @xmath227 is temperature , @xmath153 is entropy , and @xmath228 is an arbitrary constant which is equal to zero for an adiabatic process .
this path is equivalent to the assumption that the same amount of heat is transferred to the system in each equal temperature increment .
the equation of state , which the fluid has when it follows this path is called the _ polytropic equation of state_. mathematically , it is given as @xmath229 , where @xmath148 and @xmath230 are the constants . by keeping this equation of state in mind , we get the following expression for enthalpy @xcite @xmath231 using eq .
( [ 66 ] ) , eq .
( [ 24 ] ) has taken the form @xmath232 where @xmath233 is the arbitrary constant .
if we take our boundary at infinity ( i.e @xmath234 ) then eq .
( [ 67 ] ) will be @xmath235 where @xmath236 . on comparing eq .
( [ 67 ] ) and eq .
( [ 68 ] ) , we obtain @xmath237 now at the sonic point , the enthalpy can be given as @xmath238 which further on using eq .
( [ 66 ] ) gives rise to @xmath239 since @xmath240 , at the spacial infinity it will be @xmath241 .
on combining eqs .
( [ 70 ] ) and ( [ 71 ] ) we obtain @xmath242^{\frac{1}{\gamma-1}}.\end{aligned}\ ] ] at the critical point @xmath135 , together with ( [ 37 ] ) and ( [ 72 ] ) , eq .
( [ 69 ] ) becomes @xmath243 where @xmath244 note that eq .
( [ 73 ] ) has only one unknown @xmath135 .
so , if we solve this equation with boundary values of @xmath245 and @xmath246 , we obtain the position of critical values of @xmath135 , @xmath247 and @xmath145 .
similarly , as in isothermal case we can compute two critical points @xmath146 in the polytropic case also .
furthermore , eq . ( [ 72 ] ) can also be solved numerically to obtain the function @xmath192 .
the rate of change in the mass of a black hole is called mass accretion rate and is generally represented by @xmath248 . basically , it measures the mass of a black hole per unit time .
it is defined as the area times flux of a black hole at the event horizon . in this section
, we will discuss the effect of radius on the accretion rate .
the general expression to calculate it is @xmath249 @xcite , which refers to relativistic statement of the flux of mass - energy density . in non - relativistic treatment , @xmath250 is defined as the flux of rest - mass density _ or _
@xmath251 , where @xmath252 is the rest - mass density , @xmath253 , @xmath254 is the average mass per baryon and @xmath255 is the baryon density .
then it follows form the perfect fluid energy - momentum tensor ( [ 3 ] ) that @xmath256 @xcite . as our dynamical system
is conserved so we have @xmath257 and @xmath258 . from these conservation equations defined in eqs .
( [ 17 ] ) and ( [ 24 ] ) we obtain @xmath259 where @xmath260 is an arbitrary constant . now assuming the equation of state @xmath261 , the relativistic energy flux ( or continuity ) equation gives @xmath262 integration of this equation yields @xmath263 = - a_1,\end{aligned}\ ] ] where @xmath264 is an integration constant
, @xmath265 is the matter density at infinity and the minus sign is taken because @xmath118 .
if we combine the eq .
( [ 58 ] ) with ( [ 56 ] ) we obtain @xmath266,\end{aligned}\ ] ] where @xmath267 is an arbitrary constant .
if we take the boundary condition at infinity , then the constant @xmath267 becomes @xmath268 , where @xmath269 .
furthermore , on equatorial plane due to the spherical symmetry , the equation of mass flux @xmath270 can also be written as @xmath271 where @xmath272 is an integration constant . if we divide the eq .
( [ 56 ] ) with ( [ 60 ] ) , we get an another useful relation @xmath273 where @xmath274 is a constant such that @xmath275 @xcite . using ( [ 56 ] )
, the black hole s rate of change of mass takes the following form @xmath276 then , it becomes @xmath277 due to the boundary condition at infinity .
this result is valid for any fluid which obeys the equation of state of the form @xmath278 .
so , the accretion rate for the black hole will be @xmath279 at black hole horizon @xmath280 .
thus , evaluating eq.([61 - 2 ] ) at @xmath280 , one can obtain the black hole mass rate for observers at the black hole horizon . .
the other parameters are fixed and are taken as @xmath180 , @xmath290 , @xmath291 , @xmath292 and @xmath293 whereas , the vertical line shows the location of horizon corresponding to monopole parameter in each case.,title="fig:",width=222,height=204 ] [ 1 ] .
the other parameters are fixed and are taken as @xmath180 , @xmath290 , @xmath291 , @xmath292 and @xmath293 whereas , the vertical line shows the location of horizon corresponding to monopole parameter in each case.,title="fig:",width=222,height=204 ] [ 2 ] + .
the other parameters are fixed and are taken as @xmath180 , @xmath290 , @xmath291 , @xmath292 and @xmath293 whereas , the vertical line shows the location of horizon corresponding to monopole parameter in each case.,title="fig:",width=222,height=204 ] [ 1 ] .
the other parameters are fixed and are taken as @xmath180 , @xmath290 , @xmath291 , @xmath292 and @xmath293 whereas , the vertical line shows the location of horizon corresponding to monopole parameter in each case.,title="fig:",width=222,height=204 ] [ 1 ] in figure [ f5 ] we plot the mass accretion rate for different values of @xmath0 .
it can be seen that for all values of @xmath0 the accretion rate increases but when it reaches to the maximum limit , it will decrease .
also , we see the large monopole parameter @xmath0 greatly suppresses the maximum accretion rate .
in @xmath294 , the first solution of the einstein field equations with a global monopole was derived @xcite . here
, we have derived the line element ( metric ) for the reissner - nordstrm-(anti)-de - sitter spacetime with global monopole . by assuming certain conservation laws for the adiabatic system
, we explored the behaviour of fluid .
generally , the equation of state helps us to identify about what kind of fluid is accreting onto black hole . here
, we have focussed mainly on the ultra - stiff , ultra - relativistic , radiation and sub - relativistic fluids .
different kind of fluids with the unique value of state parameter have different kind of evolution onto the black hole .
we have not considered the cases for the vacuum energy or cosmological constant since their accretion does not change that much evolution of the black hole .
further , we have assumed that only a single test fluid accretes onto black hole at a certain time and discussed the isothermal as well as the polytropic flows . though we have derived the general expressions for all type of fluids satisfying the isothermal and polytropic equation of state but for isothermal equation of state we have also find the solutions of the mentioned fluids at the sonic point .
however , in case of polytropic fluids there exists no global solutions @xcite .
furthermore , we have determined the general analytical expression for the mass accretion rate @xmath250 and found that @xmath250 depends on the mass and charge of the black hole and large monopole parameter @xmath0 suppresses the maximum accretion rate frequently .
a number of extensions to our study is possible .
for instance , one may attempt to consider a non - adiabatic system , so that the fluids may have the effects due to viscosity and energy transfer .
it is also interesting to extend this result to the region between the sonic point and boundary .
a similar analysis can be done for the spherically symmetric but non - static geometry of the fluid where the velocity and energy density of the fluid varies with the time and position .
uc was supported by the scientific research projects coordination unit of akdeniz university ( bap ) .
we would like to thank the referees for giving insightful comments to improve this work .
the curvature invariants are given by @xmath295 the above curvature invariants are well defined everywhere except at @xmath95 .
99 jamil m , rashid m a and qadir a 2008 _ _ eur .
j. c__**58 * * 325 jamil m 2009 _ _ eur .
j. c__**62 * * 609 bondi h 1952 _ _ mon . not .
r. astron . soc.__**112 * * 195 michel f c 1972 _ _ astrophys .
space sci.__**15 * * 153 shapiro s l and teukolsky s a 1983 _ black holes , white dwarfs and neutron stars : the physics of compact objects " _ ( wiley , new york ) babichev e o , dokuchaev v i and eroshenko yu n 2013 _ _ phys . usp.__**56 * * 1155 debnath u 2015 _ _ eur .
j. c__**75 * * 129 karkowski j and malec e 2013 _ phys .
d _ * 87 * 044007 mach p and malec e 2013 _ _ phys .
d__**88 * * 084055 mach p , malec e and karkowski j 2015 _ _ phys .
rev . d__**88 * * 084056 bahamonde s and jamil m 2015 _ _ eur .
j. c__**75 *
* 508 barriola m and vilenkin a 1989 _ _ phys . rev .
lett.__**63 * * 341 punsly b 1998 _ _ ap.j.__*498 * 640 bronnikov k a , meierovich b e and podolyak e r 2002 _ _ j . exp . and theor .
phys.__**95 * * 392 yu h 2002 _ _ phys .
rev . d__**65 * * 087502 rahaman f and bhui b c 2005 _ _ fizika b__**14 * * 349 pitelli j p m and latelier p s 2009 _ _ phys . rev . d__**80 * * 104035 letelier p s 1979 _ _ phys .
d__*20 * 1294 ganguly a , ghosh s g and maharaj s d 2014 _ _ phys . rev . d__**90 * * , 064037 xu - dong s , ju - hua c and yon - jiu w 2014 _ _ chin . phys .
b__**23 * * 060401 ficek f 2015 _ _ class .
quantum grav.__**32 * * 235008 poisson e and israel w 1990 _ _ phys
. rev . d__**41 * * 1796 pringle j e and king a b 2007 _ astrophysical flows _ ( cambridge university press ) ahmed a k , azreg - ainou m , faizal m and jamil m 2016 _ _ eur . phys .
j. c__**76 * * 280 ahmed a k , azreg - ainou m , bahamonde s , capozziello s and jamil m 2016 _ _ eur .
j. c__**76 * * 269 | in this paper , we investigate the accretion on the reissner - nordstrm anti - de - sitter black hole with global monopole charge . we discuss the general solutions of accretion using the isothermal and polytropic equations of state for steady state ,
spherically symmetric , non - rotating accretion on the black hole . in the case of isothermal flow , we consider some specific fluids and derive their solutions at the sonic point as well .
however , in case of polytropic fluid we calculate the general expressions only , as there exists no global ( bondi ) solutions for polytropic test fluids .
in addition to this , the effect of fluid on the mass accretion rate are also studied . moreover
, the large monopole parameter @xmath0 greatly suppresses the maximum accretion rate .
+ * keywords * : accretion ; black hole ; cosmological constant ; global monopole charge . |
the set of positive maps acting on a finite dimensional hilbert space is a long - standing subject of mathematical interest . in spite of many efforts ( see @xcite and references therein ) ,
the structure of this set in spaces of arbitrary dimension is still not well understood .
of particular interest are positive maps , which are not completely positive @xcite .
the theorem of jamiokowski implies @xcite that any such map can be represented by an operator , acting on a bi - partite hilbert space , which is not positive , but is block - positive .
non completely positive maps recently attracted a considerable attention of the physics community @xcite .
positive maps have mainly been studied in view of their possible application to characterize quantum entanglement @xcite and in connection to entanglement witnesses @xcite .
an entanglement witness is a hermitian operator @xmath0 such that @xmath1 for any separable state @xmath2 , while negativity of @xmath3 implies that the state @xmath4 is entangled .
note that a hermitian operator @xmath0 may be considered as an observable , so the expectation value @xmath3 can be measured in an experiment @xcite . from a mathematical perspective any entanglement witness is a block positive operator which is not positive . in the present paper we aim to clarify the relation between positive maps and positive polynomials .
definitions and basic information can be found in section [ sec.defs.blockpos ] . in section [ sectestingbp ]
, we explore the link between positive maps and positive polynomials and we address problems related to early contributions on the subject . in particular , we analyze implications of the work of jamiokowski @xcite and show why the results of these papers do not allow one to formulate a conclusive test for positivity of a given map . on the other hand , in some particular cases such results can be obtained . in sections
[ secabc ] and [ sec4x4test ] , we investigate two families of maps and working with the corresponding polynomials we find explicit conditions for positivity .
furthermore , we demonstrate how positive maps relate to the existence of positive polynomials which are not sums of squares and we formulate an open problem concerning entanglement witnesses in @xmath5 dimensional spaces .
let @xmath6 , @xmath7 be finite dimensional spaces over @xmath8 , both equipped with hermitian inner products ( @xmath9 , @xmath10 ) .
let @xmath11 denote the algebra of linear operators on @xmath6 .
we denote with @xmath12 the set of positive elements of @xmath11 .
a linear map @xmath13 is called _ positive _ if and only if it maps elements of @xmath12 to elements of @xmath14 .
it is well known @xcite that the set of positive maps is isomorphic to the set of _ block positive operators _ ( block positive over @xmath8 ) .
therefore , instead of asking whether a given map is positive , in this work we will be concerned with the equivalent question whether the corresponding operator is block positive , so it can serve as an entanglement witness . a hermitian operator @xmath15 on @xmath16 is called _ block positive over @xmath8 _ if it satisfies the following condition , @xmath17 note that condition is not invariant with respect to global unitary transformations on @xmath18 , so this definition depends on the particular form of the decomposition of @xmath18 .
it will also be useful to introduce the concept of block positivity for real linear spaces .
let @xmath19 and @xmath20 be finite dimensional vector spaces over @xmath21 ( @xmath22 , @xmath23 ) .
let @xmath15 be a linear operator on @xmath24 . in analogy to
, we say that @xmath15 is _ block positive over @xmath21 _ if it satisfies @xmath25 condition does not imply symmetry of @xmath15 , but we may always assume that @xmath15 is symmetric because the antisymmetric part of @xmath15 in vanishes .
thence @xmath26 is a symmetric bilinear form on @xmath24 . in index notation
, condition reads @xmath27 where @xmath28 and @xmath29 are the coordinates of @xmath30 and @xmath31 with respect to the orthonormal bases @xmath32 , @xmath33 of @xmath19 , @xmath20 ( resp . ) which we are using .
obviously , is a positivity condition for a real multivariate polynomial of degree @xmath34 . if the polynomial @xmath35 is a _ sum of squares _
( sos ) of some other polynomials @xmath36 , then we must have @xmath37 where the real coefficients @xmath38 ( @xmath39 , @xmath40 ) are arbitrary and the range of the index @xmath41 is finite .
indeed , the polynomials @xmath36 must be homogeneous and of degree @xmath42 .
they can not have terms of the form @xmath43 , neither of the form @xmath44 , since there are no terms @xmath45 nor @xmath46 in the sum @xmath35 .
thus we conclude that if @xmath47 for some polynomials @xmath36 , then @xmath48 .
but looks just like a quadratic form on @xmath24 , written in the product basis @xmath49 . it is tempting to say that implies positive semidefinitness of @xmath15 , but _ this is not true_. nevertheless , a similar result can be proved if we assume that @xmath15 is _ symmetric with respect to partial transpose _ , @xmath50 , where @xmath51 denotes the transposition .
putting this in a different way , @xmath15 should satisfy @xmath52 for any operator @xmath15 being a sos and expressed by eq .
, we may define the following operator @xmath53 , @xmath54 it is easy to see that @xmath55 for all @xmath56 , @xmath57 . in appendix [ apphia ]
we show that this property together with and imply @xmath58 . but
@xmath53 is of the special form , which we did not assume about @xmath15 .
more precisely , @xmath53 is proportional to a sum of a semipositive definite operator @xmath59 with matrix elements @xmath60 and its partial transposition @xmath61 with matrix elements @xmath62 .
we conclude that @xmath47 implies @xmath63 for the operators @xmath15 with the property . a hermitian operator @xmath15 is called _ decomposable _ @xcite iff @xmath64 , where @xmath65 .
when holds , one can easily prove that is equivalent to decomposability of @xmath15 .
thus we arrive at the following conclusion , [ indecompprop ] let @xmath66 be finite dimensional linear spaces over @xmath21 .
let @xmath67 be the set of * block positive , indecomposable * operators on @xmath24 which are symmetric with respect to transposition and * partial transposition*. denote with @xmath68 the set of * positive real polynomials * of the form @xmath69 which are * not sos*. there is a linear isomorphism between @xmath67 and @xmath68 .
the isomorphism in question is @xmath70 .
we still need to show that @xmath71 is one - to - one . to this end
, we assume the equality @xmath72 for some operators @xmath73 .
choose some @xmath74 , @xmath75 .
considering the coefficients at @xmath76 in the two polynomials , we get @xmath77 .
thanks to the partial transpose symmetry of @xmath15 and @xmath59 , we get @xmath78 .
this tells us that @xmath71 is injective . on the other hand , every polynomial of the form @xmath79 is an image by @xmath71 of the partial transpose symmetric operator @xmath80 .
the operator @xmath80 must be an element of @xmath67 for @xmath79 to be an element of @xmath68 ( cf .
the discussion above ) .
thus we conclude that @xmath71 is surjective .
it was demonstrated by choi @xcite and strmer @xcite that there exist positive maps which are not decomposable .
the example by choi can be easily used to show that there exist , by proposition [ indecompprop ] , positive polynomials of the form @xmath69 which are not sos @xcite .
proposition [ indecompprop ] gives a general motivation to investigate block positive operators over @xmath21 on account of their connection to sums of squares .
it may also be expedient to study the real case in order to develop intuitions about block positivity over @xmath8 .
it should , however , be kept in mind that and _ are not the same_. despite an apparent similarity , the block positivity over @xmath81 should not be perceived as a simple generalization of the block positivity over @xmath21 . in general both definitions of block positivity
do not coincide , what can be demonstrated by the following example of a real symmetric matrix , @xmath82.\ ] ] this matrix represents an operator on @xmath83 written in the standard product basis , @xmath84 .
it is easy to show that @xmath15 satisfies inequality , but it does not satisfy condition .
hence the matrix @xmath15 in is block positive over @xmath21 , but is not block positive over @xmath8 .
moreover , when considering operators with unit trace , one can easily show that the set of such block positive operators over @xmath8 is compact whereas block positivity over @xmath21 does not imply compactness . in spite of this basic difference between the two notions of block positivity
, there exist families of matrices for which conditions and turn out to be equivalent see section [ secabc ] .
although the block positivity condition is simple to understand , it does not seem easy to check .
the early papers by jamiokowski @xcite suggest that the problem can be solved effectively . even though this conclusion is in some sense true
, we show a weak point of the argument presented in these papers . for convenience of the reader ,
let us recall the details of the reasoning presented in @xcite .
first , we write condition in index notation , @xmath85 next , we introduce _ blocks _ , @xmath86 we can interpret them simply as matrices or as operators on @xmath6 and @xmath7 , respectively .
block positivity condition can be rewritten as @xmath87 where `` @xmath88 '' refers to semipositive definiteness .
we shall concentrate on the right hand side of .
semipositivity of @xmath89 is equivalent to the following set of inequalities , @xmath90 where @xmath91 is the minor of @xmath89 involving the columns and the rows with the numbers @xmath92 .
it follows from the discussion in @xcite that the functions @xmath93 are homogeneous real polynomials of an even degree in the variables @xmath94 , @xmath95 .
thus is a set of positivity conditions for real homogeneous polynomials of an even degree .
if we could solve these conditions explicitly , we would answer the question whether a given matrix is block positive .
that was the idea presented in @xcite by jamiokowski , who suggested considering @xmath96 ( @xmath97 ) as a polynomial in the variable @xmath98 with coefficients in @xmath99 $ ] .
he obtained positivity conditions for such a polynomial in a disjunctive normal form , @xmath100 where @xmath101\forall_{i , j}$ ] .
because the same procedure could be applied to any of the @xmath102 s ( considered as elements of @xmath103\left[x_{n-1}\right]$ ] ) , it was claimed that the number of variables in can be iteratively reduced so as to yield quantifier free formulas .
the problem with this argument is that eq . does not turn out to be equivalent to @xmath104 so one can not use the procedure iteratively . to the best of our knowledge
, no simple method is known to check positivity of a general multivariate polynomial .
it is in principle possible to eliminate quantifiers @xcite from formulas like @xmath105 , but the outcome involves _ zeros of univariate polynomials of an arbitrary high degree _ , which can not in general be expressed in terms of the coefficients of the polynomials .
the known quantifier elimination procedures are laborious and should not be expected to provide a constructive solution to the problem .
thus we have to conclude this section by repeating the accepted statement that the question of explicit conditions for block positivity remains open .
fortunately , there exist some particular cases for which positivity conditions turn out to be useful in checking block positivity .
let @xmath106 . consider the following family of matrices , @xmath107}=\left[\begin{array}{cccc } \frac{1}{2}&a&0&0\\ \bar a&\frac{1}{2}&b&0\\
0&\bar b&\frac{1}{2}&c\\ 0&0&\bar c&\frac{1}{2}\\ \end{array}\right],\ ] ] which represent operators on @xmath108 .
we are going to test condition using the method suggested in the previous section .
the blocks ( [ blocks ] ) with respect to the subsystem described by @xmath109 are @xmath110.\ ] ] obviously , @xmath111 is semipositive definite for all @xmath112 if and only if @xmath113 .
that is , @xmath114 keeping @xmath115 and @xmath116 constant , we can maximize the absolute value of the term @xmath117 by choosing the phases of @xmath118 and @xmath119 such that the phase of @xmath120 is the same as the phase of @xmath121 .
so done , we see that condition is equivalent to @xmath122 in inequality , we substituted @xmath30 for @xmath115 and @xmath31 for @xmath116 .
it is now easy to see that is the same as @xmath123 we extended the domain of @xmath124 in to @xmath21 , which is permissible because @xmath125 does not increase if we change the sign of @xmath30 or @xmath31 from plus to minus . substituting @xmath126 , @xmath127 in we obtain @xmath128 where @xmath129 , @xmath130 .
condition can be easily solved in the two following situations : * @xmath131 * @xmath132 in the case a ) , condition simplifies to @xmath133 we observe that @xmath134 must hold in order that be true . keeping this in mind , we can rewrite as @xmath135}.\ ] ] where we substituted @xmath136 . as a positivity condition for a quadratic function , can be easily solved explicitly .
together with the condition on @xmath137 , we obtain @xmath138 which is an equivalent form of . in the case
b ) , it is even simpler to get the conditions on @xmath139 , @xmath140 and @xmath141 equivalent to .
we have @xmath142 . either for @xmath143 or for @xmath144 , we obtain @xmath145 and @xmath146 is not changed by the substitution @xmath144 . hence we can rewrite as @xmath147 this is equivalent to @xmath148 , which is easy to solve explicitly in terms of @xmath139 , @xmath140 and @xmath141 .
we get @xmath149 in the case of general @xmath150 , @xmath141 and @xmath151 , condition is equivalent to the following system of four inequalities , @xmath152 which have to be satisfied for all real @xmath153 .
the expression on the left hand side of condition iv ) is a polynomial @xmath154 of degree four in the variable @xmath155 .
this condition means that @xmath156 is nonnegative in the interval @xmath157 $ ] .
given particular values of @xmath150 , @xmath141 and @xmath151 , nonnegativity of @xmath156 in @xmath157 $ ] can be easily checked using the sturm sequences @xcite .
it is also possible to produce general conditions on @xmath150 , @xmath141 , @xmath151 in this way , but the resulting formulas would be too complicated to reproduce them here and not suitable for further analysis .
an analogous problem of block positivity over @xmath21 can also be solved for the family of matrices .
most of the work has already been done above .
we only need to observe that the passage from to is possible also when @xmath150 , @xmath141 , @xmath151 and @xmath118 , @xmath119 are real numbers .
this is true because the maximal value of @xmath158 for fixed @xmath159 , @xmath160 is @xmath161 .
thus the condition turns out to be equivalent to block positivity over @xmath21 of the matrices of the form with @xmath162 .
later analysis follows as in the case b ) discussed above . in this way we arrive at two important conclusions .
firstly , symmetric matrices of the form are block positive over @xmath21 if and only if they are block positive over @xmath8 .
secondly , the block positivity condition takes the form with @xmath163 , @xmath164 . on the other hand , positivity conditions for the family of matrices
are easily obtained , @xmath165 we can compare them with the block positivity condition in a picture .
is equivalent to the block positivity over @xmath21 .
it is assumed here that @xmath162 , but formulas and apply also for @xmath166 complex , provided that @xmath167 with @xmath168 . ]
it is clear from figure [ fig : author ] that the conditions and are not equivalent , and the set of positive matrices of the family forms a proper subset of the set of block positive matrices .
a similar investigation can be performed for a related family of matrices , @xmath169\ ] ] with real parameters @xmath170 and @xmath171 .
the block positivity conditions for @xmath172 can be obtained using the methods presented in this section .
in particular , taking @xmath173 with @xmath166 real , we get a symmetrization of the family , @xmath174 . \ ] ] deriving conditions for positivity and block positivity of the matrices @xmath175 , it turns out that in this case both properties do coincide , unlike in the example discussed above . in the light of proposition [ indecompprop ] , this fact can be understood as a consequence of the following theorem @xcite , [ thm.calderon]let @xmath176 , @xmath177 , @xmath178 @xmath179 that is , any positive biquadratic form in @xmath5 variables is the sum of squares of quadratic forms according to proposition [ indecompprop ] and the caldern s result , _ any operator @xmath15 on @xmath180 which is symmetric with respect to partial transpose and block positive over @xmath21 , is decomposable as well_. more than that , we know from the discussion preceding proposition [ indecompprop ] that @xmath15 can be written in the form @xmath181 with @xmath182
. in the case of @xmath183 , @xmath59 must be of the form with @xmath184 , @xmath185 and @xmath186 .
as can be checked by direct computation , the characteristic polynomials of @xmath172 and @xmath187 are the same .
it follows that @xmath188 , which in turn leads us to the conclusion that the matrix @xmath189 is block positive if and only if it is positive . to explain our observation about @xmath190
, we could also have used the strmer - woronowicz theorem @xcite , which implies that _ an operator @xmath15 on @xmath191 ( or on @xmath192 ) is block positive if and only if it is decomposable_. this suggests a possible connection between the caldern s and the strmer - woronowicz theorems .
on the other hand , the first theorem holds for all @xmath193 ( @xmath176 ) whereas the latter works for @xmath194 only .
the theorem of caldern allows us to find some further implications for the subject of positive maps .
[ thm.either ] let @xmath176 .
either all block positive operators on @xmath195 with * real matrices * are * decomposable * or there exists an operator @xmath15 on @xmath195 with real matrix elements @xmath196 such that @xmath79 is the * sum of squares * of bilinear forms , but @xmath15 is * not decomposable*. let @xmath15 be an operator on @xmath195 with real matrix elements .
if @xmath15 is block positive on @xmath195 , it must be block positive on @xmath193 . from calder ' on s theorem
it follows that @xmath79 is the sum of squares of bilinear forms . if this implies decomposability of @xmath15 , any block positive operator on @xmath195 with real matrix elements is decomposable . if not , there exists an indecomposable operator @xmath15 such that @xmath79 is sos .
both the mutually exclusive possibilities in proposition [ thm.either ] are interesting and it will be good to know which of them is true for which @xmath197 ( of course , the answer is known for @xmath198 - every positive map is decomposable ) .
we hope that stronger results of similar kind can also be obtained and they should give better insights into the structure of positive and indecomposable maps .
we want to illustrate the abstract discussion presented in section [ sectestingbp ] with a concrete example . to that aim , following @xcite , we derive sufficient and necessary conditions for an arbitrary operator @xmath15 on @xmath191 to be block positive let the matrix elements of @xmath15 be @xmath196 ( @xmath200 ) .
the blocks with respect to the first subsystem have the matrix elements @xmath201 .
positivity of @xmath202 is equivalent to the requirements that @xmath203 and @xmath204 .
nonnegativity of the trace of @xmath202 for all @xmath205 means that @xmath206 obviously , is a positivity condition for a quadratic form on @xmath207 and we can write it explicitly as @xmath208 the expression for the determinant of @xmath202 reads @xmath209 where we substituted @xmath30 for @xmath210 , @xmath211 for @xmath212 and we introduced @xmath213 the @xmath214 s are homogeneous polynomials in the matrix elements @xmath196 .
it is easy to see that non - negativity of for all @xmath215 is equivalent to @xmath216 thus we showed that in the case of a symmetric matrix @xmath15 of order @xmath34 condition is equivalent to plus .
the inequalities are explicit conditions on the matrix elements @xmath196 , but in we need some additional work to dispose of the quantifier @xmath217 .
there is no single method of doing it , but the one which seems most economical to us is by using the following theorem @xcite , [ thm.sturm]let @xmath218 be a real univariate polynomial with no multiple roots in @xmath21 . let @xmath219 be the first derivative of @xmath220 . define @xmath221 where @xmath222 is the remainder obtained when dividing @xmath223 by @xmath224 .
define @xmath225 as the number of sign changes in the sequence @xmath226 with zeros skipped .
assume @xmath227 , @xmath228 , @xmath229 and @xmath230 .
the number of zeros of @xmath231 in the interval @xmath232 equals @xmath233 , @xmath234 the sequence of functions is the same as in the euclid s algorithm applied to @xmath220 and @xmath235 .
when the signs are changed as in , the sequence is called the _ sturm sequence _ of @xmath220 . we know that @xmath236 must terminate at some @xmath237 , which is the greatest common divisor of @xmath220 and @xmath235 .
if we go to the limits @xmath238 , @xmath239 in theorem [ thm.sturm ] , we easily obtain the number of real roots of @xmath220 , [ cor.sturminfty ] let @xmath218 be a real univariate polynomial with no multiple roots in @xmath21 and @xmath219 - its first derivative .
let @xmath240 ( @xmath241 ) be defined like in ( [ fml.defoffn.1 ] ) and assume @xmath242 where @xmath243 , @xmath244 @xmath245
. denote with @xmath246 the number of sign changes in the sequence @xmath247 and with @xmath248 the number of sign changes in @xmath249 the number of real zeros of @xmath220 equals @xmath250 , @xmath251 let us take for @xmath220 the polynomial @xmath252 which appears in .
we shall now assume that it has no multiple roots in @xmath21 .
then we can use corollary [ cor.sturminfty ] to check positivity of @xmath220 .
the sequence @xmath253 consists of at most five polynomials , @xmath254 if we make an additional _ normality _ assumption , which says that the degrees of @xmath255 drop one by one in the successive lines of , it is easy to write down positivity conditions for @xmath220 , @xmath256 the expressions for @xmath257 , @xmath258 and @xmath259 can also be easily obtained in the present situation .
we get @xmath260 where @xmath261 according to , the normality assumption is equivalent to @xmath262 .
if these conditions hold , we can rewrite as @xmath263 this is an explicit condition for @xmath220 to be positive . of course
, it was obtained under the assumption that @xmath264 are normal .
nevertheless , we can use as a starting point for an all - purpose nonnegativity test for polynomials of degree less or equal four . indeed , suppose that the sequence @xmath264 is not normal .
that is , at least one of the numbers @xmath265 happens to be zero .
if @xmath266 , the nonnegativity question becomes trivial .
we get that @xmath267 is nonnegative if and only if @xmath268 the case in which @xmath269 but @xmath270 can be analyzed using a little more sophisticated techniques ( see appendix [ apppol4b ] ) .
all in all , we arrive at the following nonnegativity conditions for @xmath220 , @xmath271 where @xmath272 means @xmath273 . we can write @xmath272 explicitly as @xmath274 where @xmath275 obviously , conditions and together with the definitions , , , , , , , , , and provide us with a method to test block positivity over @xmath21 of @xmath199 matrices .
we see that lengthy calculations are involved , even though the studied example is the simplest possible one .
it is also clear that the iterative procedure proposed in @xcite could not work with conditions like , let alone .
we have re - examined the method @xcite of establishing positivity of a map with help of multivariate polynomials and we conclude that in the general case this problem remains open .
the same can be said about the equivalent problem of checking whether a given operator acting on a composite hilbert space is block positive .
nevertheless , for certain family of operators checking the positivity of the associated polynomials allowed us to find concrete criterion for block positivity .
such concrete examples are provided in sections [ secabc ] and [ sec4x4test ] . by giving the example , we touched upon the relation between the block positivity conditions over @xmath8 and over @xmath21 .
we also outlined connections between block positivity , indecomposability and the sums of squares ( propositions [ indecompprop ] and [ thm.either ] , theorem [ thm.calderon ] ) .
proposition [ thm.either ] opens a discussion about the two mutually exclusive possibilities concerning indecomposable maps on @xmath195 ( cf .
section [ secabc ] ) . finally , we tried to show that polynomials , which have been thoroughly studied by mathematicians and engineers , may deserve more respect of physicists working on quantum information or on open quantum systems .
in particular , the separability problem itself can be formulated as a set of polynomial equalities @xcite .
techniques like the calculation of a grbner basis of an ideal are widely used to solve polynomial equations and they could be of importance in physical problems like the separability problem .
the authors wish to thank andrzej jamiokowski for very useful comments and for providing them with copies of his early papers @xcite .
we also acknowledge helpful remarks by dariusz chruciski and pawe horodecki .
the work was financially supported by the special grant number dfg - sfb/38/2007 of polish ministry of science and higher education and the european research project cocos .
consider the symmetric bilinear forms @xmath276 , @xmath277 . we know that @xmath278 for arbitrary @xmath56 , @xmath57 . from
and it follows that @xmath279 , @xmath280 are symmetric with respect to partial transposition , @xmath281 let us choose @xmath56 and consider the following maps , @xmath282 , @xmath283 . from
and we know that @xmath284 , @xmath285 are symmetric bilinear forms on @xmath20 . as a consequence of @xmath278 , @xmath286 for arbitrary @xmath57 . hence the quadratic forms corresponding to @xmath279 and @xmath280 are equal .
this implies @xmath287 , so we get @xmath288 now we consider the maps @xmath289 , @xmath290 . from the symmetry of @xmath279 , @xmath280 and the properties ,
, we see that @xmath291 , @xmath292 are symmetric bilinear forms on @xmath19 . as a consequence of , @xmath293 for all @xmath56 .
this implies @xmath294 for arbitrary @xmath295 . in this way we get @xmath296 , which is the same as @xmath297 of course ,
implies @xmath298 .
our aim is to figure out all the sign configurations of @xmath265 such that they meet the constraints @xmath299 and they correspond to nonnegative polynomials @xmath300 .
we also have to check that the remaining sign configurations can never give a nonnegative @xmath220 .
of course , @xmath301 implies that @xmath302 be negative for some @xmath30 , so we only need to consider @xmath303 positive .
first we show that @xmath304 can not happen for a nonnegative @xmath220 .
suppose @xmath304 .
we know that @xmath305 or @xmath306 .
let us first consider @xmath305 . because @xmath307 and @xmath308 , we can increase @xmath309 by @xmath310 and get @xmath311 for sure .
if @xmath312 turns out to be zero after this operation , we additionally increase @xmath313 by @xmath314 , which must give us @xmath315 because @xmath316 where @xmath317 does not depend on @xmath313 .
the numbers @xmath318 , @xmath319 can be made arbitrarily small , so as not to influence the sign of @xmath320 .
hence we see that @xmath321 has a normal sturm sequence and it does not satisfy because the @xmath320 corresponding to @xmath321 is positive . but @xmath322 implies @xmath323 , so @xmath220 can not be nonnegative .
we conclude that @xmath324 is impossible for @xmath308 , @xmath304 and @xmath305 . for @xmath308 , @xmath304 and @xmath306
, we only need to increase @xmath220 by a sufficiently small @xmath319 to get to the conclusion @xmath323 .
our observations mean that @xmath304 always implies @xmath323 .
let us now consider the polynomials @xmath220 for which conditions @xmath325 are satisfied .
if @xmath326 vanishes , we can get @xmath311 by increasing @xmath309 ( @xmath327 ) .
if @xmath312 turns out to be @xmath328 afterwards , any change of @xmath313 ( @xmath329 ) will give us @xmath330 ( cf .
the discussion above ) .
we can take @xmath318 and @xmath319 arbitrarily small , which allows us to avoid changing the sign of @xmath320 .
after all , we get a polynomial @xmath321 which has a normal sturm sequence and it is positive since @xmath331 for the corresponding @xmath303 , @xmath326 and @xmath320 . because @xmath318 and @xmath319 can be arbitrarily small , we see that @xmath220 is a pointwise limit of a sequence of positive polynomials .
hence @xmath220 is nonnegative .
the same conclusion can be drawn for @xmath308 , @xmath332 , @xmath306 and @xmath333 , so we should add to our set of non - negativity conditions .
we can write and as a single condition , @xmath334 the only situation which is left to analyze is that of @xmath335 . to that end , let us write @xmath320 as a polynomial in @xmath313 , @xmath336 where @xmath337 because of the assumption @xmath308 , we know that @xmath320 is not constant with respect to @xmath313 .
the idea now is to infinitesimally increase @xmath313 and see what the outcome is .
if @xmath320 becomes positive , we conclude that @xmath323 .
if it turns out to be negative ( we denote this with @xmath338 ) , we go back to the initial values of @xmath214 and ask about the signs of @xmath326 , @xmath312 . if @xmath311 , we choose @xmath314 so small that @xmath311 holds when we increase @xmath313 by @xmath319 . then is true for @xmath339 , so @xmath339 is nonnegative . since the @xmath319 in @xmath339
can be made arbitrarily small , we get @xmath324 . if @xmath305 , we increase @xmath313 by @xmath319 to get @xmath333 and then we increase @xmath309 by @xmath318 so as to get @xmath311 and not to violate @xmath304 .
after that holds for @xmath321 and again we get to the conclusion that @xmath324
. therefore we can add @xmath340 to our list of non - negativity conditions for @xmath220 .
now we only need to analyze the case @xmath341 to finish our work .
if @xmath342 , we choose @xmath314 so small that the @xmath312 corresponding to @xmath339 is also positive .
then @xmath343 and we get @xmath324 . the case @xmath306 is also simple to analyze . because @xmath344 , increasing @xmath313 causes @xmath345 to become positive , so we get @xmath346 for @xmath339 and again this leads us to @xmath324 .
thus we can add @xmath347 to our nonnegativity conditions for @xmath220 .
it is convenient to write and in a single formula , @xmath348 because of the particular form of @xmath320 , we explicitly write the condition @xmath349 as @xmath350 since condition @xmath308 implies that @xmath351 . c. michaux , a. ozturk , _ quantifier elimination following muchnik _ , universit de mons - hainaut , institute de mathmatique , preprint 10 , april 11 , 2002 , http://math.umh.ac.be/preprints/src/ozturk020411.pdf | we link the study of positive quantum maps , block positive operators , and entanglement witnesses with problems related to multivariate polynomials . for instance , we show how indecomposable block positive operators relate to biquadratic forms that are not sums of squares .
although the general problem of describing the set of positive maps remains open , in some particular cases we solve the corresponding polynomial inequalities and obtain explicit conditions for positivity .
= 2 mm |
conjunction fallacy was first empirically documented by tversky and kahneman ( 1982 , 1983 ) through a now renowned experiment in which subjects are presented with a description of someone called linda " : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ linda is 31 years old , single , outspoken , and very bright .
she majored in philosophy . as a student , she was deeply concerned with issues of discrimination and social justice , and also participated in anti - nuclear demonstrations . " _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
then , subjects are shown a list of 8 possible outcomes describing her present employment and activities , and are asked to rank the propositions by representativeness or probability .
two items were specifically tested : * linda is a bank teller " , * linda is a bank teller and is active in the feminist movement " .
empirical results show that most people judge ( 2 ) more probable than ( 1 ) . in the framework of classical probabilities , this is a fallacy the conjunction fallacy , since a conjunction can not be more probable than one of its components .
if linda being active in the feminist movement is denoted by @xmath0 and linda being a bank teller by @xmath1 , then @xmath2 should classically prevail .
the conjunction fallacy has been shown to be particularly robust under various variations of the initial experimental protocol ( cf . tversky and kahneman 1982 , 1983 , gigerenzer 1996 , kahneman and tversky 1996 , hertwig 1997 , hertwig and chase 1998 , hertwig and gigerenzer 1999 , mellers et al .
2001 , stolartz - fantino et al .
2003 , bonini et al .
2004 , tentori et al .
2004 , hertwig et al . 2008 , moro 2009 , kahneman 2011 , erceg and galic 2014 ; for a review , cf .
moro 2009 ) .
it has been observed in other cases than the linda story , about topics like sports , politics , or natural events , and in scenarios in which the propositions to be ranked are not preceded with a description .
the fallacy also persists when the experimental setting is changed , e.g. in between subjects " experiments in which ( 1 ) and ( 2 ) are presented to different subjects only .
semantic and syntactic aspects have also been discussed , in relation with possible misunderstandings , like the implicit meaning of the words probability " and and " .
careful experiments show that the conjunction fallacy persists .
the conjunction fallacy questions the fact that classical probability theory can be used to describe human judgment and decision making , and it can also be viewed as a challenge to the definition of what a rational judgment is .
thus , it is no surprise that the conjunction fallacy has been the subject of a big amount of research ( tentori and crupi 2012 give the number of a hundred papers devoted to it ) .
it has interested psychologists , economists and philosophers alike .
for instance , behavioral economists have looked at the consequences of the fallacy for understanding real life economic behavior , measuring the robustness of this bias in an economic context with incentives or in betting situations ( e.g. charness et al .
2010 , nilsson and anderson 2010 , erceg and galic 2014 ) .
they have also investigated whether the cognitive abilities of subjects are related to behavioral biases in general ( and to the conjunction fallacy in particular , cf .
oechssler et al . 2009 ) , and this has led to stimulating research with applications in finance .
epistemologists have made confirmation and bayesianism enter the debate ( e.g. tentori and crupi 2008 and 2012 , hartmann and meijs 2012 , schupbach 2012 , shogenji 2012 ) .
given that a conjunction fallacy occurs under robust experimental conditions , a natural question arises : how can this fallacy be explained ?
several accounts have been argued for , but no one has reached an uncontroversial status today ( as noted by fisk 2004 , nilsson et al 2009 , jarvstad and hahn 2011 , tentori et al .
first , tversky and kahneman originally suggested that a representativeness heuristic ( i.e. the probability that linda is a feminist is evaluated from the degree with which the instance of linda corresponds to the general category of feminists ) could account for some conjunction fallacy cases .
but it has been argued that the representativeness concept involved is informal and ill - specified ( gigerenzer 1996 , birnbaum et al 1990 ) , and suggestions to specify it in the technical sense of a likelihood value ( shafir et al 1990 , massaro 1994 ) account for limited cases only ( crupi et al .
according to another suggestion , agents actually evaluate the probability of the conjunction from some combination of the probabilities of the components , like averaging or adding ( fantino et al .
1997 , nilsson et al .
however , such explanations do not resist empirical tests , as tentori et al .
( 2013 ) have argued .
the latter propose an account of the conjunction fallacy based on the notion of inductive confirmation as defined in bayesian theory , and give experimental grounds for it
it is one of the currently promising accounts .
others have argued , also within a bayesian framework , that there are cases in which the conjunction fallacy is actually not a fallacy and can be accounted for rationally ( hintikka 2004 , von sydow 2011 , hartmann and meijs 2012 ) .
finally , another prominent proposal to account for the conjunction fallacy , on which we focus here , makes uses of so - called quantum - like " models , which rely on the mathematics of a major contemporary physical theory , quantum mechanics ( franco 2009 , busemeyer et al .
2011 , yukalov and sornette 2011 , pothos and busemeyer 2013)note that only mathematical tools of quantum mechanics are exploited , and that the models are not justified by an application of quantum physics to the brain .
the quantum - like account of the conjunction fallacy is particularly promising as it belongs to a more general theoretical framework of quantum - like modeling in cognition and decision making , which has been applied to many fallacies or human behavior considered as irrational ( for reviews , see pothos and busemeyer 2013 , ashtiani and azgomi 2015 , or bruza et al .
2015 ; textbooks include busemeyer and bruza 2012 , haven and khrennikov 2013 ) . for instance
, quantum - like models of judgments have been proposed to account for order effect , i. e. when the answers given to two questions depend on the order of presentation of these questions ( atmanspacher and rmer 2012 , busemeyer and bruza 2012 , wang and busemeyer 2013 , wang et al .
2014 ) ; for the violation of the sure thing principle , which states that if an agent prefers choosing action a to b under a specific state of the world and also prefers choosing a to b in the complementary state , then she should choose a over b regardless of the state of the world ( busemeyer et al .
2006a , busemeyer et al . 2006b , busemeyer and wang 2007 , khrennikov and haven 2009 ; for ellsberg s paradox more specifically , cf .
aerts et al .
2011 , aerts and sozzo 2013 , aerts et al .
2014 ; for allais paradox , cf .
khrennikov and haven 2009 , yukalov and sornette 2010 , aerts et al .
2011 ) ; for asymmetry judgments in similarity , i.e. that a is like b " is not equivalent to b is like a " ( pothos and busemeyer 2011 ) ; for paradoxical strategies in game theory such as in the prisoner s dilemma ( piotrowski and sladowski 2003 , landsburg 2004 , pothos and busemeyer 2009 , brandenburger 2010 ) . more generally ,
new theoretical frameworks with quantum - like models have been offered in decision theory and bounded rationality ( danilov and lambert - mogiliansky 2008 and 2010 , lambert - mogiliansky et al .
2009 , yukalov and sornette 2011 ) . as the quantum - like account of
the conjunction fallacy is one of the few promising accounts of the conjunction fallacy that are discussed today , we choose to focus on it in this paper . more specifically , we focus on the class of quantum - like models which are presented or defended in franco ( 2009 ) , busemeyer et al .
( 2011 ) , busemeyer and bruza ( 2012 ) , pothos and busemeyer ( 2013 ) and busemeyer et al .
( 2015 ) . in these models ,
an agent s belief is represented by a quantum state and not for instance by a measurement context .
our aim is to assess the empirical adequacy of these quantum - like models that are used to account for the conjunction fallacy .
we think that two points deserve particular scrutiny .
first , it is not always clear which version of the models are supposed to account for particular cases of conjunction fallacies are the simplest ones , called non - degenerate , sufficient ? or are the more general ones , called degenerate , needed ?
more recent works tend to favor degenerate models over non - degenerate ones , and non - degenerate models have received some recent criticisms ( cf . tentori and crupi 2013 and pothos and busemeyer 2013 , p. 315 - 316 ) , but a clear and definitive argument on the matter would be welcome .
second , the models have not yet been much tested on other predictions than the ones they were intended to account for .
it should be checked that they are not _ ad hoc _ by testing their empirical adequacy in general .
it is understandable that these two points have not been tested beforehand , as a new general pattern of explanation for the conjunction fallacy is hard to come up with .
but since the models have come to be seen as one of the most promising accounts , it becomes urgent to assess them empirically more thoroughly this is our goal in this paper . as for the first point discriminate between non - degenerate and degenerate models , we follow a suggestion made by boyer - kassem et al .
( 2016 ) to test so - called gr equations " , that are empirical predictions made by non - degenerate models .
such a gr test requires a new kind of experiment : not the original linda experiment , in which agents have to rank propositions , but an order effect experiment , in which two yes - no questions are asked in one order or in the other , to different agents .
existing data can not answer the question of whether the gr equations are verified , as was already noted in 2009 by franco : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ there are no experimental data on order effects in conjunction fallacy experiments , when the judgments are performed in different orders
. such an experiment could be helpful to better understand the possible judgment strategies . "
( franco 2009 , 421 ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ we fill this gap here by running several order effect experiments that collect the needed data . as for the second point
test new empirical predictions of the models , we consider two tests that apply to any version of the quantum - like models , whether degenerate or not , that are used in the account of the conjunction fallacy .
it is well - known in the literature that quantum - like models that account for the conjunction fallacy predict an order effect for the two questions associated with the conjunction ( is linda a bank teller ? " and is linda a feminist ? " ) .
actually , this predicted order effect is not a side effect of the quantum - like models , but a core feature of them : they can not account for the conjunction fallacy without it .
this enables a direct test of the quantum - like account of the conjunction fallacy , that we apply to our collected experimental data .
also , it has been shown that any quantum - like model of the kind involved in the account of the conjunction fallacy must make an empirical prediction called the
qq equality " ( wang and busemeyer 2013 , wang et al .
we thus test whether the qq equality is verified
. the failure of any of these last two tests will be enough to refute the current quantum - like account of the conjunction fallacy . here also , the needed data is not available in the literature , but can be conveniently obtained from the same above - mentioned new experimental configuration , with two yes - no questions in both order . note that our methodology is novel : we are not testing the quantum - like models against data produced by traditional conjunction fallacy experiments that the model were designed to explain , but we are testing them against other data , in a new experimental framework on which the models actually make some predictions , and it is why the experimental situation we shall consider is different from the usual linda experiment .
our experiment instantiates the mechanism that the quantum - like account claims agents follow : to evaluate a conjunction like feminist and bank teller " , agents are supposed to evaluate one characteristic after another , answering for themselves to two yes - no questions ( is linda a feminist ? " , is linda a bank teller ? " ) . in other words , the experiment we run somehow forces
agents to follow the purported quantum - like mechanism . to have more powerful tests , we have conducted several experiments , with variations of the scenario ( linda , but also others known as bill , mr
. f. and k. ) , of the protocol ( questionnaires or computer - assisted experiment ) and with or without monetary incentives .
the results we obtain show that current quantum - like models are not able to account for the conjunction fallacy .
the outline of the paper is the following . in section 2 ,
a general quantum - like model is introduced .
section 3 presents the three empirical tests that will be performed : the gr equations , order effect , and the qq equality .
the experimental protocol is presented in section 4 , and the results in section 5 .
section 6 presents the statistical analysis , and section 7 discusses the scope of the results and the future of the research on the conjunction fallacy account .
as indicated in the introduction , we focus in this paper on a family of quantum - like models based on similar hypotheses that have recently been proposed to account for the conjunction fallacy .
they are presented or defended in franco ( 2009 ) , busemeyer et al .
( 2011 ) , busemeyer and bruza ( 2012 ) , pothos and busemeyer ( 2013 ) and busemeyer et al .
( 2015 ) . for simplicity
, we choose here to summarize them with a single model with our own notations , and the correspondence with the various models from the literature can easily be made by the reader .
for illustrative purposes , we shall consider the conjunction fallacy through the linda case , but the generalization to other instances of the conjunction fallacy are straightforward . according to this literature , after reading linda s description , the subject who has to choose the more likely proposition between * linda is a bank teller " , * linda is a feminist and a bank teller "
. has the following mental process . to compare the propositions
, she evaluates each one in terms of a yes - no question : * is linda a bank teller ? " , * is linda a feminist and a bank teller ? " .
an important hypothesis of the quantum - like models is that , when the subject considers ( @xmath3 ) , she actually answers for herself successively two simple yes - no questions : * is linda a feminist ? " , * is linda a bank teller ? " .
yes " to @xmath3
amounts to answering
yes " to both @xmath4 and @xmath5 .
also , the hypothesis is made that the more probable outcome ( bank teller or feminist ) is evaluated first .
as the description of linda makes her more likely a feminist than a bank teller , this means that @xmath3 is answered by answering first @xmath4 and then @xmath5 . _ and then _
@xmath5 ( p. 418 ) .
anyway , the tests we consider in the forthcoming sections do not depend on this hypothesis . ]
let us now turn to the quantum - like framework that enable the quantitative prediction of the conjunction fallacy , @xmath6 . for pedagogical purposes ,
the non - degenerate versions of the quantum - like models are presented first , and the degenerate versions afterwards .
the belief states of agents are represented within a vector space . in the simple case where
an agent have just given an answer
( respectively no " ) to question @xmath4 , her belief state is represented by the vector @xmath7 ( respectively , @xmath8 ) . in accordance with the literature
, we shall say for short that these vectors represent the answers themselves . similarly with @xmath9 and @xmath10 for answers to question @xmath5 . the sets
( @xmath9 , @xmath10 ) and ( @xmath7 , @xmath8 ) respectively represent all possible answers to questions @xmath5 and @xmath4 , and thus each one is a basis of the same 2-dimension vector space .
the vector space is equipped with a scalar product , thus becoming a hilbert space : for two vectors @xmath11 and @xmath12 , the scalar product @xmath13 is a complex number .
the order of the vectors within a scalar product here matters : @xmath14 is the complex conjugate of @xmath13 .
the above bases are supposed to be orthogonal : @xmath15 , and of unitary norm : @xmath16 . a representation of the bases in the special case of real coefficients can be found on figure [ fig_bases - b - and - f ] [ left ] . ( 100,80)(-30,-10 ) ( 0 , 0)(1,0)50 ( 52 , -2)@xmath9 ( 0 , 0)(0,1)50 ( -5 , 55)@xmath10 ( 0 , 0)(3,1)47.4 ( 50 , 15)@xmath7 ( 0 , 0)(-1,3)15.8 ( -25 , 50)@xmath8 ( 100,80)(-30,-10 ) ( 0 , 0)(1,0)50 ( 52 , -2)@xmath9 ( 0 , 0)(0,1)50 ( -5 , 55)@xmath10 ( 0 ,
0)(3,1)47.4 ( 50 , 15)@xmath7 ( 0 , 0)(-1,3)15.8 ( -25 , 50)@xmath8 ( 0 , 0)(4,3)40 ( 42 , 33)@xmath17 ( 40,0)(40,15)(40,30 ) ( 30 , -7)@xmath18 ( 0,30)(20,30)(40,30 ) ( 2 , 33)@xmath19 ( 45,15)(42.5,22.5)(40,30 ) ( -5,15)(17.5,22.5)(40,30 ) an agent s state of belief is represented by a normalized vector @xmath17 within the hilbert space .
this vector can be decomposed in either of the two above - mentioned bases , as indicated on figure [ fig_bases - b - and - f ] [ right ] : @xmath20 with the specific values taken in figure [ fig_bases - b - and - f ] [ right ] in a hilbert space on real numbers , this equation becomes for instance : @xmath21 the belief state @xmath17 gathers all the relevant information needed to predict the behavior of the agent , in the following way .
predictions made by the quantum - like models are probabilistic .
when a question @xmath22 ( @xmath23 = @xmath1 or @xmath0 ) is asked , the probability that the agent answers @xmath24 ( ) is given by the squared modulus of the scalar product between the belief state and the vector representing the answer : @xmath25 this rule is usually called the born rule , in analogy with the quantum mechanics denomination .
it enables to compute the probability that the agent gives each of the 4 answers , in case questions @xmath5 or @xmath4 are asked ( as @xmath17 is normalized , @xmath26 ) . in the case of a real hilbert space like on figure [ fig_bases - b - and - f ] ,
a geometric interpretation of the born rule is the following : to compute the probability to answer , say , yes " to question @xmath5 , orthogonally project @xmath17 on @xmath9
this gives the length @xmath18 , and the wanted probability is just the squared of it .
so , the more @xmath17 is aligned with a basis vector @xmath27 , the larger the probability is that the agent will answer @xmath28 if question @xmath22 is posed ( note the if question @xmath22 is posed " part : in quantum - like models , the probability of an answer is only defined in the context in which the corresponding question is posed ) .
for instance , with the specific values in figure [ fig_bases - b - and - f ] [ right ] , @xmath29 , @xmath30 , @xmath31 and @xmath32 , which is consistent with the relative alignments of the basis vectors with @xmath17 .
the last postulate of the quantum - like model has to do with the way @xmath17 changes over time .
first , @xmath17 does not change unless the agent answers a question .
this conveys the fact that the agent s beliefs are not externally influenced .
this hypothesis is supposed to be relevant for cases in which the questions are posed to the agent relatively quickly .
second , when the agent answers a question @xmath5 or @xmath4 , her state of belief changes .
if her answer to question @xmath22 is @xmath24 , then her new state of belief just after giving the answer is : @xmath33 as the fraction in eq .
[ eq_projection - postulate ] is a complex number , the state of belief after an answer @xmath24 is proportional to the vector @xmath27 representing this answer . in the case of a real hilbert space like on figure [ fig_bases - b - and - f ] , after answering yes " to question @xmath5 , @xmath17 becomes either @xmath9 or @xmath34 , whatever the state of belief before the question .
in other words , after a question @xmath23 has been posed , the state of belief is bound to be along the basis vectors representing its answers .
. [ eq_projection - postulate ] can be interpreted as follows : the @xmath35 part represents the fact that @xmath17 is projected on @xmath27 , the basis vector representing the given answer ; the @xmath36 part is then just a multiplicative factor that ensures that the new state of belief is normalized .
hence , the above rule is often called the projection postulate . because of the projection postulate , the states before and after an answer are in general different .
they are the same only if the state previous to the answer is proportional to one of the basis vectors representing the possible answers to the question , i. e. when @xmath37 , where @xmath38 is a complex number such that @xmath39 ( in the real case , @xmath40 ) . in such a case , the agent answers @xmath28 to question @xmath23 with probability 1 , and eq .
[ eq_projection - postulate ] states that @xmath41 the fact that the state of belief changes when a question is answered is a real departure from the classical viewpoint .
classically , the answer is supposed to _ reveal _ a belief , which is pre - existent to the question , and is the same before and after . however , the quantum - like models predict that once a question has been answered , the same answer will be given if the same question is posed again just after .
let us now turn to the more general versions of these models , the degenerate ones .
the difference lies in the fact that an answer is not represented by a vector belonging to a 1d space , but by any subspace of dimension @xmath42 , for instance a plane .
then , the hilbert space is not of dimension 2 , but of a higher one . when question @xmath22 is posed , the probability that the agent answers @xmath24 is now defined as : @xmath43 where @xmath44 is the orthogonal projector onto the subspace representing answer @xmath28 to question @xmath22 .
the change in the state of belief is now : @xmath45 for the rest , the model is the same . the mental process that gives rise to the conjunction fallacy that has been described at the beginning of this section is graphically illustrated on figure [ fig_bases - qp - explanation ]
. the probability of considering that linda is a bank teller corresponds to the squared length of the projection of @xmath17 onto the bank teller vector @xmath9 , and @xmath46 .
for instance , with the specific values used in figure [ fig_bases - qp - explanation ] with a real hilbert space , @xmath47 and @xmath48 . on the other hand
, the probability of considering her to be feminist and bank teller corresponds to the squared length of the projection of @xmath17 onto two successive vectors , first @xmath7 and then @xmath9 , and @xmath49 . in the example of figure [ fig_bases - qp - explanation ] , @xmath50 and @xmath51 .
( 90,90)(-10,-30 ) ( 0 , 0)(1,0)50 ( 52 , -2)@xmath10 ( 0 , 0)(0,1)50 ( -5 , 55)@xmath9 ( 0 , 0)(3,1)47.4 ( 50 , 15)@xmath17 ( 0 , 0)(3,-4)30 ( 34 , -39)@xmath8 ( 0 , 0)(4,3)40 ( 42 , 33)@xmath7 ( -7 , 28)@xmath52 ( 47,15.5)(42,22)(37,28 ) ( 37,28)(18.5,28)(0,28 ) ( 45,16)(22.5,16)(0,16 ) ( -7 , 14)@xmath53 so , there exist some model configurations , like the one plotted on figure [ fig_bases - qp - explanation ] , in which the probability to be judged feminist and bank teller is higher than the probability to be judged bank teller , leading to @xmath54 in accordance with empirical results . a quantum - like model of the conjunction fallacy has been provided .
projects her state vector onto either @xmath7 or @xmath8 , but they do not suppose that answering for herself the second question @xmath5 projects the state vector onto @xmath9 or @xmath10 , because they argue that what is needed at this time is only an evaluation of the probability , and not a firm answer ( the authors acknowledge the validity of the projection postulate as soon as the agent gives a definite answer to a question ) .
so , the authors actually do not specify what the state vector is after the evaluation of the conjunction ( personal communication , 2014 ) . for simplicity ,
we have made as if the state vector was projected onto either @xmath9 or @xmath10 , like for other questions , but this has no consequence for our forthcoming tests .
see also section [ sec_expdesign ] . ]
this section presents the three empirical predictions of the above quantum - like model that we will test .
the first one applies to non - degenerate models , while the others apply to non - degenerate and degenerate models .
following boyer - kassem et al .
( 2016 ) , some specific empirical predictions can be derived for non - degenerate models , i.e. in which the answers are represented by subspaces of dimension 1 .
it can be shown that a well - known law from quantum mechanics , the law of reciprocity , holds .
consider the two questions @xmath4 and @xmath5 in one order or in the other .
the law of reciprocity states that , for @xmath55 , and @xmath56 , @xmath57 this law asserts that conditional probabilities of an answer given another answer are the same whatever the order of the questions @xmath5 and @xmath4 .
note that this law is typically quantum : it is not true in general for a classical model , in which @xmath58 , and thus @xmath59 as soon as @xmath60 .
the law of reciprocity can be instantiated in the following ways : [ eq_pb0a0=pa0b0 ] p(b_y|f_y ) = p(f_y|b_y ) , + [ eq_pb1a0=pa0b1 ] p(b_n|f_y ) = p(f_y|b_n ) , + [ eq_pb0a1=pa1b0 ] p(b_y|f_n ) = p(f_n|b_y ) , + [ eq_pb1a1=pa1b1 ] p(b_n|f_n ) = p(f_n|b_n ) .
some easy computation enable to show that the following equations , called the grand reciprocity ( gr ) equations , hold ( cf .
boyer - kassem et al .
2016 , section 3.1 ) : [ eq_mix0011 ] p(b_y|f_y ) = p(f_y|b_y ) = p(b_n|f_n ) = p(f_n|b_n ) , + [ eq_mix1001 ] p(b_n|f_y ) = p(f_y|b_n ) = p(b_y|f_n ) = p(f_n|b_y ) . these equations
[ eq_mix0011 ] and [ eq_mix1001 ] are equivalent to one another and to the law of reciprocity itself .
they state that the conditional probabilities that exist when @xmath5 is asked before @xmath4 is asked call it situation ( @xmath61 ) and in the ( @xmath62 ) situation are actually much constrained : among the eight quantities that can be experimentally measured , there is just one free real parameter . in other words ,
the non - degenerate quantum - like model presented in section [ subsec_ql - models ] actually leaves very little freedom to conditional probabilities .
the fact that the conditional probabilities are constrained by the gr equations had not been noticed beforehand for quantum - like models for the conjunction fallacy .
note that these empirical predictions are _ consequences _ of the quantum - like models that are used to explain the conjunction fallacy in the linda experiment , and that these consequences are observable in experimental situations @xmath63 and @xmath64 situations that are _ not _ the ones of the original linda experiment .
in other words , the gr equations show that a non - degenerate quantum - like model that is used to explain a linda experiment can be further tested on another kind of experiment .
we shall come back on this point in section [ sec_expdesign ] .
the interpretation of the conditional probabilities is clear : they have been defined as the probability of some answer to a second question given the answer to a first question .
this is straightforwardly consistent with the models presented in section [ sec_ql - models ] , and in accordance with classical order effect experiments .
another interpretation of the conditional probabilities could be that of an answer given some new piece of evidence , but this is not what is considered in this paper .
quantum - like models of section [ subsec_ql - models ] can predict an order effect , that is , predict that agents give different answers to the question @xmath4 followed by question @xmath5 , and to the question @xmath5 followed by question @xmath4 ( cf .
figure [ fig_order - effect ] ) .
this comes from the projection postulate that modifies the state of belief when an answer is given to a question .
this order effect property of the quantum - like models is well - known , and it has actually been used to provide a quantum - like account of order effect ( see for example conte et al . 2009 , busemeyer et al . 2009 , busemeyer et al .
2011 , atmanspacher and rmer 2012 , pothos and busemeyer 2013 , wang and busemeyer 2013 and wang et al .
2014 , boyer - kassem et al . 2016 ) thus , the same models are at the basis of the account of order effect and of the conjunction fallacy .
( 80,70)(-30,-10 ) ( 0 , 0)(1,0)50 ( 52 , -2)@xmath9 ( 0 , 0)(0,1)50 ( -5 , 55)@xmath10 ( 0 , 0)(3,1)47.4 ( 50 , 15)@xmath7 ( 0 , 0)(-1,3)15.8 ( -25 , 50)@xmath8 ( 0 , 0)(4,3)40 ( 42 , 33)@xmath17 ( 40,0)(40,15)(40,30 ) ( 40,0)(38,6)(36,12 ) ( 45,15)(42.5,22.5)(40,30 ) ( 45,15)(45,7.5)(45,0 ) more importantly , it can be shown that only models that display an order effect are able to account for the conjunction fallacy ( cf .
busemeyer et al .
2011 , busemeyer and bruza 2012 , bruza et al .
2015 p. 388
, busemeyer et al . 2015 ) .
in other words , the quantum - like models of section 2 that do not present an order effect can not predict @xmath65 , and thus can not account for the conjunction fallacy .
the reason is , in short , the following : questions @xmath5 and @xmath4 are either compatible or incompatible in the standard quantum sense . in the latter case ,
the hilbert space is ( in the simplest case ) 2d , with basis vectors like on figure [ fig_bases - b - and - f ] , and there is an order effect . in the former case ,
the hilbert space is ( in the simplest case ) 4d , with basis vectors ( @xmath66 , @xmath67 , @xmath68 , @xmath69 ) , where the vector @xmath70 stands for answer @xmath28 to question @xmath5 and answer @xmath71 to question @xmath4 , in whatever order . and
such a model displays no order effect : whatever the order of the questions , the probability of an answer @xmath28 to question @xmath5 and of an answer @xmath71 to question @xmath4 will be @xmath72 , where @xmath73 is the coordinate along the @xmath70 vector ( @xmath74 ) .
can such a model predict a conjunction fallacy to occur ? on the one side , consider the evaluation of the conjunction : the agent first considers @xmath4 ; if she answers yes " , the state vector is projected onto the plane @xmath75 .
if she now answers yes " to @xmath5 , the resulting vector is projected onto @xmath66 .
so , the probability to answer yes " to both questions is given by the square modulus of the @xmath66 component , i.e. @xmath76 . on the other side , consider the evaluation of @xmath1 , for which the agent considers @xmath5 .
if she answers yes " , the state vector is projected onto the plane @xmath77 .
the probability of such an answer is given by the squared modulus of the length of this projection , namely @xmath78 ( remember that the basis vectors are orthogonal ) .
this quantity is at least larger than @xmath76 , so a conjunction fallacy can not occur . to sum up ,
any quantum - like model of the kind considered in section 2 which claims to account for the conjunction fallacy , be it non - degenerate or degenerate , has to display an order effect on the corresponding questions .
this provides our second test ( cf .
section [ statanalysis ] for a discussion of the mathematical expression of the test ) .
the proponents themselves of the quantum - like account of the conjunction fallacy consider that the use of incompatible concepts ( or questions ) is the key feature of their model .
as incompatible questions straightforwardly imply an order effect , our order effect test is actually a direct test of the core feature of the quantum - like account . and
@xmath5 are compatible or incompatible .
but the easiest way to do so is actually to test the order effect on these two questions . ] as for the gr equations , note that the order effect is here understood as an experimental situation with two successive yes - no questions , posed in one order or in the other after a text has been read , and that no new piece of evidence is provided between the two questions .
to sum up , three features are essential for the quantum - like models under study to account for the conjunction fallacy : the born rule ( eq . [ eq_born - rule ] ) , the projection postulate ( eq . [ eq_projection - postulate ] ) , and the presence of incompatible questions entailing order effects .
the quantum - like models of section 2 , whether degenerate or not , have recently been shown to entail new testable empirical predictions ( wang and busemeyer 2013 ) : a quantum question " ( qq ) equality .
noting @xmath79 the probability of answering first @xmath28 to question @xmath22 _ and then _
@xmath71 to question @xmath80 ( this is a joint probability , not a conditional probability ) , the qq equality reads : @xmath81 this equality is of prime importance . as busemeyer et al .
( 2015 , 241 ) put it , it is an a priori , _ precise _ , _ quantitative _ , and _
parameter - free _ prediction about the pattern of order effects " .
it has served as a test of the quantum - like models that claim to account for order effect .
it turns out that it has been statistically supported across a wide range of 70 national field experiments ( containing 651 to 3,006 nationally representative participants per field experiment ) that examined question - order effects ( wang et al . , 2014 ) "
( _ ibid . _ ) .
similarly , the qq equality can be empirically tested in the case of the quantum - like models that account for the conjunction fallacy , as the models are the same .
this constitutes our third test ( further statistical details about the test are given in section 6 ) .
the three tests presented in the previous section ( gr equations , order effect , qq equality ) require to carry out an order effect experiment that shows the description of linda and then asks the questions @xmath4 and @xmath5 in both orders , ( @xmath62 ) or @xmath63 .
the former order somehow forces the agent to follow the cognitive process supposed by the quantum - like models when evaluating a conjunction .
we propose here its first experimental realization , in order to test the quantum - like models of section 2 .
the order effect experiment we are considering here is different from the original conjunction fallacy experiment .
if we want to claim that it tests anyway the quantum - like account of the conjunction fallacy , do we need to make some extra hypothesis ?
for instance , do we need to suppose that the quantum - like model for the conjunction fallacy also applies to another kind of experiment ? or do we need to assume that forcing an agent to explicitly answer the two questions will give the same results as when she answers them for herself ?
we need not , because these assumptions are already made in the papers we are considering .
first , the simple fact that the quantum - like account of the conjunction fallacy relies on models " that have a general and universal form and not only on _ ad hoc _ rules that apply to a limited number of situations , allows anyone to use these models _ ad libitum _ in any experimental situation that the model may represent .
the order effect situation , in which two questions are asked , clearly falls within that range .
so , we are allowed to apply ( and thus to test ) the quantum - like models of the conjunction fallacy in an order effect experiment .
this amounts to testing experimental predictions of the models that they make because they have a general form . as the proponents of the models write : the basic quantum model underpinning the conjunction fallacy [ ... ] makes new _ a priori _ predictions .
foremost among them is the consequence that incompatible judgments and decisions must entail order effects . "
( bruza et al .
2015 , p. 388 ) .
( recall that incompatible judgments are required in the quantum - like model of the conjunction fallacy . ) in other words , the conjunction fallacy model entails order effects , and thus can be tested on them .
this is all the more true than the authors actually claim that the quantum - like models used for the conjunction fallacy are the same as those used to explain other fallacies or phenomena , like order effect itself or similarity judgments .
all models belong to a family that are often called a theory " of quantum cognition , and they are meant to make predictions on a wide range of phenomena , in diverse experimental situations and the authors rightly claim that this is a strength of their approach .
this supports the generality of the quantum - like models used for the conjunction fallacy .
thus , it is legitimate to use them in other situations like the order effect one . besides
, these models _ have been _ applied to question order effect ( wang and busemeyer 2013 , wang et al .
2014 ) , and it is clear that no extra hypothesis than the ones presented in section 2 is needed for that . in sum , the literature claims that the very same models can be used for the conjunction fallacy and for question order effect , so we are justified in testing them on new order effect cases as linda s .
finally , recall that we consider here two successive yes - no questions , asked in both orders .
thus , the conditional probabilities are interpreted as probabilities of a second answer given a first answer .
this is fully in line with the models of the conjunction fallacy themselves .
consider for instance : in this problem there are two questions : the feminism question and the bank teller question .
for each question , there are two answers : yes or no . "
( busemeyer and bruza 2012 , p. 15 ) ;
we consider two dichotomous questions a and b , as for example a : is linda a feminist ? and
b : is linda a bank teller ? "
( franco 2009 p. 416 ) .
what we propose here is to explicitly pose these two questions . in order to strengthen our experimental tests ,
we have considered four scenarios that have been shown in the literature to give rise to conjunction fallacies , from which we have built four experimental tasks a task consists for an agent in reading a text and then sequentially answering two yes - no questions .
the first task is drawn from the case of * linda * ( tversky and kahneman 1983 ) : for the french version that was actually used in the experiments . ] * text :
linda is 31 years old , single , outspoken , and very bright .
she majored in philosophy . as a student , she was deeply concerned with issues of discrimination and social justice , and also participated in anti - nuclear demonstrations . " * @xmath4 : according to you , is linda a feminist ? "
* @xmath5 : according to you , is linda a bank teller ? "
the second task is drawn from the case of * bill * ( tversky and kahneman 1983 ) : * text : bill is 34 years old .
he is intelligent , but unimaginative , compulsive , and generally lifeless . in school , he was strong in mathematics but weak in social studies and humanities . "
* @xmath82 : according to you , is bill an accountant ? " * @xmath83 : according to you , does bill play jazz for a hobby ? "
the third task is drawn from the case of * mr .
f. * ( tversky and kahneman 1983 ) : * text : a health survey was conducted in a representative sample of adult males in france of all ages and occupations . mr .
f. was included in the sample .
he was selected by chance from the list of participants . "
* @xmath84 : according to you , has mr .
f. already had one or more heart attacks ? " * @xmath85 : according to you , is mr .
f. over 55 years old ? " the fourth task is drawn from the case of * k. * , a russian woman ( tentori et al .
2013 ) : * text : k. is a russian woman " .
* @xmath86 : according to you , does k. live in new - york ? "
* @xmath87 : according to you , is k. an interpreter ? "
so as to increase the robustness of our results , we have chosen these four tasks as they display different kinds of conjunction fallacies , in the sense of tversky and kahneman ( 1983 ) who have distinguished between m a and a b paradigms . in the former , a model m ( the text describing the person ) is positively associated with an event a ( one of the two sentences forming the conjunction ) and negatively with the other event b.
this is the case of the linda scenario : the introductory text m is positively associated with the event linda is a feminist " and negatively with the other one linda is a bank teller " . also , bill s scenario is of type m a .
differently , in the a
b paradigm , a is positively associated with b , but not with the model m. for instance , mr .
f. is over 55 years old " is positively associated with mr . f. already had one or more heart attacks " , but not with the text .
the scenario of the russian woman seems to correspond to neither paradigm : the positive association occurs between the text m and the conjunction of the two constituents a and b , and not with only one of them , so we might call it m(ab ) the fact that the woman is russian is strongly associated with the fact that she lives in new york and is also an interpreter .
conjunction fallacies and quantum - like models have been studied by scholars of various fields , and in particular by psychologists and economists ( cf .
section [ sec_intro ] ) . to keep with these two traditions ,
we have chosen not to limit ourselves to one experimental protocol which also has the advantage of increasing the robustness of the experimental findings .
we have varied the administration method , with paper questionnaires like in the psychological tradition and with computer implementations like in the economical tradition , with and without payment .
we have carried out three experiments ( cf .
table [ tab_listexp ] for a summary ) . in the first experiment ,
two tasks were successively presented to the subjects : that of mr . f. and that of bill .
the experiment was conducted in march and april 2015 at the university of tours and of nice sophia antipolis ( france ) , with a total of 496 students in medicine , economics and management . in the psychological tradition ,
the tasks were implemented with paper questionnaires , in the lecture hall at the end of classes .
because of the improvised recruitment without appointment , and because of the short length of the task , the students were not paid , like in the psychological tradition .
these tasks are noted @xmath88 and @xmath89 , with an index @xmath90 " for paper " .
the second experiment successively featured the 4 tasks introduced above in the following order : k. the russian woman , mr .
f. , bill and linda .
the experiment was conducted on april 2015 at the lameta , the experimental economics laboratory of the university of montpellier 1 ( france ) , in 19 sessions , with a total of 302 students possibly from any discipline . in the economics tradition ,
the tasks were implemented on computers ( created with the z - tree program , fischbacher 2007 ) , and students were recruited online and received a show - up fee ( 5 or 9 euros , according to their campus of origin ) to remunerate their attendance and to reduce the effect of selection bias .
these tasks are noted @xmath91 , @xmath92 , @xmath93 and @xmath94 , with an index @xmath95 " for computer " and a euro for the payment .
a third experiment involved the task of linda , in a mixed methodology .
it was conducted on october 2014 in the leen , the experimental economics laboratory of the university nice sophia antipolis , with a computerized questionnaire .
354 students were recruited on the fly at the end of the classes , and were not paid for the short task .
this task is noted @xmath96 , with an index @xmath95 " .
.experimental tasks that were carried out , together with their administration methods , the location , and the number of subjects involved .
two dashed horizontal lines separate into three groups the seven experimental tasks , corresponding to three distinct experiments.[tab_listexp ] [ cols="<,<,<,<,<,<",options="header " , ] on the basis of the above experimental results , we now would like to test three research hypotheses that have motivated the quantum - like modeling literature on conjunction fallacy , and that correspond to the building blocks of the current models presented in section 2 .
this shall provide some interpretation of the bare statistical results obtained in sections [ subsec_test - gr - equations ] to [ subsec_test - qq - equality ] .
the first two hypotheses have already been presented in the introduction and concern the validity of quantum - like models , while the third one is larger and goes beyond quantum - like models : * * hyp .
# 1 : * non - degenerate quantum - like models ( presented in section 2 ) can account for the conjunction fallacy , * * hyp .
# 2 : * non - degenerate or degenerate quantum - like models ( presented in section 2 ) can account for the conjunction fallacy , * * hyp .
# 3 : * the conjunction fallacy account can rely on a question order effect account .
the first hypothesis is the simplest and less general one .
it restricts accounts of the conjunction fallacy to the simplest versions of the quantum - like models , i.e. non - degenerate ones , where answers are represented by 1-d subspaces .
this is the hypothesis made in franco ( 2009 ) , who only considers non - degenerate models .
this hypothesis implies that the gr equations are empirically verified . as section [ subsec_test - gr - equations ]
has shown that the gr equations are never verified in our experiments , we can safely say that the first hypothesis is empirically refuted by our data .
in other words , non - degenerate quantum - like models can not account for order effects .
this refutes the proposal by franco ( 2009 ) , who has only considered non - degenerate models all other quantum - like models cited in section 2 are not refuted , since they also consider degenerate models .
the rejection of the first hypothesis echoes recent debates .
the empirical inadequacy of non - degenerate models for the conjunction fallacy has already been discussed , although the question had not been definitely settled ( cf .
tentori and crupi 2013 and pothos and busemeyer 2013 , p. 315 - 316 ) . in a similar vein
, it has been shown that non - degenerate models for order effect are not empirically adequate ( boyer - kassem et al .
overall , our result is in line with previous suggestions that degenerate models should be preferred to non - degenerate models , as the latter should be considered as toy models " only ( e.g. busemeyer and bruza 2012 , busemeyer et al .
2015 ) .
the second research hypothesis extends the first one by considering also degenerate models , that is , models in which an answer is represented by a @xmath97-d subspace , e.g a plane .
this hypothesis is shared by all papers cited in the beginning of section 2 , except franco ( 2009 ) : the conjunction fallacy can be accounted for by quantum - like models in general , be they non - degenerate or degenerate .
as argued in section [ sec_empirical - tests ] , non - degenerate and degenerate models have ( i ) to display an order effect and ( ii ) to respect the qq equality .
thus , the second hypothesis is testable by means of the test of the order effect and that of the qq equality .
table [ stathyp2final ] summarizes the findings on these matters .
both tests results are reported , the satisfaction of the qq equality in the second column and the presence of order effect in the third one .
the last column reports the joint outcomes of the two tests , that is , the outcome of the logical operator and " , because either one test or the other one are sufficient to refute the quantum - like models of conjunction fallacy considered in this paper .
recall that we have adopted a very conservative approach on the error of type i , so as to be conclusive with a high degree of certainty .
so , we can be quite sure that the second research hypothesis is rejected in at least three out of seven tasks .
our conclusion here is that the quantum - like models can not account for the general phenomenon of the conjunction fallacy .
it is the first time that such a strong result is obtained experimentally .
l | c ; 0.5pt/4pt c | c task i d & * qq * & * oe * & * qq and oe * + @xmath88 & - & yes & - + @xmath89 & - & - & - + @xmath91 & - & no & * no * + @xmath98 & - & no & * no * + @xmath93 & - & - & - + @xmath94 & - & - & - + @xmath96 & no & yes & * no * + the third hypothesis is not restricted to quantum - like models , but is concerned with the general idea that the conjunction fallacy is related to a question order effect between suitable questions ( for instance in the linda scenario between the questions @xmath99 and @xmath4 ) .
it implies that an order effect must be observed in our experiments , and thus this hypothesis is testable by means of the order effect test .
two out of seven tasks exhibit no ( or insignificant ) order effect , as shown in section [ subsec_test - order - effect ] . and
yet , the corresponding scenarios ( k. and mr .
f. ) do exhibit a conjunction fallacy .
these results suggest that the third hypothesis , according to which the conjunction fallacy can be accounted for from an order effect , seems to be experimentally refuted .
note that the consequences of the rejection of this hypothesis have an even much broader impact than the ones deriving from the rejections of previous hypotheses : not only are we rejecting the original modeling strategy exploited by the quantum - like literature based on the introduction of an order effect to explain the conjunction fallacy , but we are also preventing its adoption for any other alternative theory ( bayesian , heuristics ... ) .
the conjunction fallacy can not be reduced , in terms of mental acts , to the order effect phenomenon .
this finding sheds some new light into an important modeling issue .
we have considered the quantum - like accounts of the conjunction fallacy that have been proposed or defended by franco ( 2009 ) , busemeyer et al .
( 2011 ) , busemeyer and bruza ( 2012 ) , pothos and busemeyer ( 2013 ) and busemeyer et al .
( 2015 ) which common trait is to represent the belief of the decision - maker with the quantum state .
we have tested three empirical predictions of these models : the gr equations ( boyer - kassem et al .
2016 ) , that apply to non - degenerate versions only of the models , the existence of an order effect and the qq equality ( wang and busemeyer 2013 ) , which apply to both non - degenerate and degenerate versions of the models , hence to the most general version of the papers .
such tests can not be performed in traditional conjunction fallacy experiments , in which subjects have to rank propositions , but require an order effect experiment , in which two yes - no questions are asked in either order .
so , the tests concern empirical predictions that are not the data that the models were supposed to explain in the first place , but are predictions of the models anyway , and are directly related to the core feature of the models , namely the incompatibility between questions .
we have performed such order effect experiments , by using a robust protocol that varies the stories ( linda , bill , mr . f. , k. ) , the administration method ( paper questionnaires or computer ) , and a possible payment , with seven tasks in total and several hundreds of subjects .
our empirical results clearly reject the hypothesis that non - degenerate models can account for the conjunction fallacy ( which is the hypothesis made in franco 2009 ) .
this confirms the recent tendency from the advocates of the quantum - like approach to consider non - degenerate models as toy models only .
most importantly , our results also reject the more general hypothesis that non - degenerate or degenerate models can account for the conjunction fallacy , which is the hypothesis made in all other papers .
as we have used very conservative statistical tests , we can safely say this general hypothesis is refuted in at least three tasks out of seven .
so the present paper provides the first clear experimental rejection of the quantum - like explanation of the conjunction fallacy .
now , it may be possible that not _ all _ instances of the conjunction fallacy can be accounted for in a quantum - like fashion , but that _ some _ instances can .
for instance , our experimental results have not formally excluded that bill s scenario could be amenable to a quantum - like account .
there is room for possible future experimental research here a possible line of division to be investigated could be between ab and ma scenarios of conjunction fallacies .
but thus , the quantum - like account would loose its generality , which was its strength . moreover , if quantum - like models were to apply to some cases of conjunction fallacies , it seems very likely that it should be degenerate versions , since non - degenerate one have been strongly ruled out .
this comes with possible drawbacks or specific duties , as argued in boyer - kassem et al .
in particular , a degenerate model resorts to some extra dimensions in the hilbert space that should receive theoretical and experimental justifications so as not to be just _ ad hoc_. and more general tests on elementary dimensions can also be considered .
as our experimental results speak against the quantum - like models of the conjunction fallacy , they can be interpreted as indirect support in favor of alternative accounts of the conjunction fallacy , like bayesian ones ( e.g. tentori et al .
2013 ) , or other kinds of quantum - like models for the conjunction fallacy that have not been tested in this paper , like yukalov and sornette ( 2010 , 2011 ) .
however , our results also provide some conclusions well beyond quantum - like modeling : they show that the conjunction fallacy can not be accounted for by any model or mechanism that relies on order effect , or entails an order effect , between the two characteristics at play ( feminist " and bank teller " in linda s case ) .
quantum - like models are well - known such examples , but it must be clear that any existing or future alternative explanation that involves a question order effect is ruled out .
after the failure of quantum - like models , this places a hard constraint on alternative explanations of the conjunction fallacy .
we suggest that future works should try to theoretically inquire whether alternative explanations predict an order effect , and to experimentally test it .
even if the quantum - like models studied in this paper are not able to account for our data , a possible research strategy could be not to abandon the quantum - like modeling of the conjunction fallacy altogether , but instead to try to modify and improve it so that it finally agrees with the experimental data . in this spirit
, one could investigate whether the use of a more general measurement theory or generalized observables could be adequate .
for instance , the use of positive operator valued measures ( povms ) , from quantum physics , has been recently applied to quantum - like models of cognition ( cf . khrennikov and basieva 2014 ) .
however , it seems to face some new challenges like response replicability ( cf .
khrennikov et al .
2014 , basieva and khrennikov 2015 ) .
another quantum - like line of research that does not face this problem considers a modification of the born rule ( aerts and sassoli de bianchi , 2015 ) . as a last remark ,
our methodology has been here to test quantum - like models of the conjunction fallacy with new experimental predictions .
we think this methodology could be fruitfully extended to quantum - like models that address other fallacies , such as the disjunction fallacy or the inverse fallacy .
many thanks to corrado lagazio for suggestions on the statistical part .
we would like to thank jerome busemeyer , vincenzo crupi , dorian jullien , andrei khrennikov , katya tentori , vyacheslav yukalov and participants at the 2015 conference quantum probability and the mathematical modelling of decision making " ( fields institute , toronto ) , for useful comments and suggestions . experimental economics laboratory at the university of nice ( leen - nice lab ) and at the university of montpellier ( lameta ) have kindly hosted the experiments .
this work has benefited from the joint gis rnmshs and french cnrs grant quantumtest " .
.8cm0 cm aerts , diederik , jan broekaert , marek czachor and bart dhooghe ( 2011 ) , a quantum - conceptual explanation of violations of expected utility in economics " , _ quantum interaction .
lecture notes in computer science _ 7052 : 192 - 198 .
aerts , diederik and massimiliano sassoli de bianchi ( 2015 ) , beyond - quantum modeling of question order effects and response replicability in psychological measurements " , arxiv : 1508.03686 .
aerts , diederik and sandro sozzo ( 2013 ) , a contextual risk model for the ellsberg paradox " , _ journal of engineering science and technology review _ 4 : 246250 .
aerts , diederik , sandro sozzo , and jocelyn tapia ( 2014 ) , identifying quantum structures in the ellsberg paradox " , _ international journal of theoretical physics _ 53 , 3666 - 3682 .
allais , maurice ( 1953 ) , le comportement de lhomme rationnel devant le risque : critique des postulats et axiomes de lcole amricaine " , _ econometrica _ 21 : 503546 .
ashtiani , mehrad and mohammad abdollahi azgomi ( 2015 ) , a survey of quantum - like approaches to decision making and cognition " , _ mathematical social sciences _ 75 : 4980 .
atmanspacher , harald and hartmann rmer ( 2012 ) , order effects in sequential measurements of non - commuting psychological observables " , _ journal of mathematical psychology _ 56 : 274280 .
barker , lawrence , henry rolka , deborah rolka and cedric brown ( 2001 ) , testing for binomial random variables : which test to use ? " , _ the american statistician _ 55 ( 4 ) : 279287 .
basieva irina and andrei khrennikov ( 2014 ) , on a possibility to combine the order effect with sequential reproducibility for quantum measurements " , working paper , http://arxiv.org/abs/1502.00132[arxiv:1502.00132 ] .
birnbaum , michael h. , carolyn j. anderson and unda g. hynan ( 1990 ) , theories of bias in probability judgment . " , _ cognitive biases : advances in psychology _ 68:477499 .
bonini , nicolao , katya tentori and daniel osherson ( 2004 ) , a different conjunction fallacy " , _ mind & language _ 19 ( 2 ) : 199210 .
boyer - kassem , thomas , sbastien duchne and eric guerci ( 2016 ) , testing quantum - like models of judgment for question order effects " , _ mathematical social sciences _ 80 : 33 - 46 .
brandenburger , adam ( 2010 ) , the relationship between quantum and classical correlation in games " , _ games and economic behavior _ 69 : 175183 .
bruza , peter d. , zheng j. wang , and jerome r. busemeyer ( 2015 ) , quantum cognition : a new theoretical approach to psychology " , _ trends in cognitive sciences _
19(7 ) : 383393 .
busemeyer , jerome r. and peter d. bruza ( 2012 ) , _ quantum models of cognition and decision _ , cambridge : cambridge university press .
busemeyer , jerome r. , marvin matthew and zheng wang ( 2006a ) , a quantum information processing explanation of disjunction effects " , in r. sun and n. miyake ( ed . ) , _ proceedings of the 28th annual conference of the cognitive science society _ ,
erlbaum , p. 131135 .
busemeyer , jerome r. , emmanuel m. pothos , riccardo franco and jennifer s. trueblood ( 2011 ) , a quantum theoretical explanation for probability judgment errors " , _ psychological review _ 118 ( 2 ) : 193218 .
busemeyer , jerome r. , and zheng wang ( 2007 ) , quantum information processing explanation for interactions between inferences and decisions " , in : d. bruza , w. lawless , k. van rijsbergen and d.a .
sofge ( eds . ) , _ quantum interaction : papers from the aaai spring symposium , technical report ss-07 - 08 _ , aaai press , p. 9197 .
busemeyer , jerome r. , zheng j. wang , and ariane lambert - mogiliansky ( 2009 ) , empirical comparison of markov and quantum models of decision making " , _ journal of mathematical psychology _ 53 : 423433 .
busemeyer , jerome r. , zheng j. wang , emmanuel m. pothos , and jennifer s. trueblood ( 2015 ) , the conjunction fallacy , confirmation , and quantum theory : comment on tentori , crupi , and russo ( 2013 ) " , _ journal of experimental psychology : general _ 144(1 ) : 236243 .
busemeyer , jerome r. , zheng wang , and james t. townsend ( 2006b ) , quantum dynamics of human decision - making " , _ journal of mathematical psychology _ 50 ( 3 ) : 220241 .
charness , gary , edi karni , and dan levin ( 2010 ) , on the conjunction fallacy in probability judgment : new experimental evidence regarding linda " , _ games and economic behavior _ 68 : 551556 .
conte , elio , andrei yuri khrennikov , orlando todarello , antonio federici , leonardo mendolicchio , and joseph p. zbilut ( 2009 ) , mental states follow quantum mechanics during perception and cognition of ambiguous figures " , _ open systems & information dynamics _
16(1 ) : 117 .
crupi , vincenzo , branden fitelson , and katya tentori ( 2008 ) , probability , confirmation and the conjunction fallacy " , _ thinking & reasoning _ 14 : 182199 .
danilov , vladimir i. and ariane lambert - mogiliansky ( 2008 ) , measurable systems and behavioral sciences " , _ mathematical social sciences _
55(3 ) : 315340 .
danilov , vladimir i. and ariane lambert - mogiliansky ( 2010 ) , expected utility theory under non - classical uncertainty " , _ theory and decision _ 68(1 ) : 2547 .
ellsberg , daniel ( 1961 ) , risk , ambiguity , and the savage axioms " , _ quarterly journal of economics _
75(4 ) : 643669 .
erceg , nikola and zvonimir galic ( 2014 ) , overconfidence bias and conjunction fallacy in predicting outcomes of football matches " , _ journal of economic psychology _ 42 : 5262 .
fantino , edmund , james kulik , stephanie stolartz - fantino , and william wright ( 1997 ) , the conjunction fallacy : a test of averaging hypothesis " , _ psychonomic bulletin & review _
4(1 ) : 96101 .
fischbacher , urs ( 2007 ) , z - tree : zurich toolbox for ready - made economic experiments " , _ experimental economics _
10(2 ) : 171178 .
fisk , john e. ( 2004 ) , conjunction fallacy " , in r. f. pohl ( ed . ) , _ cognitive illusions : a handbook on fallacies and biases in thinking , judgment , and memory _ , london : psychology press .
franco , riccardo ( 2009 ) , the conjunction fallacy and interference effects " , _ journal of mathematical psychology _ 53 : 415422 .
gigerenzer , gerd ( 1996 ) , on narrow norms and vague heuristics : a reply to kahneman and tversky ( 1996 ) " , _ psychological review _
103(3 ) : 592596 . haven , emmanuel and andrei khrennikov ( 2013 ) , _ quantum social science _ ,
cambridge : cambridge university press .
hartmann , stephan and wouter meijs ( 2012 ) , walter the banker : the conjunction fallacy reconsidered " , _ synthese _ 184 : 7387 .
hertwig , ralph ( 1997 ) , judgment under uncertainty : beyond probabilities " , berlin , germany : max planck institute for human development .
unpublished manuscript .
hertwig , ralph , bjrn benz and stefan krauss ( 2008 ) , the conjunction fallacy and the many meanings of ` and ' " , _ cognition _ 108 : 740753 .
hertwig , ralph , and valerie m. chase ( 1998 ) , many reasons or just one : how response mode affects reasoning in the conjunction problem " , _ thinking & reasoning _
4(4 ) : 319352 .
hertwig , ralph , and gerd gigerenzer ( 1999 ) , the ` conjunction fallacy ' revisited : how intelligent inferences look like reasoning errors " _ journal of behavioral decision making _ 12 : 275305 .
hintikka , jaakko ( 2004 ) , a fallacious fallacy " , _ synthese _ 140 : 2535 .
jarvstad , andreas , and ulrike hahn ( 2011 ) , source reliability and the conjunction fallacy " , _ cognitive science _
35(4 ) : 682711 .
kahneman , daniel ( 2011 ) , _ thinking , fast and slow , _ new york : farrar , straus and giroux .
kahneman , daniel , and amos tversky ( 1996 ) , on the reality of cognitive illusions : a reply to gigerenzer s critique " , _ psychological review _ 103 : 582591 .
khrennikov , andrei ( 2010 ) , _ ubiquitous quantum structure . from psychology
to finance _ , heildelberg : springer .
khrennikov , andrei and irina basieva ( 2014 ) , quantum model for psychological measurements : from the projection postulate to interference of mental observables represented as positive operator valued measures " , _ neuroquantology _ 12:324 - 336 .
khrennikov , andrei , irina basieva , ehtibar n. dzhafarov , and jerome r. busemeyer ( 2014 ) , quantum models for psychological measurements : an unsolved problem " , _ plos one _ 9(10 ) : e110909 .
khrennikov , andrei and emmanuel haven ( 2009 ) , quantum mechanics and violations of the sure - thing principle : the use of probability interference and other concepts " , _ journal of mathematical psychology _ 53 : 378388 .
lambert - mogiliansky , ariane , shmuel zamir and herv zwirn ( 2009 ) , type indeterminacy : a model of the kt(kahneman tversky)-man " , _ journal of mathematical psychology _ 53:349361 .
landsburg , steven e. ( 2004 ) , quantum game theory " , _ notices of the american mathematical society _
51(4 ) : 394399 .
massaro , dominique w. ( 1994 ) , a pattern recognition account of decision making " , _ memory & cognition _ 22 : 616627 .
mellers , barbara , ralph hertwig and daniel kahneman ( 2001 ) , do frequency representations eliminate conjunction effects ?
an exercise in adversarial collaboration " , _ psychological science _ 12 : 269275 .
moro , rodrigo ( 2009 ) , on the nature of the conjunction fallacy " , _ synthese _ 171 : 124
. nilsson , hkan , and patric andersson ( 2010 ) , making the seemingly impossible appear possible : effects of conjunction fallacies in evaluations of bets on football games " , _ journal of economic psychology _ 31 : 172180 .
nilsson , hkan , anders winman , peter juslin and gran g. hansson ( 2009 ) , linda is not a bearded lady : configural weighting and adding as the cause of extension errors " , _ journal of experimental psychology : general _ 138 : 517534 .
oechssler , jrg , andreas roider and patrick w. schmitz ( 2009 ) , cognitive abilities and behavioral biases " , _ journal of economic behavior & organisation _ 72 : 147152 .
piotrowski , edward w. and jan stadkowski ( 2003 ) , an invitation to quantum game theory " , _ international journal of theoretical physics _
42(5 ) : 10891099 .
pothos , emmanuel m. and jerome r. busemeyer ( 2009 ) , a quantum probability explanation for violations of rational `` decision theory '' , _ proceedings of the royal society b _ 276:21712178 .
pothos , emmanuel m. and jerome r. busemeyer ( 2011 ) , a quantum probability explanation for violations of symmetry in similarity judgments . " , _ proceedings of the 32nd annual conference of the cognitive science society _ 28482854 .
pothos , emmanuel m. and jerome r. busemeyer ( 2013 ) , can quantum probability provide a new direction for cognitive modeling ? " , _ behavioral and brain sciences _ 36 : 255327 .
schupbach , jonah n. ( 2012 ) , is the conjunction fallacy tied to probabilistic confirmation " , _ synthese _ 184 : 1327 .
shafir , eldar b. , edward e. smith and daniel n. osherson ( 1990 ) , typicality and reasoning fallacies " , _ memory & cognition _ 18:229239 .
shogenji , tomoji ( 2012 ) , the degree of epistemic justification and the conjunction fallacy " , _ synthese _ 184 : 2948 . stolarz - fantino , stepanie , edmund fantino , daniel j. zizzo , and julie wen ( 2003 ) , the conjunction effect : new evidence for robustness " , _ american journal of psychology _ 116(1):1534 .
tentori , katya , nicolao bonini , and daniel osherson ( 2004 ) , the conjunction fallacy : a misunderstanding about conjunction " , _ cognitive science _ 28 : 467477 .
tentori , katya and vincenzo crupi ( 2008 ) , on the conjunction fallacy and the meaning of _ and _ , yet again : a reply to hertwig , benz , and krauss ( 2008 ) " , _ cognition _ 122 : 123134 .
tentori , katya and vincenzo crupi ( 2012 ) , how the conjunction fallacy is tied to probabilistic confirmation : some remarks on schupbach ( 2009 ) " , _ synthese _ 184 : 312 .
tentori , katya and vincenzo crupi ( 2013 ) , why quantum probability does not explain the conjunction fallacy " , commentary in pothos and busemeyer ( 2013 ) , p. 308 - 310 .
tentori , katya , vincenzo crupi and selena russo ( 2013 ) , on the determinants of the conjunction fallacy : probability versus inductive confirmation " , _ journal of experimental psychology : general _ 142(1 ) : 235255 .
tversky , amos and daniel kahneman ( 1982 ) , judgments of and by representativeness " , in d. kahneman , p. slovic & a. tversky ( eds ) , _ judgment under uncertainty : heuristics and biases _ , cambridge university press , p. 8498 .
( 1983 ) , extensional versus intuitive reasoning : the conjunction fallacy in probability judgment " , _ psychological review _
90(4 ) : 293315 .
von sydow , momme ( 2011 ) , the bayesian logic of frequency - based conjunction fallacies " , _ journal of mathematical psychology _ 55 : 119139
. walker , esteban and amy s. nowacki ( 2011 ) , understanding equivalence and noninferiority testing " , _ journal of general internal medicine _
26(2 ) : 192 - 196 .
wang , zheng j. and jerome r. busemeyer ( 2013 ) , a quantum question order model supported by empirical tests of an a priori and precise prediction " , _ topics in cognitive science _
5(4 ) : 689710 .
wang , zheng j. , tyler solloway , richard m. shiffrin and jerome r. busemeyer ( 2014 ) , context effects produced by question orders reveal quantum nature of human judgments " , _ proceedings of the national academy of science _
111(26 ) : 9431 - 9436 .
yukalov , vyacheslav i. and didier sornette ( 2010 ) , mathematical structure of quantum decision theory " , _ advances in complex systems _ 13 : 659698 .
yukalov , vyacheslav i. and didier sornette ( 2011 ) , decision theory with prospect interference and entanglement " , _ theory and decision _ 70 : 283328 .
the statistical test is to compare two conditional relative frequencies @xmath100 and @xmath101 , with the null hypothesis that they are equal .
the test is therefore @xmath102 where both @xmath100 and @xmath101 are observed as conditional relative frequencies .
+ testing equation [ yxtest ] is equivalent to test @xmath103 given that @xmath100 and @xmath101 are not equal to zero .
+ alternatively , we can formulate the test in terms of the log odds ratio ( or ) @xmath104 consider the first statistical test , @xmath105 we can thus test the following condition : @xmath106 or @xmath107 by expressing the conditional relative frequencies in terms of joint frequencies , that is , @xmath108 with @xmath109 and @xmath110 the @xmath100-components of the marginal frequencies of @xmath111 and @xmath112 , we obtain @xmath113 or simplifying @xmath114 we can thus test indifferently eq .
[ condprobtest ] or [ logortest ] .
+ given condition [ logortest ] , to perform the statistical test we suppose here that @xmath115 where @xmath116 is the standard error of the log odds ratio .
it is estimated as the square root of the sum of the inverse of all the joint frequencies that are considered in the estimation of the @xmath117 : + @xmath118 + finally , we also apply the continuity correction to the estimation of or , because the normal approximation to the binomial is used , which is effective in particular for small values of @xmath119 or @xmath120 : @xmath121 +
* linda " * : * text : linda a 31 ans , elle est clibataire , franche , et trs brillante .
elle est diplme en philosophie .
lorsquelle tait tudiante , elle se sentait trs concerne par les questions de discrimination et de justice sociale et avait aussi particip des manifestations anti - nuclaires . "
* @xmath4 : selon vous , linda est - elle fministe ? " * @xmath5 : selon vous , linda est - elle employe de banque ? " * text : bill a 34 ans .
il est intelligent , mais na pas dimagination , il est compulsif , et gnralement plutt teint .
lcole , il tait fort en mathmatiques , mais faible dans les sciences humaines et sociales . " * @xmath82 : selon
vous , bill est - il comptable ? "
* @xmath83 : selon vous , bill joue - t - il du jazz pour ses loisirs ? " * text : une enqute de sant a t mene en france sur un chantillon reprsentatif dhommes adultes de tous ges et de toutes professions .
dans cet chantillon , on a choisi au hasard monsieur f. * @xmath84 : selon vous , monsieur f. a - t - il dj eu une ou plusieurs attaques cardiaques ? "
* @xmath85 : selon vous , monsieur f. a - t - il plus de 55 ans ? " | human agents happen to judge that a conjunction of two terms is more probable than one of the terms , in contradiction with the rules of classical probabilities this is the conjunction fallacy .
one of the most discussed accounts of this fallacy is currently the quantum - like explanation , which relies on models exploiting the mathematics of quantum mechanics .
the aim of this paper is to investigate the empirical adequacy of major quantum - like models which represent beliefs with quantum states .
we first argue that they can be tested in three different ways , in a question order effect configuration which is different from the traditional conjunction fallacy experiment .
we then carry out our proposed experiment , with varied methodologies from experimental economics .
the experimental results we get are at odds with the predictions of the quantum - like models .
this strongly suggests that this quantum - like account of the conjunction fallacy fails .
future possible research paths are discussed . |
graded materials with spatial gradients in their structure @xcite are abundant in nature , which have received much attention as one of the advanced inhomogeneous composite materials in diverse engineering applications @xcite .
these materials can be made to realize quite different physical properties from the homogeneous materials , and thus , to some extent , more useful and interesting . for graded materials , the traditional theories @xcite for homogeneous materials do not work any longer .
recently , we presented a first - principles approach @xcite and a differential effective dipole approximation ( deda ) @xcite , to investigate the dielectric properties of the graded materials . to our interest ,
the two methods have been demonstrated in excellent agreement between each other @xcite . in the case of graded materials , the problem will become more complicated by the presence of nonlinearity inside them .
fortunately , for deriving the equivalent nonlinear susceptibility of graded particles , we have succeeded in putting forth a nonlinear differential effective dipole approximation ( ndeda ) @xcite .
as expected , this ndeda has also been demonstrated in excellent agreement with a first - principles approach @xcite .
in addition , the finite - frequency response of nonlinear composite materials have attracted much attention both in research and industrial applications during the last two decades @xcite .
when a composite with linear / nonlinear particles embedded in a linear / nonlinear host medium is subjected to a sinusoidal electric field , the electrical response in the composite will generally consist of alternating current ( ac ) fields at frequencies of higher - order harmonics @xcite .
in fact , a convenient method of probing the nonlinear characteristics of the composite is to measure the harmonics of the nonlinear polarization under the application of a sinusoidal electric field @xcite .
the strength of the nonlinear polarization should be reflected in the magnitude of the harmonics .
for the purpose of extracting such harmonics , the perturbation expansion @xcite and self - consistent methods @xcite can be used . in this work , based on the ndeda
, we shall investigate the effect of gradation ( inhomogeneity ) inside the particles ( inclusions ) on the ac responses of the graded composite by making use of a perturbation expansion method @xcite . here ,
the composite under consideration is composed of linear / nonlinear graded particles which are randomly embedded in a linear / nonlinear host medium in the dilute limit . to this end
, it is shown that the fundamental and third - order harmonic ac responses are sensitive to the dielectric - constant ( or nonlinear - susceptibility ) gradation profile within the particle .
thus , by measuring the ac responses of the graded composites , it is possible to perform a real - time monitoring of the fabrication process of the gradation profiles within graded particles . this paper is organized as follows . in the following section , we shall present the formalism , which is followed by the numerical results in section [ numericalresults ] . in section [ discon ] ,
the discussion and conclusion will be given .
let s consider nonlinear graded spherical particles with radius @xmath0 and dielectric gradation profile @xmath1 inside it , being embedded in a nonlinear host medium of dielectric constant @xmath2 , in the presence of a uniform external electric field @xmath3 along @xmath4axis . here
@xmath5 or @xmath6 ( @xmath7 or @xmath8 ) denotes the corresponding linear dielectric constant ( nonlinear susceptibility ) , @xmath9 and @xmath10 stands for the local electric field inside the particles and the host medium , respectively .
note both gradation profiles @xmath5 and @xmath7 are radial functions where @xmath11 throughout the paper , we shall consider the case of weak nonlinearity only , that is , @xmath12 and @xmath13 recently , we put forth a deda ( differential effective dipole approximation ) @xcite for calculating the equivalent dielectric constant @xmath14 @xcite of the spherical graded particle .
this deda receives the form @xmath15\cdot[\bep_1(r)+2\ep_1(r ) ] } { r\ep_1(r)}.\label{deda}\ ] ] note eq .
( [ deda ] ) is just the tartar formula , derived for assemblages of spheres with varying radial and tangential conductivity @xcite .
so far , the equivalent @xmath16 for the whole graded particle can be calculated , at least numerically , by solving the differential equation [ eq . ( [ deda ] ) ] , as long as @xmath5 ( dielectric - constant gradation profile ) is given .
once @xmath16 is determined , we can readily take one step forward to obtain the volume average of the linear local electric field inside the particles as @xmath17 hence , the deda [ eq . ( [ deda ] ) ]
offers a convenient way to obtain the local electric field [ eq . ( [ localfield ] ) ] .
it is worth remarking that the deda [ eq . ( [ deda ] ) ] is valid for arbitrary gradation profiles . to show the correctness of eq .
( [ localfield ] ) , we shall alternatively present a first - principles approach for calculating the local electric field inside the particle . for this purpose ,
let s take the power - law gradation profile ( @xmath18 ) as a model .
for this profile , the potential within the graded particle can be given by solving the electrostatic equation , @xmath19 @xcite , @xmath20 where the coefficient @xmath21 is determined by performing appropriate boundary conditions , @xmath22 and @xmath23/2.$ ] based on the relation between the linear local electric field and the potential ( @xmath24 ) , we have @xmath25\hat{{\bf z}}\},\label{fp}\end{aligned}\ ] ] where @xmath26 and @xmath27 are the unit vectors along @xmath28 , @xmath29 , and @xmath4axes , respectively .
so far , it is straightforward to obtain the volume average of the local electric field inside the particles , @xmath30 where @xmath31 is the volume of the spherical particles . in fig . 1
, we shall numerically compare eq .
( [ localfield ] ) ( local field predicted by the deda ) with eq .
( [ volume ] ) ( local field obtained from the first - principles approach ) . in a recent work @xcite , we have established an ndeda ( nonlinear differential effective dipole approximation ) , by deriving a differential equation for the equivalent nonlinear susceptibility @xmath32 , namely , @xmath33
+ \bar{\chi}_1(r)\cdot\frac{8y-3}{r } + \frac{3\chi_1(r)}{5r } \nonumber \\ & & \cdot \left(\frac{\bar{\epsilon}_1(r)+2\epsilon_1(r)}{3\epsilon_1(r)}\right)^4(5 + 36x^2 + 16x^3 + 24x^4).\label{ndeda1}\end{aligned}\ ] ] where @xmath34\cdot[\bep_1(r)-\ep_1(r)]}{\ep_1(r)[\bep_1(r)+2\ep_2]}.\ ] ] similarly , @xmath35 can be obtained , at least numerically , by solving eq .
( [ ndeda1 ] ) , once the initial conditions , namely , @xmath36 and @xmath37 are given . in
what follows , we can investigate the nonlinear ac response of the graded spherical particle by seeing it as a homogeneous particle having the constitutive relation between the displacement and the local electric field , @xcite @xmath38 where @xmath16 and @xmath35 are determined by eqs .
( [ deda ] ) and ( [ ndeda1 ] ) , respectively . for the sake of convenience , we shall represent @xmath16 by @xmath39 , @xmath40 by @xmath41 as well as @xmath42 by @xmath43 , if no special instructions .
if we apply a sinusoidal electric field like @xmath44 the local electric field @xmath45 and the induced dipole moment @xmath46 will depend on time sinusoidally , too . here
the effective dielectric constant of the system ( @xmath47 ) is given by the following dilute - limit expression , @xmath48 where @xmath49 is the volume fraction of the particles . by virtue of the inversion symmetry ,
the local electric field is a superposition of odd - order harmonics such that @xmath50 similarly , the induced dipole moment contains harmonics as @xmath51 these harmonics coefficients can be extracted from the time dependence of the solution of @xmath45 and @xmath52 in what follows , we will perform a perturbation expansion method to extract the third harmonics of the local electric field and the induced dipole moment .
it is known that the perturbation expansion method is applicable to weak nonlinearity only , limited by the convergence of the series expansion .
let s start from the dilute - limit expression for the effective linear dielectric constant ( @xmath53 ) of the system of interest , namely , eq .
( [ dl - expre ] ) where @xmath54 .
next , we obtain the linear local electric fields @xmath55 and @xmath56 , respectively , @xmath57 in view of the existence of nonlinearity inside the two components , we readily obtain the following local electric fields for the nonlinear particle and host , respectively , @xmath58 for the below series expansions , we will take @xmath59 and @xmath60 in eqs .
( [ nonlineare1 ] ) and ( [ nonlineare2 ] ) , where @xmath61 denotes the volume average of @xmath62 let s expand the local electric field @xmath63 and @xmath56 into a taylor expansion , taking @xmath64 and @xmath65 as the perturbative quantities , @xmath66 keeping the lowest orders of @xmath64 , we can rewrite eq .
( [ expan - e1 ] ) as , @xmath67 where @xmath68}{(1-p)(\bep_1 + 2\ep_2)^5}.\nonumber\end{aligned}\ ] ] because of the time - dependence of the electric field [ eq . ( [ time - depend ] ) ] , we can take one step forward to obtain the local electric field in terms of the harmonics ( @xmath69 and @xmath70 ) , @xmath71 where @xmath72 similarly , based on eq .
( [ dm ] ) , we obtain the induced dipole moment in terms of the harmonics ( @xmath73 and @xmath74 ) , @xmath75 where @xmath76 with @xmath77}{(\bep_1 + 2\ep_2)^3},\nonumber\\ k_3 & = & \frac{j_1\chi_2[(1 + 3f)\bep_1{}^3 - 18f\bep_1{}^2\ep_2 - 3(2 - 3f)\bep_1\ep_2{}^2 - 2(2 - 3f)\ep_2{}^3 ] } { ( \bep_1 + 2\ep_2)^3}\nonumber.\end{aligned}\ ] ] in the above derivation , we have used an identity @xmath78
for numerical calculations , we take @xmath79 , and other parameters : volume fraction @xmath80 external field strength @xmath81 linear part of host dielectric constant @xmath82 the validation of using the deda is shown in fig . 1 .
in this figure , we investigate the linear local electric field by using a power - law gradation profile inside the particles , in an attempt to compare the deda with the first - principles approach .
as expected , the excellent agreement is demonstrated between the deda [ eq .
( [ localfield ] ) ] and the first - principles approach [ eq . ( [ volume ] ) ] .
in addition , it is worth noting that , for a linear gradation profile within the graded particles , the first - principles approach holds as well @xcite , and the excellent agreement between the two methods can also be found ( figure not shown here ) .
next , we discuss a power - law gradation profile @xmath83,$ ] see fig . 2 .
in this figure , the harmonics of local electric field and the induced dipole moment are investigated as a function of @xmath84 for various @xmath85 in this case , increasing @xmath84 ( or decreasing @xmath86 ) leads to increasing @xmath41 ( namely , the equivalent dielectric constant of the graded particle under consideration ) and in turn yields a decreasing local electric field inside the particle .
thus , either an increase in @xmath84 or a decrease in @xmath86 leads to the weakening third - order harmonics @xmath87 and @xmath88 as displayed in fig . 2 .
the @xmath28axes of figs .
3 and 4 represent the slope ( @xmath89 ) of a linear gradation profile .
it is because during the fabrication of graded spherical particles by using diffusion , the dielectric constant at the center @xmath90 may vary slightly while that at the grain boundary can change substantially . for a linear gradation profile @xmath91,$ ]
3 shows the harmonics as a function of @xmath89 for various @xmath92 in this case , increasing @xmath89 or @xmath90 yield increasing @xmath93 and hence one obtains the decreasing local electric field . as a result ,
larger @xmath89 or @xmath94 weaker the third - order harmonics @xmath87 and @xmath88 see fig .
3 . fig .
4 displays the effect of @xmath8 on the harmonics , for a linear gradation profile @xmath91.$ ] here , increasing @xmath8 leads to increasing the local electric field inside the graded particle of interest .
therefore , the third - order harmonics @xmath87 and @xmath95 increase for increasing @xmath96 as mentioned above , as @xmath84 and @xmath89 increases , the equivalent dielectric constant of the particle should be increased accordingly , which in turn yields a decreasing local electric field , and hence , in figs . 2@xmath974
, @xmath69 decreases for increasing @xmath84 or @xmath98 on the other hand , it is found that , in figs . 2@xmath974 , @xmath73 increases for increasing @xmath84 or @xmath89 which is , in fact , due to the increasing effective dielectric constant @xmath47 [ refer to eq .
( [ dm ] ) ] .
similarly , this analysis works fairly for understanding the effect of @xmath86 and @xmath90 on @xmath69 and @xmath73 , as displayed in figs . 2 and 3 .
however , increasing @xmath8 can increase not only the local electric field inside the particles , but also the effective dielectric constant @xmath99 and hence we observe increasing @xmath69 and @xmath100 as shown in fig .
in addition , we also discuss the effect of nonlinear - susceptibility gradation profiles ( no figures shown here ) . for linear gradation profile @xmath79 ,
as @xmath101 ( or @xmath102 ) increases , the third harmonics of both the electric field and the induced dipole moment increases accordingly . on the other hand , for pow - law gradation profile @xmath103 , increasing @xmath104 ( or decreasing @xmath105 ) leads to increasing third harmonics . to understand such results
, we can again resort to the above analysis on the effect of the relevant parameters on the local field as well as the effective dielectric constant .
here some comments are in order .
we investigate the nonlinear ac responses of the graded material where linear / nonlinear graded particles are randomly embedded in a linear / nonlinear host medium in the dilute limit . as a matter of fact ,
the ndeda ( nonlinear differential effective dipole approximation ) is valid for arbitrary gradation profiles , besides the power - law and linear profiles of interest .
in particular , based on the first - principles approach , the exact solution is obtainable , for not only power - law profiles ( see section [ comparison ] ) , but also linear profiles ( refer to ref .
@xcite ) . as an extension ,
it is of particular interest to see what happens to the nonlinear ac responses of graded particles in electrorheological fluids , in which a field - induced anisotropic structure often occurs . for discussing this anisotropy effect
, we can make use of the maxwell - garnett approximation for anisotropic structures @xcite . to sum up , based on our recently - established ndeda
, we have investigated the nonlinear ac responses of a composite with linear / nonlinear graded spherical particles embedded in a linear / nonlinear host medium , and found the fundamental and third harmonic ac responses are sensitive to the dielectric - constant ( or nonlinear - susceptibility ) gradation profile within the particles .
again , for extracting the linear local electric field , the deda agrees very well with the first - principles approach .
thus , by measuring the ac responses of the graded composites , it is possible to perform a real - time monitoring of the fabrication process of the gradation profiles within graded particles .
this work has been supported by the research grants council of the hong kong sar government under project numbers cuhk 4245/01p and cuhk 403303 , the national natural science foundation of china under grant no .
10204017 ( l.g . ) , the natural science of jiangsu province under grant no .
bk2002038 ( l.g . ) , and national `` 863 '' project of china under grant number 2002aa639270 ( g.q.g . ) . | when a composite of nonlinear particles suspended in a host medium is subjected to a sinusoidal electric field , the electrical response in the composite will generally consist of alternating current ( ac ) fields at frequencies of higher - order harmonics .
the situation becomes more interesting when the suspended particles are graded , with a spatial variation in the dielectric properties .
the local electric field inside the graded particles can be calculated by the differential effective dipole approximation , which agrees very well with a first - principles approach . in this work ,
a nonlinear differential effective dipole approximation and a perturbation expansion method have been employed to investigate the effect of gradation on the nonlinear ac responses of these composites .
the results showed that the fundamental and third - harmonic ac responses are sensitive to the dielectric - constant and/or nonlinear - susceptibility gradation profiles within the particles .
thus , by measuring the ac responses of the graded composites , it is possible to perform a real - time monitoring of the fabrication process of the gradation profiles within the graded particles . *
pacs number(s ) : * 77.84.lf , 77.22.ej , 42.65.-k , 42.79.ry |
recently , some experiments have reported narrow signals in the vicinity of 1530 mev in the @xmath7 and @xmath8 invariant mass spectra which are consistent with the exotic pentaquark baryon state @xmath9 with quark content @xmath10 @xcite , while other experiments have searched for this state with negative results . the possible existence
a charm pentaquark has also been discussed , with renewed theoretical interest in calculating their expected properties @xcite following the observation of strange pentaquarks .
the results of pentaquark searches by the h1 and the zeus collaboration are presented in this report .
the analyses are based on data samples with an integrated luminosity of 75 pb@xmath11 and 126 pb@xmath11 taken by the h1 and the zeus collaborations in the years 1994 - 2000 when hera collided electrons or positrons with 27.6 gev and protons with 820 or 920 gev .
understanding the process whereby quarks and gluons convert to colourless hadrons is one of the outstanding problems in particle physics .
the measurements of the inclusive photoproduction of the neutral mesons @xmath0 , @xmath1 , @xmath2 and @xmath3 could contribute to resolving this problem .
a strange pentaquark @xmath9 candidate was seen by the zeus collaboration . in this analysis deep inelastic scattering ( @xmath12 ) events
were selected by requiring an exchanged photon virtuality @xmath13 gev@xmath14 .
the @xmath9 was reconstructed in the decay channel to @xmath15 .
( 6.0,5.7 ) in fig .
[ theta ] , the invariant mass of proton and @xmath16 is presented for events with @xmath17 gev@xmath14 .
this distribution was fitted with a polynomial background and two gaussians .
the signal peak is observed with a mass of @xmath18 mev and the measured gaussian width @xmath19 mev which is consistent with the detector resolution .
the significance of the signal is 4.6 sigma .
the second gaussian significantly improves the fit in the low mass region , and has a mass of about 1470 mev and a width of 16 mev and may correspond to the @xmath20 for which the evidence of existence is poor .
the invariant - mass spectrum was investigated for the @xmath15 and @xmath21 separately .
the signal has been seen in both charges with significance of 3 @xmath22 .
if the signal corresponds to the @xmath9 pentaquark , this provides the first evidence for its antiparticle .
the measured total cross section for the @xmath9 in the kinematic region @xmath17 gev@xmath14 , @xmath23 gev , @xmath24 and @xmath25 is @xmath26 pb . the properties of the @xmath9 candidate were studied and it was found to be produced predominantly in the forward pseudorapidity region .
a similar analysis was done by the h1 collaboration and no peak is visible near 1520 mev .
the resulting upper limit on the @xmath9 production cross section was found to vary between 40 and 120 pb over the mass range of 1.48 to 1.7 gev and does not exclude the previously measured cross section at zeus .
( 10.0,5.8 ) na49 collaboration published results of double strange pentaquark @xmath27 searches in the @xmath28 invariant mass spectrum @xcite .
they have observed a narrow peak with mass of about 1860 mev .
a similar analysis was repeated using zeus @xmath12 data .
the @xmath27 was reconstructed from @xmath29 decay .
zeus observed a clean signal of @xmath30 in @xmath12 events with @xmath13 gev@xmath14 , but no signal with a mass around 1860 mev is observed ( fig .
a similar analysis was performed for @xmath17 gev@xmath14 , the kinematic region where the @xmath9 state was most clearly observed by zeus . again
no signal is observed near 1860 mev .
the number of @xmath30 signal events reconstructed in this analysis is about the same as for na49 data .
however , it should be noted that na49 is a fixed target experiment , which has good acceptance in the forward region .
the non - observation of this signal in the central - fragmentation region in the zeus data does not necessarily contradict the observation of a signal predominantly produced in the forward region .
a search for a charm pentaquark @xmath31 candidate was carried out by h1 using @xmath12 data at @xmath13 gev@xmath14 .
the @xmath31 was reconstructed via its decay to @xmath32 , where the @xmath33 is reconstructed using the decay channel @xmath34 .
a clear and narrow resonance is observed for both @xmath35 and @xmath36 combinations with an invariant mass of @xmath37 mev ( fig .
[ thetac ] ) with a significance of 5.4@xmath22 .
the measured width of the resonance is @xmath38(stat . )
mev , consistent with the experimental resolution . a signal with compatible mass and width
is also observed in an independent photoproduction data sample .
( 6.0,3.5 ) an acceptance corrected ratio @xmath39 @xmath40 was measured .
when extrapolating to the full phase space of the decay products one observes a ratio of the visible cross section of @xmath41 @xmath40 .
the region of @xmath42 in which the signal is observed contains a richer yield of @xmath33 mesons and exhibits a harder proton candidate momentum distribution than is the case for side bands in @xmath43 .
compared to inclusive @xmath33 production the @xmath44 production seems to be suppressed in the close to central rapidity regions in laboratory and centre - of - mass frames .
a similar analysis was done by zeus using higher statistics and reconstructing @xmath33 mesons in two channels @xmath34 and @xmath45 .
no signal near 3100 mev is observed .
zeus estimated the upper limit on the acceptance corrected ratio and it is equal 0.59 @xmath40 ( 0.51 @xmath40 for both @xmath33 decay channels ) in @xmath12 .
besides exotic searches for particles the production of well - known hadrons such as pions , @xmath46 , @xmath47 , protons , charm mesons , @xmath48 , _ etc .
_ are measured by zeus and h1 .
a recent result is the cross section measurement of inclusive photoproduction of @xmath0 , @xmath1 , @xmath2 and @xmath3 mesons at h1 in the central rapidity region . in this analysis
a photoproduction data sample taken in the year 2000 corresponding to an integrated luminosity of 38.7 pb@xmath11 and average @xmath49210 gev was used .
the @xmath1 , @xmath2 and @xmath3 mesons were reconstructed though @xmath50 decay , @xmath0 meson though @xmath51 decay .
( 6.0,5.7 ) in fig .
[ sigma ] , the measured differential cross section @xmath52 of the resonances as a function of @xmath53 , where @xmath54 is a spin and @xmath55 is a mass of measured particle , was compared with the cross section of charged pions .
the resonances have a similar behavior as observed for long - lived hadrons @xcite .
the zeus collaboration observes a narrow baryonic state at a mass of about 1520 mev that is interpreted as strange pentaquark @xmath9 .
h1 does not observe this state but the upper limit on the production cross section does not exclude the zeus observation .
a resonance search has been performed in the @xmath32 invariant - mass spectrum with the h1 and the zeus detector .
h1 observes a @xmath56 candidate at mass of about 3100 mev .
zeus does not see any signal near this mass value .
the upper limit on the @xmath56 production contradicts the h1 observation .
more data from hera_ii _ which are now being taken should resolve this contradiction .
the inclusive cross section for @xmath0 , @xmath1 , @xmath2 and @xmath3 was measured and has the similar behavior as observed for long - lived hadrons @xcite .
this work was partially supported by grants rfbr 03 - 02 - 17291 , 04 - 02 - 16445 and crdf rec-011 .
i would like to thank dmitry ozerov and andrey rostovtsev for useful discussion . | recent results on spectroscopy with special focus on searches for pentaquarks are presented from the h1 and zeus collaborations .
cross sections of observed states and upper limits on the production cross section of unobserved states are extracted in order to enable comparison between experiments .
measurements of the inclusive photoproduction of the neutral mesons @xmath0 , @xmath1 , @xmath2 and @xmath3 in @xmath4 interactions at hera at an average @xmath5 collision energy of @xmath6 gev are also presented . |
it is well know that high - mass stars form in dense clusters @xcite .
the research on embedded young clusters around young intermediate- to high - mass stars has advanced significantly via near - infrared surveys ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the _ spitzer _
space telescope with its unprecedented sensitivity at mid - infrared wavelengths provides a new means for identifying young stellar objects ( ysos ) in star forming regions through detection of dusty disks and envelopes .
high sensitivity in the @xmath4 @xmath0 m bands is invaluable for distinguishing young stars from reddened background stars through infrared excess arising from circumstellar materials @xcite .
this capability is particularly valuable for studies of high - mass star forming regions for two reasons : 1 ) such regions are typically located near the galactic plane where the density of background stars is high and 2 ) the member ysos are often faint due to the large distances to these regions ( @xmath5kpc ) .
the faint magnitude of the ysos complicates not only the detection of members but also their identification since the density of background stars rises rapidly with increasing magnitudes . with the infrared array camera ( irac ) and multiband imaging photometer for _ spitzer _ ( mips ) , a significant census of ysos
can be carried out , which provides a fossil record of the distribution of star formation sites and information on the environments in which stars and planets form .
this method is particularly effective for finding extended , lower density distributions of stars surrounding the dense regions of embedded clusters .
extended structures , such as reflection nebulae and outflows , can also be mapped with the irac .
outflows emanating from the central young stars or protostars are usually detectable in rotational transitions of co and other molecules at ( sub)millimeter wavelengths . in irac observations
, outflows can be revealed from shocked h@xmath2 emission .
hydrodynamic simulations of shock models predict that h@xmath2 emission is particularly strong in the 4.5 @xmath0 m band @xcite .
since the 4.5 @xmath0 m band is relatively free of the emission from hydrocarbons ( probably in the form of polycyclic aromatic hydrocarbons , or pahs ) compared with the other three bands , and the 5.8 and 8.0 @xmath0 m bands are significantly less sensitive than the shorter wavelength bands , composite images with emission from different bands coded in different colors can be a diagnostic tool for h@xmath2 emission in outflows .
a remarkable example of such a detection can be found in the irac observations of the hh 46/47 outflow , a collimated bipolar outflow emanating from a low - mass protostar , where the southwestern lobe was clearly seen in the 4.5 @xmath0 m band as a limb - brightened bow - shock cavity @xcite .
@xcite reported the irac detection of a spectacular outflow in the high - mass star forming region dr 21 by virtue of its brightness in the 4.5 @xmath0 m band .
further investigations of the _ iso _ sws spectra of the dr 21 outflow verified that the h@xmath2 lines contribute significantly to all four irac bands and the emission in the 4.5 @xmath0 m band is mostly attributed to the h@xmath2 v=0 - 0 @xmath6 ( 4.695 @xmath0 m ) line .
outflow activities can also be revealed from scattered light in the irac bands . by comparison of the deep irac imaging of l1448 with the radiative transfer modelling
, @xcite demonstrated that the observed infrared emission in the irac bands is consistent with the scattered light from the central low - mass protostars escaping the cavities carved by molecular outflows .
@xcite reprocessed the irac observations of the hh 46/47 outflow using a deconvolution algorithm and found a wide - angle biconical component in scattered light in addition to the previously reported bow - shock cavity .
finally , in some cases hydrocarbon emission is visible from outflow cavities , presumably due to the illumination of the cavity walls by uv radiation from the central source @xcite .
we have observed a sample of nine high - mass star forming regions with the irac and mips instruments ( pid : 3528 , pi : qizhou zhang ) .
these regions were chosen according to the following criteria : 1 ) already imaged with millimeter interferometers ; 2 ) known to have molecular outflows ; 3 ) outside of the glimpse survey except iras 19410 + 2336 ; 4 ) at a distance of @xmath7kpc ( @xmath8au at the resolution of irac ) .
these regions exhibit far - infrared luminosities ranging from @xmath9 to @xmath10 ( see table [ source ] for source parameters ) .
we carry out a census of ysos toward each region , and simultaneously , zoom in on the central massive star formation sites with the aim of searching for infrared counterparts of millimeter continuum sources and infrared emission from outflows . in this paper
we present initial results of the observations . in 2 , we describe the observations .
we present observational results in 3 and subsequent discussion in 4 .
a brief summary is given in 5 .
the irac @xcite observations were obtained from february to september 2005 except iras 19410 + 2336 , which was observed in october 2004 .
each region was observed in the high dynamic range mode with 0.4 second and 10.4 second integration times per dither , and 16 dithers per map position .
the area covered by all four bands is roughly @xmath1 .
the total effective integration time per pixel is 166.4 seconds .
the 0.4 second integration frames were used to obtain photometry for bright sources .
the frames were processed by the spitzer science center ( ssc ) using the standard pipeline version s13.2 to produce the standard basic calibrated data ( bcd ) products .
additional bright source artifacts ( `` jailbar '' , `` pulldown '' , `` muxbleed '' , and `` banding '' ) were removed or mitigated using customized idl scripts developed by the irac instrument team @xcite .
there are often `` bandwidth '' artifacts in the 5.8 and 8.0 @xmath0 m images , where a decaying trail of the 4th , 8th , and in some cases even 12th pixels to the right of a bright or saturated spot is found .
currently there is no correction for this effect .
mosaics were created at the native instrument resolution of @xmath11 per pixel from the bcd frames using gutermuth s wcsmosaic idl package @xcite .
since the irac data is under - sampled and the point - spread - function ( psf ) varies in shape with position and signal , aperture photometry is currently preferred over the psf - fitting photometry .
source finding and aperture photometry were performed using gutermuth s idl photometry and visualization tool photvis , version 1.10 @xcite .
the radii of the source aperture , the inner and outer boundaries of the sky annulus were set to @xmath12 , @xmath12 and @xmath13 , respectively .
irac photometry was calibrated using large - aperture measurements of several standard stars from observations obtained in flight @xcite .
there is a @xmath14 calibration uncertainty due to a position dependent gain for point sources which is not corrected by the flat field .
it is the dominant source of uncertainty in the photometry of irac images .
fluxes at zero mag are 280.9 jy at 3.6 @xmath0 m , 179.7 jy at 4.5 @xmath0 m , 115.0 jy at 5.8 @xmath0 m , and 64.1 jy at 8.0 @xmath0 m
. magnitudes for 1 dn / s are set to 19.455 at 3.6 @xmath0 m , 18.699 at 4.5 @xmath0 m , 16.498 at 5.8 @xmath0 m , and 16.892 at 8.0 @xmath0 m , using standard aperture corrections for the radii adopted ( see table 5.7 in irac data handbook v3.0 ) .
since regions of active star formation often have bright nebulosity in the @xmath15 and @xmath16 m bands , we rejected detections with photometric uncertainties above 0.25 mag in these two bands , while for the @xmath17 and @xmath18 m bands detections with uncertainties above 0.2 mag were rejected .
the number of detected sources and the limiting magnitude for each band toward each region are listed in table [ observation ] .
the mips @xcite observations were undertaken in the photometry mode during campaigns separated by less than 1 month from the irac observations with the exception of iras 22172 + 5549 ( 1.5 months separation ) and hh 80 - 81 ( 5 months separation ) .
the target sources were mapped in the 24 and @xmath19 m bands .
the mips 70@xmath20 m images were severely saturated and not included in this paper .
the 24@xmath20 m observations consist of 400 frames with an integration of 2.62 seconds per frame , resulting in a total effective integration time of 1048 seconds .
the frames were processed by the ssc standard pipeline s16.0 for afgl 5142 , iras 05358 + 3543 , and g192.16 - 3.82 , and by pipeline s16.1 for the other regions .
mosaics were created based on the bcd frames using the ssc mopex package , with a pixel size of @xmath21 ( half of the native instrument resolution ) .
the central brightest sources in all the regions except iras 22172 + 5549 were significantly saturated .
unsaturated point sources with peaks more than 10 times the rms noise level were identified using photvis 1.10 .
aperture photometry with a 5 pixel aperture and a sky annulus from 12 to 15 pixels was performed on these sources . multiplying by @xmath22 to correct for the smaller mosaic pixels and by 1.708 to correct from a 5 pixel aperture to infinite , and using a zero mag flux of 7.3 jy
, the aperture photometry was converted to magnitudes @xcite . the detections with photometric uncertainties above 0.25 mag were rejected .
the number of detected sources and the limiting magnitude toward each region are listed in table [ observation ] .
the psf - fitting photometry is preferred in crowded fields with multiple point - sources .
while in our mips 24 @xmath0 m images , only about 10 to 20 unsaturated objects were detected in each field with a sparse distribution , and in some fields the psf - fitting photometry is limited by the lack of ideal psf - generating stars which are supposed to be bright , unsaturated , and uncontaminated by nebulosity . to further evaluate the aperture photometry quantitatively , we performed psf - fitting photometry on all the 24 @xmath0 m detections using the idl daophot package .
the reference psf was generated from one to two objects chosen from the available in the field to be most suitable for psf - generating .
we then compared the returns from the psf - fitting with the aperture photometry and found that the majority of unsaturated objects agree with each other in 0.2 mag ; a few objects show differences of 0.3 to 0.4 mag but these objects are relatively faint and located in a spatially varying background , thus the differences are probably due to background subtraction , a limitation to both aperture and psf - fitting photometry .
therefore , in table [ photometry ] we show the aperture photometry with the exception of four objects .
these four objects are blended with a neighboring brighter object ; the resulting magnitudes derived from the contaminated aperture photometry is consequently 0.5 to 0.6 mag brighter than those determined with the psf - fitting photometry .
we show the psf - fitting photometry for these objects and have annotated the table to identify these objects .
in addition , there are three slightly saturated objects in our sample ( annotated in table [ photometry ] ) ; we derived their magnitudes by manually scaling the psf , subtracting the psf from the image , and comparing the averaged residual of the psf wing with a sky value taken from an annulus of 12 to 15 pixels .
the tabulated magnitudes give an averaged residual closest to the sky value .
since raising / lowering the magnitudes by 0.1 would give an averaged residual distinctly higher / lower than the sky value , we adopt an uncertainty of 0.1 mag for these objects .
in this section , we carry out a census of ysos and classify them using their positions in near- to mid - infrared color - color diagrams .
the results of this analysis are shown in figure [ fig1 ] , where we display the colors of the detected ysos in all nine regions , and figure [ fig2 ] , where we show color - composite images of the regions with the identified ysos overlaid .
we list the number of identified ysos in table [ source ] , and present the multi - band photometry of each yso in table [ photometry ] .
although the criteria we have used to identify ysos were developed primarily for low - mass stars , we do not select sources in specific mass ranges . the census will be dominated by low - mass ysos , but intermediate - mass ysos , such as herbig ae / be stars with disks and intermediate - mass protostars , may be identified by the same method . since we are observing high - mass star forming
regions , high - mass young stars and protostars certainly exist in our fields .
candidate massive protostellar objects in the irac bands have been identified with the glimpse survey @xcite ; however , the colors of those objects are less well understood .
it is unclear whether this method is effective for identifying high - mass stars with disks or envelopes . in our sample
, there are twelve objects exhibiting obvious infrared excess in one or more color - color diagrams and also showing evidence of being an intermediate- to high - mass star ( labelled as `` high '' in the last column of table [ photometry ] and marked with blue crosses in figure [ fig2 ] ) .
these objects show no or weak dust emission in the millimeter continuum ( except the exciting star of the hii region w75n ( a ) , which is within a warm dust shell ) and appear very bright in near - infrared , suggesting the lack of dense envelopes thus more evolved than typical protostars .
part of these objects are identified as candidate exciting sources of the surrounding uv - heated reflection nebulae ( [ nebulae ] ) .
in contrast , the youngest high - mass stars in our sample are still deeply embedded in dense dusty cores which were identified as millimeter continuum sources in previous interferometric observations .
most of these millimeter sources are not detected shortward of 3.6 @xmath0 m , and in some cases even undetectable at longer wavelengths . given the extra level of complication and uncertainty in identifying and classifying these youngest high - mass objects , we exclude these objects from this section and discuss them in more detail in [ msf ] .
@xcite and @xcite have established a ysos classification scheme in the irac color - color diagram [ 3.6]-[4.5 ] vs. [ 5.8]-[8.0 ] .
this scheme was confirmed by @xcite with an irac survey of a sample of known taurus pre - main - sequence stars , even though some ambiguities were found in a few cases .
here we adopt this scheme to classify ysos detected in all four irac bands .
the [ 5.8]-[8.0 ] color is insensitive to reddening and is effective at distinguishing between highly reddened background stars with photospheric emission only and young stars with circumstellar material .
thus we require @xmath23-[8.0]>0.2 $ ] for all the identified ysos .
bright nebulosity dominated by hydrocarbon emission features can be found in all the regions and have a very red [ 5.8]-[8.0 ] color . to assess the impact of nebulosity on creating false detections , we added artificial stars to each of the mosaics .
each artificial star was 10 pixel ( @xmath24 ) to the west of an actual source that was detected in the 3.6 and 4.5 @xmath0 m bands , and the magnitude of the artificial star in all four irac bands was set to the 3.6 @xmath0 m magnitude of the actual source .
thus , the artificial stars had @xmath25-[4.5]$ ] and @xmath23-[8.0]$ ] colors equal to @xmath26 .
these sources accurately sampled the range of magnitudes and background nebulosities found in the image .
we then recovered the photometry to determine how many such stars would be mistaken as having infrared excesses due to the nebulosity .
although there were about 10 to 30 sources with @xmath23-[8.0]>0.2 $ ] , only 1 to 3 of the sources had a color of @xmath25-[4.5]>0.2 $ ] .
hence we require @xmath25-[4.5]>0.2 $ ] to eliminate pure photospheres contaminated by hydrocarbon emission nebulosity .
figure [ fig1]a shows the [ 3.6]-[4.5 ] vs. [ 5.8]-[8.0 ] diagram for objects merged from all 9 regions and detectable in all four irac bands . in this diagram , objects with @xmath27-[4.5]<0.8 $ ] and @xmath28-[8.0]<1.1 $ ] are classified class ii ; objects with @xmath25-[4.5]>0.8 $ ] and @xmath23-[8.0]>0.2 $ ] , or , @xmath25-[4.5]\geq0.4 $ ] and @xmath23-[8.0]>1.1 $ ] are classified protostars ( including class 0 and class i objects ) ; while objects with @xmath23-[8.0]>1.1 $ ] , which is consistent with that of protostars , but with @xmath27-[4.5]<0.4 $ ] , which is lower than that of protostars , are classified class i / ii . for those detected in the 24 @xmath0 m band , we re - examine them in the irac - mips diagram . from both observations and model calculations , the [ 3.6]-[4.5 ]
vs [ 5.8]-[8.0 ] diagram has been proven quite effective in separating ysos with dusty disks or envelops from purely photospheric emission stars @xcite . the principal ambiguity for the [ 3.6]-[4.5 ] vs [ 5.8]-[8.0 ] classification scheme lies in distinguishing protostars from class ii objects .
from the direction of the reddening vector in figure [ fig1]a , the identified protostars immediately above the class ii rectangle ( @xmath29-[4.5]\lesssim1.2 $ ] ) are likely reddened class ii objects . in addtion ,
some of the objects immediately right to the class ii rectangle ( @xmath30-[8.0]\lesssim1.5 $ ] ) , which are identified as protostars or class i / ii , are likely class ii objects contaminated by hydrocarbon emission . in addition
, some intrinsic characteristics of ysos make the classification ambiguous as well .
for example , geometric effects including bipolar outflow cavities , flattened pattern of envelops and pole - on geometry will all make the irac colors of protostars much bluer @xcite . in a few cases
the discrimination between class ii objects and class iii / field stars is also ambiguous ; some ysos with circumstellar disks have an evacuated or very optically thin inner hole and show little infrared excess shortward of 8.0 @xmath0 m and then can be misidentified as class iii / field stars .
the irac 5.8 and 8.0 @xmath0 m bands are far less sensitive than the 3.6 and 4.5 @xmath0 m bands ( see the limiting magnitudes in table [ observation ] ) , and are often dominated by bright nebulosity in observations of active star forming regions . for this reason , we combine the irac photometric results with data from the 2mass _ jhks _ point source catalog .
we limit the _ jhks _ photometric uncertainty to be less than 0.1 mag .
since the 4.5 @xmath0 m band is the most sensitive band to ysos among all four irac bands @xcite , we use the @xmath31 vs. @xmath32 $ ] and @xmath33 vs. @xmath34 $ ] diagrams ( figures [ fig1]b , [ fig1]c ) to identify objects with infrared excess emission but not detected in the irac longer wavelength bands . in figures
[ fig1]b and [ fig1]c ,
objects with the @xmath32 $ ] or @xmath34 $ ] colors more than @xmath35 beyond the reddening vectors are identified as objects with intrinsic infrared excess .
the extinction law in the mid - infrared wavelengths toward star - forming regions , in particular toward distant high - mass star forming regions , has not been well established .
most recently , @xcite derived the interstellar extinction laws in irac bands toward five nearby star - forming regions .
we adopt the average of the extinction laws derived by flaherty et al . for our sample . from the dispersion of the objects left to the reddening vectors in these two diagrams
, it seems the adopted reddening law is applicable for our high - mass star forming regions .
the positions offsets of the reddening vectors in these two diagrams , which were determined by @xcite for serpens , also appear to be a good approximation for our regions . however , the completeness of the ysos identification from these two diagrams are significantly limited by the modest sensitivity of the 2mass survey .
deeper near - infrared observations will not only improve the completeness but also help to derive a more accurate reddening law toward each region , which would improve the reliability of the ysos identification . including data at wavelengths longward of 20 @xmath0 m is valuable for more reliable classification of ysos @xcite . in figure [ fig1]d
, we use the [ 3.6]-[4.5 ] vs. [ 4.5]-[24 ] diagram to classify ysos which are detectable both in the two most sensitive irac bands and in the mips 24 @xmath0 m band .
sources with spectral indexes greater than -0.3 have irac - mips colors @xmath25-[4.5]>0.652 $ ] and @xmath36-[24]>4.761 $ ]
. these sources would include both the flat spectrum and class i objects in the classification scheme of @xcite .
in addition , the high sensitivity of _ spitzer _ data allows for the detection of previously identified class 0 sources in mid - infrared @xcite ; consequently , class 0 sources can also be identified by these criteria . since
an infalling envelope is required to explain the seds of flat spectrum , class i and class 0 sources @xcite , we refer to the combined set of these sources as protostars . in figure [ fig1]d , there is clearly an isolated group close to ( 0 , 0 ) .
the objects in this group are class iii / field stars whose emission is dominantly from photospheres . in this diagram
objects outside of the protostar region and not in the class iii / field star group are classified as class ii .
there is an obvious gap between protostars / class ii objects and classiii / field stars , suggesting this diagram is very effective in distinguishing ysos with circumstellar materials from purely photospheric objects . from the objects that can be classified both in the [ 3.6]-[4.5 ] vs. [ 5.8]-[8.0 ] and in the [ 3.6]-[4.5 ] vs. [ 4.5]-[24 ] diagrams
, we find that the two classification schemes are in general consistent , otherwise we adopt the latter if there is an ambiguity . with the irac - mips diagram
we also identify 15 class ii objects and 2 protostars that can not be identified in any of the other three color - color diagrams .
there are mainly two classes of extragalactic contaminants that can be misidentified as ysos @xcite .
one is star forming galaxies and narrow - line active galactic nuclei ( agn ) which have growing excess at 5.8 and 8.0 @xmath0 m due to hydrocarbon emission .
the other is broad - line agn which have irac colors very similar to that of bona fide ysos . @xcite
have developed a method based on the bootes shallow survey data to substantially mitigate these contaminants . in this method ,
hydrocarbon emission sources , including galaxies and narrow - line agn , can be eliminated by their positions in the [ 4.5]-[5.8 ] vs. [ 5.8]-[8.0 ] and [ 3.6]-[5.8 ] vs. [ 4.5]-[8.0 ] diagrams , while broad - line agn , which are typically fainter than ysos in the _ spitzer _ bands , are identified by their positions in the [ 4.5 ] vs. [ 4.5]-[8.0 ] diagram .
here we adopt this method and lower the [ 4.5 ] limit in the [ 4.5 ] vs. [ 4.5]-[8.0 ] diagram by 1 mag since our regions are far more distant . to further remove agn , we set @xmath25<16 $ ] for ysos
not detected in the 8.0 @xmath0 m band and require @xmath37<8.5 $ ] for those only identifiable in the [ 3.6]-[4.5 ] vs. [ 4.5]-[24 ] diagram .
we filter out a total of 11 extragalactic contaminants in all nine regions .
although we are undertaking deep observations toward relatively nearby high - mass star forming regions , the completeness of the ysos census is limited by a few factors .
bright extended nebulosity in the irac bands can be found in all the regions , which significantly limit the point source detection in these areas .
the ysos identification from the 2mass - irac diagram is limited by the modest sensitivity of the 2mass survey , while that from the irac - mips diagram suffers from the significant saturation in the mips 24 @xmath0 m image caused by the central luminous sources . to further evaluate the completeness of the census quantitatively , we plot histograms of the sources which are counted within 0.5 mag bins of the 3.6 @xmath0 m photometry in figure [ complete ] . in this plot ,
all the 3.6 @xmath0 m detections with signal to noise ratios sufficient to be identified by our point source filter are plotted with the gray solid lines , and sources with corresponding multi - band photometry so that they can be placed on one or more color - color diagrams for an assessment of infrared excess are plotted with the dark dashed lines .
the identified yso members are also plotted with the gray dotted lines .
there are two types of completeness in this diagram .
the first is the completeness in the 3.6 @xmath0 m band , which declines strongly where a peak is evident in the solid gray line histogram .
the second is the coupled multi - band completeness to infrared excess sources , which is approximately given where the dark dashed line diverges from the solid gray line .
since the 3.6 @xmath0 m band detects sources of fainter magnitudes than the other bands , we define the completeness magnitude as the point where the multi - band histogram declines more than 30% with respect to the 3.6 @xmath0 m histogram .
a vertical dash - dotted line in each plot indicates the completeness magnitude . for sources with the 3.6 @xmath0 m photometry brighter than the completeness magnitude
, we calculate the ratio of the number of sources that can be placed on one or more color - color diagrams for the search of infrared excess to the number of the total detections ( table [ fraction ] ) .
the actual completeness to infrared excess sources is presumably higher since they will have stronger emission in the irac bands than field stars and young stars without disks .
previous interferometric studies have identified dusty cores at the center of each region through the millimeter continuum emission , and so far five regions show multiple or a cluster of dusty cores . of these millimeter sources ,
the deep irac imaging detects six at 8.0 @xmath0 m , five at 5.8 @xmath0 m , three at 4.5 @xmath0 m , and two at 3.6 @xmath0 m .
the low detection rate is consistent with the deep irac observations of the other two nearby ( 1.7 kpc ) high - mass star forming regions , ngc 6334 i and ngc 6334 i(n ) , where only one out of eleven sma 1.3 mm sources shows a detectable infrared counterpart @xcite .
@xcite suggested that the irac band emission of candidate massive protostellar objects arises from the luminous envelopes around the protostars rather than their photospheres or disks . in our sample , the number of detected infrared counterparts increases with the increasing wavelength of the band , despite the fact that the 3.6 and 4.5 @xmath0 m bands are much more sensitive than the 5.8 and 8.0 @xmath0 m bands .
this implies a steeply rising sed in the irac bands , which is consistent with the interpretation that the emission is mostly coming from the dense envelopes .
all the six millimeter sources detected in one or more irac bands appear significantly saturated in the mips 24 @xmath0 m band , while there are two millimeter sources detected at 24 @xmath0 m without saturation but undetectable in any of the four irac bands .
we list the irac or mips 24 @xmath0 m photometry of detected millimeter sources in table [ irmm ] . in our sample , high - mass stars deeply embedded in dusty cores are early - b types estimated from far - infrared luminosities , ionizing uv radiation fluxes ( if uc hii regions detected ) , and core masses . also considering they are associated with molecular outflows indicative of active accretion , we refer to these deeply embedded b stars as proto - b stars . in all nine regions , massive molecular outflows
have been identified in millimeter lines of co and/or sio . in the irac observations ,
the central high - mass sources are found to be associated with extended nebulosity .
several mechanisms produce extended nebulosity in the irac bands .
prominent features from hydrocarbons appear in the 3.6 , 5.8 , and 8.0 @xmath0 m bands , with the strongest emission features in the 8.0 @xmath0 m band @xcite .
vibrational and rotational h@xmath2 lines appear in all four bands , with the weakest and most spatially confined emission in the 3.6 @xmath0 m band @xcite ; this h@xmath2 emission is most apparent in the 4.5 @xmath0 m band due to the high sensitivity of this band and presence of strong rotational lines @xcite .
finally scattered light may also produce nebulosity which is expected to be the strongest in the 3.6 @xmath0 m band . to distinguish between these different mechanisms ,
we employ 3.6/4.5 @xmath0 m two - color and 3.6/4.5/8.0 @xmath0 m three - color composite images to highlight emission from shocked h@xmath2 in outflows , scattered light from outflow cavities , and emission from uv heated hydrocarbons .
we identify a total of twelve outflows in eight out of the nine regions .
reflection nebulae and externally heated rimmed clouds are detected as well . in the following subsections ,
we discuss each region separately . for figures [ afgl5142 ] to [ w75n ] , the 3.6 ,
4.5 and 8.0 @xmath0 m emissions in color composites are coded in blue , green and red , respectively ; millimeter continuum cores from previous observations are marked as plus signs ; co and sio outflows from previous observations are shown in contours ; additional symbols ( arrows , dashed curves , etc . )
are merely to highlight the irac outflows from the ambient reflection nebulae or scattered light . in figure [ afgl5142]a , a short jet in the east and a long jet in the west
are detected .
the short jet coincides with the axis of one of the three jet - like molecular outflows detected in the co ( 2 - 1 ) and so ( @xmath38 ) with the sma ( outflow b in * ? ? ?
* see contours in figure [ afgl5142]b ) .
the long jet , on the other hand , coincides with the axis of an extended remnant outflow detected in co ( 2 - 1 ) with the cso telescope @xcite .
the emission from these two jets is most prominent in the 4.5 @xmath0 m band , in agreement with the previous ground - based 2.12 @xmath0 m h@xmath2 v=1 - 0 @xmath39 observations @xcite .
apparently the irac imaging of this region reveals internal driving agents of two molecular outflows . toward the center of afgl 5142
, @xcite detected five millimeter continuum sources ( mm1 to mm5 in figure [ 5142core ] ) . in figure
[ 5142core ] , a bright source residing @xmath40 northwest to mm1 ( marked with a star symbol ) is detected in the 2mass _ ks _ band , all four irac bands and mips 24 @xmath0 m band ( see the detection at @xmath41j@xmath42 , @xmath43j@xmath44 in table [ photometry ] ) .
it is classified as a protostar in the [ 3.6]-[4.5 ] vs. [ 5.8]-[8.0 ] and [ 3.6]-[4.5 ] vs. [ 4.5]-[24 ] diagrams .
the projected offset between this source and mm1 is marginally within the irac astrometry allowance .
thus it is still possible that this source is the infrared counterpart of mm1 .
however , @xcite detected a forest of molecular lines toward mm1 and suggested it to be the driving source of a jet - like molecular outflow , which implies that the protostar deeply embedded in mm1 is actively accreting .
such a protostellar core is usually obscured in @xmath45 band ( e.g. * ? ? ?
* ; * ? ? ?
* and observations of other regions in our sample ) .
considering the source is detected in the 2mass _ ks _ band , a survey with limited sensitivity , we suggest that it is most likely a different source rather than the infrared counterpart of mm1 .
but the 24 @xmath0 m emission can still arise from mm1 and/or other millimeter sources given its relatively low resolution ( @xmath46 ) , thus we list the 24 @xmath0 m photometry in table [ irmm ] .
the 24 @xmath0 m detection is located at the bright psf wing of a nearby saturated source , resulting in a potential overestimate of the flux by about @xmath47 to @xmath48 jy .
the driving sources of the multiple co / so outflows have not been unambiguously identified .
@xcite suggested mm3 may be the powering source of the co outflow coincident with the short jet in figure [ afgl5142]b .
the long jet is apparently powered by a relatively more evolved young b star ( denoted by a blue cross in figure [ fig2]a ) . in figure [ i05358 ] ,
three highly collimated jets emanating from a group of millimeter continuum sources are detected in the central part of iras 05358 + 3543 ( hereafter referred as iras 05358 ) .
jet1 coincides a well collimated outflow in co ( 1 - 0 ) reported by @xcite ( see contours in figure [ i05358 ] ) .
jet2 coincides with the axis of a high - velocity co outflow ( outflow b in @xcite ) .
several knots from these two jets are almost exactly coincident with the previously identified 2.12 @xmath0 m h@xmath2 knots @xcite .
@xcite did not report the identification of jet3 , but weak emission features coincident with this jet can be found in their 2.12 @xmath0 m h@xmath2 image .
@xcite suggested the existence of this jet , but the emission appears to be weaker since their h@xmath2 observations are less deep than that of @xcite . in the irac imaging ( figure [ i05358 ] ) ,
a chain of emission knots in this jet can be identified and even its tip can be identified despite the contamination by a nebulous knot .
there is no reported detection of an outflow toward this jet in millimeter co or sio observations .
however , in the northern lobe of the sio ( 2 - 1 ) outflow proposed by @xcite ( their fig .
8b ) , the highest contours show elongation coincident with jet3 . considering sio is a good shock tracer at ( sub)millimeter wavelengths , the elongated sio emission could be at least partially powered by jet3 . at the center of this region
, @xcite detected three millimter continuum sources at @xmath49 resolutions ( marked as mm1 to mm3 in figure [ 05358core ] ) .
the brightest emission feature in figure [ 05358core ] coincides with mm1 ; two northern `` red '' features ( denoted by two white dashed circles ) are not real but `` bandwidth '' artifacts .
an investigation of the emission in individual bands shows that emission features coincident with mm2 and mm3 are discernible in the 3.6 and 4.5 @xmath0 m bands but is not identified by our point source filter .
the emission feature coincident with mm1 is marginally discernible in the 3.6@xmath0 m band , and identified as a point source longward of 4.5 @xmath0 m .
the mips 24 @xmath0 m emission from this area is significantly saturated . from its extremely red irac colors ( [ 4.5]-[5.8 ] = 1.33 and [ 5.8]-[8.0 ] = 1.15 ) and position coincidence with mm1 , we suggest this source to be the infrared counterpart of mm1 and list its photometry in table [ irmm ] . at higher resolutions ( @xmath50 ) ,
the mm1 core splits into two sub - cores at 875 @xmath0 m and 1.2 mm @xcite .
one of the two sub - cores coincides with a point source at 7.9 @xmath0 m detected with the gemini north telescope @xcite .
the sub - structure of mm1 can not be resolved in the irac observations .
the flux of 0.586 jy at 8.0 @xmath0 m ( @xmath51=5.098 $ ] ) is consistent with the flux of 0.68 jy at 7.9 @xmath0 m reported in @xcite , suggesting that the flux of the irac source is dominated by the sub - core coincident with the 7.9 @xmath0 m detection .
mm1 ( or more precisely , one of the two sub - cores in mm1 ) is suggested to be the driving source of jet1 @xcite .
a reliable one - to - one association between jet2 , jet3 and any two millimeter sources in this region can not be established with the current data .
@xcite detected a west - east bipolar outflow in co ( 1 - 0 ) in g192.16 - 3.82 ( hereafter referred as g192 ) . the eastern lobe of this outflow is detected in the 3.6 and 4.5 @xmath0 m bands . in figure
[ g192 ] , the most prominent features of this lobe are `` green '' nebulosities within the area outlined by two dashed lines . the emission to the west of these features is confused by the bright reflection nebula , but a `` blue '' v - shaped structure , with the millimeter source at the tip , is still discernable in the three - color composite of the field ( figure [ fig2]c ) .
the overall structure of the eastern lobe in the irac imaging , i.e. , the `` green '' features as well as the v - shaped inner part , roughly coincides with but far exceeds the eastern lobe of the co outflow . the prominence in the 4.5 @xmath0 m band and coincidence with previous h@xmath52 and [ sii ]
detections @xcite strongly suggest that the `` green '' nebulosities are shock excited h@xmath2 knots .
the base of the western lobe of the co outflow is also detected in the irac imaging , with the prominent feature being a candle - flame - shaped structure ( delineated by a dashed curve in figure [ g192 ] , but more prominent in figure [ fig2]c ) ; this structure is previously detected in the @xmath45-band imaging @xcite .
one interesting aspect of this outflow is that although the eastern lobe is filled with shocked h@xmath2 knots , a collimated component can not be found either in our irac imaging or previous observations .
the co outflow appears to be driven by a wide - angle wind revealed in the irac , h@xmath52 and [ sii ] observations .
a proto - b star surrounded by an uc hii region and embedded in a dense dusty core was suggested to be the driving source of the outflow @xcite . a bright source coincident with the millimeter continuum peak
is detected in the irac bands and show very red irac colors ( figure [ 192core ] ) .
we suggest it to be the infrared counterpart of the millimeter source and list its photometry in table [ irmm ] .
a detailed comparison between the infrared and millimeter emission in the region will be presented in a future work ( shepherd et al .
, in prep . ) . an extraordinarily well - collimated and powerful radio jet was previously detected in hh 80 - 81 ( * ?
* ; * ? ? ?
* ; * ? ? ?
* see contours in figure [ h8081 ] ) .
having the radio jet right at the axis , the _ spitzer _
data for the first time reveals a bipolar cone - shaped cavity with its emission most prominent in the 8.0 @xmath0 m band with an extent of about 4pc ( figure [ h8081 ] ) .
low - mass protostellar outflows are often found to consist of two components , with an axial narrow jet - like component surrounded by a wide - angle biconical component @xcite .
it is unclear whether outflows emanating from high - mass young stars or protostars may exhibit two components similar to low - mass outflows .
our irac imaging as well as previous radio continuum observations of the hh 80 - 81 outflow provides the best case of a two - component outflow emanating from a @xmath53 source .
the remarkable prominence in the 8.0 @xmath0 m band suggests that the emission of the irac biconical structure is from hydrocarbons heated by uv photons from the central source .
single - dish observations revealed a large bipolar outflow in co ( 1 - 0 ) @xcite , with the northern lobe being blueshifted and the southern lobe redshifted .
the orientation of the co outflow is roughly consistent with the irac outflow structure . in the three - color composite image ( figure [ fig2]d ) , a curved structure in the northern lobe stands out in the 3.6 @xmath0 m emission ( blue ) .
this curved structure is very bright in previous near - infrared @xmath45-band observations and has been suggested as the wall of a parabolic cavity @xcite .
the dominant 3.6 @xmath0 m emission from this structure may be partly explained by the scattered light in this structure being enhanced by forward scattering and lower extinction . the 8.0 @xmath0 m emission within this structure is spatially limited to the inner edge of the parabolic wall .
this may suggest that part of the northern lobe cavity is shielded from the uv photons by gas and dust in the outflow and perhaps by the circumstellar environment around the central source .
two additional reflection nebulae are found on both sides of the bipolar cavity ; these appear to be heated by two young b stars ( see table [ photometry ] ) and not directly connected to the outflow cavity .
the central driving source of the outflow has been detected in the centimeter @xcite and millimeter @xcite continuum .
mid - infrared emission from this source has been detected at 4.7 to 13 @xmath0 m in the ground - based observations @xcite . in figure [ 8081core ] ,
the centroid of the emission feature in the 3.6 @xmath0 m band is about @xmath54 northeast offset from the millimeter peak .
this feature coincides with a @xmath45-band knot detected in previous ground - based observations ( their irs 2 , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . from its high polarization and relatively large fwhm
, @xcite argued that it is a nebulous knot of gas / dust rather than a point source .
the emission feature in the 4.5 @xmath0 m band shows elongation toward the position of the millimeter peak .
the centroid of the emission feature in the 5.8 and 8.0 @xmath0 m bands , discarding the `` bandwidth '' artifact in the north , coincides with the millimeter peak .
we suggest that the emission in the 3.6 and 4.5 @xmath0 m bands is mostly attributed to the irs 2 knot , while that in the long wavelength bands is mostly coming from the millimeter core and their photometry is listed in table [ irmm ] .
the mips 24 @xmath0 m emission in the central part of this region is significantly saturated .
as outlined with a dashed ellipse in figure [ i19410 ] , an elliptical nebula orientated in the northeast - southwest direction can be identified from the extended emission in the central part of iras 19410 + 2336 ( hereafter iras 19410 ) . in interferometric co observations , multiple outflows emanating from the center of this region
show very complex morphologies @xcite . the axis of one outflow proposed by beuther et al . , which is shown in contours in figure [ i19410 ] , coincides with the major axis of the irac elliptical nebula .
jet - like 2.12 @xmath0 m h@xmath2 emission along the axis of this outflow was detected as well @xcite .
the elliptical nebula , with two local cavities at the ends of the major axis , is only detectable in the 3.6 and 4.5 @xmath0 m bands and appears more prominent in the 3.6 @xmath0 m band , suggesting the emission is dominated by scattered light from the central source .
apparently the irac imaging of the iras 19410 outflow reveals a less collimated component than the 2.12 @xmath0 m h@xmath2 jet .
the driving source of this outflow is suggested to be deeply embedded in a dusty core traced by millimeter continuum emission at a resolution of @xmath55 @xcite .
the infrared emission from this millimeter source is detected in all four irac bands and the photometry is listed in table [ irmm ] .
there is a 2mass @xmath56 band detection , with the photometry of @xmath57 , at @xmath58 away from the millimeter continuum peak . as we discussed in [ msf_afgl5142 ] , it is more likely another star rather than the infrared counterpart of the millimeter source .
the mips 24 @xmath0 m emission from the source is significantly saturated . at a higher resolution ( @xmath59 ) ,
the millimeter core splits into multiple sources @xcite ; our irac observations can not resolve those sources .
a detailed comparison between the irac observations and high - spatial - resolution millimeter continuum observations of this region will be presented in a future work ( rodon et al .
, in prep . )
. in figures [ i20126]a & [ i20126]c , a bipolar outflow in iras 20126 + 4104 ( hereafter referred as iras 20126 ) is detected as a limb - brightened biconical cavity .
the cavity structure , in particular the wall of the northwestern lobe , is clearly detected in the 3.6 , 4.5 , and 5.8 @xmath0 m bands and its prominence in the 3.6 and 4.5 @xmath0 m bands suggests that the emission is dominated by the scattered light . the molecular outflow in iras 20126 has been interpreted as driven by a precessing jet based on the s - shaped locus of the 2.12 @xmath0 m h@xmath2 knots and the orientation variation between the inner jet - like sio outflow and the larger co outflow @xcite .
the jet - precessing scenario is also consistent with the presence of a double system detected at centimeter and near - infrared wavelengths @xcite .
our irac observations put new insight into the outflow property of this luminous source .
the irac biconical structure far exceeds inner sio outflow and covers part of the larger co outflow . from the limb - brightening ,
the cavity wall has pas from @xmath60 to @xmath61 .
the pa @xmath60 wall is significantly offset from the axis of the larger co outflow which is nearly in a n - s orientation @xcite , while the pa@xmath61 wall has an approximately same orientation as the jet - like sio outflow .
a natural interpretation of the cavity ( with the limb - brightening ) is that it traces a less collimated biconical component of the iras 20126 outflow , rather than being swept up by a jet precessing from n - s to the current - position of pa@xmath62 .
however , in contrast to the hh 80 - 81 outflow ( [ msf_hh8081 ] ) , the jet component in this region coincides with the wall , not the axis , of the biconical cavity .
while multiple outflows can be one possibility , another possibility can be derived by adopting the precessing scenario and the theoretical model unifying the jet - driven and wind - driven low - mass outflows @xcite : when the underlying axial jet and the co - existing wind are precessing , the dense jet is more efficient in entraining a collimated molecular outflow via strong shock activities , while the cavity evacuating by the more tenuous wind component is less efficient .
consequently , the jet - like sio outflow can be detected very close to the current - position of the precessing jet while the cavity revealed by the scattered light more or less lags behind .
a dense dusty core , which harbors a proto - b star driving the outflow , was previously detected in the millimeter continuum @xcite . in the 4.5 and 5.8 @xmath0 m bands
the emission feature coincident with the millimeter source is discernible . in the 8.0 @xmath0 m band ( figure [ i20126]b ) ,
a point source at this position is detected and the photometry is listed in table [ irmm ] . at higher resolutions ( @xmath63 ) ,
two emission regions oriented along the outflow axis and separated by a dark lane were detected at near- and mid - infrared wavelengths @xcite .
the source in the 8.0 @xmath0 m band may encompass the previously detected double sources . in figure [ i20293 ] ,
ahead of a previously detected high - velocity co outflow @xcite , a short jet is detected in the irac imaging with an extent of about @xmath64 . with the most prominent emission in the 4.5 @xmath0 m band . in orientation the irac jet
well coincides with the co outflow .
this outflow was also detected in thermal emissions of sio and ch@xmath65oh @xcite , which often trace shock activities in outflows . in the previous 2.12 @xmath0 m h@xmath2 observations ,
the overall structure was not detected but a faint knot can be found at the southeastern end of the irac jet ( knot `` c '' in fig . 2 of @xcite ) .
we suggest that the irac imaging for the first time reveals the interval driving agent of the co / sio / ch@xmath65oh outflow . a millimeter continuum source , which may harbor an intermediate - mass protostar , is suggested to be the driving source of the co / sio / ch@xmath65oh outflow @xcite . the infrared emission from the millimeter source is only detected in the mips 24 @xmath0 m band ( table [ irmm ] ) . as the case of afgl 5142 ( [ msf_afgl5142 ] ) , the 24 @xmath0 m detection is located at the bright psf wing of a nearby saturated source , resulting in a potential overestimate of the source flux by about @xmath66 to @xmath67 jy
. in figure [ w75 ] ,
multiple bow - shaped structures in w75n can be identified .
the most remarkable one , marked by the double arrows , form a large scale bipolar bow - shaped structure .
the shell of the southwestern lobe and the tip of the northeastern lobe of this structure is clearly revealed in the 4.5 @xmath0 m band .
this bow - shaped structure has previously been detected in the 2.12 @xmath0 m h@xmath2 imaging by several authors @xcite , but not reported in previous irac observations toward the dr 21/w75n region carried out during the science verification period of _ spizer _
@xcite , probably because their observations are far less deep in integration .
the southwestern lobe coincides with the boundary of the redshifted lobe of a large outflow detected in co ( 1 - 0 ) @xcite , suggesting the bow - shock - driven nature of this molecular outflow .
in addition to this remarkable structure , at least five bow - shaped structures pointing to the southeast of the region can be found in figure [ w75 ] .
these structures may be shocked h@xmath2 `` fingers '' tracing multiple bow - shocks driven from the center of the region , roughly in a fashion similar to the famous h@xmath2 `` fingers '' in the orion kl region @xcite . zooming in on the inner part of this region
, the irac imaging reveals a new bow - shaped structure ( outlined by a solid curve in figure [ w75n ] ) .
this structure is clearly detected in the 3.6 and 4.5 @xmath0 m bands and appears more prominent in the 4.5 @xmath0 m band ( figures [ w75n]a , [ w75n]b ) .
its orientation approximately coincides with the axis of a co outflow proposed by @xcite . at the center of the large northeast - southwest
bow - shaped structure , @xcite detected a group of four millimeter continuum sources ( labelled as mm1 to mm4 in figure [ w75n]a ) .
embedded in mm1 is a cluster of four centimeter continuum sources , with three within an area of @xmath68 and another about @xmath69 to the south ( figure [ w75n]c , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
of the four centimeter sources , the northernmost one was interpreted as an ionized radio jet while the other three were suggested to be uc hii regions excited by proto - b stars .
it is still unclear which is driving the large scale co outflow .
@xcite suggested the first uc hii region from the north ( namely vla 2 ) is likely to be the driving source .
the resolution of our irac observation does not allow us to shed any further light on this issue .
@xcite suggested the second uc hii region from the north ( namely vla 3 ) to be the driving source of the co outflow whose axis approximately coincides with the newly discovered small bow - shaped structure in the irac image .
however , the co outflow is very compact and confined to within @xmath46 from mm1 while the bow - shaped structure is about @xmath70 ( 0.5 pc in projection ) away from the group of the millimeter sources . in projection mm3 is closer to the axis of this bow - shaped structure and thus more likely to be the driving source of this structure .
the infrared emission from mm1 is marginally discernible in the 3.6 @xmath0 m band ( figure [ w75n]a ) . from 4.5 ( figure [ w75n]b ) to 8.0 @xmath0 m it is clearly identified as a point source , and the photometry is listed in table [ irmm ] .
the mips 24 @xmath0 m emission from the source is significantly saturated .
iras 22172 + 5912 ( hereafter iras 22172 ) is the only region in our sample toward which no outflow signature is found in the irac bands , though a bipolar outflow in co and hco@xmath71 has been detected with the ovro array @xcite .
the peak of the dust emission , which was detected in the ovro 2.6 mm and 3.4 mm continuum observations at @xmath72 resolutions , is about @xmath73 offset from the geometric center of the bipolar molecular outflow @xcite .
the irac and mips 24@xmath0 m imaging do not detect infrared emission from the millimeter source but reveals two protostars at the center of this field ( see figure [ fig2]i ) .
the southern protostar coincides with the geometric center of the co outflow and thus can be a candidate of the driving source of the outflow . for the millimeter source ,
the lack of infrared emission shortward of 24 @xmath0 m and the apparent non - association with the molecular outflow suggest the source to be at a very young evolutionary stage .
further ( sub)millimeter continuum and line observations with high sensitivity and higher resolutions are needed to verify whether it is at a prestellar stage or harbors deeply embedded protostars .
the irac imaging of the sample reveals mainly three classes of extended emission : h@xmath2 emission dominated nebulae most prominent in the 4.5 @xmath0 m band ( green ) ; scattered light dominated nebulae most prominent in the 3.6 @xmath0 m band ( blue ) ; uv heated hydrocarbon emission dominated reflection nebulae most prominent in the 8.0 @xmath0 m band ( red ) .
the first two classes have been discussed in detail in [ msf ] .
the uv heated reflection nebulae can be found in most regions .
while the first two classes of extended emission are found in outflows or outflow cavities and thus imply a very young evolutionary stage of the central source , most of the uv heated reflection nebulae are associated with young but relatively more evolved b stars ( one exception being the biconical cavity of the hh 80 - 81 outflow ) .
for the cometary nebula in afgl 5142 , the southwestern nebula in iras 05358 , the southeastern and northerwestern nebulae in hh 80 - 81 , the central nebula in iras 20293 , and the northern and central nebulae in w75n , a bright source at the center is detected ( denoted as a blue cross in figure [ fig2 ] ) . these sources show infrared excess in our censes of ysos based on color - color diagrams , and most of them were previously suggested to be intermediate- to high - mass young stars ( see annotation of tabel [ photometry ] ) .
they are most likely the exciting sources of the surrounding reflection nebulae . of these sources , the one in iras
20293 shows an uc hii region and the central one in w75n is surrounded by a relatively extended hii region .
the spectral types of the exciting stars of these two hii regions , derived from the lyman continuum photos required to produce the observed ionized emission , are b1 and b0.5 for that in iras 20293 and w75n , respectively @xcite . toward the two sources in hh 80 - 81 , compact centimeter continuum was detected as well @xcite , and the estimated rates of lyman continuum photos suggest b3 to b2 stars being the exciting sources of these two reflection nebulae .
there is no reported detection of a distinct hii region surrounding the other candidate exciting sources .
we suggest they are young b stars capable of providing strong fuv radiation to excite hydrocarbon emission but too weak euv to ionize its environment .
spectroscopic studies are needed to further investigate whether these stars have circumstellar disks and to confirm their spectral types .
there are also uv heated reflection nebulae for which an exciting source can not be identified with the existing data , but the need of fuv photons to heat hydrocarbons and the lack of significant free - free emission also suggest a b type star being the exciting source . to lower the optical depth in fuv and consequently allow the stars to heat the surrounding hydrocarbons ,
these b stars must have cleared most of the surrounding gas . on the other hand , as discussed in
[ msf ] , the central massive star formation site in each region shows millimeter continuum cores harboring deeply imbedded proto - b stars .
thus it appears that most regions in the sample are forming b stars at a range of evolutionary stages . in two regions , iras 20126 and iras 22172
, the central massive star formation sites are embedded in a bright large scale structure which is prominent in the 8.0 @xmath0 m band .
these two structures are externally heated rimmed clouds .
one of the distinctive features of massive star formation is that it occurs in clusters .
consequently , understanding the role of clustering in the formation of high - mass stars is an essential step toward a theory of massive star formation .
a controversial issue is whether clusters are necessary for the formation of high - mass stars , or whether clusters are merely the byproduct of the formation of high - mass stars in dense massive molecular cores . in a near - infrared survey of intermediate - mass herbig ae / be stars
, @xcite found a smooth transition between the low density aggregates of young stars associated with stars of spectral type a or later , and the dense clusters associated with o and early - b type stars .
they argued that the presence of dense clusters is required for the formation of high - mass stars .
the smooth transition suggested that there may be a fundamental relationship between the mass of the most massive star and the number and/or density of stars in a cluster .
however , @xcite found evidence that 4% of o stars form in isolation .
this is evidence that clusters are not necessary for the formation of o stars , although @xcite argued that the isolated o stars may actually form in small clusters . to test whether massive stars can form without low - mass stars
requires identification of young massive objects without associated clusters of low - mass stars ; only in these cases , can we rule out the possibility that an attendant cluster dispersed due to dynamical evolution . in a @xmath45-band survey of eight
high - mass star forming cores , @xcite found one core , ngc 6334 i(north ) , showing evidence of high - mass star formation without evidence for a cluster of embedded low - mass stars ( also see * ? ? ?
* ; * ? ? ?
* ) , although a cluster may still exist deeply embedded in the cloud @xcite .
it is important to search for other potential examples of massive star formation without the presence of a dense cluster . with its ability to identify ysos through infrared excesses and detect
deeply embedded protostars , _ spitzer _ is well suited to this task .
although such regions may yet form clusters , the identification of high - mass young or protostars without a dense cluster would demonstrate that low - mass stars are not required for the formation of massive stars .
our survey of young intermediate- to high - mass ( proto)stars ( @xmath74 to @xmath75 ) can directly address the role of clustering in the early stages of intermediate- and high - mass forming stars .
the _ spitzer _
data provides the ability to directly identify likely ysos by the detection of mid - infrared excesses .
the advantage of this approach over previous near - infrared source counting methods is that the sample does not significantly suffer from contamination by background stars which can be significant for high - mass star forming regions in the galactic plane @xcite . also , the mid - infrared colors are much more sensitive to protostellar objects , and candidate protostellar objects can be identified through their colors . however , the bright mid - infrared nebulosity in high - mass star forming regions , the @xmath49 angular resolution of the irac data , and the modest sensitivity of the 2mass photometry limit the detection of sources , in particular in regions of bright nebulosity .
we expect the census of ysos to be incomplete in all of the regions .
furthermore , we expect that a certain fraction of the stars do not have disks and will not be identified in our analysis ; the fraction of stars without disks is approximately 30% for 1 myr
old low - mass stars @xcite .
we group the objects by their total far - infrared luminosity , as measured from the iras point source catalog ; this luminosity will be dominated by the most massive objects in the regions . in all the regions , associated low - mass ysos are detected .
the @xmath76 regions , afgl 5142 , g192 , iras 05358 and iras 20293 , show parsec sized clusters with around 20 ysos surrounded by a more extended and sparse distribution of young stars and protostars . the @xmath77 regions iras 19410 and
hh 80 - 81 also show clusters and extended components . except afgl 5142
, the clusters associated with these regions do not show concentrations of stars toward the central massive objects .
this may be due to the decrease in completeness toward the bright nebulosities in the central regions .
finally , the @xmath78 region w75n shows a cluster that is both richer and more spatially extended .
this region also shows a paucity of sources in the bright nebulous center of the region , further indicating the observations are significantly limited by completeness . in figure [ fig17 ]
we plot the number of sources as a function of the source luminosity . to mitigate the effects of incompleteness on our samples , we plot the number of sources with 3.6 @xmath0 m magnitudes lower than 13 and 14 . given a typical @xmath79 $ ] color of 0.25 , an extinction of @xmath80 at 1.8 kpc , and an age of 1 myr , these limits correspond to masses of 1.2 and 0.5 @xmath81 @xcite .
the two plots show a trend of increasing number of associated ysos with increasing total luminosity . in particular
, the number of ysos in the @xmath78 region ( w75n ) is significant higher than those in @xmath77 regions , which again are somewhat higher in number than the @xmath76 regions .
it is possible that a trend of increasing incompleteness with higher far - infrared luminosities is decreasing the slope of this trend .
there are two main exceptions to this trend : iras 22172 and iras 20126 .
iras 22172 appears to be part of a larger , potentially more evolved region which has undergone significant gas clearing .
the clearing of the gas would also lower the fraction of light absorbed by dust and re - emitted in the infrared ; hence this region shows a large number of stars for its measured iras far - infrared luminosity .
the other exception is iras 20126 , which has the lowest number of associated ysos in the sample . unlike the other regions , iras 20126 shows no obvious cluster in the field
this is unlikely to be solely the effects of incompleteness , with the average signal of about 2.5 mjy / sr at 4.5 @xmath0 m and 52 mjy / sr at 8.0 @xmath0 m in the central parsec area , compared to 6.5 mjy / sr at 4.5 @xmath0 m and 67 mjy / sr at 8.0 @xmath0 m for hh 80 - 81 and 3.8 mjy / sr at 4.5 @xmath0 m and 72 mjy / sr at 8.0 @xmath0 m for iras 19410 .
more rigorous studies of this sample await deeper near - infrared data to complement the _ spitzer _ 3.6 and 4.5 @xmath0 m imaging .
however , the analysis here already illustrates several features .
first , there does seem to be a trend of the number of associated ysos with luminosity .
second , these regions show both clusters in the central 1 pc as well as more extended halos of sources around the clusters . finally , there does seem to be a significant dispersion in the number of associated ysos for a given total far - infrared luminosity , with one luminous source , iras 20126 , showing no obvious cluster .
the lack of a cluster warrants future work ; if confirmed this would demonstrate that the formation of @xmath77 sources is not dependent on the presence of deep clusters . for a sample of nine high - mass star forming regions associated with co and/or sio outflows
, the deep irac imaging reveals outflows and outflow cavities toward eight regions , illustrating that the irac imaging can be an effective tool for outflow detection . for the jets in iras 05358 ,
afgl 5142 , iras 20293 , the bow - shock shells in w75n , and the nebulosities in the eastern lobe of the g192 outflow , the prominence in the 4.5 @xmath0 m band and coincidence with the ground - based 2.12 @xmath0 m h@xmath2 detections ( if detected in the 2.12 @xmath0 m h@xmath2 line ) suggest the emission is mostly attributed to shocked h@xmath2 emission , although the contribution from hi br@xmath52 ( 4.052 @xmath0 m ) , co v=1 - 0 ( 4.45 - 4.95 @xmath0 m ) lines can not be ruled out without spectroscopic observations .
a detailed investigation of the emission in the biconical cavity in iras 20126 shows strong evidence for the scattered light being dominant in the 3.6 and 4.5 @xmath0 m bands ( qiu et al .
, in prep . ) .
the elliptical structure in iras 19410 appears relatively diffuse and most prominent in the 3.6 @xmath0 m band .
the emission in this outflow can be dominated by the scattered light as well .
the clear prominence of the biconical structure in hh 80 - 81 suggested is dominated by uv heated hydrocarbon emission .
both theoretical models and high spatial resolution observations suggest an evolutionary scenario for low - mass outflows : as low - mass stars evolve from class 0 through class i to class ii stages , the mass loading in the winds and density structure in the cores conspire the widening of outflows @xcite . while for massive outflows , whether an evolutionary picture exists is unknown @xcite .
the detected outflows in the sample show dramatically different morphologies .
we detect highly collimated jets , bow - shock shells , and biconical cavities .
given the extreme complexity of the co observations @xcite and relatively confused emission in the irac imaging compared with the other regions , we leave out the iras 19410 outflow in the following discussion .
considering jet1 in iras 05358 , the short jet in afgl 5142 , and the jet in iras 20293 , the irac observations reveal the internal driving agents of well collimated co outflows in these regions .
the powering sources of jet1 in iras 05358 and the short jet in afgl 5142 are suggested to be deeply embedded proto - b stars , toward which only very weak radio continuum was detected @xcite .
the jet in iras 20293 is most likely driven by an intermediate - mass protostar showing no detectable centimeter emission @xcite . in contrast , the large scale co outflows in w75n and g192 appear to be poorly collimated within @xmath82 pc from the central driving sources . in w75n , proper - motion observations of water masers by @xcite delineated a non - collimated , shell outflow at a 160 au scale expanding in multiple directions with respective to the uc hii region vla 2 , which may drive the large scale co outflow as suggested by @xcite . the bipolar bow - shaped structure beyond @xmath83 pc detected in the 2.12 @xmath0 m h@xmath2 and the irac 4.5 @xmath0 m imaging could be the remanent of a collimated component . for the g192 outflow
, no collimated component can be found within @xmath84 pc from the driving source .
the proposed central driving sources of these two poorly collimated outflows are found to be surrounded by uc hii regions .
if we adopt the centimeter free - free emission as an indicator of the relative evolution between these sources , there seems to be a trend of evolution for the related outflows : both internal driving agents and entrained molecular outflows appear to be highly collimated for the youngest sources ( e.g. , iras 05358 , afgl 5142 , iras 20293 ) ; as the central source evolves to form a significant uc hii region , only poorly collimated outflow structures can be found around the central driving source ( e.g. , w75n , g192 ) . for the iras 20126 outflow ,
it remains ambiguous whether it is merely composed of a precessing jet and consequently larger structures are all entrained / swept - up by this jet or it has a jet - like component as well as a biconical component with a moderate opening angle . toward the central driving source of this outflow ,
very weak centimeter continuum was detected , presumably suggesting the source being at a very young evolutionary stage .
for the hh 80 - 81 outflow , our _ spitzer _ data and previous centimeter continuum reveal a remarkable two - component outflow with a biconical cavity surrounding an axial radio jet .
strong centimeter continuum was detected toward the central driving source of the hh 80 - 81 outflow , presumably suggesting the source being relatively more evolved compared with that of iras 20126 .
the outflow cavity of hh 80 - 81 also shows emission in the mid - infrared hydrocarbon features , suggesting that uv radiation from the central sources are heating the cavity walls .
therefore , if the irac biconical cavity in iras 20126 is a complementary less collimated component of the previously detected sio jet , there will be a large evolutionary time scale for the existence of both a highly collimated jet - like component and a less collimated biconical component for outflows from proto - b stars . with the existing data it is very difficult to determine whether the central source of iras 20126 is more evolved than those of afgl 5142 and iras 05358 , and whether the central source of hh 80 - 81 is younger than those of w75n and g192 .
presumably it is likely that the iras 20126 and hh 80 - 81 outflows , with the former being younger , represent evolutionary stages between that of the afgl 5142 & iras 05358 outflows , for which both internal driving agents and entrained gas appear well collimated , and that of the w75n & g192 outflows , for which only poorly collimated structures can be found around the central driving sources .
however , such a trend is only a tentative interpretation based on current observations of a small sample .
the question whether there is an evolutionary picture for massive outflows is far from conclusive , and remains an observational challenge . in high - mass star forming regions , multiple outflows with complicated structures
are often detected .
due to relatively larger distances and crowded clustering mode , it is very difficult with the current facilities to reliably identify the driving source of a massive outflow . also resolving the launching zone of a massive outflow , which is of great importance for investigate driving mechanisms , requires extremely high spatial resolution and high sensitivity .
in addition , observations of outflows in more luminous objects ( @xmath85 ) , in particular high spatial resolution observations , are still rare ( the combined bolometric luminosity of uc hii regions in mm-1 in w75n is @xmath86 ) .
extensive and comprehensive studies of outflows in this luminosity regime would be crucial for testing possible evolutionary scenarios of massive outflows and understanding formation processes of o type stars .
we have described initial results of mid - infrared imaging observations toward nine high - mass star forming regions made with the irac and mips cameras onboard _
spitzer_. the regions were selected for the presence of luminous ( @xmath87 l@xmath88 ) young objects driving molecular outflows .
the observed fields are approximately @xmath1 , corresponding to physical widths of @xmath89 pc for the source distances ranging from @xmath90 kpc . using the @xmath91 m _ spitzer _ photometry in combination with near - infrared 2mass data ,
we identify a total of 417 ysos with infrared excesses attributed to circumstellar disks or envelopes , including at least 12 candidates of intermediate- to high - mass young stars . in most regions ,
the spatial distribution of these ysos shows both a cluster component centered on the sites of massive star formation as well as a more extended distribution of ysos outside the cluster .
the number of ysos grows with the total far - infrared luminosity , with two significant exceptions : iras 22172 appears to be part of a larger , more evolved region ; iras 20126 appears to be a @xmath92l@xmath93 protostar with only 19 associated ysos and no central cluster .
zooming in on the central massive star formation sites , we search for infrared counterparts toward millimeter continuum sources identified in interferometer observations .
we detected counterparts for eight millimeter continuum sources in the irac bands or mips 24 @xmath0 m band .
these eight sources appear to be natal cores of proto - b stars driving molecular outflows .
all of the regions of massive star formation have bright associated nebulosity from scattered light , shocked h@xmath2 and/or uv heated hydrocarbon emission features .
this nebulosity limits the completeness of our yso survey in the central areas .
the deep irac imaging detects features associated with 12 outflows in eight of the surveyed regions .
compared with previous ground - based observations , our irac observations confirm previous 2.12 @xmath0 m h@xmath2 jets or bow - shock shaped structures in afgl 5142 , iras 05358 , and w75n ( the large bow - shock shaped structure ) , and previous h@xmath94/[sii ] features in g192 ; reveal new structures in hh 80 - 81 , iras 20126 , and iras 19410 ; detect a new h@xmath2 jet in iras 20293 and a new bow - shock shaped structure in w75n ( the small bow - shock shaped structure ) . in one case ,
hh 80 - 81 , the inner surface of the outflow cavity shows strong 8 @xmath0 m emission indicative of hydrocarbons heated by uv radiation .
based on the morphological variations of the detected outflows in conjunction with previous observations , we outline a possible evolutionary picture for massive outflows .
uv heated reflection nebulae dominated by hydrocarbon emission in the 8.0 @xmath0 m band can be found in most regions .
they may imply the presence of relatively more evolved young b stars .
externally heated bright rimmed clouds are found in iras 20126 and iras 22172 .
we are grateful to luis f. rodrguez for providing us the vla cm data on hh 80 - 81 .
this work is based on observations made with the _ spitzer _ space telescope , which is operated by the jet propulsion laboratory , california institute of technology under a contract with the national aeronautics and space administration ( nasa ) .
this publication makes use of data products form the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center / california institute of technology , funded by nasa and the national science foundation .
k. q. acknowledges the support of the grant 10128306 from nsfc .
h. b. acknowledges financial support by the emmy - noether - program of the deutsche forschungsgemeinschaft ( dfg , grant be2578 ) .
allen , l. e. , calvet , n. , dalessio , p. , merin , b. , hartmann , l. , megeath , s. t. , gutermuth , r. a. , muzerolle , j. , pipher , j. l. , myers , p. c. , & fazio , g. g. 2004 , , 154 , 363 arce , h. g. & sargent , a. i. 2006 , apj , 646 , 1070 aspin , c. & geballe , t. r. 1992 , , 266 , 219 aspin , c. , mccaughrean , m. j. , casali , m. m. , & geballe , t. r. 1991 , , 252 , 299 aspin , c. , puxley , p. j. , blano , p. r. , pia , p. k. , pickup , d. a. , paterson , m. j. , sylvester , j. , laird , d. c. , bridger , a. , daly , p. n. , & griffin , j. l. 1994 , , 292 , l9 benedettini , m. , molinari , s. , testi , l. , & noriega - crespo , a. 2004 , , 347 , 295 beuther , h. , leurini , s. , schike , p. , wyrowski , f. , menten , k. m. , & zhang , q. 2007 , , 466 , 1065 beuther , h. & schilke , p. 2004a
, science , 303 , 1167 beuther , h. , schike , p. , & gueth , f. 2004b , , 608 , 330 beuther , h. , schike , p. , & wyrowski , f. 2004c , , 615 , 832 beuther , h. , schike , p. , gueth , f. , mccaughrean , m. , andersen , m. , sridharan , t. k. , & menten , k. m. 2002 , , 387 , 931 beuther , h. , schike , p. , & stanke , t. 2003 , , 408 , 601 beuther , h & shepherd , d. s. 2005 , in cores to clusters : star formation with next generation telescopes , eds . m. s. n. kumar et al . , 105 - 119 baraffe , i. , chabrier , g. , allard , f. , & hauschildt , p. h. 1998 , , 337 , 403 calvet , n. , hartmann , l. , kenyon , s. j. , & whitney , b. a. 1994 , , 434 , 330 cesaroni , r. , felli , m. , jenness , t. , neri , r. , olmi ,
l. , robberto , m. , testi , l , & walmsley , c. m. 1999 , , 345 , 949 cesaroni , r. , felli , m. , testi , l. , walmsley , c. m. , & olmi , l. 1997 , , 325 , 725 cesaroni , r. , neri , r. , olmi , l. , testi , l. , walmsley , c. m. , & hofner , p. 2005
, , 434 , 1039 chen y. , yao , y. , yang , j. , & zeng , q. 2005 , , 629 , 288 davis , c. j. , kumar , m. s. n. , sandell , g. , froebrich , d. , smith , m. d. , & currie , m. j. 2007 , , 374 , 29 davis , c. j. , smith , m. d. , & moriarty - schieven , g. h. 1998 , , 299 , 825 de buizer , j. m. 2007 , , 654 , l147 devine , d. , bally , j. , reipurth , b. , shepherd , d. , & watson , a. 1999 , , 117 , 2919 de wit , w. j. , testi , l. , palla , f. , & zinnecker , h. 2005 , , 437 , 247 fazio , g. g. et al .
2004 , , 154 , 10 flagey , n. , boulanger , f. , verstraete , l. , miville deschnes , m. a. , grespo , a. n. , & reach , w. t. 2006 , , 453 , 969 flaherty , k. m. , pipher , j. l. , megeath , s. t. , winston , e. m. , gutermuth , r. a. , muzerolle , j. , allen , l. e. , & fazio , g. g. 2007 , , 663 , 1069 fontani , f. , cesanori , r. , testi , l. , molinari , s. , zhang , q. , brand , j. , & walmsley , c. m. 2004 , , 424 , 179 fuller , g. a. & ladd , e. f. 2002 , apj , 573 , 699 gmez , y. , rodrguez , l. f. , girart , j. m. , garay , g. , & mart , j. 2003 , , 597 , 414 greene , t. p. , wilking , b. a. , andr , p. , young , e. t. , & lada , c. j. 1994 , , 434 , 614 gutermuth , r. a. , megeath , s. t. , muzerolle , j. m. , allen , l. e. , pipher , j. l. , myers , p. c. , & fazio , g. g. 2004 , , 154 , 374 gutermuth , r. a. , megeath , s. t. , pipher , j. l. , williams , j. p. , allen , l. e. , myers , p. c. , & raines , n. 2005 , , 632 , 397 gutermuth , r. a. , myers , p. c. , megeath , s. t. , allen , l. e. , pipher , j. l , muzerolle , j. , porras , a. , winston , e. , & fazio , g. 2008 , , 674 , 336 hartmann , l. , megeath , s. t. , allen , l. , luhman , k. , calvet , n. , dalessio , p. , franco - hernandez , r. , & fazio , g. 2005 , , 629 , 881 hatchell , j. , fuller , g. a. , richer , j. s. , harries , t. j. , & ladd , e. f. 2007 , , 468 , 1009 haschick , a. d. , reid , m. j. , burke , b. f. , moran , j. m. , & miller , g. 1981 , , 244 , 76 hernndez , j. , hartmann , l. , megeath , s. t. , gutermuth , r. , muzerolle , j. , calvet , n. , vivas , a. k. , briceno , c. , allen , l , stuffer , j. , young , e. , & fazio , g. 2007 , , 662 , 1067 hofner , p. , cesaroni , r. , olmi , l. , rodrguez , l. f. , mart , j. , & araya , e. 2007 , , 465 , 197 hofner , p. , cesaroni , r. , rodrguez , l. f. , & mart , j. 1999 , , 345 , l43 hora j. l. et al .
2004 , proc .
spie , 5487 , 77 hunter , t. r. , brogan , c. l. , megeath , s. t. , menten , k. m. , beuther , h. , & thorwirth , s. 2006 , , 649 , 888 hunter , t. r. , taylor , g. b. , felli , m. , & tofani , g. 1994 , , 284 , 215 hunter , t. r. , testi , l. , taylor , g. b. , tofani , g. , felli , m. , & phillips , t. g. 1995 , , 302 , 249 jiang , z. , yao , y. , yang , j. , ishii , m. , nagata , t. , nakaya , h. , & sato , s. 2001 , , 122 , 313 kumar , m. s. n. , bachiller , r. , & davis , c. j. 2002 , , 576 , 313 kumar , m. s. n. & grave , j. m. c. 2007 , , 472 , 155 kumar , m. s. n. , keto , e. , & clerkin , e. 2006 , , 449 , 1033 lada , c. j. & lada , e. a. 2003 , , 41 , 57 lebron , m. , beuther , h. , schike , p. , & stanke , th .
2006 , , 448 , 1037 lee , c .- f .
, ho , p. t. p. , beuther , h. , bourke , t. l. , zhang , q. , hirano , n. , & shang , h. 2006 , , 639 , 292 longmore , s. n. , burton , m. g. , minier , v. , & walsh , a. j. 2006 , , 369 , 1196 mart , j. , rodgrguez , l. f. , & reipurth , b. , 1993 , , 416 , 208 mart , j. , rodgrguez , l. f. , & reipurth , b. , 1995 , , 449 , 184 mart , j. , rodgrguez , l. f. , & reipurth , b. , 1998 , , 502 , 337 mart , j. , rodgrguez , l. f. , & torrelles , j. m. 1999 , , 345 , l5 megeath , s. t. , allen , l. e. , gutermuth , r. a. , pipher , j. l. , myers , p. c. , calvet , n. , hartmann , l. , muzerolle , j. & fazio , g. g. 2004 , , 154 , 367 megeath , s. t. & tieftrunk , a. r. 1999 , , 526 , l113 molinari , s. , testi , l. , rodrguez , l. f. , & zhang , q. 2002 , , 570 , 758 moore , t. j. t. , mountain , c. m. , & yamashita , t. 1991 , , 248 , 79 muzerolle , j. , megeath , s. t. , gutermuth , r. a. , allen , l. e. , pipher , j. l. , hartmann , l. , gordon , k. d. , padgett , d. l. , noriega - crespo , a. , myers , p. c. , fazio , g. g. , rieke , g. h. , young , e. t. , morrison , j. e. , hines , d. c. , su , k. y. l. , engelbracht , c. w. , & misselt , k. a. 2004 , , 154 , 379 nissen , h. d. , gustafsson , m. , lemaire , j. l. , clnet , y. , rouan , d. , & field , d. 2007 , , 466 , 949 noriega - crespo , a. , moro - martin , a. garay , s. , morris , p. w. , padgett , d. l. , latter , w. b. , & muzerolle , j. 2004a , , 154 , 402 noriega - crespo , a. , morris , p. , marleau , f. r. , carey , s. , boogert , a. , van dishoeck , e. , evans ii , n. j. , keene , j. , muzerolle , j. , stapelfeldt , k. , pontoppidan , k. , lowrance , p. , allen , l. , & bourke , t. l. 2004b , , 154 , 352 palau , a. , ho , p. t. p. , zhang , q. , estalella , r. , hirano , n. , shang , h. , lee , c .- f . , bourke , t. l. , beuther , h. , & kuan , y .- j .
2006 , , 636 , l137 palau , a. , estalella , r. , girart , j. m. , ho , p. t. p. , zhang , q. , & beuther , h. 2007 , , 465 , 219 parker , r. j. & goodwin , s. p. 2007
, , 380 , 1271 persi , p. , tapia , m. , & smith , h. a. 2006 , , 445 , 971 persi , p. , tapia , m. , roth , m. , gmez , m. , & marenzi , a. r. 2005 , in proceedings of iau symposium no . 227 , massive star birth : a crossroad of astrophysics , eds .
cesaroni , r. et al . , 291 - 296 pipher , j. l. et al .
2004 , proc .
spie , 5487 , 234 porras , a. , cruz - gonzlez , i. , & salas , l. 2000 , , 361 , 660 pratap , p. , megeath , s. t. , & bergin , e. a. 1999 , , 517 , 799 raga , a. c. , noriega - crespo , a. , gonzlez , r. f. , & velzquez , p. f. 2004 , , 154 , 346 reach , w. t. et al .
2005 , , 117 , 978 rieke , g. h. et al .
2004 , , 154 , 25 robitaille , t. p. , whitney , b. a. , indebetouw , r. , wood , k. , & denzmore , p. 2006
, , 167 , 256 rho , j. , reach , w. t. , lefloch , b. , & fazio , g. g. 2006 , , 643 , 965 schultz , a. s. b. , colgan , s. w. j. , erickson , e. f. , kaufman , m. j. , hollenbach , d. j. , odell c. r. , young , e. t. , & chen , h. 1999 , , 511 , 282 shang , h. , li , z .- y . , & hirano , n. 2007 , in protostars and planets v. , eds . b. reipurth , d. jewitt , & k. keil , 261 - 276 shepherd , d. s. 2001 , , 546 , 345 shepherd , d. s. , borders , t. , claussen , m. , shirley , y. , & kurtz , s. 2004a , , 614 , 211 shepherd , d. s. & kurtz , s. e. 1999 , , 523 , 690 shepherd , d. s. , kurtz , s. e. , & testi , l. 2004b , , 601 , 952 shepherd , d. s. , testi , l. , & stark , d. p. 2003
, , 584 , 882 shepherd , d. s. , watson , a. m. , sargent , a. i. , & churchwell , e. 1998 , , 507 , 861 shepherd , d. s. , yu , k. c. , bally , j. , & testi , l. 2000 , , 535 , 833 smith , h. a. , hora , j. l. , & marengo , m. 2006 , , 645 , 1264 smith , m. d. & rosen , a. 2005 , , 357 , 1370 sollins , p. k. & megeath , s. t. 2004 , , 128 , 2374 sridharan , t. k. , williams , s. j. , & fuller , g. a. 2005 , , 631 , l73 stecklum , b. , feldt , m. , richichi , a. , calamai , g. , & lagage , p. o. 1997 , , 479 , 339 stern , d. , eisenhardt , p. , gorjian , v. et al .
2005 , , 631 , 163 su , y. n. , liu , s .- y . , chen , h .-
, zhang , q. , cesaroni , r. 2007 , , accepted testi , l. , palla , f. , & natta , a. 1999 , , 342 , 515 tobin , j. l , looney , l. w. , mundy , l. g. , kwon , w. , & hamidouche , m. 2007 , , 659 , 1404 torrelles , j. m. , gmez , j. f. , rodrguez , l. f. , ho , p. t. p. , curiel , s. , & vzquez , r. 1997 , , 489 , 744 torrelles , j. m. , patel , n. a. , anglada , g. , gmez , j. f. , ho , p. t. p. , lada , l. , alberdi , a. , cant , j. , curiel , s. , garay , g. , & rodrguez , l. f. 2003 , , 598 , l115 valusamy , t. , william , d. l. , & marsh , k. a. 2007 , , accepted van den ancker , m. e. , tielens , a. g. g. m. , & wesselius , p. r. 2000 , , 358 , 1035 walther , d. m. , aspin , c. , & mclean , i. s. 1990 , , 356 , 544 winston , e. , megeath , s. t. , wolk , s. j. , muzerolle , j. , gutermuth , r. , hora , j. l. , allen , l. e. , spitzbart , b. , myers , p. , & fazio , g. g. 2007 , , accepted werner , m. w , uchida , k. i. , marengo , m. , gordon , k. d. , morris , p. w. , houck , j. r. , & stansberry , j. a. 2004 , , 154 , 309 whitney , b. a. , indebetouw , r. , babler , b. l. , meade , m. r. , watson , c. , wolff , m. g. , clemens , d. p. , bania , t. m. , benjamin , r. a. , cohen , m. , devine , k. e. , , dickey , j. m. , heitsch , f. , jackson , j. m. , kobulnicky , h. a. , marston , a. p. , mathis , j. s. , mercer , e. p. , stauffer , j. r. , stolovy , s. r. , & churchwell , e. 2004 , , 154 , 315 whitney , b. a. , wood , k. , bjorkman , j. e. , & cohen , m. 2003 , , 598 , 1079 yamshita , t. , murata , y. , kawabe , r. , kaifu , n. , & tamura , m. 1991 , , 373 , 560 yamashita , t. , suzuki , h. , kaifu , n. , & tamura , m. 1989 , , 347 , 894 zhang , q. , hunter , t. r. , beuther , h. , sridharan , t. k. , liu , s .- y .
, su , y .-
, chen , h .- r .
, & chen , y. 2007 , 658 , 1152 afgl 5142 & 05 30 48.00 & + 33 47 53.8 & @xmath95 & 1.8 & 44 + iras 05358 + 3543 & 05 39 12.78 & + 35 45 50.6 & @xmath96 & 1.8 & 30 + g192.16 - 3.82 & 05 58 13.52 & + 16 31 58.3 & @xmath95 & 2.0 & 31 + hh 80 - 81 & 18 19 12.09 & -20 47 30.9 & @xmath97 & 1.7 & 48 + iras 19410 + 2336 & 19 43 11.00 & + 23 44 35.0 & @xmath98 & 2.1 & 46 + iras 20126 + 4104 & 20 14 26.00 & + 41 13 32.0 & @xmath98 & 1.7 & 19 + iras 20293 + 3952 & 20 31 10.70 & + 40 03 10.0 & @xmath96 & 2.0 & 31 + w75n & 20 38 36.45 & + 42 37 34.0 & @xmath99 & 2.0 & 77 + iras 22172 + 5549 & 22 19 08.60 & + 56 05 02.0 & @xmath100 & 2.4 & 91 + afgl 5142 & 517 & [email protected] & 545 & [email protected] & 193 & [email protected] & 128 & [email protected] & 16 & [email protected] + iras 05358 + 3543 & 377 & [email protected] & 410 & [email protected] & 122 & [email protected] & 76 & [email protected] & 12 & [email protected] + g192.16 - 3.82 & 432 & [email protected] & 449 & [email protected] & 130 & [email protected] & 75 & [email protected] & 14 & [email protected] + hh 80 - 81 & 783 & [email protected] & 862 & [email protected] & 578 & [email protected] & 402 & [email protected] & 17 & [email protected] + iras 19410 + 2336 & 750 & [email protected] & 844 & [email protected] & 411 & [email protected] & 203 & [email protected] & 21 & [email protected] + iras 20126 + 4104 & 630 & [email protected] & 808 & [email protected] & 253 & [email protected] & 154 & [email protected] & 7 & [email protected] + iras 20293 + 3952 & 646 & [email protected] & 769 & [email protected] & 282 & [email protected] & 137 & [email protected] & 11 & [email protected] + w75n & 627 & [email protected] & 798 & [email protected] & 227 & [email protected] & 115 & [email protected] & 11 & [email protected] + iras 22172 + 5549 & 713 & [email protected] & 869 & [email protected] & 355 & [email protected] & 250 & [email protected] & 19 & [email protected] + & & & & & afgl 5142 & & & & & + 05 30 36.61 & + 33 49 18.9 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 36.79 & + 33 47 18.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 37.98 & + 33 49 02.2 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 30 40.18 & + 33 46 18.2 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 05 30 42.14 & + 33 48 32.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 42.36 & + 33 48 15.1 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 30 42.67 & + 33 48 59.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 43.39 & + 33 46 00.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 05 30 45.26 & + 33 48 26.9 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 30 46.06 & + 33 47 54.2 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & @xmath102 & high + 05 30 46.26 & + 33 47 20.2 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 30 47.05 & + 33 47 31.5 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 47.04 & + 33 48 43.3 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 47.10 & + 33 45 23.1 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 47.11 & + 33 50 55.4 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ire + 05 30 47.53 & + 33 47 43.3 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & ire + 05 30 47.74 & + 33 47 32.9 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 47.78 & + 33 48 13.0 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 30 47.98 & + 33 47 54.9 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 05 30 48.04 & + 33 48 21.4 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 05 30 47.98 & + 33 47 10.0 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 48.31 & + 33 47 34.6 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 30 48.33 & + 33 48 04.6 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 30 48.41 & + 33 47 23.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 48.85 & + 33 50 21.7 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 30 49.08 & + 33 47 45.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 30 49.19 & + 33 48 18.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 30 49.21 & + 33 47 38.5 & & [email protected] & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 49.69 & + 33 46 24.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 05 30 49.78 & + 33 47 36.9 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 49.82 & + 33 48 17.2 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 30 50.26 & + 33 47 03.0 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 50.51 & + 33 48 05.3 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 05 30 51.09 & + 33 50 11.1 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 30 51.31 & + 33 48 04.7 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 51.41 & + 33 46 48.3 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 51.64 & + 33 48 42.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 51.74 & + 33 47 39.5 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 52.68 & + 33 47 59.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 53.23 & + 33 45 39.5 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 30 54.65 & + 33 47 30.8 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 30 54.92 & + 33 47 18.0 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 55.27 & + 33 46 53.4 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 30 57.18 & + 33 46 39.9 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 39 05.66 & + 35 47 03.0 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 39 07.22 & + 35 43 40.0 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 39 07.80 & + 35 43 59.1 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 39 08.11 & + 35 45 58.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 39 08.50 & + 35 46 37.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 05 39 08.91 & + 35 45 06.3 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 05 39 09.17 & + 35 45 03.2 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 05 39 09.17 & + 35 44 22.6 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 39 09.40 & + 35 46 08.2 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 39 09.54 & + 35 45 05.8 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 05 39 09.92 & + 35 45 17.2 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & @xmath102 & high + 05 39 09.99 & + 35 46 27.1 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 39 10.15 & + 35 44 28.2 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & [email protected] & & ire + 05 39 10.41 & + 35 45 03.5 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 39 11.27 & + 35 45 22.6 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 39 11.50 & + 35 43 33.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 39 11.58 & + 35 45 09.2 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & ire + 05 39 11.58 & + 35 46 22.2 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 39 11.75 & + 35 43 43.5 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 39 11.80 & + 35 44 24.9 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 39 12.13 & + 35 43 39.7 & & [email protected] & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 39 13.35 & + 35 44 17.0 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 39 13.81 & + 35 45 35.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 39 14.66 & + 35 45 23.9 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 39 14.75 & + 35 45 39.6 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 39 14.79 & + 35 46 01.9 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 39 18.15 & + 35 46 08.6 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 39 21.03 & + 35 46 09.3 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 39 22.95 & + 35 44 49.6 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 39 24.79 & + 35 44 45.4 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 05 58 05.44 & + 16 34 17.8 & [email protected] & [email protected] & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 58 09.98 & + 16 34 50.7 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i / ii + 05 58 10.21 & + 16 34 34.5 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 58 10.37 & + 16 33 57.4 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & [email protected] & ii + 05 58 11.15 & + 16 31 20.0 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i / ii + 05 58 11.91 & + 16 32 54.3 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i / ii + 05 58 12.16 & + 16 31 25.1 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 58 13.84 & + 16 31 57.1 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & ire + 05 58 13.93 & + 16 30 00.0 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 58 13.95 & + 16 31 31.0 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 58 13.99 & + 16 30 18.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 58 14.21 & + 16 30 46.1 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ire + 05 58 14.64 & + 16 31 54.7 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 58 14.74 & + 16 31 07.0 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 05 58 14.96 & + 16 30 22.9 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 58 15.23 & + 16 31 13.6 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 05 58 15.26 & + 16 31 39.2 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 05 58 15.44 & + 16 31 02.6 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 58 15.47 & + 16 33 57.7 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 58 15.60 & + 16 30 58.3 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 58 15.68 & + 16 32 30.0 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 58 15.78 & + 16 30 26.3 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 58 16.02 & + 16 31 01.2 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 58 16.38 & + 16 30 41.9 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 58 16.49 & + 16 31 32.4 & & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 05 58 16.58 & + 16 30 29.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 58 17.46 & + 16 30 15.4 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 05 58 17.53 & + 16 31 19.0 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 58 18.16 & + 16 34 39.9 & [email protected] & [email protected] & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 05 58 21.27 & + 16 29 34.2 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 05 58 22.19 & + 16 33 12.1 & & & & [email protected] & [email protected] & & & [email protected] & ii + 18 18 59.59 & -20 45 53.8 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & ire + 18 19 01.08 & -20 45 47.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 02.02 & -20 47 17.2 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 18 19 03.13 & -20 47 00.9 & [email protected] & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 03.81 & -20 45 06.3 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 18 19 04.51 & -20 45 26.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 05.77 & -20 49 20.0 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 18 19 06.24 & -20 49 19.3 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 07.77 & -20 48 45.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 08.97 & -20 47 36.0 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 18 19 09.11 & -20 47 15.0 & [email protected] & [email protected] & [email protected] & & [email protected] & & & & ire + 18 19 09.84 & -20 47 15.6 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 18 19 09.98 & -20 44 49.5 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 10.51 & -20 46 57.8 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & @xmath102 & high + 18 19 10.76 & -20 46 41.4 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & ire + 18 19 11.00 & -20 46 38.5 & & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 18 19 11.47 & -20 48 47.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 18 19 12.08 & -20 47 50.9 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & @xmath102 & high + 18 19 12.26 & -20 44 53.3 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 12.95 & -20 48 45.6 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 13.43 & -20 46 25.3 & [email protected] & [email protected] & & [email protected] & [email protected] & & & & ire + 18 19 13.64 & -20 47 50.5 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 18 19 13.79 & -20 47 40.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 18 19 14.02 & -20 45 09.7 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 14.19 & -20 47 59.8 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 18 19 14.53 & -20 48 28.4 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 14.64 & -20 46 59.2 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i / ii + 18 19 14.68 & -20 49 29.1 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 14.79 & -20 45 58.7 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 14.83 & -20 47 14.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 15.10 & -20 47 46.0 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & @xmath102 & high + 18 19 15.88 & -20 46 05.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 18 19 16.16 & -20 48 17.2 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 18 19 17.95 & -20 45 10.4 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 18 19 20.89 & -20 46 32.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 20.95 & -20 46 09.6 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 18 19 21.10 & -20 45 55.9 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 18 19 21.16 & -20 47 41.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 21.37 & -20 45 30.5 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 21.52 & -20 45 56.6 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 21.78 & -20 45 35.9 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 18 19 22.00 & -20 48 51.5 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 18 19 22.08 & -20 45 36.9 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 22.77 & -20 46 45.2 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 22.86 & -20 44 41.2 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 18 19 23.35 & -20 47 54.1 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 23.53 & -20 48 18.3 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 18 19 23.59 & -20 45 43.0 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 19 42 56.93 & + 23 46 32.0 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 19 42 57.66 & + 23 44 49.3 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 19 43 02.75 & + 23 43 18.3 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 19 43 03.34 & + 23 45 09.4 & & [email protected] & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 19 43 03.60 & + 23 46 50.6 & [email protected] & [email protected] & & [email protected] & [email protected] & & & & ire + 19 43 04.53 & + 23 42 32.8 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 19 43 05.75 & + 23 41 35.0 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 19 43 06.25 & + 23 42 44.2 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ire + 19 43 06.91 & + 23 43 58.0 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & [email protected] & high + 19 43 08.18 & + 23 43 41.4 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 19 43 08.60 & + 23 44 25.6 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & ire + 19 43 09.04 & + 23 44 08.9 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 19 43 09.18 & + 23 44 13.9 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 19 43 09.41 & + 23 47 05.9 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 19 43 09.43 & + 23 46 12.7 & [email protected] & [email protected] & & [email protected] & [email protected] & [email protected] & & & ire + 19 43 09.45 & + 23 44 34.0 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 19 43 09.78 & + 23 45 07.4 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 19 43 10.61 & + 23 44 57.6 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 19 43 10.77 & + 23 45 07.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 19 43 11.45 & + 23 44 50.6 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 19 43 11.53 & + 23 43 37.6 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 19 43 11.59 & + 23 47 28.6 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 19 43 11.60 & + 23 44 58.8 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 19 43 11.85 & + 23 43 53.5 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 19 43 11.91 & + 23 44 52.3 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 19 43 12.41 & + 23 45 25.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 19 43 12.85 & + 23 45 18.6 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 19 43 12.95 & + 23 45 01.7 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & ire + 19 43 13.33 & + 23 46 04.7 & [email protected] & [email protected] & & & [email protected] & & & & ire + 19 43 14.22 & + 23 45 26.9 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 19 43 14.72 & + 23 44 26.1 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 19 43 14.95 & + 23 45 33.9 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 19 43 15.09 & + 23 44 38.4 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 19 43 15.29 & + 23 45 26.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 19 43 15.61 & + 23 43 56.1 & & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 19 43 16.01 & + 23 42 58.2 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 19 43 16.43 & + 23 44 21.2 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 19 43 16.73 & + 23 45 18.2 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 19 43 16.87 & + 23 45 08.9 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 19 43 16.90 & + 23 45 13.1 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 19 43 21.22 & + 23 46 25.0 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ire + 19 43 21.59 & + 23 44 33.2 & & & [email protected] & [email protected] & [email protected] & [email protected] & & [email protected] & ii + 19 43 22.08 & + 23 46 06.6 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 19 43 22.12 & + 23 45 09.1 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 19 43 22.10 & + 23 46 38.8 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 19 43 25.12 & + 23 43 02.0 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 14 18.59 & + 41 14 33.7 & [email protected] & [email protected] & & [email protected] & [email protected] & [email protected] & & & ire + 20 14 19.05 & + 41 13 36.3 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 20 14 20.46 & + 41 13 41.3 & [email protected] & [email protected] & & [email protected] & [email protected] & [email protected] & & & ire + 20 14 23.21 & + 41 13 25.1 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 14 24.51 & + 41 11 22.1 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 20 14 24.99 & + 41 13 32.0 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i / ii + 20 14 26.57 & + 41 15 29.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 14 27.12 & + 41 14 44.0 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 20 14 28.18 & + 41 13 17.0 & [email protected] & [email protected] & [email protected] & & [email protected] & & & & ire + 20 14 28.38 & + 41 13 51.7 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 14 33.58 & + 41 12 30.6 & [email protected] & [email protected] & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 14 34.67 & + 41 10 59.6 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 14 34.91 & + 41 12 31.3 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 20 14 37.57 & + 41 12 34.8 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 20 14 38.69 & + 41 12 23.2 & & [email protected] & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 14 39.92 & + 41 14 35.4 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i / ii + 20 14 41.17 & + 41 12 55.8 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ire + 20 14 43.33 & + 41 13 40.7 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 14 44.47 & + 41 12 59.8 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 30 57.60 & + 40 02 22.7 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ire + 20 31 00.75 & + 40 02 04.1 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 20 31 04.55 & + 40 01 46.0 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 31 08.05 & + 40 04 48.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 31 08.59 & + 40 02 37.5 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & ire + 20 31 08.68 & + 40 03 48.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 20 31 08.75 & + 40 02 47.4 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 20 31 09.52 & + 40 00 22.1 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 31 09.59 & + 40 03 04.8 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & ire + 20 31 09.85 & + 40 02 37.6 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 31 10.32 & + 40 03 16.5 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & @xmath102 & high + 20 31 10.41 & + 40 04 08.2 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 31 11.26 & + 40 03 07.6 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & @xmath102 & high + 20 31 11.78 & + 40 03 48.8 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 31 12.38 & + 40 03 10.2 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 31 12.48 & + 40 03 20.0 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 31 13.87 & + 40 06 52.4 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 31 14.52 & + 40 03 26.1 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & ire + 20 31 14.64 & + 40 03 18.6 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 20 31 15.35 & + 40 02 53.8 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 31 15.69 & + 40 03 11.2 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 31 16.27 & + 40 02 55.2 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 31 17.05 & + 40 03 40.1 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 31 17.17 & + 40 03 49.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 31 17.46 & + 40 03 28.5 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 20 31 18.23 & + 40 02 53.9 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 20 31 22.88 & + 40 03 09.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 20 31 22.98 & + 40 02 21.1 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 31 25.54 & + 40 01 59.2 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 20 31 27.93 & + 40 03 35.6 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 20 31 28.05 & + 40 03 40.5 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i / ii + 20 38 19.40 & + 42 36 57.1 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 20.48 & + 42 37 13.6 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 22.84 & + 42 37 41.9 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ire + 20 38 23.18 & + 42 36 41.6 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 23.49 & + 42 37 37.4 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ire + 20 38 25.02 & + 42 35 27.2 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 25.25 & + 42 36 26.9 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 25.89 & + 42 36 35.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 26.86 & + 42 39 29.1 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 27.34 & + 42 38 29.8 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 28.58 & + 42 37 01.1 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 31.27 & + 42 38 24.9 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 31.63 & + 42 38 34.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 20 38 31.96 & + 42 37 40.9 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ire + 20 38 32.20 & + 42 36 37.3 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ire + 20 38 32.31 & + 42 39 45.2 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 20 38 32.29 & + 42 36 31.4 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 32.67 & + 42 40 07.1 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 32.82 & + 42 39 35.3 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 33.14 & + 42 38 07.4 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 33.29 & + 42 39 32.3 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 33.29 & + 42 39 38.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 20 38 33.42 & + 42 37 06.4 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 20 38 33.60 & + 42 37 27.3 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 33.64 & + 42 36 53.2 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 33.73 & + 42 39 21.9 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 33.83 & + 42 36 15.5 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 34.06 & + 42 39 55.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 20 38 34.14 & + 42 37 58.2 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 34.17 & + 42 38 40.9 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 34.39 & + 42 39 18.1 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 34.51 & + 42 38 07.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 34.66 & + 42 36 20.5 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i / ii + 20 38 35.28 & + 42 36 50.1 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 35.37 & + 42 37 13.7 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & @xmath102 & high + 20 38 36.20 & + 42 38 07.8 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 20 38 36.26 & + 42 38 26.0 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 36.34 & + 42 39 47.6 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 36.43 & + 42 40 41.8 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 20 38 36.89 & + 42 39 52.1 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 20 38 37.01 & + 42 39 34.8 & [email protected] & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i / ii + 20 38 37.29 & + 42 39 33.0 & [email protected] & & [email protected] & [email protected] & [email protected] & & [email protected] & [email protected] & ii + 20 38 37.50 & + 42 39 10.4 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 37.52 & + 42 39 26.3 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 37.73 & + 42 37 59.4 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & @xmath102 & high + 20 38 37.87 & + 42 40 00.1 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 37.91 & + 42 39 21.4 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 38.15 & + 42 39 07.5 & & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 20 38 38.46 & + 42 39 50.0 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 38.64 & + 42 39 39.2 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 38.74 & + 42 39 52.3 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 20 38 38.82 & + 42 37 17.6 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & high + 20 38 38.83 & + 42 39 26.6 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ire + 20 38 38.86 & + 42 37 43.9 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 39.01 & + 42 39 22.0 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 39.37 & + 42 37 11.6 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & ire + 20 38 39.47 & + 42 38 56.9 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 39.89 & + 42 36 42.6 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 40.07 & + 42 38 09.2 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 20 38 40.16 & + 42 39 53.5 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & [email protected] & high + 20 38 40.42 & + 42 35 45.9 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 20 38 41.16 & + 42 39 23.1 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 20 38 42.63 & + 42 35 43.7 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 20 38 43.05 & + 42 39 04.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 43.06 & + 42 36 59.7 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 43.84 & + 42 37 13.7 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 43.86 & + 42 36 52.9 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 43.93 & + 42 37 44.6 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 44.42 & + 42 38 58.4 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 45.04 & + 42 39 40.2 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 45.56 & + 42 38 01.9 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 20 38 46.22 & + 42 40 01.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 46.69 & + 42 37 44.0 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 47.43 & + 42 38 19.9 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 48.52 & + 42 37 30.0 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 20 38 49.47 & + 42 36 08.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 20 38 51.13 & + 42 35 51.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 22 18 45.18 & + 56 04 49.3 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 45.73 & + 56 03 52.6 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & ire + 22 18 46.31 & + 56 05 03.4 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 47.26 & + 56 05 58.4 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & ire + 22 18 47.54 & + 56 05 11.2 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 48.01 & + 56 05 02.9 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 49.37 & + 56 05 46.5 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 49.40 & + 56 04 31.0 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 49.86 & + 56 06 18.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 50.13 & + 56 03 40.8 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & ire + 22 18 50.78 & + 56 05 06.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 52.22 & + 56 06 13.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 52.49 & + 56 06 48.8 & [email protected] & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 52.69 & + 56 06 05.1 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 22 18 53.62 & + 56 06 24.8 & [email protected] & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 54.35 & + 56 04 59.6 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 22 18 55.39 & + 56 05 12.9 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 55.67 & + 56 05 16.3 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 55.67 & + 56 04 55.4 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 56.14 & + 56 05 48.6 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 22 18 56.27 & + 56 05 32.3 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 57.18 & + 56 05 02.0 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 22 18 57.62 & + 56 07 07.0 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i / ii + 22 18 57.71 & + 56 04 27.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 57.82 & + 56 05 52.6 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 58.38 & + 56 04 23.3 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 22 18 58.68 & + 56 07 23.6 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ire + 22 18 58.71 & + 56 04 54.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 58.76 & + 56 05 18.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 59.21 & + 56 05 00.6 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 18 59.72 & + 56 05 13.9 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 00.08 & + 56 03 14.8 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 22 19 00.18 & + 56 05 34.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 22 19 00.26 & + 56 06 41.4 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 02.01 & + 56 07 08.5 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 02.01 & + 56 08 00.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 02.18 & + 56 07 25.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 02.26 & + 56 06 36.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 02.62 & + 56 06 17.3 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 02.77 & + 56 05 40.6 & [email protected] & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & i / ii + 22 19 03.80 & + 56 05 10.3 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 04.36 & + 56 05 50.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 22 19 04.46 & + 56 05 36.0 & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 04.50 & + 56 05 07.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 22 19 05.21 & + 56 08 25.8 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 05.72 & + 56 04 52.7 & & [email protected] & [email protected] & [email protected] & [email protected] & & & & ire + 22 19 05.82 & + 56 05 18.8 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 05.79 & + 56 07 05.9 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 05.94 & + 56 05 41.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 06.02 & + 56 05 55.3 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 07.65 & + 56 05 28.6 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & ire + 22 19 08.03 & + 56 04 11.1 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 22 19 08.30 & + 56 05 11.0 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 22 19 08.45 & + 56 05 28.7 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & ire + 22 19 09.39 & + 56 04 45.7 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 22 19 09.38 & + 56 05 00.4 & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 22 19 09.70 & + 56 05 04.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i + 22 19 09.84 & + 56 07 37.4 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 09.94 & + 56 08 20.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 10.42 & + 56 02 11.5 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 10.94 & + 56 06 07.6 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 11.45 & + 56 02 53.2 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & ire + 22 19 11.62 & + 56 02 36.6 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 22 19 12.07 & + 56 07 09.1 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & i / ii + 22 19 12.65 & + 56 02 03.6 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 13.80 & + 56 07 48.4 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & & ire + 22 19 15.06 & + 56 06 53.0 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 15.18 & + 56 07 02.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 17.51 & + 56 03 14.2 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 17.89 & + 56 04 01.7 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 18.33 & + 56 02 56.0 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 19.80 & + 56 03 01.5 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & i + 22 19 20.39 & + 56 03 49.3 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 20.78 & + 56 06 12.3 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 21.06 & + 56 06 31.0 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 21.24 & + 56 06 40.1 & & & & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 22 19 21.35 & + 56 04 48.3 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 22 19 21.58 & + 56 04 20.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 21.80 & + 56 06 03.9 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 21.81 & + 56 03 06.1 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 22 19 22.67 & + 56 05 57.0 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 22.85 & + 56 06 02.3 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 23.28 & + 56 03 54.9 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 23.81 & + 56 06 00.4 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 24.50 & + 56 03 36.0 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 24.54 & + 56 03 21.3 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 22 19 25.98 & + 56 04 10.2 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & ii + 22 19 27.48 & + 56 04 47.8 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 27.90 & + 56 06 08.2 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 28.81 & + 56 04 53.9 & & & & [email protected] & [email protected] & [email protected] & [email protected] & & ii + 22 19 31.32 & + 56 06 00.8 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & ii + afgl 5142 & 15 & 84% + iras 05358 + 3543 & 14.5 & 79% + g192.16 - 3.82 & 15 & 81% + hh 80 - 81 & 13.5 & 84% + iras 19410 + 2336 & 13.5 & 73% + iras
20126 + 4104 & 14.5 & 73% + iras 20293 + 3952 & 13 & 85% + w75n & 12.5 & 75% + iras 22172 + 5549 & 14.5 & 76% + afgl 5142 & 05 30 47.94 & + 33 47 54.2 & & & & & [email protected] & 1 + iras 05358 + 3543 & 05 39 13.04 & + 35 45 51.2 & & [email protected] & [email protected] & [email protected] & @xmath1030.626 & 2 , 3 + g192.16 - 3.82 & 05 58 13.55 & + 16 31 58.2 & [email protected] & [email protected] & [email protected] & [email protected] & @xmath1030.626 & 4 , 5 + hh 80 - 81 & 18 19 12.11 & -20 47 31.0 & & & [email protected] & [email protected] & @xmath1030.626 & 6 + iras 19410 + 2336 & 19 43 11.25 & + 23 44 03.2 & [email protected] & [email protected] & [email protected] & [email protected] & @xmath1030.626 & 7 , 8 + iras 20126 + 4104 & 20 14 25.97 & + 41 13 33.1 & & & & [email protected] & @xmath1030.626 & 9 , 10 + iras 20293 + 3952 & 20 31 12.92 & + 40 03 23.0 & & & & & [email protected] & 11 , 12 , 13 + w75n & 20 38 36.49 & + 42 37 33.5 & & [email protected] & [email protected] & [email protected] & @xmath1030.626 & 14 + | we present _ spitzer _ space telescope irac and mips observations toward a sample of nine high - mass star forming regions at a distance of around 2 kpc . based on irac and mips 24 @xmath0 m photometric results and 2mass _ jhks _ data
, we carry out a census of young stellar objects ( ysos ) in a @xmath1 field toward each region . toward seven out of the nine regions ,
we detect parsec sized clusters with around 20 ysos surrounded by a more extended and sparse distribution of young stars and protostars . for the other two regions , iras 20126 + 4104 and iras 22172 + 5549 ,
the former has the lowest number of ysos in the sample and shows no obvious cluster , and the latter appears to be part of a larger , potentially more evolved cluster .
the deep irac imaging reveals at least twelve outflows in eight out of the nine regions , with nine outflows prominent in the 4.5 @xmath0 m band most probably attributed to shocked h@xmath2 emission , two outflows dominated by scattered light in the 3.6 and 4.5 @xmath0 m bands , and one outflow standing out from its hydrocarbon emission in the 8.0 @xmath0 m band . in comparison with previous ground - based observations ,
our irac observations reveal new outflow structures in five regions .
the dramatically different morphologies of detected outflows can be tentatively interpreted in terms of possible evolution of massive outflows .
the driving sources of these outflows are deeply embedded in dense dusty cores revealed by previous millimeter interferometric observations .
we detect infrared counterparts of these dusty cores in the irac or mips 24 @xmath0 m bands .
reflection nebulae dominated by the emission from uv heated hydrocarbons in the @xmath3 m band can be found in most regions and they may imply the presence of young b stars . |
it is vital to obtain a good description of a network topology because structure fundamentally affects function @xcite .
the internet contains millions of routers , which are grouped into thousands of subnetworks , called _ autonomous systems _ ( as ) , which are then glued into a global network by the border gateway protocol .
effective engineering of the internet is predicated on a detailed understanding of issues such as the large - scale structure of its underlying physical topology , the manner in which it evolves over time , and the way in which its constituent components contribute to its overall function @xcite .
the recently introduced positive - feedback preference ( pfp ) model @xcite is an internet as - level topology generator .
the model uses two evolution mechanisms , namely the interactive growth and the positive - feedback preference .
both mechanisms are inspired by , and coincident with , practical observations on the internet historic data .
originally the phenomenological model was designed to match only the internet s rich - club connectivity @xcite and the exact form of degree distribution , including the distribution of low degrees , the maximum degree and the degree distribution s power - law exponent @xcite .
it has been a pleasant surprise that numerical evaluation against the internet measurement date has shown that the model accurately reproduces a large set of other nontrivial topology characteristics as well , including the disassortative mixing @xcite , the shortest path length @xcite , short cycles @xcite and the betweenness centrality @xcite .
the pfp model has been used in more realistic simulations of the internet , e.g. @xcite . as of this writing , an analytical solution of the pfp model
is not available yet . in this paper
we use the numerical method to analyze why and how the model is able to reproduce a fuller picture of the internet than it was designed for . in other words
, we aim to investigate the exact origin of each of the topology properties captured by the pfp model .
answers to these questions can be valuable for the ongoing effort on a mathematical solution of the model . in section
ii , we review the pfp model and its two mechanisms .
we reflect on the intuitions underlying the design of the two evolution mechanisms . in section iii
, we comparatively examine two limiting cases of the pfp model .
based on detailed numerical simulations , we identify the explicit evolution dynamics that are responsible for generating the rich - club phenomenon and the degree distribution properties .
we also reveal that the origin of the disassortative mixing is in fact already embedded in the model s two evolution mechanisms .
furthermore we explain that the pfp model resembles the internet s shortest path length and short cycles because these two characteristics are correlated with other topology properties . in section iv ,
the above results become more evident when we examine how the pfp model s topology properties react to the change of the parameters that control the model s evolution mechanisms .
our investigation leads to a number of insightful discoveries .
for example we find out that the rich - club connectivity is almost exclusively determined by the interactive growth mechanism .
we also show that the interactive growth and the positive - feedback preference jointly contribute to the model s disassortative mixing behavior , however the two mechanisms have opposite effects on the degree distribution s power - law exponent . in section v we conclude that this work represents a fuller and deeper understanding of the internet topology evolution dynamics .
this work complements the research on evolution mechanisms and structural constrains of complex networks in general .
in graph theory , degree , @xmath0 , is defined as the number of links a node has .
degree is a local property but the distribution of degree provides a global view of a network structure .
the degree distribution of the internet as - level topology is characterized by a power - law @xcite as @xmath1 a power - law degree distribution means the majority of nodes that form a network have very few links .
some nodes , however , are very well connected .
barabsi and albert ( ba ) @xcite showed that a power - law degree distribution can arise from two generic mechanisms : _ growth _ , where a network `` grows '' from a small random network by attaching new nodes to @xmath2 old nodes in the existing system ( see fig.[fig : growth]a , in which @xmath3 ) ; and _ preferential attachment _ , where a new node is attached preferentially to nodes that are already well connected .
the probability that a new node will be connected to a node with degree @xmath0 is given as @xmath4 during the last few years a large number of models have been proposed to resemble and explain the power - law degree distribution and other topology properties observed on various real networks @xcite .
recently zhou and mondragn @xcite introduced the positive - feedback preference ( pfp ) model .
numerical evaluation has shown that the pfp model accurately reproduces a large number of characteristics of the internet as - level topology .
the model uses the following two evolution mechanisms .
the interactive growth is designed to satisfy a number of observations on internet history data@xcite .
firstly the internet evolution is largely due to two processes : the attachment of new nodes to the existing system , and the addition of new internal links between old nodes already present in the existing system .
secondly the majority of new nodes appeared on the internet are each attached to no more than two old nodes . and
thirdly the ratio of links to nodes on the internet as graph is approximately three .
different from the independent growth @xcite ( see fig.[fig : growth]b ) , in which new nodes and new internal links are added independently , the two evolution processes are interdependent in the interactive growth ( see fig.[fig : growth]c and d ) .
that is to say , a new internal link always starts from an old node , we call it a _ host _ node , to whom a new node has just attached .
a heuristic explanation of this interaction is that on the internet , new customers ( new nodes ) generate extra demand for service , which triggers their service providers ( host nodes ) to develop new connections to other service providers ( new internal links ) in order to accommodate the increased traffic and improve services .
the interactive growth is described as such : starting from a small random graph , at each time step , * with probability @xmath5 $ ] , a new node is attached to a _
node , and at the same time two new internal links are added connecting the host node to two other old nodes ( see fig.[fig : growth]c ) ; * with probability @xmath6 , a new node is attached to two host nodes , and only one new internal link is added connecting one of the host nodes to another old node ( see fig.[fig : growth]d ) .
numerical simulation shows that when the probability parameter @xmath7 , the interactive growth produces the best result .
the pfp model uses the following nonlinear preference probability to choose old nodes for the interactive growth , @xmath8 it is called the positive - feedback preference ( pfp ) because a node s ability of acquiring new links increases as a feed - back loop of the node s degree . when the parameter @xmath9 , it is equivalent to the ba model s linear preference ( see eq.[eq : ba ] ) .
numerical simulation shows that the positive - feedback preference produces the best result when the parameter @xmath10 of the positive - feedback preference . accordingly the value of parameter @xmath11 is revised as @xmath12 . ] .
the positive - feedback preference is designed to satisfy the observation @xcite that during the internet evolution , the probability that a new node links with a low - degree node follows the ba model s linear preference , whereas high - degree nodes have a stronger ability of acquiring new links than predicted by the linear preference .
a preference probability can be written as @xmath13 , where @xmath14 is a degree function .
the exponential preference @xcite , which uses the nonlinear degree function of @xmath15 , @xmath16 , also gives high - degree nodes a stronger preference than the linear preference .
however as shown in fig.[fig : degreefunctions ] , only the pfp can be approximated by the linear preference for low - degree nodes .
when degree increases , the pfp grows more and more rapidly and finally exceeds the exponential preference .
.preference probability ratio @xmath17 when @xmath18 and @xmath19 . [
cols="^,^,^,^",options="header " , ] table [ table : parameter - delta ] and fig.[fig : parameter - delta ] show how topology properties of the pfp model change when the positive - feedback parameter @xmath11 increases from 0 to 0.035 while the interactive growth parameter @xmath7 .
it is clear that the positive - feedback parameter @xmath11 has a fairly limited impact on the obtained rich - club exponent @xmath20 , which almost remains the same .
when @xmath11 increases , greater preference is given to high - degree nodes , consequently the maximum degree increases and the network becomes more disassortative mixed .
as expected , the characteristic path length decreases steadily when the network s disassortative mixing is getting stronger
. when the parameter @xmath11 increases , the degree distribution power - law exponent @xmath21 slightly decreases .
this is because when the positive - feedback preference gets stronger , the richest nodes attract so many new links that they suppress the degree growth of other nodes , including those with medium to high degrees .
when @xmath10 ( and @xmath7 ) , the pfp model closely matches all the topology properties of the internet as graph . as shown in fig.[fig : parameter - delta ] , by tuning the parameter @xmath11 from 0 to 0.028 , the pfp model is capable of generating a wide range of disassortative mixing networks with the value of the assortativity coefficient @xmath22 decreasing monotonically from -0.12 to -0.28 , which encompasses most technological and biological networks reported in @xcite .
notably , the pfp model achieves this while keeping the rich - club exponent and the degree distribution exponent largely unchanged .
@xmath23 is the model s tipping point , beyond which the network structure changes dramatically , e.g. the degree distribution exponent @xmath21 decreases rapidly .
this is because the preferential selection becomes so biased that super - rich nodes start to emerge .
these extremely rich nodes dominate the network growth and make the network grow towards a star - like structure .
fig.[fig : parameter - theta][fig : parameter - alpha ] are contour plots showing how three topology properties , i.e. the rich - club connectivity exponent @xmath20 , the degree distribution exponent @xmath21 and the assortativity coefficient @xmath22 , change when both the parameters are variables .
fig.[fig : parameter - theta ] clearly shows that the value of rich - club exponent @xmath20 is sensitive to the interactive growth parameter @xmath24 and is unsensitive to the positive - feedback preference parameter @xmath11 .
when parameter @xmath24 grows , the rich - club phenomenon becomes stronger .
fig.[fig : parameter - gamma ] shows that the two mechanisms have opposite effects on the value of degree distribution exponent @xmath21 .
in general , the exponent @xmath21 increases when parameter @xmath24 increases and parameter @xmath11 decreases .
fig.[fig : parameter - alpha ] shows that the pfp model s disassortative mixing becomes stronger when either of the two parameters increases .
the assortativity coefficient @xmath22 is more sensitive to parameter @xmath11 than to parameter @xmath24 .
we notice that when @xmath25 , the network s disassortative mixing actually becomes weaker .
this is because , when @xmath25 , the network starts to acquire more internal links than external links , i.e. @xmath26 ( see eq .
[ eq : link - ratio ] ) .
the increased new internal links among the rich nodes weaken the disassortative mixing behavior of the network .
the internet has an extremely complex structure .
the pfp model demonstrates a way to simultaneously reproduce many topology properties of the internet .
the model achieves this by using two generic algorithms which are designed to replicate the evolution dynamics observed on the internet historic data . in this paper
we have used the numerical method to investigate the success of the model . based on detailed simulation results ,
we show that the rich - club phenomenon is primarily controlled by the interactive growth mechanism alone .
we point out that this is because the rich - club connectivity increases with the number of new internal links which are added between rich nodes .
the interactive growth also determines the probability of new nodes initial degrees and thus controls the distribution of low degrees obtained in the generated network .
the pfp model s maximum degree is as large as the internet because both the interactive growth and the positive - feedback preference accelerate the degree growth rate .
the interactive growth does so by the degree - leap effect , whereas the positive - feedback preference achieves so by giving stronger preference to high - degree nodes .
the two mechanisms have opposite effects on the value of power - law exponent of degree distribution .
the interactive growth increases the value whereas the positive - feedback preference slightly decreases the value .
we have explained that the origin of the disassortative mixing has been , unintentionally , embedded in the pfp model s two evolution mechanisms , which not only enlarge the gap between the high - degree nodes and the low - degree nodes , but more critically , they increase the degree difference between end nodes of a link .
our analysis suggests that the reason that the pfp model also captures other properties of the internet , such as shortest path length and triangle coefficient , is that these properties are correlated with other properties , in other words , they are constrained by other .
for example the internet is small because the rich - club functions as a super traffic hub , while the disassortative mixing behavior ensures peripheral low - degree nodes always close to the hub . by investigating the pfp model s sensitivity to the mechanism parameters , we obtain more evident details on how the two evolution mechanisms profoundly control the model s growth dynamics and therefore effectively change the generated topology properties .
for example we show that by tuning the positive - feedback parameter @xmath11 , the pfp model is able to generate a wide range of disassortative mixing networks while keeping the rich - club exponent and the degree distribution exponent almost unchanged .
this work represents a better understanding of the internet topology evolution dynamics .
it facilitates the ongoing effort on a mathematical analysis of the pfp model .
this work also complements the research on the evolution dynamics and the structural constrains of complex networks in general . | the internet structure is extremely complex .
the positive - feedback preference ( pfp ) model is a recently introduced internet topology generator .
the model uses two generic algorithms to replicate the evolution dynamics observed on the internet historic data .
the phenomenological model was originally designed to match only two topology properties of the internet , i.e. the rich - club connectivity and the exact form of degree distribution . whereas numerical evaluation has shown that the pfp model accurately reproduces a large set of other nontrivial characteristics as well .
this paper aims to investigate why and how this generative model captures so many diverse properties of the internet . based on comprehensive simulation results , the paper presents a detailed analysis on the exact origin of each of the topology properties produced by the model .
this work reveals how network evolution mechanisms control the obtained topology properties and it also provides insights on correlations between various structural characteristics of complex networks . |
in approximating smooth submanifolds by polyhedra , it is well known that pathologies may occur if the triangles of the polyhedra become too thin .
for example , the famous schwartz lantern " ( cf .
@xcite , section 3.1.3 ) shows that the areas of a sequence of polyhedra @xmath8 approximating a surface @xmath9 may fail to converge unless the ratio of the area of a general constituent triangle to the square of its diameter is bounded away from zero . on the other hand , if this
fatness " condition is imposed then the sequence enjoys convergence not only to first order ( i.e. convergence of areas ) , but also to second order in the sense that certain curvature integrals ( or , more precisely , their polyhedral analogues ) for the @xmath8 converge weakly to the corresponding integrals for @xmath9 ( @xcite ) .
the key observation is that the corresponding _ absolute _ curvature integrals of the @xmath8 remain bounded in the course of the approximation process
. however , there is a further subtlety : even if the fatness condition is imposed , the bound on the absolute curvature integrals of the @xmath8 depends on the @xmath10 norm of @xmath9 , which may be much larger than its absolute curvature integral ( cf .
section [ sect : quadratic ] below ) . in the present paper
we show that by means of a more careful approach , the @xmath8 may be chosen so that their absolute curvature integrals are bounded by universal constant multiples of the corresponding integrals for @xmath9 . in order to isolate the analytic aspects of the process , instead of a smooth surface @xmath9 we consider smooth and piecewise linear ( pl ) _ functions _ @xmath11 defined on an open subset @xmath12 , replacing curvatures of @xmath9 by invariants of the hessian of @xmath11 .
the curvature integrals of @xmath9 that we consider are artifacts of the manifold @xmath13 of unit normals to @xmath9 . using the area formula ,
it is easy to see that the area of @xmath14 is bounded above and below by constant multiples of @xmath15 , where @xmath16 is the second fundamental form . by the same token
, the area of the graph of the differential @xmath7 of a @xmath10 function is bounded above and below by constant multiples of @xmath17 , where @xmath18 is the hessian ( cf . ) . for certain classes of singular surfaces @xmath9 and functions @xmath11
, it is known that there exist integral currents @xmath14 and @xmath19 that serve , respectively , as suitable replacements for the manifold of unit normals and the graph of the differential .
we state the relevant theorem in the latter case .
[ @xcite][thm : ma ] let @xmath12 be an open set and let @xmath11 be a function with locally integrable differential @xmath7 ( i.e. @xmath11 belongs to the sobolev space @xmath20 ) .
then there is at most one integral current @xmath19 of dimension @xmath21 in the cotangent bundle @xmath22 , with the properties * @xmath23 * @xmath19 is * lagrangian * , in the sense that @xmath24 , where @xmath25 is the canonical symplectic form of @xmath26 * if @xmath27 then @xmath28 * for any compactly supported continuous function @xmath29 @xmath30 here @xmath31 is the projection .
if @xmath19 exists then we say that @xmath11 is a _ monge - ampre ( ma ) function _ , and refer to @xmath19 as the * gradient cycle * of @xmath11 .
if @xmath32 ( i.e. @xmath33 and its derivative is locally lipschitz ) then @xmath19 is given by integration over the graph of the differential @xmath7 , which is a lipschitz submanifold of @xmath26 with orientation induced by the standard orientation of @xmath12 .
thus if @xmath11 is not smooth then we may think of @xmath19 as a substitute for this graph , and its existence is an indication that @xmath11 has good second order properties .
it is not difficult ( cf .
@xcite ) to construct @xmath2 directly for a pl function @xmath0 ; for completeness , and to fix ideas , we give the construction in detail for @xmath34 in section [ sect : d of pl ] below . in these terms
our main theorem is the following .
we denote the space of pl functions on an open set @xmath35 by @xmath36 .
[ thm : main ] let @xmath37 be open and @xmath32 . given an open set @xmath38 , there exists a sequence @xmath39 , with @xmath40 uniformly , such that @xmath41 in fact it is enough to prove theorem [ thm : main ] in the case that @xmath42 : convolving with a mollifier we may approximate any @xmath32 locally by @xmath10 functions @xmath43 , in such a way that the hessians of the @xmath43 converge pointwise a.e . to that of @xmath11 . invoking the estimates for the @xmath43 and using a diagonal process
, the desired estimate for @xmath11 follows from the dominated convergence theorem . as mentioned above
, there is a weaker form of theorem [ thm : main ] that holds for any pl approximation in which the underlying triangulations are uniformly fat .
this is given here as proposition [ prop : crude ] . in this result
the masses of the @xmath44 are bounded in terms of the integral of the square of the norm of the hessian matrix @xmath18 . by the area formula , @xmath45 for @xmath42 , where @xmath46^{\frac 1 2}\ ] ] is the hilbert - schmidt norm of the hessian .
thus the bound of theorem [ thm : main ] is stronger in the event that @xmath18 has one small and one large eigenvalue over a large region of @xmath47 .
all functions @xmath11 known to be ma are _ strongly approximable _ in the sense of @xcite , i.e. there is a sequence @xmath48 of @xmath10 functions that converge to @xmath11 in @xmath49 such that the @xmath50 are locally uniformly bounded in mass .
the differential current @xmath19 then arises as the limit of the @xmath50 : the federer - fleming compactness theorem for integral currents implies that a subsequence converges , and the uniqueness result from the first paragraph above implies that the limit is uniquely determined .
a fundamental conjecture states that every ma function is strongly approximable . in view of the simple construction of @xmath19 for pl functions ,
it is natural to consider also the class of * pl strongly approximable * functions , which are defined analogously by taking the approximating functions @xmath48 to be pl instead of @xmath10 .
again one conjectures that any ma function is pl strongly approximable .
theorem [ thm : main ] implies that these two notions coincide in dimension @xmath34 .
let @xmath51 and @xmath52 .
then @xmath11 is strongly approximable iff @xmath11 is pl strongly approximable . that a pl strongly approximable function must be strongly approximable follows from a construction of u. brehm and w. khnel @xcite , which ( in our language ) shows that any pl function @xmath0 of two variables may be approximated by @xmath10 functions @xmath48 in such a way that the local masses of the @xmath50 are bounded in terms of the local mass of @xmath2 .
the other direction follows from theorem [ thm : main ] .
the basic idea of the proof of theorem [ thm : main ] arises from the analysis of pl approximations to a given quadratic form @xmath53 given in section [ sect : example ] : we interpolate linearly between the values of @xmath54 on the vertices of a square grid to obtain a pl function @xmath0 , and show that if the grid is aligned with the eigenvectors of @xmath54 then the mass of @xmath2 over an open subset @xmath55 is roughly proportional to @xmath56 . if , on the other hand , the grid is aligned haphazardly , then this mass is more like @xmath57 , corresponding to the estimate of proposition [ prop : crude ] . in order to carry this analysis over to a given @xmath10 function @xmath11 defined in a neighborhood of a set @xmath35 , we attempt to impose a subdivided square grid of small mesh size so as to align with the eigenvectors of @xmath18 .
since the eigenvectors are not constant this is obviously not possible , so instead we do this in approximate fashion in open sets where the oscillation of @xmath18 is small , the union of which yields a triangulated set @xmath58 such that @xmath59 has small area .
we also take care to ensure that the square grids remain separated from each other by a distance proportional to the mesh size .
this allows us to invoke a well - known ( albeit apparently unpublished ) result from the theory of mesh generation , stating that this partial triangulation of @xmath35 may be completed to a full triangulation with a lower bound on the fatness of the triangles : [ thm : chew ] let @xmath60 , and let @xmath61 be a closed polygonal region , and @xmath62 a system of vertices and edges in @xmath63 , such that no two vertices of @xmath64 are closer than @xmath65 and every edge of @xmath62 has length @xmath66 $ ] .
then there exists a triangulation of @xmath63 that includes @xmath62 such that all edges have length @xmath67 $ ] and all angles @xmath68 $ ] .
our estimates in the quadratic case carry over to yield the desired bounds on the mass of @xmath69 .
meanwhile the mass of @xmath70 is controlled only by the cruder bounds of proposition [ prop : crude ] , but since @xmath59 has small measure this quantity is also small .
the absence of the analogue of theorem [ thm : chew ] in dimensions larger than 2 appears to constitute the principal obstacle to extending theorem [ thm : main ] to higher dimensions . for simplicity we will identify the dual space @xmath71 with @xmath72 in the usual way , and conflate the differential @xmath7 and the gradient @xmath73 of a function @xmath11 .
we would like to thank bob jerrard for interesting and helpful discussions .
a * pl function * on an open subset @xmath74 is a continuous function @xmath0 on @xmath58 , for which there exists a triangulation @xmath62 of a polyhedral set @xmath75 such that the restriction of @xmath0 to every simplex of @xmath62 is affine .
such a @xmath62 is an * underlying triangulation * for @xmath0 .
the space of all such functions is denoted @xmath76 .
given an oriented simplex @xmath77 of dimension @xmath78 , we denote by @xmath79\!]}}$ ] the @xmath78-dimensional current given by integration over @xmath77 .
in particular , if @xmath77 is a point ( i.e. @xmath80 ) then @xmath79\!]}}$ ] is simply evaluation at @xmath77 .
such currents are clearly _ integral currents _ in the sense of @xcite , chapter 4 .
the space of integral currents of dimension @xmath78 on a smooth manifold @xmath81 is denoted @xmath82 .
we discuss only the case @xmath34 . given a pl function @xmath0 defined on an open set @xmath84 , let @xmath62 be an underlying triangulation for @xmath0 consisting of relatively open simplices @xmath85 of dimensions @xmath86 respectively .
put @xmath87 for the ( constant ) value of @xmath88 on @xmath89 , and set @xmath90\ ! ] } } \times { { [ \![}\nabla_i{]\!]}}\in \mathbb i_2(w\times { \mathbb{r}}^2)\ ] ] where the orientations of the @xmath89 are induced by the standard orientation of @xmath72 . thus @xmath91\ ! ] } } \times { { [ \![}\nabla_i{]\!]}}\\ & = \sum_j { { [ \![}\sigma_j{]\ ! ] } } \times\left ( { { [ \![}\nabla_j^+ { ] \!]}}- { { [ \![}\nabla_j^-{]\!]}}\right ) \end{aligned}\ ] ] with appropriate orientations for the edges @xmath92 , where @xmath93 are the respective values of the gradient of @xmath0 on the neighboring triangles lying to the left and to the right of @xmath92 with respect to the given orientation .
put @xmath94 for the oriented line segment from @xmath95 to @xmath96 .
since the affine functions giving @xmath0 on the neighboring triangles agree along @xmath92 , it follows that @xmath94 is perpendicular to @xmath92 .
this implies that the current @xmath97\ ! ] } } \times { { [ \![}s_j{]\!]}}\ ] ] is lagrangian , with @xmath98\!]}}\times { { [ \![}s_j{]\!]}}\ ] ] given a vertex @xmath99 , let @xmath100 be the triangles of @xmath62 having @xmath99 as a vertex , listed in counterclockwise order , with corresponding gradients @xmath101 .
let @xmath102 be the edge between @xmath103 and @xmath89 , @xmath104 , oriented to point away from @xmath99 , and @xmath105 the oriented line segment @xmath106 .
then the part of @xmath107 lying above @xmath99 is @xmath108\!]}}\times\sum_{i=1}^n { { [ \![}s_i{]\!]}}$ ] .
put @xmath109 for the unique bounded polygonal region - with - multiplicities with @xmath110\ ! ] } } = - \sum { { [ \![}s_i{]\!]}}$ ] .
finally , put @xmath111\ ! ] } } \times { { [ \![}p_k{]\!]}}$ ] .
then @xmath112 satisfies the hypothesis of theorem [ thm : ma ] for @xmath113 , so @xmath114 .
we remark that if @xmath115 is any closed oriented rectifiable curve in @xmath72 then the unique compactly supported current @xmath62 of dimension 2 with @xmath116 may be constructed as follows .
choose an arbitrary point @xmath117 , and define the map @xmath118 by @xmath119 then @xmath120\ ! ] } } \times \gamma)$ ] , i.e. the _ join _ of @xmath115 with @xmath121 .
[ def : ft ] let @xmath62 be a finite triangulation of a polygonal planar region @xmath58 .
given a function @xmath11 defined over @xmath58 , put @xmath122 for the piecewise linear function on @xmath58 that agrees with @xmath11 at the vertices of @xmath62 and is affine on each triangle of @xmath62 .
[ def : complexes ] denote by @xmath123 the triangulation of @xmath72 with vertices given by the points of the integer lattice @xmath124 and edges given by the sides of the squares of the lattice , together with their diagonals of slope @xmath125 . by a * square mesh of size @xmath126 * we mean a finite triangulation @xmath62 that is congruent to a subcomplex of @xmath127 .
the * axes * of @xmath62 are the images of the @xmath128- and @xmath129-axes of @xmath123 under the congruence map .
the * support * of @xmath62 is the union of the simplices comprising @xmath62 .
the * interior * of @xmath62 is the interior of the support of @xmath62 .
the * core * of @xmath62 is the interior of the union of the closed squares of side @xmath126 centered at the vertices of @xmath62 that lie in @xmath130 .
the * crust * of @xmath62 is the union of the closed squares that intersect the boundary of @xmath131 .
( 1,.76 ) ( 0,0 ) the proof of theorem [ thm : main ] is based on the analysis of the following situation .
given @xmath132 we put @xmath133 the differential current @xmath114 may be constructed as described in section [ sect : d of pl ] , where each @xmath134 lives over the @xmath135-skeleton of @xmath123 . in particular , @xmath136 is given by integration over @xmath137 , where @xmath138 is a polygonal region with multiplicities .
it is clear that the @xmath139 are translates of @xmath140 .
( 1,0.7693005 ) ( 0,0),title="fig : " ] ( 0.50663618,0.52356205)@xmath141 ( 0.32686205,0.62162065)@xmath142 ( 0.17977414,0.52356205)@xmath143 ( 0.32686205,0.27841546)@xmath144 ( 0.52297929,0.16401378)@xmath145 ( 0.65372413,0.27841546)@xmath146 ( 0.45760688,0.40916032)@xmath147 ( 0.45760688,0.73602238)@xmath148 ( 0.13074482,0.73602238)@xmath149 ( 0.13074482,0.40916032)@xmath150 ( 0.45760688,0.08229826)@xmath151 ( 0.78446894,0.40916032)@xmath152 ( 0.78446894,0.08229826)@xmath153 we wish to compute the mass of @xmath63 in terms of @xmath154 .
we label the six triangles surrounding the origin by @xmath155 as in figure [ fig : grid ] . denoting by @xmath87 the gradient of @xmath156 ,
we compute @xmath157 the boundary of the region @xmath63 is then the closed hexagonal path with the @xmath87 as successive vertices . in particular , if @xmath158 then @xmath63 degenerates to a rectangle with signed area equal to @xmath159 .
furthermore , in this case the support of @xmath160 lies above the horizontal and vertical segments of the mesh .
the segments in the fiber of @xmath160 over the vertical segments have length @xmath161 , and over the horizontal segments have length @xmath162 . from these observations
we easily draw the following conclusion .
[ lem : aligned ] suppose @xmath163 is a quadratic function and @xmath62 is a square mesh of size @xmath126 , such that the axes of @xmath62 are parallel to the eigenvectors of @xmath164 .
then @xmath165 the addition of the linear and constant terms has no effect on the calculations above , since they yield merely a translation of the gradient cycle by the vector @xmath166 .
taking the axes parallel to the eigenvectors is equivalent to taking @xmath158 .
scaling by @xmath126 , the relation clearly holds if @xmath167 is replaced by any of its constituent squares . on the other hand ,
if the axes of @xmath62 are not aligned with the eigenvectors of @xmath54 , then the masses of the @xmath168 may be much larger .
for example , in the case of the degenerate quadratic form @xmath169 we find that @xmath170 is as shown in figure [ fig : hex ] .
the mass of @xmath63 is then equal to @xmath171 , while @xmath172 , illustrating the fact that the bound of theorem [ thm : main ] can not be achieved by a haphazard triangulation ( as in proposition [ prop : crude ] below ) . ( 1,1 ) ( 0,0 ) ( 0.7,0.6)@xmath173 although we will not need this fact , it is interesting to note that the _ algebraic _ area of @xmath63 is always precisely equal to @xmath174 . for a given triangle @xmath175 , vertex @xmath176 and edge @xmath177 of @xmath62
, we put @xmath178 and @xmath179 where @xmath180 range over all triangles of @xmath62 incident to @xmath176 , i.e. @xmath181 is the diameter of the polygon @xmath63 lying above @xmath176 .
[ pert ] let @xmath62 be a square mesh of size @xmath126 and @xmath182 functions defined on the support of @xmath62
. then 1 .
[ 1st conclusion ] for any triangle @xmath175 of @xmath62 , @xmath183 2 . for any interior edge @xmath177 of @xmath62 , @xmath184 3 .
for any interior vertex @xmath176 of @xmath62 , @xmath185 we may assume that the axes of @xmath62 are the @xmath128- and @xmath129-axes .
( 1 ) : let @xmath175 be a triangle of @xmath62 with vertices @xmath186
. then @xmath187 from which the conclusion follows at once .
( 2 ) : by conclusion ( 1 ) , the fibers of @xmath188 are line segments whose endpoints differ by at most @xmath189 .
thus the lengths of these segments differ by at most twice that , and the conclusion follows since @xmath190 .
( 3 ) : given any point @xmath191 , the current @xmath192 may be expressed as the join of @xmath193 with the bounding polygon .
this may in turn may be expressed as the sum of the joins of @xmath193 with each of the six edges .
we take @xmath194 for some arbitrarily chosen triangle @xmath175 incident to @xmath176 . for any pair of adjacent triangles
@xmath195 incident to @xmath176 , put @xmath196 we claim that @xmath197 this follows from the relations @xmath198 which follow in turn from conclusion and lemma [ lem : triangles ] below .
[ lem : triangles]let @xmath199 . put @xmath200 for the oriented triangles @xmath201 respectively .
then @xmath202 the mass of @xmath203 is the area of the symmetric difference of the two triangles .
this symmetric difference is included in the union of the triangles @xmath204 , with total area @xmath205 .
[ lem : taylor ] let @xmath62 be a square mesh of size @xmath206 such that @xmath207 .
let @xmath208 .
suppose that the axes of @xmath62 are parallel to the eigenvectors of @xmath209 , and that @xmath210 for @xmath211
. then @xmath212\end{aligned}\ ] ] put @xmath54 for the 2nd order taylor polynomial of @xmath11 at @xmath121 , so that @xmath213 .
then @xmath214 there are @xmath215 constituent squares in @xmath167 , each of which contains one vertex and six half - edges of @xmath62 . referring to ,
it is clear that @xmath216 for every vertex @xmath176 lying in @xmath167 .
thus by lemmas [ lem : aligned ] and [ pert ] , @xmath217\\ \le & \operatorname{area}({\operatorname{core}t})\times\\ \notag & \left ( 1 + 2\sqrt 2 { \parallel h_q\parallel } + |\det h_q| + \right.\\ \notag & \left.12\sqrt 2 { \epsilon}^{-2 } \omega r_0 ^ 2 + 24 { \parallel h_q\parallel } { \epsilon}^{-2}\omega r_0 ^ 2 + 24 { \epsilon}^{-4 } \omega^2 r_0 ^ 4\right)\end{aligned}\ ] ] the result now follows from the elementary estimate @xmath218 for @xmath219 .
put @xmath220 for the smallest angle of any triangle of @xmath62 .
thus @xmath220 is a measure of the fatness " of @xmath62 .
we put also @xmath221 for the length of the longest edge of @xmath62 , and @xmath222 for the length of the shortest edge .
[ prop : crude ] given @xmath223 and @xmath224 there exists a constant @xmath225 with the following property .
let @xmath62 be a finite triangulation of an open polygon @xmath226 , with @xmath227 then for any @xmath228 @xmath229 we decompose @xmath230 as above .
clearly @xmath231 by the mean value theorem , the gradient @xmath232 of @xmath122 on a given triangle @xmath233 lies in the convex hull of the image of @xmath175 under @xmath7 .
it follows that if the triangles @xmath234 share an edge or a vertex then @xmath235 therefore by the construction section of [ sect : d of pl ] @xmath236 where the sum is taken over all edges @xmath92 of @xmath62 . since clearly @xmath237 we conclude that @xmath238 clearly the number of triangles ( or edges ) incident to any given vertex @xmath176 of @xmath62 is no greater than @xmath239 . by the construction of section [
sect : d of pl ] , the fiber of @xmath136 over @xmath176 is a polygon @xmath63 with multiplicities and with at most @xmath240 edges .
it follows that these multiplicities are at most @xmath241 , so that by @xmath242 where @xmath35 is the number of vertices in @xmath62 . putting @xmath243 for the number of edges of @xmath62 , clearly @xmath244 .
therefore @xmath245 and we obtain @xmath246 the conclusion now follows if we take @xmath247 .
givem @xmath248 we put @xmath249 for the open polygon consisting of the interior of the union of all closed squares @xmath250\times [ \frac b n,\frac { b+1 } n ] , \ a , b \in { \mathbb{z}}$ ] , that intersect @xmath35 .
it is clear that @xmath251 for @xmath21 sufficiently large , and that @xmath252 * has size @xmath256 * has axes parallel to the eigenvectors of @xmath257 , where @xmath128 is the center of @xmath9 , and * is maximal with respect to the condition that no vertex of @xmath258 lies within distance @xmath259 of the complement of @xmath9 let @xmath260 be the system of vertices and edges in @xmath249 consisting of all vertices and edges of the @xmath255 , together with the edges and vertices obtained by subdividing each edge of every square @xmath9 into @xmath261 equal segments of length @xmath259 .
this system satisfies the conditions of theorem [ thm : chew ] , and we take @xmath262 to be the triangulation of @xmath249 in the conclusion of that statement . referring to definition [ def : complexes ] , put @xmath263 .
put @xmath264 for the interior of the union of @xmath265 with all of the additional triangles in @xmath262 that were obtained by means of theorem [ thm : chew ] . thus @xmath266 .
furthermore , each @xmath267 includes a square of side @xmath268 .
it follows that @xmath269 now put @xmath270 .
it is clear that @xmath271 locally uniformly as @xmath272 .
we will show that there are @xmath273 such that the sequence pl functions @xmath274 satisfies the relation in the statement of theorem [ thm : main ] .
put @xmath284 .
let @xmath285 be a modulus of continuity for @xmath286 , i.e. a monotonically increasing function with @xmath287 , such that @xmath288 applying lemma [ lem : taylor ] with @xmath289 and @xmath290 , we obtain @xmath291\end{aligned}\ ] ] the dominated convergence theorem implies that the integral converges to the corresponding integral over @xmath35 as @xmath292 , while the second term converges to zero .
this concludes the proof .
50 brehm , u. & khnel , w. smooth approximation of polyhedral surfaces regarding curvatures .
* 12 * ( 1982 ) , 435461 cheeger , j. , mller , w. , & schrader , r. on the curvature of piecewise flat spaces .
* 92 * ( 1984 ) , 405454 chew , l.p . guaranteed quality triangular meshes .
preprint , 1989 .
available at http://www.cs.sandia.gov/ samitch / unm_math_579/course_papers / chew_guaranteed_triangular_89 - 983.pdf federer , h. _ geometric measure theory_. springer , berlin and heidelberg , 1969 .
fu , j.h.g .
monge - ampre functions i. _ indiana u. math .
j. _ * 38 * ( 1989 ) , 745771 fu , j.h.g . convergence of curvatures in secant approximations .
_ j. differential geom . _
* 37 * ( 1993 ) , 177190 .
fu , j.h.g .
an extension of alexandrov s theorem on second derivatives of convex functions .
_ * 228 * ( 2011 ) , 22582267 jerrard , r. l. some remarks on monge - ampre functions . _
singularities in pde and the calculus of variations _ , 89112 .
lecture notes , * 44*. amer .
soc . , providence , ri , 2008 .
jerrard , r.l . some rigidity results related to monge - ampre functions . _ canad . j. math . _
* 62 * ( 2010 ) , 320354 morvan , j .-
_ generalized curvatures_. springer , berlin and heidelberg , 2008 | given a piecewise linear ( pl ) function @xmath0 defined on an open subset of @xmath1 , one may construct by elementary means a unique polyhedron with multiplicities @xmath2 in the cotangent bundle @xmath3 representing the graph of the differential of @xmath0 . restricting to dimension 2 ,
we show that any smooth function @xmath4 may be approximated by a sequence @xmath5 of pl functions such that the areas of the @xmath6 are locally dominated by the area of the graph of @xmath7 times a universal constant . |
although most of the problems in the standard cosmological model can be resolved by the inflation theory , the resolution of the existence of a big bang singularity in the early universe is still elusive . based upon the string / m - theory , the pre - big bang @xcite and ekpyrotic / cyclic @xcite scenarios
have been proposed to address the issue .
recently , ellis et al proposed , in the context of general relativity , a new scenario , called an emergent universe @xcite to avoid this singularity . in this scenario ,
the space curvature is positive , which is supported by the recent observation from wmap7 @xcite where it was found that a closed universe is favored at the @xmath5 confidence level , and the universe stays , past - eternally , in an einstein static state and then evolves to a subsequent inflationary phase .
so , in an emergent theory , the universe originates from an einstein static state rather than from a big bang singularity . however , the einstein static universe in the classical general relativity is unstable , which means that it is extremely difficult for the universe to remain in such an initial static state in a long time due to the existence of perturbations , such as the quantum fluctuations .
therefore , the original emergent model does not seem to resolve the big bang singularity problem successfully as expected .
however , in the early epoch , the universe is presumably under extreme physical conditions , the realization of the initial state may be affected by novel physical effects , such as those stemming from quantization of gravity , or a modification of general relativity or even other new physics . as a result , the stability of the einstein static state has been examined in various cases @xcite , from loop quantum gravity @xcite to modified gravity ( for a review see ref .
@xcite ) , from horava - lifshitz gravity @xcite to shtanov - sahni braneworld scenario @xcite in this paper , we plan to examine the stability of the einstein static universe in the dgp brane - world model @xcite . in this braneworld
, the whole energy - momentum is confined on a three dimensional brane embedded in a five - dimensional , infinite - volume minkowski bulk .
since there are two different ways to embed the 4-dimensional brane into the 5-dimensional spacetime , the dgp model has two separate branches denoted by @xmath6 with distinct features .
the @xmath7 branch can explain the present accelerating cosmic expansion without the introduction of dark energy @xcite , while for the @xmath3 branch , dark energy is needed in order to yield an accelerating expansion , as is the case in the ldgp model @xcite and the qdgp model @xcite .
using the @xmath8 , cmb shift and sne ia observational data , lazkoz and majerotto @xcite found that the ldgp and qdgp are slightly more favored than the self - accelerating dgp model .
let us also note that a crossing of a phantom divide line , which is favored by the recent various observational data @xcite , is possible with a single scalar field @xcite in the @xmath3 branch .
in addition , inflation in the dgp model displays some new characteristics .
it should be noted , however , that only in the @xmath3 case can inflation exit spontaneously @xcite .
also , in contrast to the randall - sundrum @xcite and shtanov - sahni @xcite braneworld scenarios with high energy modifications to general relativity , the dgp brane produces a low energy modification ( for a review of the phenomenology of the dgp model , see ref .
for a homogeneous and isotropic universe which is described by the friedmann - robertson - walker ( frw ) metric .
@xmath9 the friedmann equation on the warped dgp brane can be written as @xcite @xmath10\;,\end{aligned}\ ] ] where @xmath11 is the hubble parameter , @xmath12 is the constant curvature of the three - space of the frw metric , @xmath13 is the total energy density and @xmath14 is a parameter denoting the strength of the induced gravity on the brane . for @xmath3 , the brane tension can be assumed to be positive , while for @xmath7 , it is negative .
@xmath15 is given by @xmath16^{1/2}\;,\end{aligned}\ ] ] where @xmath17 with @xmath18 defined as @xmath19 here @xmath20 is the 5-dimensional newton constant , @xmath21 the 5-dimensional cosmological constant in the bulk , @xmath22 the brane tension , and @xmath23 a constant related to weyl radiation .
for simplicity , we will neglect the dark radiation term and restrict ourselves to the randall- sundrum critical case , i.e. @xmath24 , then eq.(2 ) simplifies to @xmath25 since in the very early era of the universe , the total energy density should be very high .
thus , we will , in the following , only consider the ultra high energy limit , @xmath13@xmath26@xmath27 .
in addition , we are interested in a closed universe , so we set the constant curvature @xmath12 to be @xmath28 .
as a result , the friedmann equation reduces to @xmath29 this describes a 4-dimensional gravity with minor corrections , which implies that @xmath14 must have an energy scale as the planck scale in the dgp model .
the energy density @xmath13 of a perfect fluid in the universe satisfies the conservation equation @xmath30 where @xmath31 is the equation of state of the perfect fluid .
a constant @xmath1 is considered in the present paper , which is a good approximation if the perfect fluid is a scalar field and the variation of the potential of scalar field is very slow with time .
differentiating eq .
( [ feq ] ) with respect to cosmic time , one gets @xmath32 combining this equation with the friedmann equation given in eq .
( [ feq ] ) , we have @xmath33
the einstein static solution is given by the conditions @xmath34 and @xmath35 , which imply @xmath36 at the critical point determined by above conditions , we find , using eq .
( [ deq ] ) @xmath37 which means that in this dynamical system , there is only one einstein static state solution . in order to guarantee the physical meaning of @xmath38 ,
it is necessary that @xmath39 substituting eq .
( [ rhoes ] ) into the friedmann equation , we obtain at the critical point @xmath40 with the requirement @xmath41 . before analyzing the stability of the critical point ,
we want to express eq .
( [ deq ] ) in terms of @xmath42 and @xmath11 . to do so
, we put the friedmann equation in a different way @xmath43 thus eq . ( [ deq ] )
can be re - written as @xmath44 now we study the stability of the critical point . for convenience , we introduce two variables @xmath45 it is then easy to obtain the following equations @xmath46 @xmath47 in these variable , the einstein static solution corresponds to the fixed point , @xmath48 .
the stability of the critical point is determined by the eigenvalue of the coefficient matrix resulting from linearizing the system described by above two equations near the critical point . using @xmath49 to denote the eigenvalue
, we have @xmath50 if @xmath51 , the corresponding equilibrium point is a center point otherwise it is a saddle one . in order to analyze the stability of the critical point in detail ,
we now divide our discussions into two cases , i.e. , @xmath3 and @xmath2 .
* a*. @xmath2 in this case , @xmath52 is required to ensure that the critical point is physically meaningful .
it then follows that @xmath53 , which means that this critical point is a saddle point .
thus , there is no stable einstein static solution , and an emergent universe is not realistic in this case .
* b*. @xmath3 now the requirement for the critical point to be physically meaningful is @xmath54 .
this exactly agrees with the condition of stability ( @xmath51 ) .
hence , as long as the critical point exists , it is always stable .
so , if the scale factor satisfies the condition given in eq .
( [ aes ] ) initially and @xmath1 is within the region of stability , the universe can stay at this stable state past - eternally , and may undergo a series of infinite , non - singular oscillations , as shown in fig .
( [ fig1 ] ) . as a result
, the big bang singularity can be avoided successfully .
the evolutionary curve of the scale factor with time ( left ) and the phase diagram in space ( @xmath42 , @xmath55 ) ( right ) for the case @xmath3 in planck unit and with @xmath56 .
, title="fig:",width=264 ] the evolutionary curve of the scale factor with time ( left ) and the phase diagram in space ( @xmath42 , @xmath55 ) ( right ) for the case @xmath3 in planck unit and with @xmath56 .
, title="fig:",width=264 ]
now , we have shown that an stable einstein static state exists in the @xmath3 branch . however , in order to have a successful cosmological scenario , a graceful exit to an inflationary epoch is needed .
this is possible in the following sense . in the analysis carried out in the present paper , the equation of state @xmath1 of the perfect fluid in the universe
is assumed to be a constant , and this is a good approximation if the energy component in the early universe is only that of a minimally coupled scalar field with a self - interaction potential .
one can show that the kinetic energy and potential energy of this scalar field should be both non - zero constants for an einstein static solution @xcite .
that is to say , the scalar field rolls along a plateau potential .
however a realistic inflationary model clearly requires the potential to vary as the scalar field evolves .
thus , the constant potential is merely a past - asymptotic limit of a smoothly varying one , as pointed out in refs .
so , the essentially slowly varying potential will eventually break the equilibrium of the einstein static state and lead to an exit from the initial einstein phase to an inflationary one .
some specific forms of such a potential that implements these features have been constructed in refs @xcite .
in this paper , we have studied the existence and stability of the einstein static universe in the dgp braneworld scenario . by assuming the existence of a perfect fluid with a constant equation of state , which is a good approximation
if the perfect fluid is a scalar field and the variation of the potential of scalar field is very slow with time , we have shown that for the branch with @xmath2 , there is no stable einstein static universe , whereas , for the branch with @xmath3 , the einstein static universe exists and it is stable if the equation of state @xmath1 satisfies @xmath54 .
thus , the universe can stay at this stable state past - eternally , and may undergo a series of infinite , non - singular oscillations .
hence , in the @xmath3 branch of the dgp model , the universe can originate from an einstein static state and then enter an inflation era .
furthermore , the universe can exit , spontaneously , this inflation phase to a radiation dominated era , as shown in previous studies @xcite . as a result ,
the big bang singularity problem in the standard cosmological scenario can be resolved successfully .
this work was supported in part by the national natural science foundation of china under grants nos .
10775050 , 10705055 and 10935013 , the srfdp under grant no . 20070542002 , the fanedd under grant no .
200922 , the national basic research program of china under grant no .
2010cb832803 , the ncet under grant no.09 - 0144 , the pcsirt under grant no .
irt0964 , and the programme for the key discipline in hunan province .
m. gasperini and g. veneziano , phys . rep .
373 , 1 ( 2003 ) [ arxiv : hep - th/0207130 ] ; j. e. lidsey , d. wands and e. j. copeland , phys .
337 , 343 ( 2000 ) [ arxiv : hep - th/9909061 ] .
j. khoury , b. a. ovrut , p. j. steinhardt and n. turok , phys . rev .
d 64 , 123522 ( 2001 ) [ arxiv : hep - th/0103239 ] ; p. j. steinhardt and n. turok , science 296 , 1436 ( 2002 ) ; p. j. steinhardt and n. turok , phys .
d 65 , 126003 ( 2002 ) [ arxiv : hep - th/0111098 ] ; j. khoury , p. j. steinhardt and n. turok , phys .
92 , 031302 ( 2004 ) [ arxiv : hep - th/0307132 ] .
g. f. r. ellis and r. maartens , class .
* 21 * , 223 ( 2004 ) .
g. f. r. ellis , j. murugan and c. g. tsagas , class .
* 21 * , 233 ( 2004 ) .
e. komatsu , et al .
arxiv:1001.4538 .
d. j. mulryne , r. tavakol , j. e. lidsey and g. f. r. ellis , phys .
d * 71 * , 123512 ( 2005 ) .
l. parisi , m. bruni , r. maartens and k. vandersloot , class .
* 24 * , 6243 ( 2007 ) .
p. wu and h. yu , j. cosmol .
p. * 05 * , 007 ( 2009 ) arxiv:0905.3116 .
j. e. lidsey and d. j. mulryne , phys .
d * 73 * , 083508 ( 2006 ) . c. g. boehmer , l. hollenstein and f. s. n. lobo , phys .
d * 76 * , 084005 ( 2007 ) ; n. goheer , r. goswami and p. k. s. dunsby , class .
* 26 * , 105003 ( 2009 ) arxiv : 0809.5247 ; s. del campo , r. herrera and p. labrana , j. cosmol .
p. * 0711 * , 030 ( 2007 ) ; r. goswami , n. goheer and p. k. s. dunsby , phys .
d * 78 * , 044011 ( 2008 ) ; u. debnath , class .
quant . grav . * 25 * , 205019 ( 2008 ) ; b. c. paul and s. ghose , arxiv : 0809.4131 .
s. s. seahra and c. g. bohmer , phys .
d * 79 * , 064009 ( 2009 ) . c. g. boehmer and f. s. n. lobo , phys .
d * 79 * , 067504 ( 2009 ) arxiv : 0902.2982 j. d barrow , g. ellis , r. maartens , c. tsagas , class .
. grav . * 20 * , l155 ( 2003 ) .
t. clifton , john d. barrow , phys .
d * 72 * , 123003 ( 2005 ) .
j. d barrow , c. g tsagas , arxiv:0904.1340 .
c. g. boehmer , l. hollenstein , f. s. n. lobo , and s. s. seahra , arxiv:1001.1266 .
a. odrzywolek , phys .
d * 80 * , 103515 ( 2009 ) . c. g. boehmer , f. s. n. lobo , arxiv:0909.3986 .
p. wu and h. yu , phys .
d * 81 * ( 2010 ) 103522 , arxiv : 0909.2821 .
d. dvali , g. gabadadze and m. porrati , phys .
b * 485 * , 208 ( 2000 ) ; c. charmousis , r. gregory , n. kaloper , a. padilla , jhep * 0610 * , 066 ( 2006 ) ; r. gregory , n. kaloper , r. c. myers , a. padilla , jhep * 0710 * , 069 ( 2007 ) ; d. gorbunov , k. koyama , s. sibiryakov , phys . rev .
d * 73 * , 044016 ( 2006 ) .
c. deffayet , phys .
b * 502 * , 199 ( 2001 ) .
l. randall and r. sundrum , phys .
83 * , 4690 ( 1999 ) .
y. shtanov and v. sahni , phys .
b * 557 * , 1 ( 2003 ) .
a. lue and g. d. starkman , phys .
d * 70 * , 101501(r ) ( 2004 ) .
l. p. chimento , r. lazkoz , r. maartens and i. quiros , j. cosmol . astropart.phys .
* 0609 * , 004 ( 2006 ) .
r. lazkoz and e. majerotto , j. cosmol .
p. * 0707 * , 015 ( 2007 ) .
u. alam , v. sahni , t. d. saini , a. a. starobinsky , mon . not .
soc . * 354 * , 275 ( 2004 ) ; u. alam , v. sahni , a. a. starobinsky , j. cosmol .
p. * 0406 * , 008 ( 2004 ) ; y. wang and p. mukherjee , astrophys .
j. * 606 * , 654 ( 2004 ) ; r. lazkoz , s. nesseris and l. perivolaropoulos , j. cosmol .
p. * 0511 * , 010 ( 2005 ) ; s. nesseris and l. perivolaropoulos , j. cosmol .
p. * 0701 * , 018 ( 2007 ) ; p. wu and h. yu , phys .
b * 643 * , 315 ( 2006 ) .. h. zhang , z .- h .
zhu , phys .
d * 75 * , 023510 ( 2007 ) .
m. bouhmadi - lopez , r. maartens and d. wands , phys .
d * 70 * , 123519 ( 2004 ) .
r. cai and h. zhang , j. cosmol .
p. * 0408 * , 017 ( 2004 ) .
e. papantonopoulos and v. zamarias , j. cosmol .
p. * 0410 * , 001 ( 2004 ) .
h. zhang and r. cai , j. cosmol .
p. * 0408 * , 017 ( 2004 ) .
h. zhang and z. zhu , phys .
b * 641 * , 405 ( 2006 ) .
s. del campo , r. herrera , phys .
b * 653 * , 122 ( 2007 ) .
a. lue , phys . rept . * 423 * , 1 ( 2006 ) .
k. maeda , s. mizuno and t. torii , phys .
d * 68 * , 024033 ( 2003 ) arxiv : gr - qc/0303039 . | the stability of an einstein static universe in the dgp braneworld scenario is studied in this letter .
two separate branches denoted by @xmath0 of the dgp model are analyzed .
assuming the existence of a perfect fluid with a constant equation of state , @xmath1 , in the universe , we find that , for the branch with @xmath2 , there is no a stable einstein static solution , while , for the case with @xmath3 , the einstein static universe exists and it is stable when @xmath4 .
thus , the universe can stay at this stable state past - eternally and may undergo a series of infinite , non - singular oscillations .
therefore , the big bang singularity problem in the standard cosmological model can be resolved . |
the colours calculated for stars later than mid a - type are affected by treatment of convection .
small systematic errors were found in the colours calculated using atlas6 ( kurucz , 1979 ) , which could be due to convection or missing opacity in the models ( relyea & kurucz , 1978 ) .
recent improvements in opacity in atlas9 models ( kurucz , 1991 ) ought to ensure that opacity is now a less dominant source of the discrepancies .
this leaves convection as a possible source of the discrepancies .
we have compared mixing length theory ( see castelli et al . , 1997 ) with the turbulent convection model of canuto & mazzitelli ( 1991 , 1992 ) .
the atlas9 code was used to calculate @xmath0 colours .
the computations were identical , except for treatment of convection .
we considered three cases : 1 . mixing - length theory with approximate overshooting ( mlt_ov ) , 2 . mixing - length theory without approximate overshooting ( mlt_noov ) , 3 .
the kupka ( 1996 ) implementation of the canuto & mazzitelli theory ( cm ) .
here we present a summary of the main findings from the comparisons between the three treatments of convection outlined above .
full details can be found in smalley & kupka ( 1997 ) .
the ultimate test of any model colours is to compare them to the colours of stars whose atmospheric parameters have been obtained from direct , model - independent methods .
the fundamental stars presented by smalley & dworetsky ( 1995 ) were used to test the colours obtained from the three grids .
we adopted the following methodology in the testing procedure : the observed @xmath0 photometry of the fundamental star was de - reddened using moon s ( 1985 ) uvbybeta programme .
for each grid in turn , values of @xmath1 and @xmath2 were obtained by interpolation within the ( @xmath3 , @xmath4 ) colour grid .
we then compared either @xmath1 or @xmath2 with the corresponding fundamental value .
this way we could compare @xmath1 and @xmath2 independently , since very few stars have fundamental values for both parameters . figure 1 shows the comparison between the fundamental @xmath1 values and those obtained from the 3 grids .
the cm model gives the best agreement .
the mlt_noov grid is somewhat discrepant , but still within the error bars .
the mlt_ov grid is widely discrepant .
figure 2 shows the comparison between the fundamental @xmath2 values and those obtained from the 3 grids .
there is little difference between cm and mlt_noov grids , but mlt_ov grid is clearly discrepant .
however , the cm models give less scatter and no hint of a possible systematic trend at low @xmath1 observed in the mlt_noov models . from the comparisons with fundamental stars we find that the cm grids give the best agreement .
the mlt_noov grids are only marginally less successful in recovering the fundamental values .
the mlt_ov grids are however significantly discrepant .
since there are very few stars with fundamental values of @xmath1 or @xmath2 , the comparisons were also performed using non - fundamental stars .
whilst this allows for significantly more stars in the comparisons , there is a very real danger that model - dependent systematic errors could bias any result .
however , similar results were found using stars which have values of @xmath1 obtained using the infrared flux method ( blackwell & lynas - gray 1994 ) , once known spectroscopic binaries had been excluded .
open cluster stars can be used as surface gravity standards by fitting evolutionary models to the cluster photometry . in this case
the stellar interior calculations can be influenced by the treatment of convection ( e.g. stothers & chin 1995 , canuto et al . 1996 ) . nevertheless , similar results were found for hyades stars with @xmath2 obtained from evolutionary models .
overall , the cm model has the greatest success in recovering the @xmath1 and @xmath2 obtained from both fundamental and non - fundamental methods .
however , the mlt_noov grids are only marginally less successful than the cm grids .
the mlt_ov grids are clearly discrepant .
the colours of late - a and f stars can be significantly affected by metallicity , due to the vast amount of metal lines .
the @xmath5 metallicity index can be used to estimate [ m / h ] in late - a and f stars , usually via empirical relations ( smalley 1993 ) . in all the three cases considered here ,
the theoretical @xmath5 values are found to be discrepant .
none of the model results based on the different convection treatments agrees completely with the standard main - sequence reference line , and all are discrepant for late - type stars .
the problem may not be due to convection alone , but could be related to microturbulence ( smalley & kupka 1997 ) , because microturbulence is probably closely related to the small - scale part of the photospheric convective flow ( holweger & strenburg 1993 ) .
investigations into the causes of this discrepancy are being undertaken .
from a comparison with the observed @xmath0 colours , we have found that the cm grid gives results that are generally superior to those with mlt theory without overshooting ( mlt_noov ) .
models with overshooting ( mlt_ov ) are found to be clearly discrepant . | we present a discussion on the effects of convection on the @xmath0 colours of a and f stars . the mixing - length theory used in atlas9
is compared to the turbulent convection theory of canuto & mazzitelli ( 1991 , 1992 ) .
comparison with fundamental stars reveals that colours calculated using the canuto & mazzitelli convection theory are generally in better agreement than those obtained using mixing - length theory . |
the main motivation of this article is the presence of spiral waves in excitable media , especially in cardiac tissue .
spiral waves arise as stable spatio - temporal patterns in various chemical , physical systems and biological systems , as well as numerical simulations of reaction - diffusion systems on excitable media with various geometries .
excitable media are extended non - equilibrium systems having a uniform rest state that is linearly stable but susceptible to finite perturbations .
spiral waves have been observed experimentally , for instance , in catalysis of platinum surfaces @xcite , belousov - zhabotinsky chemical reactions @xcite , rayleigh - bernard convection @xcite , slime - mould cells @xcite and the most important , cardiac tissue @xcite .
numerical simulations of the spiral waves have been done by @xcite , for example .
+ it is now believed that spiral and scroll waves that appear in the heart muscle can lead to cardiac arrhythmias ( abnormal rhythms in the heart ) , giving rise to atrial fluttering or ventricular fibrillation . in normal hearts cardiac arrhythmias are rare , but in diseased hearts cardiac arrhythmias can become more common . for example , if chambers of the heart become abnormally large , they are susceptible to serious arrhythmias in which waves are believed to circulate in a fashion that is similar to the circulation of the belousov - zhabotinsky waves in a chemical medium .
real human hearts are enormously complex three - dimensional structures . in this article
, we assume that the geometry of the excitable media is a sphere , which in the case of cardiac tissue is clearly an approximation .
+ in the planar case , a rigidly rotating spiral wave is an example of wave pattern rotating around a center and being well approximated by an archimedean spiral wave far from rotation center .
near the rotation center , there is a core region of the spiral wave , where the front of the wave has a tip , whose structure is considered to be the most important in understanding the behavior of the whole spiral wave @xcite .
barkley @xcite was the first who performed a numerical linear stability analysis for the basic - time periodic spiral wave solution in a reaction - diffusion system on the unbounded plane and showed evidence of a hopf bifurcation .
in particular , a simple pair of eigenvalues was shown to cross the imaginary axis while three neutral eigenvalues lie on the imaginary axis and the remainder of the spectrum is bounded into the left - half plane . later , using an ad hoc model , barkley @xcite was the first to realize the key importance of the group @xmath2 of all planar translations and rotations in describing the dynamics and bifurcations of planar spiral waves .
+ it is well known now that the tip of the spiral wave rotates steadily or meanders or linearly drift in plane @xcite . from a mathematical point of view , rigidly rotating spiral waves are examples of rotating waves , meandering spiral waves are examples of modulated rotating waves and linearly drifting spiral waves are examples of modulated rotating waves .
+ the first rigorous mathematical theory of the planar spiral waves was done by wulff @xcite . in her thesis ,
wulff studied the external periodic forcing of rotating waves which leads to modulated rotating waves or modulated travelling waves , and also proved an @xmath2-equivariant hopf theorem for the bifurcation from rotating waves to modulated ( rotating or travelling ) waves in autonomous systems .
the external periodic forcing of rotating waves was studied using a contraction mapping theorem on scales of banach spaces , and the proof for the @xmath2-equivariant hopf theorem for rotating waves was based on liapunov - schmidt reduction on scales of banach spaces . in both cases
, it is shown that modulated travelling waves emanate if the rotation frequency is a multiple of the external frequency , respectively the modulus of the eigenvalue leading to hopf bifurcation .
the main difficulty comes from the fact that the group @xmath2 is noncompact and the action of @xmath2 on the usual spaces of functions is not smooth .
the proofs in @xcite are based on the basic assumption that the linearization at the rotating wave in the co - rotating frame does not exhibit continuous spectrum near the imaginary axis .
+ later , sandstede , scheel and wulff @xcite proved a finite - dimensional center bundle reduction theorem near a relative equilibrium @xmath3 of an infinite - dimensional vector field on a banach space @xmath4 on which acts a finite - dimensional lie group ( not necessarily compact ) .
this generalizes krupa s results on bifurcation from relative equilibria @xcite , from compact groups to noncompact groups and from finite dimensions to infinite dimensions .
using the results of @xcite , the hopf bifurcation from one - armed rotating spiral wave to meandering waves can be studied @xcite .
+ in the case of a one - armed rotating spiral wave , the relative equilibrium @xmath5 is diffeomorphic to @xmath2 , which is diffeomorphic to @xmath6 . near hopf bifurcation ,
the reduced differential equations on the center bundle @xmath7 are given by @xmath8 where @xmath9 , @xmath10 , @xmath11 .
the rotating wave @xmath12 corresponds to @xmath13 at @xmath14 and to the solution @xmath15 for the first two differential equations in ( [ e : intro0 ] ) .
+ in case of a supercritical hopf bifurcation we have @xmath16 , @xmath17 .
it follows that for any small @xmath18 there is a meandering spiral wave which becomes a drifting linear wave if @xmath19 , for some @xmath20 , where @xmath21 is the frequency that appears due to the hopf bifurcation and @xmath22 is the @xmath23 periodic solution of the third differential equation in ( [ e : intro0 ] ) that appears due to the supercritical hopf bifurcation for @xmath13 at @xmath14 .
similar results for m - armed spiral waves were also obtained . + in @xcite , a @xmath24-equivariant semilinear system of parabolic equations ( where @xmath24 is a finite - dimensional possibly non - compact lie group ) is studied . in particular , the periodic forcing of relative equilibria and resonant periodic forcing of relative equilibria to relative periodic orbits , as well as hopf bifurcation from relative equilibria to relative periodic orbits are treated using liapunov - schmidt reduction .
resonant drift phenomena are also studied .
then , these results are applied to the planar spiral waves .
+ for the skew - product finite - dimensional system of differential equations on the center manifold near a relative equilibrium @xcite , the normal form method which further simplifies the system is presented in @xcite . the normal form for the case @xmath25 is obtained .
then , the known results regarding the meandering and drifting of planar spiral waves are recovered , as well as new results regarding the relative homoclinic and heteroclinic trajectories to relative equilibria of @xmath2-actions . + in @xcite
, scheel shows that for a large class of reaction - diffusion systems on the plane , m - armed spiral waves bifurcate from a homogeneous equilibrium when the latter undergoes a hopf bifurcation .
this was done using spatial dynamics .
+ all previous results are valid for planar spiral waves .
the interest to consider spiral waves on non - planar surfaces is motivated by the applicability to problems in physiology , biology and chemistry .
therefore , the study of spiral waves by experiments and numerical simulations of reaction - diffusion systems on a sphere have recently been undertaken . in the case of spiral waves on a sphere
, the dynamics is expected to be quite different because any spiral wave starting from a rotating center can not end at a point .
the number of tips of a wave front can not be odd , and therefore , the dynamics of spiral waves may acquire a new feature qualitatively different from the planar case . + the dynamics of spiral waves in an excitable reaction - diffusion systems on a sphere was numerically investigated by @xcite , and @xcite who employ a spectral method using spherical harmonics as basis functions .
maselko @xcite , as well as maselko and showalter @xcite performed experiments with belousov - zhabotinsky chemical waves propagating on the surface of a sphere .
they observed that a spiral wave winds outward from a meandering source at the north pole and undergoes self - annihilation as it winds into itself at the south pole .
+ in @xcite , the evolution of spiral waves on a circular domain and on a spherical surface is studied by numerical integration of a reaction - diffusion system .
two different asymptotic regimes are observed for both domains .
the first regime is a rigid rotation of an excitation wave around the symmetry axis of the domain .
the second one is a compound rotation including a drift of the rotation center of the spiral wave either along the boundary of the disk or along the equator of the sphere . in this case
the shape of the wave and its rotation velocity are periodically changing in time .
simplified analytical estimates are presented to describe the rigid rotation .
+ in @xcite , numerical integration of an excitable reaction - diffusion system on a sphere is presented .
the evolution of counter - rotating double spiral waves on the sphere is studied and it is shown that the tips of the spiral can either perform a meandering motion or rigidly rotate around a fixed center , depending on the system control parameter .
it is observed that the rotation of the spiral wave on spherical surface is similar to that obtain on the planar surface , except that in the absence of the boundary on a spherical surface some parts of the wave can undergo self - annihilation in contrast to the spiral wave behavior on bounded planar surfaces .
+ in @xcite , the evolution of spiral waves on a spherical surface is studied by integration of a reaction - diffusion system with a global feedback .
it is shown that depending on intensity , sign , and/or time delay in the feedback loop a global coupling can be effectively used either to stabilize the rigid motion of a spiral wave or to completely destroy spiral waves and to suppress self - sustained activity in a confined domain of an excitable medium . + in @xcite , the dynamics of chemical spiral waves in an excitable reaction - diffusion system on a sphere is numerically investigated employing a spectral method using spherical harmonics as basis functions .
different types of spiral waves symmetric or antisymmetric source - source or nearly antisymmetric source - nonsource have been obtained depending on whether the medium is homogeneous or inhomogeneous , and it has been observed that the tips can either rotate steadily or change their shapes .
+ the influence of the topological constraints and the inhomogeneity in the excitability on the geometry and dynamics of spiral waves on a thin spherical shell of excitable media are presented in @xcite . also , rigidly rotating waves on spherical domains
have been studied using kinematical theory by @xcite . + in @xcite , the geometrical stability of the symmetric counter - rotating spiral waves propagating on the unit sphere is studied . by the use of the eikonal equations , it is demonstrated that these solutions are stable under small perturbations normal to the wave front lying on a unit sphere .
+ in @xcite , the authors showed that stationary rotating solutions on a sphere of the eikonal equation under some boundary conditions on tips must be symmetric with respect to the equator with spiral winding out from source tips at polar points . + in @xcite , using the eikonal approximation to a reaction - diffusion system on a sphere , the authors prove existence of a class of counter - rotating double spiral solutions which are highly asymmetric with respect to the equator .
they also derive a power law , linking the angular rotation of the spiral waves with the velocity of plane waves in medium . + in @xcite , renardy considered bifurcations from rotating waves of semi - linear equations that are equivariant under a general compact lie group and applied his results to the laser equations .
his results do not cover the resonance case , which will be covered in our paper .
the theorems were proved using a generalized implicit function theorem on scales of banach spaces .
+ in @xcite , rand examined modulated rotating waves in rotating fluids and applied his results to the taylor - couette problem , where modulated rotating waves , so called modulated wavy vortices occur .
+ in this article we have studied only the trivial isotropy case , because as far as we are aware , there have been observed no m - spiral waves ( @xmath26 ) on spherical surfaces .
also , in @xcite it was numerically verified that there is a critical size of the sphere below which self - sustained spiral waves can not exist .
therefore , we use a sphere of an arbitrary , but fixed radius @xmath27 .
+ in this article , we study non - resonant and resonant supercritical hopf bifurcation from a rotating wave .
the relation between spiral waves and rotating waves , respectively modulated rotating waves will be presented somewhere else .
while we have concentrated on the case of hopf bifurcation , it should be mentioned that we would obtain similar results for _ periodic forcing _ of a rotating wave .
+ in section [ s : rds ] , we define the representation of the lie group @xmath0 on the usual function space .
it is a smooth unitary representation on the fractional spaces relative to the appropriate sectorial operator .
we also associate an @xmath0-equivariant semiflow to the reaction - diffusion system on a sphere , and we recall the notion of tip position function . in section [ s : rw_mrw ] , we recall the notions of relative equilibrium , relative periodic orbit , rotating wave and modulated rotating wave . we also define the notions of frequency vector of a rotating wave and primary frequency vector of a modulated rotating wave , and present some of their properties .
section [ s : hopfbif ] deals with a supercritical hopf bifurcation from a rotating wave .
decomposition of a modulated rotating wave obtained by hopf bifurcation into a primary frequency part and the associated periodic part is proved . in section [
s : frequency ] , the form of the reduced differential equations on @xmath1 , as well as formulae for a primary frequency vectors and the associated periodic part are obtained using baker - campbell - hausdorff formula in the lie algebra @xmath1 , the properties of the exponential map @xmath28 and the properties of the adjoint representations of @xmath0 and @xmath1 . as a consequence , there are three types of motions for the tips of the modulated rotating waves obtained by a hopf bifurcation from a rotating wave . namely , the tip can either quasi - periodically meander on the sphere , such that the primary frequency vector is of order @xmath29 near the frequency vector of the initial rotating wave , or quasi - periodically meander very slowly ( order @xmath30 ) about the primary frequency vector , which can be orthogonal or not to the frequency vector of the initial rotating wave undergoing hopf bifurcation . in the orthogonal case
, the tip motion can be approximated with a very slow drift ( order @xmath30 ) along an equator of the sphere . section [ s : examples_transition ] presents some examples that illustrate the theoretical results obtained in sections [ s : hopfbif ] and [ s : frequency ] . in section [ s :
res_drift ] , the resonant hopf bifurcation from a rotating wave with two parameters is studied . near the resonant hopf bifurcation
, we proved that there exists generically a unique branch of modulated rotating waves with primary frequency vectors orthogonal to the frequency vector of the rotating wave undergoing hopf bifurcation .
then , section [ s : numerical ] deals with some numerical results that illustrate the theoretical results obtained in sections [ s : hopfbif ] and [ s : frequency ] .
the proofs of all theorems are presented in section [ s : proof_transition ] .
three appendices are included , presenting @xmath31 formula in @xmath1 , some properties of the exponential map @xmath32 and of the adjoint representations of @xmath0 and @xmath1 , the equivariant center manifold reduction theorem and some computations done in section [ s : numerical ] .
let @xmath34 and @xmath35 be the unit sphere in @xmath36 . we consider a reaction - diffusion system of the form @xmath37 @xmath38 , @xmath39 , ,
@xmath40 with @xmath41 , + @xmath42 with @xmath43 for @xmath44 , @xmath45 , , @xmath46 the diffusion coefficients , @xmath47 is laplace - beltrami operator on @xmath33 and @xmath48 , @xmath49 , , @xmath50 .
also , @xmath51 .
+ we study the reaction - diffusion system @xmath52 on the function space @xmath53 let @xmath54 and @xmath55 be the fractional spaces of @xmath56 relative to the sectorial operator @xmath57 . using @xcite , to the reaction - diffusion system ( [ e : rd ] ) we associate a local semiflow on an open set of @xmath56 .
let @xmath58 be a @xmath59 function such that @xmath60 , where @xmath61 .
then the reaction - diffusion system ( [ e : rd ] ) defines a sufficiently smooth local semiflow @xmath62 on the function space @xmath63 .
the flow is @xmath59 if @xmath64 and @xmath65 if @xmath66 .
namely @xcite , for any @xmath67 , let @xmath68 be the sufficiently smooth solution ( @xmath59 or @xmath65 respectively ) of the initial value problem given by the reaction - diffusion system ( [ e : rd ] ) and by the initial condition @xmath69 , defined on the maximal interval of existence @xmath70 .
let @xmath71 .
then , the local semiflow @xmath72 is defined by @xmath73 , for any @xmath74 .
any reaction - diffusion system of the above type has the property that is @xmath0-equivariant , where the precise meaning of @xmath0-equivariance is defined below .
[ def : action1 ] the _ representation _
@xmath75 of @xmath0 on @xmath56 is the function @xmath76 , defined by @xmath77 the linear action @xmath78 associated to the unitary representation @xmath75 ( @xcite ) is defined by @xmath79 we denote @xmath80 . from now on , when we talk about @xmath0-equivariance , we mean equivariance with respect to the action @xmath81 defined in ( [ e : action ] ) .
using @xcite , we obtain the following two propositions : @xcite [ thm : smoothness ] the restriction of the representation @xmath75 to @xmath63 , where @xmath54 , is smooth .
@xcite[thm : equivariant ] let @xmath58 be a @xmath59 function such that @xmath60 and @xmath82 .
the local semiflow @xmath62 is @xmath0-equivariant with respect to the action @xmath81 restricted to @xmath63 . for any @xmath67
we follow the time evolution of @xmath12 under the local semiflow @xmath62 using the concept of tip motion of @xmath12 .
[ def : tip ] for @xmath83 , a @xmath65 function @xmath84 that is @xmath0-equivariant is called a _ tip position _ function . by the _ tip _ of @xmath85
, we understand the point @xmath86 .
also , for any @xmath87 , @xmath88 with @xmath89 is called the _ tip motion _ of @xmath12 .
if the function @xmath90 is defined only on an open subset of @xmath63 , then the function @xmath90 is called a _ local tip position _ function .
the tip position function is usually used to follow the time evolution of @xmath12 under the local semiflow @xmath62 , when @xmath91 is a rotating wave or a modulated rotating wave .
we consider that north hemisphere of @xmath92 is the set @xmath93\}$ ] , and south hemisphere is the set @xmath94 .
+ for any @xmath24-equivariant dynamical system , where @xmath24 is a lie group , we can talk about the concepts of relative equilibrium and relative periodic orbit . if @xmath95 for some integer @xmath96 , we can also talk about rotating waves and modulated rotating waves .
@xcite [ def : re ] let @xmath87 be such that the stabilizer of @xmath12 is @xmath97 .
the orbit group of @xmath12 is called a _ relative equilibrium _ for ( [ e : rd ] ) if there exists a matrix @xmath98 such that @xmath99 sometimes , when @xmath100 is a relative equilibrium , we call @xmath12 a _ relative equilibrium_. if @xmath101 , then any solution of the reaction - diffusion system ( [ e : rd ] ) of the form @xmath102 , where @xmath103 is called a _ rotating wave _ for ( [ e : rd ] ) . because the action @xmath81 restricted to @xmath63 is smooth , then the relative equilibrium from definition [ def : re ] is a sufficiently smooth manifold in @xmath63 diffeomorphic to @xmath0 since @xmath104 .
[ def : frequency_rw ] let @xmath105 be a rotating wave as in definition [ def : re ] , where @xmath103 . then , the vector @xmath106 is called the _ frequency vector _ of the rotating wave .
if the vector @xmath107 is in the north hemisphere , then @xmath108 is called the _ frequency _ of the rotating wave .
otherwise , @xmath109 is called the _ frequency _ of the rotating wave
. then , the rotating waves in @xmath100 have their anti - symmetric matrices associated with the frequency vectors on the same adjoint orbit .
the following property shows the importance of the frequency vector associated to any rotating wave : @xcite[prop : tipmrw ] for a rotating wave @xmath91 , the tip motion @xmath88 is a circle on the sphere @xmath110 with the center on the line having the direction of the frequency vector of @xmath91 , and this is independent of the choice of the tip position function .
@xcite [ def : rpo ] let @xmath87 be such that the stabilizer of @xmath12 is @xmath97 .
the set defined by @xmath111 is called a _ relative periodic orbit _ for @xmath52 if it is not a relative equilibrium and there exist a number @xmath112 and a matrix @xmath98 such that @xmath113 if @xmath114 for any @xmath20 , then any solution of the reaction - diffusion system ( [ e : rd ] ) of the form @xmath115 , with @xmath103 is called a _ modulated rotating wave _ for ( [ e : rd ] ) . if @xmath116 for some @xmath117 , then @xmath118 is a @xmath75-periodic solution of the reaction - diffusion ( [ e : rd ] ) . because the action @xmath81 of @xmath0 on @xmath63 is smooth , then the relative periodic orbit defined in definition [ def : rpo ] is a sufficiently smooth manifold in @xmath63 .
there are two properties related to modulated rotating waves , namely : @xcite [ rem : rot ] 1 .
@xmath119 if and only if @xmath120 .
2 . there are periodic solutions of the reaction - diffusion system ( [ e : rd ] ) that are modulated rotating waves in the sense of definition [ def : rpo ] .
then their period is an integer multiple of the corresponding @xmath75 from definition [ def : rpo ] and @xmath121 .
[ def : frequency_mrw ] let @xmath105 be a modulated rotating wave as in definition [ def : rpo ] , where @xmath103 .
then , the vector @xmath106 is called a _ primary frequency vector _ of the modulated rotating wave and the positive number @xmath75 is called the _ relative period _ of the modulated rotating wave . if the vector @xmath122 is in the north hemisphere , then @xmath123 is called the _ primary frequency _ of the modulated rotating wave .
otherwise , @xmath109 is called the _ primary frequency _ of the modulated rotating wave .
a primary frequency vector @xmath124 of a modulated rotating wave @xmath125 is unique up to an integer multiple of @xmath126 .
then the modulated rotating waves in @xmath127 have the anti - symmetric matrices associated with primary frequency vectors on the same adjoint orbit ( if we consider primary frequency vectors to be unique up to a multiple of @xmath128 , @xmath20 of the corresponding unit primary frequency vector ) .
if @xmath125 is a @xmath75-periodic solution of the reaction - diffusion system ( [ e : rd ] ) such that @xmath129 and @xmath130 for all @xmath131 , then a primary frequency vector of @xmath125 can be any vector @xmath132 with @xmath133 for @xmath20 .
the following property shows the importance of the primary frequency vector associated to any modulated rotating wave : @xcite [ prop : tipmmrw ] for a modulated rotating wave @xmath134 , the tip motion @xmath88 has the property that @xmath135 for @xmath20 are points of a circle on the sphere @xmath110 with the center on the line having the direction of the primary frequency vector of @xmath91 ; this is independent of the choice of the tip position function .
in this section , we will consider a supercritical hopf bifurcation from an equilibrium . because the bifurcating periodic solution has amplitude of order @xmath136
, the following definition will simplify the language .
let @xmath137 be a smooth manifold , @xmath4 a normed space or the empty set , @xmath138 an integer and @xmath139 for @xmath140 small .
we say that the function @xmath141 is _ @xmath142 _ on @xmath143 if the function @xmath144 defined by @xmath145 is smooth on @xmath146 , where @xmath147 .
+ we say that @xmath141 is _ @xmath65-@xmath142 _ if @xmath148 is @xmath65 , where @xmath20 , @xmath149 .
we say that @xmath141 is _ sufficiently @xmath142 _
if @xmath148 is sufficiently smooth . let us consider the following reaction - diffusion system @xmath150 @xmath38 , @xmath39 ,
, @xmath40 with @xmath151 , @xmath152 , @xmath43 , @xmath153 , @xmath45 , , @xmath46 the diffusion coefficients and @xmath48 , @xmath49 , , @xmath154 a @xmath59 function with @xmath155 .
we study the reaction - diffusion system ( [ e : rdparam ] ) on the function space @xmath56 defined in section [ s : rds ] .
let @xmath87 be a relative equilibrium that is not an equilibrium for ( [ e : rdparam ] ) at @xmath14 and such that the stabilizer of @xmath12 is @xmath97 .
let @xmath156 .
consider @xmath157 the linearization of the right - hand side of ( [ e : rdparam ] ) with respect to the rotating wave @xmath156 at @xmath14 in the co - rotating frame , that is @xmath158 suppose that : 1 .
@xmath159 is a spectral set with spectral projection @xmath160 , and dim@xmath161 ; 2 .
the semigroup @xmath162 satisfies @xmath163 for some @xmath164 and @xmath165 .
theorems [ thm : cmr_reparam ] ( see appendix [ s : app2_cmr ] ) can be applied . then
, there exist sufficiently smooth functions @xmath166 and @xmath167 such that any solution of @xmath168 on @xmath169 corresponds to a solution of the reaction - diffusion system ( [ e : rdparam ] ) on @xmath170 under the diffeomorphic identification for @xmath171 small .
also , @xmath172 , @xmath173 and @xmath174 , where @xmath175 is the projection onto @xmath176 along @xmath177 .
an example of tip position function is @xmath178 defined by @xmath179 , where @xmath180 is fixed . also , we have the following result .
suppose we omit the parameter @xmath181 .
let @xmath182 be a modulated rotating wave as defined in definition [ def : rpo ] with @xmath183 and @xmath184 , where @xmath185 is defined in theorem [ thm : cmr_reparam ] .
suppose that @xmath186 corresponds to @xmath187 and @xmath182 corresponds to @xmath188 , where @xmath187 and @xmath188 are solutions of the second differential equation in ( [ e : reduceddiffeqparam1 ] ) .
using @xcite , we have : @xcite[prop : stability_rw ] 1 .
@xmath186 is orbitally stable ( respectively unstable ) if @xmath187 is stable ( respectively unstable ) in the second differential equation of ( [ e : reduceddiffeqparam1 ] ) .
@xmath182 is orbitally stable ( respectively unstable ) if @xmath188 is stable ( respectively unstable ) in the second differential equation of ( [ e : reduceddiffeqparam1 ] ) .
let dim@xmath189 .
suppose that a supercritical hopf bifurcation with eigenvalues @xmath190 takes place in the second differential equation of ( [ e : reduceddiffeqparam1 ] ) in @xmath176 at @xmath13 for @xmath14 @xcite .
namely , 1 .
@xmath173 ; 2 .
@xmath191 has eigenvalues @xmath190 ; without loss of generality , we assume that @xmath192 for @xmath193 small ; 3 .
@xmath194 has the eigenvalues @xmath195 with @xmath196 such that @xmath197 ; 4 .
the branch of periodic solutions @xmath198 bifurcating from @xmath13 satisfies @xmath199 for @xmath200 small ; let @xmath201 for all @xmath89 .
let @xmath202 be the period of the solution @xmath198 near @xmath13 that appears for @xmath18 small due to the supercritical hopf bifurcation , where @xmath203 for @xmath200 small .
let us define @xmath204 where @xmath205 is defined in theorem [ thm : cmr_reparam](see appendix [ s : app2_cmr ] ) .
+ since @xmath206 is sufficiently @xmath142 and @xmath207 is sufficiently smooth , it follows that the function @xmath208 is sufficiently @xmath142 .
+ writing @xmath209 , we have @xmath210 for @xmath211 and @xmath200 small .
the functions @xmath212 , @xmath213 , @xmath214 , @xmath215 are @xmath216-periodic in @xmath217 for @xmath200 small .
+ let @xmath218 for @xmath211 and @xmath200 small be the solution of the initial value problem @xmath219 where @xmath220 is defined in ( [ e : xg ] ) .
[ lem : hopf_group ] suppose that the assumptions made in this section hold .
consider the initial value problem @xmath221 where @xmath222 defined by ( [ e : xg ] ) is sufficiently @xmath142 and @xmath216-periodic in @xmath217 for @xmath18 small .
then , there exists a sufficiently @xmath142 solution @xmath223 of the initial value problem ( [ e : auxhopf ] ) , where @xmath224 and @xmath225 is a @xmath226-periodic function such that @xmath227 .
we call @xmath225 the _ periodic part _ of @xmath228 associated to @xmath229 . using lemma [ lem : hopf_group ] we get the following result : [ thm : hopf_rw ] suppose that the assumptions made in this section hold
. then , there exists a sufficiently @xmath142 branch @xmath230 of solutions for the reaction - diffusion system ( [ e : rdparam ] ) such that @xmath231 and for @xmath18 small , @xmath230 is either an orbitally stable modulated rotating wave with a primary frequency vector @xmath229 and the secondary frequency @xmath232 , or an orbitally stable periodic solution with the period @xmath226 .
a slightly modified version of this theorem was proved by @xcite .
they used the liapunov - schmidt reduction on scales of banach spaces .
we use the equivariant manifold reduction and lemma [ lem : hopf_group ] .
+ the bifurcation diagram is illustrated in figure [ figure : figura1 ] .
let @xmath233 . there exist @xmath234 such that the set @xmath235 is an orthonormal basis in @xmath36 satisfying @xmath236 since @xmath1 is isomorphic to @xmath36 and @xmath237}=\overrightarrow x \times \overrightarrow y$ ] , @xmath238 is a basis of the lie algebra @xmath1 such that @xmath239= x_{2 } , [ x_{1 } , x_{2}]= x_{0}^{1 } \mbox { and } [ x_{2 } , x_{0}^{1}]= x_{1}.\ ] ] let us consider the reaction - diffusion system ( [ e : rdparam ] ) . + throughout this section , we will suppose that @xmath87 is a relative equilibrium that is not an equilibrium for ( [ e : rdparam ] ) at @xmath14 and such that the stabilizer of @xmath12 is @xmath97 . also , let @xmath156 .
+ suppose that theorem [ thm : hopf_rw ] holds .
recall that a primary frequency vector @xmath229 of @xmath240 is given by @xmath241 for @xmath18 small .
the theorems stated in this subsection can be proved without using @xmath31 formula in @xmath1 @xcite .
+ let us define the smooth following function @xmath243)= \begin{cases } y & \text{if $ \frac{1}{\left | y \right |}\overrightarrow{y}$ ,
$ \frac{1}{\left | x_{0 } \right |}\overrightarrow{x_{0}}$ are in the same hemisphere or $ y = o_{3}$}\\ ( -\frac{2\pi}{\left | y \right |}+1)y & \text{if $ \frac{1}{\left | y \right |}\overrightarrow{y}$ ,
$ \frac{1}{\left | x_{0 } \right |}\overrightarrow{x_{0}}$ are in different hemispheres and $ y \neq o_{3}$},\\ \end{cases}\ ] ] where points @xmath187 and @xmath244 are identified , and @xmath245 $ ] is an equivalence class of @xmath246 ( see appendix [ s : app1_bch ] for definition of @xmath246 )
. then @xmath247)}=e^{[y]}$ ] for any @xmath248 \in d$ ] .
@xcite[thm : frequencygeneral1 ] suppose that the hypotheses of theorem [ thm : hopf_rw ] are satisfied .
then , there exists a unique sufficiently @xmath142 branch @xmath242 such that @xmath249 for @xmath200 , @xmath250 is a primary frequency vector of @xmath251 for @xmath18 small and @xmath252 .
the branch @xmath242 does not have the property that @xmath253 .
the next corollary shows that we can construct a branch of primary frequency vectors @xmath254 such that @xmath255 .
@xcite[cor : frequencygeneral2 ] suppose that the hypotheses of theorem [ thm : hopf_rw ] are satisfied .
then , there exists a branch @xmath256 for @xmath200 such that @xmath257 is a primary frequency vector of @xmath240 for @xmath18 small and @xmath258 . moreover , 1 .
if @xmath259 for all @xmath20 , then the branch @xmath260 for @xmath200 is sufficiently @xmath142 and @xmath261x_{0}^{1}+o(\sqrt{\lambda})x_{1}+ o(\sqrt{\lambda})x_{2 } \mbox { for small } \lambda \geq 0.\ ] ] ( the branch @xmath228 contains only modulated rotating waves for @xmath200 small ) .
2 . if @xmath262 for some @xmath20 , @xmath263 , then @xmath264 for @xmath200 is continuous and @xmath265 therefore , we can not have frequency -locking phenomena for the modulated rotating waves @xmath228 .
if @xmath262 for some @xmath117 , @xmath263 , it may be possible that the branch @xmath256 be discontinuous at @xmath14 or / and at any @xmath18 such that @xmath266 .
in the rest of the section we give a way of constructing @xmath242 and @xmath256 for @xmath200 using the @xmath31 formula in @xmath1 presented in theorem [ thm : bch_formula ] .
theorems [ thm : aux1 ] , [ cor : aux2 ] and [ thm : globalzeta ] are used to construct the branch @xmath242 .
recall that a primary frequency vector @xmath229 of @xmath240 is given by @xmath267 for @xmath18 small , we show that there exists a sufficiently @xmath142 function @xmath268 such that @xmath269 for at least @xmath270 $ ] and @xmath200 small .
[ thm : aux1 ] suppose that the hypotheses of theorem [ thm : hopf_rw ] are satisfied .
let us consider the following initial value problem @xmath271 \overrightarrow { x^{g}(t , \lambda)},\\ z(0 ) & = & o_{3}. \end{array}\ ] ] then , 1 .
there exists a positive integer @xmath272 independent of @xmath273 such that the initial value problem @xmath274 has a unique sufficiently @xmath142 solution @xmath275 on @xmath276 $ ] and @xmath200 small ; 2 .
@xmath277 + for any @xmath278 $ ] and @xmath200 small ; the integer @xmath272 that appears in theorem [ thm : aux1 ] can be found and , in general , it is not @xmath279 .
therefore , in general , the function @xmath275 is not defined on the entire interval @xmath280 $ ] , for @xmath200 small .
[ cor : aux2 ] suppose that the hypotheses of theorem [ thm : hopf_rw ] are satisfied and the positive integer @xmath272 is the one obtained in theorem [ thm : aux1 ] .
then , for any @xmath153 , @xmath45 ,
, @xmath281 the initial value problem @xmath282 \overrightarrow { x^{g}(t , \lambda)},\\ z(i\frac{2\pi}{n \left | \omega_{\lambda}\right | } ) & = & o_{3 } \end{array}\ ] ] has a unique sufficiently @xmath142 solution @xmath283 on @xmath284 $ ] and @xmath200 small .
combining theorem [ thm : aux1 ] and corollary [ cor : aux2 ] , we obtain the following result : [ thm : globalzeta ] suppose that the hypotheses of theorem [ thm : hopf_rw ] are satisfied .
then , there exists a sufficiently @xmath142 function @xmath268 such that @xmath269 for @xmath211 and @xmath200 small , where @xmath285 is the solution of the initial value problem ( [ e : auxfreq ] ) .
in fact , the function @xmath286 satisfies the following initial value problem for @xmath211 and @xmath200 small : @xmath287 \overrightarrow { x^{g}(t , \lambda ) } ( \mbox { mod } 2\pi),\\ \overrightarrow { z(0 ) } & = & \overrightarrow{o_{3}}. \end{array}\ ] ] in fact , the function @xmath286 obtained in theorem [ thm : globalzeta ] is constructed as follows : 1 .
on the interval @xmath288 $ ] : @xmath289 for all @xmath290 $ ] and @xmath200 small , where @xmath153 , @xmath45 , , @xmath272 and @xmath291 is the solution of the initial value problem ( [ e : odes_frequency1 ] ) on the interval @xmath292 $ ] .
2 . on the interval
@xmath293 $ ] for any integer @xmath294 and @xmath200 small : @xmath295 using the function @xmath286 obtained in theorem [ thm : globalzeta ] , we construct the branches @xmath242 ( @xmath200 ) as follows : [ prop : frequency1 ] suppose that the hypotheses of theorem [ thm : hopf_rw ] are satisfied
. then , the branch @xmath242 for @xmath200 from theorem [ thm : frequencygeneral1 ] is defined by : @xmath296 where @xmath286 is given in theorem @xmath297 ; [ cor : frequency2 ] then , the branch @xmath256 for @xmath200 from corollary [ cor : frequencygeneral2 ] is defined by : 1 .
if @xmath259 for all @xmath20 , then we define @xmath298 for @xmath18 small .
if @xmath299 for some @xmath117 , @xmath263 , then we define : + @xmath255 ; + if @xmath300 , then we define @xmath301 for @xmath200 small ; + if @xmath302 , then we define @xmath303 , where @xmath304 , @xmath305 for @xmath18 small . using lemma [ lem : hopf_group ] for @xmath306 small
, we can associate @xmath307 to each modulated rotating wave or periodic solution of period @xmath308 obtained by theorem [ thm : hopf_rw ] , where @xmath225 , @xmath309 are @xmath226-periodic in @xmath217 for @xmath18 small , as well as @xmath310 .
also , let @xmath311 and @xmath312 be such that @xmath313 , that is @xmath314 .
[ thm : periodic_part ] suppose that the hypotheses of theorem [ thm : hopf_rw ] are satisfied . then , 1 . if @xmath259 for all @xmath20 , we have @xmath315 where @xmath316 is @xmath317-periodic in @xmath217 and sufficiently @xmath142 for @xmath89 and @xmath200 small .
if @xmath318 for some @xmath20 , @xmath263 , we have @xmath319 where @xmath320 is sufficiently @xmath142 for @xmath89 and @xmath200 small , and @xmath321 for @xmath200 small . in fact , @xmath322 ( respectively @xmath323 ) for @xmath89 and @xmath200 small , where @xmath286 is given in theorem [ thm : globalzeta ] .
using theorems [ prop : frequency1 ] , [ cor : frequency2 ] and [ thm : periodic_part ] we get the following result : [ thm : mrw ] suppose that the hypotheses of theorem [ thm : hopf_rw ] are satisfied .
then , 1 . if @xmath259 for all @xmath20 , we have @xmath324t}e^{\sqrt{\lambda}y(t,\lambda)}\psi(\sqrt{\lambda}r(t,\lambda)),\ ] ] where @xmath325 and @xmath326 are sufficiently @xmath142 and @xmath226-periodic in @xmath217 for @xmath200 small .
if @xmath318 for some @xmath20 , @xmath263 , we have @xmath327 t}e^{x_{0}t+\sqrt{\lambda}h(t,\lambda)}\psi(\sqrt{\lambda}r(t,\lambda)),\ ] ] where @xmath326 and @xmath320 are sufficiently @xmath142 and @xmath326 is @xmath226-periodic in @xmath217 for @xmath200 small , as well as @xmath328 for @xmath200 small .
in fact , we have that the initial value problem ( [ e : reduceddiffeqparam1 ] ) is equivalent with the following initial value problem @xmath329 \overrightarrow { x_{g}(q , \lambda ) } \left ( \mbox { mod } 2\pi \right ) , \\ \dot{q } & = & x_{n}(q , \lambda),\\ \overrightarrow{z(0 ) } & = & \overrightarrow{o_{3}}. \end{array}\ ] ] we could have considered that the differential equation @xmath330 has a @xmath331-periodic solution @xmath332 instead of assuming that a supercritical hopf bifurcation takes places at @xmath333 for @xmath334 .
all the results remain valid .
throughout this section , we let @xmath335 . here
we present some examples of functions @xmath220 for which we can get the closed form solution @xmath218 of the initial value problem ( [ e : auxfreq ] ) .
they obey theorem [ thm : mrw ] . [
ex : exhoph1 ] let @xmath336 be a sufficiently smooth periodic function of period @xmath337 such that @xmath338 for @xmath200 small . for any @xmath211 and @xmath339 small , we define @xmath340 it is clear that @xmath212 is a sufficiently @xmath142 , @xmath341-periodic function such that @xmath342
. then , the initial value problem ( [ e : auxfreq ] ) has the solution @xmath343 clearly , @xmath344 is not orthogonal to @xmath345 .
[ ex : exhoph2 ] let @xmath346 and @xmath336 be a sufficiently smooth periodic function of period @xmath337 such that @xmath347 for @xmath200 small .
+ for any @xmath211 and @xmath200 small , we define @xmath348 it is clear that @xmath212 is a sufficiently @xmath142 , @xmath349- periodic function such that @xmath342 . then , the initial value problem ( [ e : auxfreq ] ) has the solution @xmath350 clearly , @xmath351 is orthogonal to @xmath345 .
[ ex : exhoph3 ] let @xmath346 and @xmath336 be a sufficiently smooth periodic function of period @xmath337 such that @xmath352 for @xmath200 small .
+ for any @xmath211 and @xmath200 small , we define @xmath353 it is clear that @xmath212 is a sufficiently @xmath142 , @xmath341-periodic function such that @xmath342
. then , the initial value problem ( [ e : auxfreq ] ) has the solution @xmath354 clearly , @xmath355 is not orthogonal to @xmath345 .
[ ex : exhoph4 ] let @xmath356 for @xmath117 , @xmath263 and @xmath357 be a sufficiently smooth periodic function of period @xmath337 such that @xmath358 for @xmath359 , @xmath360 small .
+ for any @xmath211 , @xmath200 small and @xmath360 small , we define @xmath361 it is clear that @xmath362 is a sufficiently @xmath363 , @xmath341 -periodic function such that @xmath364 .
then , the initial value problem ( [ e : auxfreq ] ) for two parameters @xmath181 , @xmath365 has the solution @xmath366 clearly , for @xmath367 , we have orthogonality , that is @xmath368 is orthogonal to @xmath345 .
[ ex : exhopf5 ] let @xmath346 and @xmath336 be a sufficiently smooth periodic function of period @xmath337 such that @xmath352 for @xmath200 small .
+ for any @xmath211 and @xmath200 small , we define @xmath369 it is clear that @xmath212 is a sufficiently @xmath142 , @xmath341-periodic function such that @xmath342 .
then , the initial value problem ( [ e : auxfreq ] ) has the solution @xmath370 this clearly represents a non - uniformly rigid rotation about the line containing the vector @xmath371 .
this is a degenerate situation for the case of one parameter . also , the case of a @xmath372-periodic solution is a degenerate situation for one parameter .
therefore , we do not present it here .
+ these examples will be used in section [ s : numerical ] . as we have seen in examples [ ex : exhoph2 ] and [ ex : exhoph3 ] , if we have a single parameter @xmath181 and @xmath262 for some @xmath20 , we can have three different types of solutions for the initial value problem ( [ e : auxfreq ] ) . in example([ex : exhoph4 ] ) , we have two parameters @xmath200 and @xmath365 , as well as @xmath262 for some @xmath20 and we have a branch @xmath373 for which @xmath374 is orthogonal to @xmath375 . in next section , we prove that this is generically valid .
we consider the following reaction - diffusion system @xmath376 where @xmath377 , @xmath39 , , @xmath378 with @xmath41 , @xmath379 with @xmath43 for @xmath153 , @xmath45 , , @xmath46 are the diffusion coefficients and @xmath380 , @xmath49 , , @xmath381 are sufficiently smooth functions such that @xmath382 for @xmath193 , @xmath360 small . + we study the reaction - diffusion system ( [ e : rdparams ] ) on the function space @xmath56 defined in section [ s : rds ] .
let @xmath383 be the @xmath0-equivariant sufficiently smooth local semiflow defined as in section [ s : rds ] .
let @xmath67 be a relative equilibrium that is not an equilibrium for ( [ e : rdparams ] ) at @xmath384 and such that the stabilizer of @xmath12 is @xmath97 .
consider @xmath157 the linearization of the right - hand side of ( [ e : rdparams ] ) with respect to the rotating wave @xmath385 at @xmath386 in the co - rotating frame , that is @xmath387 suppose that : 1 .
@xmath159 is a spectral set with spectral projection @xmath160 , and dim@xmath388 2 .
the semigroup @xmath162 satisfies @xmath389 for some @xmath164 and @xmath165 .
theorems [ thm : cmr_reparam ] with parameters @xmath181 , @xmath390 can be applied .
+ there exist sufficiently smooth functions @xmath391 and @xmath392 such that any solution of @xmath393 on @xmath169 corresponds to a solution of the reaction - diffusion system ( [ e : rdparam ] ) on @xmath394 under the diffeomorphic identification for @xmath395 and @xmath396 small .
also , @xmath397 , @xmath398 and @xmath399 , where @xmath175 is the projection onto @xmath176 along @xmath177 .
let dim@xmath400 .
suppose that a supercritical hopf bifurcation with eigenvalues @xmath190 takes place in the second differential equation of ( [ e : reduceddiffeqparam2 ] ) in @xmath176 at @xmath13 for @xmath401 , that is : 1 .
@xmath398 ; 2 .
@xmath402 has eigenvalues @xmath190 ; without loss of generality , we assume that @xmath403 for @xmath193 , @xmath360 small ; 3
. @xmath404 has the eigenvalues @xmath405 with @xmath406 ) and @xmath407 .
this implies that @xmath408 for some sufficiently smooth curve @xmath409 with @xmath410 .
this curve represents the hopf points and without loss of generality , we suppose that @xmath411 for @xmath360 small ; 4 .
the branch of periodic solutions @xmath412 bifurcating from @xmath333 generically satisfies @xmath413 .
for @xmath306 small and @xmath360 small , let @xmath414 be the period of the periodic solution @xmath415 near @xmath333 , that appears due to the supercritical hopf bifurcation , where @xmath416 for @xmath200 small and @xmath417 small . + by theorem [ thm : hopf_rw ] , there exists a sufficiently @xmath142 branch @xmath418 for @xmath200 small and @xmath419 small such that for @xmath360 small , @xmath420 and such that for @xmath421 small and @xmath360 small , @xmath418 is an orbitally stable modulated rotating wave or periodic solution of period @xmath422 .
let @xmath423 be the sufficiently @xmath142 branch such that for @xmath360 small , @xmath424 , and for @xmath18 small and @xmath360 small , @xmath425 is a primary frequency vector corresponding to @xmath418 .
we have the following two results : [ thm : resonant_drift1 ] suppose the previous assumptions hold and that @xmath318 for some @xmath20 , @xmath263 .
let @xmath426 then , if @xmath427 , there exists a sufficiently @xmath142 curve @xmath428 for @xmath200 small such that @xmath429 , and there exists a sufficiently @xmath142 branch of orbitally stable modulated rotating waves @xmath430 for @xmath18 small , with a primary frequency vector @xmath431 orthogonal to @xmath432 .
[ cor : resonant_drift2 ] suppose that theorem [ thm : resonant_drift1 ] hold .
then , 1 .
the branch @xmath433 is sufficiently @xmath142 for @xmath200 small and @xmath434 for @xmath200 small .
2 . @xmath435 , for @xmath436 and @xmath200 small , where @xmath437 is a sufficiently @xmath142 function such that @xmath438 for @xmath200 small .
3 . there exists a branch @xmath439 with @xmath200 that is discontinuous at @xmath14 , but @xmath440 is continuous for @xmath200 small and @xmath441 for @xmath200 small . the branch @xmath428 for @xmath200 small can be found using the @xmath31 formula in @xmath1 .
the parameter space in the case of the resonant hopf bifurcation is illustrated in figure ( [ figure : figura2 ] ) .
the results in sections [ s : hopfbif ] , [ s : frequency ] and [ s : res_drift ] show that a rotating wave on a sphere generically undergoes a transition to quasi - periodic meandering ( modulated rotating wave ) at a hopf bifurcation .
furthermore , the primary frequency vector of the modulated rotating wave is determined by both the critical hopf eigenvalues and the frequency of the rotating wave undergoing the bifurcation . in particular , resonances between the critical hopf eigenvalue and the frequency of the rotating wave undergoing the bifurcation lead to orthogonal drift .
while we have concentrated on the case of hopf bifurcation , it should be mentioned that we would obtain similar results for _ periodic forcing _ of a rotating wave : equation ( [ e : auxfreq ] ) represents the dynamical equations of a periodically forced rotating wave . in a spherical heart , for example
, one could alter the dynamics of a rotating spiral wave simply by applying a ( weak ) periodic forcing ( e.g. pacing ) , thereby inducing meandering . by an appropriate choice of the forcing frequency
, one could then control the direction of the primary frequency vector ( i.e. steer the spiral tip to another location ) .
of course , this assumes perfect spherical symmetry .
the case when geometrical imperfections appear can also influence the dynamics of rotating waves and it will be presented somewhere else .
theorem [ thm : mrw ] shows that , generically , by a supercritical hopf bifurcation of a rotating wave , not only we get modulated rotating waves ( that is a quasi - periodic tip motion ) , but also we get a quasi - periodic meandering tip motion .
it is possible that this meandering motion is not of epicycle - type .
we illustrate this using the examples presented in section [ s : examples_transition ] .
+ let @xmath442 , @xmath443 , @xmath444 , + @xmath445 , @xmath446 , + @xmath447 .
+ we have seen that the study of the reaction - diffusion system ( [ e : rdparam ] ) on @xmath448 reduces to the study of the finite - dimensional system ( [ e : reduceddiffeqparam1 ] ) on the center manifold @xmath449 of the relative equilibrium @xmath100 , with @xmath450 .
let us denote @xmath451 for any @xmath452 and @xmath193 small .
+ if we parameterize @xmath0 by euler angles , that is @xmath453 , where @xmath454 , @xmath455 $ ] , @xmath456 , the finite dimensional system ( [ e : reduceddiffeqparam1 ] ) becomes @xmath457,\\ \dot{\theta } & = & -f^{y}(q,\lambda)\sin \phi+ f^{x}(q , \lambda)\cos \phi,\\ \dot{\psi } & = & \frac{1}{\sin \theta}\left [ f^{y}(q,\lambda)\cos \phi + f^{x}(q,\lambda)\sin \phi \right ] , \\ \dot{q } & = & x_{n}(q,\lambda ) . \end{array}\ ] ] we consider the following initial value problem associated with the system ( [ e : numerical1 ] ) : @xmath458,\\ \dot{\theta } & = & -f^{y}(q,\lambda)\sin \phi+ f^{x}(q , \lambda)\cos \phi,\\ \dot{\psi } & = & \frac{1}{\sin \theta}[f^{y}(q,\lambda)\cos \phi + f^{x}(q,\lambda)\sin \phi],\\ \phi(0 ) & = & 0,\\ \theta(0 ) & \neq & 0,\\ \psi(0 ) & = & 0,\\ \dot{q } & = & x_{n}(q,\lambda ) .
\end{array}\ ] ] we choose @xmath459 near @xmath187 in ( [ e : numerical11 ] ) .
+ we consider @xmath460 and a supercritical hopf bifurcation takes place in @xmath461 at @xmath333 for @xmath334 with eigenvalues @xmath190 .
let @xmath198 be the periodic solution of the second differential equation in ( [ e : reduceddiffeqparam ] ) , that appears by a supercritical hopf bifurcation for @xmath18 small .
let @xmath462 be its period , where @xmath463 for @xmath200 small .
let @xmath464 for any @xmath465 .
if we substitute @xmath198 in @xmath207 and in the first three equations of the system ( [ e : numerical11 ] ) , we get @xmath466 for @xmath200 small and @xmath465 .
if we write @xmath467 , @xmath468 and @xmath469 , we get that @xmath470 is @xmath372-periodic . + the following initial value problem
is obtained @xmath471,\\ \dot{\theta } & = & -f^{yy}(t,\lambda)\sin \phi+ f^{xx}(t , \lambda)\cos \phi,\\ \dot{\psi } & = & \frac{1}{\sin \theta}[f^{yx}(t,\lambda)\cos \phi + f^{xx}(t,\lambda)\sin \phi],\\ \phi(0 ) & = & 0,\\ \theta(0 ) & \neq & 0,\\ \psi(0 ) & = & 0 . \end{array}\ ] ] we choose @xmath459 near @xmath187 in ( [ e : numerical2 ] ) .
+ we use a maple program that integrates numerically the system ( [ e : numerical2 ] ) and finds numerically a primary frequency vector ( @xcite , if its norm is not @xmath187 or @xmath472 ) of the modulated rotating wave that appears by a supercritical hopf bifurcation , as discussed in section [ s : hopfbif ] . then ,
for a choice of the point @xmath180 near @xmath473 , we represent the tip motion @xmath474 for @xmath200 small .
recall that @xmath475 .
+ we consider the following cases : 1 .
@xmath476 is given in example ( [ ex : exhoph1 ] ) from section [ s : examples_transition ] ; @xmath477 ; @xmath478 , @xmath479 , @xmath480 , @xmath481 , @xmath482 , @xmath483 .
we get figures [ figure : figtest - fig1 ] to [ figure : figtest - fig2 ] for @xmath484 , @xmath485 . + since @xmath486 for any @xmath20
, we have the nonresonant case . on figures [ figure : figtest - fig1 ] and [ figure : figtest - fig2 ] , it is visualized the tip motion given by ( [ e : motion1 ] ) of the modulated rotating waves @xmath228 obtained by a supercritical hopf bifurcation that takes place in @xmath461 at @xmath333 for @xmath334 . on figure [ figure : figtest - fig1 ] we consider @xmath484 . on figure [
figure : figtest - fig2 ] we consider @xmath487 .
+ on the second of figure [ figure : figtest - fig1 ] we plot the points @xmath488 for some @xmath489 , @xmath279 , @xmath45 , and we can see that they are points of a circle on the sphere @xmath110 with the center on the line having the direction of the primary frequency vector of @xmath251 , where @xmath484 .
+ the grey line is the line containing the frequency of the rotating wave undergoing the hopf bifurcation , @xmath375 .
the black lines are the lines corresponding to the primary frequency vector associated to the modulated rotating wave .
one of them is computed numerically and the other is the exact one .
we can see that they are very close .
@xmath476 is given in example ( [ ex : exhoph2 ] ) from section [ s : examples_transition ] ; @xmath490 ; @xmath491 , @xmath492 , @xmath481 , @xmath493 , @xmath494 .
+ we get figures [ figure : figtest - fig3 ] to [ figure : figtest - fig4 ] for @xmath487 , @xmath495 . + since @xmath262 for @xmath496 , we have the resonant case . on figures [ figure : figtest - fig3 ] and [ figure : figtest - fig4 ] , it is visualized the tip motion given by ( [ e : motion1 ] ) of the modulated rotating waves @xmath228 obtained by a supercritical hopf bifurcation that takes place in @xmath461 at @xmath333 for @xmath334 . on figure [ figure : figtest - fig3 ]
we consider @xmath487 . on figure [
figure : figtest - fig4 ] we consider @xmath497 .
we can see that the primary frequency vectors associated to the modulated rotating waves are orthogonal to @xmath375 .
we call this phenomenon resonant drift . + on the second of figure [ figure : figtest - fig3 ] we plot the points @xmath498 for some @xmath489 , @xmath279 , @xmath45 , and we can see that they are points of a circle on the sphere @xmath110 with the center on the line having the direction of the primary frequency vector of @xmath251 , where @xmath487 .
+ the grey line is the line containing the frequency of the rotating wave undergoing the hopf bifurcation , @xmath375 .
the black line is the line corresponding to the primary frequency vector associated to the modulated rotating wave .
@xmath476 is given in example ( [ ex : exhoph3 ] ) from section [ s : examples_transition ] ; @xmath477 ; @xmath499 , @xmath480 , @xmath481 , @xmath500 , @xmath501 + we get figures [ figure : figtest - fig5 ] to [ figure : figtest - fig6 ] for @xmath487 , @xmath502 .
+ since @xmath262 for @xmath496 , we have the resonant case . on figures [
figure : figtest - fig5 ] and [ figure : figtest - fig6 ] , it is visualized the tip motion given by ( [ e : motion1 ] ) of the modulated rotating waves @xmath228 obtained by a supercritical hopf bifurcation that takes place in @xmath461 at @xmath333 for @xmath334 . on figure [ figure : figtest - fig5 ] we consider @xmath487 . on figure [ figure : figtest - fig6 ] we consider @xmath503 .
we can see that the primary frequency vectors associated to the modulated rotating waves are not orthogonal to @xmath375 .
this can happen in the resonant case if we consider only one parameter @xmath181 .
+ on the second of figure [ figure : figtest - fig5 ] we plot the points @xmath488for some @xmath489 , @xmath279 , @xmath45 , and we can see that they are points of a circle on the sphere @xmath110 with the center on the line having the direction of the primary frequency vector of @xmath251 , where @xmath487 . + the grey line is the line containing the frequency of the rotating wave undergoing the hopf bifurcation , @xmath375 . the black line is the line corresponding to the primary frequency vector associated to the modulated rotating wave .
@xmath504 \mbox{$x_{tip}(\mathbf{\phi}(t , u_{0.01},0.01))$ } & \mbox{$x_{tip}(\mathbf{\phi}(i\frac{2\pi}{20.01},u_{0.01},0.01))$,$i=1,\ldots , 5 $ } \end{array}$ ] @xmath505 \mbox{$x_{tip}(\mathbf{\phi}(t , u_{0.05},0.05))$ } \end{array}$ ] @xmath506 \mbox{$x_{tip}(\mathbf{\phi}(t , u_{0.05},0.05))$ } & \mbox{$x_{tip}(\mathbf{\phi}(i\frac{2\pi}{20.05},u_{0.05},0.05))$,$i=1,\ldots , 5 $ } \end{array}$ ] @xmath507 \mbox{$x_{tip}(\mathbf{\phi}(t , u_{0.1},0.1))$ } \end{array}$ ] @xmath508 \mbox{$x_{tip}(\mathbf{\phi}(t , u_{0.05},0.05))$ } & \mbox{$x_{tip}(\mathbf{\phi}(i\frac{2\pi}{20.05},u_{0.05},0.015))$ , $ i=1 , \ldots , 5 $ } \end{array}$ ] @xmath509 \mbox{$x_{tip}(\mathbf{\phi}(t , u_{0.25},0.25))$ } \end{array}$ ] the nonuniformly rigid rotation shown in example [ ex : exhoph4 ] is not generically .
therefore , we do not present it here .
let us omit the parameter @xmath181 . since @xmath0 is a compact manifold and @xmath208 is a sufficiently @xmath142 function , the initial value problem ( [ e : auxhopf ] ) has a unique sufficiently @xmath142 solution that is globally defined .
let @xmath510 be this solution .
since the exponential map @xmath511 is surjective , there exists a matrix @xmath512 such that @xmath513 + let us make the following change of variable @xmath514 .
+ then the first equation of ( [ e : auxhopf ] ) becomes : @xmath515 or since @xmath516 , we get @xmath517 or since @xmath518 it follows that @xmath519 the initial condition @xmath520 becomes @xmath521 . + let us consider now the initial value problem @xmath522 the initial value problem ( [ e : auxhopf_b ] ) has a unique global solution since @xmath0 is a compact manifold .
+ since @xmath510 is the solution of the initial value problem ( [ e : auxhopf ] ) , it follows that @xmath523 is a solution of ( [ e : auxhopf_b ] ) and @xmath524 .
+ let us define @xmath525 we will check that @xmath526 is another solution of ( [ e : auxhopf_b ] ) .
we have @xmath527 and for any @xmath211 , @xmath528 where we have used the fact that @xmath208 is a @xmath75-periodic function .
+ therefore , @xmath529 for @xmath211 , that is , @xmath530 is @xmath75-periodic .
+ so , @xmath531 for @xmath211 .
+ taking into account the parameter @xmath18 , since @xmath212 is sufficiently @xmath142 for @xmath211 and @xmath18 , we have that @xmath285 is sufficiently @xmath142 for @xmath89 and @xmath18 small .
if we work with @xmath200 , then @xmath218 is sufficiently @xmath142 for @xmath89 and @xmath200 small .
we apply theorems [ thm : cmr_reparam ] .
then , it is sufficient to study the differential equations ( [ e : reduceddiffeqparam1 ] ) .
+ the supercritical hopf bifurcation in @xmath532 for the second differential equation in ( [ e : reduceddiffeqparam1 ] ) at @xmath13 for @xmath14 implies that there exists a unique sufficiently @xmath142 branch of periodic solutions @xmath533 for @xmath306 near @xmath13 .
+ recall that we have denoted by @xmath534 the period of @xmath533 , where @xmath203 is a sufficiently smooth function for @xmath200 small .
+ if we substitute @xmath533 in the first of the differential equations ( [ e : reduceddiffeqparam ] ) , we get the differential equation @xmath535 where we recall that the function @xmath212 is sufficiently @xmath142 and @xmath216-periodic for @xmath211 and @xmath200 small ( and it is defined by ( [ e : initialdata ] ) ) . + we consider the following initial value problem on @xmath0 : @xmath536 because any solution of the differential equation ( [ e : eq1 ] ) with @xmath537 , where @xmath538 is given by @xmath539 and @xmath540 is the solution of the initial value problem ( [ e : periodiceq ] ) . + for @xmath18 small , by applying lemma [ lem : hopf_group ] , we get @xmath541 where @xmath225 has period @xmath542 and @xmath543 . + consequently , by using theorems [ thm : cmr_reparam ] , there exists a solution @xmath228 of the reaction - diffusion system ( [ e : rdparam ] ) of the form : @xmath544 with @xmath545 and @xmath546 appears in the statement of theorem [ thm : cmr_reparam ] ( see appendix [ s : app2_cmr ] ) .
if we write @xmath547 then the function @xmath548 is @xmath542-periodic in @xmath217 , since @xmath549 and @xmath198 are @xmath542-periodic in @xmath217 . also , we get @xmath550 we check that , at least generically , @xmath551 for any @xmath552 and @xmath306 small . +
we prove that @xmath229 is a primary frequency vector of @xmath228 .
+ suppose that there exists a @xmath553 such that @xmath554 and for @xmath555 we have @xmath551 .
let @xmath556 , where @xmath557 and @xmath558 .
let @xmath559 , where @xmath560 . using the definition and @xmath0-equivariance of the semiflow @xmath62
, we get @xmath561 . + therefore ,
@xmath562 and @xmath563 .
it follows that @xmath564 and so @xmath565 .
thus , @xmath566 and @xmath567 because @xmath568 .
+ also , we get @xmath569 . + let @xmath570 .
@xmath571 is @xmath542-periodic , because @xmath549 is @xmath542-periodic and @xmath567 .
thus , @xmath572 then @xmath573 , and it follows that @xmath574 .
+ we have that @xmath575 is close to @xmath12 and @xmath576 , therefore , generically , @xmath577 . since @xmath578
, it follows that @xmath579 and , since @xmath568 , @xmath580 and @xmath581 .
because @xmath546 is a local diffeomorphism ( see @xcite ) and @xmath206 , it results that @xmath582 .
we know that @xmath198 is a solution of the differential equation @xmath583 .
therefore , it follows that @xmath198 is @xmath584-periodic .
but it has the period @xmath542 . that is , @xmath585 contradicting the fact that @xmath553 . + if @xmath586 for some @xmath20 , then @xmath240 is a periodic solution with the period @xmath542 .
+ otherwise , @xmath240 is a modulated rotating wave with a primary frequency vector @xmath250 .
+ by proposition [ prop : stability_rw ] , the orbital stability of the modulated rotating wave or of the periodic solution obtained above is the same as the stability of the periodic solution @xmath533 .
+ therefore , since the periodic solution @xmath533 for @xmath18 is stable , then the corresponding modulated rotating wave or periodic solution is orbitally stable as well .
+ since @xmath285 and @xmath198 are sufficiently @xmath142 and @xmath546 is a local diffeomorphism , we get that @xmath240 is sufficiently @xmath142
. the same result can be inferred if we take into account that @xmath587 .
we will use the following result ( see @xcite ) : [ lem : scalar ] let @xmath588 be a continuous function on an open connected set @xmath589 and such that the initial value problem for the scalar equation @xmath590 has a unique solution @xmath591 on @xmath592 . if @xmath593 is continuous and @xmath594 for @xmath592 and @xmath595 , then the solutions of @xmath596 , @xmath597 exists on @xmath598 and @xmath599 for @xmath592 .
there exists a constant @xmath600 independent of @xmath200 small such that @xmath601 for @xmath211 and @xmath200 small . for @xmath200 small and any @xmath602
, we have @xmath603\overrightarrow { x^{g}(t , \lambda)}\right \| & \leq & \nonumber \\ \left [ \left \| i_{3 } \right \| + \frac{1}{2 } \left \|
z \right \| + \left ( \frac{1}{\left | z \right | ^{2}}+\frac{\cos \frac{\left | z \right |}{2 } } { 2\sin \frac{\left | z \right |}{2}\left| z \right |}\right ) \left \| z \right \| ^{2 } \right ] m,\end{aligned}\ ] ] @xmath604\overrightarrow { x^{g}(t , \lambda ) } \right \| & \leq & \nonumber \\
m \left [ \sqrt{3}+\frac{1}{2}\sqrt{2}\left | z \right |+ \left ( \frac{1}{\left | z \right | ^{2}}+\frac{\cos \frac{\left | z \right |}{2 } } { 2\sin \frac{\left | z \right |}{2}\left| z \right|}\right ) 2\left | z \right | ^{2}\right ] \end{aligned}\ ] ] or , if we use the fact that the function @xmath605 is decreasing on @xmath606 $ ] , @xmath604\overrightarrow { x^{g}(t , \lambda ) } \right \| & \leq & \nonumber \\ m\left [ \sqrt{3}+\frac{1}{2}\sqrt{2}\left | z \right |+(2 + 2\cdot 1)\right ] \leq m(6+\left | z \right |).\end{aligned}\ ] ] the initial value problem @xmath607 has solution on an maximal interval @xmath608 $ ] , where @xmath609 is defined such that @xmath610 on @xmath611 $ ] ; @xmath609 is independent of @xmath181 .
+ therefore , using lemma [ lem : scalar ] with @xmath612 , it follows that the sufficiently @xmath142 solution @xmath275 of the initial value problem ( [ e : odes_frequency ] ) is defined for any @xmath613 $ ] and @xmath200 small , and that @xmath614 for any @xmath613 $ ] and @xmath200 small . + if we choose a positive integer @xmath615 such that @xmath616 , then the initial value problem ( [ e : odes_frequency ] ) has a solution @xmath275 defined and sufficiently @xmath142 on @xmath617 $ ] and @xmath200 small . since @xmath618
, then we can choose @xmath272 independent of @xmath181 for @xmath200 small , where @xmath619 .
+ this ends the proof of ( 1 ) of theorem [ thm : aux1 ] .
+ to prove the formula ( [ e : zeta0 ] ) , we write the taylor expansion of + @xmath620 : @xmath621 if we substitute ( [ e : zetataylor1 ] ) , ( [ e : zetataylor2 ] ) and ( [ e : initialdata ] ) in the first of the differential equations given by ( [ e : odes_frequency ] ) with @xmath622 , identify the orders of @xmath623 in both sides and then take @xmath335 , we get the formula ( [ e : zeta0 ] ) for @xmath275 . for the sake of simplicity , we omit @xmath623 .
+ the result is @xmath624\overrightarrow{x_{0}x^{1}_{0}+ \epsilon x_{1}x_{1}+\epsilon x_{2}x_{2}}.\nonumber\\\end{aligned}\ ] ] since @xmath625 is sufficiently smooth , we get @xmath626 . + therefore ,
@xmath627 and @xmath628 . in the right - hand side of ( [ e : tayloraux ] )
, the coefficient of @xmath629 is @xmath630 and thus @xmath631 , where @xmath632 , @xmath633 , @xmath634 are defined in ( [ e : initialdata ] ) .
therefore , @xmath635 this gives the formula ( [ e : zeta0 ] ) after we relabel @xmath636 , @xmath637 and @xmath638 to @xmath639 , @xmath639 and @xmath640 .
this ends the proof of ( 2 ) of theorem [ thm : aux1 ] .
we have the same notations as in the proof of theorem [ thm : aux1 ] . from the proof of theorem
[ thm : aux1 ] , we know that the initial value problem @xmath641 has the solution @xmath642 defined on @xmath608 $ ] with @xmath610 on @xmath611 $ ] .
let @xmath643 $ ] .
consider the initial value problem @xmath644 the solution to this problem satisfies @xmath610 for any @xmath645 $ ] .
+ we repeat the previous argument for @xmath646 for @xmath647 , @xmath648 , , @xmath281 . from this and lemma
[ lem : scalar ] with @xmath649 for @xmath647 , @xmath648 , , @xmath281 , we get the conclusion of corollary [ cor : aux2 ] .
recall that @xmath650 for @xmath200 small . 1 . on the interval
@xmath651 $ ] : + consider the initial value problem @xmath652 let us make the change of variable @xmath653 near @xmath654 in the initial value problem ( [ e : auxhopf1 ] ) .
+ then , using proposition [ prop : properties_exp ] ( 2 ) we get the following initial value problem in @xmath148 : + @xmath655 @xmath656 @xmath657 or , using again proposition [ prop : geometrical_interp ] ( 1 ) we get @xmath658 using proposition [ prop : geometrical_interp ] ( 3 ) , we have that @xmath659 ( [ e : auxadjoint ] ) will proved later . +
if we put ( [ e : auxadjoint ] ) into ( [ e : zeta13 ] ) , we get @xmath660 we will prove later that for any @xmath661 , @xmath662 since we are looking for @xmath663 , the system ( [ e : zeta2 ] ) becomes @xmath664 \overrightarrow{x^{g}(t , \lambda)},\\ z(0 ) & = & o_{3 } , \end{array}\ ] ] where we are looking for a solution @xmath148 such that @xmath602 on some interval @xmath665 $ ] .
+ we now prove ( [ e : check1 ] ) .
it is enough to check that @xmath666=i_{3}.\ ] ] using proposition [ prop : geometrical_interp ] ( 3 ) , that is @xmath667 and @xmath668 , we have @xmath669\\ \shoveleft { = i_{3}+\frac{1}{2}z+\left ( \frac{1}{\left | z \right | ^{2}}-\frac{\cos \frac{\left | z \right |}{2}}{2\sin \frac{\left | z \right |}{2}\left | z \right | } \right ) z^{2}+ \frac{\cos\left | z \right | -1 } { 2\left | z \right | ^{2}}z^{2}}\\ \shoveleft{+\frac{\cos\left | z \right | -1 } { \left | z \right | ^{2}}z+\frac{\cos \left | z \right | -1 } { \left | z \right | ^{2 } } \left ( \frac{1}{\left | z \right | ^{2}}-\frac{\cos \frac{\left | z \right |}{2}}{2\sin \frac{\left | z \right |}{2}\left | z \right |}\right ) z^{3 } + \frac{\left | z \right | -\sin\left | z \right | } { \left | z \right | ^{3}}z^{2}+ \frac{\left | z \right | -\sin\left | z \right | } { 2\left | z \right | ^{3}}z^{3}}\\ \shoveleft{+\frac{\left | z
\right | -\sin\left | z \right | } { \left | z \right | ^{3 } } \left ( \frac{1}{\left | z \right | ^{2}}-\frac{\cos \frac{\left | z \right |}{2}}{2\sin \frac{\left | z \right | } { 2}\left| z \right |}\right ) z^{4}}\\ \shoveleft{=i_{3}+\left [ \frac{1}{2}+\frac{\cos\left
| z \right |-1 } { \left | z \right | ^{2}}- \frac{\cos\left | z \right | -1 } { \left | z \right | ^{2}}+\frac{\cos \left | z \right | -1 } { \left | z \right
| ^{2 } } \frac{\left | z
\right |\cos \frac{\left | z \right |}{2}}{2\sin \frac{\left | z \right |}{2}}-\frac{1}{2}+ \frac{\sin \left | z \right |}{2\left | z \right | } \right ] z}\\ \shoveleft{\left [ \frac{1}{\left | z \right | ^{2}}-\frac{\cos \frac{\left | z
\right |}{2}}{2\sin \frac{\left | z \right |}{2}\left | z \right |}+ \frac{\cos\left | z \right | -1 } { 2\left | z \right | ^{2}}+ \frac{\left | z \right | -\sin\left | z \right | } { \left | z \right | ^{3}}- \frac{\left | z \right | -\sin\left | z \right | } { \left | z \right | } \frac{1}{\left | z \right | ^{2}}+\right . }
\\ \shoveleft { \left .
+ \frac{\left | z \right | -\sin\left | z \right | } { \left |
z \right | } \frac{\cos \frac{\left | z \right |}{2}}{2\sin \frac{\left | z \right |}{2}\left | z \right | } \right ] z^{2 } } \\ \shoveleft { = i_{3}+\left [ \frac{-\sin^{2 } \frac{\left | z \right
|}{2}}{\left | z \right |}\frac{\cos\frac{\left | z
\right |}{2}}{\sin \frac{\left | z \right |}{2}}+ \frac{\sin \left | z \right |}{2\left | z \right | } \right ] z + \left [ \frac{\cos ^{2}\frac{\left
| z \right |}{2}}{\left | z \right | ^{2}}- \frac{2\sin \frac{\left | z \right |}{2}\cos \frac{\left
| z\right |}{2}}{2\left | z \right |^{2 } } \frac{\cos \frac{\left | z \right |}{2}}{\sin \frac{\left | z \right |}{2}}\right ] z^{2}=i_{3}}.\end{gathered}\ ] ] the proof of ( [ e : auxadjoint ] ) follows by using proposition [ prop : geometrical_interp ] ( 3 ) : @xmath670 where we have used the taylor expansions for _ sine _ and _ cosine_. + we apply theorem [ thm : aux1 ] to get @xmath671 , for any @xmath617 $ ] and @xmath200 small , where @xmath275 is the sufficiently @xmath142 solution of ( [ e : odes_frequency ] ) .
we make the change of variable @xmath672 .
we get that the solution of the initial value problem ( [ e : auxhopf1 ] ) is given by @xmath673 where @xmath225 is the solution of the initial value problem @xmath674 using the above argument and corollary [ cor : aux2 ] , we have that @xmath675 for any @xmath676 $ ] and @xmath200 small , where @xmath677 is the sufficiently @xmath142 solution of ( [ e : odes_frequency1 ] ) for @xmath678 .
+ if we continue this , we get for @xmath200 small and any @xmath679 $ ] , @xmath680 where @xmath681 is given by @xmath682 for any @xmath683 $ ] and small @xmath200 for @xmath153 , @xmath45 , , @xmath272 , where @xmath291 is the solution of the initial value problem ( [ e : odes_frequency1 ] ) on the interval @xmath684 $ ] .
+ then @xmath685 $ ] is defined by formula ( [ e : zetaformula00 ] ) .
+ also , from ( [ e : zeta0 ] ) in theorem [ thm : aux1 ] we get @xmath686 for any @xmath687 $ ] and small @xmath200 for @xmath153 , @xmath45 , ,
+ because the @xmath31 formula in @xmath1 is smooth from @xmath688 into @xmath246 ( see theorem [ thm : bch_formula ] ) and @xmath689 for @xmath153 , @xmath45 ,
, @xmath272 are sufficiently @xmath142 , then @xmath685 $ ] is sufficiently @xmath142 on @xmath690 $ ] and @xmath200 small .
2 . on the interval
@xmath691 $ ] for any integer @xmath692 .
+ since @xmath693 it is easy to see that we can define for any @xmath694 $ ] and @xmath359 small , @xmath695=bch(z^{0}(t(\lambda),\lambda ) , z^{0}(t - t(\lambda ) , \lambda)).\ ] ] because the @xmath31 formula in @xmath1 is smooth from @xmath688 into @xmath246 , @xmath696 is sufficiently @xmath142 and @xmath542 is sufficiently smooth , then @xmath685 $ ] is sufficiently @xmath142 for any @xmath694 $ ] and for @xmath200 small .
+ we then repeat the above argument .
it is clear that the function @xmath697 $ ] is sufficiently @xmath142 , and @xmath698 for @xmath211 and @xmath200 small .
let us define @xmath699)$ ] .
therefore , @xmath275 is sufficiently @xmath142 and @xmath700}= e^{z_{1}(t,\lambda)}$ ] .
+ thus , @xmath275 is the solution of the initial value problem ( [ e : odes_frequency_final ] ) . if we define @xmath701 then @xmath702 .
+ since @xmath703 is sufficiently @xmath142 for @xmath436 and @xmath200 small and @xmath704 for @xmath200 is sufficiently smooth , we get that @xmath242 is sufficiently @xmath142 for @xmath200 small .
the branch @xmath242 defined by ( [ e : freqfreq ] ) is such that @xmath705 for @xmath200 and it is the branch whose existence is stated in theorem [ thm : frequencygeneral1 ] . + using proposition [ prop : frequency1 ]
we know that @xmath242 is sufficiently @xmath142 for @xmath706 small . therefore , @xmath707 for @xmath200 small is sufficiently @xmath142 . using taylor formula for @xmath708 around @xmath709 up to the term of order @xmath623 , and then taking @xmath335
, we get @xmath710 for @xmath200 small . also , @xmath711 is continuous for @xmath200 .
+ the branch @xmath256 is constructed below .
let @xmath712 , where @xmath713 , @xmath20 , @xmath714 , and @xmath715 .
+ 1 . if @xmath259 for all @xmath20 , then @xmath716 .
+ since @xmath717 , we get that + @xmath718 . therefore , @xmath719 . since @xmath242 is sufficiently @xmath142 , then @xmath720 for @xmath200 small .
+ then @xmath721 is well - defined for @xmath200 small .
+ we check that @xmath255 .
using @xmath722 , it yields @xmath723 .
+ since @xmath720 and @xmath242 , @xmath232 are sufficiently @xmath142 , then @xmath256 is sufficiently @xmath142 .
+ using the taylor formula for @xmath724 around @xmath709 up to the term of order @xmath623 and then taking @xmath725 , we get @xmath726 for @xmath200 small , thus we get the formula ( [ e : freqtaylor01 ] ) . +
we have only modulated rotating waves because @xmath727 for any @xmath20 implies @xmath728 for any @xmath20 and for @xmath200 small by the continuity of @xmath729 .
+ 2 . if @xmath299 for some @xmath117 , @xmath263 , then @xmath730 .
+ let @xmath255 .
+ if @xmath300 , then we define @xmath301 for @xmath18 small .
+ if @xmath302 , then we define @xmath303 , where @xmath304 , @xmath305 for @xmath18 small .
+ if @xmath318 for some @xmath20 , @xmath263 , then we check that @xmath731 for all @xmath200 small .
for @xmath18 small such that @xmath732 , we have @xmath733 .
+ for @xmath18 small such that @xmath734 , we have @xmath735 .
+ hence , for @xmath200 small , we get @xmath736 .
+ since @xmath737 and @xmath738 , we get @xmath739 ) and then @xmath740 for @xmath200 small , thus we get the formula ( [ e : freqtaylor02 ] ) . + also , @xmath264 is continuous since @xmath242 and @xmath542 are continuous for @xmath200 small . by using proposition [ prop : properties_exp ] ( ( 6 ) and ( 7 ) ) , we have @xmath741 for @xmath200 small . 1 .
suppose that @xmath259 for any @xmath20 .
+ we have @xmath742 for @xmath89 and @xmath200 small .
+ since @xmath256 and @xmath286 are sufficiently @xmath142 for @xmath211 and @xmath200 small and the @xmath31 formula in @xmath1 is smooth , it follows that @xmath743 is sufficiently @xmath142 for @xmath89 and @xmath200 small .
+ since @xmath744 , it follows that @xmath745 . then , @xmath746 for all @xmath465 and @xmath200 small .
the periodicity of @xmath747 results from the following remark : + if @xmath748 , @xmath749 is a continuous @xmath75-periodic function in @xmath217 for @xmath200 small and @xmath750 is a continuous function such that @xmath751 , then @xmath437 is a @xmath75-periodic function in @xmath217 for @xmath752 small . + we have @xmath753 and @xmath754 , @xmath755 are in a neighborhood of @xmath756 for @xmath752 small .
then , by using the fact that the exponential map @xmath757 is a local diffeomorphism at @xmath756 , it follows that @xmath758 or @xmath759 for @xmath200 small . + writing the taylor formula for @xmath760 at @xmath709 up to order of @xmath623 and taking @xmath761 , we get @xmath762 for @xmath89 and @xmath200 small .
suppose that @xmath318 for some @xmath20 , @xmath263 .
+ we have @xmath763 for @xmath89 and @xmath200 small .
since @xmath242 and @xmath286 are sufficiently @xmath142 for @xmath89 and @xmath200 small , and @xmath31 formula in @xmath1 is smooth , it is clear that @xmath764 is sufficiently @xmath142 for @xmath89 and @xmath200 small .
+ since @xmath765 , we get @xmath766 . since @xmath767 and @xmath768 , it yields that @xmath769 .
we denote @xmath770 .
therefore , @xmath771 for @xmath89 and @xmath200 small .
also , it is clear that @xmath772 . since @xmath225 is @xmath773-periodic in @xmath217 and @xmath774
, it follows that @xmath775 for @xmath200 small . since @xmath776 , the conclusion of theorem [ thm : mrw ] result by applying theorems [ prop : frequency1 ] , [ cor : frequency2 ] and [ thm : periodic_part ] .
let us define @xmath777 .
+ let us denote @xmath622 , @xmath778 and @xmath779 .
let @xmath780 .
+ we have to prove that there exist sufficiently smooth curves @xmath781 satisfying @xmath782 let us define the function @xmath783 by @xmath784 we will prove the existence of the smooth curves using the implicit function theorem for the function @xmath785 at the point @xmath786 .
+ we now prove that : 1 .
@xmath787 + and 2 .
@xmath788_{x_{0}^{1},x_{1},x_{2}}$ ] has rank 3 .
we have that @xmath789 since @xmath790 .
+ recall that we have @xmath791 then @xmath792 , where @xmath793 since @xmath794 , then @xmath795 .
also @xmath796 . using proposition [ prop : properties_exp ] ( 2 ) , we get @xmath797.\end{gathered}\ ] ] we have @xmath798 , @xmath799 and @xmath800\\&+ \left
| x_{0 } \right | \left [ -2\pi \cdot sgn(\omega_{bif})\right ] \frac{(\omega_{0,\mu})^{'}|_{\mu= 0}}{\omega_{bif}^{2}}x_{0}^{1}\\ & = \frac{2\pi}{\left | \omega_{bif } \right | } \left [ ( a^{'}(0)-k\cdot ( sgn(\omega_{bif}))^{2}(\omega_{0,\mu } ) ^{'}|_{\mu= 0 } ) x_{0}^{1}+b^{'}(0 ) x_{1}+c^{'}(0 ) x_{2 } \right ] , \end{split}\ ] ] where we have used @xmath801 . therefore , ( [ e : eq22 ] ) becomes @xmath802\right ) \\ & = \frac{2\pi}{\left | \omega_{bif } \right | } \left [ ( a^{'}(0)- k ( \omega_{0,\mu})^{'}|_{\mu= 0})x_{0}^{1}+b^{'}(0)x_{1}+c^{'}(0)x_{2}\right ] \\ & + \sum_{n= 1}^{\infty}{\frac{(-1)^{n}}{(n+1 ) !
} \left ( ad(2| k |\pi x_{0}^{1})\right ) ^{n}\left ( \frac{2\pi}{\left | \omega_{bif}\right | } \left [ ( a^{'}(0)- k ( \omega_{0,\mu})^{'}|_{\mu= 0})x_{0}^{1}+b^{'}(0)x_{1}+ c^{'}(0)x_{2}\right ] \right ) } \\ & = \frac{2\pi}{\left | \omega_{bif } \right | } \left [ ( a^{'}(0)- k ( \omega_{0,\mu})^{'}|_{\mu= 0})x_{0}^{1}+ ( b^{'}(0)+ * ) x_{1}+(c^{'}(0)+*)x_{2}\right ] , \end{split}\ ] ] where we have used @xmath803 and where @xmath804 denotes other terms .
these term appear only in the coefficients of @xmath805 and @xmath806 since @xmath807\right ) \\ & = \frac{2\pi}{\left | \omega_{bif } \right | } \left [ 2| k| \pi x_{0}^{1 } , ( a^{'}(0)-k(\omega_{0,\mu})^{'}|_{\mu= 0 } ) x_{0}^{1 } + b^{'}(0)x_{1}+c^{'}(0)x_{2}\right ] \\ & = \frac{4| k |\pi^{2}}{\left | \omega_{bif}\right |}(-c^{'}(0)x_{1}+ b^{'}(0)x_{2 } ) , \end{split}\ ] ] and for any integer @xmath808 , @xmath809\right ) \\= \left [ 2| k |\pi x_{0}^{1},\left ( ad(2| k |\pi x_{0}^{1})\right ) ^{n-1}\left ( \frac{2\pi}{\left | \omega_{bif}\right | } \left [ ( a^{'}(0)- k ( \omega_{0,\mu})^{'}|_{\mu= 0})x_{0}^{1}+b^{'}(0)x_{1}+c^{'}(0)x_{2}\right ] \right ) \right ] \end{gathered}\ ] ] will contain only linear combinations of @xmath805 and @xmath806 . + then with respect to the basis @xmath238 , + @xmath810 where @xmath804 denotes other terms ( below we show that in fact @xmath811 and @xmath812 ) .
+ let @xmath813\right ) $ ] . using proposition [ prop : geometrical_interp ] ( 1 )
, we get @xmath814\right ) } \\ & = \frac{2\pi}{\left | \omega_{bif}\right |}\sum_{n= 0}^{\infty } \frac{(-1)^{n}}{(n+1 ) ! } ( 2\pi |k| x_{0}^{1})^{n } \left [ ( a^{'}(0)-k(\omega_{0,\mu})^{'}|_{\mu= 0})\overrightarrow x_{0}^{1}+b^{'}(0)\overrightarrow x_{1}+ c^{'}(0 ) \overrightarrow x_{2}\right ] . \end{split}\ ] ] if we apply the relation ( [ e : auxadjoint ] ) for @xmath815 , then taking into account that @xmath816 , we get @xmath817 if we substitute ( [ e : mike2 ] ) into ( [ e : mike1 ] ) , it follows that @xmath818 \left [ ( a^{'}(0)- k ( \omega_{0,\mu})^{'}|_{\mu= 0})\overrightarrow x_{0}^{1}+ b^{'}(0)\overrightarrow x_{1}+c^{'}(0)\overrightarrow x_{2}\right ] \\ & = \frac{2\pi}{\left | \omega_{bif}\right |}\left [ ( a^{'}(0)- k(\omega_{0,\mu})^{'}|_{\mu= 0 } ) \overrightarrow x_{0}^{1}+b^{'}(0)\overrightarrow x_{1}+ c^{'}(0)\overrightarrow x_{2}\right . \\
& \left .
+ ( a^{'}(0)- k(\omega_{0,\mu})^{'}|_{\mu= 0})(x_{0}^{1})^{2}\overrightarrow x_{0}^{1}+ b^{'}(0)(x_{0}^{1})^{2 } \overrightarrow x_{1}+ c^{'}(0)(x_{0}^{1})^{2 } \overrightarrow x_{2}\right ] . \end{split}\ ] ] since @xmath819 , @xmath820 and @xmath821 , we get @xmath822 .
+ in order to have the rank of @xmath823 equal to 3 , it is necessary and sufficient that @xmath824 we note that depending on the function @xmath825 we can establish if @xmath826 and/or @xmath827
. + _ * another proof of theorem [ thm : resonant_drift1 ] * _ we have that @xmath828 , where @xmath829 is defined as in proposition [ prop : frequency1 ] . + let @xmath830 . then , since @xmath829 is sufficiently @xmath142 , we get that @xmath831 , @xmath832 , @xmath833 are sufficiently @xmath142 and @xmath834 .
+ since @xmath835 , then @xmath836 , that is @xmath837 .
+ we want to find a sufficiently @xmath142 branch @xmath373 such that @xmath429 and @xmath838 , for @xmath200 small .
if @xmath839 , then by applying the implicit function theorem we get the existence of the required sufficiently @xmath142 branch .
+ now we show that @xmath839 is equivalent with @xmath840 .
+ we have that @xmath841 for @xmath842 small , then we get @xmath843 for some @xmath844 for @xmath842 small .
+ thus , @xmath845t(0,\mu)= \\ \left [ a(\mu)x_{0}^{1}+b(\mu)x_{1}+c(\mu)x_{2}\right ] t(0,\mu)+ \frac{2l(\mu)\pi}{\sqrt{a(\mu)^{2}+o(\mu^{2})}}\left [ a(\mu)x_{0}^{1}+ b(\mu)x_{1}+c(\mu)x_{2}\right ] $ ] for @xmath846 small . +
this implies that @xmath847 which for @xmath848 gives @xmath849 , where we have used the fact that @xmath850 implies @xmath851 , and thus , @xmath852 . since @xmath853 is continuous , we get @xmath854 for @xmath846 small . + then by taking into account that @xmath855 , ( [ e : eqauxres ] ) becomes @xmath856 by differentiation of ( [ e : eqauxres1 ] ) , it follows that @xmath857 .
+ + the conclusions of corollary [ cor : resonant_drift2 ] result in the same way as in the proof of theorems [ prop : frequency1 ] and [ thm : periodic_part ] , except the scaling of the primary frequencies @xmath858 that results from remark 3.5 in @xcite .
it is known that @xmath0 is diffeomorphic as a manifold to the real projective space @xmath859 ( see @xcite ) . if @xmath860 , then we define @xmath861 . also , we define @xmath862 . a model for the space @xmath859 is the set @xmath863 .
in fact , d is the quotient set @xmath864 , where @xmath865 is the equivalence relation @xmath866 and @xmath867 .
the set @xmath246 is considered with the quotient topology . sometimes , instead of @xmath868 , @xmath869 , we use @xmath870 , in which case we denote the equivalence class @xmath871=[\overrightarrow y]$ ] .
the projection map @xmath872 , @xmath873 $ ] is smooth .
there exists a unique smooth function @xmath874 such that @xmath875 .
[ def : bch1 ] for any @xmath4 , @xmath876 , we define @xmath877 , where @xmath878 is defined above .
clearly , we have @xmath879 and @xmath880 , for any @xmath4 , @xmath876 .
[ thm : bch_formula ] the @xmath31 formula in @xmath1 has the form @xmath881 \mbox { for } x , \ , y \in so(3),\ ] ] where @xmath882 and @xmath883 @xmath884 @xmath885 @xmath886 @xmath887 @xmath888 where @xmath889 is the angle between the two vectors @xmath890 and @xmath891 , @xmath892 @xmath893 @xmath894 @xmath895 where @xmath896 for any @xmath897 , such that @xmath898 or @xmath899 . the functions @xmath900 , @xmath901 , @xmath902 are smooth on @xmath688 and @xmath903 , @xmath904 , @xmath905 are continuous on @xmath688 . also , the function @xmath31 is smooth from @xmath688 into @xmath246 .
[ prop : geometrical_interp ] 1 .
@xcite for any @xmath906 , @xmath876 , @xmath907 2 . for any @xmath103 and any @xmath512 , we have 1 .
@xmath908 2 .
@xcite @xmath909 3 .
@xcite for any @xmath512 , we have : @xmath910 the properties of the exponential map @xmath32 are : [ prop : properties_exp ] 1 .
@xcite the exponential map @xmath511 is surjective .
@xcite the exponential map @xmath757 is a smooth function on @xmath1 and its differential is given by : @xmath911 or @xmath912 ( we will use the first formula later ) .
moreover , it is a local diffeomorphism near any @xmath512 if and only if the operator @xmath913 has no eigenvalues of the form @xmath914 with @xmath915 , that is if and only if @xmath916 for @xmath20 , @xmath263 .
@xcite @xmath917 if and only if @xmath918 , where @xmath512 and @xmath919 is a @xmath920 function .
@xcite the rodrigues formula holds : @xmath921 .
5 . @xcite the exponential map @xmath757 maps any @xmath922 , @xmath923 to the right - handed rotation @xmath103 with angle @xmath924 around @xmath890 .
@xcite @xmath925 if and only if @xmath926 for some @xmath20
@xcite @xmath927 if and only if either @xmath928 or @xmath929 for some @xmath20 .
we consider the reaction - diffusion system ( [ e : rdparam ] ) .
[ thm : cmr_reparam ] let @xmath87 be a relative equilibrium that is not an equilibrium for ( [ e : rdparam ] ) at @xmath14 and such that the stabilizer of @xmath12 is @xmath97 .
let @xmath157 be the linearization of the right - hand side of ( [ e : rdparam ] ) with respect to the rotating wave @xmath156 at @xmath334 in the co - rotating frame , that is @xmath158 suppose that : 1 .
@xmath159 is a spectral set with spectral projection @xmath160 , and dim@xmath930 2 .
the semigroup @xmath162 satisfies @xmath931 for some @xmath932 and @xmath933 .
then there exists a sufficiently smooth parameter - dependent center manifold @xmath449 for the relative equilibrium @xmath12 .
let @xmath176 be the orthogonal complement of @xmath177 in @xmath934 .
+ then the center manifold @xmath449 is diffeomorphic to @xmath169 for @xmath193 small .
furthermore , there exist sufficiently smooth functions @xmath166 and @xmath167 such that any solution of @xmath935 on @xmath169 corresponds to a solution of the reaction - diffusion system ( [ e : rdparam ] ) on @xmath170 under the diffeomorphic identification @xmath936 , with @xmath546 a local chart from @xmath937 to @xmath137 , for @xmath171 small .
also , @xmath172 , @xmath173 and @xmath174 , where @xmath175 is the projection onto @xmath176 along @xmath177 .
theorems [ thm : cmr_reparam ] is valid for more than one parameter .
we present the proof of ( [ e : numerical1 ] ) : + if we parameterize @xmath0 by euler angles , that is @xmath938 , where @xmath454 , @xmath455 $ ] , @xmath456 , and we substitute this into the equation @xmath939 , we get @xmath940\end{gathered}\ ] ] or @xmath941 by computation , it follows that @xmath942 if we substitute ( [ e : auxb ] ) into ( [ e : auxa ] ) , we get @xmath943 if we identify the corresponding coefficients of @xmath944 , @xmath945 and @xmath946 from both sides of the equation ( [ e : auxc ] ) , we get the system @xmath947 then , by multiplying the second equation in ( [ e : auxd ] ) by @xmath948 and the third equation in ( [ e : auxd ] ) by @xmath949 and adding them , we get @xmath950 . similarly , by multiplying the second equation in ( [ e : auxd ] ) by @xmath949 and the third equation in ( [ e : auxd ] ) by @xmath951 and adding them , we get @xmath952 $ ] .
if we substitute @xmath953 given by the above relation into the first equation of the system ( [ e : auxd ] ) , we get @xmath954 $ ] .
i wish to thank you to my ph.d . supervisor , victor g. leblanc , for his help to write this article .
i am also grateful to the natural sciences and engineering research council of canada for a pgs - b scholarship .
d. barkley , linear stability analysis of rotating spiral waves in excitable media , phys .
68 ( 13 ) ( 1992 ) 2090 - 2093 .
d. barkley , euclidean symmetry and the dynamics of rotating waves , phys .
72 ( 1994 ) 164 - 167 .
t. brcker , t. tom dieck , representations of compact lie groups , springer - verlag , new york , 1985 .
comanici , spiral waves on spherical domains : a dynamical systems approach , ph.d .
thesis , university of ottawa , canada , 2004 .
j. davidsen , l. glass , r. kapral , topological constraints on spiral wave dynamics in spherical shells with inhomogeneous excitability , phy .
e 70 ( 2004 ) 056203 ( 6 pages ) .
davidenko , a.v .
pertsov , r. salomonsz , w. baxter , j. jalife , stationary and drifting spiral waves of excitation in isolated cardiac muscle , nature 355 ( 1992 ) 349 - 351 .
h. dym , h.p .
mckean , fourier series and integrals , academic press , new york / london , 1972 .
k. engo , on the bch - formula in @xmath1 , bit numerical mathematics 41 ( 2001 ) 629 - 632 .
b. fiedler , b. sandstede , a. scheel , c. wulff , bifurcation from relative equilibria of noncompact group actions : skew products , meanders , and drifts , doc .
math . 1 ( 1996 ) 479 - 505 .
b. fiedler , d. turaev , normal forms , resonances , and meandering tip positions near relative equilibria of euclidean group actions , arch .
rational mech .
analysis 145 ( 1998 ) 129 - 159 .
m. field , local structure of equivariant dynamics , in : singularity theory and its applications warwick part 1989 ii ( m. roberts , i. stewart eds . ) , lecture notes in mathematics , vol .
1463 , springer - verlag , heidelberg , 1991 , pp .
143 - 166 .
a. friedman , partial differential equations , holt , rinehart and winston , new york , 1969 .
g. gerisch , naturwissenschaften 58 ( 1983 ) 420 .
m. golubitsky , v.g .
leblanc , i. melbourne , meandering of the spiral tip : an alternative approach , j. nonlinear sci . 7
( 1997 ) 557 - 586 .
m. golubitsky , v.g .
leblanc , i. melbourne , hopf bifurcation from rotating waves and patterns in physical space , j. nonlinear sci .
10 ( 2000 ) 69 - 101 .
m. golubitsky , d.g .
schaeffer , singularities and groups in bifurcation theory , vol .
i , springer - verlag , berlin , 1985 .
m. golubitsky , i. stewart , d.g .
schaeffer , singularities and groups in bifurcation theory , vol .
ii , springer - verlag , berlin , 1988 .
j. gomatam , f. amdjadi , reaction - diffusion on a sphere : meandering of spiral waves , physical review e 56 ( 4 ) ( 1997 ) 3913 - 3919 .
j. gomatam , d.a .
hodson , the eikonal equation : stability of reaction - diffusion waves on a sphere , physica d 49 ( 1991 ) 82 - 89 .
p. grindrod , j. gomatam , the geometry and motion of reaction - diffusion waves on closed two - dimensional manifolds , j. math .
biol . 25 ( 1987 ) 597 - 610 .
hale , ordinary differential equations , wiley , new york , 1969 .
e. hebey , sobolev spaces on riemannian manifolds , lectures notes in mathematics , vol .
1635 , springer - verlag , berlin , 1996 .
d. henry , geometric theory of semilinear parabolic equations , lecture notes in mathematics , vol .
840 , springer - verlag , berlin , 1981 .
hirsch , s. smale , differential equations , dynamical systems , and linear algebra , acamedic press , san diego , 1974 .
w. jahnke , w.e .
skaggs , a.t .
winfree , chemical vortex dynamics in the belousov - zhabotinsky reaction and in the two - variable oregonator model , j. chem .
93 ( 1989 ) 740 - 749 .
r. kapral , k. showalter , chemical waves and patterns , kluwer academic publishers , netherlands , 1995 .
j. keener , j. sneyd , mathematical physiology , vol .
8 , interdisciplinary applied mathematics , springer - verlag , new york , 1998 .
m. krupa , bifurcations of relative equilibria , siam j. math .
21 ( 6 ) ( 1990 ) 1453 - 1486 .
g. li , q. ouyang , v. petrov , h.l .
swinney , transition from simple rotating chemical spirals to meandering and traveling spirals , phys .
77 ( 1996 ) 2105 - 2108 .
lions , e. magenes , non - homogeneous boundary value problems and applications , springer - verlag , berlin , new york , 1972 .
marsden , t.s .
ratiu , introduction to mechanics and symmetry , springer - verlag , new york , 1999 .
j. maselko , symmetrical double rotor spiral waves on spherical surfaces , j. chem .
, faraday trans .
94 ( 16 ) ( 1998 )
2343 - 2345 . j. maselko , k. showalter , chemical waves on spherical surfaces , letters to nature 339 ( 1989 ) 609 - 611 .
p. mcquillan , j. gomatam , rotating chemical waves on the sphere , j. physical chemistry 100 ( 13 ) ( 1996 ) 5157 - 5159 .
s. nettesheim , a. von oertzen , h.h .
rotermund , g. ertl , reaction diffusion patterns in the catalytic co - oxidation on pt(110)-front propagation and spiral waves , j. chem .
98 ( 1993 ) 9977 - 9985 .
onischchik , lie groups and lie algebras , vol .
i , springer - verlag , berlin / heidelberg , 1993 .
j. palis , w. de melo , geometric theory of dynamical systems , an introduction , springer - verlag , new york , 1989 .
a. pazy , semigroups of linear operators and applications to partial differential equations , springer - verlag , new york , 1983 .
plapp , e. bodenshatz , core dynamics of multi - armed spirals in rayleigh - benard convection , physica scripta t67 ( 1996 ) 111 - 116 .
d. rand , dynamics and symmetry .
predictions for modulated waves in rotating fluids , arch .
. mech . and anal . 79 ( 1982 ) 1 - 38 .
m. renardy , bifurcation from rotating waves , arch .
mech . and anal . 79 ( 1982 ) 49 - 84 .
w. rossmann , lie groups , an introduction through linear groups , oxford university press , oxford , new york , 2002 .
a. scheel , bifurcation to spiral waves in reaction - diffusion systems , siam j. math .
29 ( 6 ) ( 1998 ) 1399 - 1415 .
selig , geometrical methods in robotics , springer - verlag , new york , 1996 .
b. sandstede , a. scheel , c. wulff , dynamics of spiral waves on unbounded domains using center - manifold reductions , journal of differential equations 141 ( 1997 ) 122 - 149 .
m. sugiura , unitary representations and harmonic analysis , an introduction , north - holland , amsterdam , 1990 .
m. taylor , partial differential equations , vol .
i - iii , springer - verlag , new york , 1996 .
a. terras , harmonic analysis on symmetric spaces and applications i , springer - verlag , new york , 1985 .
varadarajan , lie groups , lie algebras , and their representations , prentice - hall , 1974 .
varadarajan , an introduction to harmonic analysis on semi - simple lie groups , vol .
16 , cambridge studies in advanced mathematics , cambridge university press , cambridge , 1989 .
s. wiggings , introduction to applied non - linear dynamical systems and chaos , springer - verlag , new york , 1990 . c. wulff ,
theory of meandering and drifting spiral waves in reaction - diffusion systems , ph.d .
thesis , freie universitt , berlin , 1996 . c. wulff , transitions from relative equilibria to relative periodic orbits , documeta math . 5 ( 2000 ) 227 - 274 .
h. yagisita , m. mimura , m. yamada , spiral wave behaviors in an excitable reaction - diffusion system on a sphere , physica d 124 ( 1998 ) 126 - 136 .
zykov , s.c .
mller , spiral waves on circular and spherical domains of excitable medium , physica d 97 ( 1996 ) 322 - 332 .
zykov , a.s .
mikailov , s.c .
mller , controlling spiral waves in confined geometries by global feedback , physical review letters 78 ( 17 ) ( 1997 ) 3398 - 3401 . | in this article , we study parameter - dependent systems of reaction - diffusion equations on the sphere , which are equivariant under the group @xmath0 of all rigid rotations on a sphere .
two main types of spatial - temporal patterns that can appear in such systems are rotating waves ( equilibrium in a co - rotating frame ) and modulated rotating waves ( periodic solution in a co - rotating frame ) . the transition from rotating waves to modulated rotating waves on spherical domains
is explained via a supercritical hopf bifurcation from a rotating wave , @xmath0-symmetry and finite - dimensional equivariant center manifold reduction .
the baker - campbell - haussdorff formula in the lie algebra @xmath1 is used to get reduced differential equations on @xmath1 , a formula for a primary frequency vector , as well as a formula for the periodic part associated to any modulated rotating wave obtained by a supercritical hopf bifurcation from a rotating wave . as a consequence , there are three types of motions for the tips of these modulated rotating waves . in the resonant case ( with two parameters ) , we obtain that the primary frequency vectors of a branch of these modulated rotating waves are generically orthogonal to the frequency vector of the initial rotating wave undergoing hopf bifurcation . + _ _ equivariant center manifold , rotating wave , modulated rotating wave , hopf bifurcation |
growing clusters are ubiquitous in nature and they exhibit fascinating structures and patterns .
examples range from natural fractals , such as snowflakes and soots , to artificial structures such as networks , for example the internet and social networks .
various growth models have been studied extensively by physicists over the last three decades @xcite . in these models growth
starts from a single seed site and proceeds via absorbing new particles into the cluster according to certain specified rules .
different growth rules give rise to different growth models , examples being the eden model @xcite , invasion percolation @xcite , diffusion limited aggregation @xcite and the growing network models @xcite which have recently received much attention .
there are two reasons why many of these growth models are often studied on a cayley tree ( or on the bethe lattice ) @xcite .
first , the tree structure of the bethe lattice mimics a euclidean lattice in the limit of high dimensions where the mean field theory often becomes exact .
secondly , the absence of loops on the cayley tree often allows one to obtain exact analytical solutions which are very difficult to obtain on a regular @xmath10-dimensional lattice .
there is yet another compelling motivation for studying these growth models on a cayley tree and this comes from computer science . `
storing and search ' of data is a very important area of computer science @xcite .
incoming data to a computer is usually stored on a cayley tree by using various data storage algorithms and the tree so grown is called a ` search tree ' @xcite .
different algorithms lead to different search trees and in some cases , as explained below , the rules of growth of a search tree can be shown to be exactly equivalent to a ` physical ' growth model on the tree .
thus the study of these physical growth models on a cayley tree provides important insights into data storage in computer science . as a first example of this equivalence between a physical growth model and a search tree ,
we show here that the eden model on a binary cayley tree is exactly equivalent to the random binary search tree ( rbst ) .
consider the eden model on a binary cayley tree where the growth starts from the root @xcite . at the first step ,
a particle gets absorbed at the root , thus forming a cluster of size @xmath11 .
this cluster has now two empty neighbors which defines the perimeter of the cluster . at the next step ,
a new particle will get absorbed at any of these two perimeter sites chosen with equal probability , thus forming a cluster of size @xmath12 .
the subsequent growth occurs following the same rule , namely a new particle gets absorbed at any of the perimeter sites chosen with equal probability . in fig .
( 1 ) , we show a cluster after @xmath13 steps where the black sites denote the cluster and the shaded sites denote the current perimeter sites that are available for subsequent growth . an eden cluster of size @xmath13 on a cayley tree .
the black sites form the cluster and the shaded sites form the perimeter . at the next step
, growth can occur at any of the @xmath14 shaded perimeter sites with equal probability @xmath15 .
, width=384 ] fig .
( 2 ) shows all possible eden clusters of size @xmath16 and their associated statistical weights .
all possible eden clusters of size @xmath16 on a tree and their associated statistical weights w.,width=384 ] on the other hand , a binary search tree in computer science is constructed by the following simple algorithm @xcite .
imagine that we have a data string consisting of @xmath17 items which are labeled by the @xmath17 integers : @xmath18
. these could be the months of the year or the names of people etc .
let us assume that this data appears in a particular order , say @xmath19 for @xmath20 integers .
this data is first stored on a binary tree following the simple dynamical rule : the first item @xmath21 is stored at the root of the tree ( see fig .
( [ fig : bst ] ) ) .
the next item in the string is @xmath13 .
we compare it with @xmath21 at the root and since @xmath22 , we store @xmath13 in the left daughter node of the root .
had it been bigger than the root item @xmath21 , we would have stored it in the right daughter node .
the next item in the string is @xmath14 .
we again start from the root , see that @xmath23 , so we go to the left branch . there
we encounter @xmath13 and we find @xmath24 , so we go the right daughter node of @xmath13 .
this process is continued till all the @xmath20 items are assigned their nodes and we get a unique binary search tree ( bst ) ( see fig .
( [ fig : bst ] ) ) for this particular data string @xmath19 .
the binary search tree associated with the data string @xmath19 .
, width=384 ] usually the data arrives at a computer in random order . to study this situation ,
one considers the simplest model called the ` random binary search tree ' ( rbst ) model where one assumes that the incoming data string can arrive in any of the @xmath25 possible orders or sequences , each with equal probability @xcite .
for each of these sequences , one has a binary tree .
for example , in fig .
( 4 ) , we show the binary trees for @xmath26 along with their associated probabilities .
all possible random binary search trees for a data of size @xmath26 and their associated statistical weights.,width=384 ] comparing fig .
( 2 ) and fig . ( 4 ) , one sees immediately that the eden trees after 3 steps have exactly the same configurations and statistical weights as the random binary search trees with data size @xmath26 .
this analogy can be easily extended to all @xmath17 .
the key point is that after @xmath27 steps there are @xmath27 occupied sites in the eden cluster and @xmath28 perimeter sites ( this is easy to understand as the addition of a new occupied site eliminates one old perimeter site while creating two new perimeter sites ) .
the probability of subsequent growth at step @xmath28 at any of these perimeter sites is @xmath29 .
thus the statistical weight of a cluster of @xmath17 sites formed by a specific history of growth is simply @xmath30 , which is the same as in the rbst model .
thus the eden model on the cayley tree is exactly equivalent to the rbst .
another popular search tree model is known as the ` digital search tree ' ( dst ) which is constructed by the following rule @xcite .
consider again a binary cayley tree each node of which can contain at most one entry .
one starts with an empty tree and the data is stored sequentially .
the first data item is stored in the root of the tree .
the next one arrives at the root and finding it occupied , moves to any of the two empty daughter nodes chosen at random and occupies that node .
then the next item arrives and again it starts at the node , chooses any of its two daughters randomly and moves there .
if the chosen daughter is empty it occupies it .
if the chosen daughter is already occupied , it again chooses one of its two descendants at random and moves there .
thus at any stage , a new entry starts at the root and performs a random walk ( to the left or to the right daughter with equal probability ) down the tree till it finds an empty node and occupies the node .
thus one obtains again a growing tree where at any stage growth can occur at any of the perimeter sites , but now the growth probability at a perimeter site @xmath31 is @xmath32 where @xmath33 is the distance of the perimeter site from the root .
the dst is an important tree structure in computer science and has been studied extensively . in particular , it turns out the dst is a natural tree representation @xcite of the data compression algorithm due to ziv and lempel @xcite .
recently it was shown that a diffusion limited aggregation model introduced by bradley and strenski @xcite in physics is exactly equivalent the the dst model in computer science and a variety of exact results were obtained by exploiting this connection @xcite .
the examples above illustrate a profound link between growth models and the dynamics of search tree formation in computer science .
note that the two search tree models discussed above , the rbst and the dst , can be considered as special cases of a general growth model where growth occurs ( i.e. a new particle gets absorbed ) at any of the available perimeter sites @xmath31 with a growth probability @xmath34 where @xmath1 is a constant positive parameter and @xmath33 is the distance of the perimeter site @xmath31 from the root of the tree .
the rbst ( equivalently the eden model ) corresponds to @xmath3 so that all perimeter sites have equal probability to absorb a particle .
the dst , on the other hand , corresponds to @xmath4 as discussed above .
it is then useful and interesting to study this general growth model parametrized by @xmath1 and ask if there are any qualitative changes in the statistical properties of the growth clusters as one varies the parameter @xmath1 continuously . indeed ,
aldous and shields studied a continuous - time version of this generalized growth model @xcite .
note that in the two models discussed above time is _ discrete _ and is equal to the number of particles in the tree .
in the version of the model studied by aldous and shields , time is considered _ continuous _ and growth occurs at any of the available perimeter sites
say the site @xmath31 with a _ rate _ proportional to @xmath35 where @xmath1 is a positive parameter .
in this continuous - time model , the total number of particles in the tree at time @xmath36 is thus a random variable , unlike in the discrete time version .
thus , while the discrete - time model has a constant particle number ensemble , the continuous - time model has a constant time ensemble , much like the canonical and the grand canonical ensemble in statistical physics .
asymptotically at long times , both the discrete - time and the continuous - time versions of the model are expected behave in a similar fashion .
henceforth , we will consider in this paper only the continuous - time version la aldous - shields , since it is , from a technical point of view , easier to study than its discrete - time counterpart .
the question naturally arises whether the statistical properties of the growing clusters in this model undergo any qualitative change of behavior as one tunes the parameter @xmath1 continuously .
indeed , aldous and shields established rigorous probabilistic bounds to show that the nature of the fluctuations ( variance ) in the number of particles in the tree at time @xmath36 is qualitatively different for @xmath7 and @xmath6 . while for @xmath6 the central limit theorem holds and the total number of particles has a limiting gaussian distribution @xcite , for @xmath7 the central limit theorem breaks down due to the appearance of anomalously large fluctuations
thus , there is a sharp _ phase transition _ in the nature of the fluctuations at a critical value @xmath5 .
however , the mechanism responsible for this phase transition and even the explicit quantitative behavior of the fluctuations above , below , or at the critical point were not easy to obtain within the rigorous probabilistic analysis of aldous and shields .
the principal purpose of this paper is to provide a detailed quantitative understanding of this rather ` peculiar ' phase transition .
our method , completely different from the original approach of aldous and shields , employs a backward fokker - planck formalism .
the advantage of this method is that one can obtain exact asymptotic results explicitly .
moreover , our analysis also shows that the mathematical mechanism behind this phase transition is similar to the phase transitions found recently in the variance of the number of nodes needed to store data on a @xmath8-ary search tree ( where @xmath8 is the number of branches ) at the critical value @xmath37 @xcite and also in the variance of the number of splitting events in a @xmath38-dimensional fragmentation model at the critical value @xmath39
@xcite .
the layout of the paper is as follows . in the next section ( ii ) , we define the model precisely and summarize the main results .
we study here a generalized aldous - shields model where the growth takes place on a cayley tree with @xmath8 branches . in section iii
, we derive the evolution equations for the mean and variance of the number of occupied sites as a function of time via a backward fokker planck technique .
a simple scaling analysis is then carried out to determine the temporal growth exponents . in section
iv a more thorough analysis of the evolution equations is provided that enables us to obtain explicitly not just the growth exponents , but also exact expressions for various amplitudes and prefactors that include interesting log - periodic oscillations .
we conclude with a summary and a discussion of open questions in the last section .
we consider a generalized aldous - shields model where growth occurs on a cayley tree ( rooted at @xmath40 ) with @xmath8 branches ( see fig . [
fig : mary ] ) .
aldous and shields studied @xcite only the binary case @xmath41 .
initially the tree is empty and growth occurs in continuous time starting from the root @xmath40 . at any instant @xmath36 ,
one first identifies the available perimeter nodes .
a node @xmath31 at time @xmath36 is a perimeter node if it is empty at @xmath36 but its parent node is occupied at @xmath36 ( see fig .
[ fig : mary ] ) .
subsequently , in a small time interval @xmath42 , a perimeter node @xmath31 either absorbs a particle with probability @xmath43 or remains unoccupied with probability @xmath44 , where @xmath33 is the depth of the perimeter node @xmath31 , i.e. its distance from the root @xmath40 .
this growth process occurs simultaneously at all the perimeter nodes .
thus , the total number of particles @xmath45 in the tree rooted at @xmath40 is clearly not a fixed number at a given time @xmath36 , instead it is a random variable in the sense that the value of @xmath45 differs from one history of evolution to another .
we are interested in computing the statistics of @xmath45 at large times @xmath36 . the growth of the aldous - shields model on a tree with @xmath8 branches and rooted at @xmath40 .
the filled circles are occupied nodes and the shaded ones are the perimeter nodes where growth can occur subsequently .
for example the site marked @xmath31 is a typical perimeter site at a distance @xmath46 from the root @xmath40 of the tree.,width=384 ] in this model , we have two parameters @xmath8 and @xmath1 .
it is useful to first summarize our main results .
using a backward fokker - planck approach we derive an exact evolution equation for the generating function , @xmath47 where the angle brackets denote an average over all histories of the evolution process and @xmath48 is the probability distribution of @xmath28 at time @xmath36 .
we show that @xmath49 evolves via the equation @xmath50 starting from the initial condition @xmath51 . by differentiating @xmath49 with respect to @xmath52 and putting @xmath53
, one can also derive the evolution equations for all the moments of @xmath45 .
the equation ( [ eqg ] ) is nonlinear and nonlocal in time for generic values of @xmath1 and @xmath8 , and is thus difficult to solve exactly , except for the @xmath3 case when it becomes local .
however , we were able to compute exactly the asymptotic large time behaviors of the mean and the variance of @xmath45 for arbitrary @xmath8 and @xmath1 .
below we present our results for the three different cases @xmath3 , @xmath54 and @xmath55 separately . *
the case @xmath3 : * in this case our model is precisely the continuous - time version of the eden model .
this case @xmath3 is exactly solvable since the evolution equation ( [ eqg ] ) becomes local in time .
we solved for @xmath49 and obtained the following explicit result for the distribution @xmath48 for all @xmath8 and @xmath36 @xmath56^{n } \label{csol1}\ ] ] where @xmath57 is the standard gamma function .
the mean number of particles @xmath58 increases exponentially in time for all @xmath59 , @xmath60 .
\label{eqc1}\ ] ] for the special case @xmath61 ( a line with a constant rate of deposition ) , @xmath62 and the distribution @xmath63 , obtained from eq .
( [ csol1 ] ) by taking the limit @xmath64 , is purely poissonian as expected .
* the case @xmath54 : * since the growth rate at a perimeter node @xmath31 is proportional to @xmath35 where @xmath33 is the distance of the node from the root @xmath40 , it is clear that for @xmath54 , farther a perimeter node is from the root , the larger is its probability to get occupied .
thus the cluster grows in a rather ramified manner where long branches grow faster than the short branches . in this case
we expect that the mean number of sites grows at least exponentially .
but since physically this case is of little interest , we do not discuss it further in this paper . * the case @xmath55 : *
we now come to the physically most relevant case @xmath55 .
we show that in this case there is a sharp phase transition in the asymptotic statistics of @xmath45 across the critical line @xmath9 in the @xmath65 plane with @xmath59 and @xmath55 .
we calculated exactly the asymptotic time dependence of the mean @xmath66 and the variance @xmath67 , for all values of @xmath59 and @xmath55 .
we show that while the fluctuations are normal ( i.e. the variance of @xmath45 is proportional to its mean ) for @xmath68 , they are anomalously large for @xmath69 . even though we have two parameters @xmath1 and @xmath8 , it turns out that the asymptotic behaviors can be described in terms of the single growth parameter @xmath70 where @xmath71 since @xmath55 . in terms of @xmath72 ,
the phase transition takes phase at the critical value @xmath73 .
the normal phase for @xmath68 corresponds to @xmath74 and the anomalous phase for @xmath69 corresponds to @xmath75 . the phase transition as a function of @xmath76 .
the critical point at @xmath73 separates the phase ( @xmath77 ) with normal fluctuations from the phase ( @xmath78 ) with anomalously large fluctuations.,width=384 ] more precisely , we find that for large @xmath36 , the mean @xmath79 grows as a power law ( up to corrections periodic in @xmath80 ) , @xmath81 and we provide an explicit expression for the amplitude @xmath82 .
the variance , on the other hand , has different behaviors for @xmath69 and @xmath68 or equivalently for @xmath78 and @xmath77 .
we show that the variance at large times @xmath36 , again up to log - periodic corrections , grows as @xmath83 we also provide exact expressions for the amplitudes @xmath84 , @xmath85 and @xmath86 . in order to use these continuous - time results for the discrete - time model where the ` time ' is same as the number of particles , it is instructive to eliminate the explicit @xmath36 dependence in the results for the variance and instead express it as a function of the mean number of particles . eliminating @xmath36 between eqs .
( [ eqm ] ) , ( [ eqv2 ] ) , ( [ eqvc ] ) and ( [ eqv1 ] ) , we get @xmath87 explicit expressions for the amplitudes @xmath88 , @xmath89 and @xmath90 are likewise provided .
we thus see that for @xmath91 the fluctuations of @xmath45 about its mean value , denoted by @xmath92 , are of order @xmath93 as is the case for a normal gaussian or poisson distribution .
however for @xmath94 we find that @xmath95 for large @xmath96 . for @xmath94
, we have that @xmath97 and hence the relative fluctuations about the mean become larger as we cross the threshold @xmath98 from below . the phase @xmath91 , or equivalently
@xmath68 corresponds to a region of slower growth where the central limit theorem holds and the distribution of @xmath45 is asymptotically normal . on the other hand , @xmath94 marks a phase where rapid growth tends to occur along a single branch resulting in anomalously large fluctuations .
thus the statistics of @xmath45 in this phase is dominated by extreme fluctuations .
the nature of this phase transition is thus very similar to the ones recently reported in @xmath8-ary search trees @xcite and a related fragmentation model @xcite .
we end this section with a remark on the usage of the term ` phase transition ' .
the ` phase transition ' observed in this model refers to the abrupt change of the variance ( and also that of the full distribution ) of the number of particles @xmath45 in the tree as one changes the parameter @xmath99 through its critical value @xmath73 .
this may not correspond to the traditional definition of ` phase transition ' used in equilibrium statistical mechanics , e.g. the divergence of a correlation length as one approaches a critical point as in second order phase transition .
the ` phase transition ' in the aldous - shields model is closer to the change of behavior one observes in the diffusion of a lvy walker . a lvy walker jumps , at each step , by a random length @xmath2 drawn from a power law distribution , @xmath100 for large @xmath2 with @xmath101 ( required for normalization ) .
it is well known@xcite that the root mean square displacement of the particle after @xmath28 steps @xmath102 for large @xmath28 only when @xmath103 , i.e. one gets normal diffusion and the asymptotic position of the walker is distributed normally . on the other hand , for @xmath104 one gets anomalously large diffusion , @xmath105 for large @xmath28 and the asymptotic distribution of the position of the walker is non - gaussian .
thus there is a change of behavior at the critical value @xmath106 .
the change of behavior in the variance of the number of particles in the aldous - shields model at the critical value @xmath73 is thus similar in nature to the the change of behavior seen for the lvy diffusion at @xmath106 , rather than the standard ` phase transition ' observed in critical phenomena .
in this section we will derive the evolution equation for the probability distribution , and in particular the evolution equations for the mean and the variance , of the total number of occupied sites @xmath45 at time @xmath36 in a tree rooted at the site @xmath40 .
the root @xmath40 , by definition , has level or depth @xmath107 .
the method of derivation is based on a backward fokker planck formalism which involves considering the future evolution of @xmath45 conditioned on what happens in the first infinitesimal time interval @xmath108 . since
our final aim is to derive a recursion relation for the evolution process , it is convenient to first derive the evolution equation for the number of particles @xmath109 in a subtree rooted , say , at any arbitrary site @xmath31 . by definition ,
@xmath109 includes the particle at the root @xmath31 .
the number of particles in the full tree is just a special case when the site @xmath31 is chosen to be the original root @xmath40 of the full tree .
we now count the ` local ' time @xmath36 ( for this subtree ) from the instant the site @xmath31 becomes a potential growth site , so that , by definition , @xmath110 . clearly , at any given @xmath36 , the distribution @xmath111 of @xmath109 depends only on @xmath36 and @xmath33 , the depth of the site @xmath31 .
this means that one can write @xmath112 where @xmath113 is any arbitrary function .
consider now the site @xmath31 with its descendants @xmath114 , where @xmath31 is a potential growth site at @xmath115 . by definition @xmath31
is unoccupied at @xmath115 and thus @xmath116 are not potential growth sites at @xmath115 . in the first infinitesimal time interval @xmath108
there are two possibilities : ( i ) either no particle fills the potential growth site @xmath31 , thus @xmath117 .
this happens with probability @xmath118 .
( ii ) the other possibility is that the potential growth site @xmath31 is filled by a particle with probability @xmath119 and as a consequence the number of particles in the subtree rooted at @xmath31 is increased by one and the daughter nodes @xmath116 all become potential growth sites .
mathematically we can write the above evolution in the following way : in the time interval @xmath108 @xmath120 i \label{master}\ ] ] where @xmath121 is a random variable which takes the value @xmath11 with probability @xmath122 and @xmath123 with probability @xmath124 . taking the expectation of eq .
( [ master ] ) with respect to @xmath121 and the subsequent growth process in the remaining time @xmath125 we obtain , upon taking the limit @xmath126 , @xmath127.\ ] ] we now use the property that the statistics of the number of particles in a subtree rooted at level @xmath31 depends only on @xmath33 as encoded in eq .
( [ inv ] ) to obtain @xmath128.\label{eqm1}\ ] ] where we have defined @xmath129 and have used the fact that by definition @xmath130 if @xmath131 is a daughter of the node @xmath31 .
note that the root @xmath40 of the full tree has depth @xmath132 .
thus the mean @xmath133 of the total number of particles in the full tree rooted at @xmath40 is simply , @xmath134 .
thus to obtain @xmath79 , our strategy is to find the solution @xmath135 of eq .
( [ eqm1 ] ) for arbitrary @xmath33 and eventually put @xmath136 .
the next step is to notice that as one goes down a level in the tree the growth rate is reduced by a factor of @xmath1 which amounts to rescaling time by a factor of @xmath137 . in the notation of eq .
( [ inv ] ) this means that one may write @xmath138 next we put @xmath136 in eq .
( [ eqm1 ] ) , use the definition @xmath139 and also the scaling property in eq .
( [ scal ] ) to obtain , @xmath140 this equation , supplemented by the boundary condition @xmath141 , then describes the evolution of the mean number of occupied sites .
an equation for the variance of @xmath45 can be derived in a similar fashion .
the starting point is obtained by squaring the stochastic evolution equation ( [ master ] ) and then taking the expectation over @xmath121 and the evolution in the remaining time @xmath125 .
this yields @xmath142c^{-l_a}\delta t . \label{eqvs}\ ] ] we now use the fact that the subtrees rooted at sites at the same level are statistically independent and so @xmath143 now defining the variance of the number of sites occupied in the tree rooted at @xmath123 as @xmath144 and using the scaling relation eq .
( [ scal ] ) , after some elementary algebra , we obtain @xmath145 the boundary condition for this equation is clearly @xmath146 .
another way to obtain the equations for @xmath96 and @xmath147 is by deriving directly an evolution equation for the generating function @xmath49 of @xmath45 defined as in eq .
( [ defg ] ) .
following exactly the same backward fokker - planck strategy as used for the mean , it is straightforward to show that @xmath49 evolves by the nonlinear nonlocal equation ( [ eqg ] ) .
the moment equations , and equations for the higher moments , eq .
( [ eqm ] ) and eq .
( [ eqv ] ) can be obtained by differentiating eq .
( [ eqg ] ) with respect to @xmath52 the appropriate number of times and setting @xmath148 at the end .
the evolution equation ( [ eqg ] ) is difficult to solve explicitly for generic values of @xmath1 and @xmath8 since it is a nonlinear ( for @xmath149 ) and nonlocal ( for @xmath150 ) equation .
however , exact results can be derived in a few cases that we consider below . the asymptotic solution for the mean and variance for generic @xmath8 and @xmath1
will be presented later in the next section .
the case @xmath3 corresponds to the eden model where growth occurs at any of the available perimeter sites with equal rates . for @xmath3 , eq .
( [ eqg ] ) becomes local in time @xmath36 and can be explicitly solved .
we find that for all @xmath151 @xmath152^{-1/(m-1)}. \label{c1solgf}\ ] ] expanding the r.h.s .
( [ c1solgf ] ) in powers of @xmath153 as in eq .
( [ defg ] ) , one can then read off the distribution @xmath48 explicitly as in eq .
( [ csol1 ] ) .
the mean number of particles grows exponentially for all @xmath59 as in eq .
( [ eqc1 ] ) .
similarly , one can compute the variance @xmath154 .
we find @xmath155 for a fixed @xmath8 , if one takes the limit of large @xmath36 and large @xmath28 keeping the product @xmath156 fixed in eq .
( [ csol1 ] ) , one finds an asymptotic distribution @xmath157 . \label{distc1}\ ] ] thus , to the leading order , the distribution @xmath48 decays exponentially for large @xmath28 over a characteristic size @xmath158 that grows exponentially with time @xmath36 .
interestingly , the distribution has a sub - leading power law tail ( in addition to the leading exponential tail ) @xmath159 where the exponent @xmath160 depends continuously on @xmath8 . for the special case @xmath61 , where we have just a line of sites and the particles arrive at an empty available site at a constant rate @xmath11 , we get from eq .
( [ eqc1 ] ) , @xmath62 .
the full distribution , from eq .
( [ csol1 ] ) , becomes a poisson distribution @xmath161 as expected .
the case @xmath162 corresponds to the case where particles arrive at a constant rate at the root @xmath40 and then each carries out a random walk down the tree until it finds a free site to occupy . during its downward journey in the tree
the particle , after arriving at any occupied site , chooses one of its @xmath8 descendants at random .
this is precisely the algorithm for constructing a @xmath8-ary digital search tree @xcite .
if the rate at which the particles arrive at the root @xmath40 is one then the total number of particles in the tree at time @xmath36 , @xmath45 , is clearly a random variable with poisson distribution @xmath163 where @xmath164 is a positive integer .
this yields @xmath165 which we see immediately are the solutions to eq ( [ eqm ] ) and eq . ( [ eqv ] ) .
furthermore we see that the generating function @xmath49 for a poisson distribution is given by @xmath166 it is easy to check that indeed this solves eq .
( [ eqg ] ) in the case @xmath167 . the late time asymptotic behavior of eq .
( [ eqm ] ) and eq .
( [ eqv ] ) for @xmath55 and @xmath59 may be deduced quite simply by making a self - consistent ansatz for the late time behavior of @xmath96 and @xmath147 .
first consider eq .
( [ eqm ] ) .
we make the ansatz @xmath168 substituting this into eq .
( [ eqm ] ) we may neglect the derivative term on the l.h.s . and assuming that @xmath169 ( i.e. @xmath55 ) , matching the coefficients of @xmath170 gives @xmath171 which yields @xmath172 for non - trivial tree structures we are always in the situation where @xmath173 and for the above solution to make sense we require that @xmath174 to have a positive exponent @xmath72 .
while this simple minded scaling approach yields the correct power law growth of @xmath175 for @xmath55 , it does not provide us the value of the amplitude @xmath82 . to derive an exact expression for @xmath82
, we need to solve the full nonlocal equation ( [ eqm ] ) at late times , and this will be carried out in the next section .
let us make a similar power law ansatz for the late time behavior of @xmath154 @xmath176 substituting this into eq .
( [ eqv ] ) and neglecting the derivative term we obtain @xmath177 asymptotically there are two ways to satisfy this equation . first if we assume a priori that @xmath178 then the first two terms in eq .
( [ match1 ] ) must cancel leading to @xmath179 , i.e. @xmath180 . the a posteriori condition that this solution is valid is thus @xmath181 , which means @xmath182 .
the second possibility is that all three terms contribute and thus @xmath183 . in this case
we find that @xmath184 and in obtaining the last equality in eq .
( [ reab ] ) we have used @xmath185 . however for this solution to make sense we must have that @xmath186 because @xmath154 is clearly positive , consequently eq .
( [ reab ] ) can only hold when @xmath78 ( since @xmath55 ) .
this simple minded scaling approach thus indicates that there is a phase transition in the late time behavior of the variance @xmath154 at the critical parameter value @xmath73 . for @xmath77 , we have @xmath187 where the amplitude @xmath84 can not be determined by the scaling approach . on the other hand , for
@xmath78 the scaling approach indicates @xmath188 and moreover it provides a relationship between the amplitudes @xmath86 and @xmath82 ( of the mean ) via eq .
( [ reab ] ) .
the critical point @xmath189 thus separates the region of normal growth @xmath77 ( or equivalently @xmath68 ) , where @xmath190 , from the the region @xmath78 ( i.e. @xmath69 ) where the variance grows anomalously faster @xmath191^{2 - 2/\alpha}$ ] . in the next section
, we will see that the analysis of the full nonlocal equations ( [ eqm ] ) and ( [ eqv ] ) indeed corroborates theses scaling results , and in addition produces exact expressions for all the amplitudes . before proceeding to the full analysis of eqs .
( [ eqm ] ) and ( [ eqv ] ) in the next section for generic @xmath55 and @xmath8 , it is instructive to note that analytic progress is also possible for eq .
( [ eqm ] ) in the case where @xmath99 is a positive integer .
this includes , in particular , the critical point @xmath192 .
we make the following ansatz @xmath193 where the term @xmath194 in the above sum is omitted in order to respect the initial condition @xmath195 .
matching powers of @xmath36 on substituting this ansatz into eq .
( [ eqm ] ) yields @xmath196 we thus see that if there exists a positive integer @xmath197 such that @xmath198 then @xmath199 for all @xmath200 and we have found the solution to eq . ( [ eqm ] ) in these cases . at late times the leading order behavior is thus dominated by the term containing @xmath170 and we get @xmath201 in particular , at the critical point @xmath192 , we get for large @xmath36 @xmath202 thus , at this special point @xmath192 , we have even managed to compute the amplitude @xmath203 of the mean @xmath204 exactly . in the case
@xmath94 , the behavior of the variance @xmath154 now follows immediately from eq .
( [ reab ] ) .
finally , exactly at the critical point @xmath205 , we may asymptotically solve eq .
( [ eqv ] ) with the ansatz @xmath206 to yield @xmath207 where in the last line of eq .
( [ critint2 ] ) we have used @xmath203 .
the full solutions to the nonlocal and nonlinear differential equations of the type in eqs .
( [ eqm],[eqv ] ) are rather difficult to obtain completely . here we obtain the exact asymptotic solutions following an approach similar to the one used by flajolet and richmond @xcite in solving a class of difference - differential equations arising in the context of digital search trees .
we start by the analysis of eq .
( [ eqm ] ) assuming @xmath55 and @xmath59 . taking the laplace transform of eq .
( [ eqm ] ) we obtain @xmath208 where @xmath209 and we have used the initial condition @xmath210 .
the above may be written as @xmath211 now as @xmath212 should go to zero as @xmath213 and we are considering the case @xmath55 , we solve eq .
( [ eqms1 ] ) by iteration finding @xmath214 note that taking the limit @xmath215 is not straightforward in eq .
( [ eqms2 ] ) .
this is because if we set @xmath216 in the sum on the r.h.s of eq .
( [ eqms2 ] ) , the sum diverges since @xmath59 .
following @xcite we introduce the function @xmath217 thus @xmath218 . on the other hand , @xmath219 thus , one can rewrite the product @xmath220 . using this in eq .
( [ eqms2 ] ) we get @xmath221 where @xmath222 the next step is to take the mellin transform of @xmath223 defined as @xmath224 substituting @xmath223 from eq .
( [ hs1 ] ) in the definition in eq .
( [ hs2 ] ) we get @xmath225 where @xmath226 and in evaluating the sum over @xmath227 we have assumed @xmath228 or equivalently @xmath229
. we also notice that @xmath230 has no poles for @xmath231 , and that the poles of @xmath232 are at @xmath233 where @xmath234 runs over all integers .
all the poles of @xmath235 are thus to the left of the line @xmath236 . the inversion formula for the mellin transform is given by @xmath237 where the above limits denote an integration up the imaginary axis to the right of all the poles of @xmath238 , therefore we chose limits with @xmath239 .
the contour may be closed in the left half plane ( we assume that the integrand vanishes in the region @xmath240 ) and we can thus evaluate the inverse mellin transform in terms of the residues of the poles to the left of @xmath241 , _ i.e _ @xmath242 \label{imt1}\ ] ] where @xmath243 denotes the residue at the pole in question . the large time behavior of @xmath79 is determined by the small @xmath244 behavior of @xmath212 . now at small @xmath244 the dominant behavior clearly comes from the poles @xmath245 running up the imaginary axis , any pole coming to the left of this line of poles will be higher order in @xmath244 . we evaluate the residues in eq .
( [ imt1 ] ) , substitute the resulting @xmath223 in eq .
( [ ms1 ] ) and then take the limit @xmath215 to obtain the following asymptotic result @xmath246,\label{eqhs}\ ] ] where we have used the fact @xmath247 .
note that from eq .
( [ eqh ] ) and eq .
( [ qdef ] ) , we have @xmath248 where the last equality follows from an identity due to ramanujan @xcite .
note that this identity explicitly shows that the function @xmath230 has simple poles at the negative integers and zero but no poles for @xmath249 as was stated before . to extract the leading asymptotic behavior of @xmath79 for large @xmath36 ,
let us first divide the laplace transform @xmath212 into two parts , @xmath250 where @xmath251 denote the first term on the r.h.s . of eq .
( [ eqhs ] ) and @xmath252 corresponds to the remaining sum over @xmath253 .
subsequently the inverse laplace transform @xmath254 can also be divided into two parts .
the term @xmath251 has a pure algebraic form , thus its inverse @xmath255 has a pure power law growth , @xmath256 where the constant @xmath82 can be evaluated as follows . if @xmath255 has the form in eq .
( [ mpt1 ] ) , its laplace transform is @xmath257 . comparing this with the first term in eq .
( [ eqhs ] ) gives @xmath258 where we have used @xmath259 , the definition @xmath260 and the explicit form of @xmath261 from eq .
( [ rameq ] ) . here
we note that when @xmath72 is an integer , it can be verified that eq .
( [ a1 ] ) agrees with eq .
( [ eqmint ] ) derived for discrete values of @xmath72 in the previous section .
the second contribution to @xmath212 , @xmath262 is given by a fourier series in @xmath263 .
the inverse laplace transform of this term is difficult to obtain fully but it is easy to see that it gives rise to a late time behavior of the form @xmath264 where @xmath265 is a periodic function of @xmath266 . the final asymptotic result for large @xmath36 is thus @xmath267
. \label{eqmtb}\ ] ] this exact result thus not only confirms the dominant power - law scaling predicted in section ( iii ) up to log - periodic oscillations , but also provides an explicit formula for the amplitude @xmath82 as in eq .
( [ a1 ] ) .
for example , let us consider the binary case @xmath41 .
for the case , when @xmath3 , the formula in eq .
( [ a1 ] ) gives @xmath268 , thus @xmath269 for large @xmath36 . on the other hand , for @xmath41 , when @xmath5 ( the critical point ) , one can show from eq .
( [ a1 ] ) that @xmath270 and @xmath271 for large @xmath36 .
we now examine the asymptotic behavior of the variance @xmath154 for large @xmath36 using a similar formalism .
the evolution equation ( [ eqv ] ) for the variance @xmath154 is similar to that for the mean @xmath79 in eq .
( [ eqm ] ) except that the source term in eq .
( [ eqv ] ) is @xmath272 , different from the source term @xmath11 in eq .
( [ eqm ] ) .
solution of eq .
( [ eqv ] ) thus requires an explicit knowledge of how @xmath79 behaves with time .
taking the laplace transform , @xmath273 in eq .
( [ eqv ] ) and using @xmath274 we obtain @xmath275 where @xmath276 rearranging eq .
( [ eqs ] ) gives @xmath277 which can be iterated to yield @xmath278 using @xmath185 and the function @xmath279 defined in eq .
( [ qdef ] ) we can rewrite eq .
( [ vs2 ] ) as @xmath280 where @xmath281 the next step is to take the mellin transform @xmath282 of eq .
( [ h1s1 ] ) which gives , after a change of variable @xmath283 in the integration @xmath284 let us first assume that the integral @xmath285 exists ( the conditions for which will be stated later ) .
then , for @xmath286 , the geometric sum in eq .
( [ h1x1 ] ) converges ( since @xmath55 ) and we get @xmath287 inverting this mellin transform we get @xmath288 , \label{h1s2}\ ] ] where the poles are at @xmath289 with @xmath290 . in eq .
( [ h1s2 ] ) the integration is along the imaginary axis to the right of all the poles and then we close the contour over the left half plane . evaluating the residues and substituting the results in eq .
( [ vs3 ] ) we get @xmath291 , \label{vs4}\ ] ] where @xmath292 is given by eq .
( [ h1x2 ] ) assuming that it exists .
we now need to invert the laplace transform in eq .
( [ vs4 ] ) to evaluate @xmath154 . for large @xmath36 , as usual , the dominant contribution will come from the small @xmath244 behavior of @xmath293 . using @xmath247 and assuming @xmath294 exists for all @xmath295 , it is clear from the small @xmath244 behavior of @xmath293 in eq .
( [ vs4 ] ) that for large @xmath36 @xmath296 , \label{vta1}\ ] ] where @xmath297 is a periodic function in @xmath266 and the amplitude @xmath84 can be read off as @xmath298 having obtained the results in eqs .
( [ vta1 ] ) and ( [ b1 ] ) , we need to investigate when they are valid .
these results are valid as long as the integral @xmath299 in eq .
( [ b1 ] ) exists .
the existence of this integral depends on the small @xmath244 behavior of the source function @xmath300 defined in eq .
( [ ss ] ) . using the asymptotic behavior of @xmath79 from eq .
( [ eqmtb ] ) we find that for large @xmath36 @xmath301 , \label{dmdt2}\ ] ] where @xmath302 is a periodic function in @xmath266 .
substituting this large @xmath36 behavior of @xmath272 in eq .
( [ ss ] ) , it follows that , in the case @xmath303 , the integral converges to a nonzero constant as @xmath215 . on the other hand , for @xmath304 , the integral diverges as @xmath305 as @xmath215 .
up to the log - periodic oscillations , the leading behavior of @xmath300 for small @xmath244 can be summarized as follows @xmath306 where @xmath307 is a positive constant for @xmath304 . also , @xmath308 is a constant that depends on the full form of @xmath79 and not just on its asymptotic behavior since for @xmath309 the integral is convergent . substituting this small @xmath244 behavior of @xmath300 into the integral giving @xmath299 in eq .
( [ b1 ] ) and using @xmath247 , it is clear that the integral exists ( no divergence from the small @xmath244 limit ) only for @xmath77 . for @xmath78 ,
the integral does not exist since the integrand for small @xmath244 scales as @xmath310 .
thus the results in eqs .
( [ vta1 ] ) and ( [ b1 ] ) hold only for @xmath77 .
for @xmath78 , the above analysis breaks down and we need to employ a different method .
we now go back to our starting equations ( [ vs3 ] ) and ( [ h1s1 ] ) .
it turns out that for @xmath78 , we can actually extract the leading small @xmath244 behavior directly from these two equations .
we directly substitute in eq .
( [ h1s1 ] ) the leading small @xmath244 behavior of @xmath311 from eq .
( [ aless ] ) where @xmath312 .
additionally we use @xmath247 . eqs .
( [ vs3 ] ) and ( [ h1s1 ] ) then yield in the @xmath215 limit @xmath313 where we have used @xmath78 which ensures that the sum in eq .
( [ vs5 ] ) is convergent . inverting the laplace transform
, we then get the large @xmath36 behavior of @xmath154 for @xmath78 @xmath314 where the constant @xmath82 is given in eq .
( [ a1 ] ) .
note that this result in eq .
( [ vs6 ] ) for @xmath78 is in perfect agreement with the self - consistent scaling approach used in section - iii . at the critical point @xmath192 ,
the analysis is more delicate .
however , from the scaling approach of section - iii , we already know that for large @xmath36 , @xmath315 with @xmath316 as in eq .
( [ critint2 ] ) .
thus , the asymptotic behavior of the variance @xmath154 can be summarized as @xmath317 where the three amplitudes are given by @xmath318 where @xmath82 is given in eq .
( [ a1 ] ) .
note that computing the amplitude @xmath84 explicitly requires an integration over the full source function @xmath300 which is not so easy . eliminating the time @xmath36 between @xmath319 and @xmath154 in eq .
( [ varat ] ) , one can express the variance @xmath147 as a function of the mean @xmath96 for large @xmath96 as in eqs .
( [ eqvm2 ] ) , ( [ eqvmc ] ) and ( [ eqvm1 ] ) and one can read off the constants @xmath88 , @xmath89 and @xmath90 in terms of @xmath84 , @xmath85 and @xmath86 and the amplitude @xmath82 of the mean given in eq .
( [ a1 ] ) .
let us end this section with a remark on the mathematical mechanism responsible for the phase transition in the variance of the number of particles in this aldous - shields model .
we note that the exact evolution equations ( [ eqm ] ) and ( [ eqv ] ) respectively for the mean and the variance are very similar they are both linear and nonlocal in time , the only difference is in the source term .
for the mean @xmath79 in eq .
( [ eqm ] ) , the source term is a constant @xmath11 ( the first term on the r.h.s of eq .
( [ eqm ] ) ) . on the other hand , for the variance , the source term @xmath272 in eq .
( [ eqv ] ) depends on the evolution of the mean .
thus , the mean feeds into the variance equation as an external source term leading to a competition between the growth induced by this external source term and the growth induced internally by the remaining two terms on the r.h.s of eq .
( [ eqv ] ) .
this competition between the external and the internal source is finally responsible for the phase transition in the asymptotic growth of @xmath154 . for @xmath77
, the internal source term wins out and the variance grows similarly as the mean , @xmath320 , leading to the normal phase . on the other hand , for @xmath78
, the external source term wins out leading to a faster growth @xmath321^{2 - 2/\alpha}$ ] characterizing anomalously large fluctuations .
we note that a similar mechanism namely a _ competition between the internal source and the external driving _ "
was shown to be responsible for phase transitions in fluctuations in a class of fragmentation problems studied recently@xcite . in these fragmentation problems
, it was shown that the mean and the variance evolved via similar looking equations@xcite , except that they differed in their respective source terms the variance equation had a source term driven by the mean , in much the same way as in the aldous - shields model discussed here .
thus , it seems that this phase transition in fluctuations is quite generic as it occurs in a large class of problems and the mathematical mechanism responsible for it is as identified above .
in this paper we have studied analytically a growing tree model introduced by aldous and shields . in this model
, growth occurs in continuous time .
one starts at @xmath115 with an empty cayley tree with @xmath8 branches rooted at @xmath123 and the tree grows , starting from the root site , by absorbing particles in continuous time .
each site can occupy at most one particle . at a given instant @xmath36
, growth can occur only at the perimeter sites with a rate @xmath0 where @xmath1 is positive parameter and @xmath2 is the distance of the perimeter site from the root of the tree . for @xmath3
this model is isomorphic to a continuous - time eden model on a tree and also corresponds to the random binary search tree problem in computer science . for @xmath4
this model corresponds to the digital search tree problem in computer science .
we have introduced a backward fokker - planck approach that enabled us to study analytically the statistics of the total number of particles @xmath45 in the tree at large time @xmath36 .
we have shown that at large @xmath36 , while the mean number of particles grows as a power law in time , @xmath322 with @xmath99 for all @xmath55 , the variance @xmath154 of the number of particles has two different behaviors depending on the value of the parameter @xmath72 . while for @xmath77 @xmath190 for large @xmath36 , for @xmath78 the variance grows anomalously quickly : @xmath323 ^{2 - 2/\alpha}$ ] .
we have identified the mathematical mechanism behind this phase transition at the critical value @xmath73 and shown that it is qualitatively similar to the phase transitions recently encountered in a search tree problem and also in a related fragmentation problem .
essentially , for @xmath77 , the typical value of @xmath45 grows in the same way as the average and the distribution is asymptotically normal whereas for @xmath78 , the typical value does not grow the same way as the average and the distribution is characterized by large fluctuations caused by the faster growth of a single branch of the tree .
we obtained detailed analytical results for the first two moments of the number of particles for generic values of the parameter @xmath72 .
however , we were able to calculate the full asymptotic distribution of the number of particles only for two specific values of @xmath72 , namely for @xmath324 ( @xmath162 ) and @xmath325 ( @xmath3 )
. fortunately these two representative values , where an exact solution is possible , fall respectively on either side of the critical point @xmath73 .
our exact solution shows that for @xmath324 ( @xmath326 ) the distribution @xmath48 is poisson and hence is asymptotically normal for large @xmath28 . on the other hand for @xmath327 ,
the asymptotic distribution is certainly non - gaussian , @xmath328 $ ] where the exponent @xmath160 depends on @xmath8 .
the calculation of the distribution for other values of @xmath72 remains a challenging problem . while we have studied this growth model on a tree because of its connections to the search tree problems as mentioned in the introduction
, it is of general interest to study this growth problem on a regular euclidean lattice , e.g. on a hyper - cubic lattice in @xmath10 dimensions . in this lattice model
, the cluster will grow similarly from a seed site at the origin . at a given instant
, growth can occur at any of the available surface sites with a rate @xmath329 where @xmath330 is the euclidean distance of the surface site from the origin .
one can then investigate the statistics of the total number of particles in the cluster after time @xmath36 .
it is easy to make a scaling argument for the growth of the mean number of particles @xmath79 . assuming that the cluster is compact with a typical radius @xmath331 at time @xmath36 , we have @xmath332^{d}$ ] .
also , the mean number of surface sites @xmath333^{d-1}$ ] . by the growth rule , @xmath334 for large @xmath36 .
this predicts @xmath335 for large @xmath36 and hence the mean number of particles grows very slowly as @xmath336^{d}$ ] for large @xmath36 .
an interesting open question for future studies is whether , in finite dimensional lattice models , the variance exhibits a phase transition , similar to that seen on the tree , for some critical value of the parameter @xmath1 ? for a recent review on search trees for physicists , see s.n .
majumdar , d.s .
dean and p.l .
krapivsky , understanding search trees via statistical physics , in _ proceedings of the statphys 22 _ ( bangalore , india , 2004 ) , pramana - j
64 * , 1175 - 1189 ( 2005 ) ( also available at http://xxx.arxiv.org/cond-mat/0410498 ) ; see also s.n .
majumdar and p.l .
krapivsky , extreme value statistics and traveling fronts : application to computer science , phys .
e * 65 * , 036127 ( 2002 ) .
m. ghorbel and t. huillet , fragment size distributions in random fragmentations with cut - off , stat .
71 * , 47 - 60 ( 2005 ) ; t. huillet , statistical aspects of random fragmentation , j. comput .
appl . math .
* 181 * ( 2 ) , 364 - 387 ( 2005 ) . | we study analytically the late time statistics of the number of particles in a growing tree model introduced by aldous and shields . in this model
, a cluster grows in continuous time on a binary cayley tree , starting from the root , by absorbing new particles at the empty perimeter sites at a rate proportional to @xmath0 where @xmath1 is a positive parameter and @xmath2 is the distance of the perimeter site from the root . for @xmath3
, this model corresponds to random binary search trees and for @xmath4 it corresponds to digital search trees in computer science . by introducing a backward fokker - planck approach
, we calculate the mean and the variance of the number of particles at large times and show that the variance undergoes a ` phase transition ' at a critical value @xmath5 . while for @xmath6 the variance is proportional to the mean and the distribution is normal , for @xmath7 the variance is anomalously large and the distribution is non - gaussian due to the appearance of extreme fluctuations .
the model is generalized to one where growth occurs on a tree with @xmath8 branches and , in this more general case , we show that the critical point occurs at @xmath9 . * keywords : * search trees , growth models , phase transitions |
for more than four decades , since their introduction in 1965 by montroll and weiss @xcite , continuous time random walks ( ctrw s ) have been applied to virtually any field in which one wishes to provide a dynamical description on the microstructure of a given random system .
a huge number of examples and applications can be found in the literature , of which we only cite a handful : transport in disordered media @xcite , random networks @xcite , self - organized criticality @xcite , electron tunneling @xcite , earthquake modeling @xcite , hydrology @xcite , time - series analysis @xcite and finance @xcite .
the ctrw generalizes the ordinary random walk since in the latter the steps of the random walker are made at equal intervals of time , while in the ctrw the interval between steps is a continuous random variable . in this sense ctrw
is related to several other extensions of random walks in continuous time , like semi - markov processes or markov renewal processes @xcite , although the seeds of this idea can be traced back to the 1920 s with the pure birth poisson process @xcite .
a great number of developments of the ctrw are based on the assumption that the magnitude of the steps ( or jumps ) and the time intervals between them ( also called sojourns ) are a two - dimensional set of independent and identically distributed random variables . while in many cases this is a convenient assumption which allows for simple developments , there are some other cases in which independent walks are clearly insufficient to explain some aspects of the physical reality , and correlations between consecutive step sizes and/or waiting times must be considered .
we have met with such a case in our study of financial time series regarding their extreme time statistics @xcite where jump magnitudes ( transaction - to - transaction returns ) show short - range memory , but it seems to be also relevant in earthquake modelling , where some evidences pointing to the presence of cross - correlations between simultaneous and sequential earthquake magnitudes and recurrence times have been reported @xcite .
successive recurrence times appear to be positively self - correlated as well but , oddly enough , consecutive magnitudes seems to be independent in this case .
our goal in this paper is to address the theoretical setting of non - independent ctrw s , a class of random walks that can account for many physical situations . in this kind of walks jumps and sojourns are no longer independent random variables and their value at a given step may depend on previous steps .
we develop the formalism for the markovian case in which the magnitudes of a given jump and time interval only depend on the preceding step .
we also obtain a complete solution to the problem when the correlation between consecutive steps solely depend on the sign of the previous jump ( that is , whether the previous jumps is increasing or decreasing but not on its magnitude ) .
particular examples corresponding to this solvable case allow us to visualize and quantify some interesting consequences of the existence of correlations between steps as , for instance , the transitions from unimodal to bimodal distributions .
finally , we also develop the necessary perturbation techniques to deal with more general situations .
the paper is organized as follows : in sect .
[ sec2 ] we outline the traditional ctrw based on the assumption of independence between events . in sect .
[ sec3 ] we present the general setting for non - independent ctrw s that are still amenable to analytical treatment . in sect .
[ sec3b ] we present an exact solution to the problem .
section [ sec4 ] is devoted to consider weak dependent models for which we develop a complete perturbation technique .
conclusions are drawn in sect .
although this work is essentially technical some even more technical aspects are in an appendix .
suppose that a given random process @xmath0 evolves following a ctrw . in this picture
any realization of @xmath0 consists of a series of step functions and @xmath0 changes at random times @xmath1 while it remains fixed in place between successive steps .
the interval between these successive steps is a random variable @xmath2 which we call sojourn or waiting time . at the conclusion of the @xmath3th sojourn
@xmath0 experiences a random change , or jump , given by @xmath4 in the usual model of the ctrw waiting times @xmath5 and random jumps
@xmath6 constitute a two - dimensional set of identically distributed random variables , @xmath7 , described by the corresponding joint probability density function ( pdf ) @xmath8 , @xmath9 as usual , two marginal pdf s can be derived from @xmath8 : @xmath10 and @xmath11 just by integrating over the opposite variable : @xmath12 in this set - up , each pair of random variables @xmath7 is independent of any other pair @xmath13 , @xmath14 , but it is still allowed any degree of correlation between @xmath6 and @xmath5 themselves , @xmath15 , can be thought as a generalization of a semi - markov process , since @xmath16 is , in general , not a markov chain but a markov discrete process with continuous states , and @xmath5 is a random variable depending on both @xmath16 and @xmath17 through @xmath6 . ] .
some ctrw processes belonging to this category can be found in @xcite for instance , and sometimes are also named as _ non - independent _ models @xcite , since the label _ independent _ is then reserved for the particular case in which @xmath6 and @xmath5 are also _ mutually _ independent random variables , _
i.e. _ when @xmath18 this divergence in the existing notation may therefore induce to misinterpretation .
we prefer to call eq .
( [ rho_full_indep ] ) the _ fully - independent _ ctrw , and keep the term independent ctrw for any memoryless process with arbitrary joint pdf .
the chief objective of the ctrw formalism is to obtain the probability density function of @xmath0 .
this pdf , called the propagator , is defined by @xmath19 where in what follows we shall assume that the initial jump occurred at @xmath20 .
as is well known the propagator obeys the following renewal equation @xcite @xmath21 the function @xmath22 is the propagator prior the first jump and , since the trajectories of @xmath0 consist in series of step functions , we write @xmath23 where @xmath24 is the probability that no transaction has occurred before time @xmath25 @xmath26 we can solve eq .
( [ renewal_0 ] ) in terms of the joint fourier - laplace transform : @xmath27 the solution is @xmath28 where @xmath29 and @xmath30 are the joint fourier - laplace transforms of the functions @xmath22 and @xmath31 .
we easily see from eqs .
( [ p0_0])-([cumulative ] ) that the explicit form of @xmath29 is @xmath32 where @xmath33 is the laplace transform of the pausing - time density @xmath34 .
as we have explained , the ctrw outlined in the preceding section relies on the assumption that the pairs @xmath7 are independent two - dimensional random variables . however , as we have mentioned in sect .
[ sec1 ] there are many situations in which the assumption of independence may be doubtful @xcite
. we shall thus generalize the ctrw formalism in order to account for memory effects due to correlations between different sojourns and/or jump increments . among the many ways of doing this extension we choose a simple , but yet general , method which consists in assuming that the joint density of @xmath35 consecutive changes , @xmath36 , fulfills the markov property : @xmath37 where @xmath38 this means that the pair @xmath7 depends on all the previous transitions only through the immediately preceding one @xmath39 .
in an analogous way to the independent walk described by eq .
( [ renewal_0 ] ) , the integral equation governing the evolution of the return pdf is given in this case by the renewal equation @xmath40 where @xmath41 is the propagator prior the first jump and similarly to eq .
( [ p0_0 ] ) we write @xmath42 where @xmath43 is the cumulative distribution of the waiting time and it is related to the transition density @xmath44 by @xmath45 observe that in this case the process @xmath0 is not markovian because its pdf @xmath46 depends on both the magnitude of the previous jump @xmath47 and its sojourn time @xmath48 ; in other words , the probability distribution of the process at a given time depends on two previous times @xmath49 and @xmath50 .
we should note at this point that even the independent ctrw is , in general , a non - markovian process .
the only case in which the independent ctrw is markovian is when it is a fully - independent ctrw , and the set of random times @xmath51 is poissonian , that is , when the pausing time density @xmath34 obeys the exponential law @xcite : @xmath52 let us remark that the dependent ctrw outlined above is always non - markovian even for this poissonian density .
we also note that in the case of independent increments discussed in sect .
[ sec2 ] we have @xmath53 and eq .
( [ nmrenewal ] ) reduces to eq .
( [ renewal_0 ] ) . the integral equation given in eq .
( [ nmrenewal ] ) is the general equation that governs the evolution of the random process @xmath0 and it must be solved if we want to obtain the propagator for this non - independent case .
contrary to the independent case , eq .
( [ nmrenewal ] ) can not be solved , for any form of the joint density @xmath54 by means of transform methods .
indeed , the laplace transform with respect to @xmath25 of eq .
( [ nmrenewal ] ) is @xmath55 where the hat over @xmath56 and @xmath57 denotes the time laplace transform .
note that @xmath58 where we have exchanged the order of integration and performed the change of variables @xmath59 .
then @xmath60 finally , the fourier transform with respect to @xmath61 of this equation yields @xmath62 which is the farthest we can go without specifying @xmath44 .
the integral equation ( [ nmrenewal_2 ] ) can not be solved for any arbitrary form of @xmath54 . however , for the independent case in which @xmath63 and @xmath64 we recover from eq .
( [ nmrenewal_2 ] ) the solution given by eq .
( [ solution_0 ] ) .
another case in which the level of difficulty is somewhat reduced is when waiting times and jumps are not related to each other . in such a case
the joint density factorizes as : @xmath65 we will also assume the further simplification @xmath66 in which correlations between consecutive waiting times have been neglected and we have also assumed that the waiting time pdf , @xmath67 , does not depend on the magnitude of the jumps ; a situation that , as mentioned above , has been detected in some financial time series @xcite .
the renewal equation for the propagator of @xmath0 now reduces to ( cf .
( [ nmrenewal])-([nmp0 ] ) ) @xmath68 the ( time ) laplace transform of this equation yields @xmath69 where , in terms of the density @xmath33 , the waiting time distribution function @xmath70 can be written as @xmath71 let us now suppose that the conditional jump density @xmath72 has the form @xmath73 , \label{conditional_jump}\ ] ] where @xmath74 is an arbitrary parameter and @xmath75 is the unconditional jump density cf .
( [ h]) that is related to @xmath72 by the constraint @xmath76 note that this constraint impedes us to consider arbitrary functional forms for @xmath72 and @xmath75 .
for instance , if we set @xmath77/2 $ ] , @xmath78 , as in the case of the random walker with shrinking step sizes @xcite , eq .
( [ unconditional ] ) implies @xmath79 .
the distinctive point here is that , unlike geometric random walks , our ( unconditional ) pdf of @xmath6 does not depend on @xmath3 .
function @xmath80 represents the correlation between previous and current jumps and @xmath74 governs its strength .
note that in order to meet eq .
( [ unconditional ] ) together with normalization , @xmath81 the correlation @xmath80 must satisfy ( see next section for a general discussion on this issue ) @xmath82 with the form of @xmath72 given in eq .
( [ conditional_jump ] ) the integral equation for the propagator , eq .
( [ nmrenewal_3b ] ) , reads @xmath83\hat{p}(x - x',s|x')dx ' . \label{nmrenewal_3c}\ ] ] we will solve this equation for any even jump density @xmath84 and when the correlation function has the following form @xmath85 meaning that the dependence between current and previous jumps is only through their signs . in other words , the correlation depends on whether consecutive jumps are increasing or decreasing but not on their magnitude . from eq .
( [ conditional_jump ] ) we see that in this case , since @xmath72 must be positive definite , @xmath86 . in fact , this model might be interpreted as the simplest _ persistent _ ctrw @xcite , in which the probability that the process does not change its direction of movement is equal to @xmath87 , but where neither jumps nor sojourns are affected by this persistence .
this is the kind of memory we adopted in @xcite in order to model the observed anti - correlation , with origin in the bid / ask bounce effect : tick - by - tick price changes tend two oscillate back and forth between to values due to the bid / ask spread . let us note that the functional form of @xmath80 given in eq .
( [ g_sign ] ) implies that any dependence on @xmath88 is only through @xmath89 .
thus @xmath90 which allows us to write @xmath91 where @xmath92 is the heaviside step function . the substitution of eq .
( [ total_p ] ) into eq .
( [ nmrenewal_3c ] ) yields for @xmath93 the following set of coupled integral equations @xmath94 @xmath95 now the fourier transform with respect to @xmath61 turns this set into a system of algebraic equations : @xmath96 , \label{p+}\ ] ] @xmath97 , \label{p-}\ ] ] where @xmath98 is the `` half '' fourier transform of @xmath99 .
obviously @xmath100 where @xmath101 is the `` complete '' fourier transform of @xmath99 .
we note that in writing eqs .
( [ p+])-([p- ] ) and eq .
( [ sum_h ] ) we have imposed the symmetry of @xmath99 assumed in eq .
( [ even_h ] ) . solving for eqs .
( [ p+])-([p- ] ) we have @xmath102 where we have used the fact that for real jump densities @xmath99 the following identity @xmath103 holds .
the final solution to the problem is thus given by the combination of eqs .
( [ total_p ] ) and ( [ fourier_laplace_solution ] ) . therefore , under the assumptions given in eqs .
( [ independent_0 ] ) and ( [ conditional_jump ] ) and the special form of the correlation given in eq .
( [ g_sign ] ) , we have been able to obtain an exact expression for the fourier - laplace transform of the propagator valid for any forms of the waiting time density @xmath34 and jump density @xmath104 , provided that the latter is an even function of @xmath88 with no bias .
another interesting quantity is the unconditional propagator @xmath105 defined by @xmath106 in the analyzed case in which @xmath107 can be decomposed as in eq .
( [ total_p ] ) we have @xmath108 . \label{uncond_p}\ ] ] from eqs .
( [ fourier_laplace_solution ] ) and ( [ uncond_p ] ) we get @xmath109 note incidentally that when @xmath110 this expression reduces to @xmath111 which agrees with the solution of the independent case discussed in sect .
[ sec2 ] ( cf .
( [ solution_0 ] ) ) . aside from the unconditional pdf @xmath105 , which provides maximal information about the evolution of @xmath0
, there is another quantity of considerable practical interest : the ( unconditional ) variance of @xmath0 .
this quantity has the advantage that it does not require the knowledge of the entire jump distribution @xmath104 .
it suffices to know the pdf @xmath34 and the following two moments of @xmath104 : @xmath112 let @xmath113 be the unconditional second moment of the process : @xmath114 and let us denote by @xmath115 its laplace transform @xmath116 this can be written in terms of the joint fourier - laplace transform of @xmath105 by @xmath117 recall that a direct consequence of the unbiased assumption expressed in eq .
( [ even_h ] ) is that the odd moments of @xmath104 are equal to zero .
this implies that all odd moments of process @xmath0 vanish as well ; in particular this means that the variance of @xmath0 coincides with its second moment . the combination of eqs . ( [ fl_uncond ] ) and ( [ fl_moment_def ] ) leads , after some manipulations , to the relation
2\epsilon\mu_1 ^ 2\frac{\hat{\psi}^2(s)}{s[1-\hat{\psi}(s)][1-\epsilon\hat{\psi}(s)]}. \label{fl_moment}\ ] ] let us return to the propagator and particularize to the case of poissonian waiting times for which @xmath119 and @xmath120 now eq . ( [ fl_uncond ] ) reads @xmath121 whose inverse laplace transform yields the unconditional characteristic function @xcite @xmath122 } \left\{\cosh[\lambda t\tilde{k}(\omega)/2]+ ( 1-\epsilon)\frac{\tilde{h}(\omega)}{\tilde{k}(\omega)}\sinh[\lambda t\tilde{k}(\omega)/2]\right\ } , \label{cf}\ ] ] where @xmath123 as to the second moment is concern eq .
( [ fl_moment ] ) can be inverted at once with the result @xmath124 . \label{variance}\ ] ] observe that in the independent case ( @xmath110 ) the variance shows an ordinary diffusion behavior while correlations introduce a richer dynamics . in order to invert eq .
( [ cf ] ) and thus obtaining an expression for the propagator @xmath105 we must chose a functional form for the jump density @xmath104 .
we will select the two - sided exponential density , @xmath125 for two main reasons . on the one hand it can be of interest in finance , the field that motivates this work in the first instance . even though it is well established that pdf s of financial returns
show a power - law decay @xcite , there are an increasing number of evidences pointing to the fact that small and moderate returns are better described through a laplace law see @xcite and references therein . on the other hand ,
one of the main motivations of this sections is the introduction and subsequent analysis of a case for which we can obtain closed expressions .
it is clear from eq .
( [ cf ] ) that when the characteristic function of @xmath104 is intricate this goal will be well out of reach . in our case
@xmath101 is the inverse of a polynomial : @xmath126 we incidentally note that now the variance of the process is given by eq .
( [ variance ] ) where @xmath127 and @xmath128 .
when @xmath129 one can show that the fourier inversion of eq .
( [ cf ] ) reads ) -([explicit_0 ] ) , as well as all the numerical results presented below , it is somewhat simpler to start from eq .
( [ fl_uncond_poisson ] ) and use eq .
( [ f_laplacian ] ) to invert first the fourier transform and finally the laplace transform . ]
@xmath130i_0\left[\frac{(1-\epsilon)}{2\lambda\epsilon}\sqrt{u^2-\epsilon\lambda^2\gamma^2|x|^2}\right]du , \label{explicit}\end{aligned}\ ] ] ( @xmath129 ) , where @xmath131 are modified bessel functions .
although the case @xmath132 is contained in eq .
( [ explicit ] ) it can be written more explicitly as @xmath133 the recovery of the independent case @xmath110 from eq .
( [ explicit ] ) is a delicate issue that deserves a special treatment . in this case one has @xmath134 ( @xmath110 ) .
we can now graphically explore some of the most relevant properties of the unconditional propagator in the analyzed example . however , as we will show , our example will share those traits with any process that presents poissonian waiting times , _
i.e. _ for which eq .
( [ variance ] ) stands . in fig .
[ fig1 ] we plot the regular part of the probability density function ( in @xmath135 units ) for different values of @xmath74 . in particular we present the cases of ( i ) a strong anti - correlated process ( @xmath74 close to @xmath136 ) , fig .
[ fig1].a ; ( ii ) a weak anti - correlated process , fig .
[ fig1].b ; ( iii ) the independent case ( @xmath110 ) , fig .
[ fig1].c ; ( iv ) a weak correlated process , fig .
[ fig1].d ; ( v ) a strong correlated process ( @xmath74 close to @xmath137 ) , fig .
[ fig1].e ; and finally , ( vi ) the completely persistent case ( @xmath132 ) , fig .
[ fig1].f .
the visible effect of anti - persistent memory in the process is that the probability density function becomes narrower around @xmath138 .
thus , for larger negative values of @xmath74 , the system tends to remain longer near the origin , and the process exhibits sub - diffusive behavior for small timescales .
in fact , from eq .
( [ variance ] ) we will have for @xmath139 that @xmath140 and therefore the process is sub - diffusive for @xmath141 , diffusive in the independent case ( recall that this statement is valid for all timescales ) , and super - diffusive when @xmath142 . [ cols= " > , < , > , < " , ]
obtaining exact expressions for the propagator is usually quite involved , not to say impossible , in many practical situations and we have to resort to approximations based on perturbation solutions of the integral equation ( [ nmrenewal ] ) . with this purpose in mind
we will work with the following form of the joint density that generalizes the jump density given in eq .
( [ conditional_jump ] ) : @xmath143 , \label{wrho}\ ] ] where @xmath31 is the unconditional joint density satisfying @xmath144 and the function @xmath145 indicates the correlations between the waiting time and the jump of the current sojourn and those of the preceding sojourn .
@xmath74 is a parameter measuring the strength of this correlation .
in what follows we will suppose that @xmath74 is small , _
i.e. _ , the model is weakly dependent .
function @xmath146 is not arbitrary and must satisfy two consistency conditions . indeed , from the normalization of the densities @xmath44 and @xmath147 : @xmath148 immediately follows that @xmath149 on the other hand , plugging eq .
( [ wrho ] ) into eq .
( [ un_rho ] ) and taking again into account normalization we get @xmath150 we also observe that if @xmath151 and @xmath152 are integrable functions with respect to @xmath153 and @xmath154 respectively , then for consistency conditions , eqs .
( [ consistency_1])-([consistency_2 ] ) , to hold it suffices that : @xmath155 the starting point of our analysis is the renewal equation for the joint fourier - laplace transform of the propagator @xmath156 . substituting then eq . ( [ wrho ] ) into eq .
( [ nmrenewal_2 ] ) we have @xmath157 \tilde{p}(\omega , s|x',t')dx ' . \label{w_eq_1}\end{aligned}\ ] ] assuming that @xmath74 is small we look for a solution to this equation in the form @xmath158 in order to proceed further we need to know the dependence on @xmath74 of the propagator prior the first sojourn . from eqs .
( [ nmp0])-([nmpsi ] ) and eq .
( [ wrho ] ) we write @xmath159dx',\ ] ] but @xmath160 and using the consistency condition ( [ consistency ] ) we finally get @xmath161 hence @xmath57 is independent of @xmath74 .
one can easily see by substituting eqs .
( [ series ] ) and ( [ wp0 ] ) into eq .
( [ w_eq_1 ] ) that the zeroth - order @xmath162 corresponds to the independent case @xmath163 while for @xmath164 we have the recursive integral equations : @xmath165 where @xmath166 and @xmath167 in the appendix [ appendix_b ] we show that the solution to eq .
( [ int_eq_n ] ) is given by the recursive expression ( @xmath164 ) @xmath168\tilde{p}^{(n-1)}(\omega , s|x',t')dx ' , \label{recursive_n}\end{aligned}\ ] ] where @xmath169 and @xmath170 for @xmath171 we have @xmath172dx',\end{aligned}\ ] ] which , after defining @xmath173 and ( cf
( [ g ] ) ) @xmath174 can be written as @xmath175\tilde{q}(\omega , s).\ ] ] therefore , the joint fourier - laplace transform of the propagator up to first order in @xmath74 is @xmath176+{\rm o}(\epsilon^2)\biggr\}\frac{\hat{\psi}(s)}{1-\tilde{\rho}(\omega , s)}. \label{first_order}\ ] ] from the above expression of the conditional propagator we can also get the unconditional propagator defined by @xmath177 from eq .
( [ first_order ] ) and taking into account the normalization of @xmath178 we have @xmath179+{\rm o}(\epsilon^2)\biggr\}\frac{\hat{\psi}(s)}{1-\tilde{\rho}(\omega , s)}.\ ] ] but from eqs .
( [ consistency_2 ] ) and ( [ tilde_g ] ) one can easily see that @xmath180 hence @xmath181\frac{\hat{\psi}(s)}{1-\tilde{\rho}(\omega , s)}. \label{uncond_1}\ ] ] we will finally present an instrumental example of the use of the perturbation technique just developed . we assume , as in sect .
[ sec3b ] , that the joint density @xmath54 factorizes as ( cf .
( [ independent_0 ] ) and ( [ conditional_jump ] ) ) @xmath182 , \label{independent_0b}\ ] ] where @xmath183 and @xmath104 is an even function of @xmath88 .
contrary to the solvable case discussed in sect .
[ sec3b ] in which the correlation @xmath80 depends solely on the signs of consecutive jumps ( cf .
( [ g_sign ] ) ) , we now assume that the correlation depends also on jump sizes .
we thus suppose @xmath184 ( @xmath185 ) . with this correlation
we will evaluate the expressions for the propagators @xmath107 and @xmath105 as given respectively by eqs .
( [ first_order ] ) and ( [ uncond_1 ] ) for their joint fourier - laplace transform up to first order in @xmath74 . to this end
we need the explicit expressions for the auxiliary quantities @xmath186 and @xmath187 which appear in eqs .
( [ first_order ] ) and ( [ uncond_1 ] ) .
using eqs .
( [ independent_0b ] ) and ( [ corr_pet ] ) , the expression for @xmath186 defined in eq .
( [ tilde_g ] ) can be written as @xmath188 but taking into account the symmetry of @xmath75 expressed by eq .
( [ even_h ] ) we can write @xmath189 hence @xmath190 where @xmath191 is the fourier sine transform of @xmath192 . proceeding in an analogous way we see that the expression for @xmath187 defined in eq .
( [ tilde_g ] ) is given by @xmath193 ^ 2.\ ] ] but as we have just seen ( cf .
( [ h_s ] ) ) @xmath194 whence @xmath195 substituting eqs .
( [ tilde_g(2 ) ] ) and ( [ tilde_g(2 ) ] ) into eq . ( [ first_order ] )
yields @xmath196 + { \rm o}(\epsilon^2)\biggr\}\frac{\hat{\psi}(s)}{1-\tilde{\psi}(s)\tilde{h}(\omega)}. \ ] ] for poissonian sojourns
( [ poissonian ] ) ) we can take the inverse laplace transform of this expression which yields the conditional characteristic function : @xmath197 } + \epsilon\biggl\{2i\frac{\tilde{h}_{\rm s}(\omega , a)}{\tilde{h}(\omega)}\bigl[-e^{-\lambda t}+ e^{-\lambda t[1-\tilde{h}(\omega)]}\bigr]e^{-a|{\xi}|}{{\rm sgn}}({\xi } ) \nonumber \\ & -&4\biggl[\frac{\tilde{h}_{\rm s}(\omega , a)}{\tilde{h}(\omega)}\biggr]^2 \bigl[e^{-\lambda t}+[\lambda t \tilde{h}(\omega)-1]e^{-\lambda t[1-\tilde{h}(\omega)]}\bigr]\biggr\ } + { \rm o}(\epsilon^2 ) .
\label{first_order(3)}\end{aligned}\ ] ] the unconditional characteristic function can be analogously obtained through eqs .
( [ uncond_1 ] ) and ( [ tilde_g(2 ) ] ) or else directly by substituting eq . ( [ first_order(3 ) ] ) into @xmath198 by either way one chooses the final result is @xmath199 } -4\epsilon\biggl[\frac{\tilde{h}_{\rm s}(\omega ,
a)}{\tilde{h}(\omega)}\biggr]^2 \bigl[e^{-\lambda t}+[\lambda t \tilde{h}(\omega)-1]e^{-\lambda t[1-\tilde{h}(\omega)]}\bigr ] + { \rm o}(\epsilon^2 ) . \ ] ] for this example the variance of the process , @xmath200 , is easily seen to be @xmath201 where @xmath202
we have presented a generalization of the ctrw which include correlations between consecutive sojourns and jumps .
we have derived the general equations governing the time evolution of the dependent walk and we have exactly solved them in some particular instances .
we have also developed a general perturbation technique aimed to treat , within any desired degree of accuracy , weakly dependent models .
that is , those models in which there is a low correlation between consecutive events . due to the extensive analytical apparatus and technical aspects contained in this paper , which may obscure the main objective and perhaps discourage potential users of the technique presented , we shall now summarize the key expressions of our development .
the model is based on a two - dimensional markov series of jumps and sojourns , with a conditional joint density @xmath203 defined in eq .
( [ rho_mark ] ) .
the main objective of ctrw is obtaining the so - called propagator @xmath46 , that is , the ( conditional ) probability density function of the process @xmath0 , provided we know the value of the last jump size , @xmath204 , and waiting time , @xmath154 .
the propagator obeys a renewal equation ( cf .
( [ nmrenewal ] ) ) and its fourier - laplace transform satisfies the integral equation ( [ nmrenewal_2 ] ) .
we have been able to find an exact solution to this equation when the joint density has the following form ( cf .
( [ independent_0 ] ) and ( [ conditional_jump ] ) ) @xmath205 , \label{c2}\ ] ] for which the correlation between jumps depend on whether they are increasing or decreasing but not on their magnitude .
in such a case the fourier - laplace transform of the unconditional propagator is given by ( cf .
( [ fl_uncond ] ) ) @xmath206 in the case of poissonian sojourns and laplacian jumps we can invert this expression and obtain the propagator @xmath105 for different values of @xmath74 ( see eqs .
( [ explicit])-([explicit_0 ] ) ) . from these expressions
we can see some interesting properties due to the existence of correlations such are the transitions from unimodal to bimodal distributions ( cf .
[ fig1 ] and fig .
[ fig2 ] ) .
although we have been able to solve eq .
( [ nmrenewal_2 ] ) in the special case provided by eq .
( [ c2 ] ) , a general solution to the problem for any form of the joint density @xmath54 seems to be out of reach .
however , in many practical situations the degree of dependence between current and past events is weak .
in such cases it is possible to derive a perturbation technique which allows for an approximate solution to the above equation to any desired degree of accuracy .
we thus write @xmath207,\ ] ] where @xmath208 is the unconditional joint density , @xmath209 indicates correlation and @xmath74 , now a small quantity , measures the strength of such a correlation .
the function @xmath146 is not arbitrary and
must obeys some consistency conditions ( cf . eq .
( [ consistency ] ) ) in order to keep the normalization of the @xmath54 s .
the propagator can be written in the form of an infinite series @xmath210 where @xmath162 is the propagator when no correlation is present , that is , it corresponds to the propagator of the independent ctrw and is given by eq .
( [ solution_0 ] ) .
the rest of terms @xmath211 ( @xmath164 ) obey the integral equation eq .
( [ int_eq_n ] ) whose solution is given by eq .
( [ recursive_n ] ) , what allows us to compute @xmath211 if we know @xmath212 .
obviously by repeating this operation one can obtain @xmath213 for any @xmath164 and , hence , an approximate expression for the propagator to any desired degree of accuracy although , in many cases , the lowest order @xmath171 will suffice .
in such a case the explicit expression for the propagator is ( cf .
( [ tilde_g])-([first_order ] ) ) @xmath214 + { \rm o}(\epsilon^2)\biggr\}\frac{\hat{\psi}(s)}{1-\tilde{\rho}(\omega , s)},\ ] ] and for the unconditional propagator defined in eq .
( [ uncond_def ] ) we have @xmath181\frac{\hat{\psi}(s)}{1-\tilde{\rho}(\omega , s)}.\ ] ] we end this work by recalling that our first motivation to treat the problem of dependent ctrw s arose from our work in econophysics . in dealing with extreme time statistics of financial time series , in particular with the mean exit times of the process out of a given interval , we noticed that the observed behavior can not be properly described by the traditional ( _ i.e. , _ independent ) ctrw but one needs some degree of correlation between present and past events @xcite . in spite of this specific motive
we certainly believe that a general development of the dependent ctrw at least for a markovian joint density may be of broad interest because the independence assumption in the traditional ctrw is just a first approximation for many physical phenomena that are amenable to be studied within the ctrw framework @xcite . in any case in forthcoming works we will apply it to financial time series .
the authors acknowledge partial support from direccin general de investigacin under contract no .
fis2006 - 05204 .
we will solve the recursive integral equation ( [ int_eq_n ] ) : @xmath215 where @xmath164 . to this end
we multiply both sides of eq .
( [ b1 ] ) by @xmath216 and integrate over @xmath88 and @xmath154 , we obtain @xmath217 where @xmath218 is the joint fourier - laplace transform of @xmath178 , @xmath219 and @xmath220 setting @xmath221 and @xmath222 in eq . ( [ b2 ] ) we get @xmath223 which introduced back to eq .
( [ b2 ] ) yields @xmath224 from the definition of @xmath225 given in eq .
( [ b3 ] ) we see that the fourier - laplace inversion with respect to @xmath226 and @xmath227 of this quantity is @xmath228 and a similar expression for the inversion of @xmath229 .
therefore , the inversion of eq .
( [ b4 ] ) reads @xmath230 by combining the definition of @xmath231 , given in eq .
( [ b3b ] ) when @xmath221 and @xmath222 , with that of @xmath232 given in eqs .
( [ q^n])-([q^0 ] ) , we write @xmath233 where @xmath234 finally , substituting eq .
( [ q^n ] ) and eq .
( [ b6 ] ) into eq .
( [ b5 ] ) we obtain the recursive solution given in eq .
( [ recursive_n ] ) .
999 e. w. montroll and g. h. weiss , j. math . phys .
* 6 * , 167 ( 1965 ) .
e. w. montroll and m. f. shlesinger in _ nonequilibrium phenomena ii : from stochastics to hydrodynamics _ , j. l. lebowitz and e. w. montroll eds .
( north - holland , amsterdam , 1984 ) .
g. h. weiss , _ aspects and applications of the random walk _ ( north - holland , amsterdam , 1994 ) .
b. berkowitz and h. scher , phys .
lett . * 79 * , 4038 ( 1997 ) .
m. bogu and .
corral , phys .
lett . * 78 * , 4950 ( 1997 ) .
e. gudowska - nowak and k. weron , phys .
e * 65 * , 011103 ( 2001 ) .
a. helmstetter and d. sornette , phys .
e * 66 * , 061104 ( 2002 ) .
. corral , phys .
lett . * 97 * , 178501 ( 2006 ) .
b. berkowitz , g. kosakowski , g. margolin , and h. sher , ground water * 39 * , 593 ( 2001 ) .
m. dentz and b. berkowitz , water resources research * 39 * , 1111 ( 2003 ) .
p. grigolini , l. palatella , and g. raffaelli , fractals * 9 * , 439 ( 2001 ) .
r. kutner , comp .
. comm . * 147 * , 565 ( 2002 ) ; chem . phys .
* 284 * , 481 ( 2002 ) .
e. scalas , r. gorenflo , and f. mainardi , physica a * 284 * , 376 ( 2000 ) ; f. mainardi , m. raberto , r. gorenflo , and e. scalas , physica a * 287 * , 468 ( 2000 ) .
m. raberto , e. scalas , and f. mainardi , physica a * 314 * , 749 ( 2002 ) .
e. scalas , r. gorenflo , and f. mainardi , phys .
e * 69 * , 011107 ( 2004 ) .
e. scalas , physica a * 362 * , 225 ( 2006 ) .
r. kutner and f. switaa , quantitative finance * 3 * , 201 ( 2003 ) .
j. masoliver , m. montero , and g. h. weiss , phys .
e * 67 * , 021112 ( 2003 ) .
p. repetowicz and p. richmond , physica a * 344 * , 108 ( 2004 ) .
j. masoliver , m. montero , j. perell , and g. h. weiss , j. econom .
behav . organ . * 61 * , 577 ( 2006 ) .
j. masoliver , m. montero , j. perell , and g. h. weiss , physica a * 379 * , 151 ( 2007 ) .
d. r. cox and h. d. miller , _ the theory of stochastic processes _ ( wiley , new york , 1965 ) .
g. udny yule , philos . trans .
sci . * 213 * , 21 ( 1925 ) .
w. feller , _ an introduction to probability theory and its applications _
( wiley , new york , 1950 ) . m. montero , j. perell , j. masoliver , f. lillo , s. miccich , and r. n. mantegna , phys .
rev e * 72 * , 056101 ( 2005 ) .
. corral , tectonophysics * 424 * , 117 ( 2006 ) .
l. krapivsky and s. redner , am .
* 72 * , 591 ( 2004 ) .
g. e. roberts and h. kaufman , _ table of laplace transforms _ ( sauders , philapelphia , 1966 ) .
p. gopikrishnan , v. plerou , l. a. n. amaral , m. meyer , and h. e. stanley , phys .
e * 60 * , 5305 ( 1999 ) ; v. plerou , p. gopikrishnan , l. a. n. amaral , m. meyer , and h. e. stanley , phys .
e * 60 * , 6519 ( 1999 ) .
a. c. silva and v. m. yakovenko , physica a * 382 * , 278 ( 2007 ) . | the usual development of the continuous time random walk ( ctrw ) assumes that jumps and time intervals are a two - dimensional set of independent and identically distributed random variables . in this paper
we address the theoretical setting of non - independent ctrw s where consecutive jumps and/or time intervals are correlated .
an exact solution to the problem is obtained for the special but relevant case in which the correlation solely depends on the signs of consecutive jumps . even in this simple case some interesting features arise such as transitions from unimodal to bimodal distributions due to correlation .
we also develop the necessary analytical techniques and approximations to handle more general situations that can appear in practice . |
the mathematical modeling of the spread of epidemics is a subject of continuing theoretical and practical interest @xcite .
this is enhanced by the fact that the same or similar models are used for describing other phenomena such as plant and animal dispersal , and successional dynamics in ecology @xcite .
the level of description provided by a model can be purely macroscopic and deterministic or individual and stochastic @xcite . in the first case one
uses ( partial- ) differential equations to describe the time evolution of different subpopulations ; e.g. , susceptible , infectious and recovered . in the second case one typically uses stochastic dynamics on a lattice ( or more general graphs ) where the variables at each node represent the state of an individual or a small spatial region .
the time evolution of these variables is stochastic , e.g. , an infected individual at site @xmath0 has a certain probability per unit time ( rate ) @xmath1 to infect a susceptible individual at a neighboring site @xmath2 .
these systems fall into the category of what mathematicians call interacting particle systems @xcite and physicists call stochastic lattice gases @xcite - systems of great interest also in the study of equilibrium phase transitions , phase segregation kinetics , etc .
, fields very different from epidemiology and ecology . the connection between these modes of description and various intermediate ones has been investigated extensively in recent years , e.g. , see @xcite .
mathematically this involves the use of the so called hydro - dynamical scaling limit .
this uses a rigorous separation of space and time scales to derive deterministic macroscopic equations from the microscopic dynamics of stochastic lattice systems .
other approaches are based on more heuristic methods such as the mean field approximation ( mfa ) and improvement thereof @xcite the present work falls in the latter category .
we apply a pair approximation ( pa ) scheme to a microscopic stochastic epidemic model in which individuals recovered from an infection enjoy a period of immunity before again becoming susceptible at a rate @xmath3 : the sirs model .
the pa approximation was used by durrett and levin @xcite for the simpler susceptible - infected - susceptible ( sis ) model where recovered individuals immediately become susceptible again .
they compared the results of the pa and mfa with those of the stochastic sis model and found that the pa gave a quantitative improvement over the mfa . here
we consider the general sirs model .
we obtain the behavior of the stochastic model from extensive computer simulations .
we then solve the pa and mfa models analytically for the stationary state and numerically for the time dependent case .
we find that the pa gives considerably better agreement with the simulations than the mfa both for the time evolution and for the steady state . for the latter
the pa reproduces the qualitative difference between the one and higher dimensional phase diagram of this model found in ref .
this is reminiscent of the relation between the mfa and the bethe - peierls approximation ( which the pa closely resembles ) for equilibrium lattice systems @xcite .
we first recall the stochastic lattice model of the sirs epidemic process @xcite . a site @xmath4 of a @xmath5-dimensional lattice
can be occupied by an individual in a state of @xmath6 ( healthy and susceptible ) , @xmath7 ( infected ) , or @xmath8 ( recovered , i.e. , healthy and immune ) .
the system evolves according to the following transition rates , @xmath9 where @xmath10 is the number of infected ( nearest ) neighbors of @xmath4 , @xmath1 is the infection rate , @xmath11 is the recovery rate and @xmath3 is the rate at which immunization ceases .
the limit @xmath12 corresponds to the case where a recovered site passes instantaneously through the state @xmath8 ; this is the sis model , also known as the contact process .
we shall choose time units in which @xmath11=1 .
one can obtain some rigorous qualitative information about this and related models via probabilistic approaches such as those used in interacting particle systems @xcite .
of particular interest is the behavior of the stationary state on an infinite lattice which is a good approximation for the quasi - steady state behavior of large systems : see appendix [ mcsimulation ] . .
this information is encoded in the phase diagram of the stationary state which depends on the infection rate @xmath1 , the recovery rate @xmath3 and the topology of the lattice . for small @xmath1 ,
the only stationary state is one in which all sites are in the susceptible ( disease - free ) state while for large @xmath1 there is ( for the infinite system ) also a stationary state containing non zero fraction of @xmath7 and @xmath8 individuals . the critical infection rate @xmath13 is defined as the smallest value of @xmath1 , for a given @xmath3 , above which the infection can persist forever . for the sis or contact process ( @xmath14 ) , the critical infection value is known with high accuracy , @xmath15 in @xmath16 and @xmath17 in @xmath18 @xcite . considerably less
is known about the phase diagram of the sirs model .
interestingly there is a qualitative difference in the behavior of @xmath13 in one and in higher dimension when @xmath19 .
it has been shown that @xmath20 is finite when @xmath21 while @xmath22 when @xmath16 @xcite . to go beyond qualitative results we need to carry out simulation or make some approximations .
this is the subject of the rest of the paper .
the time evolution of the single site probabilities in the stochastic sirs epidemic process can be written in the following form .
[ allmaster ] @xmath23 here @xmath24 is the neighborhood ( nearest neighbor sites ) of a site @xmath4 , @xmath25 is the probability of having a state @xmath26 at site @xmath4 at time @xmath27 and @xmath28 is the joint probability to have state @xmath26 at site @xmath4 and state @xmath29 at site @xmath30 , at time @xmath27 .
we always have @xmath31 .
( [ mastera])- ( [ masterc ] ) are , as is usual for moment equations , not a closed system .
one can extend them by including equations for the time evolution of @xmath32 which in turn involve higher moments of the spatial correlations .
this leads to an infinite hierarchy . to solve such a hierarchy one usually resorts to some approximation scheme which expresses the higher order moments in terms of the lower order ones and truncates the equations at some point ; this is referred to as the moment closure method @xcite .
both the mfa and pa are such schemes . in the mfa eqs .
( [ mastera])-([masterc ] ) are closed by assuming that @xmath33 , i.e. , it neglects correlations between different sites .
this leads to a pair of coupled equations which have been studied in @xcite . in the pa scheme @xmath25 and @xmath34
are kept as unknowns while the higher - order moments are expressed , via some appropriate approximation , in terms of these quantities . to carry out the pa we complement eq .
( [ allmaster ] ) by equations for the second moments @xmath28 for nearest neighbor sites @xmath4 and @xmath30 based on the transition rule that we have described in eq .
( [ transition ] ) .
these are [ alltwomoment ] @xmath35 where @xmath36 is the set of nearest neighbor sites of @xmath30 excluding the site @xmath4 .
@xmath37 is the joint probability to have state @xmath26 at site @xmath4 , state @xmath29 at site @xmath30 and state @xmath38 at site @xmath39 at time @xmath27 . for a derivation of eq .
( [ alltwomoment ] ) see appendix [ derivation_twomoment ] . to close the system ( [ allmaster ] ) and ( [ alltwomoment ] ) and
derive a set of autonomous equations for @xmath25 and @xmath28 we approximate the triad joint probability @xmath37 for @xmath4 and @xmath39 nearest neighbors of @xmath30 , by the product of two pair probabilities @xmath28 and @xmath40 divided by the probability @xmath41 @xcite , i.e. , we set @xmath42 note that we have made use here of the structure of the hypercubic lattice .
in such lattices , three adjacent sites , @xmath43 can not form a triangle but form only linear chains .
this is not so in other lattices , e.g. , the triangular lattice , where other configurations need also be considered . while there are other choices for a pa , the approximation in eq .
( [ mypa ] ) allows one to get the steady state solutions analytically . with other pair approximations @xcite
, one has to solve the resulting differential equations numerically , making it impossible to obtain analytic expressions for the critical curve . to actually carry out computations with the pa we will assume from now on that our system is spatially uniform
the site @xmath4 , in eqs .
( [ allmaster ] ) and ( [ alltwomoment ] ) , can now be chosen to be the origin .
we also define : @xmath44 , @xmath45 where @xmath46 is the number of nearest neighbors of a site in the @xmath5-dimensional cubic lattice .
the truncated equations for the pa - sirs can now be written , by using the exact eqs .
( [ allmaster])- ( [ alltwomoment ] ) and the approximate eq .
( [ mypa ] ) , as a closed set of five coupled equations , [ allpasirs ] @xmath47 note that we always have @xmath48 which determines @xmath49 and @xmath50 . in the limit @xmath12 , @xmath51 and @xmath52 as well as their time derivatives will go to zero .
this yields @xmath53 and @xmath54 @xcite . in this limit
( [ allpasirs ] ) reduces to the pa equations of the sis considered in @xcite , [ pasis ] @xmath55 as already noted the mfa approximates the joint probability @xmath56 in eq .
( [ pasirsa ] ) by the product , @xmath57 .
this leads to the closed set of mfa of equations for the sirs @xcite , [ mfpasirs ] @xmath58 for @xmath12 , @xmath59 and @xmath60 .
( [ mfpasirs ] ) then reduces to the mfa for the sis .
let us first consider the steady state solutions of the pa - sis obtained by setting the l.h.s of eq .
( [ pasis ] ) equal to zero @xcite .
this gives for the critical value of the pa - sis epidemic process @xmath61 .
for @xmath62 , both @xmath63 and @xmath56 @xmath64 as @xmath65 for all initial states .
when @xmath66 there is , in addition to the disease - free state corresponding to @xmath67 , also a stationary state consisting of a finite fraction of infected individuals : [ stationarysis ] @xmath68}{z(z-1)\lambda-1}. \label{stationarysisb } \end{aligned}\ ] ] it is these non - zero steady states which are approached as @xmath65 when starting from any initial state with @xmath69 .
the steady state solutions of the pa - sirs system is obtained by setting the l.h.s . of eq .
( [ allpasirs ] ) equal to zero .
setting @xmath70 this yields , [ steadypasirs ] @xmath71 where @xmath72 are the approximate probabilities for having states @xmath26 and @xmath29 on neighboring sites .
after further simplifications , we find that @xmath4 has to satisfy the cubic equation , @xmath73 both the derivation of eq .
( [ pequation ] ) and the explicit expressions for @xmath74 , @xmath75 and @xmath76 as functions of @xmath1 and @xmath3 are given in appendix [ pcoefficients ] .
the root @xmath77 corresponds to the all healthy steady state , which is always a solution .
the critical curve @xmath13 is determined by the existence of a root of eq .
( [ pequation ] ) such that @xmath4 and all other stationary probabilities , are strictly positive .
it turns out that this strictly positive root is unique .
thus when @xmath78 , @xmath77 is the only steady state solution . for @xmath79 , there is also a steady state in which the infection is endemic : @xmath80 and @xmath81 , see appendix [ pcoefficients ] .
the critical curve @xmath13 is obtained in appendix [ pcoefficients ] .
it is given by the equation , @xmath82 as @xmath12 , @xmath83 , the critical point of the pa - sis epidemic process .
on the other hand , as @xmath3 approaches zero , the critical curve shows different behavior depending on the dimension of the lattice : @xmath84 diverges to infinity for @xmath16 , while @xmath84 is finite for @xmath85 .
the pa thus reproduces the qualitative difference between the one and higher dimensional phase diagram of the sirs model found in ref .
@xcite . the mfa , eq .
( [ mfpasirs ] ) , yields the mean field critical value , @xmath86 independent of @xmath3 . in the coexistence region
@xmath87 the mean field stationary states are @xmath88 and @xmath89 .
both the steady state and critical value of the mfa and pa fail to correctly represent the results of the stochastic sirs process for small @xmath3 : see figs .
[ fig1 ] and [ fig2 ] .
note in particular that @xmath90 of the stochastic sirs process is considerably larger than that of the mfa or pa for large @xmath1 and small @xmath3 .
this is due to the fact that the susceptible sites can be surrounded by recovered ones and thus protected from contacting infected ones in the stochastic case .
we compare in figs . [ fig3 ] - [ fig6 ] the steady state values of @xmath91 and @xmath72 obtained from the mfa and pa with the results from the stochastic sirs process as a function of @xmath1 at fixed values of @xmath3 .
clearly the pa gives results closer to those obtained from the stochastic model .
for the methods used to obtain the steady state results from the numerical simulation , see appendix .
[ mcsimulation ] . figs .
[ fig3 ] and [ fig5 ] show that both the mfa and pa overestimate @xmath92 as well as @xmath93 , @xmath94 .
this is due to the strong tendency of infected sites in the stochastic model to cluster into localized islands , reducing the contacts between @xmath6 and @xmath7 .
this is partially taken into account by the pa as seen by the behavior of @xmath95 and @xmath96 in figs .
[ fig3 ] and [ fig5 ] .
this clustering effect is also observed in the stochastic sis process @xcite .
it is more pronounced in one dimension .
note that @xmath95 becomes zero both at @xmath97 , when @xmath98 , and at @xmath99 when @xmath100 , reaching a peak at a positive value of @xmath1 which depends on @xmath3 . for large values of @xmath3 ,
the steady state values of @xmath91 and @xmath72 obtained from the pa , or the mfa agree well with the numerical simulation , away from the critical @xmath13 . moreover the pa yields steady state curves remarkably similar to those from the numerical simulation , see figs . [ fig4 ] and [ fig6 ] .
to study the stability of the stationary pa state , eq .
( [ allpasirs ] ) is linearized about the steady state values @xcite , see appendix [ pajacobian ] .
this leads to the study of the roots of the characteristic fifth order polynomial @xmath101 , obtained from @xmath102 where @xmath103 is the jacobian of the linearized pa - sirs system .
if @xmath104 , the solution of the linearized equation is stable , i.e. , a small perturbation around the steady state will decay back to the steady state .
we used the routh - hurwitz conditions @xcite to obtain the sign of the real part of eigenvalues of the jacobian . as expected , the positive steady state solution is stable for @xmath105 .
the zero steady state solution is stable for @xmath78 and unstable for @xmath106 .
the eigenvalues of @xmath101 have non - zero imaginary parts in some regions of the parameter space . in such regions
the pa - sirs system in eq .
( [ allpasirs ] ) will converge to the steady state in a damped oscillatory manner .
such oscillations are seen in fig .
[ fig7 ] and [ fig8 ] .
to study the time evolution of an epidemic following an initial infection of a healthy population we performed dynamical monte carlo simulations @xcite as well as solutions of eqs .
( [ allpasirs ] ) and ( [ mfpasirs ] ) . for the stochastic evolution we started with infected sites placed either randomly or in a cluster and followed the time evolution averaged over @xmath107 realizations of the sirs process . to obtain the time evolution of the mfa and
pa we solved eq .
( [ mfpasirs ] ) and eq .
( [ allpasirs ] ) numerically by using a 4th - order runge - kutta method .
we plot the results in figs .
[ fig7 ] and [ fig8 ] . to set the unit of time of the simulation we started with a fully infected state , @xmath108 and @xmath109 and obtained the exponentially decaying pattern of @xmath63 .
we then set the slope ( death rate ) of the graph , @xmath110 vs @xmath27 , from the numerical simulation equal to those from the mfa and pa .
starting with a small value of @xmath111 , @xmath63 displays an initial `` exponential '' growth in both the mfa and pa .
similar growth patterns are observed in all @xmath112 , @xmath113 .
this is explained by the initially abundantly available susceptible population .
once the susceptible population is reduced , the infected population reaches a maximum and then decreases to the steady state endemic level .
note the damped oscillatory pattern in figs .
[ fig7 ] and [ fig8 ] for this choice of the parameters ( @xmath1,@xmath3 ) .
the numerical simulation of the stochastic time evolution does not show the pronounced growth patterns of the pa and mfa when the initial fraction of infected sites is small , as seen in fig .
the formation of clusters of infected sites makes the infected population grow more slowly in the stochastic model . when the initial fraction of infected population increases to more than one percent the stochastic model shows significant change in its growth pattern , becoming similar to the pa and mfa .
if however the same fraction of infected sites are initially placed in a single cluster the stochastic epidemic process exhibits slower growth patterns , similar to those starting with a small fraction of initially infected sites .
these studies confirm that the clustering of infected sites in the stochastic model reduces both the speed of growth and the maximum fraction of infected sites . in realistic situations
the population is not well mixed so we would expect growth patterns more similar to that of the stochastic epidemic model , starting with a fraction of infected sites initially placed in a single cluster .
we investigated the stochastic sirs epidemic process and compared the results with those obtained from the deterministic mfa and pa .
these approximations close the hierarchy of dynamical equations by expressing the higher order moments in terms of the lower order ones .
the pa is found to improve over the mfa both for the stationary and for the time dependent states .
the time evolution of the system shows damped oscillatory behavior in some parameter ranges .
work supported by nsf dmr-01 - 279 - 26 and by afosr af 49620 - 01 - 1 - 0154 and by dimacs grants nsf dbi 99 - 82983 and nsf eia 02 - 05116 .
eq . ( [ twomomenta ] ) is derived by considering all transitions leaving or entering the pair configuration @xmath114 .
we list them as follows : a pair @xmath115 changes to a pair @xmath114 with a rate @xmath3 .
a pair @xmath114 changes to a pair @xmath116 with a rate @xmath1 and also changes to a pair @xmath117 with a rate 1 .
a triad configuration @xmath118 transits to a triad @xmath119 with a rate @xmath1 such that a pair configuration @xmath120 is changed to @xmath114 .
a triad @xmath121 changes to a triad @xmath122 with a rate @xmath1 .
the equations for @xmath123 and @xmath124 in eq .
( [ alltwomoment ] ) can be obtained in a similar way .
the relation @xmath125 can be used to obtain the other joint probabilities @xmath34 which are not shown in eq .
( [ alltwomoment ] ) .
the steady states in eq .
( [ steadypasirs ] ) are obtained by setting the l.h.s . of eq .
( [ pasirsa])-([pasirsd ] ) equal to zero .
in addition we set eq .
( [ pasirse ] ) equal to zero and replace a single site and joint probabilities with the steady states in eq .
( [ steadypasirs ] ) .
after simplifications , we obtain eq .
( [ pequation ] ) with the coefficients , @xmath126 the critical curve @xmath13 is given by setting @xmath127 .
only for @xmath105 does the quadratic factor of eq .
( [ pequation ] ) have a positive root .
the jacobian of the linearized pa - sirs is written , @xmath128 where @xmath129 , @xmath130 , @xmath131 , and @xmath132 .
the jacobian matrix b of linearized mf - sirs is given by @xcite @xmath133 the characteristic polynomial of the second order , @xmath134 , is obtained from @xmath135 .
the necessary and sufficient ( routh - hurwitz ) conditions @xcite for @xmath136 is @xmath137 and @xmath138 . in the coexistence region where @xmath139 , @xmath140 and @xmath141 for all @xmath142 . in the no - coexistence region where @xmath143 , @xmath144 and @xmath145 for all @xmath142 .
both in the coexistence and no - coexistence region , the real part of the eigenvalues is negative and thus the mean field steady states are stable .
now we turn our attention to the oscillatory behavior .
the eigenvalues of the characteristic polynomial @xmath146 is given by , @xmath147 in the range of @xmath148 the imaginary part of the eigenvalues is non - zero : @xmath149 . in this range of @xmath1 , the steady states correspond to the stable spiral and the system converges to the steady state in damped oscillatory pattern . even in the damped oscillatory region , any oscillation is hardly visible in the large @xmath3 limit and becomes noticeable only in small @xmath3 limit .
the numerical simulations described here used lattices with periodic boundary conditions . in one dimension ,
rings of @xmath150 sites were used . in two dimensions ,
torii of @xmath151 sites were employed . to obtain the steady state of the sirs process a random initial configuration of susceptible and infected sites
is evolved according to the transition rates in eq .
( [ transition ] ) . in practice a site
is randomly chosen and a random number ( @xmath152 $ ] ) is also chosen : if it is greater than the given transition probability for that site , which is equal to the rate@xmath153 , its state is updated : @xmath154 is chosen to be so small that transition probability is not greater than 1 for a range of ( @xmath155 ) @xcite .
otherwise its state remains the same . for a finite system
the only true stationary state of the sirs process is the absorbing state corresponding to @xmath156 , @xmath157 . to learn about the active state from simulations of a finite system we study the quasi - stationary state .
these are determined from averages over the surviving representatives of @xmath107-@xmath158 independent realizations of the sirs process with the same parameter @xmath159 , beginning with random initial distribution of the @xmath7 s .
surviving sample averages converge to stationary values as @xmath160 . to obtain the steady states and critical curve
we extrapolated quasi - stationary values of finite systems to those of the infinite system .
the finite size scaling theory @xcite can be used to obtain the critical curve @xmath161 .
we can assume a scaling function of the surviving probability : @xmath162 .
at criticality , @xmath163 , the survival probability of the infection , starting from a single infected site , has a power law behavior in time . in the subcritical region
, it decays exponentially while in the supercritical region it reaches non - zero steady state in a short time .
the power law behavior of the survival probability at criticality enables one to extract the critical curve @xmath161 from the time evolution data of the sirs process .
this dynamical monte carlo simulation is reliable when the system size is sufficiently large so that the evolution of the system is approximately confined , for the duration of the simulation to a region smaller than the size of the system @xcite .
however we found that this surviving probability oscillates wildly when @xmath3 is small .
because of this the dynamical monte carlo method is not used to determine the critical curve near @xmath164 .
phase diagram of the sirs process in two dimensions .
the coexistence phase of @xmath6-@xmath7-@xmath8 and the no - coexistence phase are separated by the critical curve from the simulation ( open circles with dotted line for eye - guidance ) , the pa ( thick solid line ) and the mfa ( long dashed line ) .
the critical curve is obtained on periodic square lattice of different sizes @xmath165 from simulations extrapolated to infinite system : @xmath166 .
, width=377,height=377 ] phase diagram of the sirs process in one dimension .
the critical curve from numerical simulations of ring lattice of different sizes @xmath165 is extrapolated to infinite system : @xmath167 .
the same symbols are used as in the fig .
, width=377,height=377 ] first and second order moments of the steady state sirs in two dimensions at @xmath168 .
the steady - state values of the density of infection in fig .
[ fig3](a ) and the second moments in fig .
[ fig3](b)-(f ) are drawn from the numerical simulation ( open circle with dotted line for eye - guidance ) , the pa ( thick solid line ) , and the mfa ( long - dashed line ) .
for the numerical simulation we used a system of size @xmath169.,width=377,height=377 ] time - evolution of the first and the second order moments of the sirs process in two dimensions .
all sub - graphs are from numerical simulations ( open circles ) , the pa ( solid line ) , and the mfa ( dashed line ) at @xmath168 and @xmath173 .
a periodic square lattice of @xmath174 sites is used in the numerical simulations averaged over @xmath175 realizations starting with random initial distribution with 1 percent of infected sites . ,
width=377,height=377 ] time - evolution of fraction of infected sites of the sirs process in two dimensions at @xmath168 and @xmath173 .
a periodic square lattice of @xmath169 is used in numerical simulation averaged over @xmath175 realizations .
main : simulation starts with 1@xmath176 of infected sites placed either randomly ( filled circles ) or in a single cluster ( open circles ) on a lattice .
both the pa and mfa takes an initial value 0.01 for @xmath111 .
inset : simulation starts with different fractions of infected sites randomly placed in a lattice : 0.1 , 1 and 5 percents of the system.,width=377,height=377 ] | we investigate the time - evolution and steady states of the stochastic susceptible - infected - recovered - susceptible ( sirs ) epidemic model on one- and two- dimensional lattices .
we compare the behavior of this system , obtained from computer simulations , with those obtained from the mean - field approximation ( mfa ) and pair - approximation ( pa ) .
the former ( latter ) approximates higher order moments in terms of first ( second ) order ones .
we find that the pa gives consistently better results than the mfa . in one dimension
the improvement is even qualitative . |
the galactic center , being the closest galactic nucleus at a distance of only 8.5kpc , contains @xmath1% of the entire galactic molecular gas within a @xmath0500pc region known as the central molecular zone ( cmz ; e.g. morris et al .
1996 ; oka et al . 1998 ; jones et al .
this gas is subject to a unique and extreme environment that is dominated by shocks from cloud - cloud collisions , large fluxes of uv photons and cosmic rays , turbulence , as well as strong magnetic and tidal fields .
these conditions are reflected in extreme physical parameters : large temperatures , densities , dispersions , and ionization fractions of the gas , and in a peculiar chemistry that is enhanced in abundances of complex molecules .
similar conditions are also found in high - z galaxies at cosmic times when the bulk of stars were created ( e.g. kruijssen & longmore 2013 ) . to understand the astrophysics of star formation processes in this environment
, one must have accurate measurements of these ism conditions , as all of these physical parameters determine the capability of the molecular gas to cool , collapse , and eventually to form stars .
of particular importance is the temperature of the gas , as it is a direct indicator of the efficiencies of shielding and cooling .
`` the survey of water and ammonia in the galactic center '' ( swag ; survey page at : https://sites.google.com/site/atcaswag ) is an approach to measure some of the most important quantities in the cmz .
swag data consists of @xmath2 spectral lines in the 21 - 26ghz range with 4ghz of continuum at @xmath00.8pc ( 20 ) spatial and 0.4kms@xmath3 spectral resolution .
the data are a good compromise between spatial resolution and surface brightness sensitivity , but the interferometric observations do filter out the diffuse gas on the largest scales .
the maps therefore reveal the clumpy , filamentary structures of the cmz in great detail with little confusion .
the swag lines include multiple ammonia transitions ( fig.[fig1 ] top ) , which are being used as a thermometer , structure , and dynamic tracer for the molecular gas .
water masers ( fig.[fig1 ] bottom ) are visible at sites where young stars currently form ( in addition to envelopes of evolved stars ) .
we also obtain a rich dataset of density , photo - dominated region and shock tracers ( including c - c@xmath4h@xmath5 , hnco , ch@xmath4oh ) , radio recombination lines to map the ionized components , and continuum maps to measure the strength , spectral index , and polarization of the emission . together , swag is a comprehensive survey to couple the physical state ( density , temperature , ionization , shocks , magnetic field strength ) of the molecular gas with the dynamical structure of each cloud embedded in the galactocentric streams .
swag will also reveal the heating / cooling balance and chemistry of the gas in diverse environments and relates those parameters to each cloud s evolutionary stage of star formation . | swag ( `` survey of water and ammonia in the galactic center '' ) is a multi - line interferometric survey toward the center of the milky way conducted with the australia telescope compact array .
the survey region spans the entire @xmath0400pc central molecular zone and comprises @xmath042 spectral lines at pc spatial and sub - km / s spectral resolution .
in addition , we deeply map continuum intensity , spectral index , and polarization at the frequencies where synchrotron , free - free , and thermal dust sources emit .
the observed spectral lines include many transitions of ammonia , which we use to construct maps of molecular gas temperature , opacity and gas formation temperature ( see poster by nico krieger et al . , this volume ) .
water masers pinpoint the sites of active star formation and other lines are good tracers for density , radiation field , shocks , and ionization .
this extremely rich survey forms a perfect basis to construct maps of the physical parameters of the gas in this extreme environment . |
the present paper investigates the asymptotic distribution of the incomplete gauss sum @xmath0 where @xmath1 , @xmath2 , and @xmath3 ; the weight function @xmath4 is periodic with period one .
the case @xmath5 corresponds to the classical gauss sum .
the main example of an incomplete gauss sum in the literature is the case when @xmath4 is the characteristic function of a subinterval of the unit interval @xcite .
it is natural to assume that @xmath6 and @xmath7 are coprime , i.e , @xmath8 . here
@xmath9 is the multiplicative group of integers mod @xmath7 .
the order of @xmath9 is denoted by @xmath10 ( euler s totient function ) .
if @xmath11 are not coprime , say @xmath12 for some @xmath13 , we set @xmath14 , @xmath15 , and observe that @xmath16 where @xmath17 the case when @xmath11 are not coprime can therefore be reduced to the coprime case . functional equations and pointwise estimates of incomplete gauss sums have been studied extensively by oskolkov @xcite , and the aim of the present paper is to complement his results by establishing limit theorems for their value distribution at random argument .
the existence of a limit distribution of the classical theta sum @xmath18 for @xmath19 uniformly distributed in @xmath20 has been proved by jurkat and van horne @xcite ( for its absolute value ) and the second author @xcite ( for its full distribution in the complex plane ) ; we refer the reader also to the recent study by cellarosi @xcite .
a striking feature of the limit distribution of theta sums is that it has a heavy tail : the probability that @xmath21 has a value greater than @xmath22 , decays , for large @xmath22 , as @xmath23 . at rational @xmath19
, the theta sum of course reduces to an incomplete gauss sum where @xmath4 is the characteristic function of an interval , and we will see below ( remark [ rem1 ] ) that in this case the limit distribution has compact support
the exact opposite of a heavy tail .
we denote by @xmath24 the fourier series of @xmath4 .
we will focus for the major part of this paper on gauss sums with differentiable weight functions @xmath4 in the space @xmath25 and only later extend our results to general riemann integrable functions , under an additional assumption on @xmath7 .
the limit distributions of incomplete gauss sums will be characterized by the following random variables : * @xmath26 takes the four values @xmath27 with equal probability .
* @xmath28 takes the values @xmath29 with equal probability . *
@xmath30 takes the values @xmath31 with equal probability .
* @xmath32 , @xmath33 , @xmath34 are random variables given by the fourier series @xmath35 respectively , with @xmath19 uniformly distributed on @xmath36 .
note that , for @xmath37 , the functions in are differentiable and thus continuous . if @xmath4 satisfies the functional relation @xmath38 , then its fourier coefficients are related via @xmath39 , and hence @xmath40 and @xmath41 .
if @xmath42 is real - valued for all @xmath43 , then @xmath44 and @xmath45 .
hence the probability density describing the distribution of the imaginary part of the random variables @xmath32 , @xmath33 , @xmath34 is symmetric .
furthermore , @xmath46 , and thus the real and imaginary part of @xmath34 have the same distribution .
we define @xmath47 if @xmath48 , and @xmath49 if @xmath50 . the symbol @xmath51 denotes convergence in distribution .
[ thm1 ] fix a subset @xmath52 with boundary of measure zero , and let @xmath37 . for each @xmath1 , choose @xmath53 at random with uniform probability . then , as @xmath54 along an appropriate subsequence as specified below , we have : [ cols="^,^,^",options="header " , ] the following technical estimate allows us to extend theorem [ thm1 ] to non - smooth @xmath4 as long as @xmath7 has a bounded number of divisors @xmath55 .
[ rilem ] fix a positive integer @xmath56
. then there exists a constant @xmath57 such that , for every riemann integrable function @xmath58 , we have @xmath59 lemma [ rilem ] is proved in section [ variasec ] .
together with chebyshev s inequality it implies the following extension of theorem [ thm1 ] .
[ thm2 ] fix a subset @xmath52 with boundary of measure zero , and let @xmath58 be riemann integrable
. then the conclusions of theorem [ thm1 ] and corollary [ cor1 ] remain valid for any sequence of @xmath54 with a bounded number of divisors .
the proof of theorem [ thm2 ] is supplied in section [ extens ] .
[ rem1 ] for @xmath37 , the functions in are continuous and bounded .
arkhipov and oskolkov @xcite prove that boundedness ( but not continuity ) still holds even if @xmath4 is the characteristic function of a subinterval of @xmath36 .
this implies that the limit distribution has compact support and that therefore corollary [ cor2 ] remains valid in this case , subject to the addtional assumption @xmath60 .
[ rem2 ] it seems plausible that the hypothesis on the number of divisors of @xmath7 can be removed from theorem [ thm2 ] , at least in the case of weights @xmath4 of bounded variation or indeed all riemann integrable functions . to illustrate theorem [ thm2 ] ,
we have computed numerically the value distribution of the real and imaginary parts of incomplete gauss sums for different values of @xmath7 , see figures [ fig5012][fig5014 ] and section [ secnumerics ] .
where @xmath4 is the characteristic function of the interval @xmath61 $ ] , @xmath62 and @xmath6 uniformly distributed in @xmath9 .
there are thus @xmath63 sample points distributed across 40 bins , which means we have on average 53.4 values in each bin .
the histogram seems consistent with fluctuations of the order of the square - root of that number , i.e. , approximately @xmath64 of the height of each bin .
the histogram on the right shows the imaginary part of the corresponding quantities .
the continuous curves represent a numerical approximation to the real and imaginary part of the probability density of the random variable @xmath65.,title="fig:",scaledwidth=49.0% ] where @xmath4 is the characteristic function of the interval @xmath61 $ ] , @xmath62 and @xmath6 uniformly distributed in @xmath9 .
there are thus @xmath63 sample points distributed across 40 bins , which means we have on average 53.4 values in each bin .
the histogram seems consistent with fluctuations of the order of the square - root of that number , i.e. , approximately @xmath64 of the height of each bin .
the histogram on the right shows the imaginary part of the corresponding quantities .
the continuous curves represent a numerical approximation to the real and imaginary part of the probability density of the random variable @xmath65.,title="fig:",scaledwidth=49.0% ] , now for @xmath66 .
the number of sample points is now @xmath67 .
the continuous curves represent a numerical approximation to the real and imaginary part of the probability density of the random variable @xmath68.,title="fig:",scaledwidth=49.0% ] , now for @xmath66 .
the number of sample points is now @xmath67 .
the continuous curves represent a numerical approximation to the real and imaginary part of the probability density of the random variable @xmath68.,title="fig:",scaledwidth=49.0% ] with the same @xmath4 as in figure [ fig5012 ] and @xmath69 , @xmath6 uniformly distributed in @xmath9 , where @xmath70 .
the histogram on the right shows the imaginary part of the corresponding quantities .
the continuous curves represent a numerical approximation to the real and imaginary part of the probability density of the random variable @xmath71.,title="fig:",scaledwidth=49.0% ] with the same @xmath4 as in figure [ fig5012 ] and @xmath69 , @xmath6 uniformly distributed in @xmath9 , where @xmath70 .
the histogram on the right shows the imaginary part of the corresponding quantities .
the continuous curves represent a numerical approximation to the real and imaginary part of the probability density of the random variable @xmath71.,title="fig:",scaledwidth=49.0% ]
legendre s quadratic residue symbol is defined for an odd prime @xmath72 by @xmath73 following jacobi , we extend the definition to arbitrary odd integers @xmath72 multiplicatively : let @xmath72 a positive odd integer with prime factorization @xmath74
. for @xmath75 we define @xmath76 and @xmath77 . the generalized quadratic residue symbol ( or jacobi symbol ) @xmath78 is characterized by the following properties ( cf .
@xcite ) : ( ) @xmath79 if @xmath80 .
if @xmath72 is an odd prime , @xmath78 coincides with the ordinary quadratic residue symbol . if @xmath81 , @xmath82 defines a character modulo @xmath72 .
if @xmath83 , @xmath84 defines a character modulo a divisor of @xmath85 , whose conductor is the conductor of @xmath86 over @xmath87 .
@xmath88 .
@xmath89 .
in particular @xmath90=1 , if @xmath91 .
we assume from now on that @xmath92 the classical gauss sum @xmath93 can be evaluated explicitly in terms of the jacobi symbol : @xmath94 with @xmath95 as defined before theorem [ thm1 ] .
the following theorem is implicit in the papers of fiedler , jurkat and krner @xcite and oskolkov @xcite [ fethm ] for @xmath37 , @xmath96 g_1(p , q ) \ , g_\varphi\big(-\frac{\overline{4p}}{q}\big ) & \text{if $ q\equiv 1\bmod 2$}\\[5pt ] 2 g_1(2 p , q/2 ) \ , g_\varphi^-\big(-\frac{\overline{8p}}{q/2}\big ) & \text{if $ q\equiv 2\bmod 4$. } \end{cases}\ ] ] ( in the first and second case , @xmath97 denotes the inverse of @xmath98 , in the third the inverse @xmath99 @xmath100 . )
since the fourier series of @xmath4 is absolutely convergent , we may assume without loss of generality that @xmath101 with @xmath102 fixed .
the proof is a then simple exercise in completing the square ( cf .
@xcite ) : \(i ) @xmath103 : for @xmath104 even , @xmath105 for @xmath106 odd , @xmath107 and therefore must be zero .
\(ii ) @xmath108 : for @xmath7 odd , we may use the inverse of 2 mod @xmath7 : @xmath109 \(iii ) @xmath110 : we deduce claim ( iii ) from the previous case : define @xmath111 and @xmath112 .
clearly @xmath113 , @xmath114 and @xmath115 .
then @xmath116 hence , for @xmath106 odd , @xmath117 is even and eq .
yields that the right hand side of equals @xmath118 furthermore , @xmath119 , since @xmath120 . if @xmath106 is even , @xmath121 this term therefore vanishes .
the functional equations of incomplete gauss sums stated in theorem [ fethm ] lead us to consider the joint distribution of @xmath122 and @xmath123 on the torus @xmath124 .
the following statement is the second main ingredient in the proof of theorem [ thm1 ] .
[ eqdthm ] let @xmath125 .
then the following convergence holds uniformly in @xmath126 as @xmath54 : 1 . for any sequence of @xmath7 , @xmath127 2 . if @xmath103 is not a square then , for every @xmath128 , @xmath129 3 . if @xmath103 then , for every @xmath130 , @xmath131 4 .
if @xmath108 is not a square then , for every @xmath130 , @xmath132 the statement of theorem [ eqdthm ] also holds for the test function @xmath133 where @xmath134 is the characteristic function of a subset @xmath52 with boundary of measure zero and @xmath135 .
this follows from a standard approximation argument .
the proof of theorem [ eqdthm ] exploits the classic weil bounds on ( twisted ) kloosterman and sali sums .
these directly yield explicit bounds on the rate of convergence in theorem [ eqdthm ] for smooth test functions @xmath136
. the following two lemmas will be helpful in proving theorem [ eqdthm ] .
[ lem7]assume @xmath1 is not a square
. then there exists @xmath137 such that @xmath138 and @xmath139 .
let @xmath140 be the prime factorization of @xmath7 , with @xmath141 .
since @xmath7 is not a square , at least one of the @xmath142 is odd .
suppose now this happens for index @xmath143 , i.e. , @xmath144 is odd , for some @xmath145 .
fix an integer @xmath146 such that , if @xmath147 then @xmath148 and otherwise @xmath149 .
then @xmath150 if @xmath147 and otherwise @xmath151 . in both cases
of course @xmath152 .
furthermore , let @xmath153 be integers so that for @xmath154 , @xmath155 is a quadratic residue mod @xmath156 , and for @xmath157 , it is not .
hence @xmath158 if @xmath157 and @xmath159 otherwise .
by the chinese remainder theorem there is an integer @xmath160 such that @xmath161 and @xmath162 .
using quadratic reciprocity , we have for @xmath163 @xmath164 since @xmath160 . using the multiplicativity of the jacobi symbol we obtain @xmath165 [ lem6 ] 1 .
if @xmath103 is not a square and @xmath128 , then @xmath166 2 . if @xmath103 and @xmath130 , then @xmath167 3 .
if @xmath108 is not a square and @xmath130 , then @xmath168 \(i ) we have @xmath169 for @xmath170 , we have of course , @xmath171 in the case @xmath172 , we need to show that @xmath173 by lemma [ lem7 ] , we can find an @xmath174 such that @xmath139 and @xmath175 .
since @xmath139 we have @xmath176 and thus for every @xmath177 there is @xmath178 such that @xmath179 .
therefore , @xmath180 since @xmath181 , we have @xmath182 . hence , @xmath183 the case @xmath184 is the complex conjugate of the case @xmath172 . for @xmath185 , @xmath186 since @xmath187 or @xmath188 if @xmath189 or @xmath190 , respectively .
\(ii ) the proof follows from the @xmath185 part of the proof for ( i ) , since the assumption that @xmath7 is not a square is not relevant in this case .
\(iii ) we have for @xmath191 , @xmath192 for @xmath170 , the sum is obviously equal to @xmath10 .
the case @xmath172 corresponds to the well known identity @xmath193 we start with the most difficult case ( ii ) . to prove this claim ,
it suffices ( in view of lemma [ lem6 ] ( ii ) and weyl s criterion ) to show that @xmath194 for every fixed @xmath195 , uniformly in @xmath126 .
we have @xmath196 for @xmath170 the inner sum is the kloosterman sum @xmath197 for which we have the classical weil bound @xmath198 , see @xcite .
since @xmath199 and @xmath43 are fixed and @xmath200 , @xmath201 is bounded above .
furthermore @xmath202 for any @xmath203 .
since @xmath204 for any @xmath203 , we see that @xmath205 tends to zero , uniformly in @xmath126 as @xmath54 , as required .
the case @xmath172 ( @xmath184 ) leads to the twisted kloosterman sum @xmath206 ( and the complex conjugate of @xmath207 ) .
here we have the same bound as for kloosterman sums , @xmath208 , see @xcite ( but also the more recent @xcite ) .
we conclude that the contribution of the @xmath209 term also tends to zero uniformly in @xmath210 .
the case @xmath185 leads to @xmath211 and is thus reduced to kloosterman sums .
this proves the case ( ii ) . case ( iii )
reduces to the same estimates as in case ( ii ) @xmath185 .
case ( iv ) is analogous , but here the estimates reduce to bounds on sali sums @xmath212 which are the same as the above for the ( twisted ) kloosterman sums .
case ( i ) of course follows from the classical weil bound for kloosterman sums .
* case 1a : * @xmath213 , @xmath7 not a square .
we need to show that for any bounded continuous @xmath214 we have @xmath215 in view of theorem [ fethm ] , this is equivalent to @xmath216 since @xmath32 and @xmath217 are continuous , the latter statement follows from theorem [ eqdthm ] ( ii ) and subsequent remark , if we choose the test function @xmath218 * case 1b : * @xmath213 , @xmath7 is a square .
we proceed as in case 1b , and note that the condition @xmath219 ( @xmath220 ) is equivalent to @xmath221 ( @xmath222 ) .
the statement follows from theorem [ eqdthm ] ( iii ) .
* case 2a : * @xmath223 , @xmath7 not a square . in this case , the statement to be proved reduces ( again using theorem [ fethm ] ) to @xmath224 which follows from theorem [ eqdthm ] ( iv ) with @xmath225 . * case 2b : * @xmath223 , @xmath7 is a square .
analogous to case 2a , except that we employ theorem [ eqdthm ] ( i ) .
* case 3a : * @xmath226 , @xmath7 not a square .
following the same strategy as above , we deduce that the claim of the theorem is equivalent to @xmath227 as in the proof of theorem [ fethm ] ( iii ) , we substitute @xmath228 and @xmath229 , i.e. , @xmath111 and @xmath112 . note that this map describes a bijection @xmath230 .
hence is equivalent to @xmath231 which is again implied by theorem [ eqdthm ] ( iv ) .
* case 3b : * @xmath226 , @xmath7 is a square .
analogous to case 3a , except that we use theorem [ eqdthm ] ( i ) .
the following lemma says that the sequence probability measures defined by the value distribution of incomplete gauss sums is tight . by the helly - prokhorov theorem
, this means that the sequence is relatively compact , i.e. , every sequence contains a convergent subsequence .
[ lema ] fix @xmath56 . for every @xmath203 there exists @xmath248 such that @xmath249 for any riemann integrable @xmath58 . we have , by chebyshev s inequality @xmath250 the claim now follows from lemma [ rilem ] .
chebyshev s inequality also implies the following .
[ lemb ] fix @xmath56 .
let @xmath58 be riemann integrable .
then , for every @xmath203 , @xmath251 there exists @xmath233 and such that @xmath252 this follows immediately from ; note that @xmath253 and @xmath254 is riemann integrable .
we now turn to the proof of theorem [ thm2 ] .
we restrict ourselves to case 1a where @xmath103 ; the other cases are analogous .
the relative compactness implied by lemma [ lema ] can be stated as follows .
any sequence of @xmath54 with @xmath60 contains a subsequence @xmath255 with the property : there is a probability measure @xmath256 on @xmath257 such that for any @xmath258 and any bounded continuous function @xmath214 we have @xmath259 the probability measure @xmath256 may depend on the choice of subsequence , @xmath4 and on @xmath260 .
let us now show that for every @xmath261 ( infinitely differentiable and of compact support ) the limit @xmath262 exists .
@xmath263 must then be equal to the right hand side of , which in fact means that @xmath256 is unique and the full sequence of @xmath7 converges . to prove the existence of @xmath263 , note first of all that since @xmath261 we have @xmath264 for some constant @xmath265 .
therefore , for @xmath240 , @xmath266 , @xmath267 as in lemma [ lemb ] , we have @xmath268 since the limit @xmath269 exists by theorem [ thm1 ] , the sequence @xmath270 defines a cauchy sequence .
using this fact , the bound and the triangle inequality , we see that @xmath271 is a cauchy sequence , too , and hence @xmath263 exists . as mentioned earlier , this means that @xmath263 must then be equal to the right hand side of , which in fact means that @xmath256 is unique and the full sequence of @xmath7 converges for every bounded continuous @xmath217 .
the bound furthermore implies that @xmath272 , as @xmath273 in @xmath274 .
this completes the proof of theorem [ thm2 ] .
the key step in the proof of theorem [ thm2 ] is the estimate on the mean - square given in lemma [ rilem ] .
we have @xmath232 take @xmath233 such that @xmath234 for all @xmath235
. then @xmath236 where ( recall ) @xmath237 the fourier series of this function is @xmath238 where @xmath239 are the fourier coefficients of @xmath240 .
we have thus shown that @xmath241 by corollary [ cor2 ] , for every fixed @xmath174 , @xmath242 the right hand side is bounded by @xmath243 .
the convergence is uniform in @xmath174 , if we assume @xmath240 has a finite fourier series . in this case , we therefore have @xmath244 since @xmath4 is riemann integrable , given any @xmath203 , there exist @xmath240 with finite fourier series such that ( a ) @xmath245 ( as required in ) and ( b ) @xmath246 .
this proves that the right hand side of is arbitrarily close to @xmath247 .
the computations used in figures [ fig5012][fig5014 ] were carried out with mathematica .
we encoded the real and imaginary part of the incomplete gauss sum @xmath275 ( where @xmath4 is the characteristic function of the interval @xmath276\subset[0,1]$ ] ) as ....
regauss[p _ , q _ , t _ ] : = if[gcd[p , q ] = = 1 , re[sum[exp[2*pi*i*h^2*p / q ] , { h , 1 , t}]/ sum[exp[2*pi*i*h^2*p / q ] , { h , 1 , q } ] ] - t / q , infinity ] imgauss[p _ , q _ ,
t _ ] : = if[gcd[p , q ] = = 1 , im[sum[exp[2*pi*i*h^2*p / q ] , { h , 1 , t}]/ sum[exp[2*pi*i*h^2*p / q ] , { h , 1 , q } ] ] , infinity ] .... the probability density of real / imaginary part of @xmath32 and @xmath33 in figures [ fig5012 ] and [ fig5013 ] was plotted via the smoothhistogram command , where we truncated the fourier series @xmath65 and @xmath68 at @xmath280 and sampled @xmath19 at 300,000 random points in @xmath281 $ ] .
as the distribution of real and imaginary part of @xmath34 are the same , we only computed @xmath282 in figure [ fig5014 ] , truncated at @xmath283 and with 500,000 sample points . | it is well known that the classical gauss sum , normalized by the square - root number of terms , takes only finitely many values .
if one restricts the range of summation to a subinterval , a much richer structure emerges .
we prove a limit law for the value distribution of such incomplete gauss sums .
the limit distribution is given by the distribution of a certain family of periodic functions .
our results complement oskolkov s pointwise bounds for incomplete gauss sums as well as the limit theorems for quadratic weyl sums ( theta sums ) due to jurkat and van horne and the second author . |
radiative corrections to the graviton / matter vertex have been investigated since the 70 s by berends and gastmans @xcite in the qed case , who quantified the impact of these corrections to the bending of photons and fermions in an external localized gravitational field .
these results were based on the study of 3-point functions ( gravitational form factors ) with two fermions ( @xmath1 ) or two photons ( @xmath2 ) in external gravity , and one insertion of the energy momentum tensor ( emt ) @xmath3 of the matter fields . the very small size of these corrections , which affect the angles of deflection , combined with the experimental difficulty to improve on their direct measurements , call into question the possibility of using these effects as a possible test of general relativity .
perhaps , it is for this reason that the study of the electroweak corrections to the bending angles have not drawn any attention on these processes until the analysis of @xcite and , more recently , of @xcite , which concern the fermion / graviton vertex . in the photon case
, the @xmath2 vertex in qed has been re - investigated in the analysis of the conformal anomaly action in @xcite , later extended to the electroweak case in @xcite . in the @xmath4 case
, the study has been separated into the flavour - changing @xcite and flavour diagonal sectors @xcite , the first of the two playing a significant role in the analysis of possible flavour transitions in the presence of a scalar component in the external gravitational field , as pointed out in @xcite .
this aspect will be re - addressed in our current study .
the issue concerning the size of these corrections and their relevance can be resolved by noticing that the very same interactions become significant in the presence of sizeable gravitational backgrounds .
such are those due to black holes , which are ubiquitous in galactic centers , and surging to remarkable relevance in the analysis of galaxy structure formation .
obviously our formalism , which simply extends to the electroweak case the original analysis of @xcite , remains valid only in the case of small fluctuations around a flat spacetime metric , where curvature effects are ignored , and does not allow to handle backgrounds characterized by a strong curvature .
it could , however , be extended to include further corrections related to the curvature scale of certain backgrounds , in the limit in which gravity is treated as an external classical source . for this reason
, we think that our analysis allows to highlight some of the basic features of the interaction of the standard model lagrangian , once this is immersed in a weak gravitational background , clarifying some of the specific features of this coupling at a perturbative level .
in particular , the renormalizability of the theory , which remains intact in this extended framework , is an important feature of these computations , not shared by other backgrounds .
one of the specific features of our results is their possible application to the neutrino sector , which has received some attention in recent years ( see for instance @xcite ) .
we will provide indeed the expression for the qed and electroweak corrections to the scattering of a neutrino off an external source .
these results are the starting point in the derivation of the equation relating the angle of deflection to the impact parameter of the collision , in the case of a static heavy source of the gravitational field .
the use of an impact parameter formalism allows to merge the usual quantum corrections , computed in the exact electroweak theory , with the semiclassical method in which the incoming asymptotic quantum state is described by a particle approaching the source at a distance @xmath5 , with @xmath5 denoting the impact parameter .
this method has been used in @xcite and in the more recent literature on the computation of corrections to the gravitational scattering of quantum fields both in static @xcite and in rotating ( lense - thirring ) backgrounds @xcite . except for the case of the qed corrections , already discussed in @xcite and
that we recompute , a general analysis of the weak corrections and of the modified lense equations , which may impact the structure of cosmic shears , is left to future work .
the primary goal of our work is to extend our previous analysis of the @xmath1 vertex , by computing the radiative corrections to the cross section in an external gravitational field , in the limit of a source of large mass , i.e. in a scattering off a potential . in this case
we discuss both the limits of low and high momentum transfers . at low momentum transfers
we derive an effective hamiltonian which describes the flavour - changing transitions .
we also discuss the cross section for the exchange of a scalar component for the gravitational field , with the inclusion of the electroweak effects .
our expressions for fermion scattering are specialized both to ordinary and to weakly interacting fermions , such as neutrinos or dark matter particles , by the inclusion of suitable form factors which account for interactions in the inner core of the source .
we conclude our analysis with some perspectives and possible extensions of our work , to be left for further studies . being this work the third in a sequel @xcite
, most of our previous notations and results will be necessary in order to proceed with the applications discussed here .
in this section we briefly review the main features of the coupling of the standard model in an external gravitational field , focusing our attention on the fermion sector .
we recall that the dynamics of the standard model plus gravity is described by the action = _ sm + _ g + _ i with @xmath6 containing the standard model ( sm ) lagrangian , while _ g & = & -d^4 x r + _ i & = & d^4 x r h^h [ thelagrangian ] denote respectively the einstein gravitational term of the action and a term of improved action @xmath7 , involving the higgs doublet @xmath8 .
the latter is responsible for generating a symmetric and traceless energy - momentum tensor ( emt ) @xcite .
@xmath6 is obtained by extending the ordinary lagrangian of the standard model to a curved metric background .
@xmath7 vanishes in the flat space time limit , due to the vanishing of the ricci scalar in the same limit .
its corresponding emt @xmath9 ) , however , is non - vanishing . with the inclusion of @xmath7 , varying the action @xmath10 with respect to the fields of the sm allows to generate an emt which is symmetric without any additional symmetrization .
we remark that @xmath11 is conformal at the value @xmath12 and guarantees the renormalizability of the model , including all green functions at any order in @xmath13 containing only insertions of external graviton fields @xmath14 , with @xmath15 .
we will be using the flat metric @xmath16 , with an expansion of the form & & g_= _ + h _
+ o(^2 ) + & & g^=^ - h^ + o(^2 ) + & & = 1+h + o(^2 ) , where @xmath17 is the trace of the metric fluctuations .
the interaction between the gravitational field and matter , at this order , is mediated by diagrams containing a single power of the emt @xmath18 and multiple fields of the standard model .
the tree level coupling is summarized by the action _
int =- d^4 x t _ h^ , [ inter ] where @xmath19 denotes the symmetric and covariantly conserved emt of the standard model lagrangian , embedded in a curved space - time background and defined as t_= |_g= .
the complete emt of the standard model , including ghost and gauge - fixing contributions can be found in @xcite .
the fermionic part of the emt is obtained using the vielbein formalism .
indeed the fermions are coupled to gravity by using the spin connection @xmath20 induced by the curved metric @xmath21 .
this allows the introduction of a derivative @xmath22 which is covariant both under gauge and diffeomorphism transformations .
generically , the lagrangian for a fermion @xmath23 takes the form l_f = ( |^(d _ ) - ( d _ | ) ^ - m | ) , [ dir ] where the covariant derivative is defined as @xmath24 , with @xmath25 denoting the gauge field . the spin connection takes the form _
= ^a b v_a^ v_b ; where @xmath26 is the vielbein , the semicolon denotes the gravitationally covariant derivative and @xmath27 are the generators of the lorentz group in the spinorial representation .
the latin indices are lorentz indices of a local free - falling frame .
the connection can be expanded as _
= ^m n. with v_^m=_^m + h^m_+o(^2 ) . at leading order the interacting dirac lagrangian ( [ dir ] ) generates the terms _ f&= & _ 0-h_t^(0 ) with @xmath28 denoting the free dirac term _ 0=(|-| ) - m t^(0)_=((|__-_| _ ) + ( ) ) -__0 that we couple to the external field @xmath29 . in momentum space
the corresponding vertex takes the form ( see fig .
[ ff ] ) v^(0 ) =( ^(p_1 + p_2)^+ ^(p_1 + p_2)^- 2 ^(_1+_2 - 2 m ) ) , from which we have omitted an overall factor @xmath30 coming from the interaction lagrangian , which will be reinserted at the level of the transition amplitude . only diagonal transition amplitudes in flavour space are induced at leading order , while the inclusion of the electroweak corrections allows flavour - changing transitions , which will play a key role in our analysis .
we will denote with @xmath31 the transition amplitude between the intial and the final state induced by the insertions of the emt vertex , with initial and final wave functions included , while @xmath32 will be denoting the corresponding vertex in momentum space , i.e. ^=|u(p_2)v^u(p_1 ) , having labeled the momenta of the incoming and outgoing fermion with @xmath33 and @xmath34 respectively .
we will similarly use the notation |u(p_2 ) o u(p_1 ) to denote the matrix element of any operator @xmath35 in momentum space . using these notations
, the same matrix element can be expressed in position space in terms of the operator @xmath36 , inserted on the dirac eigenstates given by incoming ( @xmath37 ) and outgoing ( @xmath38 ) plane waves p_2|t^(x)|p_1=|_f(p_2)v^_i(p_1 ) e^i qx , where @xmath39 is the 4-momentum transfer .
we have introduced plane waves given by _
i(p_1)=_i u(p_1 ) , _ i= , |u(p_1)u(p_1)=1 , and similarly for @xmath38 , while @xmath40 denotes a finite volume normalization of the two scattering states .
the @xmath41 ( @xmath42 ) and @xmath43 ( @xmath44 ) are the energy and corresponding mass of the incoming ( outgoing ) particle respectively .
the scattering matrix element is then written as i_if=-_v d^4 x p_2 relevant region of integration and with the gravitational field being external , which gives p_2
|h_(x ) t^(x)| p_1 & = & h_(x ) |(p_2)v^(p_1 ) e^i qx . the scattering amplitude which can be immediately expressed in momentum space as i _ fi&= & - h_(q ) |(p_2)v^(p_1 ) + & = & - h_(q ) _
i_f ^ [ sfi ] in terms of the gravitational fluctuations in momentum space @xmath45 , defined as in ( [ h2 ] ) .
+ for a static external field the energies of the incoming / outgoing fermions are conserved ( @xmath46 ) both in the flavour - diagonal and non - diagonal cases , as one can immediately realize from ( [ h1 ] ) .
it is obvious that the interaction region depends both on the type of scattering particle and on the geometry of the source .
we will consider the potential scattering of a dirac fermion off an external static source , which acts as a perturbation on the otherwise flat spacetime background .
the source is characterized by an emt @xmath47 and the fluctuations are determined by solving the linearized equations of general relativity .
we start analyzing the flavour - conserving transitions at leading order . in the rest of the paper , for simplicity
, we will omit the flavour index from each spinor , together with the explicit flavour dependence on the masses of the internal flavours , which appear at 1-loop order .
the linearized equations take the form ( h_- _ h)=-t^ext _ and can be rewritten as h_= s _
, s_=- ( t^ext_-_t^ext ) [ oneq ] , where we have denoted with @xmath48 the emt trace of the external source .
the external field is obtained by convoluting the static source with the retarded propagator g_r(x , y)= normalized as g_r(x , y)=^4(x - y ) .
the solution of ( [ oneq ] ) takes the form h^ext_(x)=d^4 y g_r(x , y)s_(y ) , with the emt of the external localized source t^ext_=^3 ( ) , which for a particle of mass @xmath49 at rest at the origin with @xmath50 ) takes the form [ emt ] t^ext_=m^0_^0_^3 ( ) .
this gives s_=|s _
^0_^0 _ and the field generated by a local ( point - like , @xmath51 ) mass distribution has a typical @xmath52 @xmath53 behaviour h^l_(x ) & = & |s_. [ hh ] the fluctuations are normalized in such a way that @xmath54 has mass dimension 1 , as an ordinary bosonic field , with @xmath55 of mass dimension @xmath56 .
the fourier transform of @xmath57 in momentum space is given by h_(q_0,)&=&d^4 x e^i qx h_(x ) + which for a static field reduces to the form h_(q_0,)=2
( q_0 ) h _ ( ) , [ h1 ] with h_()h_0 ( ) |s _ [ h2 ] defined in terms of the scalar form factor which is related to the geometrical structure of the external source .
in the case of a point - like source , we obtain h_(q_0,)&= & 2 ( q_0 ) |s_d^3 + & = & 2 ( q_0)()|s _ , [ trans ] which gives h_0 ( ) ( ) , h _ ( ) ( ) |s_. [ h0 ] at the same time , the leading order interaction is redefined as v^(0 ) o_v^(0 ) , with o_v^(0 ) =( p^^+ p^^- 2 ^ ( - 2 m ) ) , which will turn useful in the computation of the radiative corrections . using the matrix element order scattering amplitude can be rewritten as i^(0)_i f&=&d^4 x h_(x)|_f(x ) t^(0)_i(x ) + & = & ( -)_i_f(i h_()|u(p_2)o_v^(0 ) u(p_1 ) ) 2(q_0 ) .
the averaged squared amplitude then takes the form |i^(0)_i f|^2&=&(-)^2 ( _ i _
f)^2(2(q_0 ) ) ( ) ^2 _ 0 , [ averaged ] where we have introduced a factor ( 1/2 ) for the fermion spin average . as usual ,
we have extracted , from the square of the delta function , the transition time @xmath58 , using @xmath59 .
we have defined _ 0&=&tr|s _ |s _ + & = & e^2 + & = & 8f^(0)(x , ) in the flavour - diagonal case , with @xmath60 given by f^(0)(x,)=^2 + + + ^2 , [ given ] with @xmath61 and @xmath62 the 3-momentum of the incoming fermion .
we have used the relations @xmath63 in the elastic limit , while @xmath64 and @xmath65 . to compute the differential cross section
we define the transition probability from the initial state ( @xmath66 ) into a set of final states ( @xmath67 ) of differential @xmath68 given by d w= dn_f normalized with respect to the incident flux @xmath69 with the final state density , with a volume normalization @xmath26 given by d n_f= d^3
_ f=||e_2 de_2 d , which allows to define the cross section as the differential transition rate per unit time @xmath70 . combining all the contributions together and integrating over the energy of the final state we obtain & & |_l + & = & ( ) ^2 f^(0)(x , ) . [ leading ] the result in eq .
( [ leading ] ) is in agreement with the corresponding one evaluated in ref .
@xcite . in the following sections we will refer to this cross section as to the _ localized _ one , denoted in ( [ leading ] ) also as @xmath71 , being generated by a gravitational point - like source located at the origin .
this will turn useful in order to make a distinction between the cross sections of localized and of extended sources .
the computation of the angle of deflection for the fermion involves a simple semiclassical analysis , in which one introduces the impact parameter representation of the same cross section .
assuming that the incoming particle is moving along the @xmath72 direction , with the source localized at the origin , and denoting with @xmath73 the azymuthal angle , we have the relation = [ semic ] between the impact parameter @xmath5 and the scattering angle @xmath73 , measured from the @xmath74direction .
this semiclassical approach allows to relate the quantum interaction between the particle and the source .
we recall , if not obvious , that in the classical description the relation between @xmath5 and the deflection angle @xmath75 is obtained from energy and angular momentum conservation . in this respect eq .
( [ semic ] ) takes a similar role .
the relation between @xmath5 and @xmath73 as a function of the incoming energy ( @xmath76 ) , at least for small deflection angles , which correspond to large impact parameters , can be found analytically , but it should be integrated numerically otherwise . in the case of the point - like cross section @xmath77 , for instance ,
one obtains the differential relation = - 2 ( ) ^2 f^(0)(x , ) which gives b^2()=(g m)^2 ( + 2 ( 4 + 3 x ) ( ( ) ) ) . in the small @xmath73 limit
we get the relations b ~g m ( + + ( 1 + ) ) + o(x^2 ) + o ( ) , [ blocal ] which allows us to identify the deflection angle as _
d ~4 . [ impact ] notice that in general @xmath78 for a scattering at high energy , being the fermion mass small compared to the incoming energy or three - momentum .
onbviously , the bending angle of neutrinos , charged leptons and dark matter particles will be affected by the geometry of the source and by the strength of the fermion interaction with the source .
this will vary if , for instance , the particles are allowed to enter the core of the gravitational source , as in the neutrino and the dark matter cases . for this reason we will proceed by determining the structure of these corrections , providing more general cross sections beyond the point - like approximation . once we allow a finite size for the mass distribution , expression ( [ leading ] ) gets modified .
this is the case , for instance , for beams of neutrinos or dark matter particles which may have , in principle , impact parameters smaller than the source radius . in this case
, it is intuitively clear that in the limit of a weak gravitational background , gauss theorem in gravity implies that only the region of the source of radius smaller than the particle impact parameter is relevant in the scattering process .
this can be shown rigorously directly from the spherically symmetric schwarzschild metric .
however , the proof - in the weak field limit - of factorization of the 3-dimensional euclidean metric from the schwarzschild solution , requires some technical steps which are left to an appendix ( appendix [ schw ] ) .
this allows to avoid the use of polar projectors in the scattering amplitude - for spherically symmetric backgrounds - which would render the treatment more involved from the computational side .
our discussion , here , is limited to the case of a uniform mass distribution , but it can be generalized to any distribution of spherical symmetry , in the weak field limit . using this result , it follows that the metric fluctuations of a distributed source can be described ( in @xmath79 units ) by the field h_= - |s _ with the potential of the distributed source @xmath80 , given by _ d(r)=-(r - r ) -(3 - ) ( r - r ) .
[ pot ] notice that in the expression above we have separated the contribution of the external core region from the internal one , with @xmath81 being the radius of the source .
the modifications to the expression of the cross section in the various cases , due to the different nature of the scattering particles , can be taken into account by the insertion of the appropriate form of the fluctuation tensor in momentum space , obtained from this generalized geometrical setting . *
* non - weakly interacting fermions * for this reason , as in the case of the localized gravitational source , for non - weakly interacting fermions we restrict the integration region of the scattering amplitude in the form i_if =- d x_0_>r d^3 p_2 |h_(x ) t^(x)| p_1 , where we have restricted the integration region to @xmath82 , i.e. outside the radius of the gravitational source . therefore , according to ( [ pot ] ) , it is convenient to denote the contributions to the gravitational field coming from the external region as @xmath83 . at leading order , for interactions
which involve only the region of the external core we have i _ fi&= & - h_e(q_0 , ) _
^ [ sfi ] with h_e(q_0,)&= & 2(q_0 ) h_e()|s _ , + h_e()&= & _ > r d^3 ( - ) e^i + & = & ( || r ) [ out ] obtained as fourier transform of the potential in the restricted region .
it is then clear that , if we allow only interactions out of the core region , as we should for non - weakly interacting fermions , the modification of the cross section and of the expression of the bending angle , at leading order , are given by _
e=^2 ( || r ) |_l , [ sigmam ] having used for @xmath71 the result given in ( [ leading ] ) .
* * neutrinos and dark matter fermions * moving to the case of neutrino/ dark matter scatterings , as in the previous cases , the interaction region covers all the 3-dimensional space . beside @xmath84 , it is convenient to introduce also the expression of the field generated inside the core of the source @xmath85 and its corresponding form factor @xmath86 . in this case
the toal form factor @xmath87 is given by h()= h_i ( ) + h_e ( ) [ all ] with h_i()&= & _
< rd^3(- ) e^i + & = & - while @xmath88 is given by ( [ out ] ) . using ( [ all ] ) ,
the gravitational field in momentum space is given by h_(q_0,)= - ( || r ( cross section at leading order is proportional to the result for the localized one _ g = a()^2 |_l [ distr ] with the modulating factor given by a()= - ( || r
( ||r)- ( ||r ) ) .
the one - loop corrections to the graviational fermion vertex in the framework of the sm , have been computed in refs . @xcite and @xcite for the flavor diagonal and off - diagonal transitions , for the case of external on - shell fermions .
exact results at one - loop order in all masses and momenta are provided for the corresponding form factors in refs . @xcite and @xcite .
+ the expression up to 1-loop order of the contributions to the fermion - fermion - graviton vertex @xmath89 can be organized in terms of a flavour - conserving ( or diagonal ) sector contribution and of a flavour - changing part .
the diagonal sector receives contributions at tree level ( i.e. from @xmath90 ) , and at 1-loop from the virtual exchange of a photon ( @xmath91 ) , a gluon ( @xmath92 ) , a @xmath93 ( @xmath94 ) and a @xmath95 , the latter denoted as @xmath96 . in the flavour - changing ( non diagonal )
sector the corrections start at 1-loop and are generated only from the exchange of a virtual @xmath95 , denoted as @xmath97 . as usual
, the radiative corrections can be splitted into vertex corrections and the related counter - terms .
we use the expression _
counter - terms _ to refer to the sum of those diagrams coming from the self - energy contributions to the external fermion legs , plus the corresponding diagrams coming from their wave - function renormalization . in the case of the flavor - changing transitions ,
the latter contributions are absent at one - loop level in the emt counterterms , due to the fact that the fermion - graviton vertex is flavor - conserving at tree - level @xcite .
because of the energy momentum conservation , the radiative corrections induced by sm fields ( no graviton loop corrections ) to the emt are finite at any order in perturbation theory , which implies the nonrenormalization of the newton constant .
it is convenient to separate the diagonal ( @xmath98 ) from the non - diagonal ( @xmath99 ) parts using the notation t^_*f
t^d,_*ff * + t^nd_*ff * with @xmath100 expressed as [ diag ] t^d_*f f*=_*f f * t^d , where the flavour - conserving transitions are t^d= ^(0 )
_ + t^_g + t^ _
+ t^_h + t^_z while the flavour - changing matrix elements take the form [ hatt ] t^nd _ * f f * = t^ nd_w,*ff * since these can be generated only by the w exchange .
the contributions given above are expanded in terms of a set of operators @xmath101 and intermediate form factors @xmath102 , multiplied by the ckm matrix elements , related to internal the flavour transitions in the internal vertices of the diagrams .
these are accounted for by the @xmath103 matrix elements , with the indices @xmath104 and @xmath105 corresponding to the flavour of the external and internal - loop fermions respectively .
generically , the flavour structure can be completely absorbed into some final form factors @xmath106 [ ftof ] f^w * f f*_l(q^2 ) = _ * n * v_*f n*^ * v_*n f * f_l(q^2 , x_f ) , [ fw ] as we are going to specify below .
+ in the followings we will drop the flavour dependence on @xmath107 , just to simplify the expressions , and the distinction between the ( flavour ) diagonal and non diagonal cases will be made evident by use of the superscript @xmath99 ( non diagonal ) in the flavour - changing cases . in the diagonal case the contributions coming from the electromagnetic and weak transitions will be summarized by the form factors @xmath108 , with @xmath109 and @xmath110 with @xmath111 respectively . in the non diagonal case , the @xmath112 , with @xmath113 ,
will completely describe all the corresponding transitions .
the flavour structure will be omitted also from the emt in both cases .
we will also denote with @xmath114 the masses of the initial and final state fermions .
all the expressions of the form factors discussed in this work can be found in @xcite . in the diagonal sector ,
vector - like ( @xmath26 ) matrix elements are expanded onto a basis of four tensor operators @xmath115 as [ vectorbasis ] o^_v 1 & = & ^ p^+ ^ p^ , + o^_v 2 & = & m ^ , + o^_v 3 & = & m p^ p^ , + o^_v 4 & = & m q^ q^ , where @xmath116 is the mass of the corresponding fermion , generated by the photon , gluon and higgs exchanges .
a second set of operators chiral operators @xmath117 appears in the exchange of @xmath95 s and @xmath93 gauge bosons in the virtual corrections [ chiralbasis ] o^_c 1 & = & ( ^ p^+ ^ p^ ) p_l , + o^_c 2 & = & ( ^ p^+ ^ p^ ) p_r , + o^_c 3 & = & m ^ , + o^_c 4 & = & m p^ p^ , + o^_c 5 & = & m q^ q^ , + o^_c 6 & = & m ( p^ q^+ q^ p^ ) ^5 , where @xmath118 , whose contributions are summarized by the amplitudes [ akin ]
^_c l|u(p_2)o^_c lu(p_1 ) .
we have imposed the symmetry constraints on the external fermion states ( of equal mass and flavour ) and the conservation of the emt .
explicitly , the various contributions are given by t^_g & = & i c_f _ k=1 ^ 4 f_k(q^2 ) |u(p_2 ) o^_v k u(p_1 ) , + t^ _ & = & i q^2 _ k=1 ^ 4 f_k(q^2 ) |u(p_2 ) o^_v k u(p_1 ) , + t^_h & = & i m^2 _ k=1 ^ 4 f^h_k(q^2 ) |u(p_2 ) o^_v k u(p_1 ) , with @xmath119 and @xmath120 for qcd .
the remaining diagonal contributions , generated by the virtual exchange of a @xmath93 or of the @xmath95 s gauge bosons , are expanded on a tensor basis of six elements as t^_z & = & i _
k=1 ^ 6 f^z_k(q^2 ) |u(p_2 ) o^_c k u(p_1 ) , + t^_w & = & i _
k=1 ^ 6 f^w_k(q^2 ) |u(p_2 ) o^_c k u(p_1 ) .
the exact expressions for the form factors @xmath121 , @xmath122 appearing above , can be found in ref.@xcite . for convenience
we introduce the linear combinations ^(a)_k=
i ( c_f + q^2 ) f_k(q ) + i m^2
f_k^h(q ) k=1,2, 4 [ ra ] where the form factors @xmath123 are those generated by the exchange of a photon or a gluon , while the @xmath124 are those due to a virtual higgs .
the chiral operators , instead , multiply the linear combinations ^(b)_l= i ( f^z_l(q ) + f_l^w(q ) ) l=1,2, 6 [ rb ] where @xmath125 and @xmath126 denote the @xmath95 s and @xmath93 form factors , contributing to the flavour - diagonal case . in eq .
( [ fw ] ) we have set @xmath127 , where @xmath128 and @xmath129 stand for the masses of the fermion @xmath67 and of the @xmath95 gauge bosons respectively . in the following we will be using the short - hand notation @xmath130 to denote the 10 form factors above ( @xmath131 . notice that in the case of an elastic scattering of a neutrino , only @xmath132 and @xmath133 contribute to the cross section . as we have mentioned above
, the flavour - changing contributions appear at 1-loop order induced by the w - boson exchange .
they include the vertex corrections plus the corresponding flavor - changing counter terms , that consists of the flavor - changing self - energy insertions on the external fermion legs @xcite .
the total contribution in this sector can be expressed in terms of a set of operators @xmath134 as [ hatt ] t^nd
_ w = -i _ k=1 ^ 12 f^nd_k(q^2 )
|u^(p_2 ) o^_k u^(p_1 ) [ tfc ] with the tensor operator basis given by ll o_1^ = ( ^ p^+^ p^ ) p_l & o_7^= ^ m_- + o_2^= ( ^ q^+^ q^ ) p_l & o_8^= p^p^ m_- + o_3^= ^ m_+ & o_9^= q^q^ m_- + o_4^= p^p^ m_+ & o_10^= ( p^q^+q^p^ ) m_- + o_5^= q^q^ m_+ & o_11^= ( ^ p^+^ p^ ) p_r + o_6^= ( p^q^+q^p^ ) m_+ & o_12^= ( ^ q^+^ q^ ) p_r , [ basis ] where @xmath135 .
this is the most general rank-2 tensor basis that can be built out of two momenta , @xmath136 and @xmath137 , a metric tensor and dirac matrices @xmath138 and @xmath139 @xcite .
notice that in the equal mass case @xmath140 @xmath141 is proportional to the identity matrix and @xmath142 is a pseudoscalar contribution . in the massless case ,
relevant for neutrino flavour transitions , only @xmath143 and @xmath144 survive .
the exact analytical expressions for the form factors @xmath145 , with on - shell fermions , have been computed in @xcite and at the leading order in @xmath146 in @xcite .
using these notations , the scattering matrix element , in the diagonal case , can be expressed in the form [ diags ] i_i f&= & -_i_f h_^d _ + & = & -_i_f h _ ( ^(0 ) _ + _ k=1 ^ 4 ^(a)_k^_v k + _
l=1 ^ 6 ^(b)_l^_c l ) with @xmath147 given in ( [ diag ] ) and the terms @xmath148 defined in eqs.([ra]),([rb ] ) . by a simple redefinition of the operators , o_d^io_v i , i=0, ,4 o_d^io_c i-4 , i=5,
,10 r_d^ir _
i^(a ) , i=0,
,4 r_d^ir_i-4^(b ) , i=5, ,10 and defining = _ i , j=0 ^ 10o_d^i o_d^j ^i_d ^j_d with _
o_d^j tr|s _ |s _ the cross section takes the form =( ) h_0(q)^2 with @xmath149 defined in ( [ h0 ] ) . by computing the expression for @xmath150 defined above ,
we obtain for the differential cross section & = & ( f^(0)(e , m , ) + 2 e f_d^(1)(e , m , ) + c_f 2e f_d^(1)(e , m , ) .
+ & & .+ 2 e f_d^(3)(e , m , ) ) , [ crs ] where we have dropped the @xmath151 terms in all the couplings .
we have chosen to present ( [ crs ] ) given in terms of the energy of the incoming / outgoing particle @xmath152 ( @xmath153 ) and of its mass @xmath116 , beside the scattering angle @xmath73 .
@xmath154 , the leading order term , is given by eq .
( [ given ] ) , which takes the equivalent form as a function of @xmath155 and @xmath73 f^(0)(e , m , ) = . [ f0 ] for the rest , we have denoted with @xmath156 and @xmath157 the electromagnetic and weak corrections .
the strong corrections , in the quark case , differ from the electromagnetic ones just by the colour factor , proportional to the casimir of @xmath158 , @xmath159 . * * electromagnetic corrections * coming to the the electromagnetic corrrections , these can be organized in the form f_d^(1)(e , m , ) = _
k=1 ^ 4x_k(e , m , ) f_k in terms of the four form factors @xmath121 and coefficient functions @xmath160 .
the @xmath161 s have been given in @xcite and , being a function of @xmath146 , can be expressed as functions of the variables @xmath155 and @xmath73 . in the following
we will omit the @xmath146 dependence in the form factors @xmath108 .
for the expressions of the @xmath162 term we obtain x_1(e , m,)= + x_2(e , m,)&=&+ + x_3(e , m,)&=&++m^2 ^4 ( ) + x_4(e , m,)&=&-m^2 ^4 ( ) * * weak corrections * finally , the general expression of the function @xmath157 which contains the electroweak corrections to the unpolarized cross section is given by f_d^(3)(e , m,)&=&w_0 ( e , m , ) + w_1(e , m , ) + w_2(e , m,)^4 ( ) + where w_0(e , m , ) & = & f_3^h ( 4 e^2 m^4 - 2 m^6)+f_1^h ( 4 e^2 m^2 - 2 m^4)+f_5^w ( 2 e^2-m^2)+f_6^w ( 2 e^2-m^2 ) + & & + f_8^w ( 4 e^2 m^2 - 2 m^4)+4 e^2 f_8^z m^2 + 2 e^2 f_5^z+2e^2 f_6^z - f_2^h m^4-f_7^w m^2 - 2 f_8^z m^4 + & & -f_5^z m^2-f_6^z m^2-f_7^z m^2 w_1(e , m , ) & = & f_3^h ( 4 m^6 - 8 e^2 m^4)+f_4^h ( 4 e^2 m^4 - 2 m^6)+f_1^h ( 2 m^4 - 8 e^2 m^2)+f_5^w ( m^2 - 4 e^2 ) + & & + f_6^w ( m^2 - 4 e^2)+f_8^w ( 4 m^4 - 8 e^2 m^2)+f_9^w ( 4 e^2 m^2 - 2 m^4)-8 e^2 f_8^z
m^2 + & & + 4 e^2 f_9^z m^2 - 4 e^2 f_5^z-4 e^2 f_6^z+f_2^h m^4+f_7^w m^2 + & & + 4 f_8^z m^4 - 2 f_9^z m^4+f_5^z m^2+f_6^z m^2+f_7^z m^2 w_2(e , m,)=f_3^h m^4-f_4^h m^4+f_8^w m^2-f_9^w m^2+f_8^z m^2-f_9^z m^2 where the analytical expressions for the form factors @xmath163 is given in @xcite . in the above expressions we omitted the @xmath146 dependence in the form factors @xmath164 . in this case the scattering amplitude on an external gravitational field @xmath54 , for on - shell fermions ,
can be written as i_i f&= & -i_i_f h_^nd + & = & i _
i_f h__k=1 ^ 12 f^nd_k(q^2 ) ^_k where , for convenience , we have introduced the notation akin to eq .
( [ akin ] ) _
k|u(p_2 ) o^_k u(p_1 ) .
the exact analytical expressions for the @xmath121 at one - loop , can be found in ref.@xcite .
then , the flavour - changing changing cross section takes the form =( ) ^2 h_0(q)^2 _ w(m_1,m_2 ) , where _ w(m_1,m_2 ) & & _ i , j=1 ^ 12o_i^ o_j^ f^nd_i ( f^nd_j)^ + & = & tr|s _ |s_. the expression of @xmath165 , for general values of the mass parameters , is rather involved . in the case of equal external fermion masses
@xmath166 we obtain _
w(m ) & = & _ i=1 ^ 12c_i |f^nd_i|^2 + 2 _
i=1 ^ 11 _ j = i+1 ^ 12e ( f^nd_i ( f^nd _ j)^)d_i j , [ non1 ] with the nonvanishing coefficients being given in appendix [ cg ] .
the cross section above simplifies considerably in the case of massless external fermions , taking the simplified form [ flav ] =( ) h_0 ^ 2(q)|f^nd_1(q^2)|^2 ^2 , which involves a single form factor , @xmath167 .
we will come back , in a following section , to investigate the behaviour of this cross section at low momentum transfers . + from ( [ flav ] ) one can consider the possibility of generating flavour - changing transitions between massless external leptons in a gravitational background , which in the neutrino case would appear as a gravitational contribution to their flavour oscillations .
implicit in ( [ flav ] ) is the presence of a complex factor proportional to the ckm matrix element @xmath168 , that we have included into the structure of @xmath169 .
this factor mediates transitions between a massless external fermion ( * i * ) and massive internal ones ( * n * ) , which we sum over .
unitarity of @xmath26 will rule out such transitions only for mass degenerate internal fermions .
in this and in the following section we will expand the form factors at low momentum transfers , starting from the electromagnetic case and then moving to the weak case .
this will simplify considerably the expressions of the corresponding cross sections , which take , in this limit , rather simple forms .
we start our analysis by observing that the existence of a ward identity for the conservation of the energy momentum tensor of the various ( qed , qcd and electroweak ) sectors of the standard model has important implications on the structure of most of the form factors which appear in the operatorial expansions presented before .
consider the renormalized ( with self - energies on the external lines ) one - loop correction only , @xmath170 , with on - shell fermions . in this case
the emt ward identity is given by [ wi2 ] q _
t^(p_1 , p_2 ) = 0 .
this describes the conservation of the emt and emerges from the requirement of invariance under general coordinate transformations .
for this reason one can reasonably expect that some consequences of the equivalence principle can be extracted from it . indeed one can differentiate eq .
( [ wi2 ] ) with respect @xmath171 [ diffwi1 ] 0 = = t^(p_1 , p_2 ) + q _
t^(p_1 , p_2 ) , and then perform the zero momentum transfer limit setting @xmath172 , ( @xmath173 ) .
the last term in eq .
( [ diffwi1 ] ) vanishes and we have t^(p_1 , -p_1 ) t^(p , q=0 ) = 0 .
we then conclude that the ward identity implies the vanishing of the one - loop corrected @xmath174 matrix element for on - shell fermions in the zero momentum transfer limit .
+ this has some interesting consequences on the form factors . in this respect
one has to take into account , at the same time , the tensor expansion of the correlator . notice that in the @xmath175 limit , the tensor structures @xmath176 , @xmath177 , @xmath178 go to zero and nothing can be said on the corresponding form factors .
but for the remaining elements of the basis , which survive for @xmath175 , the corresponding form factors must vanish for @xmath179 .
the simplest case in which one can perform the small @xmath180 limit of the form factors is the electromagnetic one . in this case
we expand in powers of @xmath181 and obtain the expressions up to @xmath182 f_1&=&- + ( 9 + 12 + 65 - 36 2 ) + f_2&= & + ( -40 + 36 2 + 33 ) + f_3&= & - + ( 2 - 36 2 -57 ) + f_4 & = & + ( -40 + 36 2 + 33 ) + [ fem ] the values at the leading order expansion in @xmath146 of the form factors in eq .
( [ fem ] ) are in agreement with the corresponding results of ref.@xcite .
notice the presence of an infrared cutoff @xmath183 in the form factors in eq .
( [ fem ] ) , where @xmath183 is the photon mass used to regularize the infrared divergences .
this is due to the exchange of a massless photon , generated by the 3-point scalar integral expanded around zero @xmath184 c_0 ( q^2=0 , m^2 , m^2 ) = \ { - 2 x_s + } where the ellipsis mark some infrared finite terms .
we have defined x_s = - = .
it can be shown by a direct computation in qed , that the infrared singular parts of the virtual corrections of the matrix element @xmath185 , with @xmath186 being the graviton , once summed , are proportional to the tree - level vertex .
a second infrared singular contribution comes from the renormalization counterterm of the masless photon , which alse needs to be taken into account . as shown in @xcite , the inclusion of the real photon emissions from the fermion legs in the collinear limit ( the process @xmath187 ) , regulated by a cutoff @xmath183 and integrated over the photon , allows to remove the singularity in @xmath188 . using these expressions
, the virtual plus soft corrections to the cross section - at leading order in @xmath181 - take the form = ( f^(0)(e , m , ) + _ d^(1)(e , m , ) ) , where @xmath154 is given by ( [ f0 ] ) and ^(1)_d(e , m , ) & = & a_1 ( ) + a_2 ^ 3 ( ) + a_3 ^ 5()with a_1=- a_2=- a_3=- |_1| .
now , we focus on the kinematical region where @xmath189 and perform an expansion of the @xmath190 in powers of @xmath191 , where @xmath192 is the transfered momentum and @xmath193 stands for any characteristic electroweak scale , identified with the @xmath194 or higgs boson mass .
because of the conservation of the matter energy momentum tensor @xmath195 on shell ( which holds only in the absence of gravitational field corrections ) , it can be easily show that all the form factors @xmath196 in eq .
( [ t ] ) , except for @xmath197 , vanish in the limit @xmath198 and turn out to be of order @xmath199 .
we will start first our discussion with the flavour - conserving transitions , and then we will address the flavour - changing ones .
here we provide the expressions for @xmath174 at the leading term in the @xmath200 and external masses @xmath201 expansion . in the case of @xmath93 and @xmath95 exchanges , it is convenient to express eq .
( [ t ] ) in terms of rescaled operators as t^_z , w(p_1,p_2 ) & = & i _
k=1 ^ 6 |f^w , z_k |u(p_2 ) |o^_c k u(p_1 ) , [ t ] where @xmath202 for @xmath203 and |o^_c 4 = o^_c 4 , |o^_c 5 = o^_c 5 , |o^_c 6 = o^_c 6 , [ chiralbasis2 ] such that all the form factors @xmath204 are now dimensionless .
analogously , for the higgs boson corrections , we define the rescaled operators t^_h(p_1,p_2 ) & = & i _ k=1 ^ 4 |f^h_k |u(p_2 ) |o^_v k u(p_1 ) , [ t ] where @xmath205 for @xmath206 , and |o^_v 3 = o^_v 3 , |o^_v 4 = o^_v 4 , with all form factors @xmath207 dimensionless . in the case of the @xmath93 and higgs boson corrections
, we will provide the results for the form factors by retaining the external fermion - mass dependence up to the second order @xmath208 , with @xmath209 . in the case of @xmath95 corrections
, we will give only the results at the leading order in the external masses , while retaining the exact analytical dependence in the internal loop masses .
then , for the case of the @xmath95 and @xmath93 corrections we obtain ward identity , |f^w_5 = -|f^w_3,|f^w_6 = |f^w_2-|f^w_1 , and the @xmath210 functions are given by g_1(x)=g_a(x ) , g_2(x)=x_w g_d(x ) , g_3(x)=-g_b(x ) , g_4(x)=g_c(x ) , while , due to the ward identities , we have |f^w_5 = -|f^w_3 |f^w_6 = |f^w_2-|f^w_1 .
the explicit expressions of the @xmath211 functions can be found in appendix [ cg ] . in the case of the form factors
@xmath212 involving a virtual higgs boson we obtain |f^h_1 & = & + & = & -|f^h_2 , with @xmath213 and @xmath214 , where @xmath66 and @xmath67 indicate the external and internal flavour indices respectively , while @xmath215 stands for the corresponding ckm matrix element .
now , we perform the expansion of the twelve form factors @xmath216 appearing in eq .
( [ tfc ] ) at the first order in @xmath200 , as in the flavour - conserving transitions , by retaining only the leading contribution in the external masses . due to the fact that in the flavour - changing case @xmath217 , some of the form factors will not vanish in the @xmath198 limit .
therefore , it is convenient to keep the notation of form factors @xmath216 as in the basis of eq .
( [ tfc ] ) , where they have different dimensions .
finally , the non - vanishing expressions of the form factors , up to the order @xmath146 , where we neglect terms of order @xmath218 and @xmath219 , are given by ll f_1(q^2)=q^2_f _ f g_a(x_f ) & f_2(q^2)=-(m_1 ^ 2-m_2 ^ 2)_f _ f
g_a(x_f ) + f_3(q^2)=- q^2 _ f _ f g_b(x_f ) & f_4(q^2)= _ f _ f g_c(x_f ) + f_5(q^2)= _ f _ f ( g_b(x_f ) - g_g(x_f ) ) & f_6(q^2)=- _ f _ f g_c(x_f ) + f_7(q^2)=2 ( m_1 ^ 2-m_2 ^ 2)_f _ f g_a(x_f)&f_10(q^2)=- _ f _ f ( g_a(x_f ) + g_h(x_f ) ) + f_11(q^2)=q^2_f _ f g_d(x_f ) & f_12(q^2)=- ( m_1 ^ 2-m_2 ^ 2)_f _ f g_d(x_f ) [ ff ] where @xmath220 is the weak coupling , and @xmath221 , with @xmath222 the corresponding ckm matrix element .
the expressions for the functions @xmath223 are reported in appendix b. we do not report the expression of the form factors @xmath224 since they are vanishing at the leading order in our approximation .
the unitarity of the ckm is expressed by the relation @xmath225 .
one can easily check that in the adopted approximation , the form factors in eq .
( [ ff ] ) satisfy the ward identity reported in @xcite .
in this section we turn to the computation of the weak corrections to the diagonal and non - diagonal cross sections in the limit of small momentum transfers . at the same time we will derive the expression of the effective hamiltonian describing the transition from the initial to the final state in the presence of an external gravity source . in particular , we will include in our analysis also the case in which the graviton has a small mass @xmath226 , detailing the modifications on the cross sections and on the hamiltonian respect to the case of general relativity ( gr ) . the difference between the massive and
the massless graviton cases can be worked out more directly starting from the standard scattering amplitude expressed in terms of the feynman propagator . here
we discuss the amplitude in this form , performing the heavy source limit at the final stage .
we recall that in momentum space , the 2-particle covariant gravitational scattering amplitude amplitude is given by = 8 g ( _ p_h^(q^2 ) ^ext _ ) , [ mamp ] where @xmath227 is the graviton propagator in momentum space , and where @xmath39 , and @xmath228 is the fourier transform of the emt associated to the external source . in the einstein theory ,
the graviton propagator in a covariant gauge is given by p_h^(q^2)= ( ^^ + ^^ - ^^ + q^(q ) ) , [ einstprop ] where the last term @xmath229 , which is gauge dependent , is a tensor made by linear combinations of an even number of @xmath137 momenta with open indices , such as for instance @xmath230 or @xmath231 .
nevertheless , due to the conservation of the energy - momentum tensor , the contribution of @xmath229 vanishes when contracted with @xmath232 or @xmath233 , leading to a gauge - invariant result . in a scenario
in which the graviton has a small mass @xmath226 , the corresponding graviton propagator in the unitary gauge takes a modified form p_g^(q^2)= ( ^^ + ^^ - ^^ ) , [ propgm ] where @xmath234 . as one can see
, there is a discontinuity in the above propagator in the massless graviton limit , which differs by a finite term from the massless case .
the scattering amplitude in the weak field approximation , in general relativy ( gr ) is given by _
gr&= & , [ mc ] while for a massive graviton ( mg ) it is given by _
[ mc ] at this point it is convenient to decompose the amplitude as & = & m^tree + , where @xmath235 stands for the usual tree - level contribution and @xmath236 contains the one - loop ew corrections to the emt .
now we will analyze the two cases of flavour - conserving and flavour - changing transitions . by applying the on - shell relations for the external fermion states and the condition that the external source is conserved , we obtain for the ew one - loop contributions @xmath237 we obtain ^w , z&=&_k=1 ^ 4 \{2 |f^w , z_k |u(p_2 )
|o^_c k u(p_1 ) _ ^ext -c|f^w , z_k |u(p_2 )
|o^ _ c k u(p_1 ) _ ^ext } [ heffwz ] for the corrections involving the exchanges of @xmath95 and @xmath93 gauge bosons in the loop and ^h&=&_k=1 ^ 3 \{2 |f^h_k |u(p_2 ) |o^_v k u(p_1 ) _ ^ext -c|f^h_k |u(p_2 ) a higgs field .
the parameter @xmath238 allows to describe both the massless and the massive gravity scenarios , with @xmath239 and @xmath240 respectively , while the effective scale @xmath241 is given by _ eff&=&()^1/4 10 ^ 11 gev . as we can see , the characteristic energy scale @xmath241 of the local interactions is almost 8 order of magnitudes smaller than the planck scale . in the case of massive gravity scenarios
we have assumed that @xmath242 and neglected terms of order @xmath243 in the graviton propagator .
these results show that the contribution of the ew corrections to the flavour - diagonal ( fd ) gravitational scattering of fermions on an external gravitational source , in the kinematic region @xmath244 , can be described by a local effective hamiltonian given by ^_eff&= & ( _ k=1 ^ 4|f^w , z_k q_c_k + _ k=1 ^ 3|f^h_k q_v_k ) , where @xmath245 and @xmath246 are local 8-dimensional operators for the @xmath194 and higgs boson corrections respectively , defined as q_c_k~[|(x ) |o^_c k ( x ) ] t_^ext(x ) , q_v_k~[|(x ) |o^_v k ( x ) ] t_^ext(x ) , with @xmath247 standing for a generic sm fermionic field . the characteristic energy scale @xmath248 is defined above .
the tree - level matrix elements of the effective hamiltonians given above , taken between on - shell external fermion states , match the results obtained in eqs.([heffwz ] ) and ( [ heffh ] ) by using the full standard model theory . following the case of the flavour - conserving transitions
, we consider now the flavour - changing scattering amplitude on an external gravitational source f_1(p_1 ) + t^ext f_2(p_2 ) + t^ext , where we will assume the two fermions @xmath249 and @xmath250 having different flavour and masses @xmath217 . despite the presence of the @xmath251 pole in the graviton propagator
, the flavour - changing gravitational scattering in general relativity turns out to be local .
indeed , we find ^fc_m_g=0 & = & _ f _ f |u(p_2)\ { ( ^ p^+^ p^)p_l g_a(x_f ) - g^m_+(cg_a(x_f)-(2c -1 ) g_b(x_f ) ) .
g^ m_- g_a(x_f ) } u(p_1 ) ^ext _ [ mgeneral ] . the second line in eq .
( [ mgeneral ] ) is zero if c=1 ( massless graviton ) and @xmath252 , showing that in the einstein theory of general relativity the @xmath251 pole cancels out .
this is a general result that holds also in the exact case , as can be easily proved using the ward identities . as we can see from eq .
( [ mgeneral ] ) , the flavour - changing gravitational scattering on an external source - for the @xmath253 flavour transitions - can be described by a local effective hamiltonian given by ^fc_eff|_m_g=0= ( q _ 1^_f _ f
g_a(x_f ) + q _ 3_f _ f ( g_b(x_f)-g_a(x_f ) ) ) , [ heffg ] where the @xmath254 are local eight - dimensional flavour - changing operators , which are defined as q _ 1,3= [ |_2(x ) o^_1,3 _ 1(x ) ] t_^ext(x ) , with the basis @xmath255 defined in eq . ( [ basis ] ) . in the expression above , @xmath256 and @xmath257 ,
stand for the fermion fields corresponding to the initial and final fermions with flavour 1 and 2 respectively , similarly to the diagonal case , but now of different flavours .
the amplitude for the flavour - changing gravitational scattering , in the case in which the graviton has a small but non - vanishing mass , has a remarkable difference with respect to the massless graviton case . due to the different structure of the massive graviton propagator ( c=2/3 )
, we obtain ^fc_m_g0 & = & _ f _ f |u(p_2)\ { ( ( ^ p^+^ p^)p_l - 23 g^m_+ ) g_a(x_f ) .
+ & + & . 23 g^ m_- g_a(x_f ) } u(p_1 ) ^ext _ [ mgmass ] . as we can see from eq .
( [ mgmass ] ) , the @xmath251 pole in the last term of the amplitude does not vanish in the limit @xmath258 .
the residue of the pole is proportional to @xmath259 , which in the case of on - shell external fermions is equal to @xmath260 .
this has important implications in the case of a flavour - changing scattering for massive gravity or in the interaction with a graviscalar potential .
indeed , the last term in eq .
( [ mgmass ] ) generates an off - diagonal contribution ( in flavour space ) to the newton potential , inducing long - distance flavour - changing interactions . on the other hand ,
the local part of the interaction can be described by the following effective hamiltonian which is similar to the one in eq .
( [ heffg ] ) , namely ^fc_eff|_m_g0= ( q _ 1^_f _ f
g_a(x_f ) - q_3^_f _ f g_a(x_f ) ) , [ heffg ] where we have assumed the graviton mass to be @xmath261 and neglected terms of order @xmath243 in the graviton propagator .
the behaviour of the cross sections at low momentum transfers is particularly interesting in the flavour - changing case , due to the possibility of generating transitions in the presence of a gravitational source . as such they may appear as additional contributions , though largely suppressed , to ordinary flavour oscillations
. we will start by considering the case of massive gravity , where this phenomenon has been pointed out before @xcite . * * massive gravity scenario * + we consider the anelastic flavour - changing scattering of fermions on an external gravitational source in the framework of a massive gravity scenario .
generalization to the case of graviscalar interactions are straightforward .
in particular , we consider the scattering on a heavy point - like particle @xmath262 f_1(p_1 ) + t f_2(p_2 ) + t , [ mfc ] where we assume that the particle t is very heavy and acts as a generator of a static gravitational field .
the fermions @xmath263 have masses @xmath264 respectively .
we assume that the gravitational field has a small mass in order to generate a long distance potential .
the amplitude of the process can be taken from eq .
( [ mgmass ] ) by retaining only the last term in parenthesis , proportional to the @xmath265 pole , and neglecting terms of order @xmath199 .
then , the leading contribution to the amplitude for the process in ( [ mfc ] ) reduces to _ 12 & & _ f _ f |u(p_2)\ { m_- g_a(x_f ) } u(p_1 ) ^ext _ , where @xmath266 , and @xmath228 stands for the fourier transform of the energy momentum tensor associated to the heavy external particle @xmath262 .
if we consider the external source @xmath228 to be a point - like heavy source as in ( [ emt ] ) , we then have _ 12 & = & _ f _ f g_a(x_f ) \{|u(p_2 ) m_- u(p_1 ) } , [ m12 ] now , we neglect the contribution of the @xmath226 in the denominator of eq .
( [ m12 ] ) , since we assume that @xmath267 . by summing over the polarizations and
mediated over the initial ones , we get & = & ( g_n m)^2 g_f^2(m_1 ^ 4-m_2 ^ 4)|c|^2 [ m12q ] + where
= , = , c=_f _ f g_a(x_f ) . and
@xmath136 indicates the modulus of the incoming 3-momentum , namely @xmath268 , with @xmath152 the energy of the incoming particle . + now from eq .
( [ m12q ] ) we can see that the angular distribution is fully peaked in the forward direction @xmath269 , and there are no infrared divergences in the @xmath270 limit due to the fact that @xmath217 .
notice that the term @xmath271 plays the role of a small parameter as the term @xmath272 does in the elastic case .
+ finally , by integrating over the scattering angle , the total cross section takes the form _
12&= & _ 12 ^ 0 \ { 1 + } ( 1 + ) , [ sigmafc ] where @xmath273 is the leading term in the zero order expansion in @xmath271 .
this is given by _
12 ^ 0&= & , [ sigma0 ] which depends only on the mass splittings and on other fundamental constants , but not on the energy .
we stress that the above results are obtained in the approximation of small momentum transfers , @xmath274 , hence the result in eq .
( [ sigmafc ] ) holds only in the low energy regime @xmath275 . *
* massless gravity * + in the case of ordinary gravity , it is interesting to consider massless external fermions .
the flavour - changing cross section , whose general expression has been given in eq .
( [ flav ] ) , in this limit takes the simplified form [ flav1 ] =( ) ^2 |_f _
f g_a(x_f)|^2 ^2 .
notice that if all the internal fermions which appear in the loop corrections are mass degenerate - and this covers also the complete massless case - then the unitarity of the ckm matrix causes the cross section to vanish .
on the other hand , eq . ( [ flav1 ] ) shows that an external gravitational field may induce , in general , flavour - changing transitions . the result in eq .
( [ flav1 ] ) holds in the range of @xmath274 .
before coming to our comments and conclusions , we present the expression of the related cross sections at large momentum transfers , in the cases in which we drop the masses of the external fermions . in the non - diagonal case
this study , in the weak case , has been presented in @xcite , while in the qed case the diagonal cross section at large @xmath146 has been given in @xcite .
we consider the limit of large and spacelike momenta @xmath65 , where we drop all the contributions proportional to the fermion mass @xmath116 .
the expansion is organized in terms of @xmath276 , with the only non vanishing form factors being given by f^z_1 & = & ^2 ( -5 a^2-(a+v)^2 ( - ) ( 2 ( -)+3)-12 a v-5 v^2 ) + f^z_2 & = & ^2 ( -5 a^2-(a - v)^2 ( - ) ( 2 ( -)+3)+12 a v-5 v^2 ) + f^z_3 & = & 2 m_z^2 ( ( v^2 - 7 a^2 ) ( -)+4 ( v^2 - 3 a^2 ) ) + f^z_4&= & - ( 4 a^2 ( - ) ( ( -)+4)+25 a^2+v^2 ) + f^z_5&= & - ( 2 ( v^2 - 7 a^2 ) ( -)-25 a^2 + 7 v^2 ) + f^z_6&= & ( 4 ^2(-)+6 ( -)+11 ) . + a similar limit on the form form factors with a virtual @xmath95 gives f^w_1 & = & + & & ) ] f^w_2&= & - + f^w_3 & = & + & & where we have set @xmath213 , and with the remaining form factor contributions being zero . also , we have defined v = i_3 -2 s_w^2 e_qa = i_3 c^2= v^2 + a^2 , with @xmath277 denoting the third component of weak isospin of the fermion and @xmath278 is the weinberg angle .
notice that we have kept generic the value of the ricci coupling @xmath11 , obtained from the term of improvement for the higgs doublet . using the expressions of the form factors given above , in the massless fermion limit , at high momentum transfers @xmath279 , the diagonal cross section
is then given by & = & + & & 2 ^ 2 ( ) .
+ in this specific limit the result does not depend on the parameter of improvement of the emt , @xmath11 .
we have presented the structure of the radiative corrections of a scattering of a fermion in an external gravitational field in the standard model , extending the original analysis of berends and gastmans @xcite which covered the qed case .
we have used mostly a potential approch , where the gravitational source is assumed to be heavy , such as for a large astrophysical entity , like a star , a planet or a galaxy . in these cases the external field is treated as generating an external schwartzschild background .
this limit is met in a huge variety of physical scatterings , such as those of neutrinos and dark matter particles , but also of leptons .
our results are rather general and may cover , in principle , a wide variety of cases , having included in our analysis also the corrections coming from a finite geometrical form factor of the external source .
the size of these corrections is obviously quite small , but the angles of deflections which underline their expressions , as shown in @xcite , becomes sizeable for ultra high energy cosmic rays , for collisions grazing black holes . in particular , the quadratic growth ( with energy ) of the angle of deflection of
a fermion - discussed in @xcite in the qed case - is therefore expected to contribute significantly to gravitational lensing , also in the weak case .
these features are of potential interest in the astrophysical context , especially for cosmic rays , which are characterized by huge energies , up to the greisen - zatsepin - kuzmin ( gzk ) cutoff @xcite .
+ + * acknowledgements * we thank mirko serino for discussions .
in this appendix we present a rigorous derivation of the potential given in eq .
( [ pot ] ) . in the case of a spherically symmetric and stationary source ,
the ansatz for the metric in polar coordinates , which takes the form [ sch1 ] ds^2=e^(r ) d t^2 - e^(r ) dr^2 - r^2 d- r^2 ^ 2d^2 introduces two undeterminate functions which depend on a radial coordinate @xmath280 , @xmath281 and @xmath282 .
the einstein equations in the vacuum determine these two functions in terms of a single parameter @xmath238 in the form e^(r)= , e^(r)=1- [ sch2 ] as [ ds0 ] ds^2=(1- ) dt^2 - dr^2 - r^2 d- r^2
^ 2d^2 this form needs some manipulation in order to be confronted with the retarded solution generated by @xmath47 , characterized by g_00&=&_00 + h_00=(1- ) + g_ii&= & _ ii+h_ii=-(1 + ) , with the full metric written as ds^2(1- ) dt^2 -(1 + ) dd . [ sch3 ] in order to perform this comparison we need to perform a change of coordinates r = r(1 + ) ^2 which allows to rewrite the vacuum solution ( [ sch1 ] ) in the form ds^2= ( ) ^2 dt^2 - ( 1 + ) ^4 ( dr^2 + r^2 d^2 + r^2 ^2 d^2 ) , [ sch4 ] which is characterized by a single factor multiplying the spatial part @xmath283 , as in the retarded solution , given in ( [ sch3 ] ) . taking the limit @xmath284 , which corresponds to the weak field approximation ( i.e @xmath285 ) , ( [ sch4 ] ) becomes ds^2(1- ) -(1 + ) ( dr^2 + r^2 d^2 + r^2 ^2
d^2 ) [ sch5 ] which allows us to identify @xmath286 . reintroducing @xmath287 , the speed of light , the metric generated by ( [ sch3 ] ) and ( [ sch5 ] ) in the vacuum takes the usual post - newtonian form ds^2(1 + 2)dt ^2 -(1 - 2 ) dd [ post ] in terms of the newtonian potential @xmath288 .
this gives h_00=h_ii= or , equivalently h_=- |s_. the derivation of the form of the metric in the inner region of the distributed source ( @xmath289 ) requires a further examination .
we recall that in this case , assuming that the source behaves as a perfect and stationary fluid , the emt takes the form t^=(+ p)u^u^- p g^=diag(,-p ,- p ,- p ) with @xmath290 . setting in eq .
( [ sch1 ] ) @xmath291 , einstein s equations combined with the conservation of the energy momentum tensor in the source core take the form & = & 4 r^2 ( r ) + -&=&((r ) + p(r ) ) ( ) + & = & - ( in geometrized units @xmath292 ) .
factors of @xmath186 and @xmath287 can be reintroduced by the replacements m,,p . at this point , if we assume that the density of the fluid be dominated by its rest mass density @xmath293 and we take the nonrelativistic limit by expanding the previous equations in @xmath294 , we obtain & = & 4 r^2 ( r ) + -&=&(r ) + & = & - [ ecco ] notice that we have 3 equations and 4 variables . in our case
we assume that the mass density is a constant @xmath295 and solve for @xmath296 and @xmath281 . in the weak field limit @xmath297
we set @xmath298 , and eq .
( [ ecco ] ) becomes = .
the equations can be immediately solved for a body of constant density in the spherical approximation @xmath299 as m(r)&= & r^3 + p(r)&=&=-g ^2(r^2-r^2 ) + ^int(r)&= & -(- ) .
introducing @xmath300 , the metric takes the form ds^2=(1 -3 b_n + r^2))dt^2 - dr^2 - r^2 d ^2 - r^2 ^ 2 d^2 on which we need to perform a a coordinate redefinition in order to factorize the euclidean differential length @xmath301 , as in ( [ sch4 ] ) . to identify this change of coordinates we start considering a rotationally symmetric metric akin ( [ sch1 ] ) written , for convenience , in the form ds^2= a(r ) dt^2 - b(r)dr^2 -r^2d^2 - r^2 ^ 2d^2 that we intend to rewrite as ds^2=(r ) dt^2 - ( r)(dr^2 - r^2 d^2-r^2 ^ 2d^2 ) by a radial redefinition @xmath302 .
we impose the conditions dr & = & dr + r & = & r [ twos ] with @xmath303 an unknown function of the new radial coordinate @xmath304 . in our case @xmath305 and @xmath306 .
the two conditions in ( [ twos ] ) generate a separable differential equation which can be integrated as [ solution ] _
^r=_r_0^r where @xmath307 .
notice that in this case we keep both @xmath308 and @xmath309 as external parameters , and fix them by the requirements that at the core radius @xmath81 the interior solution coincides with the schwarzschild solution in the vacuum .
the solution of ( [ solution ] ) is r(r)= which can be inverted in order to determine @xmath310 .
notice that from ( [ twos ] ) ( r)&= & + & = & which is valid beyond the weak field limit .
in the weak field limit ( r)=++o(b_n^2 ) the expression @xmath311 valid at @xmath312 is ( r)=b ( -3)+1 + o(b_n^2 ) at this point we introduce the matching conditions at @xmath312 between the interior and the exterior potentials at @xmath313 ( r=r)&= & 1 + 2=1 - 2 b_n + ( r=r)&= & 1- 2 = 1 + 2 b_n which fix the parameters as @xmath314 . at the same time
we obtain at the same order ( r)=1 + 2 , ( r)=1 -2 .
this shows that at @xmath312 the metric can be expressed as in ( [ pot ] ) as expected , on the other hand , from newtonian dynamics .
the approach that we have followed can be easily extended to higher order in @xmath315 in the `` tt '' component @xmath316 , while the expression of @xmath317 remains always valid .
+ notice that in the computation of the geometrical form factors of the source we need the integral _ r _
> d^3 ( ||-r ) which requires a regularization at large momentum , due to the presence of an oscillating factor .
the simplest way avoid any regularization is to redefine it as difference between the contribution of the entire region and that of the interior region ( @xmath318 ) , the latter being given by _ r _
< d^3 ( r- ||)= ( 1-
( || r ) ) .
therefore we obtain _ r _
> d^3 ( ||-r ) & = & d^3 - _ r _
< d^3 ( r- || ) + & = & - ( 1- ( || r ) ) + & = & ( || r ) .
the @xmath287 coefficients of the flavour - changing cross section eq .
( [ non1 ] ) , in the case of equal mass fermions , are given by f. a. berends and r. gastmans , ann .
phys . * 98 * , 225 ( 1976 ) .
g. degrassi , e. gabrielli , and l. trentadue , phys.rev . *
d79 * , 053004 ( 2009 ) , arxiv:0812.3262 . c. corian , l. delle rose , e. gabrielli , and l. trentadue , phys.rev .
* d87 * , 054020 ( 2013 ) , arxiv:1212.5029 .
c. corian , l. delle rose , e. gabrielli , and l. trentadue , phys.rev .
* d88 * , 085008 ( 2013 ) , arxiv:1303.1305 .
m. giannotti and e. mottola , phys . rev . *
d79 * , 045014 ( 2009 ) , arxiv:0812.0351 .
r. armillis , c. corian , and l. delle rose , phys . rev . *
d81 * , 085001 ( 2010 ) , arxiv:0910.3381 .
c. corian , l. delle rose , a. quintavalle , and m. serino , phys.lett .
* b700 * , 29 ( 2011 ) , arxiv:1101.1624 . c. corian , l. delle rose , and m. serino , phys.rev .
* d83 * , 125028 ( 2011 ) , arxiv:1102.4558 .
r. escribano , j. frere , d. monderen , and v. van elewyck , phys.lett . *
b512 * , 8 ( 2001 ) , arxiv : hep - ph/0105211 . o. mena , i. mocioiu , and c. quigg , astropart.phys .
* 28 * , 348 ( 2007 ) , arxiv : astro - ph/0610918 .
e. f. eiroa and g. e. romero , phys.lett . *
b663 * , 377 ( 2008 ) , arxiv:0802.4251 .
a. accioly and r. paszko , phys.rev . *
d69 * , 107501 ( 2004 ) .
f. sorge and s. zilio , class.quant.grav .
* 25 * , 225004 ( 2008 ) .
f. sorge , class.quant.grav .
* 29 * , 045002 ( 2012 ) .
j. callan , curtis g. , s. r. coleman , and r. jackiw , annals phys .
* 59 * , 42 ( 1970 ) .
j. lawrence , annals phys .
* 58 * , 47 ( 1970 ) .
g. zatsepin and v. kuzmin , jetp lett .
* 4 * , 78 ( 1966 ) . k. greisen , phys.rev.lett .
* 16 * , 748 ( 1966 ) . | we investigate the role of the electroweak corrections to the scattering cross section of standard model fermions with gravity .
we use both an approach of scattering off an external potential , where the gravitational field is treated as a classical background generated by a heavy source , and the usual interaction based on the one - graviton - exchange . in the potential appoach
we consider the fields both of a localized and of a distributed gravitational source of spherical symmetry and uniform density , separating the cases of interactions taking place both in the inner and external regions of the source .
this allows to make a distinction between interactions involving neutrinos and dark matter particles with a realistic gravity source , which cover the inner region , and the rest of the standard model fermions .
the role of the gravitationally induced flavour - changing transitions , as well as the flavour diagonal ones , are investigated in the limit of both large and small momentum transfers , deriving the structure of the corresponding hamiltonian . * fermion scattering in a gravitational background : * electroweak corrections and flavour transitions _ @xmath0dipartimento di matematica e fisica `` ennio de giorgi '' , universit del salento and + infn - lecce , via arnesano , 73100 lecce , italy
_ + |
the study of the heavy quarkonium states at finite temperature got impetus after the proposal by matsui and satz @xcite where the dissociation of quarkonium due to the color screening in the deconfined medium signals the formation of quark gluon plasma ( qgp ) @xcite .
the assumption behind the proposal is that the medium effects can be envisaged through a temperature - dependent heavy quark potentials and have been studied over the decades either phenomenologically or through lattice based free - energy calculations @xcite . in recent years
there have been important theoretical developments in heavy quarkonium physics where a sequence of effective field theories ( eft ) @xcite are derived by exploiting the hierarchy of different scales of heavy quark bound state : @xmath1 , due to its large quark mass ( @xmath2 ) .
for example , the heavy quark system can be described by non - relativistic quantum chromodynamics ( nrqcd ) obtained from qcd by integrating out the mass . to describe the bound state of two quarks
, one can further integrate out the typical momentum exchange ( @xmath3 ) between the bound quarks @xcite and leads to potential non - relativistic qcd ( pnrqcd ) which describes a bound state by a two point function satisfying the schrdinger equation through the potentials as the matching coefficients of the lagrangian .
the eft can also be generalized to finite temperature to justify the use of potential models at finite temperature @xcite but the thermal scales ( @xmath4 , @xmath5 etc . )
make the analysis complicated .
_ for example _ , when the binding energy is larger than the temperature , there is no medium modifications of the heavy quark potential @xcite but the properties of quarkonia states will be affected through the interactions with ultra - soft gluons . as a result the binding energy gets reduced and a finite thermal width is developed due to the medium induced singlet - octet transitions arising from the dipole interactions @xcite .
this temperature regime is relevant for the @xmath6(1s ) suppression at the lhc . in the opposite limit ( binding energy @xmath7 )
, the potential acquires an imaginary component @xcite .
however beyond the leading - order , the above distinctions are no more possible . in non eft
, the heavy quark potential is defined from the late - time behavior of the wilson loop @xcite and can be directly calculated either in euclidean - time lattice simulations or in perturbation theory @xcite . however the definition of the heavy - quark potential related to the finite - temperature real - time wilson loop , as employed in the lattice qcd extraction is also based on an application of eft @xcite , where the derivation proceeds on the level of nrqcd and happens to be complex @xcite . the imaginary part of the potential can be interpreted as the landau damping @xcite which describes the decaying of the @xmath0 correlation with its initial state due to scatterings in the plasma .
the separation of thermal scales in eft ( @xmath8 ) ( in weak - coupling regime ) , in practice is not evident and one needs lattice techniques to test the approach . to understand the color screening in the strong - coupling regime ,
lattice calculations of the spatial correlation functions of static quarks are needed . in principle
it is possible to study the problem of quarkonium dissolution without any use of potential models .
recently a lot of progress has been made in this direction in which the in - medium properties of different quarkonium states are encoded in spectral functions in terms of the euclidean meson correlation functions constructed on the lattice @xcite .
however , the reconstruction of the spectral functions from the lattice meson correlators turns out to be very difficult , and despite several attempts its outcome still remains inconclusive .
one remarkable feature of the studies of the lattice meson correlators is their feeble temperature dependence despite the expected color screening .
this seems to be puzzling ! not only is the determination of the effective potential still an open question but also there are other related issues such as relativistic effects , thermal width of the states and contribution from quantum corrections that need to be taken care of . the physical picture of quarkonium dissociation in a medium has undergone a theoretical refinements over the last couple of years@xcite .
experimentally , the properties of thermally produced heavy quarkonium can be observed through the energy spectrum of their decay products ( dilepton pair ) @xcite .
the dissociation of quarkonium resonances correspond to the disappearance of their peaks in the dilepton production rate .
however , merely estimating the energy levels from the potential models does not allow one to reconstruct the spectral function , which can determine the production rate @xcite .
physically a resonance dissolves into a medium through the broadening of its peak gradually , due to its interaction with the partons in the medium .
earlier it was thought that a quarkonium state is dissociated when the debye screening becomes so strong that it inhibits the formation of bound states but nowadays a quarkonium is dissociated at a lower temperature @xcite even though its binding energy is nonvanishing , rather is overtaken by the landau - damping induced thermal width @xcite , obtained from the imaginary part of the potential .
its consequences on heavy quarkonium spectral functions @xcite , perturbative thermal widths @xcite quarkonia at finite velocity @xcite , in a t - matrix approach @xcite , and in stochastic real - time dynamics @xcite have been studied .
recently the dynamical evolution of the plasma was combined with the real and imaginary parts of the binding energies to estimate the suppression of quarkonium @xcite in rhic and lhc energies . as discussed above , in - medium corrections to the potential
are always accompanied with both real and imaginary components . in the weak coupling regime ,
the landau damping caused by the imaginary component is the principal mechanism for the dissociation of heavy quark bound states .
hence any realistic calculation of the spectral functions needs to incorporate both of the real and imaginary part .
however the separation of the scales , which are related to the screening of static electric fields ( @xmath9 ) and magnetic fields ( @xmath10 ) etc .
, are not satisfied at the strong - coupling limit and thus needs to handle nonperturbatively through the lattice studies .
although the lattice studies have shown that a sizable imaginary component is visible in the potential @xcite but may not be reliable because the necessary quality of the data has not yet been achieved .
one thus needs inadvertent support from the potential models at finite temperature as an important tool to complement the lattice studies .
usually potential model studies are limited to the medium - modification of the perturbative part of the potential only .
it is found that the bulk properties of the qcd plasma phase , _
e.g. _ the screening property , equation of state @xcite etc .
deviate from the perturbative predictions , even beyond the deconfinement temperature . in the sequel ,
the phase transition in qcd for physical quark masses is found to be a crossover @xcite .
it is thus reasonable to assume that the string - tension does not vanish abruptly at the deconfinement point @xcite , so one should study its effects on heavy quark potential even above @xmath11 .
this issue , usually overlooked in the literature where only a screened coulomb potential was assumed above @xmath11 and the linear / string term was neglected , was certainly worth investigation .
sometimes one - dimensional fourier transform of the cornell potential was employed with the assumption of color flux tube @xcite in one - dimension but at finite temperature , it may not be the case since the flux tube structure may expand in more dimensions @xcite .
therefore , it would be better to consider the three - dimensional form of the medium modified cornell potential .
recently a heavy quark potential was obtained by correcting both perturbative and nonperturbative terms in the cornell potential , not its coulomb part alone , with a dielectric function encoding the effects of the deconfined medium @xcite .
the inclusion of nonvanishing string term , apart from the coulomb term made the potential more attractive which can be seen by an additional long range coulomb term , in addition to the conventional yukawa term . in the short distance limit
, the potential reduces to the vacuum one , _
i.e. _ , the @xmath0 pair does not see the medium whereas in the large distance limit , potential reduces to a long - range coulomb potential with a dynamically screened - color charge . thereafter with this potential ,
the binding energies and dissociation temperatures of the ground and the lowest - lying states of charmonium and bottomonium spectra have been determined @xcite .
the discussions on the medium modifications of quarkonium properties referred above are restricted to isotropic medium only , it was until recently where the effect of anisotropy is considered in the heavy - ion collisions @xcite .
at the very early time of collision , asymptotic weak - coupling enhances the longitudinal expansion substantially than the radial expansion , thus the system becomes colder in the longitudinal direction than in the transverse direction and causes an anisotropy in the momentum space .
the anisotropy thus generated affects the evolution of the system as well as the properties of quarkonium states . in recent years
, the effects of anisotropy on both real and imaginary part of the heavy - quark potential and subsequently on the dissociation of quarkonia states have been investigated in an anisotropic medium @xcite extensively .
recently we extended our aforesaid calculation @xcite for an isotropic medium to a medium which exhibits a local anisotropy in the momentum space by correcting the full cornell potential through the hard - loop resumed gluon propagator @xcite .
the presence of anisotropy introduces an angular dependence , in addition to inter - particle separation , to the potential which is manifested in weakening the screening of the potential . as a result
the resonances become more bound than in isotropic medium .
since the weak anisotropy represents a perturbation to the ( isotropic ) spherical potential , we obtained the first - order correction due to the small anisotropic contribution to the energy eigenvalues of spherically - symmetric potential and explored how the properties of quarkonium states change in the anisotropic medium .
for example , the dissociation temperatures are found minimum for the isotropic case and increase with the increase of anisotropy . in the present work we aim to calculate the imaginary part , in addition to the real part of the potential both in isotropic and anisotropic medium by correcting the full cornell potential , not its coulomb part alone .
therefore , we first revisit the leading anisotropic corrections to the real and imaginary part of the retarded , advanced and symmetric propagators through their self energies , and then plug in their static limit to evaluate the real and imaginary part of the static potential , respectively .
this imaginary part provides a contribution to the width ( @xmath12 ) of quarkonium bound states @xcite which in turn determines their dissociation temperatures by the criterion : dissociation point of a particular resonance is defined as the temperature where the twice of the ( real part of ) binding energy equals to @xmath12 @xcite or from the intersection of the real and imaginary part of the binding energies .
the structure of our paper is as follows .
section 2 is devoted to the formalism of the potential in both isotropic and anisotropic medium .
so we have started with a review of the retarded , advanced and symmetric propagators and self energies in keldysh representation and their evaluation in htl resummed theory in both isotropic and anisotropic medium in section 2.1 . with these ingredients , we calculate the real and imaginary part of the ( static ) potential and subsequently studied the dissociation of charmonium and bottomonium states by calculating their real and imaginary binding energies and ( thermal ) widths for isotropic and anisotropic medium in subsection(s ) 2.2 and 2.3 , respectively .
moreover we have shown our results and try to explain them in terms of various effects : the contribution of the non - perturbative ( string ) term , the anisotropy , the screening scale etc .
finally , we conclude our main results in section 3 . +
as discussed earlier , any meaningful discussion of quarkonium properties in thermal medium should include both real and imaginary parts for the temperature - dependent potential .
the hierarchy of scales assumed in weak coupling eft calculations may not be satisfied and the adequate quality of the data is not available in the present lattice calculations , so one uses the potential model to circumvent the problem . because of the heavy quark mass ( @xmath2 ) , the requirement : @xmath13 and @xmath14 is satisfied for the description of the interactions between a pair of heavy quarks and antiquarks at finite temperature , in terms of quantum mechanical potential .
so we can obtain the medium - modification to the vacuum potential by correcting its _ both short and long - distance part _ with a dielectric function @xmath15 encoding the effect of deconfinement @xcite @xmath16 where we have subtracted an @xmath17-independent term ( to renormalize the heavy quark free energy ) which is the perturbative free energy of quarkonium at infinite separation @xcite .
the functions , @xmath18 and @xmath19 are the fourier transform ( ft ) of the cornell potential and the dielectric permittivity , respectively . to obtain the ft of the potential , we regulate both terms with the same screening scale . however in the framework of debye - hckel theory , digal et al .
@xcite employed different screening functions , @xmath20 and @xmath21 for the coulomb and string terms , respectively , to obtain the free energy .
at present , we regulate both terms by multiplying with an exponential damping factor and is switched off after the ft is evaluated .
this has been implemented by assuming @xmath17- as distribution ( @xmath22 @xmath23 .
the ft of the linear part @xmath24 is @xmath25 after putting @xmath26 , we obtain the ft of the linear term @xmath27 as , @xmath28 the ft of the coulomb piece is straightforward , thus the ft of the full cornell potential becomes @xmath29 the dielectric permittivity will be calculated once the self energies and propagators are obtained in htl resummation theory .
the naive perturbative expansion , when applied to gauge fields , suffers from various singularities and even the damping rate becomes gauge dependent @xcite .
diagrams which are of higher order in the coupling constant ( @xmath30 ) contribute to leading order .
these problems can be partly avoided by using the hard thermal loop ( htl ) resummation technique @xcite to obtain the consistent results , which are complete to leading - order . at the same time the infrared behavior is improved by the presence of effective masses in the htl propagators .
the htl technique has been shown to be equivalent to the transport approach @xcite and is more advantageous because it can be naturally extended to fermionic self - energies and to higher - order diagrams beyond the semiclassical approximation .
we shall now calculate the finite temperature self energies and propagator in real - time formalism @xcite where the propagators acquire a @xmath31 matrix structure : @xmath32 where each component has zero and finite temperature part which contains the distribution functions . in equilibrium ,
the distribution functions correspond to either ( isotropic ) bose ( @xmath33 ) or fermi distribution ( @xmath34 ) function .
away from the equilibrium , the distribution function needs to be replaced by the corresponding non - equilibrium one extracted from viscous hydrodynamics .
the nonequilibrium situation arises due to preferential expansion and non zero viscosity and as a consequence , a local anisotropy in momentum space sets in .
however , we consider a system close to equilibrium where the distribution function can be obtained from an isotropic one by removing particles with a large momentum - component along the direction of anisotropy @xcite , @xmath35 , i.e. , @xmath36~. \label{fun}\ ] ] the anisotropic parameter @xmath37 is related to the shear viscosity - to - entropy density ( @xmath38 ) through the one - dimensional navier stokes formula by @xmath39 where @xmath40 denotes the expansion rate of the fluid element .
the degree of anisotropy is generically defined by , @xmath41 where @xmath42 and @xmath43 are the components of momentum parallel and perpendicular to the direction of anisotropy , @xmath44 , respectively .
the positive and negative values of @xmath45 correspond to the squeezing and stretching of the distribution function in the direction of anisotropy , respectively .
however , in relativistic nucleus - nucleus collisions , @xmath37 is found to be positive .
a useful representation of the propagators in real - time formalism is the keldysh representation where the linear combinations of four components of the matrix , of which only three are independent , give the relation for the retarded ( r ) , advanced ( a ) and symmetric ( f ) propagators , respectively : @xmath46 only the symmetric component involves the distribution functions and is of particular advantage for the htl diagrams where the terms containing distribution functions dominate .
the similar relations for the self energies are : @xmath47 resumming the propagators through the dyson - schwinger equation , the retarded ( advanced ) and symmetric propagators can be written as @xmath48 substituting the symmetric propagator @xmath49 in terms of the retarded and advanced propagator , the resummed symmetric propagator can be expressed as @xmath50 \nonumber \\ & & + d_{r}(p)\,\left[\pi _ f(p)-(1 + 2f_b)\ , \mbox{sgn}(p_0)\ , [ \pi _ r(p)-\pi _ a(p)]\right ] \ , d_a(p)~. \label{2b8}\end{aligned}\ ] ] to calculate the static potential in isotropic medium , only the temporal component ( l ) of the propagator is needed so the retarded ( advanced ) propagator in the simplest coulomb gauge can be written as @xmath51 so far the resummation is done in isotropic medium , however we now extend it in a medium which exhibits a weak anisotropy ( @xmath52 )
. therefore we first expand the propagators and self energies around isotropic limit and retain only the linear term : @xmath53 thus in the presence of small anisotropy , the temporal component of the retarded ( advanced ) propagator becomes @xmath54 whereas with the notations for the difference of propagators and self - energies : @xmath55 $ ] , @xmath56 $ ] , @xmath57 $ ] , and @xmath58 $ ] , the symmetric propagator can be obtained @xcite , @xmath59 \ , { d}^l_{a(iso)}(p)~. \label{3b7}\end{aligned}\ ] ] to solve the propagators , we will now calculate the gluon self energy from the quark and gluon loops . the contribution of the quark loop @xcite to the self energy with external and internal momenta as @xmath60 and @xmath61 , respectively ( with @xmath62 ) : @xmath63 \label{en } \end{aligned}\ ] ] gives the retarded self energy @xmath64-
tr[\gamma^{\mu}s_{21}(q)\gamma^{\nu}s_{12}(k)]\right ) .
\label{en1 } \end{aligned}\ ] ] redefining the fermionic propagators : @xmath65 , the longitudinal - part of the self energy becomes , in the limit of massless quarks @xmath66 .
\label{en2 } \end{aligned}\ ] ] in the weak - coupling limit , the internal momentum ( @xmath4 ) is much larger than the external momentum ( @xmath9 ) , so the retarded self energy in the htl - approximation simplifies into @xcite @xmath67 after convoluting the distribution function , @xmath34 for quarks in an ( weakly ) anisotropic medium from ( [ fun ] ) the retarded quark self energy becomes @xmath68 with @xmath69 here , the angle ( @xmath70 ) is between @xmath35 and @xmath71 and @xmath72 .
after decomposing into isotropic ( @xmath37=0 ) and anisotropic ( @xmath73 ) pieces , the isotropic and anisotropic terms become @xmath74 respectively . in htl - limit , the structure of gluon - loop contribution is the same as the quark - loop , apart from the degeneracy factor and distribution function , so the quark and gluon loops together give the isotropic part of retarded ( advanced ) self - energy @xmath75 with the prescriptions @xmath76 ( @xmath77 ) , for the retarded ( advanced ) self - energies , respectively whereas the anisotropic part for the retarded ( advanced ) self energies are @xmath78 where
@xmath79 (= @xmath80 ) ) is the square of debye mass .
similarly the isotropic and anisotropic terms for the temporal component of the symmetric part are given by @xmath81 thus the gluon self - energy is found to have both real and imaginary part which are responsible for the debye screening and the landau damping , respectively where the former is usually obtained from the retarded and advanced self energy and the later is obtained from the symmetric self energy alone .
so , to evaluate the real part of the static potential , the real part of the temporal component of retarded ( or advanced ) propagator ( in static limit ) is needed @xmath82 while for the imaginary part of the potential , the imaginary part of the temporal component of symmetric propagator is given by @xmath83 with these real and imaginary part of the self energies and propagators , we will obtain the ( complex ) potential in subsection(s ) 2.2 and 2.3 for isotropic and anisotropic medium , respectively .
the real part of the static potential can thus be obtained from eq.([defn ] ) by substituting the dielectric permittivity @xmath19 in terms of the physical
11"- component of the gluon propagator .
the relation between the dielectric permittivity and the static limit of the 00"-component of gluon propagator in coulomb gauge is obtained from the linear response theory : @xmath84 where the propagator @xmath85 can be separated into real and imaginary parts as @xmath86 the real and imaginary parts can be further recast in terms retarded / advanced and symmetric parts , respectively @xmath87 thus using the real part of retarded ( advanced ) propagator in isotropic medium @xmath88 the real - part of the dielectric permittivity ( also given in @xcite ) becomes @xmath89 note that this one - loop result in the linear response theory is a perturbative one , where the linear approximation in qcd holds as long as the mean - field four potential ( @xmath90 ) is much smaller that the temperature @xcite .
actually for the soft scales , the mean - field four potential is at the order of @xmath91 and the linear approximation holds in the weak - coupling limit .
however , if one assumes nonperturbative effects such as the string tension survive even much above the deconfinement point then the dependence of the dielectric function on the debye mass may get modified .
so there is a _ caveat _ about the validity of the linear dependence of the dielectric function ( @xmath92 ) on the square of the debye mass @xmath93 . for the sake of simplicity
, we put the remnants of the nonperturbative effects beyond the deconfinement temperature by a multiplication factor @xmath94 to the leading - order debye mass , to take into account the next - to - leading corrections @xcite ( the factor is also obtained by fitting with the lattice results for the color - singlet free energy @xcite ) .
@xmath95 where @xmath96 and @xmath97 correspond to the medium modifications to the coulomb and string term , respectively . after performing the momentum integration ,
the coulomb term becomes @xmath98 and the string term simplifies into @xmath99 the real part of the potential in isotropic medium becomes ( with @xmath100 ) @xmath101 which is found to have an additional long range coulomb term , in addition to the conventional yukawa term . ) and without ( @xmath102 ) non - perturbative term in the potential .
the left ( right ) panel of the figure denote the results obtained with the leading - order and lattice - fitted debye masses , respectively.,title="fig:",width=222,height=245 ] ) and without ( @xmath102 ) non - perturbative term in the potential .
the left ( right ) panel of the figure denote the results obtained with the leading - order and lattice - fitted debye masses , respectively.,title="fig:",width=222,height=245 ] [ jj ] in the small - distance limit ( @xmath103 ) , the above potential reduces to the cornell potential , _
i.e. _ the @xmath0-pair does not see the medium .
on the other hand , in the long - distance limit ( @xmath104 ) , the potential is simplified into , with high temperature approximation ( ie .
@xmath105 can be neglected ) : @xmath106 which , apart from a constant term , is coulomb - like potential by identifying @xmath107 with the square of the strong coupling ( @xmath108 ) .
however if we compare the asymptotic limit ( @xmath109 ) of our result ( 39 ) with the digal et .
al @xmath110 the difference will be seen only in the string term only and may be due to the treatment of the problem classically or quantum mechanically .
if we compare them quantitatively ( with the debye mass @xmath111= @xmath112 ) , the difference becomes tiny . to see the effect of the linear term on the potential , in addition to the coulomb term
, we have plotted the ( real - part ) potential ( in fig.[jj ] ) with ( @xmath102 ) and without string term ( @xmath113 ) .
we found that the inclusion of the linear term makes the potential attractive , compared to potential with the coulomb term only .
furthermore , to see the effects of the screening scale , we have also computed the potential with the debye mass in next - to - leading order ( 1.4@xmath114 ) which is seen less stronger than the leading - order result .
to see the effects of medium on the potential at t=0 , we have evaluated the potential at different temperatures _ viz .
_ at @xmath115 , @xmath116 and @xmath117 , where the potential is found to decrease with the temperature at large distances and becomes short - range .
thus the deconfinement is reflected clearly in the large - distance behavior of heavy quark potential at finite temperature , where the screening is operative .
thus the in - medium behavior of heavy quark bound states is used to probe the state of matter in qcd thermodynamics . the imaginary part of the potential originates from the static limit of symmetric self energy .
cutting rules at finite temperature allows one to obtain the imaginary part by cutting open one of the hard thermal loop of the htl propagator which represents physically the inelastic scattering of the off - shell gluon off a thermal gluon @xcite , i.e. @xmath119 . the imaginary part of the potential plays an important role in weakening the bound state peak or transforming it to mere threshold enhancement .
it leads to a finite width ( @xmath120 ) for the resonance peak in the spectral function , which , in turn , determines the dissociation temperature .
dissociation is expected to occur while the ( twice ) binding energy decreases with the temperature and becomes equal to @xmath121 @xcite . to obtain the imaginary part of the potential in isotropic medium ,
we write the temporal component of the symmetric propagator from ( [ f00 ] ) for @xmath122 , in the static limit , @xmath123 however the same ( [ isof ] ) could also be obtained for partons with space - like momenta ( @xmath124 ) from the retarded ( advanced ) self energy ( [ 3b13 ] ) , using the relation @xcite : @xmath125 thus the imaginary part of the symmetric propagator ( [ isof ] ) gives the imaginary part of the dielectric function in isotropic medium : @xmath126 one can then similarly find the imaginary part of the potential from the definition of potential ( [ defn ] ) @xmath127\nonumber\\ & \equiv & i m v_{1(iso)}(r , t)+im v_{2(iso)}(r , t)~,\end{aligned}\ ] ] where @xmath128 and @xmath129 are the imaginary parts of the potential due to the medium modification to the short - distance and long - distance terms , respectively : @xmath130,\nonumber\\ i m v_{2(iso)}(r , t)&=&-\frac{4\sigma}{({2\pi})^2}\int \frac{d^3 { \bf p } } { ( 2\pi)^{3/2}}(e^{i { \bf p } \cdot { \bf r } } -1 ) \frac{1}{p^2 } \left[\frac{\pi t m_d^2}{p(p^2+m_d^2)^2 } \right].\end{aligned}\ ] ] after performing the integration , the contribution due to the short - distance term to imaginary part becomes ( with @xmath131 ) @xmath132 and the contribution due to the string term becomes @xmath133 where the functions , @xmath134 and @xmath135 at leading - order in @xmath136 are @xmath137 in the short - distance limit ( @xmath138 ) , both the contributions , at the leading logarithmic order , reduce to @xmath139 thus the sum of coulomb and string term gives the imaginary part of the potential in isotropic medium : @xmath140 one thus immediately observes that for small distances the imaginary part vanishes and its magnitude is larger than the case where only the coulombic term is considered @xcite and thus enhances the width of the resonances in thermal medium .
the imaginary part of the potential , in small - distance limit , is a perturbation to the vacuum potential and thus provides an estimate for the width ( @xmath12 ) for a resonance state and can be calculated , in a first - order perturbation , by folding with the unperturbed ( 1s ) coulomb wavefunction @xmath141 the main features of our results on the thermal width in ( fig.[gama1 ] ) are : first the width always increases with the temperature .
secondly the inclusion of the non vanishing nonperturbative string term , in addition to the coulomb term , makes the width larger than the earlier result with the perturbative coulomb term @xcite only and thus the damping of the exchanged gluon in the heat bath provides larger contribution to the dissociation rate and consequently reduce the yield of dileptons in the peak .
( left ) and @xmath6 ( right ) states with and without nonperturbative ( string ) term in an isotropic medium with the debye masses in leading - order and the lattice fitted result.,title="fig:",width=222,height=245 ] [ gamaiso ] ( left ) and @xmath6 ( right ) states with and without nonperturbative ( string ) term in an isotropic medium with the debye masses in leading - order and the lattice fitted result.,title="fig:",width=222,height=245 ] [ upsi ] the effect of nonperturbative term on the width is relatively more on @xmath142 than @xmath6 state because the binding of @xmath6 ( 1s ) state is more coulombic than @xmath142 ( 1s ) state .
this may have far reaching implications on the dissociation in medium .
thirdly the width is also affected by the screening scale we chose to regulate the potential , _ namely _ the width with the higher screening scale ( 1.4 @xmath143 ) is more than the leading - order result because the width , @xmath12 increases with the increase of the debye mass . to understand the in - medium properties of the quarkonium states , one need to solve the schrdinger equation with both the real and imaginary part of the finite temperature potential .
as seen earlier , in the short - distance limit , the vacuum contribution dominates over the medium contribution whereas in the long - distance limit the real part of the potential reduces to a coulomb like potential and thus yields the real part of the binding energy in isotropic medium :
@xmath144 however in the intermediate - distance ( @xmath145 ) scale , the interaction becomes complicated and the potential does not look simpler in contrast to the asymptotic limits , thus the complex potential in general needs to be dealt with numerically to obtain the real and imaginary binding energies .
there are some numerical methods to solve the schrdinger equation either in partial differential form ( time - dependent ) or eigen value form ( time - independent ) by the finite difference time domain method ( fdtd ) or matrix method , respectively . in the later method
, the stationary schrdinger equation can be solved in a matrix form through a discrete basis , instead of the continuous real - space position basis spanned by the states @xmath146 . here the confining potential v is subdivided into n discrete wells with potentials @xmath147 such that for @xmath148 boundary potential , @xmath149 for @xmath150 . therefore for the existence of a bound state , there must be exponentially decaying wave function in the region @xmath151 as @xmath152 and has the form : @xmath153 + q_{{}_e } \exp [ \gamma_{{}_{n+2}}(x - x_{n+1 } ) ] , \ ] ] where , @xmath154 , @xmath155 and , @xmath156 .
the eigenvalues can be obtained by identifying the zeros of @xmath157 .
and @xmath6 states with the temperature ( in units of critical temperature , @xmath11 ) in the left and right panel , respectively in an isotropic medium.,title="fig:",width=222,height=245 ] [ jpsi ] and @xmath6 states with the temperature ( in units of critical temperature , @xmath11 ) in the left and right panel , respectively in an isotropic medium.,title="fig:",width=222,height=245 ] [ upsi ] [ jpsi1 ] the binding energies shown in fig .
[ jpsi1 ] have the following features : first when the nonperturbative term is included , the ( real part ) binding of @xmath0 pairs gets stronger with respect to the case where only the coulomb term is included .
secondly there is a strong decreasing trend with the temperature because the screening becomes stronger with the increase of the temperature , so the real part of the potential becomes weaker compared to @xmath158 and results in early dissolution of quarkonia in the medium .
thirdly the real part of the binding energy decreases with the increase of screening scale ( 1.4 @xmath143 ) . on the other hand the imaginary part of the binding energy increases with the temperature .
thus the study of both the binding energies is poised to provide a wealth of information about the dissociation pattern of quarkonium states in thermal medium which will be used to determine the dissociation temperatures .
we will now study the dissociation in thermal medium to calculate the dissociation temperature ( @xmath159 ) either from the intersection of the ( real and imaginary ) binding energies @xcite , or from the conservative criterion on the width of the resonance as : @xmath160 @xcite .
although both definitions are physically equivalent but they are numerically different ( table 1 ) . for example , @xmath142 is dissociated at 2.45 @xmath11 obtained from the intersection of binding energies while the condition on width gives much lower temperature ( 1.40 @xmath11 ) . correspondingly @xmath6 ( 1s )
is dissociated at 3.40 @xmath11 and 3.10 @xmath11 , respectively .
our results are found relatively higher compared to similar calculation @xcite , which may be due to the absence of three - dimensional medium modification of the linear term in their calculation .
finally we explore the sensitivity of the screening scale on the dissociation mechanism where the dissociation temperatures computed with the next - to - leading order ( 1.4 @xmath143 ) debye mass are found smaller than the leading - order result ( table 2 ) .
_ for example
_ , @xmath142 s and @xmath6 s are now dissociated at 1.33 @xmath11 and 1.91@xmath11 , respectively . the space - time evolution of qgp relys on the viscous hydrodynamical treatment where the system assumes a local thermal equilibrium , i.e. close to isotropic in momentum space , which may not be true at the very early time in the collision of two nuclei , due to large momentum - space anisotropies @xcite .
the degree of anisotropy increases as the shear viscosity increases and thus one must address it while calculating the heavy quark potential in the presence of momentum - space anisotropies . the real - part of the heavy quark potential was first considered in @xcite and then the imaginary part was obtained theoretically @xcite as well as phenomenologically @xcite .
the main effect of the anisotropy is to reduce debye screening which , in turn has the effect that heavy quarkonium states can survive upto higher temperatures .
however the aforesaid works in anisotropic medium are limited to the medium - modification of the perturbative part only and the nonperturbative string term was assumed to zero .
however , the string - tension is non vanishing even at temperatures much beyond the deconfinement point @xcite , so one should study its effect on the heavy quark potential in anisotropic medium too . like in isotropic medium
, we obtain the real - part of the potential in weakly - anisotropic medium @xcite from the anisotropic corrections to the ( temporal component ) real - part of retarded propagator ( [ rtrdprop ] ) @xmath161 \nonumber\\ & \equiv & re v_{1(aniso)}({\bf r},\xi , t)+ re v_{2(aniso)}({\bf r},\xi , t)~,\end{aligned}\ ] ] where @xmath162 and @xmath163 are the medium modifications corresponding to the coulomb and string term , respectively , are given by @xmath164\\ re v_{2(aniso)}({\bf r},\xi , t ) & = & -\frac{4\sigma}{{2\pi}^{2}}\int d^{3}\mathbf{p } ( e^{i\mathbf{p } \cdot \mathbf{r}}-1)\frac{1}{p^2}\left[\frac{1}{(p^2+m_d^2 ) } -\frac{\xi m_d^2}{6(p^2+m_d^2)^2 } ( 3\cos 2\theta_p-1)\right]~.\end{aligned}\ ] ] to perform the momentum integration , we use the transformation @xmath165 , where @xmath166 and @xmath167 are the angles between * p * and * n * , * r * and * n * , respectively and @xmath168 , @xmath169 are the angular variables between the vectors , @xmath170 and @xmath171 .
so after the integration , the coulombic contribution to the potential becomes @xmath172\right.\right .
\times \nonumber\\ & & \left .
\left.(1 - 3\cos^2\theta_r)\right]\right]~ , \label{col}\end{aligned}\ ] ] and the string contribution is @xmath173\right]~. \label{strng}\end{aligned}\ ] ] thus the real - part of the potential in the anisotropic medium becomes @xmath174 \right .
\nonumber\\ & - & \left .
\frac{\alpha m_{d } } { 2}\frac{e^{-\hat{r}}}{\hat{r } } \left[\frac{e^{\hat{r}}-1}{\hat{r}^2}-\frac{1 } { \hat{r}}-\frac{\hat{r}e^{\hat{r}}}{3}+ \frac{\hat{r}}{6}-\frac{1}{2}\right ] \right .
\nonumber\\ & + & \left ( \left .
\frac{2 \sigma } { m_{d}}\frac{e^{-\hat{r}}}{\hat{r}}\left [ 3 \frac{e^{\hat{r}}-1}{\hat{r}^2}-\frac{e^{\hat{r}}}{4}- \frac{3}{\hat{r}}-\frac{\hat{r}}{4}-\frac{5}{4}\right ] \right.\right.\nonumber\\ & -&\frac{\alpha m_{d}}{2 } \frac{e^{-\hat{r}}}{\hat{r}}\left [ \left.\left .
3 \frac{e^{\hat{r}}-1}{\hat{r}^2}-\frac{3}{\hat{r}}-\frac{\hat{r}}{2}- \frac{3}{2}\right ] \right ) \cos~2 \theta_r \right ) \nonumber\\ & = & re v_{iso}(r , t ) + v_{\rm{tensor}}(r,\theta_r , t ) .
\label{fullp}\end{aligned}\ ] ] thus the anisotropy in the momentum space introduces an angular ( @xmath175 ) dependence , in addition to the interparticle separation ( @xmath17 ) , to the real part of the potential , in contrast to the @xmath17-dependence only in an isotropic medium . the real potential becomes stronger with the increase of anisotropy ( shown in fig.[para_plot11 ] ) because the ( anisotropic ) debye mass @xmath176 ( or equivalently angular - dependent debye mass @xmath177 ) in an anisotropic medium is always smaller than in an isotropic medium . as a result the screening of the coulomb and string contribution
are less accentuated , compared to the isotropic medium .
in particular the potential for quark pairs aligned in the direction of anisotropy are stronger than the pairs aligned in the transverse direction .
[ para_plot_a ] [ para_plot_b ] [ para_plot11 ] recently the imaginary part with a momentum - space anisotropy and its effects on the thermal widths of the resonance states have been studied @xcite , with the medium - modification to the perturbative ( coulomb ) term only .
the imaginary part of the potential arises due to the singlet - to - octet transitions induced by the dipole vertex as well as due to the landau damping in the plasma , _ i.e. _ scattering of the gluons with space - like momentum off the thermal excitations in the medium .
we follow their work by including the medium corrections to both perturbative ( coulombic ) and non - perturbative ( string ) terms in a weakly anisotropic medium .
like in isotropic medium , we can obtain the imaginary part of the potential by the leading anisotropic correction to the imaginary part of the ( temporal component ) symmetric propagator as @xmath179 \nonumber\\ & & \equiv i m v_{1(aniso ) } ( { \bf r},\xi , t)+ i m v_{2(aniso ) } ( { \bf r},\xi , t ) ~,\end{aligned}\ ] ] where @xmath180 and @xmath181 are the imaginary contributions corresponding to the coulombic and linear terms in anisotropic medium , respectively : @xmath182 \right],\\ i m v_{2\rm(aniso)}({\bf r},\xi , t)&=&\frac{4\sigma}{({2\pi})^2}\int \frac{d^3\mathbf{p } } { ( 2\pi)^{3/2}}(e^{i\mathbf{p } \cdot \mathbf{r}}-1 ) \frac{1}{p^2 } \left[\frac{-\pi t m_d^2}{p(p^2+m_d^2)^2 } + \xi\left[\frac{3\pi t m_d^2}{4p(p^2+m_d^2)^2}\sin^2{\theta_p}\right.\right .
\nonumber\\~ & & \left .
-\frac{2\pi t m_d^4}{p(p^2+m_d^2)^3 } ( \sin^2\theta_p-\frac{1}{3})\right]\right].\end{aligned}\ ] ] [ jpsi ] [ upsi ] [ upsi1 ] since the isotropic contribution is already calculated in sec.2.2.2 , so the anisotropic contribution to the perturbative term in the leading - order is given by @xcite @xmath183 , \end{aligned}\ ] ] where the functions @xmath184 and @xmath185 are @xmath186 , \nonumber\\ \phi_{2}(\hat{r},\theta_r)&= & \frac{{\hat{r}}^2}{90}(-4 + 3 \cos(2\theta_r ) ) .
\end{aligned}\ ] ] similarly the imaginary contribution due to the nonperturbative ( linear ) term can also separated into the isotropic and anisotropic term , where the isotropic part is already calculated in sec.2.2.2 and hence the anisotropic part is now calculated @xmath187~. \label{v2aniso}\end{aligned}\ ] ] the function , @xmath188 is given by @xmath189~,\end{aligned}\ ] ] where @xmath190 is given by @xmath191 substituting @xmath190 into @xmath192 and decomposing into @xmath175- dependent and independent terms , the function , @xmath192 can be rewritten as @xmath193\nonumber\\ & \equiv&\psi_1^{(1)}(\hat{r})+\psi_1^{(2)}(\hat{r},\theta_r ) ~ , \label{psi_1}\end{aligned}\ ] ] where the functions @xmath194 and @xmath195 are given by @xmath196\nonumber\\ & = & \hat r^4\int \frac{dx}{x(x^2+\hat r^2)^2}\left[1-\frac{3}{2}\left ( \frac { \sin(x)}{x}+\frac{\cos(x)}{x^2}-\frac{\sin(x)}{x^3}\right)\right]\nonumber\\ & = & \frac{\hat r^2}{10}+\frac{(-739 + 420\gamma_e + 420\log(\hat r))\hat r^4}{39200}+o(\hat r^5),\end{aligned}\ ] ] and @xmath197\nonumber\\ & = & \frac{3}{2}\hat r^4\int \frac{dx}{x(x^2+\hat r^2)^2}\left[\left(\frac{\sin(x)}{x}+\frac{3\cos(x)}{x^2 } -\frac{3\sin(x)}{x^3}\right)\cos^2\theta_r \right]\nonumber\\ & = & \left(-\frac{r^2}{20}+\frac{(176 - 105\gamma_e -105\log(\hat r))\hat r^4}{14700}+o(\hat r^5)\right)\cos^2\theta_r~.\end{aligned}\ ] ] the remaining function in the imaginary part of the potential associated with the linear term ( [ v2aniso ] ) can similarly be separated into @xmath175- dependent and independent terms : @xmath198\right]\nonumber\\ & = & -\frac{4}{3}\int\frac{dz}{z(z^2 + 1)^3}\left [ \left(1- \frac { 2\sin{z\hat r}}{z\hat r}-\frac{3\cos(z\hat r)}{(z\hat r)^2}+\frac{3\sin(z\hat r)}{(z\hat r)^3}\right)\right.\nonumber\\ & & \left.+ 3\left(\frac { \sin{z \hat r}}{z\hat r}+\frac{3\cos(z\hat r)}{(z\hat r)^2}-\frac{3\sin(z\hat r)}{(z\hat r)^3}\right)\cos^2\theta_r \right]\nonumber\\ & \equiv & \psi_2^{(1)}(\hat{r})+\psi_2^{(2)}(\hat{r},\theta_r ) , \label{psi_2}\end{aligned}\ ] ] where the functions @xmath199 and @xmath200 are given by @xmath201,\end{aligned}\ ] ] and @xmath202\cos^2\theta_r .\end{aligned}\ ] ] so the functions @xmath203 and @xmath204 are finally given by @xmath205 \nonumber\\ & & -4\left[-\frac{\hat r^2}{60}+\frac{\hat r^4}{840}+o(\hat r^5)\right ] \cos^2\theta_r~,\end{aligned}\ ] ] respectively and @xmath206 is the euler - gamma constant . and @xmath6 states in the anisotropic medium.,title="fig:",width=222,height=245 ] and @xmath6 states in the anisotropic medium.,title="fig:",width=222,height=245 ] finally the short and long - distance contributions , in the leading logarithmic order @xmath207 gives the imaginary part of the potential in the anisotropic medium @xmath208\log(\frac{1}{\hat r } ) ~,\end{aligned}\ ] ] which is found to be smaller than the isotropic medium and decreases with the increase of anisotropy ( shown in fig.[upsi1 ] ) . . [ cols="^,^,^,^,^",options="header " , ] like in isotropic medium , in weakly anisotropic medium too , the imaginary part is found to be a perturbation and thus provides an estimate for the ( thermal ) width for a particular resonance state : @xmath209\nonumber\\ & = & t\left(\frac{4}{\alpha m_q^2}+\frac{12\sigma}{\alpha ^2m_q^4}\right ) \left(1-\frac{\xi}{2}\right)m_d^2 \log\frac{\alpha m_q}{2m_d}~,\end{aligned}\ ] ] which shows that the width in anisotropic medium becomes smaller than in isotropic medium and gets narrower with the increase of anisotropy ( shown in fig .
[ gamaniso ] ) .
this is due to the fact that @xmath12 is approximately proportional to the ( square ) debye mass and the debye mass decreases in the anisotropic medium because the effective local parton density around a test ( heavy ) quark is smaller compared to isotropic medium . the real part of the potential thus obtained in anisotropic medium ( [ fullp ] ) , in contrast to its counterpart ( spherically symmetric potential ) in isotropic medium eq .
( [ pis ] ) is non - spherical and so one can not simply obtain the energy eigen values by solving the radial part of the schrdinger equation alone because the radial part is no longer sufficient due to the angular dependence in the potential . other way to understand
is that because of the anisotropic screening scale , the wave functions are no longer radially symmetric for @xmath73 .
so one has to solve the potential in three dimension but in the small @xmath37-limit , the non - symmetric component @xmath210 is much smaller than the symmetric ( isotropic ) component @xmath211 and thus can be treated as perturbation .
[ ju1 ] therefore , the corrected energy eigen values come from the solution of schrdinger equation of the isotropic component plus the first - order perturbation due to the anisotropic component @xmath210 . in short - distance limit
, the vacuum contribution dominates over the medium contribution even for the weakly anisotropic medium and in the long - distance limit , the real part of potential in high temperature approximation results a coulomb plus a subleading anisotropic contribution : @xmath212 where the anisotropic contribution ( @xmath213 ) is smaller than the isotropic one ( @xmath214 ) , so the anisotropic part can be treated as perturbation .
therefore , the real part of binding energy may be obtained from the radial part of the schrdinger equation ( of the isotropic component ) plus the first - order perturbation due to the anisotropic component as @xmath215 where the first term is the solution of ( radial - part ) of the schrdinger equation with the isotropic part ( @xmath214 ) and the second term is due to the anisotropic perturbation of the tensorial component ( @xmath213 ) calculated from the first - order perturbation theory . the real and imaginary part of the binding energies for the @xmath142 and @xmath6 states are computed numerically in fig . [ ju1 ] for different values of anisotropies , with the following observations : the inclusion of the nonperturbative string term makes the quarkonium states more bound in the anisotropic medium too . secondly the binding of @xmath0 pairs becomes stronger with respect to their isotropic counterpart and increases with the increase of anisotropy because the ( real part ) potential becomes deeper due to the weaker screening .
last but not the least , as the screening scale increases the binding gets weakened even in the anisotropic medium .
in contrast to the real part of binding energy , the imaginary part of the binding energy increases with the temperature but increases with the anisotropy . with these observations ,
we have now computed the dissociation temperatures at different anisotropies ( table 1 ) , where @xmath142 is dissociated at @xmath216 and @xmath217 for @xmath218 and @xmath219 , respectively ( obtained from the intersection of binding energies ) whereas @xmath6 s are dissociated at @xmath220 and @xmath221 , respectively .
thus the presence of anisotropy enhances the dissociation point to the resonances . like in isotropic medium
, we also computed the dissociation temperatures from the criterion on the thermal width and found the temperatures become smaller . for example , @xmath142 is now dissociated at @xmath222 and @xmath223 and @xmath6 is dissociated at @xmath224 and @xmath225 , for the same anisotropies .
we have investigated the properties of charmonium and bottomonium states through the in - medium modifications to both perturbative and nonperturbative of the cornell potential , not its perturbative term alone as usually done in the literature . for this purpose
we have obtained both the real and imaginary part of the potential within the framework of real - time formalism , in both isotropic and anisotropic medium . in isotropic medium ,
the inclusion of the linear / string term , in addition to the coulomb term , makes the real part of the potential more attractive .
so , as a consequence the quarkonium states become more bound compared to the medium modification to the coulomb term alone .
moreover the string term affects the imaginary part too where its magnitude is increased by the string contribution . as a result
, the ( thermal ) width of the states are broadened due to the presence of string term and makes the competition between the screening and the broadening due to damping interesting and plays an important role in the dissociation mechanism . with these cumulative observations , we studied the dissociation in a medium where a resonance is said to be dissolved in a medium @xcite either when its ( real ) binding energy decreases with temperature and becomes equal to its width or the real and imaginary binding energy becomes equal .
we have found that the quarkonium states are dissociated at higher temperature compared to the medium - consideration of the coulomb term only .
we have then extended our exploration of quarkonium to a medium which exhibits a local anisotropy in the momentum space .
this may arise due to the rapid expansion in the beam direction compared to its transverse direction , at the early stage of the evolution in ultra - relativistic heavy - ion collisions .
for that , we have first revisited the anisotropic corrections to the retarded , advanced and symmetric propagators through their self - energies in the hard - loop resummation technique and apply these results to calculate the medium - corrections to the perturbative and nonperturbative term of the cornell potential .
we are however restricted to a medium closer to equilibrium / isotropic because although the system was initially anisotropic but by the time quarkonium resonances are formed in plasma ( @xmath226 , @xmath227 is the formation time in the rest frame of quarkonium ) , the plasma becomes almost isotropic .
the effect of nonvanishing nonperturbative term on the quarkonium properties , as seen earlier , remains the same even in the presence of momentum anisotropy .
however , the anisotropy behaves as an additional handle to decipher the properties of quarkonium states , namely , in anisotropic medium , the binding of @xmath0 pairs gets stronger with respect to their isotropic counterpart because both the real and imaginary part of the complex potential becomes deeper with the increase of anisotropy .
this is due to the fact that the ( effective ) debye mass in anisotropic medium is always smaller than in isotropic medium . as a result
the screening of the coulomb and string contributions is less accentuated and thus quarkonium states are bound more strongly than in isotropic medium .
the overall observation is that the dissociation temperature increases with the increase of anisotropy .
for example , @xmath228 is dissociated at @xmath229 , @xmath216 , and @xmath217 for the anisotropies @xmath230 0 , 0.3 , and 0.6 , respectively .
similarly , @xmath6 is dissociated at @xmath231 , @xmath220 , and @xmath232 for @xmath122 , 0.3 , and 0.6 , respectively .
our results on the dissociation temperatures are found relatively higher compared to similar calculation @xcite , which may be due to the absence of three - dimensional medium modification of the linear term in their calculation .
in fact , one - dimensional fourier transform of the cornell potential yields the similar form used in the lattice qcd in which one - dimensional color flux tube structure was assumed @xcite .
however , at finite temperature that may not be the case since the flux - tube structure may expand in more dimensions @xcite .
therefore , it would be better to consider the three - dimensional form of the medium modified cornell potential which has been done exactly in the present work . in brief
, the properties of quarkonium states are affected by the inclusion of the non - perturbative ( string ) term in the potential , in addition to the anisotropic medium effects , which needs more careful treatment due to its nonperturbative character in future . 5 t. matsui and h. satz , phys .
b * 178 * , 416 ( 1986 ) .
binoy krishna patra , v agotiya , and v chandra , eur.phys.j .
* c67 * , 465 ( 2010 ) .
a. mocsy , p. petreczky phys.rev .
d * 77 * , 014501 ( 2008 ) .
a. mocsy eur.phys.j .
c * 61 * , 705 ( 2009 ) .
n. brambilla , a. pineda , j. soto , and a. vairo , nucl .
, b * 566 * , 275 ( 2000 ) . n. brambilla , a. pineda , j. soto , and a. vairo , rev . mod
. phys . * 77 * , 1423 ( 2005 ) .
w. e. caswell and g. p. lepage , phys .
b * 167 * , 437 ( 1986 ) .
b. a. thacker and g. p. lepage , phys .
d * 43 * , 196 ( 1991 ) .
g. t. bodwin , e. braaten , and g. p. lepage , phys .
d * 51 * , 1125 ( 1995 ) .
n. brambilla , j. ghiglieri , a. vairo and p. petreczky 2008 phys .
d * 78 * , 014017 ( 2008 ) .
k. g. wilson , phys .
d * 10 * , 2445 ( 1974 ) .
y. makeenko arxiv:0906.4487v1 [ hep - th ] .
j. berges , sz .
borsanyi , d. sexty , i .- o .
stamatescu phys.rev.d * 75 * , 045007 ( 2007 ) .
a. barchielli , e. montaldi and g. m. prosperi nucl .
b * 296 * , 625 ( 1988 ) .
a. rothkopf mod .
a * 28 * , 1330005 ( 2013 ) .
m. laine , o. philipsen , p. romatschke , and m. tassler , jhep * 03 * , 054 ( 2007 ) .
a. beraudo , j. p. blaizot , and c. ratti , nucl .
a * 806 * , 312 ( 2008 ) .
f. karsch , m. g. mustafa , m. h. thoma phys.lett .
b * 497 * , 249 ( 2001 ) . a. mocsy and p. petreczky , eur .
j. c * 43 * , 77 ( 2005 ) .
wong , phys .
c * 72 * , 034906 ( 2005 ) .
a. mocsy and p. petreczky , phys .
d * 73 * , 074007 ( 2006 ) .
d. cabrera and r. rapp , phys .
d * 76 * , 114506 ( 2007 ) .
a. mocsy and p. petreczky , phys .
lett . * 99 * , 211602 , 0706.2183 ( 2007 ) . w. m. alberico , a. beraudo , a. de pace , and a. molinari , phys .
d * 77 * , 017502 ( 2008 ) .
a. mocsy and p. petreczky , phys .
d * 77 * , 014501 , 0705.2559 ( 2008 ) .
de forcrand , et.al phys .
d * 63 * , 054501 ( 2001 ) .
d pal , binoy krishna patra and d k srivastava , eur .
c17 * , 179 ( 2000 ) .
binoy krishna patra and d k srivastava , phys . lett . * b505 * , 113 ( 2001 ) .
binoy krishna patra , v.j .
menon , nucl.phys . * a 708 * , 353 ( 2002 ) .
binoy krishna patra and v j menon , eur .
j. * c37 * , 115 ( 2004 ) .
binoy krishna patra and v j menon , eur .
j. * c44 * , 567 ( 2005 ) .
binoy krishna patra and v j menon , eur.phys.j .
* c48 * , 207 ( 2006 ) . b. schenke and m. strickland , phys .
d * 76 * , 025023 ( 2007 ) .
m. martinez and m. strickland , phys .
* 100 * , 102301 ( 2008 ) ; phys.rev .
c * 78 * , 034917 ( 2008 ) .
y. burnier , m. laine , and m. vepsalainen , jhep . * 01 * , 043 ( 2008 ) .
m. laine , o. philipsen , and m. tassler , jhep .
* 09 * , 066 ( 2007 ) ; t. hatsuda nucl .
phys . a 904 - 905 ( 2013 ) 210c-216c .
peter petreczky , chuan miao , agnes mocsy , nucl .
phys . a * 855 * , 125 ( 2011 ) .
brambilla n , escobedo m a , ghiglieri j , soto j and vairo a jhep * 09 * , 038 ( 2010 ) . m. a. escobedo , j. soto and m. mannarelli phys.rev . ,
d * 84 * , 016008 ( 2011 ) .
l. grandchamp , s. lumpkins , d. sun , h. van hees and r. rapp phys.rev .
c * 73 * , 064906 ( 2006 ) .
r. rapp , d. blaschke and p. crochet prog.part.nucl.phys .
* 65 * , 209 ( 2010 ) . f. riek and r. rapp new j. phys . * 13 * , 045007 ( 2011 ) .
a. emerick , x. zhao and r. rapp 2012 eur.phys.j .
a * 48 * , 72 ( 2012 ) .
x. zhao and r. rapp nucl.phys .
a * 859 * , 114 ( 2011 ) .
y. akamatsu and a. rothkopf phys.rev .
d * 85 * , 105011 ( 2012 )
. m. strickland , and d. bazow , nucl.phys .
a * 879 * , 25 ( 2012 ) .
a. rothkopf , t. hatsuda and s. sasaki , pos lat * 2009 * , 162 ( 2009 ) ; phys .
rev . lett.*108 * , 162001 ( 2012 ) .
y. burnier and a. rothkopf , phys .
d * 86 * , 051503 ( 2012 ) .
b. beinlich , f. karsch , e. laermann , and a. peikert , eur .
j. c * 6 * , 133 ( 1999 ) .
f. karsch , e. laermann , and a. peikert , phys .
b * 478 * , 447 ( 2000 ) .
y. aoki , g. endrodi , z. fodor , s.d .
katz , k.k .
szabo nature * 443 * , 675 ( 2006 ) .
f. karsch , j phys : conference series * 46 * , 121 ( 2006 ) .
m. cheng et al . , phys rev d * 78 * , 034506 ( 2008 ) .
y. maezawa et al . , phys rev d * 75 * , 074501 ( 2007 ) .
oleg andreev , valentine i. zakharov , phys lett b * 645 * , 437 ( 2007 ) .
vijai v. dixit , mod .
a * 5 * , 227 ( 1990 ) .
helmut satz , j. phys .
g * 32*,r * 25 * , ( 2006 ) .
v. agotiya , v. chandra and b. k. patra , phys .
c * 80 * , 025210 ( 2009 ) .
v. agotiya , v. chandra and b. k. patra , arxiv : nucl - th/0910.0586 .
p. romatshke and m. strikland , phys .
rev.d * 68 * , 036004 ( 2003 ) .
a. dumitru , y. guo and m. strickland phys .
b * 662 * , 37 ( 2008 ) .
a. dumitru , y. guo , a. mocsy and m. strickland phys.rev .
d * 79 * , 054019 ( 2009 ) .
y. burnier , m. laine and m. vepsalainen phys.lett .
b * 678 * , 86 ( 2009 ) .
a. dumitru , y. guo and m. strickland phys.rev .
d * 79 * , 114003 ( 2009 ) .
m. margotta , k. mccarty , c. mcgahan , m. strickland , d. y. elorriaga phys.rev.d * 83 * , 105019 ( 2011 ) .
l.thakur , n.haque , u.kakade , b.k.patra phys .
054022 ( 2013 ) . m. a. escobedo , j. soto phys .
a * 78 * , 032520 ( 2008 ) .
m. laine , nucl .
a * 820 * , 25c ( 2009 ) .
a. dumitru , y. guo , and m. strickland , phys.rev .
d * 79 * , 114003 ( 2009 ) .
s. digal , o. kaczmarek , f. karsch , h. satz eur .
j. c * 43 * , 71 ( 2005 ) .
e. megias , e. ruiz arriola and l. l. salcedo , indian journal phys . * 85 * , 1191 ( 2011 ) .
e. megias , e. ruiz arriola and l. l. salcedo , phys .
d * 75 * , 105019 ( 2007 )
. m. h. thoma arxiv : hep - ph/0010164 .
e. braaten , r. d. pisarski , nucl .
b * 337 * , 569 ( 1990 ) .
kelly , q. liu , c. lucchesi , and c. manuel , phys .
d * 50 * , 4209 ( 1994 ) .
blaizot and e. iancu , nucl .
b * 417 * , 608 1994 .
m. e. carrington , h. defu and m. h. thoma , eur .
j. c * 7 * , 347 ( 1999 ) ; phys .
d * 58 * , 085025 ( 1998 ) .
m. martinez and m. strickland phys.rev.c * 79 * , 044903 ( 2009 ) .
a. schneider prd * 66 * , 036003 ( 2002 ) .
h. a. weldon , phys .
d * 26 * , 1394 ( 1982 ) .
j. i. kapusta and c. gale , _ finite temperature field theory principle and applications _
( cambridge university press , cambridge , 1996 ) , 2nd ed .
blaizot and e. iancu , phys . rept . * 359 * , 355 ( 2002 ) . k. kajantie , m. laine , j. peisa , a. rajantie , k. rummukainen and m. e. shaposhnikov , phys . rev .
lett . * 79 * , 3130 ( 1997 ) .
p. petreczky , eur .
j. c * 43 * , 51 ( 2005 ) .
a. dumitru , prog .
. suppl . * 187 * , 87 ( 2011 ) .
m. martinez , and m. strickland , nucl.phys .
a * 856 * , 68 ( 2011 ) .
r. ryblewski , and w. florkowski , j.phys.g * 38 * , 015104 ( 2011 ) .
o. philipsen and m. tassler , arxiv:0908.1746 [ hep - ph ] .
a. dumitru , y. guo , a. mocsy and m. strickland , phys .
d * 79 * , 054019 ( 2009 ) . | we have studied the quasi - free dissociation of quarkonia through a complex potential which is obtained by correcting both the perturbative and nonperturbative terms of the @xmath0 potential at t=0 through the dielectric function in real - time formalism .
the presence of confining nonperturbative term even above the transition temperature makes the real - part of the potential more stronger and thus makes the quarkonia more bound and also enhances the ( magnitude ) imaginary - part which , in turn contributes more to the thermal width , compared to the medium - contribution of the perturbative term alone .
these cumulative observations result the quarkonia to dissociate at higher temperatures .
finally we extend our calculation to a medium , exhibiting local momentum anisotropy , by calculating the leading anisotropic corrections to the propagators in keldysh representation .
the presence of anisotropy makes the real - part of the potential stronger but the imaginary - part is weakened slightly .
however , since the medium corrections to the imaginary - part is a small perturbation to the vacuum part , overall the anisotropy makes the dissociation temperatures higher , compared to isotropic medium . 0.1 in 0.1 in lata thakur , uttam kakade and binoy krishna patra + 0.01 in pacs : 12.39.-x,11.10.st,12.38.mh,12.39.pn : quantum chromodynamics , debye mass , momentum anisotropy , string tension , dielectric permittivity , quark - gluon plasma , heavy quark potential , decay width . |
we thank a. ashtekar for making us aware of ref . @xcite .
we acknowledge the compagnia di san paolo for partial support.e.v.moreva acknowledges the support from the dynasty foundation and russian foundation for basic research ( project 13 - 02 - 01170-d ) a. augustinus hipponensis , _ confessiones _ , xi , 14 ( 400 ) .
kuchar , in `` proceedings of the 4th canadian conference on general relativity and relativistic astrophysics '' ed .
g. kunstatter , d. vincent and j. williams ( world scientific , singapore 1992 ) . c.j .
isham , in `` integrable systems , quantum groups and quantum field theories '' ed .
ibort and m.a .
rodrguez ( kluwer , dordrecht 1993 ) , gr - qc/9210011 .
a. ashtekar , new j. phys .
* 7 * , 198 ( 2005 ) .
e. anderson , `` the problem of time in quantum gravity '' , in `` classical and quantum gravity : theory , analysis and applications '' , ed .
frignanni ( nova , new york 2012 ) , arxiv:1009.2157v3 .
z. merali , nature * 500 * , 516 - 519 ( 2013 ) .
dewitt , phys .
160 * , 1113 - 1148 ( 1967 ) .
j. b. hartle , s. w. hawking , phys .
d * 28 * , 2960 - 2975 ( 1983 ) .
d.n . page and w.k .
wootters , phys .
d , 27 , 2885 ( 1983 ) .
wootters , int .
23 , 701 ( 1984 ) .
page , `` clock time and entropy '' , in _ physical origins of time asymmetry _ , eds . j.j .
halliwell , et al .
, ( cambridge univ . press , 1993 ) , arxiv : gr - qc/9303020 .
r. gambini , r.a .
porto , j. pullin , and s. torterolo , phys .
d * 79 * , 041501(r ) ( 2009 ) .
a. peres , am . j. phys .
* 48 * , 552 ( 1980 ) . c. rovelli , int .
j. of theor .
phys . * 35 * , 1637 ( 1996 ) , arxiv : quant - ph/9609002v2 .
unruh , r.m .
wald , phys .
d * 40 * , 2598 ( 1989 ) .
r.gambini , l.p.g .
pintos , j. pullin , studies his .
phys . * 42 * , 256 - 263 ( 2011 ) . c. rovelli , phys .
d * 43 * , 442 ( 1991 ) .
a. einstein , ann .
* 17 * , 891 , ( 1905 ) , translated in _ the principle of relativity _
( methuen & co. , london , 1923 ) . m. o. scully , b .-
englert , h. , walther , nature * 351 * , 111 ( 1991 ) .
drr s. , nonn t. , rempe g. , nature * 395 * , 33 ( 1998 ) .
j. preskill , lecture notes at http://www.theory.caltech.edu/people/preskill/ph229 , pg .
r. gambini , l.p.g .
pintos , j. pullin , found .
phys . * 40 * , 93 - 115 ( 2010 ) .
egusquiza , l.j .
garay , j.m .
raya , phys .
rev . a * 59 * , 3236 ( 1999 ) .
bartlett , t. rudolph , w. spekkens , rev .
mod . phys . * 79 * , 555 ( 2007 ) .
wiseman , j. opt .
b : quantum semiclass . opt .
* 6 , * s849 ( 2004 ) .
m.genovese , phys .
rep . * 413 * , 319 ( 2005 ) .
d. f. v. james , p. g. kwiat , w. j. munro , and a. g. white , phys .
a , 64 , 052312 ( 2001 ) .
yu.i.bogdanov , g.brida , m. genovese , a.p .
shurupov , e.v .
moreva , s.p .
kulik , phys .
* 105 * , 010404 ( 2010 ) .
i. bogdanov , e. v. moreva , g. a. maslennikov , r. f. galeev , s. s. straupe and s. p. kulik , phys .
rev . a 73 , 063810 ( 2006 ) . | in the last years several theoretical papers discussed if time can be an emergent propertiy deriving from quantum correlations . here , to provide an insight into how this phenomenon can occur , we present an experiment that illustrates page and wootters mechanism of `` static '' time , and gambini et al .
subsequent refinements . a static , entangled state between a clock system and the rest of the universe is perceived as evolving by internal observers that test the correlations between the two subsystems .
we implement this mechanism using an entangled state of the polarization of two photons , one of which is used as a clock to gauge the evolution of the second : an `` internal '' observer that becomes correlated with the clock photon sees the other system evolve , while an `` external '' observer that only observes global properties of the two photons can prove it is static .
.45 the `` problem of time '' @xcite in essence stems from the fact that a canonical quantization of general relativity yields the wheeler - de witt equation @xcite predicting a static state of the universe , contrary to obvious everyday evidence .
a solution was proposed by page and wootters @xcite : thanks to quantum entanglement , a static system may describe an evolving `` universe '' from the point of view of the internal observers .
energy - entanglement between a `` clock '' system and the rest of the universe can yield a stationary state for an ( hypothetical ) external observer that is able to test the entanglement vs. abstract coordinate time .
the same state will be , instead , evolving for internal observers that test the correlations between the clock and the rest @xcite .
thus , time would be an emergent property of subsystems of the universe deriving from their entangled nature : an extremely elegant but controversial idea @xcite .
here we want to demystify it by showing experimentally that it can be naturally embedded into ( small ) subsystems of the universe , where page and wootters mechanism ( and gambini et al .
subsequent refinements @xcite ) can be easily studied .
we show how a static , entangled state of two photons can be seen as evolving by an observer that uses one of the two photons as a clock to gauge the time - evolution of the other photon .
however , an external observer can show that the global entangled state does not evolve .
even though it revolutionizes our ideas on time , page and wootters ( paw ) mechanism is quite simple @xcite : they provide a static entangled state @xmath0 whose subsystems evolve according to the schrdinger equation for an observer that uses one of the subsystems as a clock system @xmath1 to gauge the time evolution of the rest @xmath2 . while the division into subsystems is largely arbitrary , the paw model assumes the possibility of neglecting interaction among them writing the hamiltonian of the global system as @xmath3 , where @xmath4 are the local terms associated with @xmath1 and @xmath2 , respectively @xcite . in this framework
the state of the `` universe '' @xmath0 is then identified by enforcing the wheeler - de witt equation @xmath5 , i.e. by requiring @xmath0 to be an eigenstate of @xmath6 for the zero eigenvalue .
the rational of this choice follows from the observation that by projecting @xmath0 on the states @xmath7 of the clock , one gets the vectors and @xmath8 are the unitary time evolution operators of the clock @xmath1 and of the rest of universe @xmath2 respectively .
@xmath0 is the global state of the system which is assumed to be eigenstate with null eigenvalue of the global hamiltonian @xmath9 ( see text).,scaledwidth=50.0% ] @xmath10 that describe a proper evolution of the subsystem @xmath2 under the action of its local hamiltonian @xmath11 , the initial state being @xmath12 ( see fig . [
f : figu0 ] ) . therefore , despite the fact that globally the system appears to be static , its components exhibits correlations that mimics the presence of a dynamical evolution @xcite . two main flaws of the paw mechanisms have been pointed out @xcite .
the first is based on the ( reasonable ) skepticism to accept that quantum mechanics may describe a system as large as the universe , together with its internal observers @xcite .
the second has a more practical character and is based on the observation that in the paw model the calculations of transition probabilities and of propagators appears to be problematic @xcite .
an attempt to fix the latter issue has been discussed by gambini et al .
( gppt ) @xcite by extending a proposal by page @xcite and invoking the notion of ` evolving constants ' of rovelli @xcite ( a brief overview of this approach is given in the appendix ) . in this work we present an experiment which allows reproducing the basic features of the paw and gppt models .
in particular the paw model is realized by identifying @xmath13 with an entangled state of the vertical @xmath14 and horizontal @xmath15 polarization degree of freedom of two photons in two spatial modes @xmath16 , i.e. ( see following section ) @xmath17 and enforcing the wheeler - de witt equation by taking @xmath18 as local hamiltonians of the system ( @xmath19 being a parameter which defines the time scale of the model ) . for this
purpose rotations of the polarization of the two photons are induced by forcing them to travel through identical birefringent plates as shown in fig . [
f : schema ] .
this allows us to consider a setting where everything can be decoupled from the `` flow of time '' , i.e. when the photons are traveling outside the plates .
nonetheless , the clock photon is a true ( albeit extremely simple ) clock : its polarization rotation is proportional to the time it spends crossing the plates .
although extremely simple , our model captures the two , seemingly contradictory , properties of the paw mechanism : the evolution of the subsystems relative to each other , and the staticity of the global system .
this is achieved by running the experiment in two different modes ( see fig . [
f : schema]a ) : ( 1 ) an `` observer '' mode , where the experimenter uses the readings of the clock photon to gauge the evolution of the other : by measuring the clock photon polarization he becomes correlated with the subsystems and can determine their evolution .
this mode describes the conventional observers in the paw mechanism : they are , themselves , subsystems of the universe and become entangled with the clock systems so that they see an evolving universe ; ( 2 ) a `` super - observer '' mode , where he carefully avoids measuring the properties of the subsystems of the entangled state , but only global properties : he can then determine that the global system is static .
this mode describes what an ( hypothetical ) observer external to the universe would see by measuring global properties of the state @xmath13 : such an observer has access to abstract coordinate time ( namely , in our experimental implementation he can measure the thickness of the plates ) and he can prove that the global state is static , as it will not evolve even when the thickness of the plates is varied . and pbs@xmath20 respectively .
the blue boxes ( a ) represent different thicknesses of birefringent plates which evolve the photons by rotating their polarization : different thicknesses represent different time evolutions .
the paw mechanism ( a ) is completely independent of the thickness , whereas the gppt mechanism ( b ) allows it to be measured by the experimenter only through the clock photon ( the abstract coordinate time is unaccessible and averaged away ) ; the dashed box ( b ) represents a ( known ) phase delay of the clock photon only ; pbs stands for polarizing beam splitter in the @xmath21 basis ; bs for beam splitter.,scaledwidth=50.0% ] in observer mode ( fig .
[ f : schema]a , pink box ) the clock is the polarization of a photon .
it is an extremely simple clock : it has a dial with only two values , either @xmath22 ( detector 1 clicked ) corresponding to time @xmath23 , or @xmath24 ( detector 2 clicked ) corresponding to time @xmath25 .
[ here @xmath26 , where @xmath19 is the polarization rotation rate of the quartz plate , since the polarization is flipped in this time interval . ]
the experimenter also measures the polarization of the first photon with detectors 3 and 4 .
this last measurement can be expressed as a function of time ( he has access to time only through the clock photon ) by considering the correlations between the results from the two photons : the time - dependent probability that the first photon is vertically polarized ( i.e. that detector 3 fires ) is @xmath27 and @xmath28 , where @xmath29 is the conditional probability that detector 3 fired , conditioned on detector @xmath30 firing ( experimental results are presented in fig .
[ f : results]a ) .
this type of conditioning is typical of every time - dependent measurement : experimenters always condition their results on the value they read on the lab s clock ( the second photon in this case ) .
the experimenter has access only to physical clocks , not to abstract coordinate time @xcite . in our experiment
this restriction is implemented by employing a different phase plate a ( of random thickness unknown to the experimenter ) in every experimental run . in super - observer mode
( fig .
[ f : schema]a , yellow box ) the experimenter takes the place of a hypothetical observer external to the universe that has access to the abstract coordinate time and tests whether the global state of the universe has any dependence on it .
hence , he must perform a quantum interference experiment that tests the coherence between the different histories ( wavefunction branches ) corresponding to the different measurement outcomes of the internal observers , represented by the which - way information after the polarizing beam splitter pbs@xmath31 . in our setup , this interference is implemented by the beam splitter bs of fig .
[ f : schema]b .
it is basically a quantum erasure experiment @xcite that coherently `` erases '' the results of the time measurements of the internal observer : conditioned on the photon exiting from the right port of the beam splitter , the information on its input port ( i.e. the outcome of the time measurement ) is coherently erased @xcite .
the erasure of the time measurement by the internal observers is necessary to avoid that the external observer ( super - observer ) himself becomes correlated with the clock .
however , the super - observer has access to abstract coordinate time : he knows the thickness of the blue plates , which is precluded to the internal observers , and he can test whether the global state evolves ( experimental results are presented in fig .
[ f : results]b ) . in addition
, we also test the gppt mechanism showing that our experiment can also account for two - time measurements ( see fig .
[ f : schema]b ) .
these are implemented by the two polarizing beam splitter pbs@xmath31 and pbs@xmath20 .
pbs@xmath31 represents the initial time measurement that determines when the experiment starts : it is a non - demolition measurement obtained by coupling the photon polarization to its propagation direction , while the initialization of the system state is here implemented through the entanglement . pbs@xmath20 together with detectors 1 and 2 represents the final time measurement by determining the final polarization of the photon . between these two time measurements both the system and the clock
evolve freely ( the evolution is implemented by the birefringent plates a ) . in the gppt mechanism , the abstract coordinate time ( the thickness of the quartz plates a ) is unaccessible and must be averaged over @xcite .
this restriction is implemented in the experiment by avoiding to take into account the thickness of the blue quartz plates a when extracting the conditional probabilities from the coincidence rates : the rates obtained with different plate thickness are all averaged together .
the formal mapping of the gppt mechanism to our experiment is detailed in the appendix . as before ,
the time dependent probability of finding the system photon vertically polarized is @xmath27 and @xmath28 .
however , a clock that returns only two possible values ( @xmath32 and @xmath33 ) is not very useful . to obtain a more interesting clock , the experimenter performs the same conditional probability measurement introducing varying time delays to the clock photon , implemented through quartz plates of variable thickness ( dashed box b in fig .
[ f : schema]b ) .
[ even though he has no access to abstract coordinate time , he can have access to systems that implement known time delays , that he can calibrate separately .
] now , he obtains a sequence of time - dependent values for the conditional probability : @xmath34 and @xmath35 , where @xmath36 is the time delay of the clock photon obtained by inserting the quartz plate b with thickness @xmath37 in the clock photon path .
the experimental results are presented in fig .
[ f : resultsgp ] , where each colour represents a different delay : the yellow points refer to @xmath38 ; the red points to @xmath39 , etc .
they are in good agreement with the theory ( dashed line ) derived in the appendix .
the reduction in visibility of the sinusoidal time dependence of the probability is caused by the decoherence effect due to the use of a low - resolution clock ( our clock outputs only two possible values ) , a well known effect @xcite .
): circles and squares represent @xmath27 and @xmath28 respectively , namely the probabilities of measuring @xmath14 on the subsystem 1 as a function of the clock time @xmath32 , @xmath33 ; circles and triangles represent @xmath40 and @xmath41 , the probabilities of measuring @xmath15 on the subsystem 1 as a function of the clock time . as expected from the paw mechanism , these probabilities are independent of the abstract coordinate time @xmath42 , represented by different phase plate a thicknesses ( here we used a 957@xmath43 m thick quartz plate rotated by 15 different equiseparated angles ) .
the inset shows the graph that the observer himself would plot as a function of clock - time : circles representing the probabilities of finding the system photon @xmath14 at the two times @xmath32 , @xmath33 , the triangles of finding it @xmath15 .
( b ) super - observer mode : plot of the conditional fidelity between the tomographic reconstructed state and the theoretical initial state @xmath13 of eq . as a function of the abstract coordinate time @xmath42 .
the fidelity @xmath44 ( which measures the overlap between the theoretical initial state @xmath13 and the final state @xmath45 after its evolution through the plates ) is conditioned on the clock photon exiting the right port of the beam splitter bs .
the fact that the fidelity is constant and close to one ( up to experimental imperfections ) proves that the global entangled state is static .
, scaledwidth=50.0% ] that the upper photon is @xmath14 ( namely that detector 3 clicked ) as a function of the time @xmath46 recovered from the lower photon .
the points with matching colors represent @xmath47 and @xmath48 : yellow , red , blue , etc .
, for @xmath49 , respectively . here nine different values of @xmath50 are obtained from a 1752@xmath43 m thick quartz plate rotated by nine different angles from the vertical ( 14,16,18,20,21.5,23,25,27,29 degrees ) .
the dashed line is the theoretical value .
its reduced visibility is an expected effect of the use of imperfect clocks @xcite.,scaledwidth=40.0% ] in summary , by running our experiment in two different modes ( `` observer '' and `` super - observer '' mode ) we have experimentally shown how the same energy - entangled hamiltonian eigenstate can be perceived as evolving by the internal observers that test the correlations between a clock subsystem and the rest ( also when considering two - time measurements ) , whereas it is static for the super - observer that tests its global properties .
our experiment is a practical implementation of the paw and gppt mechanisms but , obviously , it can not discriminate between these and other proposed solutions for the problem of time @xcite . in closing
, we note that the time - dependent graphs of fig .
[ f : resultsgp ] have been obtained without any reference to an external time ( or phase ) reference , but only from measurements of correlations between the clock photon and the rest : they are an implementation of a ` relational ' measurement of a physical quantity ( time ) relative to an internal quantum reference frame @xcite . the experimental setup ( fig .
[ f : exp ] ) consists of two blocks : `` preparation '' and `` measurement '' .
the preparation block produces a family of biphoton polarization entangled states of the form : @xmath51 by exploiting the standard method of coherently superimposing the emission of two type i crystals whose optical axes are rotated of 90@xmath52 @xcite .
the measurement block can be mounted in different configurations corresponding to `` observer '' and `` super - observer '' ones of paw and gppt scheme ( fig.1 ) . in general ,
each arm of the measurement block contains interference filters ( if ) with central wavelength @xmath53 nm ( fwhm @xmath54 nm ) and a polarizing beam splitter ( pbs ) . before the pbs the polarization of both photons evolves in the birefringent quartz plates a ( blue boxes in fig .
[ f : schema ] ) as @xmath55 , where @xmath56 is the material s optical thickness .
* `` observer '' mode in paw scheme * ( fig .
[ f : schema ] , block a ) : in this mode , the polarization of the photon in the lower arm is used as a clock : the first polarizing beam splitter pbs@xmath31 acts as a non - demolition measurement in the @xmath21 basis of the polarization of the second photon , finally detected by single - photon avalanche diodes ( spad ) 1 , 2 . in this mode , the experimenter has no access to an external clock , he can only use the correlations ( coincidences ) between detectors : the time - dependent probability of finding the first photon in @xmath57 is obtained from the coincidence rate between detectors 1 - 3 ( corresponding to a measurement at time @xmath32 ) , or 2 - 3 ( corresponding to a measurement at time @xmath33 ) : appropriately normalized , these coincidence rates yield the conditional probabilities @xmath29 .
the impossibility to directly access abstract coordinate time ( the thickness of the plates ) is implemented by averaging the coincidence rates obtained for all possible thicknesses of the birefringent plates a : the plate thickness does not enter into the data processing in any way . + * `` super - observer '' mode in paw scheme * ( fig.1b ) : this mode is employed to prove that the global state is static with respect to abstract coordinate time , represented by the thickness of the quartz plates a. the @xmath58 beam splitter ( bs ) in block b performs a quantum erasure of the polarization measurement ( performed by the polarizing beam splitter pbs@xmath31 ) conditioned on the photon exiting its right port . for temporal stability , the interferometer is placed into a closed box .
the output state is reconstructed using ququart state tomography @xcite ( the two - photon polarization state lives in a four - dimensional hilbert space ) , where the projective measurements are realized with polarization filters consisting of a sequence of quarter- and half - wave plates and a polarization prism which transmits vertical polarization ( fig.4 ) .
the fidelity between the tomographically reconstructed state and the theoretical state @xmath13 is reported in fig .
[ f : results]b .
* gppt two - time scheme * here a second pbs preceding detectors allows a two - time measurement . to obtain a more interesting time dependence than the probability at only two times ,
we delay the clock photon with an additional birefringent plate b ( dashed box in fig .
[ f : schema ] ) , a 1752@xmath43m - thick quartz plate rotated at nine different angles , placed in the lower arm , and we repeat the same procedure described above for different thicknesses of the plate b. this represents an internal observer that introduces a ( known ) time delay to his clock measurements .
the results are shown in fig .
[ f : results ] . )
crystals ( placed into a temperature - stabilized closed box t ) pumped by a 700 mw ar laser , later eliminated by a filter ( uwf ) .
the basic state amplitudes are controlled by a thompson prism ( v ) , oriented vertically , and a half - wave plate @xmath59/2 at angle @xmath60 .
two 1 mm quartz plates , that can be rotated along the optical axis , introduce a phase shift between horizontally and vertically polarized photons .
the beam splitter ( bs ) is used to split the initial ( collinear ) biphoton field into distinct spatial modes .
it prepares the singlet bell state @xmath61 of eq .
( 1 ) ( using @xmath62 ; @xmath63 , and an additional half - wave plate @xmath59/2 at @xmath64 in the transmitted arm ) .
* measurement block * : we implement paw or gppt as in fig.1 . in paw superobserver mode
the final state is checked by quantum state tomography @xcite , realised by registering the coincidence rate for 16 different projections achieved through half and quarter wave plates and a fixed analyzer ( v ) .
, scaledwidth=40.0% ] in this appendix we detail how our experiment implements the gambini et al .
( gppt ) proposal @xcite for extending the paw mechanism @xcite to describe multiple time measurements .
we also derive the theoretical curve of fig .
[ f : resultsgp ] .
time - dependent measurements performed in the lab typically require two time measurements : they establish the times at which the experiment starts and ends , respectively .
the paw mechanism can accommodate the description of these situations by supposing that the state of the universe will contain records of the previous time measurements @xcite .
however , this observation in itself seems insufficient to derive the two - time correlation functions ( transition probabilities and time propagators ) with their required properties , a strong criticism directed to the paw mechanism @xcite .
the gppt proposal manages to overcome this criticism .
it is composed of two main ingredients : the recourse to rovelli s ` evolving constants ' to describe observables that commute with global constraints , and the averaging over the abstract coordinate time to eliminate any dependence on it in the observables .
our experiment tests the latter aspect of the gppt theory
. measurements of a physical quantity at a given clock time , say @xmath46 , are described by the conditional probability of obtaining an outcome on the system , say @xmath65 , given that clock time - measurement produces the outcome @xmath46 .
this conditional probability is given by @xcite @xmath66}{\int dt\;{\mbox{tr}[p_{t}(t)\rho ] } } \labell{con}\;,\end{aligned}\ ] ] where @xmath67 is the global state , @xmath68 is the projector relative to a result @xmath46 for a clock measurement at coordinate time @xmath42 and @xmath69 is the projector relative to a result @xmath65 for a system measurement and @xmath46 for a clock measurement at coordinate time @xmath42 ( working in the heisenberg picture with respect to coordinate time @xmath42 ) . clearly , such expression can be readily generalized to arbitrary povm measurements .
( a similar expression , but in the schrdinger picture , already appears in @xcite . ) the integral that averages over the abstract coordinate time @xmath42 in embodies the inaccessibility of the time @xmath42 by the experimenter : he can access only the clock time @xmath46 , an outcome of measurements on the clock system . a generalization of this expression to multiple time measurements
is expressed by @xcite @xmath70}{\int dt\;\int dt'\;{\mbox{tr}[p_{t_f}(t)p_{d_i , t_i}(t')\;\rho\ ; p_{d_i , t_i}(t ' ) ] } } \nonumber\;,\end{aligned}\ ] ] which gives the conditional probability of obtaining @xmath71 on the system given that the final clock measurement returns @xmath72 and given that a `` previous '' joint measurement of the system and clock returns @xmath73 , @xmath74 .
( this expression can also be formulated as a conventional state reduction driven by the first measurement @xcite . ) in our experiment to implement the gppt mechanism ( fig .
[ f : schema]b ) we must calculate the conditional probability that the system photon is @xmath14 ( namely detector 3 clicks ) given that the clock photon is @xmath15 after the first polarizing beam splitter pbs@xmath31 ( initial time measurement ) and is @xmath15 or @xmath14 after the second polarizing beam splitter ( final time measurement ) .
the initial time measurement succeeds whenever one of photodetectors 1 or 2 click : this means that the clock photon chose the @xmath15 path at pbs@xmath31 .
( our experiment discards the events where the first time measurement at pbs@xmath31 finds @xmath14 , although in principle one could easily take into account these cases by adding a polarizing beam splitter and two photodetectors also in the @xmath14 output mode of pbs@xmath31 . )
the final time measurement is given by the click either at photodetector 1 or 2 : the clock dial shows @xmath75 and @xmath76 , respectively . using the gppt mechanism of eq .
, this means that the time dependent probability that the system photon is vertical ( detector 3 clicks ) is given by @xmath77}{\int dt\;\int dt'\;{\mbox{tr}[p_{t_f = t_k}(t)p_{d_i , t_i}(t')\rho p_{d_i , t_i}(t ' ) ] } } \nonumber\;,\end{aligned}\ ] ] where @xmath78 is the joint projector connected to detector 3 and detector @xmath79 or @xmath80 and @xmath81 is the projector connected to the first time measurement .
the latter projector is implemented in our experiment by considering only those events where either detector 1 or detector 2 clicks , this ensures that the clock photon chose the @xmath15 path at pbs@xmath31 ( namely the initial time is @xmath74 ) and that the system photon was initialized as @xmath24 at time @xmath74 .
( in principle , we could consider also a different initial time @xmath82 by employing also the events where the clock photons choose the path @xmath14 at pbs@xmath31 . ) introducing the unitary abstract - time evolution operators , @xmath83 , the numerator of eq
. becomes @xmath84=\int dt\ ; \mbox{tr}[p_{d=3,t_f = t_k}u_{t } p_{d_i , t_i } \rho p_{d_i , t_i}u^\dag_{t } ] , & & \nonumber\;\end{aligned}\ ] ] where we use the property @xmath85 and we dropped one of the two time integrals by taking advantage of the time invariance of the global state @xmath67 ( which has been also tested experimentally in the super - observer mode ) .
gambini et al .
typically suppose that the clock and the rest are in a factorized state @xcite , but this hypothesis is not strictly necessary for their theory @xcite : we drop it so that we can use the same initial global state that we used for testing the paw mechanism .
using the same procedure also to calculate the denominator of eq . , we can rewrite this equation as @xmath86}{\mbox{tr}[p_{t_f = t_k}\bar\rho ] } \;,\end{aligned}\ ] ] where @xmath87 is the time - average of the global state after the first projection , namely @xmath88 where the averaging over the abstract coordinate time @xmath42 is used to remove its dependence from the state . in our experiment
such average is implemented by introducing random values of the phase plates a ( unknown to the experimenter ) in different experimental runs . in our gppt experiment
there are two possible values for the initial projector @xmath81 : either the clock photon is projected on the @xmath15 path after pbs@xmath31 ( corresponding to an initial time @xmath74 ) or it is projected onto the @xmath14 path ( corresponding to an initial time @xmath89 ) .
we will consider only the first case , which corresponds to a click of either detector 1 or 2 : we are post - selecting only on the experiments where the initial time is @xmath74 . in this case
, the global initial state will be @xmath90 which is evolved into the vector @xmath91 [ \cos \omega t|v\>_r+\sin \omega t|h\>_r]$ ] where @xmath92 is the global hamiltonian defined in the main text and @xmath93 is the time delay introduced by the plate b of fig . [
f : schema]b . moreover , the projectors in eq .
are @xmath94 where @xmath95 and @xmath96 .
the projector @xmath78 corresponds to the joint click of detectors @xmath97 and 3 , while @xmath98 corresponds to the click of detector @xmath97 and either one of detectors 3 or 4 . in other words ,
eq . can be written as @xmath99 where @xmath100 is the joint probability of detectors @xmath101 and @xmath97 clicking .
for example , @xmath102 is the joint probability that detector 3 and 2 click , namely that both the clock and the system photon were @xmath14 . considering only the component @xmath103 of the state @xmath104 ,
this is given by @xmath105 where we have calculated the integral over @xmath42 of eq .
using a change of variables @xmath106 . proceeding analogously for all the other joint probabilities , namely replacing the projectors into
, we find the probability for detector 3 clicking ( namely the system photon being @xmath14 ) conditioned on the time @xmath72 read on the clock photon as @xmath107 which is plotted as a function of @xmath93 in fig .
[ f : results]b ( dashed line ) . since @xmath108
, we have plotted the points relative to @xmath109 as displaced by @xmath110 with respect to the points relative to @xmath111 , so that the two curves and are superimposed in fig .
[ f : results ] . |
following cosmological recombination at a redshift @xmath9 , the baryonic gas filling up the universe became predominantly neutral . given that this gas is known to be mostly ionized today
, one arrives at two of the major questions in current extragalactic astronomy : _ (
i ) when were the cosmic hydrogen and helium re - ionized ? _ and _ ( ii ) which sources dominated this reionization process ? _ the answers to both questions are likely to be different for hydrogen and helium ( see review by barkana & loeb 2001 ) .
recent observations provide preliminary answers to the first question .
the absorption spectra of sdss quasars at @xmath5 indicate that the neutral fraction of hydrogen increases significantly at @xmath10 ( becker et al .
2001 ; fan et al .
2002 ) , and the uv spectrum of quasars implies that helium is fully ionized only at @xmath11 ( jacobsen et al 1994 ; tytler 1995 ; davidsen et al . 1996 ; hogan et al . 1997 ; reimers et al . 1997 ; heap et al .
2000 ; kriss et al . 2001 ; smette et al .
the latter observations also indicate , through a cross - correlation between the hydrogen and helium forest of absorption lines ( kriss et al .
2001 ; smette et al .
2002 ) , that the ionization of helium at @xmath12 had significant contributions from both quasars and stars .
no analogous evidence exists for hydrogen at @xmath10 . in this paper
we make theoretical predictions for the reionization histories of hydrogen and helium .
we calculate the contribution from quasars using a model that matches all the existing observational data on the quasar luminosity function at high redshifts ( see wyithe & loeb 2002 for details ) .
we model the less constrained stellar contribution using two free parameters , which are in turn constrained by existing observational data .
since the recombination times of both hydrogen and helium are shorter than the age of the universe in overdense regions of the intergalactic medium ( igm ) , it is possible that these species experienced more than one epoch of reionization .
one of the goals of our detailed study is to identify the parameter values for which multiple reionization epochs are possible .
the reionization history has important implications for the temperature and polarization anisotropies of the cosmic microwave background ( cmb ) .
anisotropies on scales smaller than the angular size of the horizon at hydrogen reionization are suppressed by a factor @xmath13 where @xmath7 is the line - of - sight optical depth for electron scattering ( hu & white 1997 ; haiman & loeb 1997 ) , and secondary anisotropies are added ( hu 2000 and references therein ) .
reionization is the primary source of polarization anisotropies on large angular scales ( zaldarriaga & seljak 1997 ; kamionkowski et al .
1997 ; hu 2000 ) .
polarization anisotropies might be detected in the near future by map on large angular scales ( kaplinghat 2002 ) or by ground - based experiments on small angular scales .
so far , the available data on the temperature anisotropies of the cmb provides an upper limit of @xmath14 ( wang , tegmark & zaldarriaga 2002 ; bond et al .
we will use this upper limit to constrain the free parameters of our model . throughout the paper we assume
density parameters values of @xmath15 in matter , @xmath16 in baryons , @xmath17 in a cosmological constant , and a hubble constant of @xmath18 ( or equivalently @xmath19 ) . for calculations of the press - schechter ( 1974 )
mass function ( with the modification of sheth & tormen 1999 ) we assume a primordial power - spectrum with a power - law index @xmath20 and the fitting formula to the exact transfer function of cold dark matter , given by bardeen et al .
( 1986 ) . unless otherwise noted we adopt an rms amplitude of @xmath21 for mass density fluctuations in a sphere of radius @xmath22mpc
the simplest estimate of the epoch of reionization is based on the following simple considerations . given
a co - moving density of ionizing photons @xmath23 in a homogeneous but clumpy medium of comoving density @xmath24 ( where the size of the hii region is much larger than the scale length of clumpiness ) , the evolution of the volume filling factor @xmath25 of ionized regions is ( haiman & loeb 1997 ; madau et al . 1999 ; barkana & loeb 2001 ) @xmath26 where @xmath27 is the case b recombination coefficient , @xmath28 is the scale factor , @xmath29 is the comoving electron density , and @xmath30 is the clumping factor .
this equation describes statistically the transition from a fully neutral universe to a fully ionized one , and yields reionization redshifts for hydrogen of around 7 - 12 .
large uncertainties arise in both the source term and in the value of the clumping factor ( because more rapid recombinations lead to a slower evolution of @xmath31 ) . a more realistic description of reionization in a clumpy medium is provided by the model of miralda - escude et al .
( 2000 ) . in what follows ,
we draw primarily from their prescription and refer the reader to the original paper for a detailed discussion of its motivations and assumptions .
the model assumes that reionization progresses rapidly through islands of lower density prior to the overlap of individual cosmological ionized regions .
following overlap , the remaining regions of high density are gradually ionized .
it is therefore hypothesized that at any time , regions with gas below some critical overdensity @xmath32 are ionized while regions of higher density are not .
the assumption of homogeneity in equation ( [ fillfactor ] ) implies that the volume filling factor equals the mass filling factor .
therefore , within the model of miralda - escude et al .
( 2000 ) we replace @xmath33 by @xmath34 , where @xmath35 is the fraction of mass in regions with overdensity below @xmath36 , and @xmath37 is the volume weighted probability distribution for @xmath38 .
the product @xmath39 can be rewritten @xmath40 where @xmath41 is the fraction of the volume with @xmath42 .
the quantity @xmath43 is therefore the number of recombinations per atom in the igm per second .
note that the term analogous to the clumping factor is calculated from the volume weighted distribution .
the mass fraction @xmath44 ( or equivalently @xmath36 ) therefore evolves according to the equation @xmath45 this equation assumes that all ionizing photons are absorbed shortly after being emitted , so that there is no background ionizing field , and no loss of ionizing photons due to redshift .
we therefore implicitly assume that the mean free path of ionizing photons is much smaller than the hubble length .
this should be valid at redshifts not too much smaller than the overlap redshift .
the integration of equation ( [ postoverlap ] ) requires knowledge of @xmath37 .
miralda - escude et al . (
2000 ) found that a good fit to the volume weighted probability distribution for the density as seen in n - body simulations has the functional form @xmath46}\delta^{-\beta}d\delta,\ ] ] with @xmath47 and @xmath48 , 2.35 and 2.48 , and @xmath49 , 0.599 and 0.611 at @xmath50 , 3 and 4 . at @xmath51
they assume @xmath52 , which corresponds to the distribution of densities of an isothermal sphere , and solve for @xmath53 and @xmath54 by requiring the mass and volume to be normalized to unity .
we repeat this procedure to find @xmath37 at higher redshifts .
the proportionality of @xmath55 to the scale factor is expected for the growth of structure in an @xmath56 universe or at high redshift otherwise , and its amplitude should depend on the amplitude of the power - spectrum . the simulations on which the distribution in miralda - escude et al .
( 2000 ) was based assumed @xmath57 in matter , @xmath58 in a cosmological constant and @xmath59 , close to the values used in this paper .
equation ( [ postoverlap ] ) provides a good description of the evolution of the ionization fraction following the overlap of individual ionized bubbles , because the ionization fronts are exposed to the mean ionizing radiation field . however prior to overlap , the prescription is inadequate , due to the large fluctuations in the intensity of the ionizing radiation .
a more accurate model to describe the evolution of the ionized volume prior to overlap was suggested by miralda - escude et al .
( 2000 ) . in our notation
the appropriate equation is @xmath60}{dz } = \frac{1}{n^0}\frac{dn_{\gamma}}{dz } - \alpha_{\rm b}(1+z)^3r(\delta_{\rm crit})n_{\rm e}q_{\rm i}\frac{dt}{dz}.\ ] ] or @xmath61\frac{q_{\rm i}}{f_{\rm m}(\delta_{\rm crit})}.\ ] ] in this expression , @xmath31 is redefined to be the volume filling factor within which all matter at densities below @xmath62 has been ionized . within this formalism
, the epoch of overlap is precisely defined as the time when @xmath31 reaches unity .
however , we have only a single equation to describe the evolution of two independent quantities @xmath25 and @xmath63 .
the relative growth of these depends on the luminosity function and spatial distribution of the sources .
the appropriate value of @xmath62 is set by the mean free path of the ionizing photons .
more numerous sources can attain overlap for smaller values of @xmath64 .
assuming @xmath62 to be constant with redshift , we find that results do not vary much ( less than 10% in the optical depth to electron scattering ) for values of @xmath62 ranging from a few to a few tens . at high redshift , these @xmath62 correspond to mean free paths comparable to the typical separations between galaxies or quasars .
we assume @xmath65 ( which lies between the values for galaxies and quasars ) throughout the remainder of this paper .
next we describe the joint evolution of the filling factors and ionized mass fractions for hydrogen and helium using generalizations of equations ( [ postoverlap ] ) and ( [ preoverlap ] ) .
let @xmath66 , @xmath67 and @xmath68 be the filling factors of ionized hydrogen , singly ionized helium and doubly - ionized helium , and @xmath69 , @xmath70 and @xmath71 be the mass - fractions of ionized hydrogen , singly ionized helium and doubly ionized helium , within the volumes @xmath66 , @xmath67 and @xmath68 respectively . for later use
we also define is the mass - filling factor . the mass filling factors for h@xmath72 and he@xmath73 equal @xmath74 and @xmath75 respectively .
prior to the overlap of heiii regions , @xmath76 .
however following the overlap of heiii regions , @xmath77 . ]
the fractions of ionized mass in the universe @xmath78 , @xmath79 and @xmath80 .
we assign the comoving densities of hydrogen and helium ; @xmath81 and @xmath82 , respectively , and adopt the case b recombination coefficients for @xmath83 , @xmath84 and @xmath85 at a temperature of @xmath86k ( osterbrock 1974 ; barkana & loeb 2001 ) ; @xmath87 , @xmath88 and @xmath89 .
the ionizing radiation field is described by the comoving densities of ionizing photons for ionized hydrogen , singly ionized helium and doubly ionized helium , namely @xmath90 , @xmath91 and @xmath92 , respectively .
these quantities are computed through integration of the corresponding spectra in the frequency intervals @xmath93hz for hi , @xmath94hz for hei ( he@xmath95 ) , and @xmath96hz for heii ( he@xmath97 ) .
two classes of sources , quasars and stars , contribute to the ionizing radiation field .
we discuss these in turn in the following subsections . to calculate the comoving density of ionizing photons emitted by quasars , @xmath98 , we integrate over both the quasar luminosity function and over the mean quasar spectrum ( specified in schirber & bullock 2002 ) . given the b - band quasar luminosity function @xmath99 at redshift @xmath100 in units of @xmath101 , we find for example @xmath102 where @xmath103 is the planck constant , and @xmath104 is the luminosity in units of @xmath105 of a quasar with a b - band luminosity @xmath106 . at redshifts below @xmath107 we use the luminosity function of boyle et al .
( 2000 ) . at @xmath108
we use the theoretical luminosity function of wyithe & loeb ( 2002 ) ; the reader is referred to the original paper for more details .
this quasar evolution model successfully describes all known properties of the high redshift quasar luminosity function ( fan et al .
2001a , b ) , and reproduces measurements of the black - hole / dark matter halo mass relation ( ferrarese 2002 ) as well as estimates of the quasar duty cycle ( steidel et al . 2002 ) . since quasars at high redshift appear similar to typical examples at low redshift ( fan et al .
2001b ) , the luminosity function fully determines the ionizing radiation field .
we therefore have almost no freedom in our calculation of the ionizing radiation field from quasars at high redshifts .
this allows us to explore the effect of different stellar radiation fields on the reionization history of the universe . to compute the comoving densities of stellar ionizing photons , @xmath109
, we integrate over model spectra for stellar populations in the frequency ranges quoted above to find ( for each species ) the total number of ionizing photons , @xmath110 , emitted per baryon incorporated into stars .
we find @xmath111 where @xmath112 is the local number density of baryons , @xmath113 is the escape fraction for the ionizing radiation , @xmath114 is the fraction of baryons within a galaxy that are incorporated in stars , and @xmath115 is the fraction of baryons in the universe at redshift @xmath100 that collapsed and cooled inside galaxies .
to compute @xmath115 we require the dark matter collapse fraction in halos @xmath116 with virial temperatures larger than @xmath117 , given by @xmath118 where @xmath119 is the co - moving mass density in the universe , and @xmath120 is the press - schechter ( 1974 ) mass function for dark matter halos . in regions of the universe
that are neutral , the critical virial temperature ( the temperature at which atomic cooling becomes efficient ) is @xmath121k ( a circular velocity of @xmath122km@xmath123s@xmath124 ) . in reionized regions ,
infall of gas is suppressed below a virial temperature of @xmath125k ( @xmath126km@xmath123s@xmath124 , thoul & weinberg 1996 ) . given @xmath127 and
@xmath128 we can take the baryonic collapse fraction @xmath129 this prescription for calculating the stellar ionizing field will be referred to hereafter as case a. note that @xmath130 could be different for hydrogen and helium ; however , we adopt a single value for it assuming that this value is dictated by the geometry of the optically - thick gas that allows limited escape routes which are common for all ionizing photons ( wood & loeb 2000 and references therein ) .
we also consider a second evolution for the stellar ionizing field , denoted hereafter as case b. a recent study ( kauffmann et al .
2002 ) has shown that in a large sample of local galaxies , the ratio @xmath131 ( where @xmath132 and @xmath133 are the total stellar and dark matter halo masses respectively ) scales as @xmath134 for @xmath135 , but is constant for larger stellar masses .
note that the star - formation efficiency is proportional to @xmath136 .
since star - formation is thought to be regulated by supernova feedback ( dekel & silk 1986 ) , the important quantity is the depth of the galactic potential well , or equivalently the halo circular velocity . using the stellar mass tully - fisher relation of bell & de jong ( 2001 )
, we find the threshold circular velocity @xmath137 that at @xmath138 corresponds to a stellar mass of @xmath139 . in this case
we define @xmath140 as the star - formation efficiency in galaxies with circular velocities larger than @xmath141 , and calculate the effective product of the dark matter collapse fraction and star - formation efficiency @xmath142 where @xmath143 for @xmath144 and @xmath145 for @xmath146 .
the value of @xmath147 is calculated from @xmath148^{-1/2}\left(\frac{1+z}{10}\right)^{-3/2}m_{\odot},\ ] ] where @xmath149 , @xmath150 and @xmath151 $ ] .
finally , the value of @xmath152 required for calculation of @xmath153 may then be computed in analogy with case a , and substituted into equation ( [ star_ionize ] ) . in both cases
a and b we make the distinction between the ionizing radiation field due to a possible early population of zero - metallicity stars , and the metal enriched stars observed at lower redshifts .
it is thought that the primordial initial mass function favored massive stars ( bromm , copi & larson 1999 , 2001 ; abel , bryan & norman 2000 ; mackey , bromm , hernquist 2002 ) .
the possible existence of this population is very important for reionization because the spectrum of these stars would result in an order of magnitude more ionizing photons per baryon incorporated into stars ( bromm , kudritzki , & loeb 2001 ) .
the formation of the very massive stars was suppressed as the material out of which stars form was enriched with metals .
the fraction of the ionizing photons produced by metal - free stars depends on several unknown parameters , including the mixing efficiency of metals , the environments in which new stars form , and most importantly , the threshold metallicity above which star formation is transformed from being dominated by massive stars to a scalo ( 1998 ) initial mass function ( imf ) .
the threshold metallicity is believed to be small ; bromm et al .
( 2001 ) argue for a threshold @xmath154 of the solar metallicity value .
the efficiency of mixing of metal enriched outflows from star forming galaxies to the surrounding igm is even more uncertain ; scannapieco , ferrara & madau ( 2002 ) find that the mass weighted mean metallicity can reach values greater than @xmath155 of the solar value at redshifts as high as 20 , and note that the average metallicity scales with star formation efficiency , supernovae rate , and the fraction of supernovae energy that is channeled into outflows .
the average metallicity increases roughly exponentially with redshift as it is modulated by the exponential growth in the collapse fraction of baryons at high redshifts .
since the first metals were produced by supernovae , it is reasonable to suppose that the enrichment of the igm with metals at redshift @xmath100 is proportional to the mass of baryons that has formed stars by that redshift .
the average metallicity at a redshift @xmath100 can then be written as @xmath156 where @xmath157 is a constant .
note that the baryonic collapse fraction @xmath158 , which is computed from equation ( [ fb ] ) is affected by the reionization history .
the formation of metal - rich stars is assumed to be proportional to the average metalicity , with the fraction reaching unity when the average metalicity of the igm reaches @xmath159 . before @xmath160 reaches @xmath159 we may approximate the fraction of star formation at redshift @xmath100 that is in zero - metallicity stars to be @xmath161 , while after @xmath160 reaches @xmath159 we may assume all star formation to have a scalo ( 1998 ) imf .
for a given star formation efficiency and escape fraction , the above prescription yields a characteristic redshift for the build - up of the threshold metal enrichment .
the exponential growth in the collapse fraction with redshift implies that the evolution in the mode of star formation resembles a step - function .
we therefore define a transition redshift @xmath6 below which metal - rich stars with a scalo ( 1998 ) imf dominate the production rate of ionizing photons . through most of the paper we present results in terms of @xmath0 , but
return to a justification of this choice in
[ justification ] . at redshifts above @xmath6 ,
we use values of @xmath110 calculated for each of the three species @xmath83 , @xmath162 and @xmath163 assuming massive ( @xmath164 ) zero - metallicity stars and the generic spectrum calculated by bromm , kudritzki & loeb ( 2001 ) .
the resulting values of @xmath165 are @xmath166 for h@xmath72 , @xmath167 for @xmath162 and @xmath168 for @xmath163 . at lower redshifts we assume metal enriched stars ( 1/20th solar metallicity ) with a scalo ( 1998 ) mass - function , and
use spectral information from the stellar population model of leitherer et al .
this results in values of @xmath169 for h@xmath72 , @xmath170 for @xmath162 and @xmath171 for @xmath163 .
the typical distance that an ionizing photon of neutral helium can travel through neutral hydrogen having a cross - section @xmath172 before the optical depth reaches unity is the absorption mean - free - path @xmath173 , where @xmath174 we find @xmath175^{-3}$]kpc .
this distance is much shorter than @xmath176mpc which is the typical size of an hii region around a @xmath177 halo prior to reionization ( barkana & loeb 2000 ) .
therefore , at @xmath178 the @xmath162 propagation front can only lead the @xmath83 propagation front by a small fraction of the size of the hii region , and the excess @xmath179 ionizing photons photo - ionize additional hydrogen . on the other hand , most ionizing photons for hydrogen ( @xmath180hz )
are not absorbed by helium , and so the expansion of the hii region is not inhibited by the presence of helium . obviously , due to the absence of singly ionized helium the @xmath181 front can not propagate beyond the @xmath182 front . following these considerations ,
we impose two restrictions on the evolution of @xmath67 and @xmath183 .
first , @xmath184 .
second , @xmath185 .
following helium overlap , we impose the same restrictions on the evolution of the mass fraction , i.e. @xmath186 and @xmath187 .
the full equations governing the evolution of the filling factors and mass fractions are presented in the appendix .
the equations are split into three cases describing three different epochs .
the evolution of the filling factors prior to hydrogen overlap are described in case i. case ii describes the evolution of the filling factors and of the ionized hydrogen mass fraction following the overlap of hii regions , but prior to the overlap of heii regions .
finally , case iii describes the evolution of the ionized mass - fractions following the overlap of cosmological heii regions .
figure [ fig1 ] provides a schematic representation of the three epochs in reionization history .
in this section we derive reionization histories for different values of @xmath6 and @xmath188 .
these histories are shown in figures [ fig2 ] ( for case a star - formation ) and [ fig3 ] ( for case b star - formation ) .
for case a we show results for combinations of @xmath189 and 0.005 , and @xmath190 and 20.0 .
we note the following features .
large values of @xmath191 allow the massive zero metallicity stars to reionize both hydrogen and helium in the universe at very early times .
if the zero metallicity stars do not reionize helium , quasars result in the overlap of heii regions by @xmath192 , consistent with observations showing transmission just blueward of the helium ly@xmath193 line at @xmath194 ( jacobsen et al 1994 ; tytler 1995 ; davidsen et al .
1996 ; hogan et al .
1997 ; reimers et al . 1997 ; heap et al . 2000 ; kriss et al . 2001 ; smette et al .
2002 ) as well as evidence for a temperature rise in the igm at @xmath195 indicating the reionization of helium at that time ( schaye et al . 2000 ; theuns et al . 2002 ) .
interestingly , we also find that it is possible for hydrogen and/or helium to have been reionized twice .
this situation arises if the zero - metallicity stars reionize the universe prior to @xmath6 , but below @xmath0 the recombination rate is sufficiently high that the softer spectra of normal stars can not maintain overlap .
overlap of hii regions is then re - achieved at a later time as the collapse fraction grows and the density of the igm is lowered . a second overlap of heii regions results from the rise in quasar activity at @xmath196 .
similar results are found in case b , for values of @xmath191 that are a factor of @xmath197 larger than in case a. we would like to reiterate the point that multiple reionization epochs occur for some sets of parameters as a result of multiple peaks in the emissivity of ionizing photons as a function of cosmic time . in such cases the first emissivity peak occurs at @xmath6 , at which point it drops due to the softer spectra of metal enriched stars .
later , when the collapsed fraction of baryons has reached a sufficiently high value , the resulting increase in emissivity further reionizes the igm . as we will show in [ justification ] , this double peaked behavior is still seen if a smoothly varying fraction of collapsed objects at metallicities below the threshold produce the hard spectra rather than sources emitting prior to a fixed @xmath6 .
a double peaked emissivity will always arise so long as the relevant metal enrichment time is less than the time required for the collapse fraction to change by a factor equal to the ratio between the emissivities of the first generation of stars and subsequent metal enriched stars . from numerical simulations . ]
there are two small , sharp jumps in the reionization histories that are artifacts , but warrant explanation .
the first is in the evolution of @xmath198 immediately following the overlap of heiii regions .
this small jump ( note that the curve is smooth , not discontinuous ) arises because our formalism demands @xmath199 at this epoch .
however following the overlap of he@xmath73 , the he@xmath95 front is free to rapidly expand , powered by the strong ionizing radiation field , and the mass fraction @xmath200 therefore grows until the he@xmath95 front reaches the h@xmath72 front .
the second jump occurs at @xmath107 where we switch from computation of the ionizing flux due to quasars using the model quasar luminosity function , to calculation using the empirical luminosity function .
neither is a perfect description around this redshift , and there is a discontinuous jump in the luminosity density , and hence in the ionizing radiation density , resulting in a jump in the reionization histories . at early times the rise in temperature associated with reionization curtails star - formation .
the effect of this is clearly seen in panel ( a ) of figure [ fig2 ] . here
the zero metallicity stars rapidly increase the ionized fraction , but as @xmath66 approaches unity , the ionizing flux is reduced , and the rate of reionization slows .
this feature is not seen in cases where overlap occurs at lower redshift ( e.g. panel ( d ) in figure [ fig2 ] ) .
many combinations of the parameters @xmath6 and @xmath188 predict significant reionization without overlap due to the early presence of metal - free stars , resulting in a peak in the ionized fraction centered on the transition redshift , followed by recombinations before further reionization due to normal stars and quasars .
we term these peaks _ failed overlaps_. in section
[ visability ] we show that these failed overlaps leave a significant imprint on the visibility function for electron scattering of cmb photons
. different reionization histories result in different densities of electrons as a function of redshift .
one simple probe of the reionization history is the optical depth to electron scattering @xmath7 , which depend on the mass filling factors @xmath201 @xmath202(1+z)^3,\ ] ] where @xmath203 is the thomson cross - section . in the following section we plot contours of @xmath7 as a function of @xmath6 and @xmath191
in this section we discuss the constraints that current observations place on @xmath6 and @xmath188 .
first we discuss the variation of @xmath7 with @xmath0 and @xmath191 . the optical depth for electron scattering
is constrained from recent cmb anisotropy experiments to a value @xmath3 ( wang & tegmark 2002 ) . an upper limit on @xmath204 implies that there is maximum redshift beyond which the universe was neutral . as a second constraint , we note that the spectra of quasars at @xmath2 do not show a gunn - peterson ( 1965 ) trough , indicating that hii regions had achieved overlap earlier than that time . as we have seen in the previous section , the universe is reionized earlier for higher values of @xmath191 , which result in higher values of @xmath7
. hence @xmath191 can not be too large so as not to violate the limit @xmath3 .
on the other hand , if the product @xmath191 is too small , then the ionizing radiation field will be too weak for the hii regions to overlap by @xmath205 . if the transition redshift @xmath6 is large , then @xmath191 may obtain higher values since the harder spectra of metal free stars are only available to the ionizing radiation field at high redshifts when the igm is dense and the collapse fraction is small
. however if @xmath6 is low , then @xmath191 may take smaller values and still result in overlap of hii regions by @xmath205 .
in addition to the above constraints , observations suggest that the igm was metal enriched by @xmath205 ( songaila 2001 ) to a level that would lead to a scalo stellar mass function ( bromm et al .
2001 ) if new galaxies formed out of it . as a result ,
@xmath6 is limited to be larger than six .
figures [ fig4 ] and [ fig5 ] show contours of @xmath7 ( solid lines ) as a function of @xmath6 and @xmath191 for case a and b star - formation respectively . also shown are contours of @xmath66 at @xmath51 ( dashed lines ) , and a vertical line at @xmath206 ( dot - dashed line ) .
the regions excluded by the aforementioned constraints are shaded grey . for our case a star - formation model we find that @xmath207 if @xmath208 . for a higher @xmath6 , we find that larger values of @xmath188 are possible .
if @xmath209 , @xmath191 may be as low as @xmath210 , though for @xmath211 we find @xmath212 .
we find that the lower limits on @xmath188 are larger by a factor of @xmath213 for our case b star - formation model .
however since the intensity of the stellar radiation field is curtailed by feedback at high redshift in this model , we find that @xmath191 can be as large as @xmath214 for @xmath215 , or even larger for @xmath216 .
interestingly , the contours for @xmath7 and @xmath217 are not parallel .
a significant range of @xmath7 exist for different @xmath6 ( particularly where case a star - formation is assumed ) .
this indicates that measurement of optical depth does not determine the overlap redshift .
we discuss the reason for this puzzling result in [ visability ] . in [ histories ] we mentioned that hydrogen and/or helium could be reionized twice , and showed examples of reionization histories that exhibit this behavior . in figures
[ fig6 ] and [ fig7 ] we show contours of the overlap redshift for hydrogen ( left panels ) , and helium ( right panels ) , for case a and case b star - formation models .
the redshift of the most recent overlap is described by the dark contours . at high values of @xmath6 and low values of @xmath188 ,
the overlap redshift is almost independent of star - formation since then quasars dominate the ionizing radiation field .
this is particularly true for helium , which is ionized by quasars at @xmath192 in the absence of any stellar ionizing radiation .
cases where there was an additional earlier overlap are denoted by the grey contours .
the region of the @xmath218 $ ] plane that results in multiple overlap epochs is shaded grey .
we see that this area covers a significant portion of parameter space , and therefore that it is quite possible the universe was reionized twice .
the contours representing the upper limit of @xmath3 , and the lower limit for the overlap redshift @xmath219 are also shown ( thick dashed lines ) . as mentioned earlier , parameters in the region between these limits
are not excluded by current observations .
the visibility function @xmath220 describes the probability that an observed photon was scattered between conformal times @xmath221 and @xmath222 @xmath223(1+z)^2e^{-\tau_{\rm es}},\ ] ] where @xmath224 , @xmath7 is the optical depth from redshift 0 to @xmath100 and the @xmath201 are mass filling factors .
it is commonly assumed that the visibility function is peaked during the recombination era , with a second , lower amplitude , but broader peak coinciding with the epoch of overlap of hii regions .
however we have found that overlap may have occurred on multiple occasions , and that many combinations of the parameters @xmath6 and @xmath188 predict significant reionization without overlap ( _ failed overlap _ ) at high redshift due to the early presence of metal - free stars . during these failed overlaps , examples of which
are shown in figures [ fig2 ] and [ fig3 ] , the ionization fraction can be significant .
following a failed overlap at @xmath6 , recombinations dominate the evolution and the ionization fraction drops before further reionization ensues due to normal stars and quasars .
the higher cosmic density at early times might more than compensate for the smaller ionization fraction , so that the peak of the visibility function does not coincide with the epoch of reionization .
this is indeed what we find .
figures [ fig7 ] and [ fig8 ] show visibility functions corresponding to the sample reionization histories shown in figures [ fig2 ] and [ fig3 ] .
generally , if the early metal - free stars succeed in achieving reionization , then the peak in the visibility function coincides with this overlap epoch ( e.g. case a in figures [ fig2 ] and [ fig3 ] )
. however the optical depth @xmath7 is large and many of these cases are already excluded by cmb measurements . of the 8 cases shown in figures [ fig2 ] and [ fig3 ] , those that exhibit failed overlap have visibility peaks that coincide with this failed overlap , rather than with the actual reionization epoch .
_ thus , the visibility function may probe the nature of the early generation of stars rather than the reionization epoch itself .
_ the reionized gas during a failed overlap will be in bubbles .
gruzinov & hu ( 1998 ) have pointed out that a universe which is ionized in patches induces additional cmb anisotropies on arc - minute scales .
thus , patchy reionization due to a failed overlap at high redshifts will modify the shape of the small - scale power spectrum of cmb anisotropies .
however , the amplitude of the added secondary anisotropies was predicted to be at a level well below the feature in the spectrum detected more recently by the cosmic background imager ( padin et al 2001 ) .
this feature was explained by bond et al .
( 2002 ) as being due to the foreground sunyaev - zeldovich effect from x - ray clusters .
as mentioned in [ stars ] we have approximated the evolution in the mode of star formation as a step function .
that is , we have assumed that the production rate of ionizing photons in the universe is transformed sharply at @xmath6 from being dominated by massive , metal - free stars to being dominated by metal - rich stars with a scalo imf . in this section
we justify this approach by comparing its results to those obtained with a scheme involving gradual metal enrichment of the igm as described in
[ stars ] by equation ( [ gradualenrich ] ) .
figure [ fig10 ] shows two sample reionization histories computed using the gradual enrichment scheme , together with the resulting visibility function .
the examples correspond to panels a ) and c ) for case b star formation in figure [ fig3 ] .
the thick grey line in these panels shows the fraction of star formation resulting in a scalo ( 1998 ) mass function as a function of redshift . in each of these cases the value of @xmath157 ( listed in the figure )
produces around 50% scalo ( 1998 ) imf star formation at the previously used redshift @xmath6 . note the different values of @xmath225 , and the different histories of the collapsed mass - fraction .
the features discussed for the corresponding examples in figures [ fig3 ] and [ fig9 ] are still present in figure [ fig10 ] . these include the double overlap of he iii in a ) , and the non - coincidence of the overlap redshift of hii and the peak in the visibility function in b ) .
furthermore , we find unchanged values of @xmath7 .
these calculations appear to justify our use of a single parameter @xmath6 to characterize the reionization histories and visibility functions .
we have explored the reionization histories of hydrogen and helium due to stars and quasars .
the results were analyzed as a function of the two free parameters in our model , namely : ( i ) the transition redshift , @xmath0 , above which the stellar population is dominated by massive , zero metallicity stars ; and ( ii ) the product of the escape fraction of ionizing photons and the star - formation efficiency , @xmath226 .
the quasar model was not varied since it provides an excellent fit to all existing data on the luminosity function of quasars up to redshift @xmath227 ( wyithe & loeb 2002 ) .
figures [ fig2 ] and [ fig3 ] show sample reionization histories for different choices of @xmath6 and @xmath188 .
we find that a wide range of @xmath228 is allowed for @xmath229 if the star - formation efficiency does not depend on galaxy mass ( figure [ fig4 ] ) or @xmath230 if the efficiency is suppressed in low mass galaxies ( figure [ fig5 ] ) .
this wide range satisfies the constraints that overlap of the hii regions must be achieved by @xmath5 ( fan et al 2002 ) and that the optical depth for electron scattering must be limited to @xmath3 .
the allowed range leads generically to @xmath8 ( see figures [ fig4 ] and [ fig5 ] ) .
the map satellite is expected to have sufficient sensitivity to detect values of @xmath7 as small as @xmath231 ( kaplinghat et al .
2002 ) , well below this range of expected values .
a major fraction of the allowed range of @xmath191 and @xmath0 leads to an early peak in the ionized fraction due to metal - free stars at high redshifts
. often this peak results in a small but significant filling factor that is subsequently reduced ( temporarily ) due to recombination . in a restricted range of the allowed parameter values
, we find that either hydrogen or helium experience two overlap epochs , separated by recombination ( see shaded regions in figures [ fig6 ] and [ fig8 ] ) .
that helium might have been reionized twice due to the presence of early very massive stars was previously pointed out by oh et al .
the first overlap phase is caused by the population of zero - metallicity , massive stars and the second is dominated by the quasars for helium or by stars and quasars for hydrogen . even if early overlap is not achieved , the peak in the visibility function for scattering of the cmb often coincides with the early partial ionization peak rather than with the actual reionization epoch ( see figures [ fig7 ] and [ fig9 ] ) .
the resulting value of @xmath7 is therefore larger than expected based the reionization redshift alone .
thus , the cmb visibility function may be probing the nature of the early generation of stars rather than the reionization epoch itself .
future cmb experiments , such as map and planck , will provide tighter constraints on @xmath7 and will reduce the range of allowed @xmath188 in our model ( see figures [ fig4 ] and [ fig5 ] ) . any additional observational data on the composition of the stellar population or the abundance of hei or heii at @xmath232
will provide stronger lower limits on @xmath6 .
we thank rennan barkana and an anonymous referee for useful detailed comments on the manuscript .
jsbw is supported by a hubble fellowship grant from the space telescope science institute , which is operated by the association of universities for research in astronomy , inc .
, under nasa contract nas 5 - 26555 .
al acknowledges generous support from the institute for advanced study at princeton and the john simon guggenheim memorial fellowship .
this work was also supported in part by nsf grant ast-0071019 .
abel , t. , bryan , g.l . & norman , m.l .
, 2000 , apj , 540 , 39 bardeen , j.m . , bond , j.r . , kaiser , n. , & szalay , a.s .
1986 , apj , 304 , 15 barkana , r. , & loeb , a. 2001 , phys .
, 349 , 125 becker , r. h. et al .
2001 , , 122 , 2850 bell , e. f. & de jong , r. s. 2001 , , 550 , 212 bond , j.r . ,
cole , s. , efstathiou , g. , & kaiser , n. 1991 , apj , 379 , 440 bond , j. r. et al .
2002 , apj , submitted ; astro - ph/0205386 boyle , b. j. , shanks , t. , croom , s. m. , smith , r. j. , miller , l. , loaring , n. , & heymans , c. 2000 , , 317 , 1014 bromm , v. , coppi , p. s. , & larson , r. b. 1999 , , 527 , l5 bromm , v. , coppi , p. s. , & larson , r. b. 2001 , astro - ph/0102503 bromm , v. , ferrara , a. , coppi , p. s. , & larson , r. b. 2001 , , 328 , 969 bromm , v. , kudritzki , r. p. , & loeb , a. 2001 , , 552 , 464 davidsen , a. f. , kriss , g. a. , & zheng , w. 1996 , nature , 380 , 47 dekel , a. & silk , j. 1986 , , 303 , 39 fan , x. , narayanan , v. k. , strauss , m. a. , white , r. l. , becker , r. h. , pentericci , l. , & rix , h. 2002 , , 123 , 1247 fan , x. , et al .
2001a aj , 121 , 54 fan , x. , et al .
2001b aj , 122 , 2833 ferrarese , l. 2002 , astro - ph/0203469 gruzinov , a. & hu , w.1998 , , 508 , 435 gunn , j. e. & peterson , b. a. 1965 , , 142 , 1633 haiman , z. & loeb , a. , 1997 , apj , 483 , 21 heap , s. r. , williger , g. m. , smette , a. , hubeny , i. , sahu , m. s. , jenkins , e. b. , tripp , t. m. , & winkler , j. n. 2000 , , 534 , 69 hogan , c. j. , anderson , s. f. , & rugers , m. h. 1997 , aj , 113 , 1495 hu , w. 2000 , , 529 , 12 hu , w. & white , m. 1997 , , 479 , 568 jacobsen , p. et al .
1994 , nature , 370 , 35 kamionkowski , m. , kosowsky , a. , & stebbins , a. 1997 , phys .
d55 , 7368 kaplinghat , m. , chu , m. , haiman , z. , holder , g. , knox , l. , skordis , c. , 2002 , astro - ph/0207591 kriss , g. a. et al .
2001 , science , 293 , 1112 leitherer , c. et al . 1999 , ,
123 , 3 mackey , j. , bromm , v. , hernquist , l. , 2002 , astro - ph/0208447 madau , p. , haardt , f. , & rees , m. j. 1999 , , 514 , 648 miralda - escud ' e , j. , haehnelt , m. , & rees , m. j. 2000 , , 530 , 1 oh , s. p. , nollett , k. m. , madau , p. , & wasserburg , g. j. 2001 , , 562 , l1 osterborck , d. e. 1974 , astrophysics of gaseous nebulae , ( freeman : san fracisco ) , ch .
2 padin , s. et al .
2001 , , 549 , l1 press , w. h. , & schechter , p. 1974
, apj , 187 , 425 reimers , d. et al .
1997 , a & a , 327 , 890 scalo , j. 1998 , asp conf .
ser . 142 : the stellar initial mass function ( 38th herstmonceux conference ) , 201 scannapieco , e. , ferrara , a. , & madau , p. 2002
, , 574 , 590 schaye , j. , theuns , t. , rauch , m. , efstathiou , g. , & sargent , w. l. w. 2000 , , 318 , 817 schirber , m. , bullock , j.s . , 2002 , astro - ph/0207200 sheth , r. k. & tormen , g. 1999 , , 308 , 119 smette , a. , heap , s. r. , williger , g. m. , tripp , t. m. , jenkins , e. b. , & songaila , a. 2002 , , 564 , 542 songaila , a. 2001 , , 561 , l153 ; erratum : 2001 , , 568 , l139 steidel , c. , et al . , 2002 , astro - ph/0205142 theuns , t. , bernardi , m. , frieman , j. , hewett , p. , schaye , j. , sheth , r. k. , & subbarao , m. 2002 , , 574 , l111 thoul , a.a .
& weinberg , d.h . , 1996 , apj , 465 , 608 tytler , d. 1995 , in qso absorption lines , proc .
eso workshop g. meylan ed . ,
( springer : berlin ) , p. 289
wang , x. , tegmark , m. , & zaldarriaga , m. 2002 , , 65 , 123001 wood , k. & loeb , a.2000 , , 545 , 86 wyithe , j.s.b , loeb , a. , 2002 , astro - ph/0206154 zaldarriaga , m. , & seljak , u. 1997 , phys .
the filling factors @xmath66 , @xmath67 and @xmath183 , and ionized mass fractions @xmath233 , @xmath70 and @xmath234 evolve according to six coupled odes .
it is convenient to write the odes and constraints in terms of the following quantities : @xmath235 in the above definitions , the @xmath236 and @xmath237 are source terms describing the ionizing radiation field , while the @xmath238 and @xmath239 are sink terms describing the recombination rate . in the recombination terms ,
the @xmath29 are the electron densities available for recombination ( equal to the mean cosmological density multiplied by the volume filling factor and the mass fraction ) . these are therefore @xmath240\\ \nonumber n_{\rm e}^{\rm he^+ } & = & \left[n_{\rm h}^{\rm 0 } + n_{\rm he}^{\rm 0}\right]\\ \nonumber \mbox{and}\\ n_{\rm e}^{\rm he^{++ } } & = & \left[n_{\rm h}^{\rm 0 } + 2n_{\rm he}^{\rm 0}\right].\end{aligned}\ ] ] we write the generalization of equations ( [ postoverlap ] ) and ( [ preoverlap ] ) for the three species in terms of the above quantities
. there are 3 cases corresponding to the period before hii regions overlap , the period following the overlap of hii regions but preceding the overlap of heii regions , and the period following the overlap of heii regions : prior to the overlap of hii regions , the mass fractions @xmath244 and their derivatives are set by the critical value of @xmath62 chosen as the ionization density threshold .
the filling factors @xmath66 , @xmath67 and @xmath183 evolve according to a generalization of equation ( [ postoverlap ] ) . hence @xmath245 - s_{\rm k}^{\rm h^{+}}\\ \nonumber \frac{dq_{\rm he^{\rm + } } }
{ dz}&= & s_{\rm c}^{\rm he^{+ } } - \frac{1}{n_{\rm he}^{\rm 0}f_{\rm he^{+}}}\delta\frac{dn_{\rm \gamma}^{\rm he^{+}}}{dz}- s_{\rm k}^{\rm he^{+ } } -\frac{dq_{\rm he^{\rm + + } } } { dz}\\ \frac{dq_{\rm he^{\rm + + } } } { dz}&=&s_{\rm c}^{\rm he^{++ } } - \frac{1}{n_{\rm he}^{\rm 0}f_{\rm he^{++}}}\delta\frac{dn_{\rm \gamma}^{\rm he^{++}}}{dz}- s_{\rm k}^{\rm he^{++}}.\end{aligned}\ ] ] in the above , @xmath246 and @xmath247 are the excess ionizing photon rates having frequencies @xmath94hz for hei ionization , and @xmath96hz for heii ionization .
the values of these excess source terms are determined from the constraints on the relative evolution of the different ionization fronts .
as these are excess rates , their values must be positive . if @xmath248 then @xmath249 is limited to be smaller than @xmath250 since the @xmath179 front can not propagate beyond the h@xmath95 front .
therefore , if @xmath251 we require @xmath252 , and since @xmath253 in this case we find @xmath254.\end{aligned}\ ] ] in addition to the above constraint , the @xmath163 filling factor @xmath255 is limited to be less than @xmath256 . therefore , if @xmath257 we require @xmath258 so that the fronts propagate at the same rate , and @xmath259 since the he@xmath72 front can not propagate if the heii region is filling all of the available volume .
these constraints yield @xmath260\\ \nonumber & -&n_{\rm he}^{\rm 0}f_{\rm he^{+}}\frac{n_{\rm he}^{\rm 0}f_{\rm he^{++}}+n_{\rm h}^{\rm 0}f_{\rm h^{+}}}{2n_{\rm he}^{\rm 0}f_{\rm he^{++}}+n_{\rm h}^{\rm 0}f_{\rm h^{+}}}\left [ ( s_{\rm c}^{\rm he^{++}}-s_{\rm c}^{\rm he^{+ } } ) - ( s_{\rm k}^{\rm he^{++}}-s_{\rm k}^{\rm he^{+}})\right]\\ \nonumber & & \mbox{and}\\ \delta\frac{dn_{\rm \gamma}^{\rm he^{++}}}{dz}&=&n_{\rm he}^{\rm 0}f_{\rm he^{++}}\left [ ( s_{\rm c}^{\rm he^{++}}-s_{\rm c}^{\rm he^{+ } } ) - ( s_{\rm k}^{\rm he^{++}}-s_{\rm k}^{\rm he^{+ } } ) + \frac{1}{n_{\rm he}^{\rm 0}f_{\rm he^{+}}}\delta\frac{dn_{\rm \gamma}^{\rm he^{+}}}{dz}\right].\end{aligned}\ ] ] following the overlap of hii regions , the mass fraction of hydrogen @xmath69 is free to evolve according to equation ( [ preoverlap ] ) ( with modified source term to account for the excess ionizing photons ) .
the filling factors @xmath66 , @xmath67 and @xmath183 again evolve according to a generalization of equation ( [ postoverlap ] ) . note that if @xmath265 , substitution of equation ( [ postoverlap ] ) into equation ( [ preoverlap ] ) yields @xmath266 . thus , post hii overlap the value of @xmath267 remains at unity .
however by continuing to follow all the equations we allow for the possibility of a recombination epoch following an early reionization .
we have @xmath268 as before , the values of the excess ionizing photon rates are determined from constraints on the co - evolution of the ionization fronts . in the case
where @xmath248 , @xmath249 is limited to be smaller than @xmath269 .
however as mentioned above , if @xmath261 then the above equations imply @xmath270 .
we therefore have @xmath271 , and since @xmath272 in this case we find @xmath273.\ ] ] finally , following the overlap of heii regions , the mass fractions @xmath69 , @xmath200 and @xmath234 are free to evolve according to equation ( [ preoverlap ] ) ( with modified source terms to account for the excess ionizing photons ) . the filling factors @xmath66 , @xmath67 and @xmath183 still evolve according to generalizations of equation ( [ postoverlap ] ) . at the point of he@xmath73 overlap our formalism demands that there is no he@xmath72 in the universe since @xmath276 .
however , following he@xmath73 overlap the he@xmath95 front ( which must lead the he@xmath73 front ) is free to propagate into the denser igm .
therefore , following the overlap of he@xmath73 , @xmath70 becomes the fraction of helium in the universe ( equal to zero initially ) that is singly ionized .
the recombination rate @xmath277 is calculated @xmath278 where @xmath279 and @xmath280 . as before , if @xmath281 substitution of equation ( [ postoverlap ] ) into equation ( [ preoverlap ] ) yields @xmath266 .
the filling factors @xmath66 and @xmath183 therefore remain at unity .
we have @xmath282- s_{\rm kf}^{\rm h^{+}}\\ \nonumber \frac{df_{\rm he^{+}}}{dz } & = & s_{\rm cf}^{\rm he^{+ } } - \frac{1}{n_{\rm he}^{\rm 0}}\delta\frac{dn_{\rm \gamma}^{\rm he^+}}{dz } - s_{\rm kf}^{\rm he^{+ } } - \frac{df_{\rm he^{++}}}{dz}\\ \nonumber \frac{df_{\rm he^{++}}}{dz } & = & s_{\rm cf}^{\rm he^{++ } } - \frac{1}{n_{\rm he}^{\rm 0}}\delta\frac{dn_{\rm \gamma}^{\rm he^{++}}}{dz}- s_{\rm kf}^{\rm he^{++}}\\ \nonumber \frac{dq_{\rm h^+}}{dz}&=&s_{\rm c}^{\rm h^{+ } } + \frac{1}{n_{\rm h}^{\rm 0}f_{\rm h^{+}}}\left[\delta\frac{dn_{\rm \gamma}^{\rm he^+}}{dz}+\delta\frac{dn_{\rm \gamma}^{\rm he^{++}}}{dz}\right ] - s_{\rm k}^{\rm h^{+}}\\ \nonumber \frac{dq_{\rm he^{\rm + } } } { dz}&= & s_{\rm c}^{\rm he^{+ } } - \frac{1}{n_{\rm he}^{\rm 0}f_{\rm he^{+}}}\delta\frac{dn_{\rm \gamma}^{\rm he^{+}}}{dz}- s_{\rm k}^{\rm he^{+ } } -\frac{dq_{\rm he^{\rm + + } } } { dz}\\ \frac{dq_{\rm he^{\rm + + } } } { dz}&=&s_{\rm c}^{\rm he^{++}}- \frac{1}{n_{\rm he}^{\rm 0}f_{\rm he^{++}}}\delta\frac{dn_{\rm \gamma}^{\rm he^{++}}}{dz}- s_{\rm k}^{\rm he^{++}}.\end{aligned}\ ] ] the excess ionizing photon rates are determined as follows . if @xmath283 then @xmath200 is limited to be smaller than @xmath284 since the he@xmath95 front can not propagate beyond the h@xmath95 front . therefore ,
if @xmath285 we require @xmath286 , and since @xmath253 in this case , we find @xmath287.\ ] ] the @xmath163 mass fraction @xmath234 is limited to be less than @xmath288 . therefore , if @xmath289 we require @xmath290 so that the fronts propagate at the same rate , and @xmath291 since the he@xmath72 front can not grow if the @xmath163 is filling all of the available volume .
these constraints yield @xmath292\\ \nonumber & -&n_{\rm he}^{\rm
0}\frac{n_{\rm he}^{\rm 0}+n_{\rm h}^{\rm 0}}{2n_{\rm he}^{\rm 0}+n_{\rm h}^{\rm 0}}\left [ ( s_{\rm cf}^{\rm he^{++}}-s_{\rm cf}^{\rm he^{+ } } ) - ( s_{\rm kf}^{\rm he^{++}}-s_{\rm kf}^{\rm he^{+}})\right]\\\nonumber & & \mbox{and}\\ \delta\frac{dn_{\rm \gamma}^{\rm he^{++}}}{dz}&=&n_{\rm he}^{\rm 0}\left [ ( s_{\rm cf}^{\rm he^{++}}-s_{\rm cf}^{\rm he^{+ } } ) - ( s_{\rm kf}^{\rm he^{++}}-s_{\rm kf}^{\rm he^{+ } } ) + \frac{1}{n_{\rm he}^{\rm 0}}\delta\frac{dn_{\rm \gamma}^{\rm he^{+}}}{dz}\right].\end{aligned}\ ] ] | we compute the reionization histories of hydrogen and helium due to the ionizing radiation fields produced by stars and quasars .
for the quasars we use a model based on halo - merger rates that reproduces all known properties of the quasar luminosity function at high redshifts .
the less constrained properties of the ionizing radiation produced by stars are modeled with two free parameters : ( i ) a transition redshift , @xmath0 , above which the stellar population is dominated by massive , zero - metallicity stars and below which it is dominated by a scalo mass function ; ( ii ) the product of the escape fraction of stellar ionizing photons from their host galaxies and the star - formation efficiency , @xmath1 .
we constrain the allowed range of these free parameters at high redshifts based on the lack of the hi gunn - peterson trough at @xmath2 and the upper limit on the total intergalactic optical depth for electron scattering , @xmath3 , from recent cosmic microwave background ( cmb ) experiments .
we find that quasars ionize helium by a redshift @xmath4 , but can not reionize hydrogen by themselves before @xmath5 . a major fraction of the allowed combinations of @xmath1 and @xmath6 lead to an early peak in the ionized fraction due to metal - free stars at high redshifts .
this sometimes results in two reionization epochs , namely an early hii or heiii overlap phase followed by recombination and a second overlap phase .
even if early overlap is not achieved , the peak in the visibility function for scattering of the cmb often coincides with the early ionization phase rather than with the actual reionization epoch .
consequently , @xmath7 does not correspond directly to the reionization redshift .
we generically find values of @xmath8 , that should be detectable by the map satellite . |
there are few generic predictions in brane world models as possible extensions of the standard model ( sm ) .
the fields propagating in the multidimensional bulk manifest themselves as kk towers of states on the brane , where we are supposed to live .
these kk tower of states are an example of such generic predictions .
another phenomenon of this kind is the presence of a new scalar field ( or fields ) on `` our '' brane ( tev - brane ) associated with spin-0 fluctuations of the metric component corresponding to extra space dimension .
the size of the extra space dimension should be stabilized in order to get a physically meaningful picture .
this leads to the violation of the dilatation invariance , and the spin-0 component of metric fluctuations being combined in some manner with the additional scalar field introduced for stabilizing the size renders a new scalar field called the radion field .
this radion field plays the role of the dilaton in the system at hand . the randall - sundrum setup @xcite with stabilizing scalar field as proposed by goldberger and wise @xcite or worked out by dewolfe , freedman , gubser and karch @xcite , taking into account the backreaction by solving exactly the coupled equations for the metric and the scalar field , provides an example of a concrete realization of a stabilized brane world model . in a number of studies it was argued that the lowest scalar state , usually called the radion @xcite , might be significantly lighter than all the other kk excitations @xcite .
the phenomenology of the radion follows from the simple effective lagrangian @xmath4 where @xmath5 is the trace of the sm energy - momentum tensor , @xmath6 stands for the radion field , and @xmath7 , the notation @xmath8 is also often used . @xmath9 and
another often used parameter @xmath10 are related by @xmath11 . ]
is a dimensional scale parameter .
since the radion plays the role of the dilaton , it interacts with the massless photon and gluon fields via the well - known dilatation anomaly in the trace of the energy momentum tensor . at the lowest order in the sm couplings the trace has the following form , where the fields , as supposed in most of the studies , are taken on the mass - shell , @xmath12 and the sum
is taken over all sm fermions , @xmath13 and @xmath14 are the well - known qcd and qed @xmath15-functions .
the interaction vertices of the radion with the sm fields are similar to those of the sm higgs boson except for the anomaly enhanced interactions with gluons and photons .
this leads to the corresponding relative enhancement in gluon and photon decay modes and the gluon - gluon fusion production channel of the radion .
various aspects of the decay and production properties of the radion have been touched upon in a number of studies , including a possible mixing of the radion with the higgs boson , which is important for the radion and higgs phenomenology @xcite - @xcite .
after the discovery of the higgs - like boson at the lhc @xcite few studies have been curried out investigating a possible impact of the radion as the second scalar particle on the interpretation of the observed boson @xcite - @xcite .
the details depend on particular assumptions of the scenario under consideration , such as which fields propagate in the bulk and which are localized on the brane , what is the higgs - radion mixing mechanism and what is the mixing parameter etc . but generically the light radion with mass below or above the observed 126 gev boson is still not fully excluded by all the ew precision constrains and the lhc data .
the fermion part of lagrangian ( [ l1 ] ) for on - shell fermions has a very simple form @xmath16 being the same as for the higgs boson with the replacement of the scale @xmath9 by the higgs vacuum expectation value , @xmath17 .
however , for the case of off - shell fermions lagrangian ( [ lferm_onshell ] ) needs to be modified .
the corresponding lagrangian can be written symbolically in the following form ( the explicit form of the lagrangian is given in appendix 1 ) @xcite @xmath18 + ... \label{lferm_offshell}\ ] ] where @xmath19 are the sm covariant derivatives .
note that lagrangian ( [ lferm_offshell ] ) is non - trivial even for massless fermions .
the aim of this short study is to investigate a possible influence of these additional terms on the processes with tree level diagrams involving the radion and off - shell fermion lines .
the extra terms in the lagrangian lead to a momentum dependence in the vertices potentially giving additional contributions if virtual fermions participate in the process . for the main radion decay processes the fermions are on - shell , and these additional terms in the vertices vanish , but this might not be the case for the radion production .
however , it will be explicitly demonstrated that due to the gauge invariance all the additional contributions to the amplitudes of physical processes are canceled out , and the result for the matrix element is the same as computed with only higgs - like lagrangian ( [ lferm_onshell ] ) .
one should stress that in order to get the mentioned cancellations the trace @xmath20 of the sm energy momentum tensor should be computed including the sm covariant derivatives as given in appendix 1 .
the corresponding feynman rules containing new 4-point vertices are presented in appendix 2 .
we begin with a simple example of the radion production in the radion - strahlung process .
then we give a more general explanation demonstrating how the cancellation takes place and show a few more examples for the main radion production processes at the lhc and at a linear collider .
we begin with a very simple example , the radion production in association with @xmath21 boson in @xmath22 collisions called the radion - strahlung by analogy with the well - known higgs production process , the higgs - strahlung . even neglecting the very small electron mass
there are four contributing feynman diagrams at the lowest order shown in fig.[fig : radion - strahlung - diagrams ] .
note that the presence of the last 4-point diagram is a consequence of the gauge invariance .
the feynman rules for the vertices involved are given in appendix 2 .
correspondingly the diagrams give the following contributions to the process amplitude @xmath23 @xmath24\,\frac{\cancel{k}}{k^2}\ , \gamma_{\mu}\,e^-(p_1)\,\varepsilon(p_z)\,r(p_r)\ ] ] @xmath25 \,e^-(p_1)\,\varepsilon(p_z)\,r(p_r)\ ] ] @xmath26 where @xmath27 , @xmath28 , @xmath29 , @xmath30 , @xmath31 .
keeping in mind the dirac equation @xmath32 and the equality @xmath33 it is easy to see that the sum of two diagrams d2 and d3 is exactly canceled out by the last diagram d4 , so @xmath34 therefore the matrix element squared @xmath35 for the radion production is exactly equal to that for the higgs boson , and the cross section takes the well - known form with the replacement @xmath17 and @xmath36 @xmath37 where @xmath38 .
as one can see all the additional contributions are canceled out .
the same property of the cancellation of additional to the higgs - like contributions takes place for the associated radion and @xmath39 boson production , for example , @xmath40 where there are also 4 contributing diagrams similar to those in fig.[fig : radion - strahlung - diagrams ] .
the observed cancellation follows from the structure of the fermion current with the emission of the radion and a number of gauge bosons .
let us begin with a simple case with just one gauge boson emission as shown in fig.[fig : radion-1v - diagrams ] .
it corresponds to the radion - strahlung process considered in the previous section with z boson emission .
however the cancellation property is valid for any emitted sm gauge boson and for any massive sm fermion . in order to see that it is instructive to rewrite the fermion - radion vertex in the following way @xmath41 = \ ] ] @xmath42 = \ ] ] @xmath43 , \label{vertex - f - r}\ ] ] where @xmath44 is the inverse function to the propagator @xmath45 note that the last term proportional to @xmath46 in vertex ( [ vertex - f - r ] ) is exactly the same as for the higgs boson with the obvious replacement @xmath47 .
the diagrams in fig.[fig : radion-1v - diagrams ] can be expressed as follows @xmath48\ , u_{in}(p_{in})\ ] ] @xmath49\ , s(q_1)\,\gamma_{\mu } \ , u_{in}(p_{in})\ ] ] @xmath50 where the constant c includes the product of all the factors of the vertices involved , @xmath51 is the lorentz part of the fermion - boson and fermion - boson - radion vertices , which is the same in two vertices coming from the same covariant derivative . the initial and
the final spinors correspond to the same fermion for the neutral fermion current and they are different for the charged current .
once more one can easily see that using the dirac equation for @xmath52 and @xmath53 fermion states and summing up @xmath54 the only remaining contribution is proportional to the fermion mass and is the same as that for the higgs boson .
this property of the cancellation of all the contributions additional to the higgs - like part can be generalized to an arbitrary number , say @xmath55 , of gauge bosons emitted from the fermion line in association with the radion as shown in fig.[fig : radion - nv - diagrams ] .
+ the sum of all diagrams can be split into parts containing the fermion - fermion - radion vertex and the boson - fermion - fermion - radion vertex @xmath56 where @xmath57 [ -\frac{3}{2}\,(\,s^{-1}(q_l)\ , + \,s^{-1}(k_l)\ , ) + m_{f_l } ] & \\ [ \prod_{j=1 + 1}^n \ , s(k_{j-1 } ) \,\gamma_{\mu_j}^j]\ , f_{in}(p_{in } ) , & \end{split } \label{ml}\ ] ] for @xmath58, ...
,@xmath59 @xmath60\ , [ \prod_{j=1}^n \ , s(k_{j-1})\gamma_{\mu_j}^j ] f_{in}(p_{in } ) , \label{ml0}\ ] ] @xmath61 [ -\frac{3}{2}\,(\,s^{-1}(q_n)\ , + \,s^{-1}(p_{in})\ , ) + m_{f_{in } } ] f_{in}(p_{in } ) , \label{mln}\ ] ] @xmath62 [ -3\,\gamma_{\mu_l}^l]\,[\prod_{j = l+1}^{n } \ , s(k_{j-1})\gamma_{\mu_j}^j ] f_{in}(p_{in } ) , \label{mprime - l}\ ] ] for @xmath63, ... ,@xmath59 @xmath64 [ \prod_{j=2}^n \ , s(k_{j-1})\gamma_{\mu_j}^j ] f_{in}(p_{in } ) \label{mprime - l1}\ ] ] @xmath65\,[-3\,\gamma_{\mu_n}^n ] f_{in}(p_{in } ) , \label{mprime - ln}\ ] ] @xmath66 is the sm fermion and gauge boson vertex , @xmath67 , @xmath68 , @xmath69 , @xmath70 is the vector boson momentum , the upper index @xmath71 standing for any sm gauge vector boson ( z , w , photon , gluon ) , @xmath72 is the radion momentum , @xmath73 , @xmath74 , @xmath75 are the masses of the internal , initial and final state fermions which may be different in the case of the charged current .
note that the factors @xmath76 and @xmath77 are different due to the difference by one of the number of the propagators and vertices involved into expressions ( [ ml ] ) , ( [ ml0 ] ) , ( [ mln ] ) and ( [ mprime - l ] ) , ( [ mprime - l1 ] ) , ( [ mprime - ln ] ) .
this leads to a relative ( @xmath78 ) sign between the two contributions . from formulas ( [ ml])([mprime - ln ] )
one can write simple relations taking into account the relation @xmath79 and the dirac equations for @xmath52 and @xmath53 fermions @xmath80 @xmath81 @xmath82 where the higgs - like contributions @xmath83 , @xmath84 , @xmath84 labeled with the @xmath85 symbol being proportional to the fermion masses have the form which follows from ( [ ml]),([ml0]),([mln ] ) @xmath86 [ m_{f_l } ] [ \prod_{j=1 + 1}^n \ , s(k_{j-1 } ) \,\gamma_{\mu_j}^j]\ , f_{in}(p_{in } ) , \end{split } \label{mlh}\ ] ] @xmath87\ , [ \prod_{j=1}^n \ , s(k_{j-1})\gamma_{\mu_j}^j ] f_{in}(p_{in } ) , \label{ml0h}\ ] ] @xmath88 [ m_{f_{in } } ] f_{in}(p_{in } ) , \label{mlny}\ ] ] summing up the left and the right hand sides of equation ( [ relation1 ] ) from @xmath89 to @xmath59 and adding the left and the right hand sides of equations ( [ relation2 ] ) and ( [ relation3 ] ) one gets the following equality @xmath90 the result in formula ( [ main ] ) means that the sum of all the contributions leads to only the higgs - like type of the contribution and all the other parts are canceled out explicitly .
the main result of the previous section can be easily generalized to the loop case .
first of all , one should mention that in the way of proving equality ( [ main ] ) nothing special was required for boson lines in the vertices .
the boson lines in fig.[fig : radion - nv - diagrams ] could correspond to real particles or virtual propagators in the loops . in this sense
all the above considerations are valid for both real and virtual gauge bosons emitted from the fermion current . in the case of a fermion loop
one can follow a similar logic as was given above . in the loop case
there is an additional fermion propagator instead of the external spinors at the tree level .
correspondingly the expressions for various fermion loop contributions with emission of @xmath55 gauge bosons have the following form by analogy with ( [ ml]),([mln ] ) and ( [ mprime - l]),([mprime - l1]),([mprime - ln ] ) . @xmath91 [ -\frac{3}{2}\,(\,s^{-1}(q_l)\ , + \,s^{-1}(k_l)\ , ) + m_{f_l } ] \right . \\
[ \prod_{j=1 + 1}^n \ , s(k_{j-1 } ) \,\gamma_{\mu_j}^j]\,s(p ) \right\ } , \end{split } \label{ml - loop}\ ] ] for @xmath58, ... ,@xmath59 , @xmath92 [ -\frac{3}{2}\,(\,s^{-1}(q_n)\ , + \,s^{-1}(p_{in})\ , ) + m]\,s(p ) \right\ } , \label{mln - loop}\ ] ] @xmath93 [ -3\,\gamma_{\mu_l}^l]\,[\prod_{j = l+1}^{n } \ , s(k_{j-1})\gamma_{\mu_j}^j]\,s(p ) \right\ } , \label{mprime - l - loop}\ ] ] for @xmath63, ...
,@xmath59 @xmath94 [ \prod_{j=2}^n \ , s(k_{j-1})\gamma_{\mu_j}^j ] \,s(p ) \right\ } , \label{mprime - l1-loop}\ ] ] @xmath95\,[-3\,\gamma_{\mu_n}^n ] \,s(p ) \right\}. \label{mprime - ln - loop}\ ] ] where the momenta are expressed as @xmath96 , @xmath97 .
note that the contribution @xmath98 is not present now since it coincides with @xmath99 . in the same manner as in the previous section
we get @xmath100 @xmath101 from relations ( [ relation1-loop ] ) and ( [ relation3-loop ] ) it is easy to show that @xmath102 and therefore one gets once again the equality @xmath103 which demonstrates that all the contributions except for the higgs - like type are canceled out in the case of a fermion loop .
in all the main production processes of the radion , the lowest kk mode of a combination of the spin-0 metric fluctuation and the scalar stabilizing field , there are contributing feynman diagrams involving off - shell fermions .
the radion is emitted from various fermion currents containing off - shell fermion propagators in the radion - strahlung or the vector boson fusion production in @xmath22 collisions as well as in the vector boson fusion , associated radion and vector boson production , associated radion and the top - quark pair production processes in hadron collisions at the lhc .
the off - shell fermions participate in fermion loops for gluon - gluon fusion production process at the lhc and in @xmath104 , @xmath105 , @xmath106 decay modes of the radion .
the additional to the higgs boson case non - trivial vertices ( even for massless fermions ) of fermion - radion interactions follow from the structure of the trace of the gauge invariant sm energy - momentum tensor as given in the appendices 1,2 .
we have shown that all the contributions to perturbative amplitudes of physical processes due to corresponding additional terms are canceled out for both massless and massive off - shell fermions .
we demonstrated , first , the cancellation in a simple example of the radion - strahlung process in @xmath22 collisions .
then we presented a general proof of the cancellation for an arbitrary fermion current radiating the radion and any number of the sm gauge bosons .
this proof was generalized to the case of the amplitudes containing closed fermion loops and an arbitrary number of the gauge bosons .
the proof also means that the terms with the covariant derivatives of the fermion fields in lagrangian ( [ lferm_offshell ] ) can be replaced by the mass terms in accordance with the equation of motion for the fermion fields , which is not _ a priori _ justified in the gauge theory , where the quantization is performed with the help of the path integration .
thus the additional fermion - radion terms in the interaction lagrangian do not alter any production and decay properties of the radion compared to those of the higgs boson .
one should mention that the observed cancellation property , being a result of the gauge invariant structure of the sm energy - momentum trace , is also valid for any scalar particle ( not only the radion ) which interacts with the sm particles via the trace .
the work was supported by grant 14 - 12 - 00363 of russian science foundation .
the authors are grateful to m. smolyakov and i. volobuev for useful discussions and critical remarks .
* trace of the sm energy - momentum tensor . * + the trace of the sm energy - momentum tensor calculated as the variation of the sm lagrangian with respect to the metric @xcite can be written in the unitary gauge as follows : @xmath107 + 4m_{f } \bar{f}f\right\ } + \frac{4h}{v } \sum _ { f}m_{f } \bar{f}f-3ea_{\mu } \sum _ { f}q_{f } \bar{f}\gamma ^{\mu } f \\ & - & \frac{3}{2}\frac{m_{z}}{v } z_{\mu } \sum _ { f}\bar{f}\gamma ^{\mu } \left[a_{f } + b_{f } \gamma _ { 5 } \right]f - \frac{3}{\sqrt{2}}\frac{m_{w}}{v } \left(w_{\mu } ^{- } \bar{\nu } _ { j } u_{jk}^{pmns } \gamma ^{\mu } \left[1-\gamma _ { 5 } \right]e_{k } + h.c .
\right ) \\ & - & \frac{3}{\sqrt{2}}\frac{m_{w}}{v } \left(w_{\mu } ^{- } \bar{u}_{j } \gamma ^{\mu } \left[1-\gamma _ { 5 } \right]v_{jk}^{ckm } d_{k } + h.c .
\right ) - 3g_{c } \left(\bar{u}_{j } \gamma ^{\mu } \hat{g}_{\mu } u_{j } + \bar{d}_{j } \gamma ^{\mu } \hat{g}_{\mu } d_{j } \right ) \\\end{aligned}\ ] ] @xmath108 where @xmath109 is the pontecorvo - maki - nakagawa - sakata matrix , @xmath110 is the cabibbo - kobayashi - maskawa matrix .
the last two terms take the anomalies into account , they describe the interaction of the radion with the photon and the gluon fields .
* interaction vertices of radion and sm fields * + + radion and two fermions : + + @xmath111@xmath112 + 4m_{f } \right\}=\frac{i}{\lambda _ { r } } \left\{\frac{3}{2 } \left[\left(\rlap{$/$}p_{\bar{f } } + m\right)-\left(\rlap{$/$}p_{f } -m\right)\right]+m_{f } \right\}$ ] + + radion , two fermions and photon : + + @xmath113= @xmath114 + + radion , two fermions and @xmath21-boson : + + @xmath115 @xmath116 $ ] + + radion , two fermions and @xmath117-boson : + + @xmath118 @xmath119 $ ] + + @xmath120 @xmath121 $ ] + + radion , two fermions and higgs - boson : + + @xmath122 @xmath123 + + radion and two higgs - bosons : + + @xmath124 @xmath125 + + radion and three higgs - bosons : + + @xmath126 @xmath127 + + radion and four higgs - bosons : + + @xmath128 @xmath129 + + radion and two @xmath21-bosons : + + @xmath130 @xmath131 + + radion , two @xmath21-bosons and higgs - boson : + + @xmath132 @xmath133 + + radion , two @xmath21-bosons and two higgs - bosons : + + @xmath134 @xmath135 + + radion and two @xmath117-bosons : + + @xmath136 @xmath137 + + radion , two @xmath117-bosons and higgs - boson : + + @xmath138 @xmath139 + + radion , two @xmath117-bosons and two higgs - bosons : + + @xmath140 @xmath141 + + w. d. goldberger and m. b. wise , phys . lett .
b * 475 * , 275 ( 2000 ) [ hep - ph/9911457 ] ; c. csaki , m. graesser , l. randall and j. terning , phys .
d * 62 * , 045015 ( 2000 ) [ hep - ph/9911406 ] . c. charmousis , r. gregory and v. a. rubakov , phys .
d * 62 * , 067505 ( 2000 ) [ hep - th/9912160 ] .
e. e. boos , y. s. mikhailov , m. n. smolyakov and i. p. volobuev , nucl .
b * 717 * , 19 ( 2005 ) [ hep - th/0412204 ] .
g. f. giudice , r. rattazzi and j. d. wells , nucl .
b * 595 * , 250 ( 2001 ) [ hep - ph/0002178 ] ; k. -m .
cheung , phys .
d * 63 * , 056007 ( 2001 ) [ hep - ph/0009232 ] ; g. d. kribs , econf c * 010630 * , p317 ( 2001 ) [ hep - ph/0110242 ] ; m. chaichian , a. datta , k. huitu and z. -h .
yu , phys .
b * 524 * , 161 ( 2002 ) [ hep - ph/0110035 ] .
t. g. rizzo , jhep * 0206 * , 056 ( 2002 ) [ hep - ph/0205242 ] ; d. dominici , b. grzadkowski , j. f. gunion and m. toharia , nucl .
b * 671 * , 243 ( 2003 ) [ hep - ph/0206192 ] ; j. f. gunion , m. toharia and j. d. wells , phys . lett .
b * 585 * , 295 ( 2004 ) [ hep - ph/0311219 ] .
z. chacko , r. franceschini and r. k. mishra , jhep * 1304 * , 015 ( 2013 ) [ arxiv:1209.3259 [ hep - ph ] ] ; z. chacko , r. k. mishra and d. stolarski , jhep * 1309 * , 121 ( 2013 ) [ arxiv:1304.1795 [ hep - ph ] ] ; g. -c .
cho , d. nomura and y. ohno , mod .
a * 28 * , 1350148 ( 2013 ) [ arxiv:1305.4431 [ hep - ph ] ] ; n. desai , u. maitra and b. mukhopadhyaya , jhep * 1310 * , 093 ( 2013 ) [ arxiv:1307.3765 [ hep - ph ] ] ; p. cox , a. d. medina , t. s. ray and a. spray , jhep * 1402 * , 032 ( 2014 ) [ arxiv:1311.3663 [ hep - ph ] ] ; m. geller , s. bar - shalom and a. soni , phys .
d * 89 * , 095015 ( 2014 ) [ arxiv:1312.3331 [ hep - ph ] ] ; d. -w .
jung and p. ko , phys .
b * 732 * , 364 ( 2014 ) [ arxiv:1401.5586 [ hep - ph ] ] .
e. e. boos , v. e. bunichev , m. n. smolyakov and i. p. volobuev , phys .
d * 79 * , 104013 ( 2009 ) [ arxiv:0710.3100 [ hep - ph ] ] .
landau , e.m .
lifshitz , `` the classical theory of fields '' , ch . 11.94 , pergamon press , oxford ( 1975 ) . | the appearance of the radion field , which is associated with the spin-0 metric fluctuations combined in some manner with the scalar stabilizing field , and of the radion , the corresponding lowest kaluza - klein ( kk ) mode , is a generic prediction of stabilized brane world models . in such models
the radion plays the role of the dilaton , and its mass may be somewhat smaller than that of all the kk modes of other particles propagating in the multidimensional bulk . due to its origin , the radion couples to the trace of the energy - momentum tensor of the standard model , the interaction lagrangian of the radion and the standard model fermions being similar to that of the sm higgs - fermion interactions except for additional terms , which come into play only in the case of off - shell fermions . in the present paper
it is shown that all the contributions to perturbative amplitudes of physical processes due to these additional terms are canceled out for both massless and massive off - shell fermions .
thus the additional fermion - radion terms in the interaction lagrangian do not alter any production and decay properties of the radion compared to those of the higgs boson .
* higgs - radion similarity in production processes involving off - shell fermions + * e. boos@xmath0 , s. keizerov@xmath0 , e. rahmetov@xmath1 , k. svirina@xmath2 + + _ @xmath1skobeltsyn institute of nuclear physics , lomonosov moscow state university _ + _ leninskie gory , 119991 , moscow , russia _ + _ @xmath3faculty of physics , lomonosov moscow state university , leninskie gory _
+ _ 119991 , moscow , russia _ |
the equivariant gromov - witten theory of @xmath0 has been studied extensively . in @xcite ,
okounkov - pandharipande completely solved the equivariant gromov - witten theory of the projective line and established a gw / h correspondence between the stationary sector of gromov - witten theory of @xmath0 and hurwitz theory . in @xcite , givental derived a quantization formula for all genus descendant potential of the equivariant gromov - witten theory of @xmath0 ( and more generally , @xmath1 ) . in the non - equivariant limit , these results imply the virasoro conjecture of @xmath0 .
the norbury - scott conjecture @xcite relates ( non - equivariant ) gromov - witten invariants of @xmath0 to eynard - orantin invariants @xcite of the affine plane curve @xmath2 . in @xcite , p. dunin - barkowski ,
n. orantin , s. shadrin , and l. spitz relate the eynard - orantin topological recursion to the givental formula for the ancestor formal gromov - witten potential , and prove the norbury - scott conjecture using their main result and givental s quantization formula for all genus descendant potential of the ( non - equivariant ) gromov - witten theory of @xmath0 .
it is natural to ask if the norbury - scott conjecture can be extended to the equivariant setting , such that the original conjecture can be recovered in the non - equivariant limit .
our first main result ( theorem [ main ] in section [ sec : main - results ] ) relates equivariant gromov - witten invariants of @xmath0 to the eynard - orantin invariants @xcite of the affine curve @xmath3 where @xmath4 are complex parameters , @xmath5 are equivariant parameters of the torus @xmath6 acting on @xmath0 , and @xmath7 is the novikov variable encoding the degree of the stable maps to @xmath0 ( see section [ sec : tgw ] ) .
the superpotential of the @xmath8-equivariant landau - ginzburg mirror of the projective line is given by @xmath9 so theorem [ main ] can be viewed as a version of all genus equivariant mirror symmetry for @xmath0 .
we prove theorem [ main ] using the main result in @xcite and a suitable version of givental s formula for all genus _ equivariant _ descendant gromov - witten potential of @xmath1 @xcite ( see also @xcite ) .
our second main result ( theorem [ full - descendant ] in section [ sec : main - results ] ) gives a precise correspondence between ( a ) genus-@xmath10 , @xmath11-point descendant equivariant gromov - witten invariants of @xmath0 , and ( b ) laplace transforms of the eynard - orantin invariant @xmath12 along lefschetz thimbles .
this result generalizes the known relation between the a - model genus-@xmath13 @xmath14-point descendant gromov - witten invariants and the b - model oscillatory integrals . taking the non - equivariant limit @xmath15
, we obtain @xmath16 which is the superpotential of the ( non - equivariant ) landau - ginzburg mirror for the projective line .
we obtain all genus ( non - equivariant ) mirror symmetry for the projective line . in the stationary phase @xmath17 ,
the curve becomes @xmath18 and theorem [ main ] specializes to the norbury - scott conjecture @xcite .
( see section [ sec : ns ] for details . )
let @xmath19 , @xmath20 and @xmath21 , we obtain @xmath22 which reduces to @xmath23 in the large radius limit @xmath24 .
the @xmath25-equivariant mirror of the affine line @xmath26 is given by @xmath27 in the large radius limit , we obtain a version of all genus @xmath25-equivariant mirror symmetry of the affine line @xmath26 . in particular ,
let @xmath28 and @xmath29 , we obtain the lambert curve @xmath30 in this limit , theorem [ main ] specializes to the bouchard - mario conjecture @xcite relating simple hurwitz numbers ( related to linear hodge integrals by the elsv formula @xcite ) to eynard - orantin invariants of the lambert curve .
( see section [ sec : bm ] for details . ) in @xcite , borot - eynard - mulase - safnuk introduced a new matrix model representation for the generating function of simple hurwitz numbers , and proved the bouchard - mario conjecture . in @xcite ,
eynard - mulase - safnuk provided another proof of the bouchard - mario conjecture using the cut - and - joint equation of simple hurwitz numbers .
recently , a new proof of the elsv formula and a new proof of the bouchard - mario conjecture are given in @xcite .
we thank p. dunin - barkowski , b. eynard , m. mulase , p. norbury , and n. orantin for helpful conversations .
the research of the authors is partially supported by nsf dms-1206667 and nsf dms-1159416 .
let @xmath6 act on @xmath0 by @xmath31= [ t_1^{-1 } z_1 , t_2^{-1 } z_2].\ ] ] let @xmath32:={\mathbb{c}}[{\mathsf{w}}_1,{\mathsf{w}}_2]$ ] be the @xmath8-equivariant cohomology of a point : @xmath33 $ ] .
the @xmath8-equivariant cohomology of @xmath0 is given by @xmath34/\langle ( h-{\mathsf{w}}_1)(h-{\mathsf{w}}_2)\rangle\ ] ] where @xmath35 .
let @xmath36 $ ] and @xmath37 $ ] be the @xmath8 fixed points
. then @xmath38 .
the @xmath8-equivariant poincar dual of @xmath39 and @xmath40 are @xmath41 and @xmath42 , respectively .
let @xmath43}{\mathbb{c}}[{\mathsf{w } } , \frac{1}{{\mathsf{w}}_1-{\mathsf{w}}_2}]\ ] ] then @xmath44 , @xmath45 so @xmath46 is a canonical basis of the semisimple algebra @xmath47 } { \mathbb{c}}[{\mathsf{w}},\frac{1}{{\mathsf{w}}_1-{\mathsf{w}}_2}].\ ] ] we have @xmath48 @xmath49 where @xmath50 cup product with the hyperplane class is given by @xmath51 suppose that @xmath52 or @xmath53 , so that @xmath54 is non - empty . given @xmath55 and @xmath56 , we define genus @xmath10 , degree @xmath57 , @xmath8-equivariant descendant gromov - witten invariants of @xmath0 : @xmath58^{{\mathrm{vir } } } } \prod_{j=1}^n \psi_j^{a_j}{\mathrm{ev}}_j^*(\gamma_j ) \in { \mathbb{c}}[{\mathsf{w}}]\ ] ] where @xmath59 is the evaluation at the @xmath60-th marked point , which is a @xmath8-equivariant map .
we define genus @xmath10 , degree @xmath57 primary gromov - witten invariants : @xmath61 let @xmath62 .
if @xmath53 , define @xmath63 suppose that @xmath64 .
given @xmath65 , we define @xmath66 in particular , if @xmath67 then @xmath68 where we use the fact @xmath69 , and the identity @xmath70 we use the second line of to extend the definition of the correlator in the first line of to the unstable cases @xmath71 : @xmath72 suppose that @xmath64 or @xmath73 . given @xmath74 , we define @xmath75 let @xmath21 . then for @xmath76 , @xmath77 as a @xmath32$]-module , @xmath78 .
the ring structure is given by the quantum product @xmath79 defined by @xmath80 or equivalently , @xmath81 where @xmath82 is the product in @xmath83 , and @xmath21 .
in particular , @xmath84 the @xmath8-equivariant quantum cohomology of @xmath0 is @xmath85/\langle ( h-{\mathsf{w}}_1)\star(h-{\mathsf{w}}_2)-q\rangle\ ] ] where @xmath35 , @xmath86 .
the ( non - equivariant ) quantum cohomology of @xmath0 is @xmath87/\langle h\star h -q\rangle\ ] ] let @xmath88 then @xmath89 so @xmath90 is a canonical basis of the semi - simple algebra @xmath91\ ] ] where @xmath92 is defined by .
we also have @xmath93 where @xmath94 quantum multiplication by the hyperplane class is given by @xmath95 finally , we take the non - equivariant limit @xmath19 , @xmath96 .
we obtain : @xmath97 @xmath98 @xmath99 these non - equivariant limits coincide with the results in ( * ? ? ?
* section 2 ) .
let @xmath101 be the flat coordinates with respect to the basis @xmath102 , and let @xmath103 be the canonical coordinates with respect to the basis @xmath104 .
then @xmath105 @xmath106 the above equations determine the canonical coordinates @xmath107 and @xmath108 up to a constant in @xmath109 $ ] .
givental s a - model canonical coordinates @xmath110 are characterized by their large radius limits : @xmath111 for @xmath112 and @xmath113 , define @xmath114 by @xmath115 and define the @xmath100-matrix to be @xmath116.\ ] ] then @xmath117 = \left[\begin{array}{cc } dt^0 & dt^1\end{array}\right]\psi,\ ] ] @xmath118 let @xmath119\ ] ] be the inverse matrix of @xmath100 , so that @xmath120 then @xmath121 let @xmath122 i.e. @xmath123 .
we take the non - equivariant limit @xmath19 , @xmath96 : @xmath124 @xmath125 these non - equivariant limits agree with the results in ( * ? ? ?
* section 2 ) .
the @xmath126-operator is defined as follows . for any cohomology classes
@xmath127 , @xmath128 the @xmath8-equivariant @xmath129-function is characterized by @xmath130 for any @xmath131 .
let @xmath132 we consider several different ( flat ) bases for @xmath133 : 1 .
the canonical basis : @xmath134 , @xmath135 .
2 . the basis dual to the canonical basis with respect to the @xmath8-equivariant poincare pairing : @xmath136 , @xmath137 .
3 . the normalized canonical basis @xmath138 , @xmath139 , which is self - dual : @xmath140 , @xmath141 .
the natural basis : @xmath142 , @xmath143 .
the basis dual to the natural basis : @xmath144 , @xmath145 . for @xmath146 ,
define @xmath147 then @xmath148 is the matrix of the @xmath126-operator with respect to the ordered basis @xmath149 : @xmath150 for @xmath113 and @xmath112 , define @xmath151 then @xmath152 is the matrix of the @xmath126-operator with respect to the ordered bases @xmath153 and @xmath154 : @xmath155 we have @xmath156 by @xcite , the equivariant @xmath129-function is @xmath157 for @xmath158 , define @xmath159 then @xmath160 so @xmath161 following givental , we define @xmath162 then @xmath163 by givental @xcite , the matrix @xmath165 is of the form @xmath166 where @xmath167 and is unitary , and @xmath168 introducing formal variables @xmath169 where @xmath170 define @xmath171^{{\mathrm{vir } } } } \prod_{j=1}^n { \mathrm{ev}}_j^*(u_{a_j } ) \psi_j^{a_j}\prod_{i=1}^m{\mathrm{ev}}_{i+n}^*({\mathbf{t}}).\end{aligned}\ ] ] we define the total descendent potential of @xmath0 to be @xmath172 consider the map @xmath173 which forgets the map to the target and the last @xmath174 marked points .
let @xmath175 be the pull - backs of the classes @xmath176 , from @xmath177 .
then we can define @xmath178^{{\mathrm{vir } } } } \prod_{j=1}^n { \mathrm{ev}}_j^*(u_{a_j } ) \bar\psi_j^{a_j}\prod_{i=1}^m{\mathrm{ev}}_{i+n}^*({\mathbf{t}}).\ ] ] let the ancestor potential of @xmath0 be @xmath179 the quantization of the @xmath126-operator relates the ancestor potential and the descendent potential of @xmath0 via givental s formula . concretely , we have ( see @xcite ) @xmath180 where @xmath181 denotes @xmath182 at @xmath183 and @xmath184 is the quantization @xcite of @xmath126 . for our purpose , we need to describe a formula for a slightly different potential : @xmath185the descendent potential with arbitrary primary insertions .
now we first describe a graph sum formula for the ancestor potential @xmath186 . given a connected graph @xmath187 , we introduce the following notation . 1 .
@xmath188 is the set of vertices in @xmath187 .
@xmath189 is the set of edges in @xmath187 .
@xmath190 is the set of half edges in @xmath191 .
@xmath192 is the set of ordinary leaves in @xmath187 .
@xmath193 is the set of dilaton leaves in @xmath187 . with the above notation , we introduce the following labels : 1 .
( genus ) @xmath194 .
( marking ) @xmath195 .
this induces @xmath196 , as follows : if @xmath197 is a leaf attached to a vertex @xmath198 , define @xmath199 .
3 . ( height )
@xmath200 . given an edge @xmath201 ,
let @xmath202 be the two half edges associated to @xmath201 .
the order of the two half edges does not affect the graph sum formula in this paper . given a vertex @xmath198 ,
let @xmath203 denote the set of half edges emanating from @xmath204 .
the valency of the vertex @xmath204 is equal to the cardinality of the set @xmath203 : @xmath205 .
a labeled graph @xmath206 is _ stable _ if @xmath207 for all @xmath198 .
let @xmath208 denote the set of all stable labeled graphs @xmath209 .
the genus of a stable labeled graph @xmath210 is defined to be @xmath211 define @xmath212 given @xmath112 , define @xmath213 we assign weights to leaves , edges , and vertices of a labeled graph @xmath214 as follows . 1 .
_ ordinary leaves .
_ to each ordinary leaf @xmath215 with @xmath216 and @xmath217 , we assign : @xmath218 ( \sum_{\alpha=1,2}\frac{{\mathbf{u}}^\alpha(z)}{\sqrt{\delta^\alpha(q ) } } r_\alpha^{\,\ \beta}(-z ) ) .\ ] ] 2 .
_ dilaton leaves .
_ to each dilaton leaf @xmath219 with @xmath220 and @xmath221 , we assign @xmath222(-\sum_{\alpha=1,2}\frac{1}{\sqrt{\delta^\alpha(q ) } } r_\alpha^{\,\ \beta}(-z)).\ ] ] 3 . _
_ to an edge connecting a vertex marked by @xmath112 to a vertex marked by @xmath223 and with heights @xmath224 and @xmath225 at the corresponding half - edges , we assign @xmath226 \bigl(\frac{1}{z+w } ( \delta_{\alpha,\beta}-\sum_{\gamma=1,2 } r_\gamma^{\,\ \alpha}(-z ) r_\gamma^{\,\ \beta}(-w)\bigr).\ ] ] 4 .
_ vertices .
_ to a vertex @xmath204 with genus @xmath227 and with marking @xmath228 , with @xmath11 ordinary leaves and half - edges attached to it with heights @xmath229 and @xmath174 more dilaton leaves with heights @xmath230 , we assign @xmath231 we define the weight of a labeled graph @xmath214 to be @xmath232 then @xmath233 now we describe a graph sum formula for @xmath185the descendant potential with arbitrary primary insertions . for @xmath158 , let @xmath234 then @xmath235 , @xmath236 is the normalized canonical basis of @xmath237 , the @xmath8-equivariant quantum cohomology of @xmath0 .
define @xmath238 then @xmath239 is the matrix of the @xmath126-operator with respect to the ordered basis @xmath240 : @xmath241 we define a new weight of the ordinary leaves : 1 . _
ordinary leaves .
_ to each ordinary leaf @xmath215 with @xmath216 and @xmath217 , we assign : @xmath242 ( \sum_{\alpha,\gamma=1,2 } \frac{{\mathbf{u}}^\alpha(z)}{\sqrt{\delta^\alpha(q ) } } { s}^{{{\hat{\underline \gamma}}}}_{\,\ { { \hat{\underline \alpha}}}}(z ) r(-z)_\gamma^{\,\ \beta } ) .\ ] ] we define a new weight of a labeled graph @xmath214 to be @xmath243 then @xmath244 we can slightly generalize this graph sum formula to the case where we have @xmath11 _ ordered _
variables @xmath245 and @xmath11 _ ordered _ ordinary leaves .
let @xmath246 and let @xmath247^{{\mathrm{vir } } } } \prod_{j=1}^n { \mathrm{ev}}_j^*((u_j)_{a_j } ) \psi_j^{a_j}\prod_{i=1}^m{\mathrm{ev}}_{i+n}^*({\mathbf{t}}).\ ] ] define the set of graphs @xmath248 as the definition of @xmath249 except that the @xmath11 ordinary leaves are _
ordered_. let @xmath250 be the ordinary leaves in @xmath251 and for @xmath252 let @xmath253 ( \sum_{\alpha,\gamma=1,2 } \frac{{\mathbf{u}}^\alpha_j(z)}{\sqrt{\delta^\alpha(q ) } } { s}^{{{\hat{\underline \gamma}}}}_{\,\ { { \hat{\underline \alpha}}}}(z ) r(-z)_\gamma^{\,\ \beta } ) .\ ] ] define the weight @xmath254 then @xmath255
let @xmath256 be coordinates on @xmath25 .
the @xmath8-equivariant superpotential @xmath257 is given by @xmath258 where @xmath259 and @xmath260 . in this section ,
we assume @xmath261 is a positive real number .
the jacobian ring of @xmath262 is @xmath263/\langle \frac{\partial w_t^{\mathsf{w}}}{\partial y } \rangle = { \mathbb{c}}[y , y^{-1},q,{\mathsf{w}}]/\langle y-\frac{q}{y } + { \mathsf{w}}_1-{\mathsf{w}}_2\rangle\ ] ] let @xmath264 the jacobian ring is isomorphic to @xmath265 if one identifies @xmath266 with @xmath267 @xmath268/\langle(b-{\mathsf{w}}_1)(b-{\mathsf{w}}_2)-q\rangle.\ ] ] the critical points of @xmath269 are @xmath270 , @xmath271 , where @xmath272 endow a metric on @xmath273 by the residue pairing @xmath274 by direct calculation , we have @xmath275 we denote @xmath276 , @xmath277 and @xmath278 by @xmath279 .
these calculations show the following well - known fact .
there is an isomorphism of frobenius manifold @xmath280}{\mathbb{c}}[{\mathsf{w}},\frac{1}{{\mathsf{w}}_1-{\mathsf{w}}_2 } ] \cong { { \mathrm{jac } } } ( w^{\mathsf{w}}_t)\otimes_{{\mathbb{c}}[{\mathsf{w}}]}{\mathbb{c}}[{\mathsf{w}},\frac{1}{{\mathsf{w}}_1-{\mathsf{w}}_2}].\ ] ] we denote @xmath281}{\mathbb{c}}[{\mathsf{w}},\frac{1}{{\mathsf{w}}_1-{\mathsf{w}}_2}]$ ] by @xmath282 .
the dubrovin - givental connection is denoted by @xmath283 on @xmath284 .
the isomorphism of frobenius structures automatically ensures their canonical coordinates are the same up to a permutation and constants .
we fix the b - model canonical coordinates in this subsection by the critical values of the superpotential @xmath262 , and find the constant difference to the a - model coordinates that we set up in earlier sections .
let @xmath285 be the graph of the equivariant superpotential .
it is a covering of @xmath25 given by @xmath286 .
let @xmath287 be the compactification of @xmath25 with @xmath288 as its coordinate .
at each branch point @xmath289 , @xmath290 and @xmath291 have the following expansion @xmath292 where @xmath293 .
note that we define @xmath294 by @xmath295 , which differs from the definition of @xmath296 in @xcite by a factor of @xmath297 .
the critical values are @xmath298 since @xmath299 we have @xmath300 recall that @xmath301 , so in the large radius limit @xmath24 , we have @xmath302 from , , and , we conclude that @xmath303 where @xmath304 on @xmath305 , let @xmath306 be the liouville form on @xmath307 .
then @xmath308 . let @xmath309 then @xmath310 is a holomorphic 1-form on @xmath26 . recall that @xmath21 and @xmath311 .
define @xmath312 then @xmath313 descends to holomorphic 1-forms on @xmath25 which extends to meromorphic 1-forms on @xmath0 .
we have * @xmath314 has simple poles at @xmath315 and @xmath316 , and @xmath317 * @xmath318 is an exact 1-form . let @xmath319 be the fundamental normalized differential of the second kind on @xmath320 ( see e.g. @xcite ) .
it is also called the bergman kernel in @xcite . in this simple case @xmath287 ,
we have @xmath321 following @xcite , given @xmath158 and @xmath322 , define @xmath323 then @xmath324 satisfies the following properties . 1 .
@xmath324 is a meromorphic 1-form on @xmath0 with a single pole of order @xmath325 at @xmath326 .
2 . in local coordinate @xmath294 near @xmath326 , @xmath327 where @xmath328 is analytic around @xmath326 .
the residue of @xmath324 at @xmath326 is zero , so @xmath324 is a differential of the second kind . the meromorphic 1-form @xmath324 is characterized by the above properties ; @xmath324 can be viewed as a section in @xmath329 . in particular
, @xmath330 is @xmath331
then we have @xmath332 we have @xmath333 for @xmath146 , @xmath113 and @xmath334 , define @xmath335 where @xmath336 is the lefschetz thimble going through @xmath326 , such that @xmath337 near its ends .
it is straightforward to check that @xmath338 is a solution to the quantum differential equation @xmath339 for @xmath158 .
we quote the following theorem [ thm : factorize - s ] near a semi - simple point on a frobenius manifold of dimension @xmath11 , there is a fundamental solution @xmath340 to the quantum differential equation satisfying the following properties 1 .
@xmath340 has the following form + @xmath341 where @xmath342 is matrix of formal power series in @xmath343 , and @xmath344 is a matrix formed by canonical coordinates .
if @xmath340 is unitary under the pairing of the frobenius structure , then @xmath342 is unique up to a right multiplication of @xmath345 where @xmath346 are constant diagonal matrices .
[ a - model s ] for equivariant gromov - witten theory of @xmath0 , the fundamental solution @xmath340 in theorem [ thm : factorize - s ] is viewed as a matrix with entries in @xmath347((z))[[q , t^0,t^1]]$ ] .
we choose the canonical coordinates @xmath348 such that there is no constant term by equation .
then if we fix the powers of @xmath349 and @xmath350 , only finitely many terms in the expansion of @xmath351 contribute .
so the multiplication @xmath352 is well defined and the result matrix indeed has entries in @xmath347((z))[[q , t^0,t^1]]$ ] . for a general abstract semi - simple frobenius manifold defined over a ring @xmath353 , the expression @xmath354 in theorem [ thm : factorize - s ]
can be understood in the following way .
we consider the free module @xmath355 over the ring @xmath356 $ ] where @xmath357 are the flat coordinates of the frobenius manifold .
we formally define the differential @xmath358 and we extend the differential to @xmath359 by the product rule .
then we have a map @xmath360 .
we consider the fundamental solution @xmath354 as a matrix with entries in @xmath359 .
the meaning that @xmath340 satisfies the quantum differential equation is understood by the above formal differential . in our case ,
the multiplication in the a - model fundamental solution @xmath354 is formal in @xmath343 as in remark [ a - model s ] . on b - model side ,
we use the stationary phase expansion to obtain a product of the form @xmath361 .
the multiplications @xmath361 on both a - model and b - model can be viewed as matrices with entries in @xmath359 . and their differentials are obvious the same with the formal differential above .
we repeat the argument in givental @xcite and state it as the following fact .
the fundamental solution matrix @xmath362 has the following asymptotic expansion where @xmath363 is a formal power series in @xmath343 @xmath364 by the stationary phase expansion , @xmath365 it follows that @xmath366 can be asymptotically expanded in the desired form ( notice that @xmath100 is a matrix in @xmath343-degree @xmath13 ) . in particular , by @xmath367 following eynard @xcite , define laplace transform of the bergman kernel @xmath368 where @xmath146 . by (
* equation ( b.9 ) ) , @xmath369 setting @xmath370 , we conclude that @xmath371 .
this shows @xmath372 is unitary . following iritani @xcite ( with slight modification ) , we introduce the following definition .
[ k - framing ] we define @xmath373 $ ] by the following two properties which uniquely characterize it . *
@xmath374 is a homomorphism of additive groups : @xmath375 * if @xmath376 is a @xmath8-equivariant line bundle on @xmath0 then @xmath377 for any @xmath378 , we define the _ k - theoretic framing _ of @xmath379 by @xmath380 where @xmath381 . by localization ,
property ( b ) in the above definition is characterized by @xmath382 where @xmath383 is the inclusion map .
the following definition is motivated by @xcite .
[ syz - dual ] let @xmath384 be an equivariant ample line bundle on @xmath0 , where @xmath385 are integers such that @xmath386 .
we define the equivariant syz t - dual @xmath387 of @xmath376 to be the oriented graph in figure 1 below . in @xmath26 ]
we extend the definition additively to the equivariant k - theory group @xmath389 . in @xmath26 and the ( non - equivariant ) syz t - dual of @xmath390 in @xmath25 .
] the following theorem gives a precise correspondence between the b - model oscillatory integrals and the a - model 1-point descendant invariants .
[ central - charge ] suppose that @xmath391 .
then for any @xmath392 , @xmath393 @xmath394 here @xmath395 .
the left hand side of is @xmath396 by the string equation , the right hand side of is @xmath397 so is equivalent to .
it remains to prove for @xmath398 , where @xmath399 .
we will express both hand sides of in terms of ( modified ) bessel functions .
a brief review of bessel functions is given in appendix [ sec : bessel ] .
the equivariant quantum differential equation of @xmath0 is related to the modified bessel differential equation by a simple transform ( see appendix [ sec : qde ] ) .
let @xmath400 be defined as in appendix [ sec : bessel ] .
@xmath401 by lemma [ integral - bessel ] , @xmath402 therefore , the left hand side of is @xmath403 recall from section [ sec : a - s ] that @xmath404 we have @xmath405 @xmath406 so the right hand side of is @xmath407 definition [ k - framing ] ( equivariant k - theoretic framing ) and definition [ syz - dual ] ( equivariant syz t - dual ) can be extended to any projective toric manifold . in @xcite ,
the first author uses the mirror theorem @xcite and results in @xcite to extend proposition [ central - charge ] to any semi - fano projective toric manifold .
the left hand side of is known as the central charge of the lagrangian brane @xmath387 .
[ abr ] the a and b - model @xmath164-matrices are equal @xmath408 by the asymptotic decomposition theorem of the @xmath340-matrix ( theorem [ thm : factorize - s]),we only have to compare at the limit @xmath409 since both @xmath410 and @xmath411 are unitary .
notice that @xmath100 has an non - degenerate limit at @xmath412 , then it suffices to show that @xmath413 the lefschetz thimble @xmath414 is @xmath415 .
while the lefschetz thimble @xmath416 could not be explicitly depicted , we could alternatively consider the thimble @xmath417 for @xmath418 of the oscillating integral @xmath419 .
the integral yields the same asymptotic answer once we analytically continue @xmath418 to @xmath334 , since the stationary phase expansion only depends on the local behavior ( higher order derivatives ) of @xmath262 at the critical points .
so letting @xmath420 for @xmath421 , or @xmath422 for @xmath423 , @xmath424 taking the limit @xmath24
@xmath425 here we use the stirling formula @xmath426 notice that @xmath427 and similar calculation shows ( letting @xmath420 if @xmath421 and @xmath422 if @xmath423 ) @xmath428 notice that the matrix @xmath372 is given by the asymptotic expansion .
this theorem does not imply @xmath429 , which are unequal .
let @xmath12 be defined recursively by the eynard - orantin topological recursion @xcite : @xmath430 when @xmath53 , @xmath431 where @xmath432 is in a small neighborhood of @xmath326 , and @xmath433 is the other point in the neighborhood such that @xmath434 .
the b - model invariants @xmath12 can be expressed as graph sums @xcite .
we will use the formula stated in ( * ? ? ?
* theorem 3.7 ) , which is equivalent to the formula in ( * ? ? ?
* theorem 5.1 ) . given a labeled graph @xmath435 with @xmath436
, we define its weight to be @xmath437 here , @xmath438 .
note that the definitions of @xmath439 , @xmath440 , @xmath441 in this paper are slightly different from those in @xcite ; for example , the definition of @xmath442 in this paper differs from equation ( 3.11 ) of @xcite by a factor of @xmath443 . in our notation (
* theorem 3.7 ) is equivalent to : [ thm : doss ] for @xmath53 , @xmath444 given a meromorphic function @xmath445 on @xmath0 which is holomorphic on @xmath446 , define @xmath447 then @xmath448 is also a meromorphic function which is holomorphic on @xmath449 . for @xmath112 ,
let @xmath450 then @xmath451 is a meromorphic function on @xmath0 with a simple pole at @xmath289 and holomorphic elsewhere .
moreover , the differential of @xmath451 is @xmath330 . for @xmath452 , define @xmath453 define @xmath454 then @xmath455 therefore , @xmath456 where the last equality follows from theorem [ central - charge ] . for @xmath158 and @xmath252 ,
let @xmath457 [ main ] for @xmath73 and @xmath53 , we have @xmath458 we will prove this theorem by comparing the a - model graph sum in the end of section 2.7 and the b - model graph sum in the previous section .
1 . vertex . by section 3.1
, we have @xmath293 .
so in the b - model vertex , @xmath459 . therefore the b - model vertex matches the a - model vertex .
2 . edge . by section 3.6 ,
we know that @xmath460\left(\frac{uv}{u+v}(\delta_{\alpha,\beta } -\sum_{\gamma=1,2 } f^\alpha_\gamma(u , q)f^\beta_\gamma(v , q))\right)\\ & = & [ z^{k}w^{l}]\left(\frac{1}{z+w}(\delta_{\alpha,\beta } -\sum_{\gamma=1,2 } f^\alpha_\gamma(\frac{1}{z},q)f^\beta_\gamma(\frac{1}{w},q))\right).\end{aligned}\ ] ] by definition @xmath461 \bigl(\frac{1}{z+w } ( \delta_{\alpha,\beta}-\sum_{\gamma=1,2 } r_\gamma^{\,\ \alpha}(-z ) r_\gamma^{\,\ \beta}(-w)\bigr).\ ] ] but we know that @xmath462 therefore , we have @xmath463 3 . ordinary leaf .
we have the following expression for @xmath464 ( see @xcite ) : @xmath465 by item 2 ( edge ) above , for @xmath466 , @xmath467 \bigl(\frac{1}{z+w } ( \delta_{\alpha,\beta}-\sum_{\gamma=1,2 } r_\gamma^{\,\ \alpha}(-z ) r_\gamma^{\,\ \beta}(-w))\bigr).\end{aligned}\ ] ] we also have @xmath468(r_\beta^{\,\ \alpha}(-z))=\delta_{\alpha,\beta}.\ ] ] therefore , @xmath469r_\beta^{\,\ \alpha}(-z)\right)w^\beta_i.\ ] ] so under the identification @xmath470 the b - model ordinary leaf matches the a - model ordinary leaf .
4 . dilaton leaf .
we have the following relation between @xmath471 and @xmath472 ( see @xcite ) @xmath473\sum_{\beta}\sqrt{-1}h^\beta_1f^\alpha_\beta(u , q).\ ] ] by the relation @xmath474 and the fact @xmath475 , it is easy to see that the b - model dilaton leaf matches the a - model dilaton leaf .
taking laplace transforms at appropriate cycles to theorem [ main ] produces a theorem concerning descendants potential .
[ full - descendant ] suppose that @xmath73 and @xmath53 .
for any @xmath476 , there is a formal power series identity @xmath477 by theorem [ central - charge ] , @xmath478 which is the analogue of in the unstable case @xmath479 . by ,
@xmath480 define the flat coordinates @xmath481 by @xmath482 and a power series in @xmath483 @xmath484 then @xmath485 notice that @xmath486 is unitary , i.e. @xmath487 .
we have @xmath488 taking the laplace transform of @xmath12 @xmath489 s_{\
\beta_i}^{{{\hat{\underline \alpha}}}}(-z_i ) \frac{w^\alpha_{k}(y_i)}{\sqrt{-2}}\bigr).\end{aligned}\ ] ] using @xmath490 s_{\ \beta_i}^{{\hat{\underline \alpha}}}(-z_i))s^{\ \kappa({\mathcal{l}}_i)}_{{\hat{\underline \alpha}}}(z_i ) ( -z_i^{-k-1}))\bigr)\\ = & ( -1)^{g-1 } \sum_{\beta_i , a_i } \llangle \prod_{i=1}^n\tau_{a_i}(\phi_{\beta_i}(0 ) ) \rrangle_{g , n } \prod_{i=1}^n \chi^{\beta_i } ( \phi_{\beta_i}(0 ) , \kappa({\mathcal{l}}_i ) ) z_i^{-a_i-1}\\ = & ( -1)^{g-1 } \llangle \frac{\kappa({\mathcal{l}}_1)}{z_1-\psi_1},\dots,\frac{\kappa({\mathcal{l}}_n)}{z_n-\psi_n } \rrangle_{g , n}.\end{aligned}\ ] ]
in this section , we consider the non - equivariant limit @xmath15 . by
* section 1.3 ) , @xmath491 is uniquely determined by : 1 .
the recursive relation : @xmath492 $ ] .
2 . the homogeneity of @xmath342 : @xmath493 is a constant matrix .
the unique solution @xmath342 satisfying the above conditions was computed explicitly in @xcite : @xmath494 by proposition [ abr ] , @xmath495 . in this subsection
, we recover the above lemma by computing the stationary phase expansion of @xmath496 .
we assume @xmath497 , where @xmath21 .
@xmath498 let @xmath499 , then @xmath500 @xmath501 @xmath502 similarly ,
@xmath503 @xmath504 therefore , @xmath505 @xmath506 \left ( { \widetilde{s}}(z)e^{-u / z}\right ) \\ = & \begin{pmatrix } \frac{((2n-1)!!)^2}{\sqrt{2}n!2^{4n}q^{\frac{n}{2}+\frac{1}{4 } } } & \frac{\sqrt{-1}(-1)^{n+1}((2n-1)!!)^2}{\sqrt{2}n!2^{4n}q^{\frac{n}{2}+\frac{1}{4 } } } \\ \frac{((2n-1)!!)^2}{\sqrt{2}n!2^{4n}q^{\frac{n}{2}-\frac{1}{4 } } } -(\frac{n}{2}-\frac{1}{4})\frac{((2n-3)!!)^2}{\sqrt{2}(n-1 ) ! 2^{4n-4 } q^{\frac{n}{2}-\frac{1}{4 } } } & \frac{\sqrt{-1}(-1)^n(2n-1)!!)^2}{\sqrt{2}n !
2^{4n } q^{\frac{n}{2}-\frac{1}{4 } } } + ( \frac{n}{2}-\frac{1}{4})\frac{\sqrt{-1}(-1)^{n+1}((2n-3)!!)^2}{\sqrt{2}(n-1 ) !
2^{4n-4 } q^{\frac{n}{2}-\frac{1}{4 } } } \end{pmatrix},\end{aligned}\ ] ] @xmath507 in this subsection , we assume @xmath508
. then @xmath509 note that when @xmath510 is not an nonnegative integer , both hand sides are zero .
when @xmath53 , @xmath12 is holomorphic near @xmath315 , and one may expand it in the local holomorphic coordinate @xmath511 .
suppose that @xmath53 .
then near @xmath315 , @xmath12 has the following expansion @xmath512 the norbury - scott conjecture corresponds to the specialization @xmath513 , i.e. @xmath514 .
define @xmath515 by @xmath516 by theorem [ main ] , it suffices to show that @xmath515 agrees with the expansion of @xmath517 near @xmath315 in @xmath518 .
we now compute @xmath515 explicitly .
@xmath519 @xmath520 @xmath521 @xmath522 @xmath523 @xmath524 @xmath525 for @xmath158 , @xmath526 where @xmath527 @xmath528 recall that @xmath529 by and , to complete the proof , it remains to show that , @xmath530 agree with the expansion of @xmath451 near @xmath315 in @xmath531 .
assume that @xmath532 .
we have @xmath533 the @xmath11-th coefficient in the expansion of @xmath534 at @xmath315 is given by the residue @xmath535 @xmath536 which agrees with @xmath537 defined in .
@xmath538 @xmath539 which agrees with @xmath540 defined in .
in this section , we will specialize theorem [ main ] to the large radius limit case . in this case ,
theorem [ main ] relates the invariant @xmath12 of the limit curve to the equivariant descendent theory of @xmath26 . after expanding @xmath451 in suitable coordinates
, we can relate the corresponding expansion of @xmath12 to the generation function of hurwitz numbers and therefore reprove the bouchard - mario conjecture @xcite on hurwitz numbers .
let @xmath19 , @xmath20 and take the large radius limit @xmath24 .
then our mirror curve becomes @xmath541 when @xmath28 , this is just the lambert curve . recall that the two critical points @xmath542 of @xmath543 are @xmath544 since @xmath545 , @xmath546 under the limit @xmath24 . in other words , @xmath270 goes out of the curve under the limit @xmath24 and @xmath547 . as a result
, @xmath548 also turns to zero under the large radius limit . under the identification @xmath549 in theorem [ main ]
, we have @xmath550 when @xmath24 . on the a - model side , since @xmath412 , the @xmath551matrix @xmath552 is diagonal . therefore , we also have @xmath553 when @xmath24 under the identification in theorem [ main ] .
this means that in the localization graph of the equivariant gw invariants of @xmath0 , we can only have a constant map to @xmath554 . since @xmath555 and @xmath556
, we can not have any primary insertions .
therefore , in the large radius limit , we get @xmath557^{{\mathrm{vir } } } } \prod_{j=1}^n { \mathrm{ev}}_j^*((u_j)^2_{a_j}\phi_2(0 ) ) \psi_j^{a_j}\\ & = & \sum_{a_1,\ldots , a_n\in { \mathbb{z}}_{\geq 0 } } \frac{1}{-{\mathsf{w}}_1}\int_{{\overline{{\mathcal{m}}}}_{g , n}}\prod_{j=1}^n ( u_j)^2_{a_j } \psi_j^{a_j}\lambda_{g}^{\vee}(-{\mathsf{w}}_1),\end{aligned}\ ] ] where @xmath558 and @xmath559 is the @xmath60-th chern class of the hodge bundle . at the same time , we also have @xmath560 .
so @xmath561 .
therefore theorem [ main ] specializes to @xmath562 now we study the expansion of @xmath563 near the point @xmath315 in the coordinate @xmath564 .
we have @xmath565 since @xmath566 , by taking the differential we have @xmath567 therefore , @xmath568 . let @xmath569 near the point @xmath315
. then we have @xmath570 therefore @xmath571 on a - model side , let @xmath572 then @xmath573 by the elsv formula @xcite , @xmath574 so @xmath575 when @xmath28 , this is just the generating function of hurwitz numbers .
let @xmath576 be the expansion of @xmath577 in the coordinate @xmath578 near @xmath315 .
then we have for @xmath73 and @xmath53 , the invariant @xmath576 for the curve @xmath579 satisfies @xmath580 in particular , when @xmath28 , the right hand side is the generating function of hurwitz numbers and the bouchard - mario conjecture is recovered .
in this section , we give a brief review of bessel functions .
the bessel s differential equation is @xmath581 the _ bessel function of the first kind _ is defined by @xmath582 the _ bessel function of the second kind _ is defined by @xmath583 when @xmath11 is an integer , @xmath584 .
@xmath585 and @xmath586 form a basis of the 2-dimensional space of solutions to the bessel s differential equation . replacing @xmath290 by @xmath587 in
, one obtains the the modified bessel differential equation @xmath588 the _ modified bessel function of the first kind _ is defined by @xmath589 the _ modified bessel function of the second kind _ is defined by @xmath590 the following integral formulas are valid when @xmath591 : @xmath592 where @xmath593 is the real line with the standard orientation : ] @xmath594 where @xmath595 is the following contour : ] therefore , @xmath596 @xmath597 for any integers @xmath598 with @xmath599 , let @xmath400 be the following contour : ] [ integral - bessel ] for any @xmath600 such that @xmath399 , we have @xmath601 we observe that @xmath602 in particular , @xmath603 this proves in case @xmath604 . if @xmath605 then @xmath606 equations and imply @xmath607 equation follows from the above equation and , .
the equivariant quantum differential equation of @xmath0 is the vector equation @xmath608 which is equivalent to the following scalar equation : @xmath609 let @xmath610 then is equivalent to @xmath611 which is the modified bessel differential equation with @xmath612 . when @xmath613 , any solution to is of the form @xmath614 where @xmath615 , and @xmath616 are functions of @xmath617 .
v. bouchard , m. mario , `` hurwitz numbers , matrix models and enumerative geometry , '' _ from hodge theory to integrability and tqft @xmath618-geometry , _
263283 , proc .
pure math . , * 78 * , amer .
soc . , providence , ri , 2008 p.
dunin - barkowski , m. kazarian , n. orantin , s. shadrin , and l. spitz , `` polynomiality of hurwitz numbers , bouchard - mario conjecture , and a new proof of the elsv formula , '' adv .
( 2015 ) , 67103 .
p. dunin - barkowski , m. mulase , p. norbury , a. popolitov , and s. shadrin , `` quantum spectral curve for the gromov - witten theory of the complex projective line , '' arxiv:1312.5336 , to appear in j. reine angew .
p. dunin - barkowski , n. orantin , s. shadrin , and l. spitz , `` identification of the givental formula with the spectral curve topological recursion procedure , '' comm .
* 328 * ( 2014 ) , no .
2 , 669700 .
e. eynard , n. orantin , `` computation of open gromov - witten invariants for toric calabi - yau 3-folds by topological recursion , a proof of the bkmp conjecture , '' comm .
* 337 * ( 2015 ) , no .
2 , 483567 .
e. eynard , n. orantin , `` computation of open gromov - witten invariants for toric calabi - yau 3-folds by topological recursion , a proof of the bkmp conjecture , '' comm .
* 337 * ( 2015 ) , no .
2 , 483567 .
a. b. givental , `` elliptic gromov - witten invariants and the generalized mirror conjecture , '' _ integrable systems and algebraic geometry _ ( kobe / kyoto , 1997 ) , 107155 , world sci .
publ . , river edge , nj , 1998 . | we study the equivariantly perturbed mirror landau - ginzburg model of @xmath0 .
we show that the eynard - orantin recursion on this model encodes all genus all descendants equivariant gromov - witten invariants of @xmath0 .
the non - equivariant limit of this result is the norbury - scott conjecture @xcite , while by taking large radius limit we recover the bouchard - mario conjecture on simple hurwitz numbers @xcite . |
bayesian approaches have inherent advantages in solving inference and decision problems , but practical applications pose challenges for computation . as these challenges have been met bayesian approaches have proliferated and contributed to the solution of applied problems .
mcgrayne ( 2011 ) has recently conveyed these facts to a wide audience .
the evolution of bayesian computation over the past half - century has conformed with exponential increases in speed and decreases in the cost of computing .
the influence of computational considerations on algorithms , models , and the way that substantive problems are formulated for statistical inference can be subtle but is hard to over - state .
successful and innovative basic and applied research recognizes the characteristics of the tools of implementation from the outset and tailors approaches to those tools .
recent innovations in hardware ( graphics processing units , or gpus ) provide individual investigators with massively parallel desktop processing at reasonable cost . corresponding developments in software ( extensions of the c programming language and mathematical applications software ) make these attractive platforms for scientific computing .
this paper extends and applies existing sequential monte carlo ( smc ) methods for posterior simulation in this context .
the extensions fill gaps in the existing theory to provide a thorough and practical foundation for sequential posterior simulation , using approaches suited to massively parallel computing environments .
the application produces generic posterior simulators that make substantially fewer analytical and programming demands on investigators implementing new models and provides faster , more reliable and more complete posterior simulation than do existing methods inspired by conventional predominantly serial computational methods .
sequential posterior simulation grows out of smc methods developed over the past 20 years and applied primarily to state space models ( particle filters " ) .
seminal contributions include baker ( 1985 , 1987 ) , gordon et al .
( 1993 ) , kong et al .
( 1994 ) , liu and chen ( 1995 , 1998 ) , chopin ( 2002 , 2004 ) , del moral et al .
( 2006 ) , andrieu et al .
( 2010 ) , chopin and jacob ( 2010 ) and del moral et al .
( 2011 ) .
the posterior simulator proposed in this paper , which builds largely on chopin ( 2002 , 2004 ) , has attractive properties .
* it is highly generic ( easily adaptable to new models ) .
all that is required of the user is code to generate from the prior and evaluate the prior and data densities .
* it is computationally efficient relative to available alternatives .
this is in large part due to the specifics of our implementation of the simulator , which makes effective use of low - cost massively parallel computing hardware .
* since the simulator provides a sample from the posterior density at each observation date conditional on information available at that time , it is straightforward to compute moments of arbitrary functions of interest conditioning on relevant information sets .
marginal likelihood and predictive scores , which are key elements of bayesian analysis , are immediately available . also immediately available
is the probability integral transform at each observation date , which provides a powerful diagnostic tool for model exploration ( generalized residuals " ; e.g. , diebold et al . , 1998 ) .
* estimates of numerical standard error and relative numerical efficiency are provided as an intrinsic part of the simulator .
this is important but often neglected information for the researcher interested in generating dependable and reliable results . *
the simulator is robust to irregular posteriors ( e.g. , multimodality ) , as has been pointed out by jasra et al .
( 2007 ) and others . *
the simulator has well - founded convergence properties .
but , although the basic idea of sequential posterior simulation goes back at least to chopin ( 2002 , 2004 ) , and despite its considerable appeal , the idea has seen essentially no penetration in mainstream applied econometrics .
applications have been limited to relatively simple illustrative examples ( although this has begun to change very recently ; see herbst and schorfheide , 2012 ; fulop and li , 2012 ; chopin et al . , 2012 ) .
this is in stark contrast to applications of smc methods to state space filtering ( particle filters " ) , which have seen widespread use .
relative to markov chain monte carlo ( mcmc ) , which has become a mainstay of applied work , sequential posterior simulators are computationally costly when applied in a conventional serial computing environment .
our interest in sequential posterior simulation is largely motivated by the recent availability of low cost hardware supporting massively parallel computation .
smc methods are much better suited to this environment than is mcmc .
the massively parallel hardware device used in this paper is a commodity graphical processing unit ( gpu ) , which provides hundreds of cores at a cost of well under one dollar ( us ) each .
but in order to realize the huge potential gains in computing speed made possible by such hardware , algorithms that conform to the single - instruction multiple - data ( simd ) paradigm are needed .
the simulators presented in this paper conform to the simd paradigm by design and realize the attendant increases in computing speed in practical application .
but there are also some central issues regarding properties of sequential posterior simulators that have not been resolved in the literature .
since reliable applied work depends on the existence of solid theoretical underpinnings , we address these as well .
our main contributions are as follows . 1 .
_ theoretical basis .
_ whereas the theory for sequential posterior simulation as originally formulated by chopin ( 2002 , 2004 ) assumes that key elements of the algorithm are fixed and known in advance , practical applications demand algorithms that are adaptive , with these elements constructed based on the information provided by particles generated in the course of running the algorithm . +
while some progress has been made in developing a theory that applies to such adaptive algorithms , ( e.g. , douc and moulines , 2008 ) , we expect that a solid theory that applies to the kind of algorithms that are used and needed in practice is likely to remain unattainable in the near future .
+ in section [ sec : algorithms ] we provide an approach that addresses this problem in a different way , providing a posterior simulator that is highly adapted to the models and data at hand , while satisfying the relatively straightforward conditions elaborated in the original work of chopin ( 2002 , 2004 ) .
numerical accuracy .
_ while chopin ( 2004 ) provides a critical central limit theorem , the extant literature does not provide any means of estimating the variance , which is essential for assessing the numerical accuracy and relative numerical efficiency of moment approximations ( geweke , 1989 ) .
this problem has proved difficult in the case of mcmc ( flegal and jones , 2010 ) and appears to have been largely ignored in the smc literature . in section [ sec : parpostsims ] we propose an approach to resolving this issue which entails no additional computational effort and is natural in the context of the massively parallel environment as well as key in making efficient use of it .
the idea relies critically upon the theoretical contribution noted above .
marginal likelihood .
_ while the approach to assessing the asymptotic variance of moment approximations noted in the previous point is useful , it does not apply directly to marginal likelihood , which is a critical element of bayesian analysis .
we address this issue in section [ sec : pml ] .
parallel implementation .
_ it has been noted that sequential posterior simulation is highly amenable to parallelization going back to at least chopin ( 2002 ) , and there has been some work toward exploring parallel implementations ( e.g. , lee et al . , 2010 ; fulop and li , 2012 ) .
building on this work , we provide a software package that includes a full gpu implementation of the simulator , with specific details of the algorithm tuned to the massively parallel environment as outlined in section [ sec : algorithms ] and elsewhere in this paper .
the software is fully object - oriented , highly modular , and easily extensible .
new models are easily added , providing the full benefits of gpu computing with little effort on the part of the user .
geweke et al .
( 2013 ) provides an example of such an implementation for the logit model . in future work
, we intend to build a library of models and illustrative applications utilizing this framework which will be freely available .
section [ ss : software ] provides details about this software .
5 . _ specific recommendations . _
the real test of the simulator is in its application to problems that are characteristic of the scale and complexity of serious disciplinary work . in section [ sec : application ] , we provide one such application to illustrate key ideas . in other ongoing work , including geweke et al .
( 2013 ) and a longer working paper written in the process of this research ( durham and geweke , 2011 ) , we provide several additional substantive applications . as part of this work ,
we have come up with some specific recommendations for aspects of the algorithm that we have found to work well in a wide range of practical applications .
these are described in section [ ss : details ] and implemented fully in the software package we are making available with this paper .
the sequential simulator proposed in this paper is based on ideas that go back to chopin ( 2002 , 2004 ) , with even earlier antecedents including gilkes and berzuini ( 2001 ) , and fearnhead ( 1998 ) .
the simulator begins with a sample of parameter vectors ( particles " ) from the prior distribution .
data is introduced in batches , with the steps involved in processing a single batch of data referred to as a cycle . at the beginning of each cycle , the particles represent a sample from the posterior conditional on information available up to that observation date . as data
is introduced sequentially , the posterior is updated using importance sampling ( kloek and van dijk , 1978 ) , the appropriately weighted particles representing an importance sample from the posterior at each step . as more data is introduced , the importance weights tend to become unbalanced " ( a few particles have most of the weight , while many others have little weight ) , and importance sampling becomes increasingly inefficient .
when some threshold is reached , importance sampling stops and the cycle comes to an end . at this point ,
a resampling step is undertaken , wherein particles are independently resampled in proportion to their importance weights .
after this step , there will be many copies of particles with high weights , while particles with low weights will tend to drop out of the sample .
finally , a sequence of metropolis steps is undertaken in order to rejuvenate the diversity of the particle sample . at the conclusion of the cycle ,
the collection of particles once again represents a sample from the posterior , now incorporating the information accrued from the data newly introduced .
this sequential introduction of data is natural in a time - series setting , but also applicable to cross - sectional data . in the latter case ,
the sequential ordering of the data is arbitrary , though some orderings may be more useful than others .
much of the appeal of this simulator is due to its ammenability to implementation using massively parallel hardware .
each particle can be handled in a distinct thread , with all threads updated concurrently at each step .
communication costs are low . in applications where computational cost is an important factor ,
nearly all of the cost is incurred in evaluating densities of data conditional on a candidate parameter vector and thus isolated to individual threads .
communication costs in this case are a nearly negligible fraction of the total computational burden .
the key to efficient utilization of the massively parallel hardware is that the workload be divided over many simd threads . for the gpu hardware used in this paper optimal usage
involves tens of thousands of threads . in our implementation
, particles are organized in a relatively small number of groups each with a relatively large number of particles ( in the application in section [ sec : application ] there are @xmath0 groups of @xmath1 particles each ) .
this organization of particles in groups is fundamental to the supporting theory .
estimates of numerical error and relative numerical efficiency are generated as an intrinsic part of the algorithm with no additional computational cost , while the reliability of these estimates is supported by a sound theoretical foundation .
this structure is natural in a massively parallel computing environment , as well critical in making efficient use of it .
the remainder of this section provides a more detailed discussion of key issues involved in posterior simulation in a massively parallel environment .
it begins in section [ subsec : compenv ] with a discussion of the relevant features of the hardware and software used .
section [ subsec : notation_assumption ] sets up a generic model for bayesian inference along with conditions used in deriving the analytical properties of various sequential posterior simulators in section [ sec : algorithms ] .
section [ sec : parpostsims ] stipulates a key convergence condition for posterior simulators , and then shows that if this condition is met there are attractive generic methods for approximating the standard error of numerical approximation in massively parallel computing environments .
section [ sec : algorithms ] then develops sequential posterior simulators that satisfy this condition . the particular device that motivates this work is the graphics processing unit ( gpu ) . as a practical matter several gpu s can be incorporated in a single server ( the host " ) with no significant complications , and desktop computers that can accommodate up to eight gpu s are readily available .
the single- and multiple - gpu environments are equivalent for our purposes .
a single gpu consists of several multiprocessors , each with several cores .
the gpu has global memory shared by its multiprocessors , typically one to several gigabytes ( gb ) in size , and local memory specific to each multiprocessor , typically on the order of 50 to 100 kilobytes ( kb ) per multiprocessor .
( for example , this research uses a single nvidia gtx 570 gpu with 15 multiprocessors , each with 32 cores .
the gpu has 1.36 gb of local memory , and each multiprocessor has 49 kb of memory and 32 kb of registers shared by its cores . )
the bus that transfers data between gpu global and local memory is significantly faster than the bus that transfers data between host and device , and accessing local memory on the multiprocessor is faster yet . for greater technical detail on gpu s ,
see hendeby et al .
( 2010 ) , lee et al .
( 2010 ) and souchard et al .
( 2010 ) .
this hardware has become attractive for scientific computing with the extension of scientific programming languages to allocate the execution of instructions between host and device and facilitate data transfer between them . of these
the most significant has been the compute unified device architecture ( cuda ) extension of the c programming language ( nvidia , 2013 ) .
cuda abstracts the host - device communication in a way that is convenient to the programmer yet faithful to the aspects of the hardware important for writing efficient code .
code executed on the device is contained in special functions called kernels that are invoked by the host code .
specific cuda instructions move the data on which the code operates from host to device memory and instruct the device to organize the execution of the code into a certain number of blocks with a certain number of threads each .
the allocation into blocks and threads is the virtual analogue of the organization of a gpu into multiprocessors and cores . while the most flexible way to develop applications that make use of gpu parallelization is through c / c++ code with direct calls to the vendor - supplied interface functions , it is also possible to work at a higher level of abstraction .
for example , a growing number of mathematical libraries have been ported to gpu hardware ( e.g. , innovative computing laboratory , 2013 ) .
such libraries are easily called from standard scientific programming languages and can yield substantial increases in performance for some applications .
in addition , matlab ( 2013 ) provides a library of kernels , interfaces for calling user - written kernels , and functions for host - device data transfer from within the matlab workspace .
we augment standard notation for data , parameters and models .
the relevant observable random vectors are @xmath2 @xmath3 and @xmath4 denotes the collection @xmath5 .
the observation of @xmath2 is @xmath6 , @xmath7 denotes the collection @xmath8 , and therefore @xmath9 denotes the data .
this notation assumes ordered observations , which is natural for time series .
if @xmath10 is independent and identically distributed the ordering is arbitrary .
a model for bayesian inference specifies a @xmath11 unobservable parameter vector @xmath12 and a conditional density @xmath13 with respect to an appropriate measure for @xmath14 .
the model also specifies a prior density @xmath15 with respect to a measure @xmath16 on @xmath17 .
the posterior density @xmath18 follows in the usual way from and @xmath19 .
the objects of bayesian inference can often be written as posterior moments of the form @xmath20 $ ] , and going forward we use @xmath21 to refer to such a generic function of interest .
evaluation of @xmath21 may require simulation , e.g. @xmath22 $ ] , and conventional approaches based on the posterior simulation sample of parameters ( e.g. geweke , 2005 , section 1.4 ) apply in this case .
the marginal likelihood @xmath23 famously does not take this form , and section [ sec : pml ] takes up its approximation with sequential posterior simulators and , more specifically , in the context of the methods developed in this paper .
several conditions come into play in the balance of the paper .
[ cond : prior_evaluate](prior distribution ) .
the model specifies a proper prior distribution .
the prior density kernel can be evaluated with simd - compatible code .
simulation from the prior distribution must be practical but need not be simd - compatible .
it is well understood that a model must take a stance on the distribution of outcomes @xmath14 _ a priori _ if it is to have any standing in formal bayesian model comparison , and that this requires a proper prior distribution .
this requirement is fundamentally related to the generic structure of sequential posterior simulators , including those developed in section [ sec : algorithms ] , because they require a distribution of @xmath24 before the first observation is introduced .
given a proper prior distribution the evaluation and simulation conditions are weak .
simulation from the prior typically involves minimal computational cost and thus we do not require it to be simd - compatible .
however , in practice it will often be so .
[ cond : lf_evaluate](likelihood function evaluation ) the sequence of conditional densities @xmath25 can be evaluated with simd - compatible code for all @xmath12 . evaluation with simd - compatible code is important to computational efficiency because evaluation of ( [ model_conditional_pdf ] ) constitutes almost all of the floating point operations in typical applications for the algorithms developed in this paper .
condition [ cond : lf_evaluate ] excludes situations in which unbiased simulation of ( [ model_conditional_pdf ] ) is possible but exact evaluation is not , which is often the case in nonlinear state space models , samples with missing data , and in general any case in which a closed form analytical expression for ( [ model_conditional_pdf ] ) is not available .
a subsequent paper will take up this extension .
[ cond : bounded_like ] ( bounded likelihood ) the data density @xmath26 is bounded above by @xmath27 for all @xmath28 .
this is one of two sufficient conditions for the central limit theorem invoked in section [ subsec : nonadaptive ] .
it is commonly but not universally satisfied .
when it is not , it can often be attained by minor and innocuous modification of the likelihood function ; section [ sec : application ] provides an example .
[ cond : prior_var](existence of prior moments ) if the algorithm is used to approximate @xmath29 $ ] , then @xmath30<\infty$ ] for some @xmath31 . in any careful implementation of posterior simulation
the existence of relevant posterior moments must be verified analytically .
this condition , together with condition [ cond : bounded_like ] , is sufficient for the existence of @xmath32 $ ] and @xmath33 $ ] .
condition [ cond : prior_var ] also comes into play in establishing a central limit theorem .
consider the implementation of any posterior simulator in a parallel computing environment like the one described in section [ subsec : compenv ] .
following the conventional approach , we focus on the posterior simulation approximation of @xmath34 . \label{post_moment}\ ] ] the posterior simulator operates on parameter vectors , or particles , @xmath35 organized in @xmath36 groups of @xmath37 particles each ; define @xmath38 and @xmath39 .
this organization is fundamental to the rest of the paper .
each particle is associated with a core of the device .
the cuda software described in section [ subsec : compenv ] copes with the case in which @xmath40 exceeds the number of cores in an efficient and transparent fashion .
ideally the posterior simulator is simd - compatible , with identical instructions executed on all particles .
some algorithms , like the sequential posterior simulators developed in section [ sec : algorithms ] , come quite close to this ideal .
others , for example gibbs samplers with metropolis steps , may not .
the theory in this section applies regardless .
however , the advantages of implementing a posterior simulator in a parallel computing environment are driven by the extent to which the algorithm is simd - compatible .
denote the evaluation of the function of interest at particle @xmath35 by @xmath41 and within - group means by @xmath42 @xmath43 .
the posterior simulation approximation of @xmath44 is the grand mean @xmath45 in general we seek posterior simulators with the following properties .
[ cond : normal_approx](asymptotic normality of posterior moment approximation ) the random variables @xmath46 are independently and identically distributed .
there exists @xmath47 for which @xmath48 as @xmath49 .
going forward , @xmath50 will denote a generic variance in a central limit theorem .
convergence is part of the rigorous foundation of posterior simulators , e.g. geweke ( 1989 ) for importance sampling , tierney ( 1994 ) for mcmc , and chopin ( 2004 ) for a sequential posterior simulator . for importance sampling
@xmath51 @xmath52 , where @xmath53 is the ratio of the posterior density kernel to the source density kernel . for to be of any practical use in assessing numerical accuracy there must also be a simulation - consistent approximation of @xmath50 .
such approximations are immediate for importance sampling . they have proven more elusive for mcmc ; e.g. , see flegal and jones ( 2010 ) for discussion and an important contribution . to our knowledge
there is no demonstrated simulation - consistent approximation of @xmath50 for sequential posterior simulators in the existing literature .
however , for any simulator satisfying condition [ cond : normal_approx ] this is straightforward .
given condition [ cond : normal_approx ] , it is immediate that @xmath54 define the estimated posterior simulation variance @xmath55 \sum_{j=1}^{j}\left ( \overline{g}_{j}^{n}-\overline{g}^{\left ( j , n\right ) } \right ) ^{2}\label{vhat}\ ] ] and the numerical standard error ( _ nse _ ) @xmath56 ^{1/2}=\left\ { \left [ j\left ( j-1\right ) \right ] ^{-1}\sum_{j=1}^{j}\left ( \overline{g}_{j}^{n}-\overline{g}^{\left ( j , n\right ) } \right)^2 \right\ } ^{1/2}. \label{nse_def}\ ] ] note the different scaling conventions in and : in the scaling is selected so that @xmath57 approximates @xmath50 in because we will use this expression mainly for methodological developments ; in the scaling is selected to make it easy to appraise the reliability of numerical approximations of posterior moments like those reported in section [ sec : application ] .
one conventional assessment of the efficiency of a posterior simulator is its relative numerical efficiency ( _ rne _ ) ( geweke , 1989 ) , @xmath58 where @xmath59 , the simulation approximation of @xmath60 $ ] .
[ prop : clt_var_approx]condition [ cond : normal_approx ] implies @xmath61 and @xmath62 ^{1/2}\overset{d}{\rightarrow } t\left ( j-1\right ) , \label{p1b}\ ] ] both as @xmath63 .
conditions [ cond : prior_evaluate ] through [ cond : prior_var ] imply @xmath64 = var\left [ \overline{g}^{\left ( j , n\right ) } \right ] .
\label{p1c}\ ] ] from condition [ cond : normal_approx ] , @xmath65 as @xmath63 . substituting in yields .
condition [ cond : normal_approx ] also implies @xmath66 as @xmath63 . since and are independent in the limiting distribution
, we have .
conditions [ cond : prior_evaluate ] through [ cond : prior_var ] imply the existence of the first two moments of @xmath67 . then follows from the fact that the approximations @xmath68 are independent and identically distributed across @xmath69 .
we seek a posterior simulator that is generic , requiring little or no intervention by the investigator in adapting it to new models beyond the provision of the software implicit in conditions [ cond : prior_evaluate ] and [ cond : lf_evaluate ] .
it should reliably assess the numerical accuracy of posterior moment approximations , and this should be achieved in substantially less execution time than would be required using the same or an alternative posterior simulator for the same model in a serial computing environment .
such simulators are necessarily adaptive : they must use the features of the evolving posterior distribution , revealed in the particles @xmath70 , to design the next steps in the algorithm .
this practical requirement has presented a methodological conundrum in the sequential monte carlo literature , because the mathematical complications introduced by even mild adaptation lead to analytically intractable situations in which demonstration of the central limit theorem in condition [ cond : normal_approx ] is precluded .
this section solves this problem and sets forth a particular adaptive sequential posterior simulator that has been successful in applications we have undertaken with a wide variety of models .
section [ sec : application ] details one such application .
section [ subsec : nonadaptive ] places a nonadaptive sequential posterior simulator ( chopin , 2004 ) that satisfies condition [ cond : normal_approx ] into the context developed in the previous section .
section [ subsec : adaptive ] introduces a technique that overcomes the analytical difficulties associated with adaptive simulators .
it is generic , simple and imposes little additional overhead in a parallel computing environment .
section [ ss : details ] provides details on a particular variant of the algorithm that we have found to be successful for a variety of models representative of empirical research frontiers in economics and finance .
section [ ss : software ] discusses a software package implementing the algorithm that we are making available .
we rely on the following mild generalization of the sequential monte carlo algorithm of chopin ( 2004 ) , cast in the parallel computing environment detailed in the previous section . in this algorithm
, the observation dates at which each cycle terminates ( @xmath71 ) and the parameters involved in specifying the metropolis updates ( @xmath72 ) are assumed to be fixed and known in advance , in conformance with the conditions specified by chopin ( 2004 ) .
[ alg : general_nonadaptive](nonadaptive ) let @xmath73 be fixed integers with @xmath74 and let @xmath75 be fixed vectors . 1 . initialize @xmath76 and
let @xmath77 .
2 . for @xmath78 1 .
correction @xmath79 phase : 1 .
2 . for @xmath81@xmath82 3 .
selection @xmath84 phase , applied independently to each group @xmath85 : using multinomial or residual sampling based on @xmath86 , select @xmath87 3 .
mutation @xmath88 phase , applied independently across @xmath89 : @xmath90 where the drawings are independent and the p.d.f . satisfies the invariance condition @xmath91 3 .
@xmath92 the algorithm is nonadaptive because @xmath93 and @xmath94 are predetermined before the algorithm starts .
going forward it will be convenient to denote the cycle indices by @xmath95 . at the conclusion of the algorithm , the simulation approximation of a generic posterior moment is .
[ prop_chopin ] if conditions [ cond : prior_evaluate ] through [ cond : prior_var ] are satisfied then algorithm [ alg : general_nonadaptive ] satisfies condition [ cond : normal_approx ] .
the results follow from chopin ( 2004 ) , theorem 1 ( for multinomial resampling ) and theorem 2 ( for residual resampling ) .
the assumptions made in theorems 1 and 2 are 1 .
@xmath96 and @xmath97 @xmath98 , 2 . the functions @xmath99 are integrable on @xmath100 , and 3 .
the moments @xmath101 exist .
assumption 1 is merely a change in notation ; conditions [ cond : prior_evaluate ] through [ cond : bounded_like ] imply assumption 2 ; and conditions [ cond : prior_evaluate ] through [ cond : prior_var ] imply assumption 3 .
theorems 1 and 2 in chopin ( 2004 ) are stated using the weighted sample at the end of the last @xmath102 phase , but as that paper states , they also apply to the unweighted sample at the end of the following @xmath103 phase . at the conclusion of each cycle @xmath104 , the @xmath36 groups of @xmath37 particles each , @xmath105 @xmath106 , are mutually independent because the @xmath107 phase is executed independently in each group . at the end of each cycle
@xmath104 , all particles @xmath108 are identically distributed with common density @xmath109 . particles in different groups are independent , but particles within the same group are not .
the amount of dependence within groups depends upon how well the @xmath103 phases succeed in rejuvenating particle diversity .
in fact the sequences @xmath110 and @xmath111 for which the algorithm is sufficiently efficient for actual application are specific to each problem . as a practical matter
these sequences must be tailored to the problem based on the characteristics of the particles @xmath112 produced by the algorithm itself .
this leads to algorithms of the following type .
[ alg : general_adaptive](adaptive ) algorithm [ alg : general_adaptive ] is the following generalization of algorithm [ alg : general_nonadaptive ] . in each cycle
@xmath104 , 1 .
[ alg : general_adaptive_cond1]@xmath113 may be random and depend on @xmath114 ; 2 .
[ alg : general_adaptive_cond2]the value of @xmath115 may be random and depend on @xmath116 . while such algorithms can be effective and are what is needed for practical applied work , the theoretical foundation relevant to algorithm [ alg : general_nonadaptive ] does not apply .
the groundwork for algorithm [ alg : general_nonadaptive ] was laid by chopin ( 2004 ) . with respect to the first condition of algorithm [ alg : general_adaptive ] , in chopin s development each cycle
consists of a single observation , while in practice cycle lengths are often adaptive based on the effective sample size criterion ( liu and chen , 1995 ) .
progress has been made toward demonstrating condition [ cond : normal_approx ] in this case only recently ( douc and moulines , 2008 ; del moral et al . , 2011 ) , and it appears that conditions [ cond : prior_evaluate ] through [ cond : prior_var ] are sufficient for the assumptions made in douc and moulines ( 2008 ) . with respect to the second condition of algorithm [ alg : general_adaptive ] , demonstration of condition [ cond : normal_approx ] for the adaptations that are necessary to render sequential posterior simulators even minimally efficient appears to be well beyond current capabilities .
there are many specific adaptive algorithms , like the one described in section [ ss : details ] below , that will prove attractive to practitioners even in the absence of sound theory for appraising the accuracy of posterior moment approximations . in some cases condition [ cond :
normal_approx ] will hold ; in others , it will not but the effects will be harmless ; and in still others condition [ cond : normal_approx ] will not hold and the relation of to will be an open question .
investigators with compelling scientific questions are unlikely to forego attractive posterior simulators awaiting the completion of their proper mathematical foundations .
the following algorithm resolves this dilemma , providing a basis for implementations that are both computationally effective and supported by established theory .
[ alg : general_hybrid](hybrid ) 1 .
[ alg : general_hybrid_step1]execute algorithm [ alg : general_adaptive ] .
retain @xmath117 @xmath118 and discard @xmath119 @xmath120 .
[ alg : general_hybrid_step3]execute algorithm [ alg : general_nonadaptive ] using the retained @xmath117 but with a new seed for the random number generator used in the @xmath107 and @xmath103 phases . [ prop : hybrid ] if conditions [ cond : prior_evaluate ] through [ cond : prior_var ] are satisfied then algorithm [ alg : general_hybrid ] satisfies condition [ cond : normal_approx ] . because the sequences @xmath121 and @xmath122 produced in step [ alg : general_hybrid_step1 ] are fixed ( predetermined ) with respect to the random particles @xmath123 generated in step [ alg : general_hybrid_step3 ] , the conditions of algorithm [ alg : general_nonadaptive ] are satisfied in step [ alg : general_hybrid_step3 ] . from proposition
[ prop_chopin ] , condition [ cond : normal_approx ] is therefore satisfied .
applications of bayesian inference are most often directed by investigators who are not specialists in simulation methodology .
a generic version of algorithm [ alg : general_adaptive ] that works well for wide classes of existing and prospective models and data sets with minimal intervention by investigators is therefore attractive . while general descriptions of the algorithm such as provided in sections [ subsec : nonadaptive][subsec : adaptive ] give a useful basis for discussion , what is really needed for practical use is a detailed specification of the actual steps that need to be executed .
this section provides the complete specification of a particular variant of the algorithm that has worked well across a spectrum of models and data sets characteristic of current research in finance and economics .
the software package that we are making available with this paper provides a full implementation of the algorithm presented here ( see section [ ss : software ] ) .
c phase termination : : the @xmath102 phase is terminated based on a rule assessing the effective sample size ( kong et al . , 1994 ; liu and chen , 1995 ) . following each update step @xmath124 as described in algorithm 1 , part 2(a)ii , compute @xmath125 ^{2}}{\sum_{j=1}^{j}\sum_{n=1}^{n}w_{jn}\left ( s\right ) ^{2}}.\ ] ] it is convenient to work with the relative sample size , @xmath126 , where @xmath40 is the total number of particles . terminate the @xmath102 phase and proceed to the @xmath107 phase if @xmath127 .
+ we have found that the specific threshold value used has little effect on the performance of the algorithm .
higher thresholds imply that the @xmath102 phase is terminated earlier , but fewer metropolis steps are needed to obtain adequate diversification in the @xmath103 phase ( and inversely for lower thresholds ) .
the default threshold in the software is 0.5 .
this can be easily modified by users , but we have found little reason to do so .
resampling method : : we make several resampling methods available .
the results of chopin ( 2004 ) apply to multinomial and residual resampling ( baker , 1985 , 1987 ; liu and chen , 1998 ) .
stratified ( kitagawa , 1996 ) and systematic ( carpenter et al . , 1999 )
samplers are also of potential interest .
we use residual sampling as the default .
it is substantially more efficient than multinomial at little additional computational cost .
stratified and systematic resamplers are yet more efficient as well as being less costly and simpler to implement ; however , we are not aware of any available theory supporting their use and so recommend them only for experimental use at this point .
metropolis steps : : the default is to use a sequence of gaussian random walk samplers operating on the entire parameter vector in a single block . at iteration @xmath128 of the @xmath103 phase in cycle @xmath104 , the variance of the sampler is obtained by @xmath129 where @xmath130 is the sample variance of the current population of particles and @xmath131 is a stepsize " scaling factor .
the stepsize is initialized at 0.5 and incremented depending upon the metroplis acceptance rate at each iteration .
using a target acceptance rate of 0.25 ( gelman et al . , 1996 ) , the stepsize is incremented by 0.1 if the acceptance rate is greater than the target and decremented by 0.1 otherwise , respecting the constraint @xmath132 .
+ an alternative that has appeared often in the literature is to use an independent gaussian sampler
. this can give good performance for relatively simple models , and is thus useful for illustrative purposes using implementations that are less computationally efficient .
but for more demanding applications , such as those that are likely to be encountered in practical work , the independent sampler lacks robustness .
if the posterior is not well - represented by the gaussian sampler , the process of particle diversification can be very slow .
the difficulties are amplified in the case of high - dimensional parameter vectors .
and the sampler fails spectacularly in the presence of irregular posteriors ( e.g. , multimodality ) .
+ the random walk sampler gives up some performance in the case of near gaussian posteriors ( though it still performs well in this situation ) , but is robust to irregular priors , a feature that we consider to be important .
the independent sampler is available in the software for users wishing to experiment .
other alternatives for the metropolis updates are also possible , and we intend to investigate some of these in future work .
m phase termination : : the simplest rule would be to terminate the @xmath103 phase after some fixed number of metropolis iterations , @xmath133 . but in practice , it sometimes occurs that the @xmath102 phase terminates with a very low @xmath134 . in this case
, the @xmath103 phase begins with relatively few distinct particles and we would like to use more iterations in order to better diversify the population of particles .
a better rule is to end the @xmath103 phase after @xmath135 iterations if @xmath136 and after @xmath137 iterations if @xmath138 ( the default settings in the software are @xmath139 , @xmath140 , @xmath141 and @xmath142 ) .
we refer to this as the deterministic stopping rule . "
+ but , the optimal settings , especially for @xmath143 , are highly application dependent . setting @xmath143 too low results in poor rejuvenation of particle diversity and the _ rne _ associated with moments of functions of interest will typically be low .
this problem can be addressed by either using more particles or a higher setting for @xmath143 .
typically , the latter course of action is more efficient if _ rne _ drops below about 0.25 ; that is , finesse ( better rejuvenation ) is generally preferred over brute force ( simply adding more particles ) . on the other hand ,
there is little point in continuing @xmath103 phase iterations once _ rne _ gets close to one ( indicating that particle diversification is already good , with little dependence amongst the particles in each group ) .
continuing past this point is simply a waste of computing resources .
+ _ rne _ thus provides a useful diagnostic of particle diversity .
but it also provides the basis for a useful alternative termination criterion .
the idea is to terminate @xmath103 phase iterations when the average _ rne _ of numerical approximations of selected test moments exceeds some threshold .
we use a default threshold of @xmath144 .
it is often useful to compute moments of functions of interest at particular observation dates . while it is possible to do this directly using weighted samples obtained within the @xmath102 phase , it is more efficient to terminate the @xmath102 phase and execute @xmath107 and @xmath103 phases at these dates prior to evaluating moments . in such cases , it is useful to target a higher _ rne _ threshold to assure more accurate approximations .
we use a default of @xmath145 here .
we also set an upper bound for the number of metropolis iterations , at which point the @xmath103 phase terminates regardless of _
the default is @xmath146 .
+ while the deterministic rule is simpler , the _ rne_-based rule has important advantages : it is likely to be more efficient than the deterministic rule in practice ; and it is fully automated , thus eliminating the need for user intervention .
some experimental results illustrating these issues are provided in section [ sec : application ] .
+ an important point to note here is that after the @xmath107 phase has executed , _
no longer provides a useful as a measure of the diversity of the particle population ( sincle the particles are equally - weighted ) .
_ rne _ is the only useful measure of which we are aware .
parameterization : : bounded support of some parameters can impede the efficiency of the gaussian random walk in the @xmath103 phase , which has unbounded support .
we have found that simple transformations , for example a positive parameter @xmath147 to @xmath148 or a parameter @xmath149 defined on @xmath150 to @xmath151 , can greatly improve performance .
this requires attention to the jacobian of the transformation in the prior distribution , but it is often just as appealing simply to place the prior distribution directly on the transformed parameters .
section 5.1 provides an example .
the posterior simulator described in this section is adaptive ( and thus a special case of algorithm [ alg : general_adaptive ] ) .
that is , key elements of the simulator are not known in advance , but are determined on the basis of particles generated in the course of executing it . in order to use the simulator in the context of the hybrid method ( algorithm [ alg : general_hybrid ] ) , these design elements must be saved for use in a second run . for the simulator described here
, we need the observations @xmath152 at which each @xmath102 phase termination occurs , the variance matrices @xmath153 used in the metropolis updates , and the number of metropolis iterations executed in each @xmath103 phase @xmath154 . the hybrid method then involves running the simulator a second time , generating new particles but using the design elements saved from the first run .
since the design elements are now fixed and known in advance , this is a special case of algorithm [ alg : general_nonadaptive ] and thus step 2 of algorithm [ alg : general_hybrid ] . and
therefore , in particular , condition [ cond : normal_approx ] is satisfied and the results of section [ sec : parpostsims ] apply .
the software package that we are making available with this paper ( http://www.quantosanalytics.org/garland/mp-sps_1.1.zip ) implements the algorithm developed in section [ ss : details ] .
the various settings described there are provided as defaults but can be easily changed by users .
switches are available allowing the user to choose whether the algorithm runs entirely on the cpu or using gpu acceleration , and whether to run the algorithm in adaptive or hybrid mode .
the software provides log marginal likelihood ( and log score if a burn - in period is specified ) together with estimates of nse , as described in section [ subsec : ml ] .
posterior mean and standard deviation of functions of interest at specified dates are provided along with corresponding estimates of rne and nse , as described in section [ sec : parpostsims ] .
rss at each @xmath102 phase update and rne of specified test moments at each @xmath103 phase iteration are available for diagnostic purposes , and plots of these are generated as part of the standard output
. for transparency and accessibility to economists doing practical applied work , the software uses a matlab shell ( with extensive use of the parallel / gpu toolbox ) , with calls to functions coded in c / cuda for the computationally intensive parts . for simple problems where computational cost is low to begin with ,
the overhead associated with using matlab is noticeable .
but , for problems where computational cost is actually important , the cost is nearly entirely concentrated in c / cuda code .
very little efficiency is lost by using a matlab shell , while the gain in transparency is substantial .
the software is fully object - oriented , highly modular , and easily extensible .
it is straightforward for users to write plug - ins implementing new models , providing access to the full benefits of gpu acceleration with little programming effort .
all that is needed is a matlab class with methods that evaluate the data density and moments of any functions of interest .
the requisite interfaces are all clearly specified by abstract classes . to use a model for a particular application , data and a description of the prior must be supplied as inputs to the class constructor .
classes implementing simulation from and evaluation of some common priors are included with the software . extending the package to implement additional prior specifications
is simply a matter of providing classes with this functionality .
the software includes several models ( including the application described in section [ sec : application ] of this paper ) , and we are in the process of extending the library of available models and illustrative applications using this framework in ongoing work .
geweke et al .
( 2013 ) provides an example of one such implementation for the logit model .
marginal likelihoods are fundamental to the bayesian comparison , averaging and testing of models .
sequential posterior simulators provide simulation - consistent approximations of marginal likelihoods as a by - product almost no additional computation is required .
this happens because the marginal likelihood is the product of predictive likelihoods over the sample , and the @xmath102 phase approximates these terms in the same way that classical importance sampling can be used to approximate marginal likelihood ( kloek and van dijk , 1978 ; geweke , 2005 , section 8.2.2 ) . by contrast
there is no such generic approach with mcmc posterior simulators . the results for posterior moment approximation apply only indirectly to marginal likelihood approximation .
this section describes a practical adaptation of the results to this case , and a theory of approximation that falls somewhat short of condition [ cond : normal_approx ] for posterior moments .
any predictive likelihood can be cast as a posterior moment by defining @xmath155 then @xmath156 : = \overline{g}\left(t , s\right ) .
\label{pl__formula}\ ] ] [ prop_pl]if conditions [ cond : prior_evaluate ] , [ cond : lf_evaluate ] and [ cond : bounded_like ] are satisfied , then in algorithms [ alg : general_nonadaptive ] and [ alg : general_hybrid ] condition [ cond : normal_approx ] is satisfied for the function of interest ( [ pl_func ] ) and the sample @xmath157 .
it suffices to note that conditions [ cond : prior_evaluate ] through [ cond : bounded_like ] imply that condition [ cond : prior_var ] is satisfied for .
the utility of proposition [ prop_pl ] stems from the fact that simulation approximations of many moments ( [ pl__formula ] ) are computed as by - products in algorithm [ alg : general_nonadaptive ] .
specifically , in the @xmath102 phase at for @xmath158 and @xmath159 @xmath160 from proposition [ prop_pl ] applied over the sample @xmath161 implies for @xmath158 and @xmath162 @xmath163 algorithms [ alg : general_nonadaptive ] and [ alg : general_hybrid ] therefore provide weakly consistent approximations to expressed in terms of the form .
the logarithm of this approximation is a weakly consistent approximation of the more commonly reported log predictive likelihood .
the numerical standard error of @xmath164 is given by , and the delta method provides the numerical standard error for the log predictive likelihood .
the values of @xmath124 for which this result is useful are limited because as @xmath124 increases , @xmath165 becomes increasingly concentrated and the computational efficiency of the approximations @xmath166 of @xmath167 declines . indeed ,
this is why particle renewal ( in the form of @xmath107 and @xmath103 phases ) is undertaken when the effective sample size drops below a specified threshold .
the sequential posterior simulator provides access to the posterior distribution at each observation in the @xmath102 phase using the particles with the weights computed at . if at this point one executes the auxiliary simulations @xmath168
@xmath169 then a simulation - consistent approximation of the cumulative distribution of a function @xmath170 , evaluated at the observed value @xmath171 , is @xmath172 } \left [ f\left ( y_{sjn}^{\left ( \ell -1\right ) } \right ) \right ] } { \sum_{j=1}^{j}\sum_{n=1}^{n}w_{jn}\left ( s-1\right)}.\ ] ] these evaluations are the essential element of a probability integral transform test of model specification ( rosenblatt , 1952 ; smith , 1985 ; diebold et al . , 1998
; berkowitz , 2001 ; geweke and amisano , 2010 ) .
thus by accessing the weights @xmath173 from the @xmath102 phase of the algorithm , the investigator can compute simulation consistent approximations of any set of predictive likelihoods ( 18 ) and can execute probability integral transform tests for any function of @xmath174 to develop a theory of sequential posterior simulation approximation to the marginal likelihood , some extension of the notation is useful . denote @xmath175 and then @xmath176 it is also useful to introduce the following condition , which describes an ideal situation that is not likely to be attained in practical applications but is useful for expositional purposes .
[ cond : mphase]in the mutation phase of algorithm [ alg : general_nonadaptive ] @xmath177 with the addition of condition [ cond : mphase ] it would follow that @xmath178 $ ] in condition [ cond : normal_approx ] , and computed values of relative numerical efficiencies would be about 1 for all functions of interest @xmath179
. in general condition [ cond : mphase ] is unattainable in any interesting application of an sequential posterior simulator , for if it were the posterior distribution could be sampled by direct monte carlo .
[ prop_ml]if conditions [ cond : prior_evaluate ] , [ cond : lf_evaluate ] and [ cond : bounded_like ] are satisfied then in algorithms [ alg : general_nonadaptive ] and [ alg : general_hybrid ] @xmath180 as @xmath63 , and @xmath181 ^{-1}\sum_{j=1}^{j } \left ( \overline{w}_{j}^{n}-\overline{w}^{\left ( j , n\right ) } \right ) ^{2}\right\ } = \mathrm{var}\left ( \overline{w}^{\left ( j , n\right ) } \right ) .
\label{prop_4a}\ ] ] if , in addition , condition [ cond : mphase ] is satisfied then @xmath182 \overset{d}{\rightarrow } n\left ( 0,v\right ) .
\label{prop_4b}\ ] ] from proposition [ prop_pl ] , condition [ cond : normal_approx ] is satisfied for @xmath183 .
therefore @xmath184 as @xmath63 @xmath185 .
the result follows from the decomposition @xmath186 condition [ cond : bounded_like ] implies that moments of @xmath187 of all orders exist . then follows from the mutual independence of @xmath188 .
result follows from the mutual independence and asymptotic normality of the @xmath189 terms @xmath190 , and @xmath50 follows from the usual asymptotic expansion of the product in @xmath187 .
note that without condition [ cond : mphase ] , @xmath191and @xmath192 are not independent and so condition [ cond : normal_approx ] does not provide an applicable central limit theorem for the product @xmath193 ( although condition [ cond : normal_approx ] does hold for each @xmath194 @xmath195 individually , as shown in section [ subsec : pl ] ) .
let @xmath196 denote the term in braces in .
proposition [ prop_ml ] motivates the working approximations @xmath197,\notag \\ & & \log \left ( \overline{w}^{\left ( j , n\right ) } \right ) \overset{\cdot } { \thicksim } n\left [ \log p\left ( y_{1:t}\right ) , \frac{{\mathrm{v}\widehat{\mathrm{a}}\mathrm{r}}\left ( \overline{w}^{\left ( j , n\right ) } \right ) } { \left ( \overline{w}^{(j , n)}\right ) ^{2}}\right ] .
\label{ml_se}\end{aligned}\ ] ] standard expansions of @xmath198 and @xmath199 suggest the same asymptotic distribution , and therefore comparison of these values in relation to the working standard error from ( [ ml_se ] ) provides one indication of the adequacy of the asymptotic approximation .
condition [ cond : mphase ] suggests that these normal approximations will be more reliable as more metropolis iterations are undertaken in the algorithm detailed in section [ ss : details ] , and more generally , the more effective is the particle diversity generated by the @xmath103 phase .
the predictive distributions of returns to financial assets are central to the pricing of their derivatives like futures contracts and options .
the literature modeling asset return sequences as stochastic processes is enormous and has been a focus and motivation for bayesian modelling in general and application of sequential monte carlo ( smc ) methods in particular .
one of these models is the exponential generalized autoregressive conditional heteroskedasticity ( egarch ) model introduced by nelson ( 1991 ) .
the example in this section works with a family of extensions developed in durham and geweke ( 2013 ) that is highly competitive with many stochastic volatility models . in the context of this paper
the egarch model is also of interest because its likelihood function is relatively intractable .
the volatility in the model is the sum of several factors that are exchangeable in the posterior distribution .
the return innovation is a mixture of normal distributions that are also exchangeable in the posterior distribution .
both features are essential to the superior performance of the model ( durham and geweke , 2013 ) .
permutations in the ordering of factors and mixture components induce multimodal distributions in larger samples .
models with these characteristics have been widely used as a drilling ground to assess the performance of simulation approaches to bayesian inference with ill - conditioned posterior distributions ( e.g. , jasra et al . , 2007 ) .
the models studied in this section have up to @xmath200 permutations , and potentially as many local modes . although it would be possible to optimize the algorithm for these characteristics , we intentionally make no efforts to do so . nonetheless , the irregularity of the posteriors turns out to pose no difficulties for the algorithm .
most important , in our view , this example illustrates the potential large savings in development time and intellectual energy afforded by the algorithm presented in this paper compared with other approaches that might be taken .
we believe that other existing approaches , including importance sampling and conventional variants on markov chain monte carlo ( mcmc ) , would be substantially more difficult . at
the very least they would require experimentation with tuning parameters by bayesian statisticians with particular skills in these numerical methods , even after using the approach of geweke ( 2007 ) to deal with dimensions of posterior intractability driven by exchangeable parameters in the posterior distribution .
the algorithm replaces this effort with a systematic updating of the posterior density , thereby releasing the time of investigators for more productive and substantive efforts . an egarch model for a sequence of asset returns @xmath201 has the form @xmath202 the return disturbance term @xmath203 is distributed as a mixture of @xmath204 normal distributions , @xmath205where @xmath206 is the gaussian density with mean @xmath207 and variance @xmath208 , @xmath209 @xmath210 and @xmath211 .
the parameters of the model are identified by the conditions @xmath212 and @xmath213 ; equivalently , @xmath214 the models are indexed by @xmath215 , the number of volatility factors , and @xmath204 , the number of components in the return disturbance normal mixture , and we refer to the specification ( [ egarch_mod1])([egarch_mod2 ] ) as ` egarch_ki ` .
the original form of the egarch model ( nelson , 1991 ) is with @xmath216 .
all parameters have gaussian priors with means and standard deviations indicated below ( the prior distribution of @xmath217 is truncated below at @xmath218 ) . indices @xmath219 and @xmath220 take on the values @xmath221 and @xmath222 . .parameters and prior distributions for the egarch models [ cols="^,^,^,^",options="header " , ]
[ t : rbar ] table [ t : rbar ] shows the outcome of an exercise exploring the relationship between the number of metropolis iterations undertaken in the @xmath103 phases and the performance of the algorithm .
the first 8 rows of the table use the deterministic @xmath103 phase stopping rule with @xmath143 = 5 , 8 , 13 , 21 , 34 , 55 , 89 and 144 .
the last row of the table , indicated by ` * ' in the @xmath143 field , uses the _
rne_-based stopping rule with @xmath144 , @xmath145 and @xmath223 .
the last column of the table reports a measure of computational performance : precision ( i.e. , @xmath224 ) normalized by computational cost ( time in seconds ) .
the ratio of precision to time is the relevant criterion for comparing two means of increasing numerical accuracy : more steps in the @xmath103 phase versus more particles .
the ratio of precision to time is constant in the latter strategy .
( in fact , on a gpu , it increases up to a point because the gpu works more efficiently with more threads , but the application here with @xmath225 particles achieves full efficiency . ) therefore adding more steps in the @xmath103 phase is the more efficient strategy so long as the ratio of precision to time continues to increase . table [ t : rbar ]
shows that in this example , adding iterations is dominant at @xmath226 and is likely dominant at @xmath227 .
ideally , one would select @xmath143 to maximize the performance measure reported in the last column of the table .
rne_-based stopping rule , shown in the last line of the table , does a good job of automatically picking an appropriate stopping point without requiring any user input or experimentation .
the number of metropolis iterations varies from one @xmath103 phase to the next with this stopping rule , averaging just over 100 in this application .
although the rne - based rule uses fewer metropolis steps overall relative to @xmath227 , total execution time is greater .
this is because the metropolis iterations are more concentrated toward the end of the sample period with the _ rne_-based rule , where they are more computationally costly .
the fact that so many metroplis iterations are needed to get good particle diversity is symptomatic of the extremely irregular posterior densities implied by this model ( we return to this issue in section [ ss : irregular ] ) . for models with posterior densities that are closer to gaussian
, many fewer metropolis iterations will typically be needed .
models for asset returns , like the egarch models considered here , are primarily of interest for their predictive distributions .
we illustrate this application using the three functions of interest @xmath228 introduced in section [ subsec : egarch_performance ] .
moments are evaluated by monte carlo approximation over the posterior distribution of @xmath229 using the particles obtained at time @xmath230 , as described in section [ sec : parpostsims ] .
the last observation of the sample , march 31 , 2010 in this application ( @xmath231 ) , is typically of interest . for illustration , we specify the same date one year earlier , march 31 , 2009 ( @xmath232 ) , as an additional observation of interest .
volatility is much higher on the earlier date than it is on the later date . at each date of interest ,
the @xmath102 phase is terminated ( regardless of the _ rss _ ) and @xmath107 and @xmath103 phases are executed . ; ( c ) @xmath226 ; ( d ) @xmath227 . ]
figure [ f : rne ] shows _ rne _ for all three test functions @xmath233 at each metropolis iteration of each cycle @xmath104 .
the figure reports results for four different @xmath103 phase stopping rules : the _ rne_-based rule ( with @xmath144 , @xmath145 and @xmath223 ) and the deterministic rule with @xmath143 = 34 , 89 and 144 .
the beginning of each @xmath103 phase is indicated by a vertical line .
the lower axis of each panel indicates the metropolis iteration ( cumulating across cycles ) ; the top axis indicates the observation date @xmath230 at which each @xmath103 phase takes place ; and the left axis indicates _
rne_. for reference , figure [ f : rne ] includes a horizontal line indicating @xmath234 .
this is the default target for the _ rne_-based stopping rule and serves as a convenient benchmark for the other rules as well . in the early part of the sample ,
the deterministic rules execute many more metropolis steps than needed to achieve the nominal target of @xmath234 . however , these require relatively little time because sample size @xmath230 is small .
as noted above , there is little point in undertaking additional metropolis iterations once _ rne _ approaches one , as happens toward the beginning of the sample for all three deterministic rules shown in the figure . toward the end of the sample , achieving any fixed _
target in the @xmath103 phase requires more iterations due to the extreme non - gaussianity of the posterior ( see section [ ss : irregular ] for a more detailed discussion of this issue ) .
rne_-based rule adapts automatically , performing iterations only as needed to meet the _ rne _ target , implying more iterations as sample size increases in this application . at observation @xmath235 ,
november 15 , 1991 , the @xmath102 phase terminates with very low @xmath134 ( regardless of @xmath103 phase stopping rule used ) , the result of a return that is highly unlikely conditional on the model and past history of returns .
the deterministic rules undertake additional metropolis iterations to compensate , as detailed in section 3.3 .
rne_-based rule also requires more iterations than usual to meet the relevant threshold ; but in this case the number of additional iterations undertaken is determined algorithmically .
cccccccccccccccccccc & & & & & & + @xmath143 & compute & e & sd & nse & rne & & e & sd & nse & rne & & e & sd & nse & rne + & time & & & & & & & & & & & & & & + + 5 & 91 & -37.405 & 0.373 & 0.027 & 0.003 & & 95.877 & 7.254 & 0.513 & 0.003 & & -1.999 & 0.823 & 0.063 & 0.003 + 8 & 127 & -37.388 & 0.390 & 0.029 & 0.003 & & 96.457 & 7.543 & 0.531 & 0.003 & & -2.204 & 0.837 & 0.060 & 0.003 + 13 & 205 & -37.347 & 0.366 & 0.020 & 0.005 & & 97.311 & 7.167 & 0.361 & 0.006 & & -2.348 & 0.774 & 0.034 & 0.008 + 21 & 304 & -37.382 & 0.352 & 0.014 & 0.009 & & 96.767 & 6.938 & 0.255 & 0.011 & & -2.530 & 0.788 & 0.030 & 0.011 + 34 & 482 & -37.339 & 0.349 & 0.011 & 0.015 & & 97.582 & 6.913 & 0.211 & 0.016 & & -2.569 & 0.812 & 0.021 & 0.023 + 55 & 776 & -37.340 & 0.353 & 0.007 & 0.034 & & 97.591 & 6.970 & 0.142 & 0.037 & & -2.563 & 0.811 & 0.011 & 0.079 + 89 & 1245 & -37.330 & 0.364 & 0.006 & 0.055 & & 97.735 & 7.133 & 0.109 & 0.065 & & -2.574 & 0.812 & 0.010 & 0.105 + 144 & 2120 & -37.332 & 0.355 & 0.004 & 0.141 & & 97.743 & 6.996 & 0.066 & 0.172 & & -2.580 & 0.816 & 0.007 & 0.230 + * & 2329 & -37.334 & 0.359 & 0.003 & 0.170 & & 97.687 & 7.037 & 0.063 & 0.192 & & -2.587 & 0.818 & 0.005 & 0.406 + + 5 & 91 & -50.563 & 0.309 & 0.014 & 0.007 & & 0.596 & 0.253 & 0.013 & 0.006 & & -2.078 & 0.816 & 0.063 & 0.003 + 8 & 127 & -50.514 & 0.309 & 0.011 & 0.012 & & 0.662 & 0.276 & 0.013 & 0.007 & & -2.274 & 0.838 & 0.059 & 0.003 + 13 & 205 & -50.517 & 0.309 & 0.009 & 0.016 & & 0.696 & 0.284 & 0.010 & 0.013 & & -2.425 & 0.771 & 0.033 & 0.008 + 21 & 304 & -50.512 & 0.310 & 0.007 & 0.031 & & 0.736 & 0.295 & 0.008 & 0.021 & & -2.599 & 0.780 & 0.028 & 0.012 + 34 & 482 & -50.492 & 0.309 & 0.006 & 0.044 & & 0.754 & 0.301 & 0.006 & 0.045 & & -2.616 & 0.795 & 0.018 & 0.030 + 55 & 776 & -50.491 & 0.310 & 0.004 & 0.097 & & 0.755 & 0.302 & 0.004 & 0.103 & & -2.610 & 0.797 & 0.010 & 0.093 + 89 & 1245 & -50.482 & 0.315 & 0.003 & 0.164 & & 0.765 & 0.308 & 0.003 & 0.162 & & -2.640 & 0.794 & 0.008 & 0.145 + 144 & 2120 & -50.483 & 0.314 & 0.002 & 0.381 & & 0.765 & 0.308 & 0.002 & 0.240 & & -2.632 & 0.800 & 0.006 & 0.276 + * & 2329 & -50.482 & 0.315 & 0.002 & 0.460 & & 0.769 & 0.307 & 0.002 & 0.588 & & -2.643 & 0.796 & 0.005 & 0.452 + table [ t : moments ] reports details on the posterior mean approximations for the two dates of interest .
total computation time for running the simulator across the full sample is provided for reference . the last line in each panel , indicated by `
* ' in the @xmath143 field , uses the _
rne_-based stopping rule .
for both dates , the _ nse _ of the approximation declines substantially as @xmath143 increases .
comparison of the compute times reported in table [ t : moments ] again suggests that increasing @xmath143 is more efficient for reducing _
nse_than would be increasing @xmath37 , up to at least @xmath226 .
some bias in the moment approximations is evident with low values of @xmath143 .
the issue is most apparent for the 3% loss probability on march 31 , 2010 .
since volatility is low on that date , the probability of realizing a 3% loss is tiny and arises from tails of the posterior distribution of @xmath229 , which is poorly represented in small samples . for example , with @xmath236 , _ rne _ is 0.013 , implying that each group of size @xmath237 has an effective sample size of only about 13 particles .
there is no evidence of bias for @xmath238 or with the _ rne_-based rule .
plots such as those shown in figure [ f : rne ] provide useful diagnostics and are provided as a standard output of the software .
for example , it is easy to see that with @xmath239 ( panel ( b ) of the figure ) not enough iterations are performed in the @xmath103 phases , resulting in low _ rne _ toward the end of the sample . with lower values of @xmath143
the degradation in performance is yet more dramatic . as @xmath143
is increased to 89 and 144 in panels ( c ) and ( d ) of the figure , respectively , the algorithm is better able to maintain particle diversity through the entire sample .
rne_-based rule does a good job at choosing an appropriate number of iterations in each @xmath103 phase and does so without the need for user input or experimentation .
in the ` egarch_23 ` model there are 2 permutations of the factors @xmath240 and 6 permutations of the components of the normal mixture probability distribution function of @xmath241 .
this presents a severe challenge for single - chain mcmc as discussed by celeux et al .
( 2000 ) and jasra et al .
( 2007 ) , and for similar reasons importance sampling is also problematic .
the problem can be mitigated ( frhwirth - schnatter , 2001 ) or avoided entirely ( geweke , 2007 ) by exploiting the special mirror image structure of the posterior distribution .
but these models are still interesting as representatives of multimodal and ill - behaved posterior distributions in the context of generic posterior simulators .
we focus here on the 6 permutations of the normal mixture in ` egarch_23 ` .
consider a @xmath242 parameter subvector @xmath243 with three distinct values of the triplets @xmath244 @xmath245 .
there are six distinct ways in which these values could be assigned to components @xmath246 of the normal mixture in the ` egarch_23 ` model .
these permutations define six points @xmath247 @xmath248 . for all sample sizes
@xmath230 , the posterior densities @xmath249 at these six points are identical .
let @xmath250 be a different parameter vector with analogous permutations @xmath251 @xmath252 . as the sample adds evidence @xmath253 or @xmath254 .
thus , a specific triplet set of triplets @xmath255 @xmath256 and its permutations will emerge as pseudo - true values of the parameters ( geweke , 2005 , section 3.4 ) . from selected posterior distributions conditional on @xmath157 ( @xmath257 , @xmath258 , @xmath259 , @xmath260 , @xmath239 ) .
] the marginal distributions will exhibit these properties as well .
consider the pair @xmath261 , which is the case portrayed in figure [ egarch_multimodal_sigma ] .
the scatterplot is symmetric about the axis @xmath262 in all cases . as sample size
@xmath230 increases six distinct and symmetric modes in the distribution gradually emerge .
these reflect the full marginal posterior distribution for the normal mixture components of the ` egarch_23 ` model ( i.e. , marginalizing on all other parameters ) that is roughly centered on the components @xmath263 , @xmath264 and @xmath265 .
the progressive decrease in entropy with increasing @xmath230 illustrates how the algorithm copes with ill - behaved posterior distributions .
particles gradually migrate toward concentrations governed by the evidence in the sample .
unlike mcmc there is no need for particles to migrate between modes , and unlike importance sampling there is no need to sample over regions eliminated by the data ( on the one hand ) or to attempt to construct multimodal source distributions ( on the other ) .
similar phenomena are also evident for the other mixture parameters as well as for the parameters of the garch factors .
the algorithm proposed in this paper adapts to these situations without specific intervention on a case - by - case basis . to benchmark the performance of the algorithm against a more conventional approach
, we constructed a straightforward metropolis random walk mcmc algorithm , implemented in c code on a recent vintage cpu using the ` egarch_23 ` model .
the variance matrix was tuned manually based on preliminary simulations , which required several hours of investigator time and computing time .
the algorithm required 12,947 seconds for 500,000 iterations .
the numerical approximation of the moment of interest @xmath266 $ ] for march 31 , 2010 , the same one addressed in table [ t : moments ] , produced the result @xmath267 and a _ nse_of @xmath268 . taking the square of _ nse _ to be inversely proportional to the number of mcmc iterations , an _
nse _ of @xmath269 ( the result for the _ rne_-based stopping rule in table [ t : moments ] ) would require about 1,840,000 iterations and 47,600 seconds computing time .
thus posterior simulation for the full sample would require about 20 times as long using random walk metropolis in a conventional serial computing environment .
although straightforward , this analysis severely understates the advantage of the sequential posterior simulator developed in this paper relative to conventional approaches .
the mcmc simulator does not traverse all six mirror images of the posterior density .
this fact greatly complicates attempts to recover marginal likelihood from the mcmc simulator output ; see celeux et al .
( 2000 ) . to our knowledge
the only reliable way to attack this problem using mcmc is to compute predictive likelihoods from the prior distribution and the posterior densities @xmath270 @xmath271 . recalling that @xmath272 , and taking computation time to be proportional to sample size ( actually , it is somewhat greater ) yields a time requirement of @xmath273 seconds ( 2.28 cpu years ) , which is almost 100,000 times as long as was required in the algorithm and implementation used here .
unless the function of interest @xmath274 is invariant to label switching ( geweke , 2007 ) , the mcmc simulator must traverse all the mirror images with nearly equal frequency .
as argued in celeux et al .
( 2000 ) , for all practical purposes this condition can not be met with any reliability even with a simulator executed for several cpu centuries .
this example shows that simple speedup factors may not be useful in quantifying the reduction in computing time afforded by massively parallel computing environments . for the scope of models set in this work that is , those satisfying conditions [ cond : prior_evaluate ] through [ cond : prior_var]sequential posterior simulation is much faster than posterior simulation in conventional serial computing environments
this is a lower bound .
there are a number of routine and reasonable objectives of posterior simulation , like those described in this section , that simply can not be achieved at all with serial computing but are fast and effortless with smc .
most important , sequential posterior simulation in a massively parallel computing environment conserves the time , energy and talents of the investigator for more substantive tasks .
recent innovations in parallel computing hardware and associated software provide the opportunity for dramatic increases in the speed and accuracy of posterior simulation .
widely used mcmc simulators are not generically well - suited to this environment , whereas alternative approaches like importance sampling are .
the sequential posterior simulator developed here has attractive properties in this context : inherent adaptability to new models ; computational efficiency relative to alternatives ; accurate approximation of marginal and predictive likelihoods ; reliable and well - grounded measures of numerical accuracy ; robustness to irregular posteriors ; and a well - developed theoretical foundation . establishing these properties
required a number of contributions to the literature , summarized in section 1 and then developed in sections 2 , 3 and 4 .
section 5 provided an application to a state - of - the - art model illustrating the properties of the simulator .
the methods set forth in the paper reduce computing time dramatically in a parallel computing environment that is well within the means of academic investigators .
relevant comparisons with conventional serial computing environments entail different algorithms , one for each environment .
moreover , the same approximation may be relatively more advantageous in one environment than the other .
this precludes generic conclusions about speed - up factors . in the example in section [ sec : application ] , for a simple posterior moment ,
the algorithm detailed in section [ ss : details ] was nearly 10 times faster than a competing random - walk metropolis simulator in a conventional serial computing environment . for predictive likelihoods and marginal likelihood it was 100,000 times faster .
the parallel computations used a single graphics processing unit ( gpu ) . up to eight gpu s can be added to a standard desktop computer at a cost ranging from about $ 350 ( us ) for a mid - range card to about $ 2000 ( us ) for a high - performance tesla card .
computation time is inversely proportional to the number of gpu s .
these contributions address initial items on the research agenda opened up by the prospect of massively parallel computing environments for posterior simulation . in the near term it should be possible to improve on the specific contribution made here in the form of the algorithm detailed in section [ ss : details ] .
looking forward on the research agenda , a major component is extending the class of models for which generic sequential posterior simulation will prove practical and reliable .
large - scale hierarchical models , longitudinal data , and conditional data density evaluations that must be simulated all pose fertile ground for future work .
the research reported here has been guided by two paramount objectives .
one is to provide methods that are generic and minimize the demand for knowledge and effort on the part of applied statisticians who , in turn , seek primarily to develop and apply new models .
the other is to provide methods that have a firm methodological foundation whose relevance is borne out in subsequent applications .
it seems to us important that these objectives should be central in the research agenda .
baker je ( 1985 ) .
adaptive selection methods for genetic algorithms . in grefenstette j ( ed . ) , proceedings of the international conference on genetic algorithms and their applications , 101111 .
malwah nj : erlbaum .
chopin n , jacob p ( 2010 ) .
free energy sequential monte carlo , application to mixture modelling . in : bernardo jm , bayarri mj , berger jo , dawid ap , heckerman d , smith afm , west m ( eds . ) , bayesian statistics 9 .
oxford : oxford university press .
chopin n , jacob pi , papaspiliopoulis o ( 2011 ) .
smc@xmath275 : a sequential monte carlo algorithm with particle markov chain monte carlo updates . working paper .
geweke j , durham g , xu h ( 2013 ) .
bayesian inference for logistic regression models using sequential posterior simulation .
lee l , yau c , giles mb , doucet a , homes cc ( 2010 ) . on the utility of graphics cards to perform massively parallel simulation of advanced monte carlo methods .
journal of computational and graphical statistics 19 : 769789 .
mcgrayne shb ( 2011 ) .
the theory that would not die : how bayes rule cracked the enigma code , hunted down russian submarines , and emerged triumphant from two centuries of controversy .
new haven : yale university press .
souchard ma , wang q , chan c , frelinger j , cron a , west m ( 2010 ) .
understanding gpu programming for statistical computation : studies in massively parallel massive mixtures .
journal of computational graphics and statistics 19 : 419438 . | massively parallel desktop computing capabilities now well within the reach of individual academics modify the environment for posterior simulation in fundamental and potentially quite advantageous ways . but to fully exploit these benefits algorithms that conform to parallel computing environments are needed .
sequential monte carlo comes very close to this ideal whereas other approaches like markov chain monte carlo do not .
this paper presents a sequential posterior simulator well suited to this computing environment .
the simulator makes fewer analytical and programming demands on investigators , and is faster , more reliable and more complete than conventional posterior simulators .
the paper extends existing sequential monte carlo methods and theory to provide a thorough and practical foundation for sequential posterior simulation that is well suited to massively parallel computing environments .
it provides detailed recommendations on implementation , yielding an algorithm that requires only code for simulation from the prior and evaluation of prior and data densities and works well in a variety of applications representative of serious empirical work in economics and finance .
the algorithm is robust to pathological posterior distributions , generates accurate marginal likelihood approximations , and provides estimates of numerical standard error and relative numerical efficiency intrinsically .
the paper concludes with an application that illustrates the potential of these simulators for applied bayesian inference .
keywords : graphics processing unit ; particle filter ; posterior simulation ; sequential monte carlo ; single instruction multiple data jel classification : primary , c11 ; secondary , c630 . |
as light passes an object , it diffracts around it and spreads out by an angle @xmath0 , so if the angle that the object subtends , @xmath1 , is less than this , the object `` disappears '' .
furthermore , if the object has mass , light that passes close - by is delayed by travelling through its potential well . the figure of merit for whether diffraction is important in gravitational lensing is @xmath2 where @xmath3 is the schwarzschild radius of the lens . for values of @xmath4
much less than unity , diffraction dominates and the peak magnification is greatly diminished , for large values of @xmath4 the diffraction fringes are closely spaced , so either the finite size of the source or the finite bandwidth of the observations can easily smear them out yielding the magnification that one expects from geometric optics .
the path length from the star at @xmath5 to the telescope through the point @xmath6 in the lens plane is @xmath7 the magnification of the star is related to the sum of all the phases from the light taking various paths , @xmath8 the schwarzschild radius of the earth is about one centimeter . beyond @xmath9 au
lensing dominates over occultation
. therefore , let s take @xmath10 .
if we take the integral over the amplitudes to be @xmath11 ( @xmath12 ) , @xmath13 the magnification increases with the value of @xmath4 with the peak magnification given by @xmath14 .
the left panel of figure [ fig : finite_sources ] depicts the magnification as function of the impact parameter of the source relative to the lens in units of the einstein radius and the value of @xmath4 .
averaging over bandwidth or over a finite source brings the oscillations toward the geometric .
however , if @xmath15 there is little magnification . for large values of @xmath4 , the size of the oscillation scales as @xmath16 the right panel of figure [ fig : finite_sources ]
gives a estimate of the radio flux density ( red ) and amplitude of its oscillation ( blue ) at 10 ghz for stars in the bulge field 1 of the ogle - ii survey ( szymanski 2005 ; udalski , kubiak & szymanski 1997 ) .
the fluxes are normalized using the values for arcturus at 1 kpc ( consult heyl ( 2010b ) for details ) .
there is a second complementary signature ( heyl 2010c ) .
as the lens passes in front of the source , its position on the sky appears to move .
the key idea to figure how where the image appears is that the gradient of the phase of the wave as measured at the telescope points toward the image .
furthermore , the situation is symmetric about the lens , so moving the telescope to the right is like moving the source to the left , and we can take the gradient of the phase with respect to the position of the source instead at the telescope , yielding the centroid of the image @xmath17\ ] ] figure [ fig : astrometry ] shows that lensing moves the centroid radially outward : the closer the passage , the greater the deviation .
minima in the displacement correspond to maxima of the magnification , and the displacement reaches the blue curve @xmath18 at minima of the magnification .
for distant asteroids @xmath19 ( heyl 2010a ) . if they are massive , lensing is also important .
if lensing is weak , the fresnel pattern gets lensed
. if lensing is strong , things are more complex .
the ska could detect oscillations from earth - mass lenses as far as 10 pc against bulge giants .
heyl , j. , 2010a , _ mnras _ , * 402 * , l39-l43 ( arxiv:0910.3922 ) . + , 2010b , _ mnras _ , * 411 * , 1780 - 1786 ( arxiv:1002.3007 ) . + , 2010c , _ mnras _ , * 411 * , 1787 - 1791 ( arxiv:1003.0250 ) .
+ schneider , p. , ehlers , j. , falco , e. 1992 , _ gravitational lenses _ , springer : berlin . + szymanski m. k. , 2005 , _ acta astron .
_ , 55 , 43 . + udalski a. , kubiak m. , szymanski m. , 1997 , _ acta astron .
_ , 47 , 319 . | diffraction is important when nearby substellar objects gravitationally lens distant stars .
if the wavelength of the observation is comparable to the schwarzschild radius of lensing object , diffraction leaves an observable imprint on the lensing signature .
the ska may have sufficient sensitivity to detect the typical sources , giant stars in the bulge .
the diffractive signatures in a lensing event break the degeneracies between the mass of the lens , its distance and proper motion . |
most of the us national power grid has ultimately been shaped and built during the quick urbanization period of the beginning of 20@xmath0 century .
increasing power demand continues to put enormous strain on the infrastructure of the aging grid , forcing utilities to maximally utilize its existing resources . as a consequence ,
many parts of the grid often operate in nearly critical regimes with significantly increased probability of large scale failures @xcite . in the modern world
it becomes crucially important to be able to read the state of the pre - critical smart grid in a timely and precise manner , to effectively control parameters of smart grids operating in pre - critical regimes minimizing the probability of their large scale failures .
unfortunately , commonly used used power system state estimation and control algorithms , working extremely well for stable operating regimes , become less efficient ( see for example @xcite ) when the operating point of the grid is close to the threshold of instability . as we shall discuss below
, a partial reason for the decrease of efficiency of state estimation procedures is related to a strong amplification of fluctuations of the power system state vector closer to the threshold of instability .
the present contribution addresses the problem of stochastic fluctuations of power system state vector close to voltage or phase - angle instabilities .
the amplification of stochastic fluctuations of the system state vector in pre - critical operating regimes is also directly related to the phenomenon of critical slowing - down often observed in power grids on the brink of failure @xcite .
naturally , detecting the onset of critical slowing - down by itself can provide an efficient way for early detection of approach to a large - scale instability @xcite .
as will be explained below , the onset of critical slowing - down is naturally associated with a strong growth of fluctuations of the system state vector .
this growth can in turn be locally identified well befiore the event of collapse from local synchrophasor measurements of voltage phase and magnitude on a given node of the smart grid .
the contributions of our work can be summarized as follows .
we provide a formal mathematical description of critical slowing - down phenomenon , and characterize it in terms of power flow jacobian as well as generator inertia and damping matrices .
second , we derive a closed form equation for the autocorrelation and autostructure functions as well as power spectral density of state vector that completely characterizes the probabilities of arbitrary system trajectories . remarkably , the autocorrelation and autostructure functions are expressed in terms of steady - state power flow jacobian and its eigenvectors . finally , we validate all our results with numerical simulation of ieee 39 and ieee 57 test systems .
to describe dynamics of system variables close to the threshold of instability , we use the structure - preserving model @xcite which reduces to the system of coupled swing equations on @xmath1 ( generator ) nodes of the power grid @xmath2 and power flow equations on @xmath3 ( consumer ) nodes of the grid @xmath4
@xmath5 @xmath6 @xmath7 here @xmath8 is a voltage phase on a bus @xmath9 , @xmath10 is inertia constant of a generator on the node @xmath9 , parameters @xmath11 describe droop controls of generators and/or load dependence on frequency fluctuations @xmath12 on the @xmath3 nodes , @xmath13 is a voltage magnitude on a bus @xmath9 ( for the @xmath1 nodes , @xmath14 ) , parameters @xmath15 encode dependence of the power load on the voltage magnitude @xmath13 .
finally , coefficients @xmath16 describe a ( weak ) dependence of the power load on the rate @xmath17 of the voltage change with time .
the power losses are important for the dynamics of system state variables , thus generally @xmath18 .
we consider the load model ( [ eq : p ] ) , ( [ eq : q ] ) with such parameters chosen that @xmath19 for simplicity , the choice of @xmath11 corresponds to @xmath20% change of the load per @xmath21% change in system frequency @xmath22 and @xmath23% change in power generation per @xmath21% change in @xmath22 @xcite , while parameters @xmath16 to @xmath24 change in power load per @xmath25 change in @xmath26 . for simplicity we assume , that all the loads have fixed power factors @xmath27 , and all fluctuate with time in the vicinity of the average ( base ) load @xmath28 , constantly deviating from the base value and returning back , so that @xmath29 @xmath30 @xcite .
the same applies to generated power , especially if the grid is exposed to intermittence of renewable energy sources .
thus , @xmath31 generally behave as stochastic processes @xcite .
the base load value @xmath32 itself changes with time but relatively slowly , with significant changes only noticeable at time scales of several hours . in principle , a given , aggregated load @xmath33 can be thought of as an aggregation of a huge number of power - consuming devices connected to the node @xmath9 of the grid , which get online and offline , connected and disconnected from the grid .
a resulting overall time profile of the aggregated active load @xmath33 is represented on the fig .
[ fig : load - time - profile ] .
the characteristic time scale @xmath34 of step - like changes depicted there is @xmath35 sec .
and the absolute value of its fourier transform .
[ fig : load - time - profile ] ] these relatively fast power load fluctuations on different nodes of the grid are statistically independent from each other , and their correlation properties of the vectors @xmath36 , @xmath37 can be considered gaussian @xcite .
the latter have mean zero and are thus completely characterized by the expectation values @xmath38 while @xmath39 and @xmath40 here @xmath41 is a diagonal covariance matrix .
note that generally @xmath42 alternatively , the fluctuations can be represented with the power spectral density @xmath43 , which is simply a fourier transform of the autocorrelation function ( [ eq : pcorr ] ) , see for example @xcite . here and in the forthcoming paragraphs the averaging corresponds to summation over possible realization of the load fluctuations . for stationary stochastic processes , the same results could be recovered by averaging over long time periods @xmath44 . whenever the random load profile is composed of a continuing sequence of independent individual `` jumps '' , the power spectral density @xmath45 behaves like @xmath46 , where @xmath47 is the characteristic time of power load correlations .
we assume that this time is much longer than the typical timescales of @xmath48 that are considered in this work , but smaller than the scale of several hours . on the time
scales correspoding to large enough frequencies , such that @xmath49 , one approximately has @xmath50 .
the same approximation has been also proposed and justified in @xcite .
importantly , this model is different from the ones considered in a number of other studies that rely on a more traditional `` white noise '' model @xcite .
the latter implies that the correlation properties of the fluctuating power loads are described by @xmath51 , independent of @xmath22 .
the power of such fluctuations of @xmath36 does not diminish with frequency , which is not realistic .
close to the equilibrium operating point , the nonlinear system of equations ( [ eq : swing ] ) , ( [ eq : p ] ) , ( [ eq : q ] ) can be linearized : @xmath52 where @xmath53 is the system state vector respreseting the deviations of voltage phases and magnitudes from the equilibrium value , @xmath54 is the power flow jacobian , the diagonal matrix @xmath55 encodes inertia of the turbines , while the diagonal matrix @xmath56 corresponds to the frequency damping
. the system ( [ eq : swinglinearized ] ) of stochastic differential equations ( sde in what follows ) will be the main subject of our study .
we discuss the general case of a power grid with a large number of @xmath1 nodes ( generators ) and @xmath3 nodes ( loads ) .
it is natural to assume that the graph of the grid under consideration does not have any specific translational / rotational symmetries , in which case the saddle - node bifurcation , corresponding to the global collapse of the power system , has a co - dimension 1 .
the autocorrelation function @xmath57 of the system state vector @xmath53 is given by @xmath58 where @xmath59 and the system matrix is @xmath60 the value of the integral ( [ eq : corrfourier ] ) is determined by the singularities of the integrand in the complex plane of @xmath22 , which in particular include zeros of the determinants @xmath61 and @xmath62 as well as poles of @xmath63 at @xmath64 .
when the autocorrelation function of the system state vector in the frequency domain is analyzed , the contribution of the the poles of @xmath63 can be filtered out by a high - pass filter . in time domain
one can study the autostructure function @xmath65 instead of @xmath66 , where the large contribution of the pole of @xmath63 is cancelled out .
naturally , among the remaining singularities , the one closest to the real @xmath22 axis undermines behavior of the autostructure function ( [ eq : corrfourier ] ) at large @xmath67 . to identify this singularity
, we recall that near a co - dimension 1 saddle node bifurcation one ( and only one ) of the eigenvalues of the power flow jacobian @xmath54 vanishes @xcite .
close to the bifurcation , the inverse of the power flow jacobian can be written as @xmath68 where @xmath69 is the eigenvalue of @xmath54 vanishing at the bifurcation , while @xmath70 and @xmath71 are the corresponding left and right eigenvectors . under these assumptions ,
the leading singularity of the integrand in ( [ eq : corrfourier ] ) coincides with a zero of @xmath61 ( or @xmath62 depending on the sign of the difference @xmath72 ) .
such singularity is a simple pole by assumption that the center manifold of the power system is one - dimensional .
constructing perturbation theory in powers of small @xmath69 and assuming the overdamped operating regime , one finds that the dominating mode determining behavior of the expectation value ( [ eq : corrfourier ] ) at @xmath73 is given by @xmath74 to the leading order in @xmath69 .
the @xmath75 contribution can be neglected as long as @xmath76 .
note that the dependence of the leading mode on the matrix @xmath55 of inertia constants appears only in the second order in @xmath69 and is negligible in the vicinity of the bifurcation point , as @xmath77 .
the frequency of the leading mode is purely imaginary to the order @xmath78 unless @xmath79 , i.e. , when the system is underdamped .
estimating the integral ( [ eq : corrfourier ] ) near the leading singularity @xmath80 , one finally finds to the leading order in @xmath69 @xmath81 @xmath82 note that the matrix elements @xmath83 and @xmath84 are simple numerical factors , so that the dominating direction in the phase space of the system where the fluctuations of the system vector @xmath53 grow coincides with the direction of the right eigenvector @xmath71 corresponding to the lowest eigenvalue @xmath69 of the power flow jacobian @xmath54 , @xcite .
according to the expression ( [ eq : corrffinal ] ) , as @xmath77 and the operating regime of the grid approaches the bifurcation , amplitude of the fluctuations of the system state vector @xmath53 grows as @xmath85 .
depends on the load model , in particular , on the dependence of the expectation value of fluctuating loads on frequency @xmath22 ( @xmath86 in our case ) . ]
simultaneously , the characteristic correlation time of these fluctuations @xmath87 grows as @xmath88 at @xmath89 .
these two effects explain the phenomenon of critical slowing - down often observed during large - scale failures of power grids @xcite .
the growth of both the amplitude of fluctuations and the correlation time imply that close to collapse dynamics of the system vector @xmath53 can be represented as a sequence of relatively long time intervals with weakly changing values of system variables @xmath53 .
however , these values significantly deviate from the equilibrium ones , @xmath90 , determined by the stationarity condition @xmath91 .
this in turn ( at least partially ) explains why it is technically hard to correctly and rapidly identify the state of the power system close to an unstable regime using the standard power flow estimated procedures @xcite : away from the threshold of instability stochastic fluctuations of the system state vector @xmath53 are suppressed , while close to this threshold they are strongly amplified .
the same applies to numerical errors of power flow estimators in operating regimes near bifurcation and prevents effective convergence of numerical schemes .
it is very tempting to use the very signatures of critical slowing - down as markers of approach to stability loss @xcite . in this section
we shall consider how the critical slowing - down affects behavior of the autocorrelation function in the frequency domain and then compare our conclusions to results of numerical simulations .
the complete dominance of a single mode close to collapse implies that in the frequency domain expression ( [ eq : autocorrfrequency ] ) reduces to @xmath92 thus , at very small angular frequencies @xmath93 one finds a quadratic behavior of the fourier - transformed autocorrelation function ( [ eq : autocorrfrequency ] ) with @xmath94 : @xmath95 , while in the intermediate range of frequencies this behavior is quartic instead : @xmath96 .
these two types of behavior are rather universal and can be seen even if the dynamics of the power system does not yet exhibit critical slowing - down .
closer to the threshold of instability , as @xmath77 , the quadratic behavior of ( [ eq : autocorrfrequencysinglemode ] ) is realized in a smaller and smaller range of frequencies @xmath97 .
this in turn leads to a stronger @xmath98 amplification of the amplitude of fluctuations at small @xmath22 : the matching point @xmath99 between @xmath100 and @xmath101 regimes is reached at smaller and smaller @xmath22 , while @xmath77
. corresponding amplification of the amplitude of fluctuations of @xmath53 particularly noticeable at small frequencies is a very good marker of approach to stability loss . in order to check these predictions ,
we have performed numerical simulations of the stochastic behavior of ieee 39 and ieee 57 test power systems close to the threshold of instability in both frequency and time domains .
bifurcation points for both power systems discussed here were first localized using continuation power flow procedure @xcite implemented in psat toolbox for matlab @xcite and then identified more precisely using matpower library for matlab @xcite . for the case of the ieee 39 test power system two chosen values of the continuation power flow parameter corresponded to the smallest eigenvalue of the power flow jacobian @xmath102 p.u .
( operating regime relatively far from the threshold of instability ) and @xmath103 p.u .
( pre - critical operating regime).the loads were allowed to fluctuate only on the buses 3 , 10 and 21 in order to check the difference in correlations of the state vector between the nodes with and without fluctuating power loads . on each of these nodes , a single realization of the active power load fluctuations was considered ( although realizations of @xmath104 of course differed between the nodes ) .
other parameters of the model were chosen similarly to @xcite , which brings the system into an overdamped operating regime .
for the case of the ieee 57 test power system the operational regimes chosen for simulations corresponded to @xmath105 p.u .
( pre - critical regime ) and @xmath106 p.u .
( normal stable operating regime ) .
parameters @xmath11 were chosen as described in the section 2 .
inertia constants @xmath10 of generators were chosen according to the relation @xmath107 , which , as explained in @xcite , effectively holds for many test power system models .
as usual , the bus @xmath21 was the slack bus .
the active power loads were allowed to fluctuate only on the nodes @xmath108 with characteristic amplitude of fluctuations @xmath109 p.u .
the results of frequency domain simulations for the ieee 39 model are presented on the figs.[fig : ieee39-freq ] , [ fig : ieee39freq2 ] .
the @xmath98 behavior along the relevant interval of frequencies is clearly seen for both cases of @xmath102 p.u . and @xmath110 p.u . , as well as an amplification of the fluctuations at small frequencies by more than an order of magnitude in the pre - critical operating regime .
the prediction of the theory for the value of the auto - correlation function at @xmath111hz is @xmath112 for the case @xmath103 p.u . and @xmath113 for the case @xmath102 p.u . , in a good agreement with the results of simulation .
as expected , we have found that the low - frequency behavior of different inter - node / single - node structure functions of the system state vector @xmath53 is rather similar irrespective whether the power load is fluctuating or fixed on the given node .
this implies that a single - node autocorrelation function of the system state vector , in particular , its behavior at small @xmath22 , can be a good indicator of the approach to the loss of stability .
the behavior at large frequencies differed noticeably , with inter - node correlations decaying more rapidly with @xmath22 for nodes with fixed power loads .
the results of simulations of the ieee 57 test power system are essentially similar . the results of simulations of the autostructure function @xmath114 in time domain demonstate its amplification at @xmath115 sec . ,
the load on both nodes @xmath116 and @xmath117 .
the regime @xmath103 p.u .
is close to the threshold of instability , while the regime with @xmath102 p.u .
is relatively far from the threshold .
amplification of the amplitude of fluctuations at small frequencies @xmath22 by more than an order of magnitude is clearly seen .
the main plot uses the log - log scale .
the inset represents behavior of the correlation function at small @xmath22 using the normal scale .
[ fig : ieee39-freq ] ] using the normal scale . the difference in amplitude of correlation functions at very small @xmath22
is fully explained by the difference in magnitudes of the corresponding components of the vector @xmath71 .
[ fig : ieee39freq2 ] ] the singular value decomposition of the system state vector @xmath118 shows that there exists a dominating mode for the pre - critical regime @xmath105 p.u . , while this is not the case for the stable operating regime @xmath106 p.u .
, where the contributions of many different modes to @xmath53 provide a contribution of the same order of magnitude into ( [ eq : corrfourier ] ) , see fig .
[ fig : eigsvd ] .
largest singular values of the system state vector @xmath53 of the ieee 57 power system for pre - critical ( @xmath105 p.u . ) and sub - critical ( @xmath106 p.u . ) operating regimes .
@xmath119 smallest eigenvalues of the power flow jacobian for the same regimes .
[ fig : eigsvd ] ] finally , we would like to emphasize that the contribution of different nodes into the right eigenvector @xmath71 of the leading mode relatively weakly depends on the operating regime .
thus , the dominant nodes of the grid contributing the most to this mode can be identified well before the threshold of instability is approached .
in the present contribution we have studied the phenomenon of critical slowing - down often observed in power grids close to the threshold of a large - scale failure such as voltage collapse or a loss of synchrony @xcite . as we have argued ,
this phenomenon is directly related to ( and can be explained by ) a strong amplification of fluctuations of the system state variables - voltage phases and magnitudes on individual nodes of the grid - for operating regimes close to collapse .
we have explained how the phenomenon of critical slowing - down can be used to effectively detect approach to the threshold of instability .
a technically simple method of detection is based on analysis of single - node ( or inter - node ) autostructure functions of the system state vector in the frequency domain , where approach to the threshold of instability is characterized by a strong amplification of low frequency part of the correlation function .
it is very important to note that the onset of critical slowing - down can be detected using limited measurements of state variables , such as in the situation of incomplete observability of the smart grid , when effective and precise measurement of inter - area modes is impossible .
this work was partially supported by nsf award eccs-1128437 and mit / sktech seed funding grant .
g. andersson , p. donalek , r. farmer , n. hatziargyriou , i. kamwa , p. kundur , n. martins , j. paserba , p. pourbeik , j. sanchez - gasca , r. schulz , a. stankovic , c. taylor , and v. vittal , `` causes of the 2003 major grid blackouts in north america and europe , and recommended means to improve system dynamic performance , '' _ ieee transactions on power systems _ , vol .
20 , no . 4 , pp . 19221928 , nov .
2005 .
p. pourbeik , p. kundur , and c. taylor , `` the anatomy of a power grid blackout - root causes and dynamics of recent major blackouts , '' _ ieee power and energy magazine _ ,
vol . 4 , no . 5 , pp .
2229 , sep .
2006 .
k. loparo and g. blankenship , `` a probabilistic mechanism for small disturbance instabilities in electric power systems , '' _ ieee transactions on circuits and systems _ , vol .
32 , no . 2 ,
177184 , feb . 1985 .
p. hines , e. cotilla - sanchez , and s. blumsack , `` topological models and critical slowing down : two approaches to power system blackout risk analysis , '' in _ 2011 44th hawaii international conference on system sciences_. ieee , jan .
2011 , pp . 110
. d. podolsky and k. turitsyn , `` random load fluctuations and collapse probability of a power system operating near codimension 1 saddle - node bifurcation , '' in _ ieee power and energy society general meeting ( pes ) _ , 2013 , pp .
g. ghanavati , p. d. h. hines , t. lakoba , and e. cotilla - sanchez , `` calculation of the autocorrelation function of the stochastic single machine infinite bus system , '' in _ 2013 north american power symposium ( naps)_. ieee , sep . 2013 , pp .
16 .
j. f. hauer , d. j. trudnowski , and j. g. desteese , `` a perspective on wams analysis tools for tracking of oscillatory dynamics , '' in _ 2007 ieee power engineering society general meeting_. ieee , jun .
2007 , pp . 110
. k. wang and m. l. crow , `` the fokker - planck equation for power system stability probability density function evolution , '' _ ieee transactions on power systems _ , vol .
28 , no . 3 , pp . 29943001 ,
i. dobson , `` observations on the geometry of saddle node bifurcation and voltage collapse in electrical power systems , '' _ ieee transactions on circuits and systems i : fundamental theory and applications _ , vol .
39 , no . 3 , pp . 240243 , mar .
1992 .
r. avalos , c. canizares , f. milano , and a. conejo , `` equivalency of continuation and optimization methods to determine saddle - node and limit - induced bifurcations in power systems , '' _ ieee transactions on circuits and systems i : regular papers _ , vol . 56 , no . 1 , pp .
210223 , jan .
2009 .
r. d. zimmerman , c. e. murillo - sanchez , and r. j. thomas , `` matpower : steady - state operations , planning , and analysis tools for power systems research and education , '' _ ieee transactions on power systems _ , vol .
26 , no . 1
1219 , feb . 2011 . | we consider stochastic electro - mechanical dynamics of an overdamped power system in the vicinity of the saddle - node bifurcation associated with the loss of global stability such as voltage collapse or phase angle instability .
fluctuations of the system state vector are driven by random variations of loads and intermittent renewable generation . in the vicinity of collapse
the power system experiences so - called phenomenon of critical slowing - down characterized by slowing and simultaneous amplification of the system state vector fluctuations . in generic case of a co - dimension 1 bifurcation corresponding to the threshold of instability it is possible to extract a single mode of the system state vector responsible for this phenomenon .
we characterize stochastic fluctuations of the system state vector using the formal perturbative expansion over the lowest ( real ) eigenvalue of the system power flow jacobian and verify the resulting expressions for correlation functions of the state vector by direct numerical simulations .
we conclude that the onset of critical slowing - down is a good marker of approach to the threshold of global instability .
it can be straightforwardly detected from the analysis of single - node autostructure and autocorrelation functions of system state variables and thus does not require full observability of the grid . |
at cern s large hadron collider ( lhc ) @xcite protons will soon collide at an astonishing 800 million times per second to provide experimental evidence for the higgs @xcite , supersymmetry ( susy ) @xcite or extra dimensions @xcite .
susy is widely considered to be one of the best candidates for physics beyond the standard model ( sm ) .
it provides an explanation for the higgs mass problem , a candidate for cold dark matter , and unification of low energy gauge couplings by introducing superpartners to sm fields ( see ref .
@xcite and references therein ) .
an alternative to susy is given by phenomenological extra - dimensional models such as large extra dimensions ( leds ) @xcite , warped braneworlds @xcite or universal extra dimensions @xcite .
scenarios with leds are specially appealing . in these models , gravity becomes strong at the tev scale , where radiative stability is achieved .
the fundamental scale of gravity , @xmath0 tev , is related to the observed planck scale , @xmath1 , by the relation @xmath2 , where @xmath3 is the volume of the extra @xmath4-dimensional space .
one of the most astounding consequences of the existence of extra - dimensions would be the production of subatomic black holes ( bhs ) in particle colliders @xcite and cosmic ray showers @xcite .
( for reviews , see refs . @xcite . )
the atlas @xcite and cms @xcite experiments at the lhc are entrusted with the task of studying events with large transverse momentum ( @xmath5 ) , a signature common to both susy and extra dimensions . while we wait for these experiments to start collecting data , it is worthwhile to look into means of distinguishing susy and extra - dimensional models @xcite
. comparisons of susy and universal extra dimensions / little higgs models in colliders have been investigated by various authors @xcite .
discrimination of susy and bh events by means of dilepton events was recently discussed by the authors in ref.@xcite . in this paper
, we revisit that analysis and extend it to include event shape variables , missing transverse momentum and visible energy .
bh and susy events are simulated with the bh generator ` catfish ` @xcite and the high - energy event generator ` pythia ` @xcite , respectively .
susy masses are set with ` isajet ` @xcite .
the analysis below will show that susy and bh events can be clearly distinguished at the lhc .
bh events tend to be more spherical than susy events because of the isotropic nature of bh decay .
thus event shape variables , such as sphericity , provide good discriminators . on the contrary , visible energy and are less effective discriminators because of the presence of invisible channels in both susy and bh models , which make the amount of comparable in the two scenarios .
the dilepton invariant mass is also an excellent discriminator ; the susy invariant mass shows a sharp cutoff at @xmath6 100 gev , which is absent in the bh model because most of the dileptons originate from uncorrelated events .
the remainder of this paper is organized as follows . in sect .
[ susylhc ] and [ bhlhc ] we briefly review the fundamentals of susy and tev bhs which are needed for our analysis , respectively .
simulations are described in sect .
[ evt_sim ] .
the analysis of visible / missing momentum and event shape variables is presented in sect .
[ shape ] , and the discrimination of susy and bh using dileptons is discussed in sect .
[ dilep ] .
conclusions are presented in sect .
[ concl ] .
the minimal supersymmetric extension of the standard model ( mssm ) @xcite is the simplest susy model . according to the mssm , all sm fermions ( bosons )
must have a bosonic ( fermionic ) partner .
superpartners have identical masses , charges and quantum numbers of their sm counterparts , differing only in their spin .
the mssm allows for the unification of electromagnetic , weak and strong forces at @xmath7 gev .
since we do not observe superpartners of sm particles at low energies , susy must be a broken symmetry .
the susy breaking scale , i.e. the mass scale at which we expect the first susy particles to appear , is generally assumed to be around 1 tev .
a method of susy breaking which is mediated by gravitational interactions is supergravity ( sugra ) . in its minimal version ,
msugra is determined by a point in the five - dimensional moduli space with parameters : * @xmath8 , the common scalar mass at @xmath9 ; * @xmath10 , the common gaugino mass at @xmath9 ; * @xmath11 , the common trilinear coupling at @xmath9 ; * @xmath12 , the ratio of the vacuum expectation values of the two higgs fields ; * @xmath13 , the sign of the higgsino mass parameter .
msugra parameters for five typical lhc points are given in table [ table1 ] @xcite .
neutralino ( @xmath14 ) , gluino ( @xmath15 ) and squark ( @xmath16 ) masses are determined by @xmath8 and @xmath10 as @xmath17 , @xmath18 , @xmath19 and @xmath20 @xcite .
0.50@|c|cccccc| lhc point & @xmath8 & @xmath10 & @xmath11 & tan @xmath21 & @xmath13 & + a & 100 & 300 & 300 & 2.1 & + & + b & 400 & 400 & 0 & 2 & + & + c & 400 & 400 & 0 & 10 & + & + d & 200 & 100 & 0 & 2 & - & + e & 800 & 200 & 0 & 10 & + & + [ table1 ] visible energy , missing transverse momentum and sphericity for the five lhc points of table [ table1 ] are shown in fig .
[ fig1_1 ] , where all susy processes except sm higgs production have been implemented .
sparticle production at point d ( open blue circles ) is higher as squarks and gluinos are lighter .
this point is usually taken as the comparison point between the lhc and other experiments , e.g. tevatron @xcite and nlc @xcite .
for the purposes of our analysis , the difference between the five points is not significant and any of them can be chosen as susy benchmark . in the following , we will consider point a. this is justified by the fact that point a allows for susy higgs production @xcite . since bhs may evaporate into higgs
( see sect .
[ bhlhc ] below ) , a meaningful comparison of susy and bh events requires the presence of the higgs channel in both models .
moreover , distinguishability of susy and bh events must be assessed by minimizing the differences between the two models . since
bh events are characterized by up to several tev of missing transverse momentum , susy points with large , such as point a , must be considered .
a symmetry of the mssm is @xmath22-parity @xcite : @xmath23 where @xmath24 ( @xmath25 ) is the baryon ( lepton ) number and @xmath26 is the particle spin .
all sm particles have @xmath27 whereas their superpartners have @xmath28 .
@xmath22-parity implies that susy particles are always pair produced from sm particles . if @xmath22-parity is conserved , the endpoint of a susy process at the lhc is a state with sm particles and two lightest stable susy particles ( lsps ) , which are generally neutralinos . being colorless and chargeless , the lsps escape the detector and are the source of missing transverse momentum , a leading signature of susy events .
if @xmath22-parity is not conserved , the missing transverse energy is reduced by the lsp decay . in the following
, we will assume that @xmath22-parity is conserved , in agreement with the mssm ( msugra ) scenario .
we end this section with a list of dominant susy interactions at lhc point a and the definition of invariant mass .
this is important for the following analysis because it enables us to select processes that could serve as potential discriminators .
the third decay chain in fig . [ diag ] is specially interesting because it allows the separation of isolated leptons from the hadronic background @xcite .
the invariant mass is defined as @xmath29 where @xmath30 is the angle between the two particles .
the method of constructing invariant masses from susy decay chains has been traditionally used to calculate sparticle masses @xcite . in sect .
[ dilep ] we will use the invariant mass of isolated dileptons with large @xmath5 as a potential susy / bh discriminator .
( for a review on lepton production at colliders , see ref .
@xcite . ) in the rest frame of the second lightest neutralino , the dilepton invariant mass is @xmath31^{1/2}\ , .
\label{invmass}\ ] ] since the momentum of the lsp is not constrained , the invariant mass distribution shows an edge at @xmath32 gev .
in led scenarios , @xmath33 collisions at the lhc could produce tev - mass bhs with characteristic lifetimes of @xmath34 seconds @xcite .
numerous studies have focused on bh signatures at the lhc @xcite and various monte carlo generators are available for simulation purposes @xcite . a quick look at bh production at the lhc
reveals the following . according to thorne s hoop conjecture @xcite ,
a bh of mass @xmath35 is formed when an object is compressed in all directions such that @xmath36 where c is the circumference of the region where the object is compacted into and @xmath37 is the schwarzschild radius for a bh of mass @xmath35 .
an upper limit on the bh mass is obtained by assuming no gravitational energy loss at formation , corresponding to the black disk ( bd ) cross - section @xmath38 .
a more realistic model assumes that all the cm energy is not available for bh formation , some being lost as gravitational radiation ( see ref .
@xcite for a more detailed discussion ) . to estimate the energy loss ,
the colliding particles are treated as two aichelburg - sexl shock waves @xcite ; the overlap of the shock waves forms a trapped - surface ( ts ) which sets a lower limit to the mass of the bh @xcite .
( for an alternative estimate of the collisional gravitational loss , see ref .
the cross - section at the lhc involves summing up the contributions from all the initial partons .
the cross section in the ts scenario is @xmath39 where @xmath40 is four - momentum transfer squared , @xmath41 are the parton distribution functions , @xmath42 is the normalized impact parameter , and @xmath43 is a form factor .
the cutoff in @xmath44 is related to the minimum allowed mass of the object , @xmath45 , and the fraction of center - of - mass energy trapped in the bh , @xmath46 , by @xmath47 $ ] .
tev bhs may carry electric or color charge and angular momentum . immediately after formation , they are expected to decay through loss of excess multipole moments ( balding phase ) , gravitational + hawking radiation @xcite ( evaporation phase ) and final @xmath4-body decay or remnant production ( planck phase ) .
sm particles are emitted on the brane and can be detected @xcite .
since the balding phase is poorly understood , simulations neglect the energy loss in this phase .
the description of the evaporation phase is also approximated ; since emissivities of rotating bhs are not known for all fields , bh generators use greybody factors for non - rotating bhs @xcite . in catfish ,
the total decay multiplicity is @xcite @xmath48 where @xmath49 are the degrees of freedom of species @xmath50 , @xmath51 and @xmath52 are the relative emissivities of ref .
@xcite , @xmath53 is the initial entropy of the bh , and @xmath54 and @xmath55 are spin - dependent normalization factors .
a more detailed discussion of the evaporation and planck phases of tev bhs can be found in refs .
@xcite .
dilepton production in bh events differ greatly from dilepton production in the mssm .
unlike susy , there is no single process of dilepton production ; dileptons are either produced by the bh directly or by the decay of heavier particles such as the @xmath56 boson , @xmath57 pairs of a combination of the two .
therefore , the bh dilepton invariant mass does not show a sharp cut - off at high energy .
susy simulations are carried out using a combination of ` isajet ` and ` pythia ` , with the former generating the mass spectrum .
bh simulations are carried out using the ` catfish ` monte carlo generator .
the setup for each simulation is summarized below : * susy : * * the mssm mass spectrum is generated with ` isajet ` ( ver .
7.75 ) ; * * the mass spectra in slha format is fed into ` pythia ` ; * * all susy processes except sm higgs production are simulated ; * * unstable sm particles and sparticles are hadronized or decayed with ` pythia ` . *
bhs : * * the cross section for a bh event is calculated in the center - of - mass frame ; * * the initial bh mass is sampled from the differential cross section ; * * the bh is decayed through hawking mechanism and final @xmath4-body event ( or remnant ) ; * * unstable quanta are hadronized or decayed with ` pythia ` .
the benchmark model for susy is lhc point a. the parameters for the bh benchmark model are fundamental planck scale @xmath58 tev , minimum bh mass @xmath59 tev , classical - to - quantum threshold @xmath60 tev , six extra - dimensions ( @xmath61 ) and two - body final decay ( @xmath62 ) .
particles produced in the initial - radiation phase are removed by imposing @xmath5 cuts of @xmath63 gev and @xmath64 gev for leptons and photons+hadrons , respectively @xcite .
in this section , we first use event shape variables to discriminate susy and bhs models .
we then complement these results by looking at isolated dilepton events .
the salient features of this analysis is that bh events tend to be more spherical than susy events due to the spherical nature of the hawking radiation .
this is specially evident for high - mass bhs .
the formation of a stable bh remnant at the end of the evaporation phase also helps to discriminate mssm and bh events because of the large amount of energy which is carried away by the remnant .
isolated dilepton events provide a further powerful means to distinguish the two models .
this is due to the fact that leptons are rarely emitted by bhs ( the hadron - to - lepton ratio is approximately 5:1 ) and are uncorrelated ; they can be emitted at any angle w.r.t .
beam axis , whereas susy dileptons typically originate from a single decay chain .
figure [ fig2 ] shows visible energy , missing transverse momentum , and transverse momentum of leptons and hadrons+photons for 10,000 mssm and bh benchmark events . even in the absence of a bh remnant ,
the amount of visible energy and is comparable for the two scenarios .
this is due to the presence of invisible channels in both models : the lsp for susy and neutrinos+gravitons for the bh .
the flavor of the decay products is a better discriminator .
mssm interactions do not produce leptons with energy above the tev since isolated leptons are produced by the decay of sparticles with typical energy of less than a few hundred gev .
on the contrary , quanta produced in the bh decay are characterized by an average energy @xmath65 , where the multiplicity @xmath66 is less than 10 for typical bhs at the lhc . since hawking evaporation
does not distinguish leptons from hadrons , hard leptons with energy up to several tev are likely to be produced during the bh decay .
this suggests that isolated leptons may provide a powerful means to discriminate the two models .
this is indeed the case , as we shall see in the next section .
the four plots in the left panel of fig .
[ fig3 ] show how variations in the bh planck phase affect the observables of fig . [ fig2 ] .
the plots compare two- and four - body decays to the formation of a bh remnant ( @xmath67 ) . by the time the remnant has formed , the bh is expected to have shed electric and color charges .
( see , however , ref .
@xcite for an alternative scenario . )
if this is the case , the bh remnant is undetectable and a source of missing transverse momentum in addition to neutrinos and gravitons which are emitted during the hawking evaporation phase .
this leads to a larger difference in between the mssm and bh models .
the visible transverse momentum in hadrons+photons is sensibly reduced in the presence of a bh remnant ; the latter carries away energy which otherwise would have been emitted in visible channels ( mostly hadrons ) during the bh decay phase .
it is interesting to note that the amount of transverse momentum in the leptonic channel is essentially unaffected by the presence of a bh remnant .
this is due to the fact that leptons are rarer than hadrons in the bh decay phase ; variations in the energy distribution of the leptonic channel are thus suppressed compared to the hadronic channel .
changes in the number of final planckian hard quanta do not produce significant differences in the distributions ; more quanta of lower energy behave statistically like less quanta with higher energy . provided that the bh decays at the end of the hawking phase , it is thus safe to set the number of planckian quanta to @xmath62 or @xmath68 , although bhs may decay in different numbers of particles on a event - to - event basis .
variations in the classical - to - quantum threshold @xmath69 are also not expected to cause significant differences in the energy / momentum distributions . a higher threshold increases the emission in the planck phase while decreasing hawking radiation .
since these phases differs only in relative greybody factors , the effect is too small to be detected .
the four plots in the right panel of fig . [ fig3 ] show visible energy , , and visible transverse momenta of leptons and hadrons+photons for different values of the fundamental planck scale .
higher values of @xmath70 lead to more massive bhs , i.e. higher multiplicity and more energetic quanta .
this causes a significant increase in missing and visible momenta .
if the value of the fundamental planck scale happens to be large , bhs are likely to be found and easily distinguished from susy through detection of highly - energetic isolated leptons and hadronic jets . missing transverse momentum of several tev
would also be observed .
event shape variables such as sphericity and @xmath71 fox - wolfram moment can be used to complement the above analysis .
bh events are more spherical because of the nature of hawking radiation and the `` jetty '' nature of susy decays .
formation of a bh remnant and high values of the fundamental scale lead to significant higher sphericity than susy ( top panels of fig .
[ fig4 ] ) . the @xmath71 fox - wolfram moment ( bottom panels of fig .
[ fig4 ] ) is stable versus changes in the bh planck phase and provides a good mssm / bh discriminator .
bh models with higher @xmath70 can be differentiated more easily from the mssm .
similar conclusions can be reached by looking at jet masses and number of jets .
the mssm generates more and lighter jets than the bh model due to copious production of quarks ( fig .
[ fig5 ] ) .
the difference is again specially significant for high values of @xmath70 and in the presence of bh remnants .
absence of sub-@xmath69 hard jets could provide strong evidence for bh remnant production .
( see the suppression of heavy jets below the classical - to - quantum threshold @xmath72 tev in the top leftmost panel of fig .
[ fig5 ] . ) the use of isolated dileptons as susy signature has been extensively discussed in the literature @xcite .
although their production is not as high as colored particles , high - energy isolated leptons provide a cleaner environment by allowing the removal of the qcd background . moreover ,
since most of bhs produced at the lhc are expected to be very light , multi - particle analysis may not provide the most effective discriminators @xcite .
the study of leptonic final states alleviates this problem . the dominant mssm interaction for opposite - sign , same - flavor ( ossf ) dileptons at lhc point a is @xcite @xmath73 with a branching ratio of 27% .
the maximum dilepton invariant mass for this interaction is @xmath74^{1/2 } \sim 100~{\rm gev}\ , . \label{mll}\ ] ] the background for this process is due to sm decays of @xmath75 , @xmath76 bosons and top quarks . this background can be removed by applying suitable cuts on transverse momentum and sphericity of the leptons @xcite : * @xmath77 gev , @xmath78 ; * isolation cut , @xmath79 gev in a cone of @xmath80 , where @xmath81 is the transverse momentum of the leptons , @xmath82 , @xmath83 $ ] is the pseudorapidity , and @xmath84 and @xmath30 are the azimuthal and polar angles of the lepton w.r.t . beam axis , respectively .
isolated leptons in bh decays come directly from the bh itself , from the decay of @xmath56 bosons and top quarks , or from a combination of the two . since the branching ratio of @xmath56 into leptons is small , @xmath85 @xcite , and the decay of top quarks into leptons is rare @xcite , production of ossf dileptons is less frequent in the bh model than in the mssm .
our analysis shows that an ossf dilepton event occurs approximately every 100 bh and 20 susy events , with a @xmath6 1:5 ratio of bh - to - susy dilepton events at fixed luminosity .
figure [ fig6 ] shows the dilepton invariant mass distribution for the mssm ( shaded pink plot ) and the bh model with final two- and four - body decay ( left and right panels , respectively ) .
as expected , the susy distribution shows a sharp edge at @xmath6 100 gev @xcite .
the bh invariant mass distribution is characterized by a peak at @xmath86 gev and a long tail up to energy of several tev .
the peak is due to dileptons events produced from the decay of @xmath56 bosons , the dominant channel for ossf dilepton production in bh .
the tail is originated by uncorrelated lepton pairs emitted directly by the bh or in top quark decays .
the leptons are hard and the reconstructed dilepton mass can have super - tev values .
the bh invariant mass for the two - body planckian decay shows a second , smaller peak at @xmath6 1 tev .
this occurs because the bh at the end of the hawking phase may decay in two ossf leptons , leading to a reconstructed dilepton invariant mass equal to @xmath87 .
if the planckian decay is a four - body process , the bh mass at the end of the hawking evaporation is distributed among four quanta .
this produces a lower , smoother reconstructed invariant mass .
the number of isolated , high-@xmath5 leptons can also be used to complement the dilepton analysis ( left panel of fig .
[ fig7 ] ) .
susy events are capable of producing up to five isolated leptons from the cascade decay of heavy sparticles .
events with @xmath88 or @xmath89 may produce four or three isolated leptons , respectively @xcite . on the contrary , events with three or more isolated leptons are very suppressed in bh decays at the lhc energy .
although multilepton events are rare , there is very little background and they could be effectively used to distinguish the mssm and the bh model .
other effective discriminators can be constructed by looking at dilepton events with same sign and/or opposite - flavor leptons .
the `` democratic '' nature of the bh decay makes all dilepton events roughly equally probable , whereas the mssm favors same - flavor dileptons .
presence of hard opposite - flavor leptons is a clear indication of bh decay ( right panel of fig .
[ fig7 ] ) .
our analysis shows that 73% of susy dilepton events are ossf , compared to only 50% in the bh model .
conversely , opposite - flavor events are twice more frequent in the bh model ( 40% ) compared to the mssm ( 21% ) .
we have discussed and compared the signatures of the mssm and the tev - bh model at the lhc . a thorough analysis of event - shape variables and dilepton events has shown that it is possible to distinguish the two models .
bh events are characterized by higher sphericity than susy processes . if a bh remnant is formed at the end of the evaporation phase , missing @xmath5 and heavy jet mass are effective signatures to discriminate bh formation from the mssm .
although event - shape variables alone can not unequivocally discriminate between susy and bhs , their knowledge may prove useful when combined with the analysis of the leptonic channel .
isolated dileptons could provide the `` smoking gun '' for detecting bhs at the lhc .
the bh dilepton invariant mass shows a tail at high energy which is absent in the sm or mssm .
this analysis can be further strengthened by looking at the number and flavor of isolated leptons .
the authors thank lucien cremaldi , alakabha datta , romulus godang , david sanders and don summers for fruitful discussions , peter skands for help with slha input file and xerxes tata for help on ` isajet ` . this work was supported ( in part ) by u.s .
doe contract de - fg05 - 91er40622 .
99 the cern website , _ www.cern.ch _
p. w. higgs , phys .
* 12 * , 132 ( 1964 ) ; + f. englert and r. brout , phys .
* 13 * , 321 ( 1964 ) ; + g. s. guralnik , c. r. hagen and t. w. b. kibble , phys .
* 13 * , 585 ( 1964 ) ; + p. w. higgs , phys .
rev . * 145 * , 1156 ( 1966 ) ; + a. djouadi , arxiv : hep - ph/0503172 .
`` atlas tdr on physics performance , vol ii , chap .
19 , '' _ higgs bosons ( 1999 ) _ ; + a. drozdetskiy [ cms and atlas collaborations ] , `` the standard model higgs boson at lhc : recent developments , '' cern cms note , cern - cms - cr-2007 - 022 ( 2007 ) ; + d. kcira , arxiv:0710.1957 [ hep - ex ] .
j. wess and b. zumino , phys .
b * 49 * , 52 ( 1974 ) ; + s. weinberg , `` the quantum theory of fields .
3 : supersymmetry , '' _ cambridge , uk : univ . pr .
( 2000 ) _ and references therein .
h. baer , c. h. chen , f. paige and x. tata , phys .
d * 52 * , 2746 ( 1995 ) [ arxiv : hep - ph/9503271 ] ; + h. baer , c. h. chen , f. paige and x. tata , phys .
d * 53 * , 6241 ( 1996 ) [ arxiv : hep - ph/9512383 ] ; + f. e. paige , arxiv : hep - ph/9801254 . n. arkani - hamed , s. dimopoulos and g. r. dvali , phys .
b * 429 * , 263 ( 1998 ) [ arxiv : hep - ph/9803315 ] ; + i. antoniadis , n. arkani - hamed , s. dimopoulos and g. r. dvali , phys .
b * 436 * , 257 ( 1998 ) [ arxiv : hep - ph/9804398 ] ; + n. arkani - hamed , s. dimopoulos and g. r. dvali , phys .
d * 59 * , 086004 ( 1999 ) [ arxiv : hep - ph/9807344 ] .
l. randall and r. sundrum , phys .
lett . * 83 * , 3370 ( 1999 ) [ arxiv : hep - ph/9905221 ] ; + l. randall and r. sundrum , phys .
* 83 * , 4690 ( 1999 ) [ arxiv : hep - th/9906064 ] .
t. appelquist , h. c. cheng and b. a. dobrescu , phys .
d * 64 * , 035002 ( 2001 ) [ arxiv : hep - ph/0012100 ] ; + j. a. r. cembranos , j. l. feng and l. e. strigari , phys .
d * 75 * , 036004 ( 2007 ) [ arxiv : hep - ph/0612157 ] .
s. p. martin , arxiv : hep - ph/9709356 ; + m. e. peskin , arxiv:0801.1928 [ hep - ph ] .
p. c. argyres , s. dimopoulos and j. march - russell , phys .
b * 441 * , 96 ( 1998 ) [ arxiv : hep - th/9808138 ] ; + t. banks and w. fischler , arxiv : hep - th/9906038 ; + s. dimopoulos and g. landsberg , phys .
lett . * 87 * , 161602 ( 2001 ) [ arxiv : hep - ph/0106295 ] ; + s. b. giddings and s. d. thomas , phys .
d * 65 * , 056010 ( 2002 ) [ arxiv : hep - ph/0106219 ] ; + e. j. ahn , m. cavagli and a. v. olinto , phys .
b * 551 * , 1 ( 2003 ) [ arxiv : hep - th/0201042 ] .
e. j. ahn and m. cavagli , gen .
grav . *
34 * , 2037 ( 2002 ) [ arxiv : hep - ph/0205168 ] ; + v. p. frolov and d. stojkovic , phys .
* 89 * , 151302 ( 2002 ) [ arxiv : hep - th/0208102 ] ; + v. p. frolov and d. stojkovic , phys .
d * 66 * , 084002 ( 2002 ) [ arxiv : hep - th/0206046 ] ; + m. cavagli , s. das and r. maartens , class .
* 20 * , l205 ( 2003 ) [ arxiv : hep - ph/0305223 ] ; + a. chamblin , f. cooper and g. c. nayak , phys .
d * 69 * , 065010 ( 2004 ) [ arxiv : hep - ph/0301239 ] ; + m. cavagli and s. das , class .
* 21 * , 4511 ( 2004 ) [ arxiv : hep - th/0404050 ] ; + a. chamblin , f. cooper and g. c. nayak , phys .
d * 70 * , 075018 ( 2004 ) [ arxiv : hep - ph/0405054 ] ; + c. m. harris , m. j. palmer , m. a. parker , p. richardson , a. sabetfakhri and b. r. webber , jhep * 0505 * , 053 ( 2005 ) [ arxiv : hep - ph/0411022 ] ; + j. tanaka , t. yamamura , s. asai and j. kanzaki , eur .
j. c * 41 * , 19 ( 2005 ) [ arxiv : hep - ph/0411095 ] ; + b. webber , _ in the proceedings of 33rd slac summer institute on particle physics ( ssi 2005 ) : gravity in the quantum world and the cosmos , menlo park , california , 25 jul - 5 aug 2005 , pp t030 _ [ arxiv : hep - ph/0511128 ] ; + l. lonnblad , m. sjodahl and t. akesson , jhep * 0509 * , 019 ( 2005 ) [ arxiv : hep - ph/0505181 ] ; + j. l. hewett , b. lillie and t. g. rizzo , phys . rev . lett . * 95 * , 261603 ( 2005 ) [ arxiv : hep - ph/0503178 ] ; + g. c. nayak and j. smith , phys . rev .
d * 74 * , 014007 ( 2006 ) [ arxiv : hep - ph/0602129 ] ; + h. stoecker , int .
j. mod .
d * 16 * , 185 ( 2007 ) [ arxiv : hep - ph/0605062 ] ; + g. l. alberghi , r. casadio , d. galli , d. gregori , a. tronconi and v. vagnoni , arxiv : hep - ph/0601243 ; + b. koch , m. bleicher and h. stoecker , j. phys .
g * 34 * , s535 ( 2007 ) [ arxiv : hep - ph/0702187 ] ; + d. m. gingrich , jhep * 0711 * , 064 ( 2007 ) [ arxiv:0706.0623 [ hep - ph ] ] . j. l. feng and a. d. shapere , phys . rev
. lett . * 88 * , 021303 ( 2002 ) [ arxiv : hep - ph/0109106 ] ; + a. ringwald and h. tu , phys .
b * 525 * , 135 ( 2002 ) [ arxiv : hep - ph/0111042 ] ; + l. a. anchordoqui , j. l. feng , h. goldberg and a. d. shapere , phys .
d * 65 * , 124027 ( 2002 ) [ arxiv : hep - ph/0112247 ] ; + e. j. ahn , m. ave , m. cavagli and a. v. olinto , phys .
d * 68 * , 043004 ( 2003 ) [ arxiv : hep - ph/0306008 ] ; + j. i. illana , m. masip and d. meloni , phys .
d * 72 * , 024003 ( 2005 ) [ arxiv : hep - ph/0504234 ] ; + v. cardoso , m. c. espirito santo , m. paulos , m. pimenta and b. tome , astropart .
* 22 * , 399 ( 2005 ) [ arxiv : hep - ph/0405056 ] ; + e. j. ahn , m. cavagli and a. v. olinto , astropart .
* 22 * , 377 ( 2005 ) [ arxiv : hep - ph/0312249 ] ; + a. cafarella , c. coriano and t. n. tomaras , jhep * 0506 * , 065 ( 2005 ) [ arxiv : hep - ph/0410358 ] ; + e. j. ahn and m. cavagli , phys .
d * 73 * , 042002 ( 2006 ) [ arxiv : hep - ph/0511159 ] ; + m. cavagli and a. roy , phys .
d * 76 * , 044005 ( 2007 ) [ arxiv:0707.0274 [ hep - ph ] ] .
m. cavagli , int .
j. mod .
a * 18 * , 1843 ( 2003 ) [ arxiv : hep - ph/0210296 ] ; + r. emparan , _ black hole production at a tev _ , arxiv : hep - ph/0302226 ; + s. hossenfelder , _ what black holes can teach us _
, arxiv : hep - ph/0412265 ; + p. kanti , int .
j. mod .
a * 19 * , 4899 ( 2004 ) [ arxiv : hep - ph/0402168 ] ; + g. landsberg , j. phys .
g * 32 * , r337 ( 2006 ) [ arxiv : hep - ph/0607297 ] .
atlas collaboration , w.w .
armstrong _ et al .
_ , technical proposal , cern / lhcc 94 - 43 , lhcc / p2 , 15 , december 1994 .
cms collaboration , g.l .
bayatian _ et al .
_ , technical proposal , cern / lhcc 94 - 38 , lhcc / p1 , 15 , december 1994 . v. buescher , m. carena , b. dobrescu , s. mrenna , d. rainwater and m. schmitt , arxiv : hep - ph/0608322 . t. g. rizzo , phys . rev .
d * 64 * , 095010 ( 2001 ) [ arxiv : hep - ph/0106336 ] ; + c. macesanu , c. d. mcmullen and s. nandi , phys .
d * 66 * , 015009 ( 2002 ) [ arxiv : hep - ph/0201300 ] ; + a. datta , k. kong and k. t. matchev , phys .
d * 72 * , 096006 ( 2005 ) [ erratum - ibid .
d * 72 * , 119901 ( 2005 ) ] [ arxiv : hep - ph/0509246 ] ; + m. battaglia , a. k. datta , a. de roeck , k. kong and k. t. matchev , _ in the proceedings of 2005 international linear collider workshop ( lcws 2005 ) , stanford , california , 18 - 22 mar 2005 , pp 0302 _ [ arxiv : hep - ph/0507284 ] ; + m. battaglia , a. datta , a. de roeck , k. kong and k. t. matchev , jhep * 0507 * , 033 ( 2005 ) [ arxiv : hep - ph/0502041 ] ; + p. konar and p. roy , phys .
b * 634 * , 295 ( 2006 ) [ arxiv : hep - ph/0509161 ] .
a. roy and m. cavagli , arxiv:0710.5490 [ hep - ph ] .
m. cavagli , r. godang , l. cremaldi and d. summers , comput .
commun . * 177 * , 506 ( 2007 ) [ arxiv : hep - ph/0609001 ] ; + m. cavagli , r. godang , l. m. cremaldi and d. j. summers , jhep * 0706 * , 055 ( 2007 ) [ arxiv:0707.0317 [ hep - ph ] ] .
t. sjostrand , s. mrenna and p. skands , jhep * 0605 * , 026 ( 2006 ) [ arxiv : hep - ph/0603175 ] ; + see also : http://projects.hepforge.org/pythia6 .
f. e. paige , s. d. protopopescu , h. baer and x. tata , arxiv : hep - ph/0312045 .
a. bartl _ et al .
_ , _ in the proceedings of 1996 dpf / dpb summer study on new directions for high - energy physics ( snowmass 96 ) , snowmass , colorado , 25 jun - 12 jul 1996 , pp sup112_. m. dittmar , arxiv : hep - ex/9901004 .
r. arnowitt and p. nath , `` prospects for observation of supersymmetry at the tevatron , '' _ ctp - tamu 13/87 , nub-2724 , hutp-87/a021 , jan 1987 . 8pp .
_ j. l. feng , m. e. peskin , h. murayama and x. tata , phys .
d * 52 * , 1418 ( 1995 ) [ arxiv : hep - ph/9502260 ] .
i. hinchliffe , f. e. paige , m. d. shapiro , j. soderqvist and w. yao , phys .
d * 55 * , 5520 ( 1997 ) [ arxiv : hep - ph/9610544 ]
. g. polesello , l. poggioli , e. richter - was , j. soderqvist , precision susy measurements with atlas for sugra point 5 , atlas internal note , phys - no-111 , october 1997 . c. diaconu , arxiv : hep - ph/0401111 . s. dimopoulos and g. l. landsberg , in _ proc . of the aps / dpf / dpb summer study on the future of particle physics ( snowmass 2001 ) _ ed .
n. graf , _ in the proceedings of aps / dpf / dpb summer study on the future of particle physics ( snowmass 2001 ) , snowmass , colorado , 30 jun - 21 jul 2001 , pp p321_. c. m. harris , p. richardson and b. r. webber , jhep * 0308 * , 033 ( 2003 ) [ arxiv : hep - ph/0307305 ] .
d. c. dai , g. starkman , d. stojkovic , c. issever , e. rizvi and j. tseng , arxiv:0711.3012 [ hep - ph ] .
thorne , in : _ magic without magic : john archibald wheeler _ , edited by j. klauder ( freeman , san francisco , 1972 ) .
v. cardoso , e. berti and m. cavagli , class .
* 22 * , l61 ( 2005 ) [ arxiv : hep - ph/0505125 ] .
p. c. aichelburg and r. u. sexl , gen .
* 2 * , 303 ( 1971 ) .
h. yoshino and y. nambu , phys .
d * 67 * , 024009 ( 2003 ) [ arxiv : gr - qc/0209003 ] ; + h. yoshino and v. s. rychkov , phys . rev .
d * 71 * , 104028 ( 2005 ) [ arxiv : hep - th/0503171 ] .
e. berti , m. cavagli and l. gualtieri , phys .
d * 69 * , 124011 ( 2004 ) [ arxiv : hep - th/0309203 ] .
s. w. hawking , commun .
phys . * 43 * , 199 ( 1975 ) [ erratum - ibid . *
46 * , 206 ( 1976 ) ] .
r. emparan , g. t. horowitz and r. c. myers , phys .
lett . * 85 * , 499 ( 2000 ) [ arxiv : hep - th/0003118 ] ; + m. cavagli , phys .
b * 569 * , 7 ( 2003 ) [ arxiv : hep - ph/0305256 ] .
p. kanti and j. march - russell , phys .
d * 66 * , 024023 ( 2002 ) [ arxiv : hep - ph/0203223 ] ; + p. kanti and j. march - russell , phys .
d * 67 * , 104019 ( 2003 ) [ arxiv : hep - ph/0212199 ] ; + c. m. harris and p. kanti , jhep * 0310 * , 014 ( 2003 ) [ arxiv : hep - ph/0309054 ] ; + d. ida , k. y. oda and s. c. park , phys .
d * 67 * , 064025 ( 2003 ) [ erratum - ibid .
d * 69 * , 049901 ( 2004 ) ] [ arxiv : hep - th/0212108 ] ; + a. s. cornell , w. naylor and m. sasaki , jhep * 0602 * , 012 ( 2006 ) [ arxiv : hep - th/0510009 ] .
v. cardoso , m. cavagli and l. gualtieri , jhep * 0602 * , 021 ( 2006 ) [ arxiv : hep - th/0512116 ] ; + v. cardoso , m. cavagli and l. gualtieri , phys .
* 96 * , 071301 ( 2006 ) [ erratum - ibid . *
96 * , 219902 ( 2006 ) ] [ arxiv : hep - th/0512002 ] .
b. koch , m. bleicher and s. hossenfelder , jhep * 0510 * , 053 ( 2005 ) [ arxiv : hep - ph/0507138 ] .
s. abdullin and f. charles , nucl .
b * 547 * , 60 ( 1999 ) [ arxiv : hep - ph/9811402 ] ; + m. chiorboli , m. galanti and a. tricomi , acta phys .
b * 38 * , 559 ( 2007 ) .
meade and l. randall , arxiv:0708.3017 [ hep - ph ] .
b. abbott _ et al .
_ [ d0 collaboration ] , phys .
lett . * 80 * , 2063 ( 1998 ) [ arxiv : hep - ex/9706014 ] .
h. baer _ et al .
_ , arxiv : hep - ph/9503479 . | we show how to differentiate the minimal supersymmetric extension of the standard model from black hole events at the large hadron collider .
black holes are simulated with the ` catfish ` generator .
supersymmetry simulations use a combination of ` pythia ` and ` isajet ` .
our study , based on event shape variables , visible and missing momenta , and analysis of dilepton events , demonstrates that supersymmetry and black hole events at the lhc can be easily discriminated . |
it is of great interest to have an accurate theoretical framework for the description of the interaction and reaction of particles at surfaces
. calculations of this nature would give insight and understanding into the varied and complex processes that take place between atoms , molecules , and solid surfaces .
in relation to this , information concerning atomic and electronic structure , energy barriers , site specificity , bond formation and scission , and the identification of reaction pathways could be obtained . indeed , experimentally , a strong motivation for chemisorption studies with well - defined single crystal surfaces is the prospect of gaining deeper insight into the elementary steps governing heterogeneously catalyzed reactions .
obviously a proper theoretical description of such processes at surfaces with accurate predictive power is of enormous technological importance , particularly for example , in the field of catalysis . in this respect ,
density functional theory is emerging as a very promising approach for describing these phenomena which involve the breaking and formation of chemical bonds , as for example , in the cases of dissociative adsorption , associative desorption , and chemical reactions .
many first - principles theoretical calculations have been performed for atoms , molecules , and small clusters using quantum chemical methods @xcite , and for bulk crystals , clean surfaces , and surfaces with adsorbates using methods originating from solid state physics @xcite .
far fewer , however , have been performed for systems containing both , that is , atoms and molecules at or above surfaces , as for example , gas surface systems , or for chemical reactions at surfaces .
thus , schemes for performing calculations for these kinds of systems typically employ either a finite cluster of atoms or a supercell geometry , reflecting the respective approaches of chemists and physicists .
the supercell method in particular , is proving to be a very successful method for these kinds of problems even though its application to such systems represents a relatively new field .
see , for example , the recent studies of hydrogen dissociation at various surfaces in refs .
@xcite . from studies performed over the last few years
, it has been demonstrated that an improved description of the exchange and correlation energy functional is necessary for a realistic description of the physical and chemical properties of certain systems .
the presently most successful and widely used approach of achieving this is to use the generalized gradient approximation ( gga ) .
such an approach results in a significantly improved description of free atoms and molecules , see for example , @xcite . also for the calculation of hydrogen dissociation , where bonds are in the process of being broken and reformed , the gga has been found to be of crucial importance for obtaining energy barriers in agreement with experiment .
cohesive @xcite and chemisorption energies obtained using the gga are also found to be considerably closer to experiment than the lda values , for example , molecules at surfaces in which there are strong charge inhomogenuities in the system , e.g. , for co adsorption @xcite .
we note that improvement brought about by using the gga is not limited to atoms and molecules in free space and at surfaces , it has also been found , for example , in a study for different phases of ice to be necessary in order to describe the hydrogen bond properly @xcite .
furthermore , it has been shown to improve the coexistence pressure for the diamond structure to @xmath2-tin phase transition in si @xcite and in describing a high - pressure phase transition of sio@xmath3 @xcite .
there is still some difficulty in treating certain systems however , as highlighted recently in a test of twenty - four different local and gradient - corrected dft methods @xcite for the energy barrier of the h + h@xmath4h@xmath3 + h reaction . in this study
it was found that the energy barrier is significantly underestimated when using the lda _ or _ a gga . in this case ,
inclusion of a self - interaction correction ( sic ) was found to be important . for surfaces , however , numerous calculations have demonstrated that dft within the lda represents a good description of the quantum - mechanical many - body interactions for well - bonded situations , where calculated electron densities , atomic geometries , and total energy differences are very reliable .
nevertheless , the role , of the gga in calculations for adsorbates at surfaces , including diffusion barriers , is not yet clear . in the present paper recent results of density functional theory calculations for two quite different adsorption systems are discussed .
the first system involves a strongly electropositive adsorbate , namely , sodium on al(111 ) for which the stability of an ordered `` four - layer '' surface alloy with a @xmath5 periodicity was identified @xcite .
this structure was recently confirmed by low energy electron diffraction ( leed ) intensity and surface extended x - ray adsorption fine structure ( sexafs ) studies @xcite .
the second system concerns the adsorption of oxygen on ru(0001 ) for which it was very recently predicted that the formation of a @xmath0 adlayer should be possible which therefore implies that the apparent saturation coverage of @xmath1 observed under ultra high vacuum conditions using molecular oxygen is due only to kinetic hindering @xcite .
the calculation method used to study the two adsorption systems is the pseudopotential plane wave method employing _ ab initio _ , fully separable pseudopotentials where the adsorbate system is modelled using the supercell approach . the calculation scheme
@xcite affords a fully self - consistent relaxation of the electrons and full relaxation of the atomic positions using damped molecular dynamics .
calculations for na on al(111 ) were carried out primarily using the local - density approximation ( lda ) for the exchange - correlation functional @xcite where the non - linear form of the core - valence exchange - correlation functional was taken into account for the alkali metal atom @xcite .
further details of the calculation can be found in ref .
one additional calculation was performed for the identified @xmath5-na / al(111 ) surface structure using the gga of perdew _ et al .
_ @xcite . in this respect ,
we note that the gga functional is used in creating the pseudopotentials as well as in the surface calculations .
thus the gga is treated in a consistent way which is to be distinguished from the inconsistent approach where lda pseudopotentials are used , with an otherwise gga treatment . for the calculations for o on ru(0001 )
we employed the gga for all calculations because for our later , related work on the catalytic oxidation of carbon monoxide reaction , use of the gga is expected to be very important .
details of the calculations are given in ref .
recent studies of na and other alkali metal atoms on al(111 ) have revealed that these systems contradict almost every expectation of alkali metal adsorption behavior based on traditional understanding ( see , for example , refs .
in particular , at room temperature where thermal energy allows activation barriers to be overcome , evidence was presented from a combined surface extended x - ray adsorption fine structure ( sexafs ) and density - functional theory ( dft ) study @xcite that an unexpected reaction takes place between the na atoms and the al atoms at the surface : the na atoms `` kick - out '' surface al atoms and adsorb in their place , occupying surface substitutional sites .
the resulting structure has a @xmath6 periodicity and a local coverage of @xmath7 , where @xmath8 is the ratio of the number of adsorbate atoms to the number of substrate atoms in an ideal layer .
the surface substitutional site for na was later confirmed by a detailed leed intensity analysis @xcite .
in addition to sodium , it was subsequently shown that potassium @xcite and rubidium @xcite also adsorb substitutionally at room temperature with the same periodicity and coverage .
adsorption of lithium on al(111 ) results in a @xmath6 structure and it has been proposed that the li atoms also adsorb substitutionally @xcite ; this , however , has not been confirmed by a structural analysis .
cesium , on the other hand , forms a @xmath9 structure at room temperature and there is indication from high resolution core level spectroscopy ( hrcls ) studies @xcite that reconstruction of the substrate also takes place , however , the exact atomic geometry of this phase is to date unknown . at higher coverages ( @xmath10 ) of na on al(111 ) , in contrast to k , rb , and cs where @xmath7 is the saturation coverage , a ( @xmath11 ) structure forms in which there are two na atoms per surface unit cell .
identification of the precise atomic geometry of this phase had proved somewhat elusive : early studies by porteus @xcite and hohlfeld and horn @xcite suggested models that involved unreconstructed al substrates , namely , in the former study , it was proposed that there were three rotated domains of @xmath12 periodicity , and in the latter study , the model was proposed to be comprised of two na layers , each with a @xmath5 periodicity . based on hrcls
it was concluded @xcite that these previous models were incorrect and that a correct model had to include strongly intermixed na / al layers . from a normal incidence standing x - ray wave - field ( nisxw ) study @xcite a model
was presented involving two layers each of stoichiometry naal@xmath3 , while from a recent scanning tunnelling microscopy ( stm ) study @xcite , it was concluded that in the surface unit cell one na atom occupies a substitutional site and the other na atom occupies a hollow site .
the , then unclear , situation regarding knowledge of the detailed atomic structure of this @xmath5 phase prompted our theoretical study @xcite . by performing calculations for a number of possible surface geometries a particularly stable structure was identified .
the validity of this structure , which constitutes a binary `` four - layer '' surface alloy , was confirmed by low energy electron diffraction ( leed ) and sexafs @xcite . the atomic structure of @xmath5-na / al(111 )
is shown in fig . 1 . the al layer containing atoms of the type al@xmath3 we refer to as the `` vacancy - layer '' .
the na atoms , na@xmath13 and na@xmath14 , occupy substitutional and fcc - hollow sites , respectively , and in between these two na layers , al atoms , al@xmath15 , are positioned in hcp - hollow sites with respect to the `` vacancy - layer '' .
the structural parameters determined by dft - lda , leed , and sexafs are given in tab .
i. it can be seen that the values agree very well ; the interlayer spacings obtained by all three methods deviate by at most by 0.13 .
we note that the sexafs analysis could not uniquely determine the surface structure because good fits between experimental and theoretical sexafs curves were also obtained for a different structural model , although it definitely excluded the nixsw- and stm - derived models .
the dft - lda and leed results exhibit , in particular , very good agreement : they both identify the presence of small ( @xmath16 0.04 ) lateral displacements of al atoms in the `` vacancy - layer '' ( al@xmath3 ) and the layer beneath it , as well as a small rumpling ( @xmath16 0.05 ) in the two al layers beneath the `` vacancy - layer '' .
the calculations for bulk al using the gga gave a lattice constant that is slightly larger than that obtained using the lda : compare 4.05 to 3.98 , i.e. , it is 0.07 or 1.83 % larger .
these values can be compared to the experimental al lattice constant of 4.02 @xcite ( at low temperature ) .
it is a general trend that gga - determined lattice constants are larger than the lda - derived values @xcite .
the effect of the gga with respect to the surface structure is found to yield na - al bond lengths somewhat longer than the lda result ; the respective average na - al bond lengths are 3.35 and 3.26 , i.e. the value determined using the gga is 2.7 % longer than the result obtained using the lda . for comparison ,
the corresponding value obtained from the leed intensity analysis is 3.34 which agrees more closely with the gga result .
the bond length of the al atom , al@xmath15 , to its nearest neighbors in the `` vacancy layer '' however , increases by only 0.7 % when using the gga as compared to the lda result . with respect to the above - mentioned lateral relaxations , and
the rumpling in al layers five and six ( i.e. the first and second al layers beneath the vacancy - layer , respectively ) , almost no change is found between the lda and gga results .
similarly , the induced work function change is unaltered with a value of @xmath171.16 ev ; the value and sign of which reflects the electropositive nature of the na atom .
we note that a similar effect was obtained in calculations for the two @xmath18-rb / al(111 ) phases which form at low temperature ( on - top adsorption site ) and room temperature ( substitutional adsorption site ) @xcite . here
the bond lengths obtained for rb in the on - top and substitutional sites using the lda were 3.30 and 3.65 respectively , and using the gga the corresponding values were 3.45 and 3.78 .
the values obtained from a leed intensity analysis @xcite were 3.34 and 3.72 respectively , for the on - top and substitutional sites .
similarly , as discussed below , the gga results for the o - ru bond lengths are slightly longer than those determined by leed .
the tendency of a gga to commonly give slightly larger interatomic distances than experiment has been previously noted @xcite , and it appears that this trend extends to bond lengths of adsorbates at surfaces as well .
the atomic structure of @xmath5-na / al(111 ) together with that of @xmath6-na / al(111 ) , for which na adsorbs substitutionally , and as mentioned above , likewise is formed for the k and rb adsorbates , indicates that a certain mass transport occurs on the surface as explained in the following .
considering first the @xmath6 structure , it is of interest to consider how it forms and where the `` missing '' 1/3 of a monolayer of al atoms go .
it seems most natural to assume that the alkali metal atoms `` kick - out '' al atoms , i.e. , a place exchange occurs between the alkali metal atoms and al atoms in the layer beneath . in relation to this , from dft calculations @xcite , a mechanism was identified which is necessary for the kick - out reaction process to be significant , namely , the re - bonding of the displaced substrate atoms at kink sites at steps .
this uses the fact that in thermal equilibrium kink sites represent the thermodynamic reservoir for substrate atoms and establishes that the chemical potential of the al atoms equals the cohesive energy . as discussed in ref .
@xcite , it could also be possible , however , that al atoms desorb _ from _ steps and diffuse to the regions between on - surface alkali metal atoms to form the substitutional structure , i.e. , the on - surface alkali metal atoms `` capture '' the al atoms .
additionally it could be considered that at first the alkali metal atoms kick - out al atoms which then diffuse to regions between on - surface alkali metal atoms , forming two terraces each with a substitutional structure .
interestingly , from a hrcls study of the temperature - dependent phase transition from the on - top to substitutional @xmath6-rb / al(111 ) adsorption structures , al atoms `` trapped '' between on - surface rb atoms have been identified @xcite .
some insight into the mechanisms involved in the formation of the @xmath6 substitutional structures can be obtained by consideration of the activation energy barrier for the phase transition .
this activation energy barrier is related to the transition temperature through the arrhenius equation . for the kick - out mechanism ,
the transition state could be considered as corresponding to a configuration involving , the alkali metal atom plus an al atom , over a vacancy in the al(111 ) surface .
this transition state geometry will yield a corresponding activation energy barrier ( and transition temperature ) which _ depends _ on the species of alkali metal adsorbate since the alkali metal atom is involved in the geometry of the transition state .
it may happen , however , that the barrier height is small or comparable to that of an `` intermediate geometry '' corresponding to that in which the alkali metal atom is in the vacancy ( adsorbed substitutionally ) and the ejected al atom in an isolated position on the surface , not yet re - bonded at a kink site at a step .
in this case the activation energy barrier ( and transition temperature ) could be determined using this intermediate geometry .
this has been assumed in earlier work to estimate transition temperatures @xcite . on the other hand , for the mechanism where we consider the possibility that al atoms desorb from steps to regions _ between _ on - surface alkali metal atoms , the rate - determining process is likely to be the desorption of al atoms from the steps .
this mechanism will be largely _ independent _ of the species of the on - surface alkali metal adatoms and correspondingly so will the activation energy barrier and transition temperature be .
interestingly , from hrcls @xcite and second harmonic generation measurements @xcite , it is found that the transition temperature _ does _ depend on the adsorbate and is higher for rb than for k , reflecting a higher activation energy barrier for rb : for rb the transition temperature was reported to be 250 k @xcite and for k temperatures of 220 k @xcite and 210 k @xcite were given .
the fact that the transition temperature does depend on the particular species of the alkali metal atom lends support to the kick - out process .
additional support comes from the stm study @xcite in which small al plateaus on the @xmath6 surface , not typically found on the clean surface , were identified .
discussion along similar lines has been presented in some detail in ref .
@xcite .
we now consider formation of the higher coverage @xmath5 phase .
the atomic composition of this phase reveals that another interesting mechanism could be active : on first consideration it would appear that no mass transport of al atoms is required in its creation , that is , the `` missing ''
al atoms from the vacancy - layer are seemingly re - bonded between the two na layers ( see fig .
1 ) . however , because the @xmath5 structure can be formed by adsorption of one - sixth of a monolayer of na onto the @xmath6 structure , in which every third surface al atom is _ missing _
, a reverse process is implied , namely , diffusion of one - third of a monolayer of al atoms back from kink sites at steps and reaction with the na atoms to form the @xmath5 phase . a simplistic illustration of this process
is provided in fig .
2 . in the stm study
@xcite `` holes '' in terraces were observed to have formed on creation of the @xmath5 phase .
the number of al atoms taken from these `` holes '' , however , was reported to correspond to only 1/12 of a monolayer , i.e. , less than the required 1/3 of a monolayer .
this suggests that the additional al atoms come from steps or that they have undergone long range mass transport and come from areas on the surface not imaged .
we now turn to the second and quite different adsorption system of o on ru(0001 ) .
from recent experiments of the catalytic oxidation of carbon monoxide performed at high gas partial pressures , evidence has been presented that ru(0001 ) can support an unusually high concentration of oxygen at the surface @xcite .
this is particularly interesting because it is well known that under ultra high vacuum conditions ( uhv ) the saturation coverage is close to @xmath10 @xcite and very little in fact is actually known about possible structures that may form under high o@xmath3 pressures . in order to investigate the structure and stability of high coverage oxygen structures , we performed density functional theory calculations for various o - adlayers on ru(0001 ) as well as for clean ru(0001 ) , as reported in ref .
the structures investigated were the two ordered phases , @xmath5 @xcite and @xmath12 @xcite , which form at room temperature under uhv conditions for coverages @xmath19 and @xmath10 , respectively , and for several higher coverage @xmath0 structures with coverage @xmath20 .
the calculations for the @xmath5 and @xmath12 phases provided a test of the accuracy of the calculations through comparison with available leed intensity analyses . from the comparisons it was found that very good agreement with respect to the preferred adsorption site and the structural parameters was obtained .
the investigation for the higher coverage structures revealed that although a @xmath0 phase is not observed to form under uhv conditions using molecular oxygen , the adsorption of oxygen on ru(0001 ) with a @xmath0 structure is appreciably exothermic indicating that it should be able to form .
this furthermore implies that its formation at low o@xmath3 pressures is prohibited by kinetic limitations due to activation energy barriers for dissociation . for the @xmath5 structure ,
the theoretically obtained binding energy of o relative to the free o atom , in the hcp - hollow site was found to be considerably more favorable than in the fcc - hollow site where the energies are 5.55 ev and 5.12 ev , respectively .
for the free o atom , we included the spin polarization energy which is calculated to be 1.521 ev @xcite .
the dft - gga result is in accord with the leed intensity determination of the hcp - hollow adsorption site for oxygen in this phase .
the theoretical o - ru bond length is 2.10 and the first ru - ru interlayer spacing is found to be contracted by 2.7 % with respect to the bulk value ( using the centers of gravity of the first and second buckled ru layers ) .
the corresponding leed - derived values are 2.03 and 2.1 % , respectively .
as noted above , the calculated bond length is somewhat longer than the experimental one .
lateral and vertical relaxations of the first two ru layers are induced by the o adatoms , the direction and magnitude of which are well reproduced by the theory ( see ref .
@xcite for a comparison ) .
figure 3 shows the atomic geometry of the @xmath5-o / ru(0001 ) structure as determined by the calculations .
it can be seen that the three ru atoms coordinated to the o adatom move laterally away from it resulting in three ru atoms in the first layer moving closer together ; the fourth ru atom does not move laterally ( it is prohibited by symmetry ) but it is vertically displaced in towards the bulk and slightly further than the other three ru atoms .
similarly in the second ru layer , three ru atoms move laterally towards each other and the fourth ru atom , below the o adatom , is vertically displaced in towards the bulk relative to the other three ru atoms in the same layer
. calculations for the @xmath12 structure show that the obtained difference in binding energy for oxygen in the fcc- and hcp - hollow sites is notably less than for that of the lower coverage @xmath5 structure ; the binding energy of o in the hcp - hollow site is 5.28 ev and that for the fcc - hollow site is 5.00 ev , a difference of 0.28 ev as compared to 0.43 ev .
again , adsorption in the hcp - hollow site is preferred , in agreement with the site determined from the leed intensity analysis @xcite .
the oxygen - induced substrate relaxations are particularly complex : the o atoms atoms adsorb in `` off '' hcp - hollow sites , i.e. , they are displaced slightly from the center of the site towards an on - top site , and `` row - pairing '' and buckling of the first two substrate layers occur .
figure 4 depicts the atomic structure as obtained from the calculations . on comparision with
the leed determined structural parameters ( see ref .
@xcite ) we find that the dft - gga calculations predicted all of these relaxations with the one exception that our results indicated row - pairing of the ru atoms in the _ second _ ru layer , as well as in the first layer , and the leed intensity analysis did not . both dft - gga and leed find that the first two ru - ru interlayer spacings , defined with respect to the centers of gravity of the buckled atomic layers , correspond to the bulk value to within 0.01 .
the two theoretical o - ru bond lengths in the @xmath12 structure , each of approximately 2.08 , are very similar to that for the lower coverage structure which was 2.10 .
the value is again slightly larger than that obtained from the leed intensity analysis , which for this structure , was 2.02 . in view of the very good agreement of the results of the dft - gga calculations and the leed intensity analyses as outlined above , various higher coverage @xmath0-o structures with coverage @xmath20 were investigated . in particular , adlayers with o in the fcc- , and hcp - hollow sites were considered .
similarly to the lower coverage structures , the hcp - hollow site was found to be energetically more favorable than the fcc - hollow site with respective binding energies of 4.84 ev and 4.76 ev ; the difference in binding energy of 0.08 ev between the two sites again being reduced as compared to those of the lower coverage structures .
calculations were also performed with a higher energy cut - off and a larger * k*-point set , namely , 60 ry and fourteen special * k*-points is the irreducible part of the brillouin zone .
the resulting binding energies differed by only 0.03 ev and by 0.05 ev for the hcp- and fcc - hollow sites , respectively , and the resulting structural parameters differed by not more than 0.02 and by 0.04 for the hcp- and fcc - hollow sites , respectively , for the two different basis sets .
these results are collected in tab .
in addition , using the larger basis set , the binding energy of o in the on - top and bridge sites was calculated .
these sites were found to be considerably less favorable that the two hollow sites .
the obtained binding energies being 3.62 , 3.93 , 4.81 , and 4.87 ev , for the on - top , bridge , fcc- and hcp - hollow sites , respectively . for the hcp - hollow site ,
the theoretical o - ru bond length of 2.03 is slightly shorter than that of the lower coverage structures ( compare 2.10 and 2.08 for the @xmath5 and @xmath12 structures , respectively ) and the first ru - ru interlayer spacing is found to be _ expanded _ by @xmath212.7 % relative to the bulk spacing .
the surface atomic geometry is depicted in fig .
the value of 4.87 ev for the binding energy of o in the hcp - hollow site on ru(0001 ) indicates that the @xmath0 adlayer structure should be able to form because it is considerably larger ( by @xmath21 1.8 ev ) than that which the o atoms have in o@xmath3 .
as noted above , under uhv conditions using o@xmath3 the saturation coverage is @xmath22 associated with the ordered @xmath12 superstructure .
thus , the reason that a @xmath23 adlayer with coverage @xmath20 does not form under uhv conditions is apparently due to a kinetic hindering for the dissociation of o@xmath3 .
it is probable that such a hindering may be overcome at high o@xmath3 pressures and elevated temperatures as used in the catalytic reactor experiments @xcite due to the high attempt frequency . alternatively ,
if _ atomic _ oxygen is offered , the @xmath0-o structure may be realized as we predicted in ref .
@xcite .
the coverage dependence of the binding energy described above , in which it becomes less favorable with increasing coverage , reflects a repulsive interaction between the adsorbates and implies that no island formation is expected to occur in the coverage regime of @xmath19 to @xmath20 .
concomitantly , the difference in binding energy between the fcc- and hcp - hollow sites becomes less with increasing coverage .
in fact , for the full monolayer this difference is quite small . however , because the full monolayer is reached successively via the other ( lower coverage ) phases , for which the hcp - hollow site is clearly favored , we expect a nearly perfect hcp - hollow site occupation , i.e. only few fcc - hollow site dislocation structures , for the @xmath0 oxygen layer . in the coverage range
investigated , the work function increases with increasing coverage in accordance with the high electronegativity of oxygen and in good agreement with experiment @xcite .
for the @xmath0-o structure , the value of the work function change is about 1.11 ev . because the @xmath0-o adlayer is appreciably exothermic it prompts the consideration of whether an even higher coverage of oxygen can adsorb at the surface . to investigate this possibility we performed calculations for three structures with an o coverage of @xmath24 .
these structures are shown in fig .
6 . in the first structure ,
depicted in fig .
6a , there are four o atoms in hcp - hollow sites in the @xmath5 surface unit cell and one in an fcc - hollow site .
this structure was found to be unstable in that the oxygen atom in the fcc - hollow site did not form a bond with the substrate and lifts off the surface . in the second geometry considered , illustrated in fig .
6b , an o@xmath3 dimer is adsorbed in the hcp - hollow site of the surface unit cell with the molecular axis perpendicular to the surface . in the remaining three hcp - hollow sites ,
o atoms are positioned .
likewise , this structure was also found to be unstable .
these results indicate that the optimum coverage chemisorbed _ on the surface _
is @xmath8=1 . in the last structure
considered , as shown in fig .
6c , there are four o atoms in hcp - hollow sites in the surface unit cell and one o atom in an octahedral ( subsurface ) site between the first and second ru layers . this structure was found to be metastable where the average adsorption energy per o atom in the surface unit cell is 4.09 ev ; alternatively stated , a free o atom gains 1.07 ev if it can adsorb in the subsurface octahedral site of the o - covered surface .
this geometry is , however , less favorable ( by an average of 0.37 ev per atom ) than that with four o atoms in hcp - hollow sites and the fifth o atom in an o@xmath3 dimer in gas - phase .
that is , while o adsorption under the o - covered surface in an octahedral site is metastable with respect to free o atoms , if o@xmath3 is used , it is energetically more favorable to retain the @xmath0 adlayer structure with the additional o atom in the o@xmath3 dimer in the gas - phase . if , however , atomic oxygen is used it is possible that the oxygen atoms may enter the subsurface region , provided they can overcome the diffusion energy barrier for passing through the first ru layer .
using density functional theory calculations it has been possible to identify new and unexpected adsorbate phases , as well as obtaining insight into the behavior of the adsorption systems in general . in the present paper this was demonstrated for the @xmath5-na phase on al(111 ) which represents a binary `` four - layer '' surface alloy that is able to form even though the constituent atoms are immiscible in bulk , and for the adsorption of oxygen on ru(0001 ) for which it was found that a @xmath0-o adlayer structure should be able to form .
as indicated in the introduction , the application of similar theoretical approaches to those mentioned in the present paper , to studies into the interaction and chemical reaction of atoms and molecules at surfaces is expected to be a particularly active area due to the relevance of these processes in heterogeneous catalysis .
for example , the pseudopotential method , m. scheffler , in festkrperproblem * 22 * ed .
p. grosse , vieweg , braunschweig 1982 , p. 115 ; m. j. gillan , j. phys . :
matter * 1 * , 689 ( 1989 ) ; j. dabrowski and m. scheffler , phys . rev .
b * 40 * , 10391 ( 1989 ) ; j. l. a. alves , j. hebenstreit , and m. scheffler , phys . rev .
b * 44 * , 6188 ( 1991 ) ; m. c. payne , h. p. teter , d. c. allan , t. a. arias , j. d. joannopoulos , rev . mod
. phys . * 69 * , 1045 ( 1992 ) ; the linear augmented plane wave ( lapw ) method , e. wimmer , h. krakauer , m. weinert , and a. j. freeman , phys .
b * 24 * , 864 ( 1981 ) ; m. weinert , w. wimmer , a. j. freeman , and h. krakauer , phys . rev
. lett . * 47 * , 705 ( 1981 ) ; the linear muffin - tin orbital ( lmto ) and augmented spherical wave ( asw ) method , m. said , f. mca , k. kambe , and m. scheffler , phys .
b * 38 * , 8505 ( 1988 ) ; m. methfessel , d. hennig , and m. scheffler , phys . rev .
b * 46 * , 4816 ( 1992 ) .
.interlayer spacings , z , between the atomic layers indicated in parenthesis ( see fig . 1 ) .
for the rumpled al layers , five and six , the interlayer distances are given with respect to the center of gravity of the respective layer . in relation to this
, we note that in refs .
@xcite , the interlayer spacings for the dft - lda results were given with respect to the positions of the three al atoms per unit cell which have the same vertical position and not with respect to the center of gravity of the rumpled layer . [ cols="<,^,^,^,^,^,^",options="header " , ] | the success of density functional theory for the description of the adsorption of atoms on surfaces is well established , and based on recent calculations using gradient corrections , it has been shown that it also describes well the dissociative adsorption of molecules at surfaces admittedly however , the data base for reactions at surfaces is still somewhat small . in the present paper the power of density functional theory calculations
is demonstrated by investigations for two different adsorption systems , namely , one with a strongly electropositive adsorbate [ na on al(111 ) ] and one with a strongly electronegative adsorbate [ o on ru(0001 ) ] . in each case , new hitherto not expected adsorbate phases have been predicted by the theory : for na on al(111 ) the stability of a `` four - layer '' surface alloy was identified while for o on ru(0001 ) it was predicted that the formation of a @xmath0-o adlayer should be possible which implies that the apparent saturation coverage of @xmath1 is due to kinetic hindering . |
high mobility two dimensional electron gases formed in gaas / algaas heterostructures allow ballistic motion of electrons over distances of several @xmath2 m at low temperatures@xcite .
ballistic system is realized at low temperatures , low electron densities and low impurity concentrations .
such samples are used in many experiments to study various novel transport properties in low dimensional mesoscopic semiconductors , and a lot of interesting phenomena have been discovered .
one of such mesoscopic systems is a two dimensional electron gas ( 2deg ) in a periodic antidot array . when the sample s typical geometrical scale , such as the distance between antidots , is larger than the fermi wave length , the dynamics of a wave packet approaches the classical limit .
semi - classical model explains some experimental results in this system , such as the quenching of the hall resistance in low fields@xcite , and the oscillatory field - dependence of the magnetoresistance in an antidot array @xcite . in the latter ,
the potential is periodically modulated , and the commensurability of the period of potential and the cyclotron radius plays an important role .
variety of interesting features appear in the diagonal and hall resistances due to this commensurability effect .
most of the features are now well understood in terms of classical orbits .
for example , the peak structure of magnetoresistance is explained by the existence of runaway orbits.@xcite detailed analyses of the hall resistivity have made clear the step - like structure of the hall resistivity accompanied by the peak of the diagonal resistivity,@xcite and the negative and vanishing hall resistivity in low field regime.@xcite the quantum transport is also studied extensively @xcite to clarify the altshuler - aronov - spivak effect @xcite and the anderson localization @xcite in antidot systems .
then , the question arises .
can we observe similar effect when the magnetic fields instead of potentials are modulated ?
the answer is yes .
many peculiar features are observed experimentally when the magnetic field is modulated one dimensionally@xcite and two dimensionally@xcite . in this paper
, we numerically calculate the magnetoresistance as well as the hall resistance in the presence of two dimensionally modulated magnetic fields . a classical billiard model is adopted and use is made of the kubo formula to obtain the transport coefficients .
such a classical model is justified , since the experimentally observed temperature dependence of the transport coefficients is weak , suggesting that the quantum interference effects may be neglected . in the next section ,
we explain the model in detail , and in the third section , we present the results .
the final section is devoted to summary and discussions .
besides a uniform magnetic field @xmath3 , the periodically modulated fields @xmath4 are applied perpendicular to the plane of 2deg in the model .
experimentally , these modulated magnetic fields are formed by ferromagnetic metals or a type ii superconductor put on the surface of the heterostructure .
we introduce dimensionless variables @xmath5 where @xmath6 is the basic lattice constant of the artificial structure of @xmath4 , @xmath7 is the fermi velocity , and @xmath8 .
we scale the magnetic field by @xmath9 , where @xmath10 is the effective mass of electron , and @xmath11 is the electron charge .
thus @xmath12 corresponds to a cyclotron radius of @xmath13 .
we ignore inelastic scattering mechanism in this system , and therefore the classical electron velocity @xmath14 is conserved .
the equations of motion is given by @xmath15 where @xmath16 is a total magnetic field perpendicular to the plane of the 2deg and scaled by @xmath17 .
it is a sum of the external uniform and periodically modulated magnetic field , @xmath18 we use two types of modulated magnetic fields @xmath4 . in one model ( type i )
, @xmath4 is finite and constant only in a finite region of radius @xmath19 and zero otherwise ( fig .
1(a ) ) . in the other model ( type ii ) ,
a smooth space variation is assumed for the modulated field ( fig .
1(b ) ) . whether we take type i model or type ii depends on how we realize the periodically modulated fields experimentally . in the first case ( type i ) , the magnetic field @xmath20 is constant and changes discontinuously in the designed region which exists at intervals . in the unit cell
whose center is at the origin , it is expressed as @xmath21 in the second case ( type ii ) , we use smooth function for @xmath4 .
@xmath22^{\beta } , \label{e5}\ ] ] where @xmath23 controls the steepness of the modulated magnetic field .
the factor @xmath24 in eq.([e5 ] ) determines the maximum value of the modulated magnetic field , and @xmath25 controls the period of modulation .
we consider here only the case where @xmath26 .
we apply the classical linear response theory@xcite to calculate the conductivity tensor , and @xmath27 is given by @xmath28 where @xmath29 is the velocity correlation averaged over initial condition . here
@xmath30 is the average time scaled by @xmath31 between two impurity scatterings in this system .
impurity scatterings are not taken into account in the calculation of the equation of motion eq.([e2 ] ) , assuming that the impurity scattering effect and the classical motion are separable as expressed in eq.([e6 ] ) .
a similar assumption was adopted in ref.@xcite to obtain magnetoresistance oscillations in an antidot array . in this paper we chose @xmath32 .
we compute @xmath29 by numerically integrating eq.([e6 ] ) to obtain @xmath27 . in the actual simulation ,
ten thousands of random initial conditions are chosen for the average , and the resulting error ( standard deviation ) of the data points is typically a few percent . from the conductivity tensor ,
the magnetoresistance @xmath33 and the hall resistance @xmath34 are expressed as @xmath35 @xmath36
2 , we show three typical trajectories found through a numerical simulation of the classical electron motion in periodic modulated magnetic systems ( type i , ii ) , chaotic , pinned and runaway trajectories@xcite .
the chaotic trajectory represents an ordinary diffusion mechanism .
most trajectories belong to this type of the chaotic trajectory which appear in all magnetic fields we examined .
the pinned trajectory is a non - diffusive trajectory which exists only near commensurability external magnetic fields ( @xmath37 ) .
it is confined around a modulated magnetic field for a long time , and does not contribute to the diffusion .
the runaway trajectory skips between the periodic modulation along @xmath0 or @xmath1 direction .
this trajectory appears suddenly from chaotic or pinned trajectories as the cyclotron diameter becomes commensurate with the lattice constant .
it does not persist for a very long time .
it is a chaotic trajectory in a strict sense .
figure 3 shows a typical result for the longitudinal resistance @xmath33 and the hall resistance @xmath34 in a type i model as a function of the externally applied field @xmath3 .
the period of the magnetic modulation is @xmath6 , the radius of the modulated field region @xmath19 is @xmath38 , and the strength of the modulated magnetic field @xmath24 is @xmath39 .
resistances are scaled by @xmath40 which is the resistance of the system without the field modulation and without the applied external field .
first , we report the behavior of the longitudinal resistance ( fig .
3 , the solid line ) .
we observe two features .
one is that the peaks of the longitudinal resistance clearly appear near the commensurability magnetic fields ( @xmath37 ) .
the other is the asymmetry of magnetoresistance about zero field .
these structures are characteristic in the periodically modulated magnetic fields .
we then change the strength of the modulated field @xmath4 .
the commensurability peak decreases with the decrease of a modulated field @xmath4 and disappears at low field ( fig .
the peaks disappear when the cyclotron diameter in the modulated region is larger than the diameter of the modulated region .
the magnetoresistance is influenced by the shape and strength of the modulation . in case of smooth modulated model ( type ii ) ,
the structure of the longitudinal resistance is similar to type i ( fig .
however , the peak of the smooth modulated model shifts to a lower magnetic field when we use the same value for @xmath24 as in type i. when the steepness parameter @xmath23 increases , the peak position approaches to the one in the type i model .
the hall resistance @xmath34 also shows peculiar features . in fig .
3 , we observe that the magnetoresistance peaks are accompanied by the abrupt increase of @xmath34 .
this is also observed in the type ii model ( fig .
5 ) . furthermore , in the type ii model
, @xmath34 shows three plateaus between the sharp changes .
there is the asymmetry of magnetoresistance about zero field .
note that the hall resistance is still positive when @xmath41 .
the step - like structure of @xmath34 in antidot systems has been analyzed in detail and explained by ishizaka and ando.@xcite they have introduced a hopping model to simplify the classical motion of electrons , and have classified the trajectories ( called _ components _ ) by the successive center coordinates of the antidots which scatter the electron .
the step of @xmath34 has been shown to arise from the cancelation of the contributions from the right going components and the left going ones . as seen from fig .
2 , the chaotic motions of the electrons in modulated magnetic fields can also be mapped onto such simplified components , though the scattering mechanism is completely different .
therefore , the step - like structure accompanied by the peak of the diagonal resistivity in this system is also explained according to the picture .
the non - vanishing value of @xmath34 at @xmath41 will be discussed in the next section .
we have also examined the other lattice structure , namely the triangular lattice . in this case
, runaway trajectory skips along @xmath42 or @xmath43 direction , and the peaks in the resistivity again appear at the commensurability fields .
the above results that peaks of the longitudinal resistance occur near the commensurability magnetic fields suggests that pinned or runaway trajectories contributes to the peaks .
the pinned trajectory at commensurate magnetic field can not explain the difference of magnetoresistance peak positions between positive and negative external field , nor the disappearance of the commensurability peaks at low modulated field , since the pinned trajectory is not influenced by the direction of the external field and the strength of the modulation .
an asymmetry of magnetoresistance about zero field and its dependence on strength of modulated magnetic field are explained only by the runaway trajectory .
this means that runaway trajectories shown in fig .
2 are essential to explain the simulation .
figure 6(a ) and ( b ) show typical runaway trajectories at fields given by eqs.([e81 ] ) and ( [ e82 ] ) , respectively .
when @xmath3 and @xmath4 are in the same direction , @xmath46 is the condition that a runaway trajectory appears , where @xmath47 is the cyclotron radius in the absence of modulation , and @xmath48 is the cyclotron radius in the modulated region . when @xmath3 and @xmath4 are in the opposite directions , the condition changes to @xmath49 .
the peak positions of the magnetoresistance obtained in the simulations are in good agreement with the field values given by eqs.([e81 ] ) and ( [ e82 ] ) , as indicated in fig . 3 by arrows . in smoothly modulated magnetic field
, electron motion is bended more strongly than in the absence of modulation . in this case
, the runaway trajectory appears in the lower field than the field expected from eqs.([e81 ] ) and ( [ e82 ] ) .
this is the reason why peaks for the type ii model shift to a lower field .
as stated in the previous section , @xmath34 remains to be positive even when the external field @xmath3 is vanishing .
to explain this behavior , we have introduced the effective magnetic fields .
we introduce an averaged magnetic field @xmath50 felt by electron as @xmath51 where @xmath52 is the position of the electron .
the obtained @xmath50 looks almost linearly dependent on @xmath3 and is well described by @xmath53 this effectively shifts the point of @xmath54 to a negative value of @xmath3 ( fig . 7 ) . here
@xmath55 and @xmath56 are constant and the values are summarized in table i. in type i model , @xmath55 increases and becomes almost equal to unity as @xmath24 increases .
this is because the electron can not stay in the modulated field region for long when @xmath24 is large .
this idea of the effective field qualitatively explains the positive value of @xmath34 at @xmath41 , since @xmath57 even if @xmath41 .
however , @xmath50 remains positive even when @xmath58 .
one might think that such vanishing of @xmath34 for positive @xmath50 can be explained by the vanishing of the hall resistivity@xcite , but we do not observe the plateau or non - monotonic behavior around @xmath58 which is typical for the vanishing @xmath34 . the fact that @xmath59 and @xmath60 means that the non - uniform magnetic fields act as scatterers as well as magnetic fields bending the electron trajectories and giving rise to the hall effect . to summarize
, we have studied the transport properties in two dimensional periodic magnetic fields by numerical simulation using classical model and explained the mechanism of magnetoresistance peaks accompanied by sharp changes of the hall resistance by the runaway trajectory .
runaway trajectories are found to play a very important role in periodically modulated magnetic fields similarly as in the antidot array systems . in real experiments
performed so far @xcite , attaching ferromagnets to produce modulated magnetic fields necessarily introduces a strain induced electric potential as well , while in our simulation , we have assumed that the effect of magnetic field modulation is much stronger than that of the scalar potential modulation and have neglected its effect .
furthermore , the field modulation is assumed to be stronger than the externally applied field .
such assumptions might be satisfied by introducing the modulated field by attaching type ii superconductor.@xcite density modulated composite fermion systems with @xmath61@xcite may also be a possible candidate .
99 for review article , see for example , c. w. j. beenakker and h. van houten : solid state phys .
* 44 * ( 1991 ) 1 .
m. l. roukes : phys .
* 59 * ( 1987 ) 3011 . c. w. j. beenakker and h.
van houten : phys .
* 63 * ( 1989 ) 1857 .
d. weiss , m. l. roukes , a. menschig , p. grambow , k. von klitzing and g. weimann : phys .
66*,(1991 ) 2790 ; for review article , see for example , d. weiss , g. ltjering and k. richter : chaos , solitons and fractals * 8 * ( 1997 ) 1337 .
k. tsukagoshi , s. wakayama , k. oto , s. takaoka and k. murase : phys .
b * 52 * ( 1995 ) 8344 .
k. tsukagoshi , t. nagao , m. haraguchi , s. takaoka , k. murase and k.gamo : j. phys .
. jpn . * 65 * ( 1996 ) 1914 .
r. fleischmann , t. geisel and r. ketzmerick : phys .
* 68 * ( 1992 ) 1367 . .
m. baskin , g. m. gusev , z. d. kvon , a. g. pogosov and m. v. ntin : jetp lett .
* 55 * ( 1992 ) 678 .
s. ishizaka and t. ando : phys .
b * 55 * ( 1997 ) 16331 .
t. nagao : j. phys .
* 66 * ( 1997 ) 3183 .
t. ando , s. uryu , s. ishizaka and t. nakanishi : chaos , solitons and fractals * 8 * ( 1997 ) 1057 .
t. nakanishi and t. ando : phys .
rev . b*54 * ( 1996 ) 8021 .
s. uryu and t. ando : in the proceeding of the 12th international conference on electronic properties of two dimensional systems ( ep2ds-12 ) , physica b * 249 - 251 * ( 1998 ) 308 ; s. uryu : ph.d .
thesis , university of tokyo ( 1997 ) .
s. izawa , s. katsumoto , a. endo and y. iye : j. phys .
. jpn . * 64 * ( 1995 ) 706 .
m. kato , a. endo and y. iye : j. phys .
* 66 * ( 1997 ) 3178 .
p.d.ye , d.weiss , k. von klitzing , k. eberl , h. nickel : appl .
* 67 * ( 1995 ) .
p. d. ye , d. weiss , r. r. gerhardts , m.seeger , k. von klitzing , k. eberl and h.nickel : phys .
* 74 * ( 1995 ) 3013
geim , s. j. bending and i.v .
grigorieva : phys .
. lett . * 69 * ( 1992 ) 2252 .
b. i. halperin , p.a .
lee and n. read : phys .
b*47 * ( 1993 ) 7312 .
rev . b*55 * ( 1997 ) 15552 .
n. nagaosa and h. fukuyama : to appear in j. phys .
fig . 1 : schematic view of the type i model ( fig .
1(a ) ) and type ii ( fig .
1(b ) ) . in case of type
i , an external constant magnetic field @xmath3 is applied perpendicular to the plane outside the region enclosed by circles and constant magnetic field @xmath24 exists inside the circles . in case of type ii
, the modulated magnetic field @xmath4 is expressed by eq.([e5 ] ) and changes smoothly .
3 : longitudinal and hall resistance in the type i model .
the modulation of periodic magnetic field forms square lattice ( @xmath62 ) .
solid line is the longitudinal resistance , while the dotted line is the hall resistance @xmath34 .
the magnetoresistance is normalized by the longitudinal resistance @xmath40 in zero field .
the modulated magnetic field @xmath63 .
typical standard deviations of the data points are indicated by error bars .
4 : magnetoresistance for various modulated magnetic field @xmath24 in the type i model . solid line shows longitudinal resistance with @xmath63 , while the dotted line is for @xmath64 , and the dashed line for @xmath65 .
typical standard deviations of the data points are indicated by error bars . at low field
, the peak structure disappears . fig . 5 : longitudinal and hall resistance in smoothly modulated fields ( type ii ) .
@xmath23 in eq.([e5 ] ) is set to be 4 .
the solid line shows longitudinal resistance and the dashed line shows hall resistance . the maximum magnetic field @xmath63 .
again , the typical standard deviations of the data points are indicated by error bars .
fig . 6 : typical runaway trajectory given by eqs.([e81 ] ) and ( [ e82 ] ) . in fig .
6(a ) , a typical runaway trajectory where @xmath3 and @xmath4 are in the same direction is shown , while in fig .
6(b ) , a typical runaway trajectory where @xmath3 and @xmath4 are in the opposite directions is shown . | ballistic transport properties in a two dimensional electron gas are studied numerically , where magnetic fields are perpendicular to the plane of two dimensional electron systems and periodically modulated both in @xmath0 and @xmath1 directions .
we show that there are three types of trajectories of classical electron motions in this system ; chaotic , pinned and runaway trajectories .
it is found that the runaway trajectories can explain the peaks of magnetoresistance as a function of external magnetic fields , which is believed to be related to the commensurability effect between the classical cyclotron diameter and the period of magnetic modulation . the similarity with and difference from the results in the antidot lattice
are discussed . |
the exceptional binary star , @xmath0 aurigae exhibits two year - long eclipses every 27.1 years .
the eclipse of the f supergiant primary star is now understood to be caused by an opaque disk hiding a companion star ( see stencel , 2012 and references therein ) . an international effort to monitor all aspects of
the 2009 - 2011 eclipse resulted in substantial archives of new data , including those involving high resolution spectroscopic monitoring , e.g. strassmeier et al .
( 2014 ) and others . here
we report on sampling of the high resolution spectroscopic archive generated during 2009 - 2011 eclipse , by the arces instrument at apache point observatory ( wang et al .
2003 ) . as discussed by strassmeier et al . ,
selected spectral lines exhibit eclipse or non - eclipse behaviors , and show differential velocity effects .
our purpose is to confirm and extend their findings , in light of continuing developments about the nature of this enigmatic binary star and disk system . among the developments resulting from study of the 2009 - 2011 eclipse was recognition of third contact phenomena , interpreted as a mass transfer stream impacting the disk ( griffin & stencel 2013 ) .
third contact occurred during spring 2011 , circa rjd 55,620 ( see stencel 2012 for eclipse contact times ) .
evidence for this stream was seen in low excitation fe i lines and , curiously , lines of rare earth elements , blueshifted up to 45 km / sec .
corroborating observations of spectral line radial velocity curves near third contact , showing similar effects , were reported by kambe et al .
( 2013 ) and sadakane et al .
strassmeier et al .
presumably would have detected similar spectral features if their robotic stella spectrometer had not been off - line during those portions of the eclipse .
arces observers maintained a steady cadence through all of the eclipse phases and thus provide a more detailed picture of third contact phenomena . in this paper
we will discuss the new spectroscopic observations ( sections 2 , 3 and 4 ) , followed by an evolutionary context ( section 5 ) , shape and cloudy models for the system ( sections 6 and 7 ) , and conclusions in section 8 .
the arces instrument is an echelle spectrograph commissioned in 1999 for use with the arc ( astrophysical research consortium ) 3.5 m telescope located at the apache point observatory [ http://www.apo.nmsu.edu/arc35m/instruments/arces/ - doi : 10.1117/12.461447 ] .
arces main optics consist of an off - axis parabolidal collimator , an echelle grating , two cross - dispersing prisms , and an f/2.7 schmidt camera with achromatic correctors . since its commission , arces has been primarily used for observations of stellar abundances and as a tool for surveying diffuse interstellar bands ( dibs ) .
important features of this instrument include : a 2.5 pixel resolution of 31,500 ( 9 km / s ) , a spectral range of 3200a-10000a , an s / n @xmath4 3000 , and remote operability .
in addition , the efficiency of the entire telescope plus spectrograph system has been determined to be greater than 2.2% at 647 nm and the spectrograph has been estimated to have an efficiency of 8% and 2% at 630 nm .
some limitations of this instrument include a slight drift throughout nightly observations of about 0.5 pixels over 10 hours .
this drift occurs mostly during the first half of the night and is likely due to thermal changes in the double - prism system as temperatures raise and lower .
this has been accounted for by interpolating values between hourly calibrations .
overall , arces has been proven to be a stable , reliable instrument that has many uses in astronomical observations . similarly , triplespec is a near - ir spectrometer built for the arc 3.5 meter telescope at apache point observatory ( apo : http://www.apo.nmsu.edu/arc35m/instruments/triplespec/1 ) .
spectral coverage ranges between 0.95 and 2.46 microns , with 2.1 spectral pixel resolution of 3500 with the 1.1 arcsec slit .
steps in the data reduction described by wang et al .
( 2003 ) include : ( 1 ) extracting spectra from raw ccd frames ; ( 2 ) applying deblazing and wavelength standardization ; ( 3 ) applying heliocentric rv - corrections , but no telluric corrections were attempted .
arces and triplespec observations of @xmath0 aurigae are listed in table 1 , including date and reduced julian date plus an estimate of the signal to noise ratio ( snr ) based on continuum signal statistics as described below . to achieve high - accuracy rv measurements with the echelle spectrograph
, we obtained a thorium - argon ( th - ar ) exposure after every science exposure .
the 9 km / s velocity resolution translates to a spectral precision of 0.15@xmath5 at 5000@xmath5 and 0.20@xmath5 at 6500@xmath5 .
observations of @xmath0 aurigae were obtained with arces beginning in 2009 feb 16 ( rjd 54878 , where rjd = jd - 2,400,000 ) and continuing every few nights , through 2011 dec 10 ( rjd 55905 ) .
these span key times during the eclipse cycle , including ingress ( 2009 aug 15 , rjd 55060 ) , second contact ( 2010 feb 15 , rjd 55250 ) , mid - eclipse ( 2010 july 30 , rjd 55400 ) , third contact ( 2011 march 15 , rjd 55620 ) and end of eclipse ( 2011 aug 30 , rjd 55800 ) . .arces and triplespec observation log for @xmath0 aurigae [ cols="^,^,^,^,^,^,^ " ,
shape 5.0 is a morpho - kinematic modeling and reconstruction tool used for astrophysical objects . with shape
, users can input any known physical parameters of their objects and can augment single bodies with clouds , disks , and other features .
it is also possible to specify chemical abundances as well as emission and absorption lines .
automatically created with each physical model is a velocity - space model , as well as a a graph of the object s velocity curve .
users also have the capability to change the frame through which the object is being viewed , by changing inclination and by rotating the object with respect to the observer .
one problem with shape was the tendency for files to be unsaved or deleted in between opening and working with them .
shape was used to interpret and model the mass transfer stream observed during third contact .
a small sphere was set in the center of the frame and surrounded by a disk , representing the b star and the accretion disk .
the primary f star is set in the background at roughly third contact and there are two separate `` clouds '' representing the observed mass transfer stream .
the cloud farther out from the disk represents the higher velocity component that is illuminated by the f star first , and then transitions into the slower , closer in cloud as the eclipse ends . the deceleration from -60 km / s to -40
km / s over the projected distance of one f star radius is indicative of an 8 solar mass central star inside the disk , assuming the disk radius scale derived by kloppenborg et al .
( 2015 ) at the 737 pc distance .
the mass function implies an f star mass of 6 solar masses ( q=0.78 ) . for the 1 kpc distance
, the implied mass inside the disk is 11 solar masses , and for the f star , 12 solar masses ( q = 1.1 ) .
one challenging aspect of the stream velocity at third contact is why it only appears between rjd 55600 and 55650 , which is only 0.3 percent of the orbital phase .
this suggests the higher speed stream closely parallels the edge of the opaque disk .
broadband polarization reported by kemp et al .
( 1986 ) spikes after mid - eclipse and during third contact , which suggests strong forward scattering of light by circum - disk material during those phases .
in addition to the optical spectra obtained by arces , described above , a co - mounted near - infrared spectrograph captured data on the same schedule ( table 1 ) .
triplespec provides wavelength coverage from 0.95 to 2.46 microns with a spectral resolution of 3500 using the 1.1 arcsec slit .
previously reported changes in the helium i 10830 line ( stencel et al .
2011 ) reveal a large increase in absorption equivalent width during mid - eclipse .
the more extensive coverage with triplespec confirms this behavior and enables study of details of the mid - eclipse , per figure [ fig : tspec ] , including a suggestion of a local minimum near rjd 55475 , roughly 1 au past mid - eclipse .
this offset could relate to the position of the disk inner wall .
line during the eclipse of @xmath0 aurigae , as seen with triplespec .
variation of the pa @xmath1 line profile during third contact phase of the eclipse of @xmath0 aurigae , as seen with triplespec.,title="fig : " ] line during the eclipse of @xmath0 aurigae , as seen with triplespec .
variation of the pa @xmath1 line profile during third contact phase of the eclipse of @xmath0 aurigae , as seen with triplespec.,title="fig : " ] equally interesting was the appearance , starting after mid - eclipse and especially during third contact , of a p cygni profile in the paschen @xmath1 line at 1.28 microns .
the blue - edge velocity is approximately 100 km s@xmath6 , presumably arising from the energetics due to the stream - disk collision .
cloudy is an internationally used spectral synthesis code designed to simulate conditions in plasmas , such as circumstellar and interstellar matter , exposed to a radiation field under a broad range of conditions .
cloudy also has the ability to predict emission and absorption spectra , which can be compared with the observed data . extending over the whole electromagnetic spectrum ,
cloudy includes the 30 lightest elements in its calculations and the relative concentrations of these elements can be modified .
the ionic and molecular emission data used in cloudy comes from the chianti database , the lamda database as well , as its own atomic and molecular database with level energies taken from nist ( turner et al .
2016 ) . for our purposes ,
cloudy was used to interpret the equivalent widths of two lines that experienced broadening and strengthening during the eclipse : he 18030and h-@xmath7 . to generate equivalent widths from the cloudy code ,
certain parameters had to be specified in order to generate a continuum file that would contain the calculated intensities over our selected range of wavelengths . for the input parameters we varied log temperature and log density , to see their effect on the line strength , while other parameters such as geometry , radius , and chemical abundances were not varied .
temperature was run from 15000k to 30000k with a step size of 2500k , or in log terms , from 4.17 to 4.47 .
the log of the hydrogen density was varied from 8 to 12 with a step size of 0.5 . for our code
, we assumed an open geometry with an inner and outer radius set at 1au and 5au , respectively .
chemical abundances were assumed to be the same as solar abundances ( grevesse et al .
the code was iterated twice and the continuum results were exported to a text file .
equivalent width was plotted as a function of log density and log temperature as can be seen in figures 10 and 11 .
the blue space shows the deviation from the observed maximum equivalent width of helium i 10830 which was 2 .
a determination for disk density was given by stencel et al .
( 2015 ) by fitting transient co lines seen near third contact .
the result was that log n = 10.5 .
figures 10 and 11 suggest that for log n = 11 then t=15000k provides a good fit for the equivalent width of both he 10830 and for h - alpha .
we conclude this temperature is consistent with a b5v central star .
[ h ] , with a step - size of 0.5 . ]
[ h ] varies with different densities and temperatures .
temperatures were run from 15000 k to 30000 k and the log of the hydrogen density ranged from 8 to 12 @xmath2 . ]
spectroscopic monitoring of @xmath0 aurigae during its recent eclipse , using the apache point observatory arces and triplespec instruments , has helped detail substructure in the disk and stream at selected phases . in particular ,
the amplitude of the disk velocity is better defined and the appearance and velocity of the stream is more well defined .
variations in the he i 10830 absorption line strength during mid - eclipse was shown to relate to the inner disk structure .
hydrogen pa@xmath1 exhibits a p cygni profile during third contact , suggestive of an expansion zone related to stream impact zone .
having defined special phases around third contact , ideally more targeted observations can be arranged during the next third contact time - frame , in particular ( rjd 65510 or 2038 march 28 ) . one hope resulting from study of these data is another attempt to determine the distance to the @xmath0 aurigae system - much debated in the literature .
recent work by guinan et al .
( 2012 ) claimed to constrain distance estimates to @xmath8 kpc , using correlations between distance measurements to stars in the vicinity of @xmath0 aur with interstellar absorption and reddening properties .
however , that study rests on several assumptions , including uniformity of the interstellar medium ( ism ) , enforcing strictly linear correlations between distance - sensitive properties , and spectral type calibration uncertainties that factor into their mesa models and distance moduli .
distances based on different dibs disagree with each other , so reliance on a single one also can be highly suspect friedman et al .
each of these assumptions contribute to a substantial cumulative uncertainty in their distance estimate .
it is instructive to note that several binary stars exist in the @xmath0 aurigae field .
for example , b - type stars are fairly common within a few degrees , and indications are that the @xmath0 aurigae primary could have arisen from a b - type main sequence star , and its disk may contain a b - type secondary star .
similarly , @xmath0 aurigae has a moderate eccentricity @xmath9 stefanik et al .
two neighbor stars , which also are guinan et al . sample stars , turn out to be high eccentricity b - star binaries . these stars are hd 31894 ( e=0.61 , estimated stellar mass of 9m@xmath10 ) and hd 31617 ( e=0.76 , estimated primary stellar mass of 9 to 15m@xmath10 ) .
the formation of such binaries represents a challenge to star forming theories .
another neighbor , hd 30353 ( a5iap ) , is suggested to be a case c mass transfer binary ( lauterborn 1970 ) .
collectively , are these clues to binary star evolution in this region of our galaxy ?
we argue that this may be so , and explore the implications thereof .
the distance moduli for these b star binaries and angular proximity to @xmath0 aurigae argue for a distance less than 1kpc .
acknowledgements : this work was supported in part by the bequest of william herschel womble in support of astronomy at the university of denver .
thanks the university of denver undergraduate research center for a partners in scholarship ( pins ) grant , and the native american scholarship fund that helped his efforts in this research .
the authors are grateful to the astronomical research consortium ( arc ) that operates the apache point observatory 3.5 meter telescopes and its arces and triplespec instruments that provided the bulk of the data reported in this paper .
* casini , r. ; bevilacqua , r. ; lpez ariste , a. , 2005 .
apj , 622 , 1265 . principal component analysis of the he i d3 polarization profiles from solar prominences .
* friedman , scott d. ; york , donald g. ; mccall , benjamin j. ; dahlstrom , julie ; sonnentrucker , paule ; welty , daniel e. ; drosback , meredith m. ; hobbs , l. m. ; rachford , brian l. ; snow , theodore p. 2011 . the astrophysical journal , 727 , 1 .
studies of diffuse interstellar bands v. pairwise correlations of eight strong dibs and neutral hydrogen , molecular hydrogen , and color excess .
* griffin , r.e.m . and
stencel , r.e .
2013 pasp 125 , 775 .
merging recent and historic spectra of aurigae : properties of the system s components , and discovery of a mass transfer stream .
* grevesse , n. ; asplund , m. ; sauvahttp://adsabs.harvard.edu / abs/2011apj ... 727 ... 33fl , a. j. ; scott , p. 2010 , astrophysics and space science , volume 328 , issue 1 - 2 , pp .
179 - 183 , the chemical composition of the sun .
* guinan , e. f. ; mayer , p. ; harmanec , p. ; boi , h. ; bro , m. ; nemravov , j. ; engle , s. ; lechta , m. ; zasche , p. ; wolf , m. ; korkov , d. ; johnston , c. 2012 .
astronomy and astrophysics , volume 546 , id.a123 .
large distance of aurigae inferred from interstellar absorption and reddening + http://adsabs.harvard.edu/abs/2012a%26a...546a.123g * hoard , d. w. ; howell , s. b. ; stencel , r. e. 2010 astrophys .
journal 714 , 549 . _ system parameter constraints for epsilon aurigae . _ http://adsabs.harvard.edu/abs/2010apj...714,549 * kambe , e. , sadakane , k. , hashimoto , o. , honda , s. and sato , b. 2013 in helio- and astero - seismology , eds .
h.shibahashi and a.e .
lynas - gray , asp conf .
vol.469 , p.203 .
line - profile variations of the primary of the epsilon aurigae eclipsing binary system .
* kemp , j. c. ; henson , g. d. ; kraus , d. j. ; beardsley , i. s. ; carroll , l. c. ; ake , t. b. ; simon , t. ; collins , g. w. 1986 astrophys .
journal 300 , l11 .
* kloppenborg , b. , stencel , r. , monnier , j. et al .
2015 apj suppl .
220 , 14 + http://adsabs.harvard.edu/abs/2015apjs..220...14k
* lauterborn , d. 1970 .
astronomy and astrophysics , vol . 7 , p. 150
evolution with mass exchange of case c for a binary system of total mass 7 m sun .
* mennickent , r. e. 2014 pasp 126 , 821 .
_ evolutionary stage of the interacting binary au mon . _ http://adsabs.harvard.edu/abs/2014pasp..126..821m * sadakane , k. , kambe , e. , hashimoto , o. , honda , s. and sato , b. 2013 pasj 65 , 1 .
disk origin narrow metallic absortion lines observed during the 2009 - 2011 eclipse of aurigae .
* skumanich , a. , lopez ariste , a. the astrophysical journal , 570:379386 , 2002 may 1 .
the physical content of the leading orders of principal component analysis of spectral profiles .
+ http://iopscience.iop.org/article/10.1086/339503/pdf * stencel , r. , 2012 , jaavso 40 , 618 , @xmath0 aurigae - an overview of the 2009 - 2011 eclipse campaign .
http://www.aavso.org/sites/default/files/jaavso/v40n2/618.pdf * strassmeier , k. , weber , m. ; granzer , t. ; schanne , l. ; bartus , j. ; ilyin , i. 2014 astron nachrichten 335 , 904 .
time - series high - resolution spectroscopy and photometry of aurigae from 2006 - 2013 .
* struve , o. and elvey , c. , 1930 , ap.j .
71 : 136 . preliminary results of spectrographic observations of 7 epsilon aurigae .
http://adsabs.harvard.edu/abs/1930apj....71..136s * turner , jake d. ; christie , duncan ; arras , phil ; johnson , robert e. ; schmidt , carl .
2016 monthly notices of the royal astronomical society , volume 458 , issue 4 , p.3880 - 3891
. investigation of the environment around close - in transiting exoplanets using cloudy . //
* van rensbergen , w. ; de greve , j. p. ; mennekens , n. ; jansen , k. ; de loore , c. , 2011 astron . astrophys .
mass loss out of close binaries .
the formation of algol - type systems , completed with case b rlof . //
wang , s .-
, hildebrand , r. , hobbs , l. , heimsath , s. , kelderhouse , g. , loewenstein , r. , lucero , s. , rockosi , c. , sandford , d. , sundwall , j. , thorburn , j. and york , d. 2003 , in m. iye and a. f. m. moorwood , editors , instrument design and performance for optical / infrared ground - based telescopes , volume 4841 of society of photo- 43 optical instrumentation engineers ( spie ) conference series , pages 1145 - 1156 .
an echelle spectrograph for the astrophysical research consortium ( arc ) 3.5 m telescope .
doi : 10.1117/12.461447.arces | context : worldwide interest in the recent eclipse of @xmath0 aurigae resulted in the generation of several extensive data sets , including those related to high resolution spectroscopic monitoring .
this lead to the discovery , among other things , of the existence of a mass transfer stream , seen notably during third contact . + aims : we explored spectroscopic facets of the mass transfer stream during third contact , using high resolution spectra obtained with the arces and triplespec instruments at apache point observatory .
+ methods : one hundred and sixteen epochs of data between 2009 and 2012 were obtained , and equivalent widths and line velocities measured , selected according to reports of these being high versus low eccentricity disk lines .
these datasets also enable greater detail to be measured of the mid - eclipse enhancement of the he i 10830 line , and the discovery of the p cygni shape of the pa-@xmath1 line at third contact .
+ analysis : we found evidence of higher speed material , associated with the mass transfer stream , persisting between third and fourth eclipse contacts .
we visualize the disk and stream interaction using shape software , and use cloudy software to estimate that the source of the enhanced he i 10830 absorption arises from a region with log n = 10 ( @xmath2 ) and temperature of 20,000 k consistent with a mid - b type central star .
+ results and next steps : van rensbergen binary star evolutionary models are somewhat consistent with the current binary parameters for the case of a 9 plus 8 solar mass initial binary , evolving into a 2.3 and 14.11 solar mass end product after 35 myr .
prior to the next eclipse , it is possible to make predictions which suggest that continued monitoring will help resolve standing questions about this binary . |
there are currently many different avenues to a theory of quantum gravity .
this paper focuses on the formulation provided by dynamical triangulations .
extensive work has been done in two dimensions and a significant body of work has also examined three and four dimensions @xcite .
dynamical triangulations are similar to the formulation of regge in that the manifold is broken up into simplexes and the curvature is identified with a deficit angle about a subsimplex with dimensionality two less than that of the manifold .
the formulations differ in their treatment of the functional integral over all possible metrics . in the regge approach , this is provided by varying the link lengths on a fixed lattice . in the dynamical triangulation approach ,
the link lengths are held fixed but the lattice is varied : all possible triangulations are summed over .
monte carlo evaluations of the sum are performed by randomly wandering through the space of triangulations .
it has been proven that any triangulation can be reached from any other triangulation ( for a given topology ) through a sequence of local moves @xcite .
a simple algorithm allows the computation of functional integrals over metrics in any desired dimension with a computer program only 350 lines long @xcite .
the simplest observables in these models are the number of subsimplexes of various dimensions ( e.g. nodes , links , triangles , etc . ) and these numbers can be used to construct an action . they are related due to topological constraints and the requirement that the triangulations yield a manifold : in three and four dimensions the number of nodes and the number of simplexes determine all of the other numbers of subsimplexes . the basic action used in these theories is @xmath0 where @xmath1 and @xmath2 are chemical potentials corresponding to newton s constant and the cosmological constant , respectively , while @xmath3 is the number of nodes and @xmath4 is the volume . in two dimensions , @xmath3 and @xmath4
are related so there is only one term in the action .
in fact , it is possible to work in an ensemble where the volume is fixed ( leaving zero terms in the action ) .
this is possible because ergodicity has been proven for a restricted set of update moves that do not alter the volume . in three and higher dimensions
there is no corresponding proof : the full ensemble including volume fluctuations must be used @xcite . to recover a viable continuum theory from the lattice theory requires two limits to be taken
first , there is the thermodynamic , large volume , limit .
it is reached by adjusting @xmath2 .
when @xmath2 is large , the volume is small .
as @xmath2 is decreased , the volume increases until a critical value of @xmath2 is reached at which the volume diverges . in practical calculations , @xmath2 must be taken slightly larger than the critical value and there are finite size effects .
the other limit that must be taken is the continuum limit .
a non - zero correlation length occurs in the continuum , where the lattice spacing is taken to zero , only when the correlation length in lattice units is infinite , so there must be a second order phase transition present in the phase diagram of the theory . in spin theories and lattice gauge theories , the block spin renormalization group approach is a natural tool with which to study continuous phase transitions @xcite .
this paper develops a renormalization group approach for dynamical triangulations .
it is used here to examine the flow of couplings , but it also provides for the systematic calculation of critical couplings and critical exponents if a second order transition can be found . as an excellent example of the potential of the renormalization group approach see the application of it to the three - dimensional ising model in @xcite . before applying this idea to the lattice quantum gravity models , it is useful to consider the differences between the ising model and a dynamically triangulated model . in the former
there is a fixed lattice , of linear dimension @xmath5 , on which the degrees of freedom are arranged in a regular way .
it is simple to draw a courser ( block ) lattice , of linear dimension @xmath6 , as a subset of the original one which partitions the spins into identically structured groups ( such as @xmath7 by @xmath7 blocks ) .
these groups can then be averaged according to a specified rule in order to convert the original spin configuration @xmath8 , composed of @xmath9 spins , into a block spin configuration @xmath10 , composed of @xmath11 spins .
in contrast , in a dynamically triangulated model the lattice is not fixed and the lattice structure itself embodies the degrees of freedom .
physically , triangulations are interpreted as embodying information about the metric structure of spacetime : points connected by a link are closer than points not connected by a link .
presumably this should be true for block triangulations as well and , in two dimensions , the condition can be imposed easily to define a block spin renormalization group transformation @xcite .
this is accomplished as follows . in two dimensions at fixed volume
the set of local update moves is a single move , the link flip .
view a link and the two triangles that share it as a square with one diagonal drawn .
a link flip simply replaces the diagonal with the alternative one .
it is the simplicity of the link flipping move that makes it easy to define the block triangulations . beginning with the original triangulation , @xmath12 , illustrated with thin lines in fig .
1 , the @xmath13 block nodes are defined as some subset of the original nodes .
a block triangulation , @xmath14 , illustrated with thick lines in the figure , is required to have the property that the minimum number of links on the original lattice required to travel from one block node to another block node connected to it by a block link must be smaller than the corresponding number of links that must be traversed to travel between the block nodes associated with the alternative diagonal .
block links are flipped until this requirement is satisfied .
once an initial block triangulation has been obtained , updates of the original lattice are followed by updates of the block lattice so as to maintain the rule .
the simplest way to obtain an initial block triangulation is to start with a regular original triangulation where it is easy to generate a course lattice with the required property .
matter can also be incorporated into this scheme .
this renormalization group transformation gives quite good results for coupling constant flows and critical couplings , but does not do well on critical exponents .
a variety of subsequent renormalization group transformations have been proposed @xcite .
the most successful is the node deletion blocking scheme which is very simple and gives excellent results @xcite .
the idea , still working in two dimensions , is to start with an original triangulation with @xmath3 nodes and @xmath15 triangles ( @xmath16 in two dimensions with @xmath17 the euler characteristic of the surface ) and to delete @xmath18 nodes in order to produce a block triangulation with @xmath19 nodes and @xmath20 triangles .
consider the deletion of a single particular node labeled @xmath21 .
the node @xmath21 belongs to a number of triangles forming a set to be referred to as @xmath22 .
the set @xmath22 forms a two - dimensional ball around the node @xmath21 .
the way @xmath21 is eliminated is by removing the entire ball @xmath22 from the manifold and then filling the resulting hole back in with new triangles , but without any new nodes .
the following algorithm is used to accomplish this . a link attached to @xmath21
is chosen at random and , if it is an allowed move , flipped . as a result
two old triangles are removed from the set @xmath22 and one new one is introduced , decreasing the total by one .
links are repeatedly randomly chosen and flipped until there are only three triangles remaining in the ball .
then the node and the remaining three triangles are replaced with one new triangle .
the original ball has been completely removed and the hole has been refilled .
2 schematically represents the procedure .
@xmath18 randomly chosen nodes are deleted in this way and the resulting lattice is interpreted as resulting from a block spin renormalization group transformation . when the standard renormalization group formalism is used to extract critical couplings and exponents for the two - dimensional ising model coupled to dynamical triangulations , the results agree with the known exact solutions .
since the node deletion procedure works so well in two dimensions , it is natural to try to apply it in higher dimensions .
as mentioned earlier , it is possible to formulate and implement the update moves for triangulations in such a way that the dimensionality of the simplexes is an easily modified variable .
the purpose of this paper is to put the node deletion scheme in this more general setting . obviously , it is simple to identify and eliminate all of the simplexes associated with a given node regardless of the dimension .
the tricky part is filling the hole back in .
the reason two dimensions at fixed volume is different from a general dimension , @xmath23 , is that the former has only one update move ( the link flip ) while the latter , using the formulation in @xcite , has @xmath24 .
if @xmath25 ranges from @xmath26 to @xmath23 , then there is a move , for each @xmath25 , that replaces @xmath27 simplexes with @xmath28 new ones .
@xmath25 is the dimension of a subsimplex that must be chosen to make the move .
for instance , a link flip has @xmath29 and @xmath30 which means a one - dimensional subsimplex ( the link ) must be chosen to make the move and two old simplexes are replaced with two new ones . in two dimensions , the simplexes associated with a chosen node are eliminated by making link flips which remove simplexes from the node s association until the minimal number are left .
the node is then eliminated .
this final step is actually an @xmath31 move . in higher dimensions ,
the set of simplexes associated with a node can be reduced using the @xmath32 moves until the minimal number is reached and a final @xmath31 moves completes the operation . by setting up the problem in this way ,
it is possible to write a program that does node deletion for a simulation of dynamical triangulations in any dimension .
changing dimensions requires the alteration of a single character in the program .
this section gives details which can be skipped , but which may be helpful to those who wish to implement the scheme . to produce a block configuration , start by copying the original configuration , along with all relevant bookkeeping , into new data structures .
the update routines described in @xcite can be modified so that they take arguments specifying which copy of the lattice information is to be changed . in this way
, exactly the same routines can be used to update the original lattice and to eliminate nodes in the blocked lattice . aside from being convenient , this is a good way to avoid errors . in the case of the block lattice ,
the update moves are not used to generate configurations distributed in accordance with the action but to fill a hole in the manifold made by eliminating a node and its associated simplexes .
create a list of all of the simplexes associated with the node that is to be eliminated .
these simplexes form a ball , @xmath22 . just as in two dimensions where links attached to the node are chosen at random and flipped ( when consistent with the geometrical constraints ) so as to reduce the number of associated simplexes , in arbitrary dimensions interior @xmath25-subsimplexes are chosen at random and the update move labeled by @xmath25
@xmath25 can not be zero until the node itself is finally removed .
the interior subsimplex in question is chosen by choosing one of the simplexes in the list at random and then choosing @xmath28 of its nodes at random .
the other @xmath33 nodes specify @xmath33 neighboring simplexes because for a marked node in a given simplex there is one other simplex that shares all of the nodes in the given simplex except for the marked one .
these neighbors play a role in the algorithm of @xcite .
each of the @xmath33 neighboring simplexes has a new node not found in the given simplex and all @xmath33 of these new nodes must be the same if the contemplated move is an allowed one .
if the contemplated move passes this test it is referred to as legal " .
the @xmath33 neighbors play an additional role here in that they must all be in the list of simplexes associated with the node .
if they are not , the chosen subsimplex is not interior to the ball . altering it
would therefore not fill the hole , but would alter its boundary . until the last ( node eliminating ) move
, @xmath25 can be any number from @xmath34 to @xmath23 . in two dimensions at fixed volume , where @xmath30 is the only possibility
, there is no choice to be made .
in general @xmath23 , many different algorithms for choosing @xmath25 are possible ; this is just the freedom to choose different kernels in the renormalization group transformation .
some may be apt and some may not .
only a single choice for the kernel is explored in this paper .
it is intended to eliminate the ball of simplexes as rapidly as possible .
since an @xmath25 move removes @xmath35 simplexes , it is likely that the quickest way to remove the ball is to choose moves with @xmath25 as small as possible .
this is not quite as obvious as it seems because , when a move is made in the interior of the ball , additional simplexes are lost : some of the new ones are no longer associated with the node and therefore do not belong in @xmath22 .
@xmath31 is not an option until the last move , so @xmath30 is probably the best move .
however , sometimes no @xmath30 move is allowable , anywhere in the ball , because of pre - existing connections on the surface of the ball due to exterior simplexes .
in fact , sometimes the only possible move is node insertion , @xmath36 .
consequently , the algorithm used here looks for and tries all possible @xmath30 moves , then does the same for all possible @xmath37 moves and so on up to @xmath38 . if no such moves are possible , a random simplex is chosen from the ball and a node is inserted .
after each interior update is made , membership in the ball ( defined by having the targeted node as a vertex ) is re - examined .
then another interior update is made .
this is repeated until the minimal number ( @xmath39 ) of simplexes are associated with the node which is then removed with an @xmath31 move . for each move
type , all of the simplexes ( in random order ) are checked for appropriate possibilities .
this is done using the fact that a legal move of type @xmath25 involves @xmath33 neighbors that share the same new node . for a chosen simplex in the ball ,
all of the new nodes from its neighbors ( i.e. those conjugate to nodes in the chosen simplex ) that are in the ball are placed in a list which is checked for multiplicity .
each new node with the proper multiplicity correspond to a possible move .
these possible moves are drawn from at random and checked to see if the geometrical constraints are satisfied . if so , the move is made . if not , the list of possible moves is drawn from again .
if the list is empty then no move at the given value of @xmath25 is possible and the procedure must be repeated for the next value of @xmath25 . instead of removing a node and its associated simplexes , one might consider removing a link and its associated simplexes or , generally , an @xmath40-dimensional subsimplex and its associated simplexes . in two dimensions ,
a node is associated with the curvature . in general , a @xmath41-subsimplex is associated with the curvature .
perhaps the appropriate generalization of removing a node in two dimensions is eliminating a @xmath41-subsimplex in @xmath23 dimensions .
if the subsimplex chosen for deletion is of dimension @xmath40 rather than zero , essentially the same procedure as above can be used to fill the hole generated by removing all of the simplexes associated with it .
the most important difference is that now only moves of @xmath42 are allowed until the very last move when @xmath43 is allowed . in particular , if @xmath44 , i.e. subsimplexes associated with the curvature are eliminated , then the lowest allowed value of @xmath25 is @xmath45 until the last move .
@xmath46 adds @xmath47 simplexes each move . in two dimensions the volume change is zero
which is consistent with the ensemble used there . in higher dimensions the volume increases . one generally expects a renormalization group transformation to produce a smaller system , one with fewer degrees of freedom , so more thought would be necessary to interpret the results in the case of @xmath44 with @xmath48 .
even @xmath49 empirically appears to produce larger volumes rather than smaller ones in three and four dimensions . throughout the rest of the paper
, @xmath40 will always be zero .
while deletion of one node is interpreted as a renormalization group transformation , it is generally useful to delete a number of nodes .
perhaps the most naive way to choose the node is to choose a simplex at random and then one of its nodes at random .
however , the node chosen in this way is chosen with a probability proportional to its order " , defined to be the number of simplexes that contain it .
choosing different nodes with different probabilities could cause unwanted distortions in the effective theory .
one remedy is to include a step which only accepts the node with probability equal to the inverse of its order . in this paper , since a large number of deletions are typically used , a list of all of the nodes is constructed and chosen from at random .
a brief study suggested that these alternatives have very similar characteristics , both in speed and in results .
note that two dimensions is special in that the number of nodes and the number of @xmath23-simplexes ( triangles ) is related : deleting a node requires the deletion of exactly two triangles .
this is useful , because in the block spin renormalization group approach there is typically a small target volume which one wishes to block to .
comparing expectation values only when they are produced on the same size lattice eliminates finite size effects . for instance , if a system of @xmath50 simplexes is blocked down to @xmath51 simplexes and the expectation values of that blocked theory are compared to those of a system of @xmath52 simplexes blocked down to @xmath51 simplexes , any differences in expectation values of operators are not due to finite size effects ( the two systems are the same size ) but to differences in the effective theories produced in the two cases .
it is easy to see that @xmath53 nodes must be eliminated in the first case and @xmath54 in the second . in higher dimensions ,
the relationship between the number of nodes and the number of simplexes is dynamically determined .
eliminating all of the simplexes associated with a given node can delete anywhere between @xmath39 simplexes and a number of them comparable to the volume . to target the volume
, it would be necessary to perform a sequence of runs , adjusting the number of deleted nodes each time until finally reaching the desired volume .
it is more convenient instead to target the node number . for a given coupling ,
the number of nodes at various volumes is either known from previous work or can be obtained with a single run . taking the difference between the starting number of nodes and the target number of nodes gives a very good estimate of the number of nodes that must be deleted from the larger lattice to produce the same number as the smaller lattice .
a minor adjustment is often still necessary , since node insertion moves are occasionally needed in the process of eliminating the ball of simplexes associated with a deleted node .
alternatively , instead of deleting a fixed number of nodes , nodes can be deleted until a target node number is reached .
the number of deleted nodes then becomes a fluctuating quantity while the final number of blocked nodes is fixed .
a very brief study found that these two schemes give identical answers .
regardless of which scheme is used , node numbers can typically be matched only to an accuracy of @xmath55 , since the number of eliminated nodes is an integer while the initial average number of nodes and the target number of nodes are not .
this can be patched with an interpolation of data from two runs , or by noting that the node density is nearly constant over such small changes in the node number .
consistent results are obtained in all cases .
node deletion in two dimensions at fixed volume has been studied in @xcite . for pure gravity , the string susceptibility exponent and
the hausdorff dimension were calculated .
flow toward a fixed point was verified using operators built from the local curvature and a measure term was added and shown to be irrelevant . for ising matter , a fixed point was obtained at the ( analytically known ) critical coupling and a variety of exponents ( the string susceptibility exponent , the magnetic exponent , and the thermal exponent ) were correctly computed .
these results demonstrate that the node deletion renormalization group transformation is quite apt in two dimensions . in two dimensions ,
once the infinite volume limit is obtained by adjusting the cosmological constant , the theory is also critical . in higher dimensions
an additional parameter ( at least ) must be tuned .
given the action , eqn .
( [ action ] ) , and tuning the cosmological constant to give infinite volume , the only variable left with which to search for a continuous phase transition , and therefore a continuum limit , is @xmath1 . in both three and four dimensions
there is a phase transition as a function of @xmath1 from a phase where the ratio of the number of nodes to the volume goes to zero as the volume increases to a phase where the same ratio approaches a non - zero constant . in three dimensions
the phase transition is first order while in four dimensions the phase transition has been thought to be second order although now there is some evidence that it is first order @xcite .
the renormalization group procedure detailed in the last section allows us to study the renormalization group flow of these higher - dimensional theories .
two operators are used to study the renormalization group flows : the volume , @xmath4 , and the logarithm of the order of the nodes , @xmath56 , defined as @xmath57 the latter operator is associated with a measure term in the sense that if a term of the form @xmath58 is added to the action , it corresponds to a continuum measure of the form @xmath59 where @xmath60 is the determinant of the metric @xcite .
the target number of nodes used here and in the following sections is the number of nodes associated with an unblocked system with @xmath61 simplexes .
smaller systems have the problem that when a large system is blocked down to that size using a fixed number of node deletions the large fluctuations in the block node number occasionally reach zero . fig .
[ fourd ] shows the expectation values of the two operators as a function of the degree of blocking for various couplings .
all of the flows begin in the lower left hand corner , where @xmath62 .
subsequent points in the plot correspond to systems with initial volumes of @xmath51 , @xmath53 , @xmath63 , and @xmath64 simplexes , all of which have been blocked down to the indicated number of nodes .
lines have been drawn through points sharing a given initial value of @xmath1 .
the values of @xmath1 are ( from right to left ) @xmath65 , and @xmath66 and are chosen to span the transition region .
all runs were for at least a million sweeps each .
errors in the blocked volume , as a function of initial volume , are typically 0.05 , 0.8 , 1 , 3 , and 4 , respectively , with somewhat smaller errors for large @xmath1 and larger errors for small @xmath1 .
the corresponding errors in @xmath67 are of order 0.05 , 0.2 , 0.4 , 0.4 , and 0.6 .
although all of the flows initially head toward the same vicinity , they then split up . for @xmath68 and @xmath69 ,
the volume quickly increases with the number of deleted nodes , meaning that the node density quickly flows to zero as the renormalization group transformation is iterated .
this is not surprising , since it is the expected large distance behavior in the crumpled phase .
for @xmath70 the flow turns toward smaller volumes although it is not clear how far it will go : for a given number of nodes , the volume has a lower bound . for @xmath71 ,
the volume stays constant for many iterations of the renormalization group , but continues to flow slowly toward increasing @xmath56 .
the slow flow near the end of the @xmath71 line suggests that a fixed point is nearby and could be reached with a small adjustment in the action . if @xmath1 is ruled to be insufficient , @xmath72 may be the appropriate parameter to vary .
the renormalization group formalism itself can determine the adjustments necessary in @xmath1 and @xmath72 to obtain a fixed point , but much higher statistics will be necessary to get a useful result ( the last two points on the @xmath71 line are only 1.5 standard deviations apart ) . in four dimensions
the peak in the node susceptibility with 8000 simplexes ( which moves toward greater @xmath1 as the volume is increased ) occurs around @xmath73 @xcite .
more recent estimates include transition couplings in the range of @xmath74 to @xmath75 depending on assumptions @xcite .
the results in the figure are compatible with these estimates .
since the first order character of the transition is much more firmly established in the case of the three - dimensional dynamical triangulation model , it is interesting to look at the renormalization group flows for this model and to see how they compare to the four - dimensional one .
[ threed ] shows the renormalization group flow lines for the three - dimensional dynamical triangulation model for six values of @xmath1 : @xmath76 and @xmath77 ( labeling the flows from right to left ) spanning the transition region .
the errors , as a function of initial volume , are typically 0.1 , 0.4 , 0.8 , 1.0 , and 1.2 for the blocked volume and 0.04 , 0.1 , 0.4 , 0.7 , and 0.7 for @xmath67 .
the flows for the first two values of @xmath1 flow strongly toward large volumes ( and therefore small node densities ) as is expected in the crumpled phase .
higher values of @xmath1 flow toward smaller volumes , similar to the smooth phase behavior in four dimensions .
there is no intermediate case for the chosen set of initial couplings and therefore no indication of a nearby fixed point .
this is consistent with the strongly first order character of the transition .
finally , we consider the addition of a measure term to the action . in @xcite , they suggest that the scaling behavior at the transition changes with the presence of a measure term , @xmath78 recent work suggests that the new coupling provides an expanded phase diagram in which there are lines of first order transitions in both three and four dimensions @xcite . a possible phase diagram is given in fig .
[ phase ] .
the question is : is there a finite value of @xmath79 ?
if so , the endpoint is presumably a second order phase transition at which a continuum limit can be taken .
the nearly second order nature of the transition in four dimensions suggests that @xmath80 in that case . in three dimensions studies of the specific heat
suggest @xmath81 @xcite . as a final illustration of the the node deletion scheme ,
the renormalization group flows are determined for @xmath82 results are shown in fig .
[ wmeas ] .
typical errors , as a function of the initial volume , are 0.1 , 0.4 , 0.6 , 1 , and 2 for the blocked volume and 0.04 , 0.1 , 0.1 , 0.2 , and 0.2 for @xmath67 .
all of the flows , for several very large couplings ( @xmath83 , @xmath84 , and @xmath85 ) , where one would expect to find the smooth phase , flow toward small node densities .
the largest value of @xmath1 at the largest blocking level does begin to move to the left .
this is not necessarily incompatible with a scenario of flows beyond the end of a first order line , but it may be that as @xmath72 is made increasingly negative it is necessary to go to larger volumes to see the transition . in this case
@xmath79 would be less than @xmath86 . | _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ a block spin renormalization group approach is proposed for the dynamical triangulation formulation of quantum gravity in arbitrary dimensions .
renormalization group flow diagrams are presented for the three - dimensional and four - dimensional theories near their respective transitions .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ .25 in pacs numbers : 04.60.+n , 05.70.jk , 11.10.gh .25 in = 10000 = 10000 |
within the growing field of sociophysics ( see @xcite for the defining paper , and @xcite-@xcite for an impression of the recent state of the art ) , two- and three states galam opinion models and their modifications ( @xcite , @xcite ) play a guiding role in analyzing the process of opinion spreading in communities .
these models are centered around the _ local majority rule _ ( l.m.r . ) , which is applied either in a deterministic or a probabilistic way . in the basic _ deterministic _ case , supporters of the ( two or three ) opinions present in a community are randomly distributed over all possible groups of a fixed size . within each group
the members adopt the opinion that has the majority in that group , after which all group members are recollected again . in case
there is no majority in a group , its members stick to their own opinion ( i.e. , _ neutral _ treatment ; the _ probabilistic _ treatment in case of a tie assigns opinions to the group members according to a certain probability distribution ) .
repeated application of this principle generates what is called _ randomly localized dynamics with a local majority rule _
( @xcite ) . in the basic _ probabilistic _ case , the community members are divided among groups of various sizes according to some probability distribution , and within each group all members adopt one of the possible opinions with a certain probability . in the basic deterministic two states opinion model , fast dynamics
occurs in which the opinion that originally has the majority eventually will obtain complete presence at the cost of the alternative opinion . in the probabilistic two states opinion model ,
the final outcome depends on the probability distributions for group sizes and local adaptation .
eventually the state of the community can be either a polarization on the initial majority or minority , or a perfect consensus on both opinions ( see @xcite , which unifies basic probabilistic two states opinion models ) . in
@xcite a three states opinion model is introduced in which the community members are randomly distributed over groups of size 3 . within each group
the l.m.r . is applied , with the additional rule that in case of a tie all members of the group adopt one of the three opinions according to some probability distribution .
it is shown that the dynamics leads to fast polarization on one of the three opinions , which may be an opinion that initially has a minor presence in the community . to add more realism to opinion models , in @xcite the basic deterministic two states galam opinion model
is extended by the introduction of so - called contrarians .
contrarian _ is a community member who , instead of the opinion it should adopt under the l.m.r .
, switches to the alternative opinion . depending on the density of contrarians as well as on group size , their presence either leads to a stabilization of the opinion dynamics in which one opinion ( the one with the lower density of contrarians ) dominates the other , to an equilibrium in which neither opinion dominates ( in case both opinions have equal densities of contrarians ) , or ( in case of relatively large densities of contrarians for both opinions ) to a dynamics in which the dominating opinion constantly alternates between the two opinions .
the incorporation of contrarians in opinion dynamics models is a step towards an explanation of the `` hung elections '' outcome in several recent voting events ( the u.s.a .
presidential elections in 2000 , the german elections in 2002 and 2005 , as well as the 2006 italian elections , see @xcite , @xcite ) , in which one strategy achieved only a very narrow gain at the cost of another strategy .
the origin of contrarian behaviour as well as its implications have been the focus of various other recent studies , see @xcite , @xcite , @xcite , @xcite . in a recent paper ( @xcite )
the basic deterministic two states galam model has been modified by introducing opinion supporters that express what in politics ( and other games ) is called _
inflexible behavior_. an inflexible community member is a supporter that under all conditions sticks to its opinion . under this terminology supporters that switch opinion when in the minority then classify as _
floaters _ , and we shall use this distinction in what follows . in @xcite the effect of inflexible behavior on opinion dynamics is studied for the case that opinion supporters repeatedly meet in groups of fixed size 3 .
it is shown that a small density of inflexibles for only one of the two opinions allows for the existence of two local attractors .
one of these local attractors is a mixed one , on which both opinions are present and on which the opinion that is supported by inflexibles is a minority .
the other attractor is a single state attractor , on which the opinion that is supported by inflexibles has complete majority , i.e. , its density equals 1 , the other opinion being absent . due to the presence of these two attractors , the outcome of the opinion dynamics thus depends on the initial condition , the basin of attraction for the mixed local attractor being relatively small compared to that for the single state attractor .
if the density of inflexibles is sufficiently large ( approximately 17@xmath0 ) , the mixed attractor disappears and the single state attractor becomes global . in case both opinions have small and equal densities of inflexibles there are two mixed local attractors .
these two attractors are symmetrically situated with regard to a separator on which both opinions are present with density @xmath1 .
a change in the density of inflexibles for one of the opinions breaks this symmetry , and a sufficiently large increase may lead to a global attractor on which the opinion with the larger density of inflexibles has the majority , see @xcite . recently , the effect of inflexibles and floaters on opinion dynamics has also been studied in @xcite and @xcite . in this paper
we combine the approaches presented in @xcite and @xcite , by allowing for groups composed of inflexibles as well as contrarian and non - contrarian opinion supporters . for clarity
we restrict ourselves to groups of fixed size at most 3 . for both opinions we assume fixed densities for the inflexibles . also , we consider the contrarians to be part of the floaters , i.e. , in a given group the contrarians first determine their opinion according to the l.m.r . , and subsequently change to become a floater ( not necessarily a contrarian ) for the alternative opinion ( which thus may be the opinion that the contrarian initially was supporting ) .
the presence of contrarians for each opinion is quantitatively expressed as a fixed fraction of the density of floaters of the respective opinion . in case of a tie in groups of size 2
we apply the neutral treatment . after an opinion update ,
all supporters for both opinions are recollected and then are redistributed again , either as an inflexible or as a non - contrarian or contrarian floater , according to the fixed densities for inflexibles and the fixed fractions of contrarians for the two opinions .
we study qualitative characteristics of the opinion dynamics generated by repeated updates .
in particular we study changes in the number of equilibria , and changes from monotone to alternating dynamics , due to changes in parameter combinations . a detailed mathematical treatment of the generated opinion dynamics , for groups of arbitrary size ,
shall be given in a forthcoming paper ( @xcite ) .
+ + _ notation _
we denote the two opinions by @xmath2 and @xmath3 .
the densities of inflexibles for the @xmath2 and @xmath3 opinion are denoted by @xmath4 and @xmath5 respectively , with @xmath6 as well as @xmath7 , and in addition @xmath8 .
since the roles of the @xmath2 and @xmath3 opinion are interchangeable in deriving the opinion dynamics , we may without loss of generality assume that @xmath9 , and we shall do so in what follows .
the fraction of contrarians among the @xmath2 floaters is denoted by @xmath10 , and @xmath11 denotes the fraction of contrarians among the @xmath3 floaters , with both @xmath12 and @xmath13 .
the size of the groups in which opinion supporters meet is denoted by @xmath14 .
the density of the @xmath2 opinion at time @xmath15 ( or after @xmath16 updates ) shall be denoted as @xmath17 .
note that for given @xmath4 and @xmath5 the density @xmath17 necessarily lies in the interval @xmath18 $ ] ( independent of @xmath14 , @xmath10 or @xmath11 ) .
with @xmath19 we denote the function that determines the density of the @xmath2 opinion after application of the l.m.r .
followed by the switch of the contrarians .
thus , @xmath20 . setting @xmath21
, @xmath22 then gives the density obtained from @xmath17 when the l.m.r .
is applied without being followed by the switch of the contrarians . in the appendix tables
are given , presenting all possible group compositions in terms of inflexibles and non - contrarian and contrarian floaters for group sizes @xmath23 to 3 , together with the effects of the l.m.r . and
the opinion changes of contrarians .
it is assumed that the community is sufficiently large to allow for the derivation of the density of each possible group composition in the ensemble of all groups of a fixed size from the densities in the community of the constituents of a group . from these tables
the expressions for @xmath24 are obtained .
+ with @xmath25 we denote the dynamics generated by repeated application of @xmath26 in subsequent timesteps .
furthermore , @xmath27 denotes an asymptotically stable equilibrium for @xmath28 , and @xmath29 refers to an asymptotically stable periodic point .
+ we now turn to the treatment of the opinion dynamics for group sizes @xmath30 and @xmath31 .
the case @xmath23 resembles a community in which each member is unaffected by other community members in determining its opinion , and the only changes in opinion come from the contrarians .
the contributions to the @xmath2 density after application of the local majority rule is obtained from the second column in table 1 in appendix [ subsection : t1 ] .
this column obviously is equal to the first one , since in groups of size @xmath32 local majority is automatically obtained , but is without effect on the opinion densities .
these contributions are : @xmath4 for the @xmath2 inflexibles , and @xmath33 for the ( non - contrarian and contrarian ) @xmath2 floaters .
their sum is @xmath34 , and we obtain that @xmath35 consequently , each @xmath36 $ ] is a neutrally stable equilibrium for the opinion dynamics . in case only (
non - contrarian and contrarian ) floaters are involved both @xmath4 and @xmath5 are equal to 0 , and we restrict ourselves to the contributions from the second , third , fifth and sixth line in the table . since the l.m.r . leaves each group of size 1 unaffected , a switch by a contrarian in this case necessarily implies a change to the opinion it initially does not support .
thus , here also a contribution to the @xmath2 density comes from the group that initially consists of only @xmath3 contrarians , as these will turn into @xmath2 floaters . in this case
we obtain for the contribution to the @xmath2 density : @xmath37 the effect of both inflexibles and non - contrarian as well as contrarian floaters is obtained by adding all the expressions in the last column : the contributions @xmath4 due to the invariant density of @xmath2 inflexibles , @xmath38 from the non - contrarian @xmath2 floaters , and @xmath39 from the contrarian @xmath3 floaters .
this yields : @xmath40 it follows that if @xmath41 , then @xmath42 is the unique equilibrium for the opinion dynamics @xmath43 .
due to its linearity as a function of @xmath44 , expression ( [ eq : l=1general ] ) implies that the dynamical characteristics of this equilibrium are governed solely by the frequencies of the contrarians .
the equilibrium is asymptotically stable if and only if @xmath45 .
for @xmath46 the equilibrium is approached monotonically , with an increase in the @xmath2 density if and only if its initial value is less then the equilibrium value . for @xmath47 ,
the function @xmath48 is constant and equals @xmath49 ; the opinion dynamics then reaches its equilibrium in one iteration . for @xmath50 ,
the equilibrium is approached alternately . for @xmath51 ,
i.e. , both @xmath52 and @xmath53 , the equilibrium equals @xmath54 and is neutrally stable ; each @xmath36 $ ] different from @xmath54 generates a neutrally stable cycle of length 2 .
+ on the equilibrium , the @xmath2 opinion has the majority if and only if the inequality @xmath55 holds .
thus , for an opinion to achieve the majority it is required that it is being supported by a sufficiently large density of inflexibles , and/or a sufficiently small frequency of contrarians among the floaters .
+ given densities @xmath4 and @xmath5 of inflexibles for the two opinions , a change in the frequencies of contrarians from 0 into small values @xmath10 and @xmath11 causes the bifurcation from a collection of neutrally stable equilibria for @xmath56 into a unique stable equilibrium for @xmath57 . the opinion which has the majority on this equilibrium is determined by inequality ( [ eq : l=1majority ] ) . in case @xmath58 , the opinion with the smaller frequency of contrarians obtains the majority .
conversely , given different frequencies @xmath10 and @xmath11 of contrarian floaters for the two opinions , in the absence of inflexibles the dynamics @xmath59 has @xmath60 as its unique stable equilibrium , on which the opinion with the smaller frequency of contrarians has the majority .
fixing sufficiently small densities @xmath4 and @xmath5 of both opinions as inflexibles , this equilibrium slightly shifts but leaves the majority unaltered . in case @xmath61 ,
in the absence of inflexibles the equilibrium @xmath62 equals @xmath1 , and the introduction of small densities of inflexibles for both opinions changes this equilibrium into one on which the opinion with the larger density of inflexibles takes the majority .
figure [ cap : cap1alt ] illustrates these conclusions .
figure [ cap : l=1overview ] gives a qualitative overview of the outcomes of the possible opinion dynamics @xmath63 .
+ figure [ cap : l=1overview ] : an overview of the opinion dynamics of @xmath64 , for values @xmath4 and @xmath5 as indicated , and with @xmath10 and @xmath11 in each pane on the horizontal and vertical axis , respectively , both ranging between 0 and 1 . in each pane the line with negative slope @xmath47 is drawn , and possibly an additional dashed line of positive or zero slope . on the line @xmath47 the function @xmath65 is constant , and the corresponding values of @xmath10 and @xmath11 separate between monotone and alternating dynamics , with the monotone dynamics occurring if @xmath46 , i.e. , below the line . the dashed line , if present , gives the values @xmath66 for which the equilibrium of the opinion dynamics equals @xmath1 , and is determined by the expression @xmath67 .
opinion @xmath2 obtains the majority if ( and only if ) @xmath68 holds , i.e. , if @xmath69 and @xmath70 lies above the dashed line .
the panes for values @xmath71 for which @xmath72 represent degenerate cases , in the sense that these values allow for only one density @xmath73 for the @xmath2 opinion to occur .
in groups of size 2 the number of members that support the @xmath2 or @xmath3 opinion may be equal , in which case a tie occurs . we shall deal with the neural treatment in case of a tie , in which each supporter keeps its own opinion .
+ appendix [ subsection : t2 ] shows the table for this case .
we obtain @xmath75 which is obvious , since in groups of size 2 no majorities can occur , and , in case of a tie , the neutral application of the local majority rule does not have any effect . incorporating the effect of non - contrarian as well as contrarina floaters , the table yields that @xmath76 thus , for groups of size 2 the effect of the neutral application of the local majority rule and the contrarians is the same as for groups of size 1 .
group size 3 is the smallest value of @xmath14 for which the local majority rule becomes effective due to possible group compositions in which a majority of one of the two opinions occurs . as a consequence ,
the generated dynamics allows for features different from those for group sizes 1 and 2 .
careful bookkeeping based on the table in appendix [ subsection : t3 ] yields that @xmath78 + for clarity we start the analysis of the generated opinion dynamics with the symmetric case of equal densities of inflexibles and equal fractions of contrarians for both opinions . taking @xmath58 and @xmath80 , we obtain that @xmath81 as an illustration to expression [ eq : l=3symm ] , figure [ cap : cap3alt2 ] shows a collection of graphs of @xmath82 as function of @xmath34 , for @xmath83 and several values of @xmath84 . from expression ( [ eq : l=3symm ] ) the analysis of the generated opinion dynamics
is straightforward .
we give an overview .
+ symmetry considerations imply that the dynamics @xmath88 has @xmath89 as an equilibrium , for any choice of @xmath90 $ ] and @xmath91 $ ] .
in addition to parameter combinations @xmath92 and @xmath84 for which this equilibrium is unique and stable , there are combinations which allow for an unstable repelling equilibrium @xmath93 in combination with two other , asymptotically stable , equilibria , or with two asymptotically stable periodic points of minimal period 2 .
details for these possibilities to appear are derived in appendix [ subsection : l=3symm ] , here we confine ourselves to the outcome .
+ let the _ critical curves _ @xmath94 and @xmath95 be defined as follows : @xmath96\times[0,1]:(3 - 4\alpha)(1 - 2\gamma)=2\},\ ] ] and @xmath97\times[0,1]:(3 - 4\alpha)(1 - 2\gamma)=-2\}.\ ] ] figure [ cap : cap4alt11 ] shows the curves @xmath94 and @xmath95 in the @xmath98-parameter space . on @xmath94
the derivative @xmath99 equals 1 , whereas on @xmath95 this derivative equals -1 .
the two corner areas in figure [ cap : cap4alt11 ] enclosed by either @xmath94 or @xmath95 are the regions of parameter combinations for which @xmath1 is unstable ; outside these regions ( including the curves ) @xmath1 is the unique asymptotically stable equilibrium for @xmath88 , independent of the initial condition .
the lower left corner region is the area for which the dynamics @xmath88 has two asymptotically stable equilibria @xmath100 .
given parameter combinations @xmath98 in this region , the opinion dynamics eventually will stabilize on an equilibrium on which the opinion with the initial majority will have maintained its majority . in case @xmath101 , this equilibrium is mixed ; if neither inflexibles nor contrarians are present for both opinions , i.e. @xmath102 , the equilibrium is a single state attractor with only one opinion present .
these results generalize those obtained in @xcite for the case of equal densities of inflexibles and no contrarians for both opinions . for parameter combinations in the upper left corner region in figure [ cap : cap4alt11 ] , the dynamics has two attracting periodic points of period 2 . here an initial majority does not guarantee the eventual majority , since the dynamics is such that both opinions alternately switch between minority and majority . thus ,
if both opinions are being supported by equal densities @xmath92 of inflexibles and equal fractions @xmath84 of contrarians among the floaters , for an opinion to obtain the majority it is necessary that @xmath92 as well as @xmath84 are sufficiently small , and that it has the initial majority .
also , with increasing @xmath92 ( @xmath84 ) , the maximum value of @xmath84 ( @xmath92 ) for which a majority is attainable decreases .
if no inflexibles are present , the fraction of contrarians among the floaters must be less than approximately @xmath103 ( @xmath104 ) for a majority to be realizable , and if the fraction of contrarians among the floaters equals 0 , the density of inflexibles must be less than @xmath105 .
+ if in the parameter space a combination @xmath98 approaches from within a corner towards one of the two critical curves , then the two additional equilibria or periodic points approach towards @xmath89 ; a withdrawal in the parameter space results in the opposite movement of the additional equilibria or periodic points .
it follows that when passing through @xmath94 , the dynamics @xmath88 undergoes a supercritical pitchfork bifurcation , and when passing through @xmath95 the dynamics undergoes a period doubling bifurcation ( flip bifurcation ) .
we now return to the general expression ( [ eq : l=3 ] ) and give an overview of the possible outcomes of the dynamics @xmath106 .
the analytical background is given in appendix [ subsection : l=3general ] .
we distinguish several cases . 1 .
@xmath47 . + for @xmath10 and @xmath11 such that @xmath47 , the function @xmath107 is quadratic in @xmath34 .
the corresponding opinion dynamics @xmath108 has a unique stable equilibrium in the interval @xmath109 $ ] . for @xmath110 ,
the function @xmath111 becomes constant and equals @xmath112 ; it allows for a unique stable equilibrium @xmath113 , on which opinion @xmath2 has the majority if and only if @xmath114 .
the following two figures distinguish between parameter combinations @xmath4 , @xmath5 and @xmath10 for which the @xmath2 opinion obtains either the majority or minority in equilibrium ( figure [ cap : equimajor ] ) , and for which the equilibrium is approached monotonically or alternately ( figure [ cap : switch ] ) .
it follows that with increasing value of @xmath10 , the region of parameter combinations @xmath115 for which opinion @xmath2 obtains the majority , decreases . in addition , if @xmath116 , the @xmath2 opinion can obtain the majority for any value of @xmath4 , provided that @xmath5 is sufficiently small ; if @xmath117 , @xmath4 must be sufficiently large and @xmath5 sufficiently small for an @xmath2 majority to occur .
+ + the different panes , distinguished by different values of @xmath10 , have @xmath4 on the horizontal axis and @xmath5 on the vertical one .
each pane shows the line @xmath72 and in addition the dashed line @xmath118 of parameter values for which the equilibrium value @xmath119 equals @xmath1 . below the dashed lines the equilibrium value lies above @xmath1 , i.e. , opinion @xmath2 then obtains the majority
the region @xmath120 is not involved in the analysis.,width=755,height=604 ] + the different panes show in white the region of parameters @xmath71 for which the equilibrium @xmath119 is approached monotonically ; the black regions indicate parameter combinations for which the equilibrium is approached alternately . for @xmath121 ,
the derivative of @xmath122 in the equilibrium equals 0 for all parameter values @xmath71 , and the equilibrium is reached in one iteration .
the line @xmath72 refers again to the degenerate cases for which the opinion dynamics of density @xmath34 is restricted to a single density @xmath123 .
parameter combinations @xmath71 for which @xmath120 are not involved in the analysis.,width=755,height=604 ] 2 .
+ the expression @xmath125 for determining the equilibria is @xmath126 the number of solutions is determined by its discriminant , which is denoted by + @xmath127 .
the expression for the discriminant is derived in appendix [ subsection : l=3general ] ; here we discuss its implications .
+ for parameter combinations @xmath128 such that @xmath129 , the equation @xmath130 has a unique real solution . if @xmath131 , there are three real solutions . however , these solutions do not necessarily have to belong to the interval @xmath109 $ ] ( but if a solution lies in this interval , it clearly is an equilibrium for the dynamics @xmath106 ) .
if @xmath132 there are three real solutions , of which at least two coincide ; if this happens in the interval @xmath133 $ ] , the parameter combination is at a bifurcation point , discriminating between dynamics with either a unique equilibrium or three equilibria .
if at the bifurcation point exactly two of the three solutions coincide , the coinciding solutions form a semistable equilibrium . + figure [ cap : p3selectionalt ] shows a collection of signplots for the discriminant , for values @xmath4 and @xmath5 as indicated , and with @xmath10 and @xmath11 for each signplot between 0 and 1 .
in addition the outcome of the analysis for parameter combinations @xmath134 is included , as well as the results of the analysis for combinations @xmath135 .
figure [ cap : p3part3 ] in appendix [ subsection : l=3general ] shows an extended version of figure [ cap : p3selectionalt ] that includes more values of @xmath4 and @xmath5 .
it shows that for both @xmath4 and @xmath5 larger than 0.3 , there is not much qualitative change in the signplots of the determinant @xmath136 .
+ , width=755,height=680 ] + figure [ cap : p3selectionalt ] : a collection of panes , for values @xmath4 and @xmath5 as indicated , and @xmath10 and @xmath11 for each pane ranging between 0 and 1 , with @xmath10 on the horizontal axis and @xmath11 on the vertical axis . in each pane
the signplot of the discriminant @xmath137 is shown for points @xmath138 for which @xmath139 .
white areas represent the parameter combinations with a positive discriminant ( i.e. , combinations for which the corresponding opinion dynamics has a unique equilibrium ) , and in dark regions the discriminant is negative ( the corresponding opinion dynamics then has 3 different equilibria , but not necessarily in the interval @xmath109 $ ] ) . in each pane
, on the line @xmath47 a density plot of the equilibrium density @xmath140 is given .
an increasing gray level indicates an increasing equilibrium value , with graylevel 0 ( black ) representing a density 0 , and graylevel 1 ( white ) representing density 1 .
furthermore , in panes for which @xmath79 holds , on the line @xmath61 in the black regions ( i.e. , a negative discriminant ) in white the points are indicated for which the equilibrium @xmath93 for @xmath106 is unstable ; other points on the lines @xmath61 ( for @xmath79 ) indicate parameter combinations for which @xmath93 is stable ( as follows from figure [ cap : cap4alt11 ] ) . + + the discriminant becomes singular for parameter combinations @xmath141 with @xmath47 . in approaching such parameter combinations for which @xmath142 , the value of @xmath137 goes to @xmath143 . for @xmath144 ,
the limit generically equals @xmath145 when this point is approached from the region @xmath146 ; the limit equals @xmath143 in case it is approached from the other side , i.e. , from the region @xmath147 .
( in case @xmath148 is approached along the zero set of @xmath136 , i.e. , in each pane in figures [ cap : p3selectionalt ] and [ cap : p3part3 ] along the boundary that distinguishes between the white and black regions and touches with the line @xmath47 , the limit clearly equals 0 . ) + our further discussion of the opinion dynamics @xmath106 is based on figures [ cap : p3selectionalt ] and [ cap : p3part3 ] . instead of a detailed analytical treatment , we continue with a number of characteristic outcomes of the opinion dynamics .
a first characteristic that draws attention in figure [ cap : p3selectionalt ] is the existence of a wedge - shaped region of parameter combinations @xmath149 with negative discriminant for sufficiently small values of all four parameters . for the cases with both @xmath79 and @xmath61 within this region
, we already found the existence of two attracting equilibria , symmetrically positioned with respect to a third , unstable equilibrium @xmath1 .
we therefore expect also to find a similar pattern of three equilibria in @xmath109 $ ] for deviations from such symmetric cases within the wedge - shaped region . in
@xcite it has been derived that this is indeed the case in the absence of contrarians , i.e. , for parameter combinations for which @xmath21 , and for @xmath4 and @xmath5 sufficiently small .
figure [ cap : qallsmall ] , which shows a number of graphs of functions @xmath150 for relatively small values @xmath4 ,
@xmath5 , @xmath10 and @xmath11 , implies the same pattern : in case the determinant @xmath136 is negative , the opinion dynamics has two attracting equilibria that are separated by an unstable one .
the two attracting equilibria differ with respect to the opinion by which they are dominated . by leaving the wedge - shaped area ,
a bifurcation in the opinion dynamics occurs on its boundary @xmath151 .
generically , when moving from inside the wedge - shaped area towards this boundary , the unstable equilibrium and one of the two stable equilibria move towards each other , and at the bifurcation point merge ( thus causing a supercritical saddle - node bifurcation ) .
once the boundary has been crossed , the region of parameters with a positive discriminant is entered , and the dynamics is left with one attracting equilibrium . on
this equilibrium opinion @xmath2 dominates if the upper part of the boundary is crossed , i.e. , when @xmath152 ; opinion @xmath3 has the majority when the right - hand side of the boundary is passed , on which @xmath153 holds .
this is also illustrated in figure [ cap : qallsmall ] .
the occurrence of such a bifurcation may lead to a drastic change in the outcome of the opinion dynamics : whereas inside the wedge - shaped region the outcome of the opinion dynamics depends on the initial condition , outside the wedge - shaped area the opinion dynamics will end on the unique equilibrium , independent of the initial condition . at the bifurcation point at the endpoint of the sharp region of the wedge - shaped area
a supercritical pitchfork bifurcation occurs , in which the three equilibria merge together into one attracting equilibrium .
figure [ cap : qallsmall ] : four panes of graphs of functions @xmath154 for relatively small values @xmath4 ,
@xmath5 , @xmath10 and @xmath11 , with @xmath71 as indicated below each pane . in each of the four panes ,
@xmath10 and @xmath11 satisfy @xmath155 .
the color coding of the graphs is as follows : red corresponds to @xmath156 , for green @xmath157 , for blue @xmath158 , for purple @xmath159 , and for magenta @xmath160 .
i.e. , in the corresponding panes in figure [ cap : p3selectionalt ] we traverse the line @xmath155 from its upper left point on the @xmath156 axis to its lower right point on the @xmath161 axis , thus passing through regions with positive , zero as well as negative discriminant . above each pane
the densities for opinion @xmath2 are presented , as obtained by the corresponding opinion dynamics @xmath162 , with initial density @xmath89 .
+ + the white regions in figure [ cap : p3selectionalt ] ( and also in figure [ cap : p3part3 ] ) are formed by the parameter combinations for which the discriminant @xmath137 is positive .
the corresponding opinion dynamics then have a unique equilibrium , which ( for the parameter combinations in figure [ cap : p3selectionalt ] ) is approached monotonically .
figure [ cap : psmallposdis ] shows a number of graphs @xmath163 for parameter combinations with a positive discriminant .
the figure indicates that for small values of @xmath10 and large values of @xmath11 opinion @xmath2 dominates in equilibrium , and that the dominion shift towards the alternative opinion if the fraction of contrarians among the @xmath2 floaters increases and that among the @xmath3 floaters decreases .
four panes of graphs of functions @xmath154 for relatively small values @xmath4 and @xmath5 , and with @xmath10 and @xmath11 satisfying @xmath164 .
the values of @xmath10 and @xmath11 are indicated below each of the four panes .
the color coding of the graphs is as follows : red corresponds to @xmath156 , for green @xmath160 , for blue @xmath165 , for purple @xmath121 , and for magenta @xmath166 .
i.e. , for given @xmath71 , we traverse the line @xmath164 from its upper left point on the @xmath156 line to its lower right point on the @xmath161 line .
the discriminant @xmath136 for the exposed parameter values is positive , indicating a unique equilibrium for the corresponding opinion dynamics.,title="fig:",width=642,height=642 ] + for given parameters @xmath4 and @xmath5 , crossing the boundary of the white area in any direction away from the lower left corner leads to the occurrence of a saddle - node bifurcation , now however outside the domain @xmath18 $ ] of the functions @xmath167 ( maintaining an attracting equilibrium in the domain ) .
therefore in the black region thus entered , the opinion dynamics also is characterized by a unique attracting equilibrium . proceeding towards the upper right corner , the line of parameter combinations @xmath138 satisfying @xmath47 is crossed . on this line
the discriminant @xmath137 is singular , and the corresponding opinion dynamics have been analyzed in [ sect : general].1 .
+ in the black area in the upper right corner , for equal and sufficiently small values @xmath58 and sufficiently large and equal values @xmath80 it has been derived earlier that @xmath168 has a unique unstable equilibrium @xmath93 , which causes the convergence of the dynamics towards an attracting periodic orbit of period 2 .
neither of the two opinions then achieves the definite majority .
the values @xmath92 and @xmath84 for which this occurs have been derived in [ subsection:4.1 ] , and are represented in figures [ cap : p3selectionalt ] and [ cap : p3part3 ] by white line segments in the upper right corners . for these parameter combinations
the discriminant of the equation @xmath169 is negative and thus has three different solutions , of which two are situated outside the domain @xmath109 $ ] .
continuity arguments imply that this behavior will be maintained for parameter combinations sufficiently close to these line segments .
figure [ cap : psmallupperright ] illustrates this . if the parameter combinations are sufficiently far removed from these line segments but @xmath10 and @xmath11 are still relatively large ( i.e. , for given @xmath4 and @xmath5 , in the upper right corner ) , the dynamics will converge alternately to a unique equilibrium .
i.e. , by moving away from the manifold determined by the constraints @xmath79 and @xmath61 with large values @xmath10 and @xmath11 , a flip bifurcation occurs in which the attracting periodic 2 orbit collapses to an attracting equilibrium point .
this is illustrated by figure [ cap : psmallupperrightconv ] .
on the attractor the dominion shifts towards opinion @xmath3 with increasing @xmath10 and decreasing @xmath11 .
figure [ cap : psmallupperright ] : four panes of graphs of functions @xmath154 for relatively small values @xmath4 and @xmath5 , and with @xmath10 and @xmath11 satisfying @xmath170 . the values of @xmath10 and @xmath11 are indicated below each of the four panes .
the color coding of the graphs is as follows : red corresponds to @xmath171 , for green @xmath172 , for blue @xmath173 , for purple @xmath174 , and for magenta @xmath52 .
i.e. , for given @xmath71 , we traverse the line @xmath170 from its upper left point on the @xmath53 line to its lower right point on the @xmath52 line . above each pane
the densities for opinion @xmath2 are presented , as obtained by the corresponding opinion dynamics @xmath162 , with initial density @xmath89 .
figure [ cap : psmallupperrightconv ] : four panes of graphs of functions @xmath154 for relatively small values @xmath4 and @xmath5 , and with @xmath10 and @xmath11 satisfying @xmath175 .
the values of @xmath10 and @xmath11 are indicated below each of the four panes .
the color coding of the graphs is as follows : red corresponds to @xmath166 , for green @xmath171 , for blue @xmath172 , for purple @xmath173 , and for magenta @xmath52 .
i.e. , for given @xmath71 , we traverse the line @xmath175 from its upper left point on the @xmath53 line to its lower right point on the @xmath52 line . above each pane
the densities for opinion @xmath2 are presented , as obtained by the corresponding opinion dynamics @xmath162 , with initial density @xmath89 .
+ + we end our discussion by presenting some additional opinion dynamics @xmath162 for parameter combinations from both the regions with positive and negative discriminant .
we consider cases with @xmath176 and with @xmath177 , i.e. , a case with small @xmath4 and intermediate @xmath5 , and one with both @xmath4 and @xmath5 intermediate .
the corresponding graphs of @xmath150 are represented in figures [ cap : pasmallbbig ] and [ cap : pabintmed ] , for several values of @xmath10 and @xmath11 .
all cases allow for a unique attracting equilibrium .
high values of both @xmath10 and @xmath11 lead to alternating convergence .
furthermore , a decrease in the fraction of contrarians among the floaters of an opinion increases the density of this opinion in equilibrium . figure [ cap : pasmallbbig ] : the left column shows four panes of graphs of functions @xmath178 , for different combinations of @xmath10 and @xmath11 .
the color coding for the different rows is given in the following table , with in the second to fifth column the values of @xmath10 : + + [ [ subsection : l=3symm ] ] @xmath77 : analysis for the fully symmetric case @xmath79 and @xmath61 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ the derivative ( with respect to @xmath34 ) @xmath179 of the function @xmath180 as given by expression ( [ eq : l=3symm ] ) in the equilibrium @xmath93 equals @xmath181 .
there are two additional equilibria @xmath182\ ] ] if and only if @xmath183 and @xmath184 .
if the two additional equilibria exist they are symmetrically positioned on opposite sides of @xmath1 , and asymptotically stable ; the equilibrium @xmath1 then is unstable , with @xmath185 . + for @xmath92 and @xmath84 such that @xmath186 and @xmath187 , the equilibrium @xmath1 also is unstable , with @xmath188 . in this case the dynamics @xmath189 has two asymptotically stable periodic points @xmath190 of minimal period 2 , symmetrically positioned with respect to @xmath1 : @xmath191.\ ] ] 1 .
@xmath110 : expression ( [ eq : l=3 ] ) equals @xmath112 , and allows for a unique stable equilibrium @xmath113 , on which opinion @xmath2 has the majority if and only if @xmath114 .
@xmath193 , @xmath47 : the function @xmath194 is quadratic in @xmath34 .
the discriminant + @xmath127 for the equation @xmath125 equals @xmath195 .
the opinion dynamics @xmath108 has a unique equilibrium + @xmath196 + + @xmath197 + + in the interval @xmath109 $ ] .
the derivative in the equilibrium equals + + @xmath198 .
3 . @xmath124 .
the expression @xmath125 for determining the equilibria now is @xmath199 its discriminant is @xmath200 with + @xmath201 , + @xmath202 , + @xmath203 , + @xmath204 , + and + @xmath205 , + + @xmath206 .
+ + figure [ cap : p3part3 ] shows , in three parts , a signplot of the discriminant for this case .
in addition the results of the analysis for combinations @xmath135 is included .
figures _ a , b , and c _ shows a collection of panes with signplots of the discriminant @xmath136 , for values @xmath4 and @xmath5 as indicated , and @xmath10 and @xmath11 for each pane ranging between 0 and 1 , with @xmath124 ; @xmath10 is on the horizontal axis and @xmath11 on the vertical axis .
the white areas represent the parameter combinations with a positive discriminant , and in dark regions the discriminant is negative . on the line @xmath47 in each pane
the discriminant becomes singular . in panes for which @xmath79 holds , on the line @xmath61 in the black regions ( i.e. , a negative discriminant ) in white the points
are indicated for which the equilibrium @xmath93 for @xmath106 is unstable ; other points on the lines @xmath61 ( for @xmath79 ) indicate parameter combinations for which @xmath93 is stable ( as follows from figure [ cap : cap4alt11]).,title="fig:",width=529,height=604 ] + 999 s. galam , y. geffen , y. shapiro , math . j.
sociology 9 , 1 , ( 1982 ) s. galam , physica a , 333 , 453 - 460 ( 2004 ) s. galam , europhys .
, 70 , 6 , 705 - 711 ( 2005 ) s. galam , f. jacobs , physica a , 381 , 366 - 376 ( 2007 ) s. galam , qual .
quant , 41 , 579 - 589 ( 2007 ) s. gekle , l. peliti , and s. galam , eur .
j. b 45 , 569 - 575 ( 2005 ) f. jacobs , s. galam , _ in preparation _
de la lama , j.m .
lpez , and h. s. wio , europhys .
72 , 5 , 851 - 857 ( 2005 ) m. mobilia , a. petersen , and s. redner , j. stat .
p08029 ( 2007 ) j.j .
schneider , international jpurnal of modern physics c , vol .
15 , 5 , 659 - 674 ( 2004 ) d. stauffer , j.s .
s martins , physica a , 334 , 558 - 565 ( 2004 ) h.s .
wio , m.s . de la lama , and j. m. lpez , physica a , 371 , 108 - 111 ( 2006 ) l. zhong , d .-
f . zheng , b. zheng , and p.m. hui , physical review , e 72 , 026134 ( 2005 ) | we assume a community whose members adopt one of two opinions .
each member appears as an inflexible , or as a non - contrarian or contrarian floater .
an inflexible adheres to its opinion , whereas a floater may change into a floater of the alternative opinion .
the occurrence of this change is governed by the local majority rule : members meet in groups of a fixed size , and a floater changes provided its opinion has the minority in its group .
a non - contrarian floater keeps the opinion it adopts under the local majority rule , whereas a contrarian floater adopts the alternative opinion .
we determine the dynamics generated by repeated opinion changes , and study its dependence on the densities of inflexibles and the frequencies of contrarians among the floaters .
we restrict ourselves to groups of size at most 3 .
20 pt * two opinions dynamics generated by inflexibles and non - contrarian and contrarian floaters * + + f. jacobs + + _ section theoretical biology + institute of biology + leiden university + kaiserstraat 63 + nl-2311 gp leiden + the netherlands + + e - mail address : [email protected]_ + + s. galam
+ + _ centre de recherche en epistmologie applique ( crea ) + ecole polytechnique + 1 rue descartes + 75005 paris + france + + e - mail address : [email protected]_ + + author for correspondence : f. jacobs + + * running title : * two opinions dynamics + + * acknowledgement * f. jacobs appreciates the support of the research underlying this paper by cost grant cost - stsm - p10 - 01215 .
* keywords : * sociomathematics , sociophysics , opinions dynamics , local majority rule , contrarian behaviour , floating behaviour + + * note concerning the current version of the text : due to the large sizes of some figures , the accompanying captions to these figures do not fit together with the figures on the same page . therefore , these captions are placed in the main text , instead of in the sequence of figures at the end of the text . * |
stochastic processes are used to model a great variety of systems in disciplines as disparate as physics @xcite , genomics @xcite , finance @xcite , climatology @xcite and social sciences @xcite . here
we will devote our attention to the special class of markovian stochastic processes that are also stationary and that can be described by a fokker - planck ( fp ) equation @xcite .
such processes are fully determined by the knowledge of their probability density function ( pdf ) @xmath6 and their conditional transition probability @xmath7 .
the simplest stationary markovian stochastic process is the ornstein - uhlenbeck ( ou ) one that is characterized by an exponential autocorrelation function @xmath8 where @xmath9 is referred to as the time - scale of the process . for a general stationary markovian additive process
the methodology of eigenfunction expansion @xcite allows us to write the autocovariance function as the infinite weighted sum of exponential functions @xmath10 each characterized by a time - scale @xmath11 , where the @xmath12 values are the eigenvalues of the fp operator . by analogy with respect to the ou process , for an additive stationary markovian process
the time - scales are thus defined as the inverse of the eigenvalues of the fp operator .
the methodology of eigenfunction expansion shows that the typology of time - scales included in a stochastic process is therefore strictly linked to the auto correlation properties of the process itself .
the generalization to a multiplicative stationary markovian process has been considered in ref .
@xcite , where we considered stationary markovian processes that include multiple time - scales @xcite .
in fact , we considered explicit examples of @xmath13 short - range correlated processes with an autocorrelation function that shows a decay modulated by some ( stretched- ) exponential function , @xmath14 short - range correlated processes with an autocorrelation function that shows a power - law decay and @xmath15 long - range correlated processes , thus showing anomalous diffusion @xcite . by using such explicit examples of stationary markovian stochastic processes , the investigations performed in ref .
@xcite mainly regarded the relationship between the stationary pdf or the conditional probability on one side and on the other side the typology of time - scales included in the process , i.e. its correlation properties .
the aim of this paper is to investigate how the inclusion of more and more time - scales into stationary markovian stochastic processes affects the first passage time distribution ( fptd ) .
first passage time issues are a classical topic in stochastic proceses research @xcite .
they also have relevant applications , for example , in many fields of finance .
just to mention , the assessment of the default risk for firms assets is essentially based on the evaluation of the fpt under the assumption that the asset value @xmath16 follows a geometric brownian motion @xcite , i.e. an ornstein - uhlembeck process for the variable @xmath17 .
there has therefore been much effort in obtaining the fptd of the ou process @xcite thus neglecting the possible implications of the correlation properties of the asset value at the level of the fptd and therefore at the level of the risk assessment . in this paper
we will start considering the effects of the autocorrelation properties of a stationary markovian stochastic process on the fptd .
deliberately , our investigations will be at a theoretical level , with the aim of understanding what are the relevant aspects of the problem .
to this end we will concentrate on the first passage time through two barriers located at @xmath18 and @xmath19 starting from position @xmath20 $ ] .
in fact , as explained below , such problem can be treated analytically .
the search for a more realistic model in the context of default risk or for other financial variables @xcite will be left for the future . in this paper
we will show that the large time dependance of the fptd crucially depends on the correlation properties of the process and it is independent from its stationary distribution .
moreover , in the case when the process is power - law correlated the fptd shows power - law tails @xmath0 . as a result ,
in the case when the process is power - law correlated the moments @xmath1 of the fptd are finite only when @xmath2 .
finally , we will give a general result connecting the fptd of an additive stochastic processes @xmath3 to the fptd of a generic process @xmath4 related by a coordinate transformation @xmath5 to the first one .
the paper is organized as follows . in section [ proc ]
we will give explicit examples of stationary markovian processes that include multiple time - scales .
we will consider short - range correlated processes that include an infinite set of time - scales bounded from above and processes that include an infinite set of time - scales that extends up to infinity .
for both cases we will consider the case when the stationary pdf is gaussian or with power - law tails . in section [ mfpt ] we will consider the mean fpt for the processes considered in section [ proc ] .
we will show how the simple knowledge of the mean fpt may be not enough to let us discern between processes with different correlation properties . in section [ fptd ]
we will derive a general expression for the fptd of stationary markovian stochastic processes . by using the explicit examples of section [ proc ] we
will then show how the correlation properties of the considered process affect the tails of the fptd .
finally , in section [ concl ] we will draw our conclusions .
in this section we will briefly illustrate the stochastic processes to be considered in this paper .
further details can be found in ref .
@xcite and ref .
@xcite .
let us consider stationary markovian processes that can be described by a fp equation @xcite : @xmath21 where @xmath22 and @xmath23 are the drift and diffusion coefficient appearing in the associated langevin equation : @xmath24 where @xmath25 is a @xmath26correlated gaussian noise term . for a stochastic process with a constant diffusion coefficient , the eigenvalue spectrum of the fp equation describing a stationary process consists of a discrete part @xmath27 and a continuous part @xmath28\lambda_c,+\infty[$ ] ( @xmath29 ) associated with eigenfunctions @xmath30 .
the fp equation with constant diffusion coefficient can be transformed into a schrdinger equation @xcite with a quantum potential @xmath31 .
the eigenvalue spectrum of the schrdinger equation is equal to the eigenvalue spectrum of the fp equation .
the relation between the eigenfunctions of the fp equation and the eigenfunctions @xmath32 of the schrdinger equationis @xmath33 . for a stationary process
such that @xmath34 only admits a null eigenvalue and a continuum part of the spectrum , one can show that : @xmath35 where @xmath36 .
this shows that the autocovariance function is the infinite weighted sum of exponential functions @xmath10 each characterized by a time - scale @xmath11 , where the @xmath12 values are the eigenvalues of the fp operator .
@xmath37 clearly shows that the typology of time - scales included in a stochastic process is strictly linked to the auto correlation properties of the process itself .
a generalization to a multiplicative stationary markovian process has been considered in ref .
@xcite .
let us consider the stochastic process described by the following langevin equation @xcite : @xmath38 where @xmath39 is a real constant and @xmath25 is a @xmath26correlated gaussian noise term .
the diffusion coefficient @xmath40 will be set to unity hereafter .
it is straightforward to show that the stationary distribution of the process is : @xmath41 the above process is a stochastic markov process described by a fp equation with constant diffusion coefficient . as such , the fp equation maps onto an equivalent schrdinger equation @xcite with an appropriate quantum potential @xmath34 . in this case
the potential is given by : @xmath34@xmath42 .
it admits a ground state with null eigenvalue and a continuum set of eigenvalues @xmath43 whose eigenfunctions are given in ref .
@xcite . by using the methodology of eigenfunction expansion
@xcite it is possible to prove that the autocovariance function @xmath44 of the above process is : @xmath45 which behaves like a power - law with an exponential truncation for large time lags : @xmath46 as @xmath47 . as a result
, this is a short - range correlated process characterized by an infinite set of time - scales bounded from above , i.e. @xmath48 .
let us consider the stochastic process described by the following nonlinear langevin equation @xcite : @xmath49 where @xmath50 and @xmath51 are real arbitrary constants and @xmath25 is a @xmath26correlated gaussian noise term .
it is straightforward to show that the stationary distribution of this process is : @xmath52 where @xmath53 and @xmath54 are real constants that can be obtained by imposing that @xmath55 is continuous and normalized to unity .
the quantum potential associated with this process is given by : @xmath56 such potential admits a ground state with null eigenvalue and a continuum set of eigenvalues @xmath57 attached to the null eigenvalue .
the eigenfunction of the ground state is @xmath58 whereas for @xmath59 it decays according to @xmath60 .
the constants @xmath61 and @xmath62 are set by imposing that @xmath63 is normalized and continuous in @xmath64 .
it is worth noting that for @xmath65 the stationary pdf @xmath66 of the stochastic process is a power - law function decaying as @xmath67 .
the normalizability of the eigenfunction of the ground state is ensured if @xmath68 . in the present study we consider stochastic processes with finite variance which implies @xmath69 . for @xmath70
the eigenfunction @xmath71 is a linear combination of bessel functions @xmath72 where @xmath73 . for @xmath74
we find @xmath75 and @xmath76 .
the coefficients @xmath77 , @xmath78 , @xmath79 and @xmath80 are fixed by imposing that @xmath71 and its first derivative are continuous in @xmath64 and that @xmath71 are orthonormalized with a @xmath26-function of the energy . by using the methodology of eigenfunction expansion @xcite it is possible to prove that the autocovariance function @xmath81 , where @xmath82 thus showing that we are dealing with a power - law correlated stochastic process . in the range @xmath83 the process is long - range correlated . as a result , this is a power - law correlated process admitting an infinite and unbounded set of time - scales .
let us now consider the process of section [ risk ] and the coordinate transformation @xcite : @xmath84 \label{risken_to_gauss}\end{aligned}\ ] ] by using the ito lemma , one can show that , starting from the process of eq .
@xmath85 , in the coordinate space @xmath86 one gets a multiplicative stochastic process whose stationary pdf is given by : @xmath87 with @xmath88 an additional arbitrary parameter . the corresponding drift and diffusion coefficients are given by : @xmath89 where @xmath90 when @xmath91 and @xmath92 when @xmath93 .
for large values of @xmath94 the drift and diffusion coefficient behave as : @xmath95 according to the eigenfunction expansion methodology @xcite , the autocorrelation function of the process defined by eq .
@xmath85 with the coordinate transformation of eq .
@xmath96 is given by @xmath97 where : @xmath98 where @xmath99 and @xmath100 are the eigenfunctions of the schrdinger equation with potential @xmath34@xmath42 associated to the stochastic process of eq .
@xmath85 . as explained in ref .
@xcite such process admits an infinite set of time - scales bounded from above and its autocorrelation function decays like @xmath101 for large time lags . as a result
, this is a gaussian distributed , short - range correlated process , characterized by an infinite set of time - scales bounded from above .
let us now consider again the process described by eq . @xmath85 . by performing the following coordinate transformation : @xmath102 and by using the ito lemma
, one can show that in the coordinate space @xmath103 one gets a multiplicative stochastic process whose stationary pdf is given by : @xmath104 with @xmath50 and @xmath105 positive real constants .
the requirement that the diffusion coefficient is continuous in @xmath106 gives the following constraint : @xmath107 correspondingly , the drift and diffusion coefficients are given by : @xmath108 according to the eigenfunction expansion methodology @xcite , the autocorrelation function of the process defined by eq .
@xmath85 with the coordinate transformation of eq .
@xmath109 is given by eq .
@xmath110 with @xmath111 replaced by @xmath112 . the relevant integrations can be performed analytically .
one can therefore show that such process admits an infinite set of time - scales bounded from above and its autocorrelation function decays like @xmath113 for large time lags . as a result , this is a short - range correlated process with a power - law decaying stationary pdf , characterized by an infinite set of time - scales bounded from above .
let us now consider the process of section [ chi ] and the coordinate transformation @xcite : @xmath114 , \hfill ~|x| > l , \\ \\
f_{ii}=\sqrt{2\,s}~{\rm{erf}}^{^{-1 } } \bigl[1-p(x ) \bigl ] , \hfill ~|x| \le l , \end{array } \right .
\nonumber \\ & & r(x)= { { 2\,n_i\,x^{1-\alpha } } \over { \alpha-1 } } , \label{chi_to_gauss } \\ & & p(x)={{2\,n_i\,l } \over { l^\alpha } } - { { n_i}\over{l^\alpha\,\cos(\sqrt{v_0 } l)^2}}\,\bigl(x - l\bigr)+ \nonumber \\ & & -~{{n_i}\over{2 \sqrt{v_0}\,l^\alpha\,\cos(\sqrt{v_0 } l)^2 } } \bigl(\sin(2 \sqrt{v_0 } x)- \sin(2 \sqrt{v_0 } l)\bigr ) .
\nonumber\end{aligned}\ ] ] by using the ito lemma , one can show that , starting from the process of eq .
@xmath115 , in the coordinate space @xmath116 one gets a multiplicative stochastic process whose stationary pdf is given by eq .
@xmath117 , with @xmath88 as an additional arbitrary parameter .
one can easily show that @xmath118 is continuos along the real axis , although its first derivative is discontinuos in @xmath119 .
the drift coefficient @xmath120 is discontinuos in @xmath119 . for large values of @xmath94
the drift and diffusion coefficient behave as : @xmath121 according to the eigenfunction expansion methodology @xcite , the autocorrelation function of the process defined by eq .
@xmath115 with the coordinate transformation of eq .
@xmath122 is given by eq .
@xmath110 with @xmath111 replaced by @xmath123 , @xmath124 and the eigenfunctions @xmath63 and @xmath71 are solutions of the schrdinger equation with the quantum potential of eq .
@xmath125 . as explained in ref .
@xcite and ref .
@xcite such process admits an infinite and unbounded set of time - scales and its autocorrelation function decays like @xmath126 with logarithmic corrections for large time lags . as a result , this is a gaussian distributed , power - law correlated process , characterized by an infinite and unbounded set of time - scales .
in order to understand how the correlation properties of a stochastic process affect its fptd we start investigating the mean fpt , which is the first moment of the fptd .
we will show in this section that processes with different correlation properties might show a similar behaviour in the mean fpt .
let us consider the mean time @xmath127 that is needed to reach for the first time position @xmath128 starting from position @xmath129 .
this is obtained by solving the equation @xcite : @xmath130 with boundary conditions @xmath131 when @xmath132 .
here @xmath22 and @xmath23 are the drift and diffusion coefficient appearing in the non - linear langevin equation associated to the considered stochastic process . in this section
we will compare the mean fpt of the two stochastic processes with gaussian pdf considered in the previous section .
indeed , the results shown here have already been considered in ref .
however , for the sake of clarity we present them here again in a shortened form . for the stochastic gaussian process of section [ riskgaussall ]
the mean fpt can be analytically computed by using the drift coefficient of eq . @xmath133 and the diffusion coefficient of eq . @xmath134 into eq .
@xmath135 . for the case when @xmath136 one
gets : @xmath137 an alternative way to obtain such result is to analytically compute the mean fpt for the stochastic process of eq .
@xmath85 and then performing the substitution : @xmath138 for large values of @xmath139 the mean fpt of eq .
@xmath140 gives : @xmath141 for the process of section [ chigaussall ] the mean fpt can not be computed by analytically solving eq .
however , one can follow the alternative way mentioned above .
specifically , we first computed analytically the mean first passage time of the process of eq .
@xmath115 by using eq .
@xmath135 . subsequently , according to eq .
@xmath122 , we performed the substitution : @xmath142 in the region @xmath143 it is possible to obtain an analytical expression showing that for large values of @xmath139 one gets : @xmath144 eq . @xmath145 and eq . @xmath146 clearly show that for large values of @xmath139 the process of section [ chigaussall ] has a mean fpt larger than the process of section [ riskgaussall ] .
this confirms that the process that incorporates more time - scales is the slowest . in the two examples considered here , the way the correlation properties of the process affect
the mean fpt is clear . in the case of eq .
@xmath146 we have a linear growth in terms of the @xmath147 function , while in the case of eq .
@xmath145 we have a power - law growth with exponent larger than unity .
let us now consider the mean fpt of the two stochastic processes with power - law decaying pdf considered in section [ proc ] . for the stochastic process of setion [ rpa ] the mean first passage time
can be analytically computed by using the drift coefficient and the diffusion coefficient of eq . @xmath148 and eq . @xmath149 into eq .
it is possible to obtain an analytical expression showing that for large values of @xmath139 one gets : @xmath150 where @xmath105 is defined in eq .
@xmath151 . for small values of @xmath139 one
gets @xmath152 .
as mentioned in the previous section , for the stochastic process of section [ chi ] the mean fpt can be analytically computed by using the drift coefficient and the diffusion coefficient of eq . @xmath115 into eq .
it is possible to obtain an analytical expression showing that for large values of @xmath139 one gets : @xmath153 for small values of @xmath139 one gets @xmath152 .
both in eq . @xmath154 and @xmath155 the parameter @xmath105 is the exponent of the pdf tail . therefore , as for the processes with gaussian pdf , the mean fpt of the process that includes more time - scales is larger than the other
. however , in the two cases considered here a power - law dependance from @xmath139 is found .
the qualitative behaviour of the mean fpt in both processes is therefore very similar , although the correlation properties of the two processes are very different .
in fact , we will show that the different correlation properties of the stochastic processes display their effect at the level of the higher order fptd moments .
let us suppose to be interested in the distribution @xmath156 of the first times at which a process reaches the absorbing barriers @xmath18 or @xmath19 starting from a generic position @xmath157 with @xmath158 . according to ref .
@xcite one gets : @xmath159 where @xmath160 is the conditional probability of finding the process in position @xmath129 at time @xmath161 , knowing that it was in position @xmath157 at time @xmath162 . according to the methodology of eigenfunction expansion @xcite , for an additive stationary markovian process described by a fpe admitting only a continuum set of eigenvalues ,
the conditional probability is given by : @xmath163 where the @xmath12-integration is extended over the continuum part of the spectrum and @xmath99 , @xmath100 are the eigenfunctions of the quantum potential @xmath164 associated to the stochastic process considered .
if the potential @xmath34 admits a discrete part of the spectrum then an additional term involving a sum over all the discrete eigenvalues should be considered . by using the above expression for the conditional probability one
finally gets : @xmath165 the role played by each time - scale in the fptd can be emphasized by rewriting eq .
@xmath166 in the form @xmath167 where @xmath168 can be considered as the weight with which each time - scale enters the fptd .
the above formulae can be used to obtain the fptd in the case of the additive stochastic processes of section [ risk ] and section [ chi ] .
let us now consider a stochastic process @xmath4 obtained by performing a coordinate transformation @xmath5 starting from a stochastic additive process @xmath3 , as for the processes of section [ riskgaussall ] , section [ rpa ] and section [ chigaussall ] .
the conditional probability is given by @xcite : @xmath169 where @xmath170 is the stationary distribution of the process @xmath4 and @xmath99 , @xmath100 are the eigenfunctions of the schrdinger equation with the potential @xmath34 associated to the additive process @xmath3 . as a result , the fptd of the transformed process @xmath4 is given by : @xmath171 a direct comparison of eq . @xmath172 and eq .
@xmath166 shows that : @xmath173 therefore , for a general stochastic process @xmath4 , the problem of finding the fptd can always be re - conducted to the problem of finding the fptd of an appropriate additive process . when the additive process admits a quantum potential @xmath34 that is exactly solvable then eq .
@xmath174 can be used .
if @xmath34 is not exactly solvable , one might either take advantage of the approximations schemes widely used in quantum mechanics to estimate the eigenfunctions or revert to numerical simulations of the process to estimate the fptd .
let us now consider the stochastic process of section [ risk ] .
moreover , let us fix @xmath175 and @xmath176 , @xmath177 as in section [ mfpt ] .
the integration over @xmath129 required in eq .
@xmath166 can be performed analytically .
it gives that @xmath178 when @xmath179 .
therefore the large time behaviour of the fptd is given by @xcite : @xmath180 according to the result of eq .
@xmath181 , the same asymptotic behaviour is shared by the fptd of the process of section [ riskgaussall ] and section [ rpa ] , although they have different pdfs .
let us now consider the stochastic process of section [ chi ] for the case when @xmath175 and @xmath176 , @xmath177 .
the integration over @xmath129 required in eq .
@xmath166 can be performed analytically .
one therefore obtains that @xmath182 .
therefore , the asymptotic behaviour of the fptd is given by : @xmath183 according to the result of eq .
@xmath181 , the same asymptotic behaviour is shared by the fptd of the process of section [ chigaussall ] , although they have different pdfs . in fig .
[ fptdintchi ] we show the function @xmath184 for the process of section [ chi ] .
the curves are obtained by using eq .
we have considered the following three set of parameters : ( i ) @xmath185 and @xmath186 ( solid line ) corresponding to @xmath187 ; ( ii ) @xmath185 and @xmath188 ( dashed line ) corresponding to @xmath189 ; ( iii ) @xmath185 and @xmath190 ( dashed dotted line ) corresponding to @xmath191 .
for all the three sets we considered @xmath192 . in the bottom part of the legend we report the result of a nonlinear fit performed in the range @xmath193 $ ] .
these results are clearly in agreement with those of eq .
@xmath194 .
the above results explicitly show that the tail of the fptd is independent from the specific stationary distribution of the stochastic process considered .
rather , it is strictly related to the correlation properties of the process .
in fact the processes of section [ chi ] and section [ chigaussall ] have the same fptd tail although their pdf is different .
similarly for the three processes of section [ risk ] , section [ riskgaussall ] and section [ rpa ] .
another relevant comment regards the moments of the fptd distribution .
the existence of power - law tails in the fptd of power - law correlated stationary markovian processes implies that the moments of the distribution @xmath1 are defined only for @xmath195 . in the case of the processes of section [ risk ] , section [ riskgaussall ] and section [ rpa ] , the exponential cut - off of eq .
@xmath196 ensures that the moments of the fptd are always finite .
that marks a relevant difference between a process admitting a bounded set of timescales and a process in which all timescales are included .
as shown in section [ powerpropagator ] , such difference is not caught by looking at the mean fpt only .
in fact both the processes of section [ rpa ] and section [ chi ] , both characterized by a power - law decaying pdf , have a power - law growing mean fpt . however , in the case of section [ rpa ] all the higher moments of the fptd exist , while in the case of section [ chi ] only the moments @xmath1 such that @xmath195 are defined . finally , the result of eq .
@xmath194 shows that the tail of the fptd for power - law correlated stationary markovian processes is different from what is expected for other non - markovian processes with anomalous diffusion .
for instance , in ref .
@xcite it is shown that for a fractional brownian motion the tails of the fptd decay like @xmath197 , where @xmath198 $ ] is defined as @xmath199 and @xmath200 is the mean square displacement of the process .
the fact that the process of section [ chi ] includes an infinite and unbounded set of time - scales is a necessary condition for the existence of power - law tails in its fptd .
this is ultimately due to the fact that the eigenvalues spectrum of the schrdinger potential of eq .
@xmath125 is given by a continuum set of eigenvalues @xmath201 attached to the null eigenvalue associated to the ground state and therefore it includes an infinite and unbounded set of time - scales .. one might therefore wander whether or not such power - law decay is merely due to the fact that the process incorporates an unbounded set of timescales or something else must be considered . in general , any schrdinger potential that asymptotically decays like @xmath202 would give a stochastic process with an infinite and unbounded set of time - scales .
in the case @xmath203 , we will show here that the corresponding fptd tail does not necessarily display a power - law decay .
let us now consider the additive stochastic process associated to the schrdinger potential @xcite : @xmath204 where @xmath50 , @xmath51 and @xmath205 are real positive constants .
this quantum potential is exactly solvable . in the region @xmath206 the eigenfunctions are given by : @xmath207 where @xmath208 is the k - bessel function of order @xmath209 , @xmath210 is the confluent hypergeometric function and @xmath211 is the kummer hypergeometric function @xcite .
the normalization constant @xmath61 is fixed by imposing that the ground state is normalized to unity .
the other normalization constants @xmath77 and @xmath212 are fixed by imposing that the odd and even eigenfunctions fulfill the normalization condition @xmath213 .
the drift coefficient @xmath22 aymptotically decays like @xmath214 .
as mentioned above , the large time behaviour of @xmath215 is determined by the small energy behaviour of @xmath216 @xcite .
one can explitely show that in the limit when @xmath12 is small then : @xmath217 where @xmath218and @xmath208 are bessel functions of order @xmath209 . as a result ,
the small energy limit of @xmath71 is essentially determined by the small energy limit of @xmath219 . by making use of eq .
@xmath220 and eq .
@xmath221 of ref .
@xcite one gets : @xmath222 where @xmath223 is the regular coulomb function with null index . by making use of eq .
@xmath224 of ref .
@xcite one gets : @xmath225 where @xmath226 is the bessel functions of order @xmath39 and the coefficients @xmath227 are defined in eq .
@xmath228 of ref .
we have thus factorized the @xmath12 dependencies from the @xmath94 dependencies . as a result
, the large time behaviour of the fptd is given by the small energy behaviour of @xmath229 . by using the results of ref .
@xcite one finally gets : @xmath230 the above result shows that the mere existence of an infinite and unbounded set of timescales is not enough to ensure an asymptotic power - law behaviour in @xmath231 . in passing , by using the same arguments it is also possible to prove that the stochastic process associated to the quantum potential of eq .
@xmath232 is not power - law correlated , even though it incorporates an infinite and unbounded set of timescales @xcite .
the aim of this paper was to investigate how the fptd is affected by the typology of time - scales included in a stochastic process , i.e. by the correlation properties of the process itself .
the motivations are both theoretical and in view of possible applications . in fact , first passage time issues have relevant applications in many fields of finance such as the assessment of the default risk for firms assets .
in this paper we have shown that the tail of the fptd crucially depends on the correlation properties of the process and it is independent from its stationary distribution .
when the process includes an infinite set of time - scales bounded from above the fptd shows tails modulated by some exponential decay . in the case when the process is power - law correlated @xmath233 the fptd shows power - law tails @xmath0 . as a result , in the case when the process is power - law correlated the moments @xmath1 of the fptd are finite only when @xmath2 .
we have also shown that such power - law behaviour is not merely due to the fact that the process includes an infinite and unbounded set of time - scales .
rather , it is also required that each time - scale enters the fptd in a peculiar way .
in fact , in order to observe a power - law tail in the fptd the weights @xmath168 must be distributed according to a power - law for small values of @xmath12 , i.e. for large time - scales .
moreover , we have given a general result showing that given ( i ) and additive stochastic processes @xmath3 and ( ii ) another process @xmath4 related by a coordinate transformation @xmath5 to the first one , the fptd of @xmath4 can be obtained from the one of the @xmath3 process by using eq .
therefore , for a general stochastic process @xmath4 , the problem of finding the fptd can always be re - conducted to the problem of finding the fptd of an appropriate additive process . when the additive process admits a quantum potential @xmath34 that is exactly solvable then eq .
@xmath174 can be used .
if @xmath34 is not exactly solvable , one might either take advantage of the approximations schemes widely used in quantum mechanics to estimate the eigenfunctions or revert to numerical simulations of the process to estimate the fptd . finally , the result of eq .
@xmath194 shows that the tail of the fptd for power - law correlated stationary markovian processes is different from what is expected for other non - markovian processes with anomalous diffusion , such as the fractional brownian motion .
this result might be used to test the markovian property of a generic stochastic process . | the aim of this paper is to investigate how the correlation properties of a stationary markovian stochastic processes affect the first passage time distribution .
first passage time issues are a classical topic in stochastic processes research .
they also have relevant applications , for example , in many fields of finance such as the assessment of the default risk for firms assets .
by using some explicit examples , in this paper we will show that the tail of the first passage time distribution crucially depends on the correlation properties of the process and it is independent from its stationary distribution .
when the process includes an infinite set of time - scales bounded from above , the fptd shows tails modulated by some exponential decay . in the case when the process is power - law correlated the fptd shows power - law tails @xmath0 and therefore the moments @xmath1 of the fptd are finite only when @xmath2 .
we will also show that such power - law behaviour is not merely due to the fact that the process includes an infinite and unbounded set of time - scales .
rather , the time - scale must enter the fptd with weights that must be distributed according to a power - law for large time - scales values .
finally , we will give a general result connecting the fptd of an additive stochastic processes @xmath3 to the fptd of a generic process @xmath4 related by a coordinate transformation @xmath5 to the first one . |
detection plays a fundamental role in the field of image processing .
an _ edge _ is a collection of connected pixels where the intensity level changes abruptly.@xcite edges are especially important for image recognition and analysis since they can capture local features and provide useful information .
so far , techniques for edge detection have been studied extensively .
edges can usually be found in parts of an image where transition occurs , either between different objects , different regions , or between objects and the background . in this view , gradients are effective descriptors of edges .
consequently , various gradient - based edge detectors were proposed during the 1960s , including the robert operators , robinson operators , prewitt operators , and sobel operators , first suggested by irwin sobel and gary feldman@xcite . a more accurate detector known as the _ canny detector _ was later proposed by j. canny , which smooths the input image with a gaussian filter and applies non - maxima suppression for optimal results.@xcite for digital images
, derivatives can be approximated with discrete differentiation .
therefore , first - order edge detectors are easy to implement and widely used . among them , the sobel operators are especially preferred because they are non - linear filters with image smoothing , and thus can produce less fragmentary edge images . nevertheless , early implementations of sobel operators , mainly software - based and dsp - based were slow and unable to support real - time operations .
the recent introduction of field programmable gate arrays ( fpgas ) have succesfully resolved the problem .
an fpga is a device that enables real - time parallel processing with pipeline technology , which in previous studies were used to implement sobel detectors@xcite .
experimental results in @xcite proved the design an effective approach to real - time edge detection .
a novel pipelined structure was also proposed to speed up the operations@xcite . however , when sobel operators are applied to low - contrast images , the results are often less satisfactory , with more isolated points and significant distortions . to this end
, we introduce _ fuzzy theory _ to enhance the edges before applying the sobel operators .
this method was studied in @xcite and @xcite . in this paper , we combine the fuzzy method and the sobel edge detector to implement a fpga - based color image detector with improved accuracy for low - contrast images . the rest of the paper is arranged as follows . the sobel edge detection filter is introduced in section [ sec2 ] .
theoretical foundations for the preprocessing procedures are elaborated in section [ sec3 ] .
section [ sec4 ] gives the fpga - based hardware implementation for this paper . in section [ sec5 ] , we present the experimental results .
future prospects are briefly discussed in section [ sec6 ] .
the sobel edge detection filter is a commonly used edge detector that computes an approximate gradient of the image intensity function . for each pixel in the image
, it obtains the vertical and horizontal components of the gradient by applying convolution with two 3-by-3 masks defined as : @xmath0 @xmath1 specifically , the horizontal and vertical gradients of the pixel @xmath2 can be written as @xmath3 it can be shown that the coefficient @xmath4 endows the operator with noise suppression capabilities.@xcite the @xmath5 norm of the corresponding gradient vector can then be computed by @xmath6 however , eq . is of high computational costs , and the evaluation of square roots will require extra modules such as a cordic unit@xcite , rendering the circuit complex . thus the @xmath7 norm is frequently used to approximate this magnitude : @xmath8 an improved version of the original sobel operator uses two extra convolution kernels that detect gradients along the diagonals : @xmath9 @xmath10 for the sake of simplicity , we shall confine ourselves to the original sobel operator with only two masks .
for this study , threshold selection is realized using _
the otsu method_. this adaptive method is considered optimal in the automatic evaluation of a suitable threshold .
the algorithm assumes two categories of pixels , object pixels and background pixels .
the algorithm finds a threshold @xmath11 such that the mean grayscale of object , @xmath12 , and the mean grayscale of the background , @xmath13 are furthest from the mean grayscale of the entire image , @xmath14 .
this threshold @xmath11 can be found by maximizing between - class variance .
its procedure can be described as follows .
@xmath15 where @xmath16 is the frequency of the pixels in the picture and @xmath17 is the gray - level of point @xmath18 in the original image .
the percentage of the object is @xmath19 and the percentage of the background is @xmath20 the means of object @xmath21 is @xmath22 and the mean of background @xmath23 is @xmath24 therefore , the overall mean is @xmath25 the final threshold value @xmath11 is @xmath26 intuitively , this formula gives the maximum between - class variance , which can roughly be viewed as the contrast between two classes of pixels , black and white .
the larger the variance is , the better the contrast will be .
therefore , the optimal segmentation threshold @xmath11 for the original image found by eq .
splits the image into object and background effectively .
this method can be realized in an fpga fabric using the fixed - point number system and supplementary modules.@xcite we introduce fuzzy theory to enhance the edge detection capabilities of the sobel detector .
the theory of fuzzy sets was proposed by zadeh in 1965 and has gained much weight in image processing .
it uses the degree of membership function to determine whether a given point is black " or white " , a concept that is usually subjective and hard to accurately describe .
formally , a _
fuzzy set _
@xmath27 in a finite set @xmath28 is represented as @xmath29 , where @xmath30 $ ] is the function that represents the belongingness or degree of membership in of an element in @xmath31 in the finite set . in this paper
, we only have two fuzzy sets @xmath27 and @xmath32 ( @xmath27 for black , @xmath32 for white ) , thus the non - belongingness of an element of set @xmath27 is @xmath33 , which is also the belongingness of fuzzy set @xmath32 .
note that the degree of memebership function is usually a family of functions determined by the threshold parameter , which in this case is obtained via the otsu method
. some commonly used common degree of membership include @xmath34 distribution , cauchy distribution , gauss distribution , and the s - shaped distribution .
the cauchy distribution is defined as @xmath35 where @xmath36 , @xmath37 is a positive even integer , and @xmath38 is the center of distribution .
another version of this distribution is @xmath39 where @xmath40 .
the s - shaped distribution is @xmath41 to employ fuzzy theory in image segmentation and edge detection , one approach is to divide an image into multiple layers by different thresholds , which can can be obtained by the maximum fuzzy entropy@xcite .
another way is to directly transform the original image into an enhanced version using the fuzzy function . in this study
, we select a universal polynomial function @xmath42 where @xmath31 is the original gray value .
this function is independent to the threshold and similar to the increase contrast " function used in adobe photoshop .
the preprocessing procedure carried out in the experiment can be summarized as follows : 1 .
calculate the threshold @xmath43 using the otsu method .
2 . apply threshold @xmath43 to get the degree of membership @xmath44 .
3 . obtain the fuzzy image from the function @xmath45&\rightarrow [ 0,255 ] \\x & \mapsto 255\cdot \mu_w(x).\end{aligned}\ ] ]
the proposed architecture is illustrated in fig .
this system is composed of three modules , the fuzzy preprocessor , the window generator and the sobel operator .
fuzzy preprocessor is the module that enhances the original image to increase contrast between dark and bright regions .
the data input of this module are the individual pixels from the original image . for this experiment ,
the polynomial degree of membership function is hard - coded into this module . the resulting image
is then stored and accessed by the window generator .
the final threshold is determined by the otsu method using a separate program , although a complete fpga solution is also possible@xcite .
since the image input is in the form of a data stream , only one pixel can be transmitted during each clock cycle . to obtain the nine necessary pixels simultaneously ,
we deploy three rows of shift registers and two fifos to buffer the data .
the clock signal ` clk ` controls the input so when the nine 8-bit arrays @xmath46 are full , a 3-by-3 window is prepared for the sobel algorithm .
the two @xmath47 fifos are generated by rams .
the structure of the window generator is shown in fig .
it consists of two fifos with length of 253 and width of 8 and three 3-bit shift register .
the three shift registers crops " a 3-by-3 window , while the two fifos are used to buffer the data between while traversing different rows of the image matrix , which has length @xmath48 .
when we shift the data line , the position of the window is also shifting . the function of the sobel operator has been explained in sec .
the inputs of this module are the nine entries of the window matrix marked @xmath49 to @xmath50 , and the output is either 1 or 0 .
1 indicates that the pixel we are calculating is an edge pixel and 0 otherwise .
the result is obtained by thresholding the gradient , given by the sum of absolute magnitudes of the two components .
it is worth noticing that the operations can be effectively reduced by half if the additions are grouped properly .
we define the following quantities , first proposed in @xcite : _ neighbouring sums _ : @xmath51 @xmath52 _ partial sums _ : @xmath53 @xmath54 _ interlaced differences _ : @xmath55 @xmath56 the original procedure described in section [ sec2 ] requires 15 summations to compute the gradient for a single pixel . as the 3-by-3 window shifts with the data stream ,
it is clear that some pixels must be accessed more than once .
now , if we save the above intermediate results , the process can be accelerated .
first , add the gray value of each pixel with its right neighbour and obtain a matrix for right - neighbour sum ( the same procedure is repeated on the columns ) . then , add each entry of the obtained matrix with its right entry as same as the first step , matrix for partial sum is obtained .
last , compute the interlaced difference for every pixel , the absolute value of interlaced difference is exactly the horizontal gradient . in this manner
only seven sums are needed to compute the gradient for each pixel , accelerating the operations significantly .
the circuit design is completed for @xmath57 8-bit rgb images in verilog hdl and simulated in modelsim 10.1c . since cmos sensors such as ov9650
can be configured to output in rgb mode , for test purposes we store the three channels of an image in three separate text files to be read by the testbench program .
the eight images in fig .
[ fig4 ] are the corresponding edge maps of fig .
[ fig_sim ] , obtained using our edge detector .
we compare the results of standalone sobel operators to that obtained with fuzzy preprocessing enabled . under the same threshold of the sobel operator ,
it is obvious that our method is able to extract more details from the original picture . in fig .
[ fig4 ] , the edge detector using non - fuzzy method produces incomplete edges of the pens . on the other hand , with fuzzy preprocessing
the different regions become more distinguished , so more edges are detected , as desired .
the reason fuzzy method produces this more ideal outcome is that fuzzy method increases the contrast of different portions of the image and thus increases the difference of the pixels on either side of the edge .
however the noise level is not significantly reduced as expected .
for example , at the upper - left corner of fig .
[ fig4](a ) , while brightness level fluctuations in this area are clearly visible , edges are barely detected .
several solutions to resolve the noise problem mentioned in section [ sec5 ] are possible .
for instance , as is shown in @xcite , the canny operator can produce an ideal outcome .
second , smoothing operation plays an important role in de - noising .
it is feasible to add a smoothing convolution module before the sobel module to reduce the noise level .
third , in this experiment we used a simple polynomial function for fuzzy preprocessing for simplicity .
a more appropriate fuzzy transformation function has yet to be found .
although the edge detection based on sobel operator and fuzzy set has been thoroughly studied , we wish to extend and optimize our work using memristors . in 1971 , chua proposed the possibility of a new electronic component called a memristor . in 2008 , hewlett - packard implemented the first instance of memristor which could be used as a basic component of a digital circuit .
memristors are capable of performing logic operations with its unique memristive features .
moreover , the memory of memristor is non - volatile to the loss of voltage .
furthermore , the voltage and resistance features can be used to compute complex logic functions@xcite .
the idea of conventional edge detection based on memristor is building xor function to test the edge in the image using memristor @xcite . moreover , a memristor - based nonvolatile latch design was proposed in@xcite , which greatly reduces the size of latch circuit .
one future prospect for edge detection is to implement a sobel edge detector using memristors .
we believe that the size of the circuit will be reduced significantly and the whole circuit will be more non - volatile to voltage fluctuations . [ cols="^,^,^,^,^,^,^",options="header " , ] table 1 : stimulation sequence for the xor function . in @xcite , an edge detector based on memristor was proposed .
a memristor has two states , high impedance and low impedeace which are represented by @xmath58 and @xmath59 .
when the forward voltage is larger than the threshold voltage , the impedance will be lowered , and when reverse voltage is larger than the threshold voltage , the impedance will change to high impedance .
[ fig5 ] is the implementation of xor operator using memristor , table 1 shows the stimulating voltage sequence of the xor function , where the final output is stored in the memristor @xmath43 .
[ fig6 ] gives the general architecture of the memristive edge detector .
@xcite has proposed a feasible scheme to produce the edge map by scanning the image horizontally and vertically , and applying the above simulating voltages .
the advantage of embracing this new architecture is that one can build a much more compact circuit with virtually no loss in functionality .
however , we believe that an xor function is probably not the most ideal method for edge detection . we intend to reorganize the circuit structure using memristive latches proposed by @xcite .
we wish to thank mr .
yunji chen for his encouragement and guidance that inspired this project .
we would also like to thank ms .
shengyuan zhou and ms .
tian zhi for her help .
none , however , is responsible for any remaining errors .
z. guo , w. xu , and z. chai , image edge detection based on fpga , " in _ distributed computing and applications to business engineering and science ( dcabes ) , 2010 ninth international symposium on distributed computing and applications to business , engineering and science _ ,
169 - 171 .
ieee , 2010 .
e. lehtonen and m. laiho , stateful implication logic with memristors , " in _ proceedings of the 2009 ieee / acm international symposium on nanoscale architectures _ , pp .
ieee computer society , 2009 .
j. pandey _ et al .
_ , a novel architecture for fpga implementation of otsu s global automatic image thresholding algorithm , " _ 2014 27th international conference on vlsi design and 2014 13th international conference on embedded systems _ , 2014 . | edge detection is a classic problem in the field of image processing , which lays foundations for other tasks such as image segmentation .
conventionally , this operation is performed using gradient operators such as the roberts or sobel operator , which can discover local changes in intensity levels .
these operators , however , perform poorly on low contrast images . in this paper
, we propose an edge detector architecture for color images based on fuzzy theory and the sobel operator .
first , the r , g and b channels are extracted from an image and enhanced using fuzzy methods , in order to suppress noise and improve the contrast between the background and the objects .
the sobel operator is then applied to each of the channels , which are finally combined into an edge map of the origin image .
experimental results obtained through an fpga - based implementation have proved the proposed method effective .
image edge detection , fuzzy systems , sobel operator , hardware implementation , memristors . |
quark gluon plasma ( qgp ) is a non - ideal system and hence the formulation of statistical mechanics ( sm)and thermodynamics ( td ) is a challenging one .
sm and td of ideal system is straight forward , but the equilibrium studies of non - ideal system is still underdeveloped because of the difficulties due to mutual interaction among the constituents . of course , there are theories like classical and quantum mayer cluster expansion @xcite , but they are not fully successful for qgp because of the strong interaction @xcite .
mayer s cluster expansion is valid for weakly coupled system , whereas qgp seems to be strongly coupled @xcite .
therefore , one goes for phenomenological models like strongly coupled quark gluon plasma @xcite , quasi - particle models @xcite , etc . with few fitting parameters .
model results are fitted with the results of the simulation of lattice gauge theory ( lgt ) for qgp @xcite . here
we study and compare different quasiparticle models and discuss sm and td consistencies of each models .
next section we discuss the quasiparticle model which was first introduced .
the modification of this model by eliminating the td inconsistency is presented in section iii . another alternative way without td and sm inconsistencies
is presented in section iv .
discussions and conclusions is presented in section v. +
quasi - particle model for qgp was first introduced by goloviznin and satz @xcite , and then peshier _ et .
al . _ for gluon plasma @xcite .
following the recent article by castorina , miller and satz @xcite , one starts from the expression for pressure , + @xmath0 @xmath1 where @xmath2 and t is the temperature .
@xmath3 is the modified bessel function of order 2 .
@xmath4 is the thermal mass which is a function of temperature .
basic idea behind the quasiparticle model is that the effects of mutual interaction among the constituents in nonideal plasma is taken in the form of effective mass or thermal mass and treat the system as ideal .
there are different phenomenological models for thermal masses and here we use the recent model , @xcite , + @xmath5 where @xmath6 and @xmath7 are parameters of the model .
@xmath8 is the phase - transition temperature to gluon plasma .
the structure of this @xmath9 is discussed , in detail , in ref .
@xcite which involves the main properties of phase transition near critical temperature and proper high temperature limit for large temperature .
all other thermodynamic quantities like energy density , speed of sound , etc . are derived from pressure using thermodynamic relations . for e.g. , the energy density , @xmath10 ,
is + @xmath11 \,\,.\ ] ] let us call this model qpm - i .
however , above expression for energy density leads to statistical mechanics inconsistency . we know that
when we treat an equilibrium system as canonical ensemble ( ce ) or grand canonical ensemble ( gce ) the energy of the system is not constant and fluctuates and we define average energy @xcite and hence energy density as + @xmath12 \,\,.\ ] ] therefore , we see that @xmath10 @xmath13 @xmath14 and hence sm inconsistency .
this model violates the standard sm of ce or gce system .
but , since @xmath9 has few parameters , we can tune them to fit lgt data for pressure and them obtain other td quantities , using td relations .
hence it is td consistent .
thus we see that this model is td consistent , but sm inconsistent . and @xmath15 as a function of @xmath16 for gluon plasma from 3 quasiparticle models qpm - i , qpm - ii , qpm - iii and compared with lgt results of ref .
@xcite ( symbols).,width=453,height=302 ] in this model , qpm - ii , we follow peshier _ et .
@xcite where one starts from pressure , just like the earlier model , + @xmath17 however , the energy density @xmath10 is + @xmath18 = \varepsilon_{sm } \,\,,\ ] ] which is different from qpm - i .
hence it is sm consistent .
but the drawback is that it is td inconsistent as pointed out by various authors @xcite .
to solve this problem , borrowing idea from the bag model of hadron spectroscopy , one introduces bag pressure term @xmath19 in the expression for pressure as , + @xmath20 and energy density , + @xmath18 + b(t ) \,\ , .
\label{eq : e}\ ] ] this @xmath19 was interpreted as temperature dependent bag pressure , vacuum pressure , residual interaction energy , etc .
then one exploit @xmath19 to solve the td inconsistency problem by demanding that pressure and energy density , eqs .
( [ eq : p ] ) and ( [ eq : e ] ) , be td consistent .
this leads to td consistent relation , + @xmath21 and hence @xmath19 is related to @xmath9 .
thus the qpm - ii is now td and sm consistent . and @xmath15 as a function of @xmath16 for gluon plasma from 3 quasiparticle models qpm - i , qpm - ii , qpm - iii at boltzmann limit and compared with lgt results of ref .
@xcite ( symbols).,width=453,height=302 ] here we follow yet another approach , qpm - iii , which follow straight forward from standard sm .
once the interaction effects of nonideal system is absorbed in thermal mass and treat the resultant system as ideal system in ce or gce .
we know that in ce or gce , one defines average energy @xcite and hence we start from energy density , + @xmath18 = \varepsilon_{sm } \,\,,\ ] ] and hence sm consistent
. pressure is obtained from the td relation , + @xmath22 on integration and thus it is td consistent also .
therefore , this model is both sm and td consistent and no need of @xmath19 . of course , one can include @xmath19 in this model also , but then @xmath19 is free function of @xmath23 . as we have shown in ref .
@xcite this model gives qpm - ii with the help of an extra constraint equation , eq .
( [ eq : tdc ] ) . [ cols="^,^,^,^,^,^,^ " , ] + * table 1 * +
thus we saw that qpm - i is sm inconsistent , but by exploiting the phenomenological parameters we can fit the lgt data @xcite of gluon plasma for pressure .
values of parameters are tabulated in table - i .
but it is interesting that even after the violation of sm , by adjusting three parameters one can get reasonably good fit as shown in fig .
1 . using the 3 fitting parameters other td quantities like @xmath10 is evaluated from the td relations and also plotted in fig . 1 and compared with lgt data and fits well as it should .
next we discuss the results of qpm - ii .
it is sm consistent and also td consistent by td consistent relation , and again reasonably good fit to lgt data for pressure and energy density are obtained by adjusting 3 parameters which are tabulated in table - i .
plots are given in fig .
1 . results of qpm - iii , which is both sm and td consistent , is also plotted in fig . 1 , by fitting energy density with lgt data by adjusting 3 parameters .
values of fitted parameters are given in table -i .
again reasonably good fits are obtained .
it is interesting to note that even the single term in the summation for the expression of thermodynamic quantities , pressure or energy density , in all three models , also fits the lgt data by different sets of fitting parameters given in table - i ( right ) .
these results are equivalent to the boltzmann limit discussed in ref .
plots are compared in fig .
2 . in the ref.@xcite
the reason for the 3 parameters in @xmath9 is discussed .
@xmath25 is related to the critical exponent of phase transition and is estimated to be @xmath26 for su(2 ) .
parameter @xmath27 is introduced based on the fact that thermal mass is supposed to be proportional to @xmath23 as @xmath28 .
we may even demand that at high temperature @xmath9 may goes as @xmath29 where @xmath30 is the plasma frequency which follows from perturbative results of quantumchromodynamics ( qcd ) .
therefore , we may take @xmath31 where @xmath32 , which is not constant as in ref .
@xcite , but running constant because of @xmath33 .
we used the qcd perturbative result , @xmath34 .
we may take 2-loop perturbative coupling constant , @xmath35 where @xmath36 from the lattice studies ref .
@xcite with which we compare our results .
assuming that su(3 ) also have the similar phase transition as su(2 ) , we may take @xmath37 for gluon plasma .
now we have only one free parameter , @xmath38 , and by adjusting it to fit to lgt data using our model qpm - iii , we again get good fit as plotted in fig . 3 for @xmath39 .
advantage here is that 2 of 3 parameters of @xmath9 is fixed by physical arguments . at present
we do nt know to estimate the value of @xmath38 .
we also tried to fit lgt data using qpm - i and qpm - ii with the same @xmath40 and @xmath37 , but by adjusting @xmath38 .
however , both of these models fail to get good fit to lgt data . it is further interesting to note that the same model , qpm - iii , also fits the recent lattice results of wuppertal - budapest group @xcite , which deals with precision su(3 ) thermodynamics for a large temperature range , reasonably well as shown in fig .
also @xmath41 and @xmath37 , but @xmath33 is the 3-loop running coupling constant used in the same reference @xcite .
our result is less than 2 % lower than that of lattice for large temperature as can be seen from fig .
4 . there is no much difference even if we use earlier @xmath33 , eq.([eq : as ] ) , with the same @xmath36 . in conclusion ,
all three quasiparticle models @xcite , considered here , fit the lgt data on gluon plasma @xcite reasonably well by adjusting three parameters of a phenomenological thermal mass proposed recently by castorina , miller and satz @xcite .
however , only one model @xcite , qpm - iii , which is both statistical mechanics and thermodynamics consistent , fits the lgt data after fixing 2 of 3 parameters by physical arguments and adjusting the remaining one free parameter .
it is remarkable that the same model , qpm - iii , which fits bielefeld lattice data @xcite , also gives good fit to the recent lattice results of wuppertal - budapest group @xcite which includes data for larger range of temperature ( @xmath42 ) .
99 g. boyd , j. engels , f. karsch , e. laermann , c. legeland , m. lutgemeier and b. petersson , phys .
. lett . * 75 * , 4169 ( 1995 ) ; nucl .
b469 * , 419 ( 1996 ) .
s. borsanyi , g. endrodi , z. fodor , s. d. katz and k. k. szabo , [ hep - lat/1204.6184v1 ] .
r. k. pathria , _ statistical mechanics _ , butterworth - heinemann , oxford ( 1997 )
. v. m. bannur , phys . lett . *
b362 * , 7 ( 1995 ) . v. m. bannur , eur .
j. * c11 * , 169 ( 1999 ) ; j. phys .
g : nucl . part .
phys . * 32 * , 993 ( 2006 ) .
b. k. patra , v. agotiya and v. chandra , eur .
j. * c67 * , 465 ( 2010 ) . v. goloviznin and h. satz , z. phys .
* c57 * , 671 ( 1993 ) .
a. peshier , b. kampfer , o. p. pavlenko and g. soff , phys . lett .
* b337 * , 235 ( 1994 ) . m. i. gorenstein and s. n. yang , phys . rev .
* d52 * , 5206 ( 1995 ) .
a. peshier , b. kampfer and g. soff , phys . rev . * d66 * , 094003 ( 2002 ) .
v. m. bannur , eur .
j. * c50 * , 629 ( 2007 ) ; jhep09 , 046 ( 2007 ) ; phys . rev . * c78 * , 045206 ( 2008 ) ; phys
. rev . * c75 * , 044905 ( 2007 ) .
m. bluhm , b. kampfer , r. schulze , d. seipt and u. heinz , phys .
* c76 * , 034901 ( 2007 ) .
f. brau and f. buisseret , phys . rev .
* d79 * , 114007 ( 2009 ) .
f , g. gardim and f. m. steffens , nucl .
a825 * , 222 ( 2009 ) .
p. k. srivastava , s. k. tiwari and c. p. singh , phys . rev .
* d82 * , 014023 ( 2010 ) . s. yin and ru - keng su , phys . rev . * c77 * , 055204 ( 2008 ) .
a - meng zhao _ et .
a25 * , 47 ( 2010 ) .
j. cao , y. jiang , w. sun , and h. zong , phys .
b711 * , 65 ( 2012 ) .
l. i. zhu and c. b. yang , nucl .
* a831 * , 49 ( 2009 ) . c. jing _ et .
* 27 * , 031201 ( 2010 ) .
m. panero , phys .
lett . * 103 * , 232001(2009 ) .
p. castorina , d. e. miller and h. satz , eur .
j. * c71 * , 1673 ( 2011 ) .
f. buisseret and g. lacroix , eur .
j. * c70 * , 1051 ( 2010 ) . v. m. bannur , phys
* b647 * , 271 ( 2007 ) . | widely studied quasiparticle models for quark gluon plasma is revisited here to understand the statistical mechanics and thermodynamics of the system .
we investigate the statistical mechanics and thermodynamics inconsistencies involved in these models and their consequences in the observables .
quasiparticle model is a phenomenological model with few parameters and by adjusting them all models fit the results of lattice gauge simulation of gluon plasma @xcite .
however , after fixing 2 of the 3 parameters of the model by physical arguments , only one quasiparticle model , which is consistent with both statistical mechanics and thermodynamics , fits the bielefeld lattice data @xcite .
the same model also fits the recent lattice results of wuppertal - budapest group @xcite , which deals with precision su(3 ) thermodynamics for a large temperature range , reasonably well .
.75 cm * pacs
nos :* 12.38.mh , 05.30jp , 12.39.mk , 12.38.gc , 05.70.ce + * keywords :* equation of state , gluon plasma , quasiparticle model . |
it has long been the dream of stellar astronomers to study individual stellar objects in distant galaxies to obtain detailed spectroscopic information about the star formation history and chemodynamical evolution of galaxies and to determine accurate distances based on the determination of stellar parameters and interstellar reddening and extinction . at the first glance
, one might think that the most massive and , therefore , most luminous stars with masses higher than 50 @xmath0 are ideal for this purpose .
however , because of their very strong stellar winds and mass - loss these objects keep very hot atmospheric temperatures throughout their life and , thus , waste most of their precious photons in the extreme ultraviolet .
as we all know , most of these uv photons are killed by dust absorption in the star forming regions , where these stars are born , and the few which make it to the earth can only be observed with tiny uv telescopes in space such as the hst or fuse and are not accessible to the giant telescopes on the ground .
thus , one learns quickly that the most promising objects for such studies are massive stars in a mass range between 15 to 40 @xmath0 in the short - lived evolutionary phase , when they leave the hydrogen main - sequence and cross the hrd in a few thousand years as blue supergiants of late b and early a spectral type . because of the strongly reduced absolute value of bolometric correction when evolving towards smaller temperature these objects increase their brightness in visual light and become the optically brightest `` normal '' stars in the universe with absolute visual magnitudes up to @xmath1 rivaling with the integrated light brightness of globular clusters and dwarf spheroidal galaxies .
these are the ideal stellar objects to obtain accurate quantitative information about galaxies .
there has been a long history of quantitative spectroscopic studies of these extreme objects . in a pioneering and comprehensive paper on deneb
@xcite was the first to obtain stellar parameters and detailed chemical composition .
this work was continued by wolf ( 1971 , 1972 , 1973 ) in studies of a supergiants in the milky way and the magellanic clouds .
@xcite using newly developed nlte model atmospheres found that at the low gravities and the correspondingly low electron densities of these objects effects of departures from lte can become extremely important .
with strongly improved model atmospheres @xcite , @xcite and @xcite continued these studies in the milky way .
most recently , @xcite and @xcite used very detailed nlte line formation calculations including ten thousands of lines in nlte ( see also @xcite , @xcite , @xcite , @xcite , @xcite ) to determine stellar parameters and abundances with hitherto unkown precision ( @xmath2 to @xmath3% , @xmath4 to @xmath5 dex , individual metal abundances to @xmath5 dex ) . at the same time , utelizing the power of the new 8 m to 10 m class telescopes , high resolution studies of a supergiants in many local group galaxies were carried out by @xcite ( smc ) , @xcite ( m33 ) , @xcite ( m31 ) , @xcite ( m31 ) , @xcite ( ngc 6822 ) , @xcite ( wlm ) , and @xcite ( sextans a ) yielding invaluable information about the stellar chemical composition in these galaxies . in the research field of massive stars
, these studies have so far provided the most accurate and most comprehensive information about chemical composition and have been used to constrain stellar evolution and the chemical evolution of their host galaxies .
the concept to go beyond the local group and to study a supergiants by means of quantitative spectroscopy in galaxies out to the virgo cluster has been first presented by @xcite and @xcite . following - up on this idea , @xcite and @xcite used the vlt and fors at 5 @xmath6 resolution for a first investigation of blue supergiants in ngc 3621 ( 6.7 mpc ) and ngc 300 ( 1.9 mpc ) .
they were able to demonstrate that for these distances and at this resolution spectra of sufficient s / n can be obtained allowing for the quantitative determination of stellar parameters and metallicities .
@xcite extended this work and showed that stellar gravities and temperatures determined from the spectral analysis can be used to determine distances to galaxies by using the correlation between absolute bolometric magnitude and flux weighted gravity @xmath7@xmath8 ( fglr ) .
however , while these were encouraging steps towards the use of a supergiants as quantitative diagnostic tools of galaxies beyond the local group , the work presented in these papers had still a fundamental deficiency . at the low resolution of 5 @xmath6 it is not possible to use ionization equilibria for the determination of @xmath2 in the same way as in the high resolution work mentioned in the previous paragraph . instead
, spectral types were determined and a simple spectral type - temperature relationship as obtained for the milky way was used to determine effective temperatures and then gravities and metallicities . since the spectral type - @xmath2 relationship must depend on metallicity ( and also gravities ) , the method becomes inconsistent as soon as the metallicity is significantly different from solar ( or the gravities are larger than for luminosity class ia ) and may lead to inaccurate stellar parameters . as shown by @xcite , the uncertainties introduced in this way could be significant and would make it impossible to use the fglr for distance determinations .
in addition , the metallicities derived might be unreliable .
this posed a serious problem for the the low resolution study of a supergiants in distant galaxies .
this problem was overcome only very recently by @xcite ( herafter kubgp ) , who provided the first self - consistent determination of stellar parameters and metallicities for a supergiants in galaxies beyond the local group based on the detailed quantitative model atmosphere analysis of low resolution spectra .
they applied their new method on 24 supergiants of spectral type b8 to a5 in the scultor group spiral galaxy ngc 300 ( at 1.9 mpc distance ) and obtained temperatures , gravities , metallicities , radii , luminosities and masses .
the spectroscopic observations were obtained with fors1 at the eso vlt in multiobject spectroscopy mode .
in addition , eso / mpi 2.2 m wfi and hst / acs photometry was used .
the observations were carried out within the framework of the araucaria project ( @xcite ) . in the following
we discuss the analysis method and the results of this pilot study .
for the quantitative analysis of the spectra kubgp use the same combination of line blanketed model atmospheres and very detailed nlte line formation calculations as @xcite in their high signal - to - noise and high spectral resolution study of galactic a - supergiants , which reproduce the observed normalized spectra and the spectral energy distribution , including the balmer jump , extremely well .
they calculate an extensive , comprehensive and dense grid of model atmospheres and nlte line formation covering the potential full parameter range of all the objects in gravity ( @xmath4 = 0.8 to 2.5 ) , effective temperature ( @xmath2= 8300 to 15000k ) and metallicity ( [ z ] = logz/@xmath9 = -1.30 to 0.3 ) .
the total grid comprises more than 6000 models .
the analysis of the each of the 24 targets in ngc 300 proceeds in three steps .
first , the stellar parameters ( @xmath2 and @xmath4 ) are determined together with interstellar reddening and extinction , then the metallicity is determined and finally , assuming a distance to ngc 300 , stellar radii , luminosities and masses are obtained . for the first step , a well established method to obtain the stellar parameters of supergiants of late b to early a spectral type is to use ionization equilibria of weak metal lines ( oi / ii ; mgi / ii ; ni / ii etc . ) for the determination of effective temperature @xmath2 and the balmer lines for the gravities @xmath4 . however , at the low resolution of 5 the weak spectral lines of the neutral species disappear in the noise of the spectra and an alternative technique is required to obtain temperature information .
kubgp confirm the result by @xcite that a simple application of a spectral type - effective temperature relationship does not work because of the degeneracy of such a relationship with metallicity .
fortunately , a way out of this dilemma is the use of the spectral energy distributions ( seds ) and here , in particular of the balmer jump @xmath10 .
while the observed photometry from b - band to i - band is used to constrain the interstellar reddening , @xmath10 turns out to be a reliable temperature diagnostic , as is indicated by fig.[fig1 ]
. a simultaneous fit of the balmer lines and the balmer jump allows to constrain effective temperature and gravity independent of assumptions on metallicity .
fig.[fig2 ] demonstrates the sensitivity of the balmer lines and the balmer jump to gravity and effective temperature , respectively . at a fixed temperature
the balmer lines allow for a determination of @xmath4 within 0.05 dex uncertainty , whereas the balmer jump at a fixed gravity yields a temperature uncertainty of 2 percent .
however , since the isocontours in fig.[fig1 ] are not orthogonal , the maximum errors of @xmath2 and @xmath4 are 5 percent and 0.2 dex , respectively .
these errors are significantly larger than for the analysis of high resolution spectra , but they still allow for an accurate determination of metallicity and distances . the accurate determination of @xmath2 and @xmath4 is crucial for the use of a supergiants as distance indicators using the relationship between absolute bolometric magnitude @xmath11 and flux weighted gravity @xmath12 defined as @xmath13 where @xmath14/10000k ( see @xcite ) .
the relatively large uncertainties obtained with this fit method may casts doubts whether @xmath12 can be obtained accurately enough .
fortunately , the non - orthogonal behaviour of the fit curves in the left part of fig.[fig1 ] leads to errors in @xmath2 and @xmath4 , which are correlated in a way that reduces the uncertainties of @xmath12 .
this is demonstrated in right part of fig.[fig1 ] , which shows the corresponding fit curves of the balmer lines and d@xmath15 in the @xmath16 @xmath2 ) plane . as a consequence ,
much smaller uncertainties are obtained for @xmath12 , namely @xmath17 0.10 dex .
kubgp give a detailed discussion of physical reason behind this .
= -1.30 to 0.30 , as described in the text .
the dashed vertical lines give the edges of the spectral window as used for a determination of metallicity.,width=302 ] knowing the stellar atmospheric parameters @xmath2 and @xmath4 kubgp are able to determine stellar metallicities by fitting the metal lines with their comprehensive grid of line formation calculations .
the fit procedure proceeds in several steps .
first , spectral windows are defined , for which a good definition of the continuum is possible and which are relatively undisturbed by flaws in the spectrum ( for instance caused by cosmic events ) or interstellar emission and absorption . a typical spectral window used for all targets
is the wavelength interval 4497 @xmath18 4607 .
fig.[fig3 ] shows the synthetic spectrum calculated for the atmopsheric parameters of target no .
21 ( the previous example ) and for all the metallicities of the grid ranging from -1.30 @xmath19 [ z ] @xmath19 0.30 .
it is very obvious that the strengths of the metal line features are a strong function of metallicity . in fig.[fig4 ]
the observed spectrum of target no . 21 in this spectral region is shown overplotted by the synthetic spectrum for each metallicity .
separate plots are used for each metallicity , because the optimal relative normalization of the observed and calculated spectra is obviously metallicity dependent .
this problem is addressed by renormalizing the observed spectrum for each metallicity so that the synthetic spectrum always intersects the observations at the same value at the two edges of the spectral window ( indicated by the dashed vertical lines ) .
the next step is a pixel - by - pixel comparison of calculated and normalized observed fluxes for each metallicity and a calculation of a @xmath20-value .
the minimum @xmath21)$ ] as a function of [ z ] is then used to determine the metallicity .
this is also shown in fig.[fig4 ] .
application of the same method on different spectral windows provides additional independent information on metallicity and allows to determine the average metallicity obtained from all windows
. a value of is [ z ] = -0.39 with a very small dispersion of only 0.02 dex .
however , one also need to consider the effects of the stellar parameter uncertainties on the metallicity determination .
this is done by applying the same correlation method for [ z ] for models at the extremes of the error box for @xmath2 and @xmath4 .
this increases the uncertainty of [ z ] to @xmath17 0.15 dex , still a very reasonable accuracy of the abundance determination .
the fit of the observed photometric fluxes with the model atmosphere fluxes was used to determine interstellar reddening e(b - v ) and extinction a@xmath22 = 3.1 e(b - v ) .
simultaneously , the fit also yields the stellar angular diameter , which provides the stellar radius , if a distance is adopted .
alternatively , for the stellar parameters ( @xmath2 , @xmath4 , [ z ] ) determined through the spectral analysis the model atmospheres also yield bolometric corrections bc , which we use to determine bolometric magnitudes .
these bolometric magnitudes then also allow us to compute radii .
the radii determined with these two different methods agree within a few percent .
@xcite in their multi - wavelength study of a large sample of cepheids in ngc 300 including the near - ir have determined a new distance modulus m - m = 26.37 mag , which corresponds to a distance of 1.88 mpc .
kubgp have adopted these values to obtain the radii and absolute magnitudes . as a function of angular galacto - centric distance @xmath23 for the a supergiants (
filled circles ) and the early b - supergiants studied by @xcite ( filled squares ) .
note that for the latter metallicity refers to oxygen only .
the dashed curve represents the regression discussed in the text .
, width=302 ]
as a first result , the quantitative spectroscopic method yields interstellar reddening and extinction as a by - product of the analysis process . for objects embedded in the dusty disk of a star forming spiral galaxy
one expects a wide range of interstellar reddening e(b - v ) and , indeed , a range from e(b - v ) = 0.07 mag up to 0.24 mag was found .
the individual reddening values are significantly larger than the value of 0.03 mag adopted in the hst distance scale key project study of cepheids by @xcite and demonstrate the need for a reliable reddening determination for stellar distance indicators , at least as long the study is restricted to optical wavelengths .
the average over the observed sample is @xmath24 = 0.12 mag in close agreement with the value of 0.1 mag found by @xcite in their optical to near - ir study of cepheids in ngc 300 . while cepheids have somewhat lower masses than the a supergiants of our study and are consequently somewhat older , they nonetheless belong to the same population and are found at similar sites .
thus , one expects them to be affected by interstellar reddening in the same way as a supergiants .
fig.[fig5 ] shows the location of all the observed targets in the @xmath25 @xmath2 ) plane and in the hrd compared to the early b - supergiants studied by @xcite .
the comparison with evolutionary tracks gives a first indication of the stellar masses in a range from 10 m@xmath26 to 40 m@xmath26 .
three targets have obviously higher masses than the rest of the sample and seem to be on a similar evolutionary track as the objects studied by @xcite .
the evolutionary information obtained from the two diagrams appears to be consistent .
the b - supergiants seem to be more massive than most of the a supergiants .
the same three a supergiants apparently more massive than the rest because of their lower gravities are also the most luminous objects .
this confirms that quantitative spectroscopy is -at least qualitatively - capable to retrieve the information about absolute luminosities .
note that the fact that all the b supergiants studied by @xcite are more massive is simply a selection effect of the v magnitude limited spectroscopic survey by @xcite . at similar v magnitude as the a supergiants those objects have higher bolometric corrections because of their higher effective temperatures and are , therefore , more luminous and massive .
fig.[fig6 ] shows the stellar metallicities and the metallicity gradient as a function of angular galactocentric distance , expressed in terms of the isophotal radius , @xmath23 ( @xmath27 corresponds to 5.33 kpc ) . despite the scatter caused by the metallicity uncertainties of the individual stars
the metallicity gradient of the young disk population in ngc 300 is very clearly visible . a linear regression for the combined a- and b - supergiant sample yields ( d in kpc ) @xmath28 = -0.06\pm{0.09 } - ( 0.083\pm{0.022})~d.\ ] ] note that the metallicities of the b supergiants refer to oxygen only with a value of log n(o)/n(h ) = -3.31 adopted for the sun ( @xcite ) . on the other hand , the a supergiant metallicities reflect the abundances of a variety of heavy elements such as ti , fe , cr , si , s , and mg . with these results kubgp extended the discussion started by @xcite to compare with oxygen abundances obtained from hii - region emission lines .
@xcite used line fluxes published by @xcite and applied various different published strong line method calibrations to determine nebular oxygen abundances , which could then be used to obtain the similar regressions as above .
the different strong line method calibrations lead to significant differences in the central metallicity as well as in the abundance gradient .
the calibrations by @xcite and @xcite predict a metallicity significantly supersolar in the center of ngc 300 contrary to the other calibrations . on the other hand , the work by kubgp yields a central metallicity slightly smaller than solar in good agreement with @xcite and marginally agreeing with @xcite , @xcite , and @xcite . at the isophotal radius , 5.3 kpc away from the center of ngc 300
, they obtain an average metallicity significantly smaller than solar [ z ] = -0.50 , close to the average metallicity in the smc .
the calibrations by @xcite , @xcite , @xcite do not reach these small values for oxygen in the hii regions either because their central metallicity values are too high or the metallicity are gradients too shallow . in light of the substantial range of metallicies
obtained from hii region emission lines using different strong line method calibrations it seems to be extremely valuable to have an independent method using young stars .
it will be very important to compare our results with advanced work on hii regions , which will avoid strong line methods , and will use direct infomation about nebular electron temperatures and densities ( bresolin et al .
, 2008 , in preparation , see also this volume ) .
kubgp discuss the few outliers in fig.[fig6 ] and claim that these metallicities seem to be real .
their argument is that the expectation of homogeneous azimuthal metallicity in patchy star forming galaxies seems to be naive .
future work on other galaxies will show whether cases like this are common or not .
massive stars with masses in the range from 12 m@xmath26 to 40 m@xmath26 evolve through the b and a supergiant stage at roughly constant luminosity ( see fig.[fig5 ] ) .
in addition , since the evolutionary timescale is very short when crossing through the b and a supergiant domain , the amount of mass lost in this stage is small .
this means that the evolution proceeds at constant mass and constant luminosity .
this has a very simple , but very important consequence for the relationship of gravity and effective temperature along each evolutionary track
. from @xmath29 follows immediately that @xmath30 thus , along the evolution through the b and a supergiant domain the _ `` flux - weighted gravity '' _
@xmath31/@xmath8 should remain constant .
this means each evolutionary track of different luminosity in this domain is characterized by a specific value of @xmath32 .
this value is determined by the relationship between stellar mass and luminosity , which in a first approximation is a power law @xmath33 and leads to a realtionship between luminosity and flux - weighted gravity @xmath34 with the definition of bolometric magnitude @xmath35@xmath36@xmath37 one then derives @xmath38 this is the _ `` flux - weighted gravity
luminosity relationship '' _ ( fglr ) of blue supergiants .
note that the proportionality constant @xmath39 is given by the exponent of the mass luminosity power law through @xmath40 for instance , for @xmath41 , one obtains @xmath42 .
note that the zero point of the relationship is chosen at a flux weighted gravity of 1.5 , which is in the middle of the range encountered for blue supergiant stars .
kubgp use the mass - luminosity relationships of different evolutionary tracks ( with and without rotation , for milky way and smc metallicity ) to calculate the different fglrs predicted by stellar evolution . very interestingly ,
while different evolutionary model types yield somewhat different fglrs , the differences are rather small .
@xcite were the first to realize that the fglr has a very interesting potential as a purely spectroscopic distance indicator , as it relates two spectroscopically well defined quantitities , effective temperature and gravity , to the absolute magnitude . compiling a large data set of spectroscopic high resolution studies of a supergiants in the local group and with an approximate analysis of low resolution data of a few targets in galaxies beyond the local group ( see discussion in previous chapters ) they were able to prove the existence of an observational fglr rather similar to the theoretically predicted one . with the improved analysis technique of low resolution spectra of a supergiants and with the much larger sample studied for ngc 300
kubgp resumed the investigation of the fglr .
the result is shown in fig.[fig7 ] , which for ngc 300 ( left diagram ) reveals a clear and rather tight relationship of flux weighted gravity @xmath12 with bolometric magnitude @xmath11 .
a simple linear regression yields @xmath43 = 8.11 for the zero point and @xmath39 = -3.52 for the slope .
the standard deviation from this relationship is @xmath44 = 0.34 mag . within the uncertainties
the observed fglr appears to be in agreement with the theory . in their first investigation of the empirical fglr @xcite
have added a supergiants from six local group galaxies with stellar parameters obtained from quantitative studies of high resolution spectra ( milky way , lmc , smc , m31 , m33 , ngc 6822 ) to their results for ngc 300 to obtain a larger sample .
they also added 4 objects from the spiral galaxy ngc 3621 ( at 6.7 mpc ) which were studied at low resolution .
kubgp added exactly the same data set to their new enlarged ngc 300 sample , however , with a few minor modifications .
for the milky way they included the latest results from @xcite and @xcite and for the two objects in m31 we use the new stellar parameters obtained by @xcite .
for the objects in ngc 3621 they applied new hst photometry .
they also re - analyzed the lmc objects using ionization equilibria for the temperature determination .
fig.[fig7 ] ( right diagram ) shows bolometric magnitudes and flux - weighted gravities for this full sample of eight galaxies revealing a tight relationship over one order of magnitude in flux - weighted gravity .
the linear regression coefficients are @xmath45 and @xmath46 , very simimilar to the ngc 300 sample alone .
the standard deviation is @xmath44 = 0.32 mag .
the stellar evolution fglr for milky way metallicity provides a fit of almost similar quality with a standard deviation of @xmath44 = 0.31 mag .
the astrophysical potential of low resolution spectroscopy of a supergiant stars for studying galaxies beyond the local group is quite remarkable . by introducing a novel method for the quantitative spectral analysis one
is able to determine accurate stellar parameters , which allow for a detailed test of stellar evolution models . through the spectroscopic determination of stellar parameters one can also constrain interstellar reddening and extinction by comparing the calculated sed with broad band photometry .
the study of ngc 300 finds a very patchy extintion pattern as to be expected for a star forming spiral galaxy .
the average extinction is in agreement with multi - wavelength studies of cepheids including k - band photometry .
the method also allows to determine stellar metallicities and to study stellar metallicity gradients .
solar metallicity is found in the center of ngc 300 and a gradient of @xmath470.08 dex / kpc . to our knowledge
this is the first systematic stellar metallicity study in galaxies beyond the local group focussing on iron group elements . in the future
the method can be extended to not only determine metallicity but also the ratio of @xmath48- to iron group elements as a function of galactocentric distance .
the stellar metallicities obtained can be compared with oxygen abundance studies of hii regions using the strong line method .
this allows to discuss the various calibrations of the strong line method , which usually yield very different results .
the improved spectral diagnostic method makes it possible to very accurately determine stellar flux weighted gravities @xmath12 = @xmath4/@xmath8 and bolometric magnitudes . above a certain threshold in effective temperature
a simple measurement of the strengths of the balmer lines can be used to determine accurate values of @xmath12 .
absolute bolometric magnitudes @xmath11 and flux - weighted gravities @xmath12 are tightly correlated .
it is shown that such a correlation is expected for stars , which evolve at constant luminosity and mass . this `` flux - weighted gravity - luminosity relationship ( fglr ) '' agreement with stellar evolution theory . with a relatively small residual scatter of @xmath44 = 0.3 mag
the observed fglr is an excellent tool to determine accurate spectroscopic distance to galaxies .
it requires multicolor photometry and low resolution ( @xmath49 ) spectroscopy to determine effective temperature and gravity and , thus , flux - weighed gravity directly from the spectrum . with effective temperature , gravity and metallicity
determined one also knows the bolometric correction , which is small for a supergiants , which means that errors in the stellar parameters do not largely affect the determination of bolometric magnitudes .
moreover , one knows the intrinsic stellar sed and , therefore , can determine interstellar reddening and extinction from the multicolor photometry , which then allows for the accurate determination of the reddening - free apparent bolometric magnitude .
the application of the fglr then yields absolute magnitudes and , thus , the distance modulus . with the intrinsic scatter of @xmath44 = 0.3 mag and 30 targets per galaxy
one can estimate an accuracy of 0.05 mag in distance modulus ( 0.1 mag for 10 target stars ) .
the results for wlm by miguel urbaneja et al . and
for m33 by vivian u et al .
presented in these proceedings are a first demonstration of the power of the method .
the advantage of the fglr method for distance determinations is its spectroscopic nature , which provides significantly more information about the physical status of the objects used for the distance determination than simple photometry methods .
most importantly , metallicity and interstellar extinction can be determined directly .
the latter is crucial for spiral and irregular galaxies because of the intrinsic patchiness of reddening and extinction .
since supergiant stars are known to show intrinsic photometric variability , the question arises whether the fglr method is affected by such variability . for the targets in ngc 300
this issue has been carefully investigated by @xcite , who studied ccd photometry lightcurves obtained over many epochs in the parallel search for cepheids in ngc 300 .
they concluded that amplitudes of photometric variability are very small and do not affect distance determinations using the fglr method .
the effects of crowding and stellar multiplicity are also important . however , in this regard a supergiants offer tremendous advantages relative to other stellar distance indicators .
first of all , they are significantly brighter . @xcite using hst acs photometry compared to ground - based photometry have studied the effects of crowding on the cepheid distance to ngc 300 and concluded that they are negligible . with a supergiants being 3 to 6 magnitudes brighter than cepheids it is clear that even with ground - based photometry only crowding is generally not an issue for these objects at the distance of ngc 300 and , of course , with hst photometry ( and in the future jwst ) one can reach much larger distances before crowding becomes important .
in addition , any significant contribution by additional objects to the light of an a supergiant will become apparent in the spectrum , if the contaminators are not of a very similar spectral type , which is very unlikely because of the short evolutionary lifetime in the a supergiant stage .
it is also important to note that a supergiants have evolutionary ages larger than 10 million years , which means that they have time to migrate into the field or that they are found in older clusters , which are usually less concentrated than the very young ob associations .
it is evident that the type of work described in this paper can be in a straightforward way extended to the many spiral galaxies in the local volume at distances in the 4 to 7 mpc range .
pushing the method we estimate that with present day 8 m to 10 m class telescopes and the existing very efficient multi - object spectrographs one can reach down with sufficient s / n to v = 22.5 mag in two nights of observing time under very good conditions . for objects brighter than @xmath50 = -8
mag this means metallicities and distances can be determined out to distances of 12 mpc ( m - m = 30.5 mag ) .
this opens up a substantial volume of the local universe for metallicity and galactic evolution studies and independent distance determinations complementary to the existing methods . with the next generation of extremely large telescopes such as the tmt , gmt or the e - elt the limiting magnitude
can be pushed to v = 24.5 equivalent to distances of 30 mpc ( m - m = 32.5 mag ) .
, 1998 , _ quantitative spectroscopy of the brightest blue supergiant stars in galaxies _ , in `` stellar physics for the local group '' , eds .
a. aparicio , a. herrero & f. sanchez ( cambridge university press ) , p. 149 | a supergiants are objects in transition from the blue to the red ( and vice versa ) in the uppermost hrd .
they are the intrinsically brightest `` normal '' stars at visual light with absolute visual magnitudes up to -9 .
they are ideal to study young stellar populations in galaxies beyond the local group to determine chemical composition and evolution , interstellar extinction , reddening laws and distances .
we discuss most recent results on the quantitative spectral analysis of such objects in galaxies beyond the local group based on medium and low resolution spectra obtained with the eso vlt and keck .
we describe the analysis method including the determination of metallicity and metallicity gradients . a new method to measure accurate extragalactic distances based on the stellar gravities and effective temperatures
is presented , the flux weighted gravity - luminosity relationship ( fglr ) .
the fglr is a purely spectroscopic method , which overcomes the untertainties introduced by interstellar extinction and variations of metallicity , which plague all photometric stellar distance determination methods .
we discuss the perspectives of future work using the giant ground - based telescopes of the next generation such as the tmt , the gmt and the e - elt . |
veritas , located at the fred lawrence whipple observatory near tucson , arizona , is an array of four 12-meter imaging atmospheric cherenkov telescopes .
each telescope has a pixelated camera comprised of 499 photomultiplier tubes with a @xmath2 field of view . designed to detect photons of astrophysical origin between 100 gev and 30 tev , veritas detects and images the secondary cherenkov light produced when gamma rays and cosmic rays initiate particle cascades in the upper atmosphere .
stereoscopic reconstruction of events using multiple telescopes allows for single - photon angular resolution of better than @xmath3 and energy resolution on the order of 15 - 25% . in its current configuration
, veritas can detect a source with @xmath4 of the crab nebula flux in less than 30 hours .
the instrument has been operating in full array mode since 2007 .
see @xcite for further details regarding the operation of veritas .
snr g78.2 + 2.1 , also known as the @xmath0-cygni supernova remnant ( snr ) , is a @xmath5 diameter shell - like radio and x - ray snr@xcite .
it is estimated to be at a distance of @xmath6 kpc@xcite and to be approximately @xmath7 years old@xcite .
it is thought to be in an early phase of adiabatic expansion into a low - density medium@xcite .
a slowly expanding h i shell immediately surrounding the radio shell was found by @xcite , which @xcite believes to have been created by the progenitor stellar wind .
the radio shell divides roughly into northern and southern arcs @xcite .
enhanced thermal x - ray emission in the north suggests shocked gas @xcite and falls in a void of co emission @xcite .
strong optical emission with sulfur lines also characterizes the region @xcite .
the @xmath0-ray satellite _ fermi _ , operating at gev energies , has discovered a @xmath0-ray pulsar at the center of the remnant @xcite .
while this pulsar has a low luminosity ( @xmath8 erg s@xmath9 ) and a spin - down age ( 76.8 kyr ) much greater than the estimated age of , the pulsar kinematics make it probable that was born with something close to its current spin period and is the remnant of s progenitor star @xcite .
diffuse @xmath0-ray emission above 10 gev is also reported by _ fermi _ over the full extent of the remnant @xcite .
a point source co - located with was previously reported in the first and second _ fermi _ catalogs @xcite , but @xcite conclude it to be artifact .
@xcite reported an extended region of emission above a few gev that they interpreted as a freshly accelerated cocoon of cosmic rays .
the cavity defining the cocoon region is outlined by ionization fronts visible in the mid- infrared .
@xmath0-cygni overlaps one end of the cocoon and is a potential source of the cocoon s trapped cosmic rays .
figure [ fig : excesswithoverlay ] displays the acceptance - corrected very - high - energy ( vhe ) @xmath0-ray excess map of the area around @xcite . an extended source is seen overlapping the northern edge of the remnant , with a detection significance of .
we use a binned , extended maximum - likelihood fit to the raw counts map to assess the morphology .
the source is modeled in the fit as a symmetric , two - dimensional gaussian convolved with the point - spread function ( psf ) ; the background is assumed to be flat before exposure effects are taken into account .
we find a fitted extension of the two - dimensional gaussian to be , with fitted centroid coordinates .
the statistical uncertainty in the centroid location is @xmath10 , with a systematic uncertainty of @xmath11 .
the systematic uncertainty considers both the telescope pointing error and systematic errors of the fit itself @xcite .
the positions of the @xmath0-ray pulsar ( ) ( from ) and the centroid of the emission above 10 gev from the remnant ( as seen by _ fermi _ ) are also shown for reference .
, width=302 ] figure [ fig : tevspec ] shows the spectrum of reconstructed @xmath0-ray events within from r.a .
@xmath12 , decl .
@xmath13 . runs where only three of four telescopes were operational have been excluded from this sample @xcite .
the photon spectrum is consistent with a differential power law in energy , @xmath14 , between the analysis threshold of and 10 tev .
the photon index is and the flux normalization at 1 tev is . the integral flux above ( ) corresponds to of the crab nebula flux above that energy @xcite .
figure [ fig : asca ] illustrates the region of enhanced x - ray emission overlapping @xcite .
it displays the @xmath15 kev exposure - corrected x - ray map using data from asca sequence # 25010000 ( data originally presented by @xcite ) , generated by co - adding data from the two gas imaging spectrometers .
a spectrum was extracted from a @xmath16 elliptical region enclosing most of the x - ray emission inside the veritas contours , centered on coordinates r.a .
@xmath17 @xmath18 @xmath19 , decl .
@xmath20 @xmath21 @xmath22 ( j2000 ) and oriented with position angle @xmath23 .
we selected background photons from an identically sized ellipse near the center of the remnant at r.a .
@xmath17 @xmath24 @xmath25 , decl .
@xmath20 @xmath26 @xmath27 ( j2000 ) , with position angle is @xmath28 .
the source and background regions are displayed in figure [ fig : asca ] . .
the stepped line shows the fit of a raymond - smith thermal plasma model with parameters as given in the text .
bottom panel residuals from the best - fit model.,width=302 ] the source spectrum between @xmath29 and @xmath30 kev is shown in figure [ fig : ascaspec ] @xcite .
the x - ray spectrum was modeled using an absorbed raymond - smith thermal plasma model with a best - fit temperature of @xmath31 kev . with this model the column density is @xmath32 , the normalization is @xmath33 m and the absorption - corrected flux is @xmath34 in the [email protected] kev band .
this result differs significantly from that given in @xcite , which claimed an additional power - law component and a large ne ix line feature .
the divergence in results appears to hinge , not on the choice of source region , which is similar , but on the choice of background region @xcite .
we selected a background region that was as close as possible to the source while @xcite chose a region @xmath2 away .
we can produce uchiyama s results by choosing a background region similar to uchiyama s .
for further details , see @xcite .
it is plausible that the vhe @xmath0-ray emission seen from @xmath0-cygni arises from particles accelerated in shocks occuring at the interaction of the supernova ejecta and the surrounding medium .
these particles could be either accelerated electrons , which would produce the emission via inverse - compton scattering , or accelerated nuclei .
should they be high - energy electrons , they would also be expected to produce x - ray synchrotron radiation , which would appear as a non - thermal power - law component in the x - ray spectrum . while our analysis of the asca x - ray spectrum does not argue for a non - thermal component , our upper limit on this component is still weak enough that we can not exclude the possibility that the tev , if not the gev , emission is due to inverse - compton scattering @xcite . on the other hand ,
it is also plausible that the vhe @xmath0-ray emission is produced by interaction of accelerated nuclei with the hi shell surrounding the remnant . estimates of the target material density required for accelerated nuclei to produce the observed vhe @xmath0-ray flux , based on @xcite , give a range of densities @xcite consistent @xcite s estimates of the gas density within the hi shell .
however , it must be noted that the shock velocities inferred from the optical and x - ray data are too low for the forward shock to be currently accelerating particles to tev energies . if the vhe @xmath0-ray emission is hadronic , it is likely due to particles accelerated when the remnant was younger that are now interacting with the shell . the relationship of the cocoon of freshly - accelerated cosmic rays detected by _ fermi
_ to and also remains unclear .
it is possible either has injected or is injecting accelerated particles into the cocoon . however , while they are shown for reference , we caution against using the cocoon contours from @xcite to judge the relationship of the cocoon to the vhe @xmath0-ray emission , since they are derived from an analysis where , which is no longer considered an independent source , was included as part of the background model .
it should also be noted that the vhe @xmath0-ray excess map in this paper was made with the ring - background estimation method , which is ill - suited to detecting a large - scale ( @xmath36 square degree ) region of @xmath0-ray emission such as the cocoon .
therefore the vhe @xmath0-ray maps shown here can not be used to set a meaningful upper limit on cocoon emission above 300 gev .
a conclusion determination of the relationship between and will have to await further data , analyzed with more sophisticated analysis techniques .
this research was supported by grants from the u.s .
department of energy , the u.s . national science foundation and the smithsonian institution , by nserc in canada , by science foundation ireland and by stfc in the uk .
we acknowledge the work of the technical support staff at the fred lawrence whipple observatory and at the collaborating institutions in the construction and operation of the instrument .
amanda weinstein and vikram dwarkadas acknowledge the support of nasa grant nnx11ao86 g .
the cygnus cocoon contours were graciously provided in electronic form by the fermi lat team . | the study of @xmath0-ray emission from galactic sources such as supernova remnants ( snr ) may provide key insights into their potential role as accelerators of cosmic rays up to the knee ( @xmath1 ev ) .
the veritas observatory is sensitive to galactic and extragalactic @xmath0-ray sources in the 100 gev to 30 tev energy range .
we report here on veritas observations of the vicinity of the cocoon of freshly accelerated cosmic rays reported by fermi , which lies between potential accelerators in the cygnus ob2 association and the @xmath0-cygni snr . a particular focus is placed on the source ver j2019 + 407 in @xmath0-cygni . |
understanding the origin of asymmetric outflows in pne and ppne requires feedback between observers , specific object modelers , and paradigm - seeking , order of magnitude theorists . here
we behave as the latter .
generally , ppne exhibit a combination of a fast bipolar outflow embedded within a slow spherically symmetric wind from the agb star @xcite .
presently , the data do not rule out all ppne having gone through a strongly asymmetric outflow and all pne having gone through an asymmetric pne phase .
the symmetry of pne would then correlate with age and the evolution from the ppne to pne could reflect a time evolution of the same physical mechanism that produces asymmetry on small scales but leads to a a nearly spherical structure on large scale as supersonic motions damp . while agb stars produce spherically symmetric outflows , ppne asymmetry arises within @xmath0 yr @xcite . for ppne @xcite , each fast wind has a typical age @xmath1yr , speed @xmath2km / s , mass @xmath3 , outflow rate , @xmath4 , momentum @xmath5 , and mechanical luminosity @xmath6 ( can be as high as @xmath7erg / s ) .
the slow ppne wind has an age @xmath8yr , a speed @xmath9km / s a mass @xmath10 , outflow rate , @xmath11 , momentum @xmath12 , and mechanical luminosity @xmath13
. for pne observations suggest e.g. @xcite an age @xmath14yr a slow wind of speed @xmath15km / s of mass @xmath16 , outflow rate , @xmath17 , momentum @xmath18 , and mechanical luminosity @xmath19 .
pne have fast winds of speed @xmath20km / s , mass @xmath21 , outflow rate , @xmath22 , momentum @xmath23 , and mechanical luminosity @xmath24 .
the ppne phase demands the most power .
the linear momenta of fast bipolar ppne outflows seems too large for radiation driving @xcite .
this motivates the need to tap rotational energy which binaries can provide .
the high fraction of binaries attributed to p / pne has led to speculation that all asymmetric pne and ppne involve binaries @xcite .
how is the rotational energy and angular momentum converted into outflows ?
differential rotation supplied by binaries can amplify magnetic fields which can in turn produce accretion powered bipolar jets .
accretion powered , magnetically mediated jets , are seemingly ubiquitous in astrophysics and can accommodate the high momentum ppne demands @xcite .
in a common envelope ( ce ) @xcite the companion drags on the envelope of the primary , transferring angular momentum and kinetic energy . provided that the envelope cooling time scale exceeds the energy supply time from the in - spiraling companion , a converted fraction @xmath25 of its loss in gravitational energy can spin up the envelope and unbind it . for a fixed envelope mass , a lower mass companion must fall deeper to unbind the envelope .
1 @xcite shows aspects of ce for low mass companions in an agb envelope . when the roche radius is reached by the in - spiraling companion
, accretion can occur .
young agb star ( left ) and inter - pulse agb star ( right ) , the solid lines show the radius where a companion must in - spiral to unbind the envelope for different values of the drag parameter @xmath26 .
the dash - dotted line shows the core - envelope boundary .
the long - dashed line marks the tidal shredding radius and the short dashed line is where the companion first fills its roche lobe.,title="fig:",height=188 ] young agb star ( left ) and inter - pulse agb star ( right ) , the solid lines show the radius where a companion must in - spiral to unbind the envelope for different values of the drag parameter @xmath26 .
the dash - dotted line shows the core - envelope boundary .
the long - dashed line marks the tidal shredding radius and the short dashed line is where the companion first fills its roche lobe.,title="fig:",height=188 ] ref @xcite discusses the secondaries for which accretion disks will form around the primary core .
brown dwarf ( bd ) ( @xmath27 ) radii increase with decreasing mass unlike their roche radii which decrease with decreasing mass .
such objects will unstably lose mass .
since the circularization radius lies outside of the primary s core , a disk can form within a few orbit times .
this contrasts the @xmath28 case for which the stellar radius decreases with decreasing mass and more strongly so than the roche radius .
supercritical companions present a circularization radius _ within _ the primary s core .
material leaving the secondary would then initially spiral swiftly into the primary rather than orbit quasi - stably .
ref.@xcite focuses on bds ( @xmath29 ) , but accretion from low mass stars and planets warrant further study : fig.1 ( left ) shows that for low values of the ( uncertain ) drag parameter @xmath26 , some low mass stars could tidally shred into a disk before reaching the radius where they unbind the envelope and subsequent in - spiraling is potentially slowed .
1 also shows that planets will shred into a disk upon inspiral and for large @xmath26 in the inter - pulse agb , even large planets might unbind the envelope .
generalizations must incorporate the fact that the companion size can be of order the tidal shredding or roche lobe radius , and the structural evolution of the primary as the secondary in - spirals .
the latter is important for radii both exterior and interior to the companion : how fast do further inspiral and angular momentum transfer subsequently occur once the outer envelope is ejected ?
if a stellar companion unbinds the envelope at a radius with roche lobe overflow but without tidal shredding , sustained accretion requires the binary to lose angular momentum .
as the secondary fills its roche lobe , it will drag on any residual inner envelope material . this could transfer the needed angular momentum , but too much drag could prevent the companion from forming a quasi - stable keplerian disk .
material that initially leaves the companion for the primary also carries magnetic fields and thus will magnetically link the primary and secondary . even though the circularization radius for @xmath30 is inside the core , the in - spiraling material still incurs differential rotation .
magnetic fields can then be amplified and can also release angular momentum .
much of the initial accreted energy in the @xmath30 case would be released upon material impact to stellar surface .
this could produce dwarf novae type bursts shrouded by the stellar envelope .
however , if the right range of mass and angular momentum are transferred in this initial accretion phase to ( 1 ) drop @xmath31 below @xmath32 , ( 2 ) keep @xmath31 filling its roche lobe , and ( 3 ) leave enough angular momentum to form a keplerian disk , then accretion could proceed as for the initial @xmath33 case of ref.@xcite . ref .
@xcite suggests a @xmath34 year delay in the time scale between ejection of cicumbinary dust tori and presence of jets in ppne .
if the companion initially has @xmath35 , the fraction of the envelope ejected as it spirals in might become the dust torus , and the delay before the jet could be the time it takes for the companion to lose enough mass to move the circularization radius outside the core .
the role of low mass stellar and planetary companions is important because although @xmath36 of nearby ( @xmath37pc ) sun - like ( late f to early k ) stars have companions more massive than jupiter at @xmath38au , 11% are stars , 5% are planets , and @xmath39 are bds @xcite .
if this @xmath36 were crudely taken as the fraction of binary induced , and thus asymmetric p / pne , this is within a factor of 2 of estimates of the total fraction of low mass stars incurring any pne @xcite .
it is also possible for accretion disks to form around the secondary @xcite .
the accretion rate inferred from the bondi wind accretion formulae is = ( m_2m_1)^2(v / v_s)^4 [ 1+(v / v_s)^2]^3/2 , where @xmath40 is the orbital speed of the secondary and @xmath41 is the slow wind speed from the primary . in general , for @xmath42 , reasonable parameters provide an accretion rate compatible with the luminosities required at the pne stage if the companion is a main sequence star .
however , the outflow velocities of the fast wind in pne would require a wd companion , as seen in the next section .
complementarily , the luminosities of the fast wind ppne outflows would also require a wd companion .
accretion disk outflows have a mechanical luminosity of order @xcite l_m~gm_*m_a2 r_i = 4.5 10 ^ 36 _
-1m_*m_-4r_i,10 , [ 1 ] where @xmath43 is the dimensionless efficiency of conversion from accretion to outflow , @xmath44 is the inner disk radius , @xmath45 is the accretion rate , @xmath46 is newton s constant , and @xmath47 is the central stellar mass . in the last term @xmath48 , @xmath49 and @xmath50/yr . for an mhd outflow , eq .
( [ 1 ] ) equals the poynting flux at the launch surface . for propagation into a region of negligible inertia ,
the asymptotic outflow speed is @xmath51 @xcite where @xmath52 is the angular speed of field anchor point and @xmath53 is the radius where the poloidal outflow speed equals the alfvn speed .
this product is typically of order 1 - 3 times the escape speed of the inner most radius of the disk and is thus at least v_out~v_esc = 1600 ( m_*r_*,10)^1/2km / s . [ 2 ] time dependent accretion outflows described with eqs .
( [ 1 ] ) and ( [ 2 ] ) are consistent with the high ppne outflow mechanical luminosity and the fast pne wind speed of sec . 1
when @xmath54 corresponds to a wd ( for accretion onto primary @xmath55 , or secondary @xmath56 ) the outflow speed from ( [ 2 ] ) does not depend explicitly on @xmath45 ( only implicitly and weakly via @xmath44 which can exceed the stellar radius for a strong field ) and is thus largely time independent .
however , this speed depends on the inertia of material blocking the outflow .
conservation of momentum gives v_obs = m_fv_f f_m_env + m_f ~80km / s , [ 3 ] where @xmath57 is the mass in one of the fast collimated outflows , @xmath58 is the envelope mass , @xmath59 is the solid angle fraction intercepted by the collimated outflow , and @xmath60 is used to incorporate @xmath61 of ref .
the numbers have been scaled to the ppne case so that here @xmath62 and for an envelope of mass @xmath63 , giving the intercepted mass of 0.2 @xmath64 for @xmath65 .
( [ 3 ] ) then implies @xmath66km / s , within the range observed @xcite .
( [ 3 ] ) represents the observed speed of the fast when blocked and loaded by the envelope . by the end of the ppne phase ,
the envelope is quite extended , reducing the optical depth and revealing material moving at the `` free streaming '' fast wind speed . assuming a dust - to - gass mass ratio of 1/100 and micron sized grains of density of 2g/@xmath67 , the optical depth from dust is _
d ~2.5 10 ^ -3(n_d2.5 10 ^ -13cm^-3 ) ( _ d10 ^ -8cm^2 ) ( r10 ^ 18 cm ) , scaled for pne . for ppne , the density increases by a factor @xmath68 and @xmath69 is down by a factor of 10 , so @xmath70 , optically thick .
the different optical depths of ppne and pne can thus explain why observed pne fast winds can have @xmath71km / s , whilst those of ppne have @xmath72km / s .
keeping in mind our discussion of low mass stellar companions in sec 2 , note that ref .
@xcite considers a companion of mass @xmath73 and a shakura - sunyaev viscosity parameter @xmath74 , for which the accretion rate then decays as @xmath75yr .. using this in ( [ 1 ] ) with @xmath76 for @xmath77 yr with @xmath78 cm and @xmath79 gives @xmath80 .
this provides the needed power demands of sec . 1 for ppne after @xmath81 yr and for pne after @xmath82 yr . because the surface density evolves , the gas opacity evolves from thomson to kramer s after @xmath83 yr @xcite , and the height to radius ratio decreases substantially as the disk cools .
a fixed @xmath74 is self - consistent with the time evolving accretion and power above .
we refer to the `` launch '' region @xcite of mhd outflows as the region where the magnetic force and energy dominates the flow and thermal energy .
this extends to a height typically no greater than @xmath84 , where @xmath85 is the innermost radial scale of the engine .
( e.g. the inner radius of an accretion disk ) . in the launch region
the bulk flow is accelerating but is sub - alfnic until reaching @xmath86 .
the `` propagation '' region describes @xmath87 where the poloidal flow speed exceeds the alfn speed , eventually approaching its asymptotic speed .
presently , only the propagation region is observationally spatially resolved . with this distinction
, we describe 3 classes of mhd outflow related work .
_ launch to propagation : _ e.g. @xcite ; here the magnetic field is imposed to be ordered on a scale at least as large as the anchoring rotator .
the base of the rotator is typically a boundary condition .
calculations address how material is accelerated by the combination of centrifugal flinging of material along quasi - rigid field lines and/or a vertical gradient in the magnetic pressure .
both a poloidal and toroidal field component at the base are required .
the flow can be collimated and supersonic by hoop stresses before upon reaching the propagation region provided that there is an ambient pressure to collimate the magnetic field .
simulations can cover from the base of the launch region to a scale typically @xmath0 times the engine scale .
_ asymptotic propagation : _ e.g. @xcite ; here a collimated jet is injected on small scales , and the subsequent propagation and shaping by the ambient medium of specified magnetic and thermal properties is studied .
the jet and ambient medium parameters are varied to assess what conditions can produce the observed asymptotic , morphologies .
these simulations do not address the magnetic field origin or acceleration mechanism . _ field origin to launch :
_ e.g.@xcite ; the previous two categories do not address where the dynamically important large scale fields comes from in the first place .
accretion of flux may be difficult in a turbulent disk , but these large scale fields can plausibly be produced by a combination of a flow dominated helical dynamo inside of the rotator , followed by magnetic buoyancy , and a magnetically dominated helical dynamo relaxation in the corona @xcite .
the latter opens up structures to the large scales needed to drive jets , much as solar coronal loops open to create solar coronal holes .
this category of work focuses on the field origin , with kinematic estimates of the subsequent launch , but not the dynamical launch itself .
it is a frontier to couple the above categories , and each has limitations when separated from the other two .
for example , a self - consistently grown strong field dynamically mediating the launch need not necessarily imply a strong magnetic influence in the propagation region : if the launch region produces a super - magnetosonic collimated outflow , then even subsequent ballistic propagation into the propagation region would still emerge as collimated .
in addition , a super - magnetosonic outflow could become turbulent and the turbulence can amplify small scale magnetic energy to @xmath88 of equipartition with this turbulence . this can be a substantial fraction of the initial bulk flow energy and such a field would be responding to the flow , not the reverse . to summarize : the physics of the launch region ( @xmath89 ) involves : ( 1 ) origin of large scale magnetic fields , field buoyancy to coronae , field relaxation into larger coronal structures , ( 2 ) physics of centrifugal + magnetic acceleration of material from small to super - alfvnic speeds , or poynting flux driven bursts of acceleration , ( 3 ) criteria for steady or bursty jets , and ( 4 ) assessment of the extent of poynting flux domination .
the physics of the propagation region ( @xmath90au observationally resolved ) involves such issues as : ( 1 ) propagation , instability formation , and sustenance of collimation in as a function of internal vs. external density and strength of magnetic fields , ( 2 ) bow shocks , cocoon physics , particle acceleration , ( 3 ) effect of cooling on morphology , and ( 4 ) interaction with ambient media
spatially resolving the launch region and measuring the magnetic field strength and geometry therein would be the gold standard for directly evaluating the role of magnetic fields in producing asymmetric p / pne .
measurements of fields in the propagation region provide primarily indirect evidence , though the detection of relatively strong fields there is particularly significant . @xcite . whether binaries supply needed rotational energy to amplify jet - mediating fields in accretion disks is a fundamental question .
the basic wind kinematics are roughly consistent with accretion onto a wd , suggesting the importance of accretion onto the primary .
coupling accretion disk physics to large scale magnetic field production , to jet launch , and jet propagation in a unified theory is work in progress that spans several subfields of theoretical astrophysics , let alone the specific application to p / pne .
however some predictions / trends can be studied : ( 1 ) it should be possible to evaluate the kinematic constraints / predictions of sec 3 in more detail and compare the distribution of inferred fast outflow speeds to what would be expected from known binary statistics of low mass main sequence stars .
this would serve to help determine the commonality of accretion onto the companion vs. the primary .
( 2 ) ce models would predict mostly oxygen rich rather than carbon rich post agb systems because typically , the companion ejects the envelope on time scales of order years , right from the beginning of the agb phase when the envelope expands , so the agb star would not have had a chance to reach the carbon dredge up .
( 3 ) crystalline dust in post - agb systems can be produced if a binary induced spiral shock anneals silicates @xcite . is this universal ? ( 4 ) ce evolution would predict equatorial outflows from companion inspiral that precedes any accretion driven poloidal jet . a delay is observed @xcite but more work is needed to predict the delay time scale .
( 5 ) are the geometry and composition of dust tori around post - agb objects consistent with the influence of a binary ? ( 6 ) are fast outflows contaminated by material that could represent accretion disk residue of shredded low mass companions ? ( 7 ) are time scales of observed outflow precession consistent with the gravitational influence of a binary on a disk ?
( 8) can double peaked line profiles be detected to identify accretion disks within the launch region ?
( 9 ) can shrouded novae outbursts from a @xmath91 companion feeding the primary be detected in x - rays ? ( 10 ) improved statistics on the fraction of bipolar ppne , the fraction of suitable precursor binaries for cee , and the fraction of stars which evolve to be ppne will constrain theories and assess whether all pne incur asymmetry . | observations suggest that many , if not all , post agb systems evolve through an aspherical outflow phase .
such outflows require a sufficient engine rotational energy which binaries can provide . via common envelope evolution
, binaries can directly eject equatorial outflows or produce poloidal outflows from magnetized accretion disks around the primary or secondary .
we discuss how accretion driven magnetohydrodynamic outflow models all make similar predictions for the outflow power and speed and we distinguish between the launch vs. propagation regimes of such outflows .
we suggest that the high velocity bipolar outflows observed in planetary nebulae ( pne ) and the lower velocity but higher power bipolar outflows observed in pre - pne ( ppne ) are kinematically consistent with time dependent accretion onto a white dwarf ( wd ) within a depleting envelope .
since the wd primary core is always present in all post - agb systems , accretion onto this core is potentially common .
previous work has focused on core accretion from sub - stellar companions , but low mass stellar companions may be more important , and further work is needed . |
@xmath3-deuteron compton scattering probes the structure of the deuteron and provides necessary information that may allow the extraction of neutron properties , such as neutron electric and magnetic polarizabilities .
the differential cross section of unpolarized @xmath3-deuteron compton scattering has been measured at incident photon energies of 49 mev and 69 mev with 10% uncertainty @xcite .
theoretical calculations based on potential models and taking the nucleon polarizabilities as inputs find reasonably good agreement with data @xcite but are not sufficient to give tight constraints on the nucleon polarizabilities . in comparison to potential model calculations ,
a model independent , parameter free and analytic computation of unpolarized @xmath3-deuteron compton scattering based on a recently developed nucleon - nucleon effective field theory @xcite was presented in @xcite .
contributions up to next - to - leading order ( nlo ) in the effective field theory expansion including diagrams contributing to the nucleon polarizabilities give excellent agreement with the data at 49 mev and reasonable agreement with the data at 69 mev . at this order ( nlo )
, the agreement with the data is comparable to the agreement between the data and potential model calculations @xcite , while inclusion of the next - next - to - leading order ( nnlo ) contributions should reduce the theoretical uncertainty from 10% to a few percent level .
in addition to @xmath3-deuteron compton scattering , the effective field theory technique also successfully describes the @xmath4 scattering phase shifts up to center - of - mass momenta of @xmath5 per nucleon @xcite in all partial waves .
the electromagnetic moments , form factors @xcite and polarizability @xcite of the deuteron as well as parity violation in the two - nucleon sector @xcite also have been explored with this new effective field theory .
the results all agree with data ( where available ) within the uncertainty from neglecting higher order effects . in the case of polarized @xmath3-deuteron compton scattering
, no experiments have been performed so far .
planned experiments at tunl to examine the gerasimov - drell - hearn ( gdh ) sum rule using a circularly polarized photon beam will be the first attempt to study vector polarized @xmath3-deuteron compton scattering .
experiments using tensor polarized targets are also technically feasible .
existing technologies like spin - exchange optical pumping @xcite , which will be applied in the blast polarized @xmath6h and @xmath7h laser driven sources , and free electron lasers can provide high quality polarized deuteron targets and photon beams .
the question is , what kind of physics can be measured in such an experiment ?
is it interesting enough to motivate the experiments ? theoretically , four vector form factors ( corresponding to the @xmath8 interaction of the deuteron field ) and six tensor form factors ( corresponding to @xmath9 ) are identified in @xcite with the lower multipole contributions calculated using dispersion relations and potential models . in this paper , we perform an analytic calculation of the differential cross section of the tensor polarized @xmath3-deuteron compton scattering cross section to first non - vanishing order in the effective field theory expansion . at this order ,
the contributions to the cross section from the @xmath9 amplitude are the interference terms between the leading electric coupling diagrams and the subleading single potential pion exchange diagrams or the subleading magnetic moment coupling diagrams .
thus the pion effects are of leading order
. we can further isolate the pion contribution by setting the photon scattering angle to be 90@xmath10 at this angle , only the pion term contributes at this order to the tensor polarized differential cross section .
this provides a clean way to study the photon pion dynamics in the two nucleon sector .
the experimental precision required to measure these effects will also be discussed in the following sections .
the process we will focus on is the low energy ( below pion production threshold ) tensor polarized compton scattering @xmath11 where the incident photon of four momentum @xmath12 in the deuteron rest frame ( lab frame ) scatters off the polarized deuteron target to an outgoing photon of four momentum @xmath13 .
the polarization of photons and final state deuteron are not detected . for convenience
, we will work in the lab frame and choose the @xmath14 direction to be the ( 0,0,1 ) direction .
the scattering amplitude can be written in terms of scalar , vector and tensor form factors @xmath15 and @xmath16 corresponding to the @xmath17 interactions of the deuteron field , @xmath18 where @xmath19 and @xmath20 are the polarization vectors of the initial and final state deuterons . using the power counting described in @xcite , the scalar form factor @xmath21 contributes to the amplitude starting at leading order ( lo ) in the effective field theory expansion , while the tensor form factor @xmath16 and the vector form factor @xmath22 contribute starting at next - to - leading order ( nlo ) and next - next - to - leading order ( nnlo ) respectively . squaring the amplitude to form the cross section ,
the lo contribution to the cross section only comes from the @xmath23 term which is independent of the deuteron target polarization , while the target polarization dependent @xmath21 and @xmath16 interference term contributes at nlo .
note that the vector form factor @xmath22 does not contribute to the cross section through the @xmath21 and @xmath22 interference term since the final state deuteron polarization is not detected .
it is useful to define the tensor polarized differential cross section @xmath24 as a tensor combination of polarized differential cross sections to eliminate the lo polarization independent effect and make the @xmath21 and @xmath16 contribution become leading .
@xmath24 is defined by @xmath25 \
\quad , \ ] ] where @xmath26 indicates the polarization of the deuteron target .
we have @xmath27 = 4.0 in 0.8 in 0.1 in .2 in to predict @xmath24 , we need expressions for @xmath21 and @xmath16 .
the calculation of the scalar form factor @xmath21 was carried out up to nlo in @xcite .
two distinct structures contributing to @xmath21 can be parameterized by electric and magnetic form factors @xmath28 and @xmath29 as @xmath30 where @xmath31 and @xmath32 are the polarization vectors of the initial and final state photons , @xmath33 and @xmath34 are unit vectors in the direction of @xmath14 and @xmath35 . at lo ( denoted by a superscript on the form factors )
, @xmath28 receives contributions from the electric coupling of the @xmath36 operator ( minimal coupling ) , as shown in fig .
[ fig : gdeleclead ] , to be @xmath37 \nonumber \\ & + & \left[\omega \rightarrow -\omega \right]\quad , \label{f0}\end{aligned}\ ] ] where cos @xmath38 is the cosine of the angle between the incident and outgoing photons and @xmath39 is the deuteron binding momentum . in eq .
( [ f0 ] ) terms suppressed by additional factors of @xmath40 ( i.e. , recoil effects ) are neglected since they only contribute to nnlo differential cross sections . at lo ,
the magnetic contribution vanishes , @xmath41 we do not show the nlo results for @xmath28 and @xmath29 here since we only calculate @xmath24 to its first non - vanishing order .
the tensor form factor @xmath16 has more distinct structures .
however , up to nlo , after neglecting pion contributions suppressed by factors of @xmath42 , @xmath16 can be parameterized by electric and magnetic form factors @xmath43 and @xmath44 as @xmath45 @xmath43 and @xmath44 are related to the form factors @xmath46 and @xmath47 defined in @xcite by a normalization factor . at lo , @xmath43 and @xmath44 vanish , @xmath48 = 4.0 in .2 in at nlo @xmath43 receives contributions from the electric coupling of the one potential pion exchange diagrams shown in fig .
[ fig : polsubpi ] and gives a renormalization scale @xmath49 independent contribution @xmath50 \log \left [ { m_\pi + 2\gamma \over \mu } \right ] \right.\nonumber\\ & & \left . \
m_\pi ^2(m_\pi ^2 + 4\gamma ^2 - 4m_n\omega ) \log \left [ { \displaystyle { m_\pi + 2\sqrt{\gamma ^2-m_n\omega -i\epsilon } \over \mu } } \right ] \right.\nonumber\\ & & \left . \
-\ 8\left [ m_\pi ^2(2m_\pi ^2 - 4\gamma ^2+m_n\omega ) -3m_n^2\omega ^2\right ] \log \left [ { \displaystyle { m_\pi + \gamma + \sqrt{\gamma ^2-m_n\omega -i\epsilon } \over \mu } } \right ] \right.\nonumber\\ & & \left . \ + \ { 32m_n\omega \left [ 2m_\pi \gamma ^2-m_n\omega ( m_\pi -\gamma ) \right ] \over m_\pi + \gamma + \sqrt{\gamma ^2-m_n\omega -i\epsilon } } \right.\nonumber\\ & & \left . \ + \ { \displaystyle { 4m_n^2\omega ^2 ( 5m_\pi ^2 - 12\gamma ^2 ) \over ( m_\pi + 2\gamma ) ^2\ } } \right.\nonumber\\ & & \left . \
-\ 2\left [ m_\pi ^3 + 4\left ( 2m_\pi ^2-m_n\omega \right ) ( m_\pi -\gamma ) \right ] \left ( \gamma -\sqrt{\gamma ^2-m_n\omega -i\epsilon } \right ) \right\}\nonumber\\ & & \left . \ + \ \left\{\omega \rightarrow -\omega \right\}\right . \ \ \ , \label{fpi}\end{aligned}\ ] ] where @xmath51 mev is the pion decay constant . in the computation of the pion diagrams , the pion propagators with photon momentum dependence
have the form @xmath52 where @xmath53 is the loop momentum .
@xmath54 is of higher order in the power counting compared with the other scales in the propagators and can be neglected .
we keep the first term in the @xmath14 expansion of the pion propagators to get the result of eq .
( [ fpi ] ) .
the error of this approximation is of order @xmath55 of eq .
( [ fpi ] ) and of even higher order in powers of @xmath55 at @xmath56 and @xmath57 . because the error is formally of nlo in power counting , strictly speaking we have not presented the complete calculation at nlo
. however , numerically the terms omitted are nnlo . = 1.2 in 0.1 in .2 in the magnetic moment
coupling diagrams shown in fig .
[ fig : maglead ] contribute to @xmath44 at nlo as @xmath58 where @xmath59 and @xmath60 are isoscalar and isovector nucleon magnetic moments in nuclear magnetons , with @xmath61 and @xmath62 .
the leading order nucleon nucleon scattering amplitudes contribute to the magnetic moment diagrams through the @xmath63 bubble chains and has the expression @xcite @xmath64 the renormalization scale @xmath49 dependence in the denominator is canceled by the @xmath49 dependence of @xmath65 and @xmath66 , as required .
values of @xmath65 and @xmath66 have been determined from nucleon nucleon scattering in the @xmath67 and @xmath68 channels @xcite to be @xmath69 and @xmath70 at @xmath71 having obtained the leading non - vanishing contributions for @xmath21 and @xmath16 , we now give the leading non - vanishing expressions for @xmath24 . at lo , @xmath24 vanishes , @xmath72 while at nlo , @xmath73(1+\cos ^2\theta ) + 2re[f_0^{lo}g_2^{nlo^{*}}]\cos \theta \right ] \quad .\ ] ] = 3.7 in .2 in there are two terms in @xmath74 the first term comes from the interference between the lo electric coupling diagrams and the nlo one potential pion exchange diagrams .
the second term comes from the interference between the lo electric coupling diagrams and the nlo magnetic moment coupling diagrams . in fig .
[ fig : dsig24969 ] , the one pion interference term contributing to @xmath75 are plotted as the dashed curves at incident photon energies of 49 and 69 mev .
the angular distributions are dominated by the @xmath76 factor that comes from the sum of the inner product squares of the incident and outgoing electric fields over the two different modes ( electric fields parallel or perpendicular to the scattering plane ) .
the small asymmetry comes from the finite size of the deuteron .
the magnetic moment interference terms are plotted as the dotted curves . the @xmath77 factor comes from the sum of the inner product of the incoming and outgoing electric fields times the inner product of the incoming and outgoing magnetic fields over the two different modes .
the sum of the two terms that form the nlo contributions to @xmath78 are plotted as solid lines .
the result is dominated by the pion interference terms .
higher order corrections could be as large as 30% of the contributions we show in this figure .
= 3.7 in .2 in = 3.7 in .2 in = 4.0 in .2 in to indicate the relative size of @xmath79 we write the @xmath80 and @xmath81 differential cross sections in terms of the unpolarized and tensor polarized differential cross sections , @xmath82 and @xmath24 , as @xmath83 the unpolarized differential cross section , @xmath84 \cos \theta \right ] \quad , \ ] ] up to nlo can be found in @xcite .
note that we have used @xmath85 @xmath86 = @xmath85 @xmath87 in eq .
( [ poldsig ] ) as photons are not circularly polarized .
this is a consequence of parity conservation . in fig .
[ fig : polrd4969 ] , we plot the @xmath80 and @xmath81 differential cross sections as dotted and dashed curves respectively at photon energies of 49 and 69 mev .
the difference between each set of dotted and dashed curves gives @xmath88 while the weighted sum gives @xmath89 at both energies , @xmath78 at backward angles can be extracted if the error bars in @xmath80 and @xmath81 differential cross sections are @xmath90 , as that of the unpolarized experiment @xcite performed four years ago . with error bars of 4% , the forward angle part of @xmath91 would also be measurable .
in these estimates , we have assumed that the higher order corrections to @xmath92 are 30% .
we can isolate the one pion interference term from @xmath93 by going to @xmath94 . as can easily be seen from eq .
( [ sig2 ] ) , at @xmath94 the magnetic interference term vanishes .
this provides a clean way to study the photon pion dynamics in the two nucleon sector . in fig .
[ fig:90deg4969 ] , the nlo differential cross sections at @xmath94 for deuteron polarizations @xmath80 and @xmath81 are plotted as the dotted and dashed curves while @xmath78 is plotted as the solid curve . again , with measurements of the @xmath80 and @xmath95 polarization made to within 7% , the pion interference term effects would be observable for photon energies between 40 and 80 mev with the higher order effect estimated to be 30% .
note that the effective field theory expansion power counting is expected to break down at photon energy of about 100 mev ( recent works by mehen and stewart @xcite suggest that in fact the breakdown scale may be much higher ) . but the neglected nlo terms which are suppressed by factors of @xmath42 could contribute a @xmath96 effect at 80 mev .
thus the 80 mev photon energy sets an upper bound for the calculations performed in this paper .
it is also interesting to compare our results with those from potential model calculations @xcite . in fig .
[ fig : f2g2 ] , we show the fist non - vanishong order ( nlo ) results of @xmath43 and @xmath44 multiplied by a constant to compare with the form factors @xmath46 and @xmath47 calculated and plotted in @xcite .
the agreement on @xmath44 is @xmath97 while the agreement on @xmath43 is @xmath98-@xmath99 .
we have presented analytic expressions for the differential cross section of tensor polarized @xmath3-deuteron compton scattering in an effective field theory expansion .
the first non - vanishing contributions are the interference terms between the leading electric coupling diagrams and the subleading single potential pion exchange diagrams , and between the leading electric coupling diagrams and the subleading magnetic moment coupling diagrams . at 90@xmath1 photon scattering angle
, only the pion term contributes at this order to the tensor polarized differential cross section .
this provides a clean way to study the photon pion dynamics in the two nucleon sector . for photon energy between 40 and 80 mev
, the one pion interference term contributions would be measurable at 90@xmath1 provided the uncertainty in the @xmath3-deuteron compton scattering experiments are @xmath2 . at backward angles
, the magnetic interference term adds to the pion term to give an effect which could be seen if the uncertainty in the measurements are @xmath100 .
the author would like to thank martin savage , roxanne springer , harald griesshammer , jerry miller , jon karakowski , and daniel phillips for helpful discussions and hartmuth arenhovel for correspondence .
this work is supported in part by the u.s .
dept . of energy under grants
de - fg03 - 97er41014 . | the differential cross section for @xmath0-deuteron compton scattering from a tensor polarized deuteron is computed in an effective field theory .
the first non - vanishing contributions to this differential cross section are the interference terms between the leading electric coupling diagrams and the subleading single potential pion exchange diagrams or the subleading magnetic moment coupling diagrams . at 90@xmath1 photon scattering angle
, only the pion term contributes at this order to the tensor polarized differential cross section .
this provides a clean way to study the photon pion dynamics in the two nucleon sector .
the effect is measurable for photon energies between 40 and 80 mev provided the uncertainty in the measured cross sections are @xmath2 .
# 1#1 # 1#1 # 1| # 1| # 1#2 # 11.5ex-16.5mu # 1 # 1#2#3#4#1 * # 2 * , # 3 ( # 4 ) |
recently , the accuracy of buried - objects detection system is increasing important which decided personal safety in mine affected areas .
few technology is employed in the real world to detect land mines because of high false alarm rate [ 1 ] . since mine detection requires low false alarm rate , improving the precision of detection system
should be studied .
many research works have been reported in the control of false alarm rate .
angular correlation function [ 2 ] and a statistical approach method [ 3 ] are employed to detect buried objects .
traditional method of detecting buried objects is to analysis the reflection echo amplitude , which is a analog signal detecting method and difficult in identifying the existence of several reflection echo in the lossy medium .
a novel digital method detecting predesignated reflection echo frequency is presented to detect buried objects .
with the development of finite time difference domain ( fdtd ) theory , it has been used to study time domain near - field measurement .
fdtd as a numerical method to solve maxwell s equations was introduced by yee in 1966 [ 4 ] , which divided both space and time into discrete grids .
the electromagnetic parameters can be figured out in combination with the boundary conditions .
the curl equations that are used in the fdtd algorithm are @xmath0 where @xmath1 is permeability , @xmath2 is permittivity and @xmath3 is electric conductivity . for a two dimensions free space time domain near - field measurements system , we assume that the dielectric media is nonmagnetic , _
i.e. _ @xmath4 , the @xmath5-field and @xmath6-field can be written as [ 5,6 ] @xmath7\ ] ] @xmath8\ ] ] @xmath9 and @xmath10\varepsilon_0dx}[h_y(i+1,j , t)-h_y(i , j , t ) ] \nonumber \\ & & -\frac{dt}{[\varepsilon_\infty+\chi_0(i , j)]\varepsilon_0dy}[h_x(i , j+1,t)-h_x(i , j , t)]\end{aligned}\ ] ] where @xmath11 is dielectric s static permittivity , @xmath12 is dielectric s optical permittivity , @xmath13 is dielectric s permeability , @xmath14 is dielectric s relaxation time , and @xmath15 ^ 2}\ ] ] is the susceptibility function .
a periodic microwave pulse has particular characters such as high peak value and low average value in near field , so it is difficult to detect the buried objects using amplitude detecting in the lossy medium .
therefor , to measure high peak value rapidly is the key point for mine detection system .
we use microwave frequency detector and high - speed sampling technology in receiver to realize the high efficiency detection .
the component of detection system is shown in figure 1 .
microwave pulse was produced by microwave signal source and launched by transmitting antenna .
the antenna can receive reflected pulse envelop at the same time . through microwave detector
, we received time domain waveform of reflect pulse envelop .
this effect is demonstrated in figure 2 .
@xmath16@xmath17 is the microwave pulse width and @xmath18 is the period of microwave pulse .
the high - speed sampling is key technology for pulse information during digital signal processing .
high - speed @xmath19/@xmath20 conversion circuit will records high level when the reflection echo of pulse is detected and records low level during other time .
finally , completed digital form time domain waveforms can be shown in the computer .
an iron ball ( @xmath21 ) with diameter is @xmath22 buried in a @xmath23@xmath23@xmath24 cubical vessel filled with dry sand ( @xmath25 ) . as is shown in the figure 3 .
there are @xmath22 between air - dielectric interface with spherical center , @xmath26 between @xmath27 with air - dielectric interface , @xmath26 between @xmath27 with @xmath28 and @xmath26 between @xmath28 with @xmath29 .
it is well known that fdtd utilizes the yee cell for calculation at nodes of the finite - difference lattice .
we set 420 grids in the @xmath30-direction and 100 grids in the @xmath31-direction with a grid spacing @xmath32 in the yee cell .
we have chosen a microwave signal source with microwave pulse width @xmath16@xmath17@xmath33@xmath34@xmath35 and pulse period @xmath36 as microwave pulse producer .
a @xmath30-band microtrip antenna is employed to both transmit and receive microwave pulse .
there is a microwave frequency detector to detect the @xmath37 microwave at the receiving end .
then , we can transform analog signals into digital signals through high - speed sampling and @xmath19/@xmath20 conversion . finally , the completed time domain near - field information of refection echo is shown in the computer monitor .
we put the microtrip antenna at @xmath27 , @xmath28 and @xmath29 to detect buried object respectively .
the analog signal simulation result is described in figure 4 and digital signal simulation result in figure 5 .
the peaks @xmath38 and @xmath39 are the detecting results at @xmath27 , @xmath40 and @xmath41 are the detecting results at @xmath28 and @xmath42 and @xmath43 are the detecting results at @xmath29 .
the paper introduces a frequency detection method using microwave frequency detector and high - speed sampling technology to detect the buried - objects in time domain near - field . from the simulation results we can see that the method can get more clear and accurate reflection echo information from the buried - objects than traditional amplitude detecting method . at the same time
, the low false alarm rate detecting method would be a new idea to detect remote objects .
guifu zhang , leung tsang , and kyung pak , angular correlation function and scattering coefficient of electromagnetic waves scattered by a buried object under a two - dimensional rough surface , journal of the optical society of american a * 15 * ( 1998 ) 2995 - 3002 r. luebbers , f. p. hunsberger , k. s. kunz , r. b. standler , and m. schneider , a frequency - dependent finite - differency time - domain formulation for dispersive materials , ieee trans electromag compat * 32 * ( 1990 ) 222 - 227 | the paper analyses the ground antenna echo using microwave frequency detector and high speed sampling technology and a new method detecting buried objects in time domain near - field is presented .
the method detecting particular reflection echo frequency of microwave pulse via digital signal processing is to reduce the false alarm rate .
simulation results show that this method has advantages of easy identification and high precision . *
key words * : near - field measurement ; mine detection ; fdtd ; microwave frequency detector ; high - speed sampling |
the milky way contains a supermassive black hole at its very center , sagittarius a * , whose mass is @xmath0m@xmath1 ( e.g. * ? ? ?
a dense stellar cluster surrounds this black hole , with a central density at least comparable to that seen in the cores of the densest globular clusters .
at least one , and possibly two , discs containing young stars at distances between @xmath2 and @xmath3pc from the black hole are also seen @xcite .
the latter stellar population is thought to have an unusually flat imf @xcite .
it has long been known that the central 0.2pc or so of the galactic center is deficient in bright red giants @xcite .
since the center of the galaxy has a high number density of stars , it is natural to suggest that stellar collisions may explain the observed depletion .
a recent crop of papers have studied the stellar population in the central regions @xcite .
they report that the early - type stars ( bright main - sequence stars ) follow a cusp - like profile whilst the surface density of late - type ( red - giant ) stars that are brighter than a k magnitude of 15.5 is rather flat out to 0.4pc or 10arcsec .
this surprising result implies that the red - giant population is depleted out to about 0.4pc from the central supermassive black hole .
we will consider here whether stellar collisions could be responsible for this observed depletion .
we will consider two different cases : 1 ) where the stellar population in the galactic center is drawn from a miller - scalo imf , and 2 ) where the imf is much flatter . in each case
, we calculate collision probabilities between the various stellar species , and produce , via monte carlo techniques , a stellar population . from this
we measure the surface density of the early and late - type stars ( to compare directly with what is observed ) making some reasonable assumptions concerning the effects of collisions on the population .
the masses of stars contributing to the observed early and late - type populations are rather different , as shown in fig .
[ mdavies_figure1 ] , where we have produced a synthetic population with a flat imf .
we see that the early - type stars are virtually all between 12 and 27m@xmath1 whereas stars contributing to the observed late - type population have much lower masses , between 1 and 5m@xmath1 .
the galactic center is a crowded environment , with number densities of stars probably of order @xmath4starspc@xmath5 or more . in such dense environments ,
collisions between stars will be frequent , and thus affect the stellar population .
the cross section for two stars , having a relative velocity at infinity of @xmath6 , to pass within a distance @xmath7 is given by @xmath8 where @xmath9 is the relative velocity of the two stars at closest approach in a parabolic encounter ( _ i.e. _ @xmath10 , where @xmath11 and @xmath12 are the masses of the two stars ) .
the second term is due to the attractive gravitational force between the two stars , and is referred to as the gravitational focusing term . in the very center of the galaxy , where the supermassive black hole dominates , and stars may be assumed to move on keplerian orbits around the black hole at speeds exceeding 1000km / s , @xmath13 and we recover the result , @xmath14 . in this
regime collisions involving larger red giants will be relatively more frequent despite their short lifetimes compared to main - sequence stars .
this is not the case in globular clusters where @xmath15 and thus @xmath16 .
one may estimate the timescale for a given star to undergo an encounter with another star , @xmath17 .
the collision timescale will therefore be a function of both the number density of stars and the makeup of the stellar population , _
i.e. _ the distribution of stellar masses and types
. the effects of collisions will depend on the types of stars involved and on the relative speed of the two stars when they collide .
collisions involving two main - sequence stars occurring at relatively low speed ( less than the surface escape speed of the stars ) are likely to result in the merger of the two objects with relatively low amounts of mass loss .
collisions occurring at much higher speeds are likely to lead to significant mass loss and even to the destruction of the stars involved .
collisions between main - sequence stars and compact objects ( _ i.e. _ black holes , white dwarfs or neutron stars ) are likely to be destructive ( dale et al . , in preparation ) . in the case of a black hole impactor
, the main - sequence star is often torn apart by tidal forces , with some material being accreted by the black hole and rest being ejected .
this is illustrated in fig .
[ mdavies_figure2 ] where we show snapshots of a collision between a 1m@xmath18 main - sequence star and a 10m@xmath18 black hole . the close passage of the black hole results in the tidal disruption of the main - sequence star .
a small fraction of the material is accreted by the black hole and the rest is dispersed . for neutron star and white dwarf impactors ,
a larger fraction of the material from the main - sequence star may form an envelope around the compact object .
however such an object is likely to be relatively short lived , perhaps appearing as a bright red supergiant , before the envelope is ejected .
thus all collisions between main - sequence stars and compact objects will act to reduce the population of luminous stars .
even though they are very frequent , collisions between red giants and main - sequence stars in fact have very little effect , as the main - sequence star passes through the envelope and effects very little mass loss @xcite .
the same is also true for white dwarf or neutron star impactors .
more interesting are encounters involving black holes . in fig .
[ mdavies_figure3 ] we show snapshots of an encounter between a 1m@xmath18 red giant and and 10m@xmath18 black hole ( @xmath19 km / s , @xmath20 r@xmath18 ) .
as the black hole passes close to the red - giant core , it gives it a jerk and the core is ejected at high speed , retaining only a small fraction of the envelope ( in this case about 13% ) .
such excessive mass loss will prevent this red giant from becoming brighter : we have thus removed it from the pool of brighter red giants . thus we see that in order to deplete the stellar population of red giants ( as indicated by the observations ) , we must consider collisions of three types : 1 ) encounters between red giants and black holes , 2 ) encounters between two main - sequence stars , and 3 ) encounters between main - sequence stars and compact objects ( _ i.e. _ black holes , white dwarfs and neutron stars ) .
the relative frequencies of encounters will depend on the imf of the underlying stellar population . with a miller - scalo type imf , where most stars are of low mass
, the majority of stars formed will still be on the main sequence today , with relatively few stars having evolved off the main - sequence to ultimately form compact remnants .
thus collisions involving two main - sequence stars will be more frequent than collisions between main - sequence stars and compact objects . collisions between red giants and black holes may be relatively frequent , at least in the very center , owing to the larger size ( and thus collisional cross section ) of red giants .
we consider first that the stellar population in the galactic center has been drawn from a miller - scalo imf @xcite where we also assume the stars have a uniform spread in ages from 0 to 14gyr .
using a density distribution for the stars given by @xmath21 and a density distribution for 10m@xmath18 black holes of @xmath22 @xcite , we compute the collision probabilities between the various stellar types as a function of galactocentric radius . the results are shown in fig .
[ mdavies_figure4 ] .
we see that collisions between red giants and black holes are the most frequent in the very central regions whilst collisions between two main - sequence stars dominate further out . for stars following a miller - scalo imf , collisions between main - sequence stars and black holes ( or other compact remnants )
are relatively less frequent and are unlikely to have a significant effect on the stellar population apart from perhaps in the very central regions .
even though collisions between main - sequence stars are relatively frequent , in the case of a miller - scalo imf with a low - mass cutoff of 0.2m@xmath18 the number of main - sequence stars in the mass range 14m@xmath18 will not be reduced .
the reason is that although some stars in this mass range will be removed as they merge with other stars to produce more massive stars , they will be replaced by stars produced by the merger of lower - mass stars .
this process will be investigated in detail in dale et al .
( in preparation ) . it could be that a stellar population with a miller - scalo imf but with a larger low - mass cutoff may lead to a depletion of the lower - mass main - sequence stars which later evolve to produce the observed population of red giants .
collisions between red giants and stellar - mass black holes are able to deplete the red - giant population within the inner 0.08pc or so but not further out .
collisions can plausibly explain the depletion of red giants of middle brightness ( @xmath23 ) but not those in the brightest band ( @xmath24 ) which are seen to be depleted to about 0.2pc @xcite .
collisions are not able to explain the observed flattening of the red - giant population seen out to 0.4pc @xcite .
in this section we consider the case where the stellar population in the galactic center is drawn from a much flatter imf than a miller - scalo imf .
observations of the young disc of stars in the galactic center suggest that they may have been drawn from such an imf with a power - low slope as low as @xmath25 @xcite . a stellar population drawn from such
a flat imf would be quite different from one drawn from a miller - scalo imf
. a much larger fraction of stars would be massive , and explode as core - collapse supernovae producing either black holes or neutron stars .
in addition , assuming stars have an approximately uniform spread in ages , a very large fraction of all stars will have evolved to become compact remnants .
( _ i.e. _ black holes , white dwarfs , or neutron stars ) .
in other words , the majority of stellar objects will in fact be compact objects .
the most common flavor of collision will then be collisions involving compact objects . in order to produce a stellar population , we assume a model where a stellar disc , similar to the one seen today , is produced about every @xmath26 years up to 14gyr ago , with the most recent disc being produced only 6 million years ago to be consistent with observations .
the disc mass is set to match the observations of the early - type stars which come entirely from the most - recent disc , whilst the total number of discs produced is set to match the observed late - type population 10arcsec from the supermassive black hole .
one should note that the formation mechanism for such discs of stars is unclear , one suggestion being that gas clouds interact tidally with the supermassive black hole @xcite .
we consider here for purposes of illustration the case of @xmath27 .
for this imf ( which is flat in log mass ) , the mass is each disc is @xmath28m@xmath18 , with the stars having a range of masses between 1.0 and 32.5m@xmath18 .
it is entirely possible that the star formation rate was higher in the past . as a simple first model
, we assume that in addition to the discs produced at a steady rate , there was an excess of discs produced sufficiently early in the history of the galaxy that essentially all the stars from these earlier discs have evolved into compact remnants
. the exact excess of compact objects will depend on the history of the disc formation rate , but here we assume that two thirds of all star formation occurred at this earlier epoch .
this will increase the number of compact objects present in the galactic center population .
as collisions involving compact remnants destroy main - sequence stars , increasing the compact object population will enhance the red - giant depletion .
the collision probability for collisions between stars of various types for the stellar population produced as described above are shown in fig .
[ mdavies_figure5 ] .
we see here that collisions between main - sequence stars and compact objects are most likely ( at least further out from the center ) .
main - sequence stars are likely to undergo such a collision , and thus be destroyed , out to a distance of about 0.2pc .
as discussed earlier , collisions between compact objects and main - sequence stars are likely to be destructive . therefore with a flat imf , and a suitably dense stellar population , the low - mass main - sequence stars which would ordinarily have evolved to become the observed late - type population are depleted within a distance of about 0.2pc of the supermassive black hole .
therefore a stellar population produced from a flat imf may be able to explain the flat surface density profile which has been observed for the late - type stars ( red giants ) as stars in the mass range 1 - 5m@xmath18 within 0.2pc have simply been destroyed before they could evolve into red giants .
equipped with the collision probabilities as a function of radius , we are now able to synthesize a stellar population for both miller - scalo and flat imfs allowing for the collisional depletion of the red - giant population . in fig . [ mdavies_figure6 ] we show the calculated surface density profiles for early - type stars and late - type stars having k magnitudes brighter than 15.5 for both imfs . in both cases , the stellar population has been normalised to the surface brightness of late - type stars 10arcsec from the galactic center , seen to be about 3 sources per square arcsec ( this normalization was also applied to the collision probability calculations and figures shown earlier ) . in both plots
we do not include the effect of stellar collisions on the population of early - type stars as the collisional depletion of these massive stars will be very small .
however it should be noted that the existence of the so - called s stars in the very center of the galaxy ( distances less than 1 arcsecond from the central black hole ) may place limits on the number density of compact objects .
the plots shown in fig .
[ mdavies_figure6 ] should be compared to fig .
11 of @xcite .
one can see from the plots that the early - type population only matches the observations in the case of the flat imf .
a miller - scalo population is clearly not able to consistently match both the observed early and late populations .
it is also clear from these plots that in the case of the miller - scalo imf , the effect of stellar collisions on the red - giant population is negligible .
however for the flat imf , appreciable red - giant depletion occurs out to 10arcseconds .
thus a flat imf would seem to have the potential to explain the observed ( flat ) surface density profile of the red giants .
one problem with the model used here is that the total stellar mass out to 0.4pc is about @xmath29m@xmath18 .
this is much higher than suggested by observations @xcite .
however , in our calculations using a flat imf here we have made the unrealistic assumption that stars remain at the radius at which they were formed .
in other words , we have not allowed for the effects of mass segregation . in reality
, the large population of heavier stellar - mass black holes will sink in the potential , forming a central sub - cluster .
this sub - cluster may then collapse to form a supermassive black hole in the manner envisaged by @xcite if a supermassive black hole is not there already , or the stellar - mass black holes may simply be fed into an existing supermassive black hole . in either case
, a large fraction of the stellar - mass black holes may end up inside the supermassive black hole .
it is interesting in this context to note that the total mass contained in stellar - mass black holes for our flat imf ( about @xmath30m@xmath18 ) is close to the observed value for the supermassive black - hole mass .
a larger population of black holes produced by a flat imf would also lead to a much higher rate of emri type events , where stellar - mass black holes spiral in to the central , supermassive black hole , emitting gravitational radiation in the process .
emris could be an important source for lisa .
our calculations here have been a simplification .
we have made the unrealistic assumption that stars remain at the radius they were formed .
in addition we have not allowed for the growth of the supermassive black hole .
early on , before the supermassive black hole has grown , the disc - mode of star formation may be different or not occur at all owing to a lack of a supermassive black hole . a natural next step in these calculations
will be inclusion of the dynamical evolution of the stellar cluster and for the growth of the central black hole ( nzoke et al .
, in preparation ) .
rpc acknowledges support from the wenner gren foundation .
jed is supported by a marie curie fellowship as part of the european commission fp6 research training network ` constellation ' under contract mrtn ct2006035890 and the institutional research plan av0z10030501 of the academy of sciences of the czech republic and project lc06014centre for theoretical astrophysics of the ministry of education , youth and sports of the czech republic . | we consider whether stellar collisions can explain the observed depletion of red giants in the galactic center .
we model the stellar population with two different imfs : 1 ) the miller - scalo and 2 ) a much flatter imf . in the former case ,
low - mass main - sequence stars dominate the population , and collisions are unable to remove red giants out to 0.4pc although brighter red giants much closer in may be depleted via collisions with stellar - mass black holes . for a much flatter imf ,
the stellar population is dominated by compact remnants ( _ i.e. _ black holes , white dwarfs and neutron stars ) .
the most common collisions are then those between main - sequence stars and compact remnants .
such encounters are likely to destroy the main - sequence stars and thus prevent their evolution into red giants . in this way
, the red - giant population could be depleted out to 0.4pc matching observations .
if this is the case , it implies the galactic center contains a much larger population of stellar - mass black holes than would be expected from a regular imf .
this may in turn have implications for the formation and growth of the central supermassive black hole . |
the existence of the scalar - isoscalar @xmath3 meson as a broad @xmath4 resonance has long been controversial . recently , an increasing number of theoretical and experimental analyzes point toward the existence of this important meson @xcite .
most of these analyzes find a @xmath3-pole position near 500-i250 mev @xcite .
a direct experimental evidence seems to emerge from the d@xmath5 decay channel observed by the fermilab e791 collaboration , where @xmath3 meson is seen as a clear dominant peak with m@xmath6=478 mev and @xmath7=324 mev @xcite .
since @xmath3 meson is a relevant hadronic degree of freedom , it must be incorporated into the analysis of hadronic processes .
although at sufficiently high energies and low momentum transfers electromagnetic production of vector mesons on nucleon targets has been explained by pomeron exchange models , at low energies near threshold scalar and pseudoscalar meson exchange mechanisms become important @xcite . in particular ,
the presently available data on the photoproduction of @xmath8 meson on proton targets near threshold can be described by a simple one - meson exchange model @xcite . in this picture ,
@xmath8 meson photoproduction cross section on protons is given mainly by @xmath3 meson exchange .
an important physical input for these studies is the coupling constant g@xmath0 .
this coupling constant was estimated by calculating the @xmath9-vertex assuming vector meson dominance of the electromagnetic current and then performing a fit to the experimental @xmath2-photoproduction data .
when this result was derived using an effective lagrangian for the @xmath9-vertex the value g@xmath0=2.71 is obtained for this coupling constant @xcite . since it plays an important role in hadronic processes involving @xmath3 meson , it is of interest to study this coupling constant from a different perspective other than vector meson dominance . in this work ,
we estimate the coupling constant g@xmath0 by employing qcd sum rules which provide an efficient method to study many hadronic observables , such as decay constants and form factors , in terms of nonperturbative contributions proportional to the quark and gluon condensates @xcite . in section 2
, we give a qcd sum rules analysis of scalar current , derive a sum rule for the overlap amplitude @xmath10 of the scalar current to the @xmath3 meson state and estimate this amplitude since it is not available experimentally . in section 3 , we derive a sum rule for the coupling constant g@xmath0 and by utilizing the experimental value of @xmath11 , the calculated value of @xmath10 and the known values of condensates we estimate this coupling constant .
let us here note that the nature of @xmath3 meson , whether it is a @xmath12 state or it is of another quark , gluon structure is still a matter of debate @xcite .
the qcd sum rules approach @xcite is a model independent method to study the properties of hadrons through correlation functions of appropriately chosen currents . in the su(2 ) flavour limit m@xmath13 , we study the scalar - isoscalar mesons by considering the scalar current correlation function @xmath14 where @xmath15 is the interpolating current for @xmath3 meson .
the two - loop expression for the scalar current correlation function @xmath16 in perturbative qcd was calculated @xcite , and it is given by the expression @xmath17\right\}~~.\ ] ] qcd - vacuum condensate contributions to the scalar current correlation function @xmath18 were obtained by the operator product method @xcite , and they were found in the @xmath19=0 limit as @xmath20 we note that the term @xmath21 is not considered in the order m@xmath22 terms , as this condensate s magnitude , which is obtained by making use of gell - mann - oakes - renner relation as @xmath23 , is independent of quark mass @xcite .
on the other hand , the correlation function @xmath18 satisfies the standard subtracted dispersion relation @xcite @xmath24 where the spectral density function is given as @xmath25 .
we parameterize the spectral density as a single sharp pole @xmath26 , where the overlap amplitude is defined as @xmath27 , plus a smooth continuum representing the higher states .
the continuum contribution to the spectral density is estimated in the form @xmath28 where s@xmath29 denote the continuum threshold and @xmath30 is given by the expression @xmath31 with @xmath32 obtained from eq .
( 2 ) and eq . ( 3 ) as @xmath33 . after performing the borel transformation
we obtain the following qcd sum rule for the overlap amplitude @xmath34 of the scalar current @xmath35(1+\frac{\alpha_{s}(m)}{\pi}\frac{17}{3}- \frac{\alpha_{s}(m)}{\pi}\int_0^{s_0/m^2 } w\ln we^{-w}dw\right\ } \nonumber
\\ + \frac{3}{2m^2}<m_q\overline{q}q>+\frac{1}{16\pi m^2}<\alpha_s g^2>- \frac{88\pi}{27 m^4}<\alpha_s(\overline{q}q)^2>~~.\end{aligned}\ ] ] for the numerical evaluation of the above sum rule , we use the values @xmath36 , @xmath37 , @xmath38 @xcite , and m@xmath39=0.5 gev . for the continuum threshold
we choose s@xmath29=1.1 , 1.2 , 1.3 gev@xmath40 and we study the m@xmath40 dependence between 0.5 gev@xmath40 and 1.4 gev@xmath40 .
the overlap amplitude as a function of m@xmath40 for different values of s@xmath41 is shown in fig . 1
from which by choosing the middle value @xmath42 in its interval of variation we obtain the overlap amplitude as @xmath43 where we include the uncertainty due to the variation of the continuum threshold and the borel parameter @xmath44 .
other sources of uncertainty are the errors attached to the estimated values of condensates as quoted above .
if we take these errors into account in our analysis as well , we then obtain the value @xmath45 for the overlap amplitude .
in order to derive the qcd sum rule for the coupling constant g@xmath0 , we consider the three point correlation function @xmath46 the interpolating current for @xmath47 meson @xcite is @xmath48 , and @xmath49 , @xmath50 are the electromagnetic and scalar currents , respectively .
the theoretical part of the sum rule is obtained by calculating the perturbative contribution and the power corrections from operators of different dimensions to the three point correlation function @xmath51 . for the perturbative contribution
we consider the lowest order bare - loop diagram .
furthermore , we consider the power corrections from the operators of different dimensions , @xmath52 , @xmath53 and @xmath54 .
we do not consider the gluon condensate contribution proportional to @xmath55 since it is estimated to be negligible for light quark systems .
we perform the calculations of the power corrections in the fixed point gauge @xcite .
we work in the limit m@xmath22=0 , and in this limit perturbative bare - loop diagram does not make any contribution . moreover , in this limit only operators of dimensions d=3 and d=5 make contributions that are proportional to @xmath52 and @xmath53 , respectively .
the relevant feynman diagrams for the power corrections are shown in fig .
2 . in order to obtain the phenomenological part , we consider a double dispersion relation for the vertex function @xmath51 , and we saturate this dispersion relation by the lowest lying meson states in the vector and scalar channels .
we obtain this way for the physical part @xmath56 where the contributions from the higher states and the continuum is denoted by dots . in this expression , @xmath57 has been determined in section 2 . the overlap amplitude @xmath58 of rho meson is defined as @xmath59 where @xmath60 is the polarization vector of the @xmath47 meson .
the matrix element of the electromagnetic current is given as @xmath61 where @xmath62 and @xmath63 is a form factor with k(0)=1 .
this expression defines the coupling constant g@xmath0 through the effective lagrangian @xmath64 describing the @xmath9-vertex @xcite . after performing double borel transform with respect to the variables @xmath65 and @xmath66 , and by considering the gauge - invariant structure ( @xmath67 ) we obtain the sum rule for the coupling constant g@xmath0 @xmath68 where we use the relation @xmath69 . for the numerical evaluation of the sum rule we use the values @xmath70 , @xmath71 @xcite , and @xmath72 , @xmath73 . for the overlap amplitude
@xmath34 we use the value @xmath45 that we have estimated in section 2 and for @xmath58 we use its experimental value @xmath74 as obtained from the measured leptonic width @xmath75 of @xmath47 meson @xcite . in order to analyze the dependence of g@xmath0 on borel parameters @xmath44 and @xmath76 , we study the independent variations of @xmath44 and @xmath76 in the interval @xmath77 as these limits determine the allowed interval for the vector channel @xcite .
the variation of the coupling constant g@xmath0 as a function of borel parameters @xmath44 for different values of @xmath76 is shown in fig .
3 , examination of which indicates that it is quite stable with these reasonable variations of @xmath44 and @xmath76 . we choose the middle value @xmath78 for the borel parameter in its interval of variation and obtain the coupling constant g@xmath0 as g@xmath79 where only the error arising from the numerical analysis of the sum rule is considered . the other sources contributing to the uncertainty in the coupling constant besides those due to variations of @xmath44 and @xmath76 are the uncertainties in the estimated values of the vacuum condensates .
if we take these uncertainties into account , we then obtain the coupling constant g@xmath0 as g@xmath80 .
our estimate of the coupling constant g@xmath0 is in agreement with its value deduced from the analysis using the vector meson dominance of the electromagnetic current for the @xmath9-vertex . on the other hand ,
an independent anaysis @xcite utilizing the experimental value of the decay rate @xmath81 and using the values for the @xmath3 meson parameters @xmath82=478 mev and @xmath83=324 mev that are obtained from the fermilab e791 experiment gives the value g@xmath84 for this coupling constant .
however , we like to note that in the present work we consider @xmath3 meson in the narrow resonance limit and do not take the finite width of @xmath3 meson into account .
n. a. trnqvist , invited talk at the conference : `` possible existence of the light @xmath3 resonance and its implications to hadron physics '' , kyoto , japan , 11 - 14 june 2000 , kek - proceedings 2000 - 4 , dec .
2000 , 224 - 231 and soryushiron kenkyu ( kyoto ) 102 , e224-e231 ( 2001 ) , ( hep - ph/0008135 ) . n. a. trnqvist , a. d. polasa , invited talk at the conference `` heavy quark at fixed target '' , rio de janeiro , brazil , october 9 - 12 , 2000 , to appear in frascati physics series , ( hep - ph/0011107 ) .
e. m. aitala et al .
( e791 collaboration ) , phys .
86 ( 2001 ) 770 .
a. i. titov , t. -s .
h. lee , h. toki and o. streltsova , phys . rev .
* c 60 * ( 1999 ) 035205 .
b. friman and m. soyeur , nucl .
phys . * a 600 * ( 1996 ) 477 .
m. a. shifman , a. i. vainstein , v. i. zakharov , nucl .
phys.*b 147 * ( 1979 ) 385 and 448 .
l. j. reinders , h. r. rubinstein , s. yazaki , phys .
rep . 127 ( 1985 ) 1 .
p. colangelo , a. khodjamirian , hep - ph/0010175 , to be published in the boris ioffe festschrift , world scientific , singapore , 20001 .
v. elias , a. h. fariborz , fang shi , t. g. steele , nucl .
phys . * a 633 * ( 1998 ) 279 .
k. g. chetyrkin , phys . lett .
* b 390 * ( 1997 ) 309 .
e. bagan , j. i. la torre , p. pascual , z. physics .
* c 32 * ( 1986 ) 43 .
b. l. ioffe , a. v. smilga , nucl . phys .
* b 232 * ( 1984 ) 109 .
a. v. smilga , sov .
* 35 * ( 1982 ) 271 . v. m. belyaev , b. l. ioffe , sov
* jetp 56 * ( 1982 ) 493 .
d. e. groom et al .
j. * c15 * ( 2000 ) 1 ; v. l. eletsky , b. l. ioffe , ya .
i. kogan , phys .
lett.*b 122 * ( 1983 ) 423 .
a. gokalp , o. yilmaz , phys . rev . *
d 62 * ( 2000 ) 093018 . * figure 1 * : : : the overlap amplitude @xmath34 as a function of borel parameter @xmath44 .
* figure 2 * : : : feynman diagrams for the @xmath9-vertex : a- bare loop diagram , b- d=3 operator corrections , c- d=5 operator corrections .
the dotted lines denote gluons .
* figure 3 * : : : the coupling constant @xmath85 as a function of the borel parameter @xmath44 for different values of @xmath76 . | we employ qcd sum rules to calculate the coupling constant g@xmath0 by studying the three point @xmath1-correlation function .
our results is consistent with the value of this coupling constant obtained using vector meson dominance of the electromagnetic current and the experimental @xmath2-photoproduction data . |
in the extensive literature devoted to the pricing and risk management of commodity derivatives , most investigations take as a starting point the specification of a price process for the commodity . the outcome of chance in the market in which the commodity is traded is usually modelled by a fixed probability space equipped with the filtration generated by a brownian motion of one or more dimensions , and it is typically assumed that the commodity price can be modelled as an ito process adapted to this filtration .
such an approach to the pricing of commodities and related derivatives is in line with the standard " modelling framework for asset pricing , within which much of modern financial engineering has been carried out .
a fundamental methodological issue arising in the standard framework is that the market filtration is fixed in an essentially _ ad hoc _ way , and that no indication is provided concerning the nature of the information it purports to convey , or why it is relevant to the price .
the information is in practice typically no more than that of the price movements themselves , so it can hardly be claimed in any useful way that the price movements are taking place `` in response '' to `` shocks '' associated with the arrival of information .
one knows , however , that in real markets , information concerning the possible future cash flows and other benefits or obligations linked to the physical possession of an asset can be crucial in the determination of trading decisions , even in situations where such information is of an imperfect nature .
the movement of the price of an asset should thus be regarded as a _ derived _
concept , induced by the flow of information to market participants .
this is the point of view put forward in the information - based asset pricing framework of brody , hughston & macrina ( 2007 , 2008a , b ) that forms the analytical basis of the present investigation .
the objective of this paper is in such a context to incorporate the role played by forward - looking information in commodity markets , and to derive a model for the prices that are used as underlyings in the valuation formulae for commodity derivatives . specifically , we make use of the concept of market information about future supply and demand as a basis for valuation of storable commodities . for a detailed treatment of theory of storage , convenience yield , and related concepts ,
the reader can be referred to geman ( 2005 ) and references cited therein . in our approach
, we shall assume that the possession of one standard unit of the commodity provides a `` convenience dividend '' equivalent to a cash flow @xmath0 .
we thus work directly with the actual flow of benefit arising from the storage or possession of the commodity , rather than the percentage convenience yield .
the point is that the percentage convenience yield is a secondary notion , since it depends on the price , which is what we are trying to determine . in what follows ,
we present a simple model for the convenience dividend process @xmath1 . additionally , we introduce a market information process @xmath2 that provides partial or speculative information about the future dividend flow .
the market filtration is then assumed to be generated jointly by these two processes . in that sense , we are explicitly constructing the market filtration in such a way that it contains information relevant to the commodity price . given the market filtration , the price of the commodity is taken to be the appropriately risk - adjusted discounted expected value of the totality of the future convenience dividends .
we model @xmath1 by an ornstein - uhlenbeck process .
we model @xmath2 by a process that consists of two terms : a `` signal '' term containing information about the future convenience dividend flow , and a `` noise '' term given by an independent brownian motion . by use of this information - based model we are able to derive closed - form expressions for both the price of the commodity , and the prices of associated derivatives . the paper is organised as follows . in [ sec:2 ]
we introduce our assumptions for the convenience dividend and for the market filtration . in [ sec:3 ] various useful facts relating the ornstein - uhlenbeck ( ou ) process
are recalled in particular , certain features of the ou bridge .
these are used in [ sec:4 ] to show in proposition 1 that the information process and the convenience dividend rate are jointly markovian , and in [ sec:5 ] to derive an expression for the commodity price .
the resulting model is sufficiently tractable to allow for monte carlo simulations of the price process to be carried out . in [ sec:6 ] we proceed to derive the stochastic differential equation satisfied by the price . in doing so , we are able to obtain an innovations representation for the associated filtering problem in closed form . in [ sec:7 ]
we work out pricing formulae for call options on the underlying spot price . in conclusion
, the model is applied in [ sec:8 ] to obtain the corresponding processes for futures prices , which are useful since futures contracts tend to be more commonly traded in commodities markets than the associated underlyings .
in the information - based approach of brody _ et al_. ( 2007 , 2008a , b ) , the starting point is the specification of : ( i ) a set of one or more random variables ( called `` market factors '' ) determining the cash flows associated with a given asset ; and ( ii ) a set of one or more random processes ( called `` information processes '' ) determining the flow of information to market participants concerning these market factors .
the setup , more specifically , is as follows .
we model the outcome of chance in a commodity market with the specification of a probability space @xmath3 .
the market is not assumed to be complete , but we do assume the existence of a preferred pricing measure @xmath4 .
then if @xmath1 represents the convenience dividend , the price of the commodity at time @xmath5 is given by @xmath6 , \label{eq:1}\end{aligned}\ ] ] where the expectation is taken with respect to @xmath4 . here the discount factor @xmath7 is given by @xmath8 in terms of the short rate @xmath9 .
for simplicity , we shall assume that the default - free interest rate system is deterministic .
the conditioning in ( [ eq:1 ] ) is taken with respect to the market filtration @xmath10 , which is generated jointly by : ( a ) the convenience dividend process @xmath11 ; and ( b ) a market information process @xmath12 of the form @xmath13 representing partial or noisy information about the future dividend flow .
the parameter @xmath14 determines the rate at which information about the future dividend stream is revealed to the market .
the @xmath4-brownian motion @xmath15 represents noise arising from rumour , baseless speculation , uninformed trading , and the like , and is assumed to be independent of the dividend process @xmath16 .
therefore , we have @xmath17.\end{aligned}\ ] ] our next modelling choice is to specify the form of the dividend process .
we consider in this paper the case in which @xmath16 is an ornstein - uhlenbeck ( ou ) process .
it is worth recalling , by way of contrast , that gibson & schwartz ( 1990 , 1991 ) assume that the percentage convenience yield should follow a mean - reverting process .
the idea in the present work is that a mean - reverting absolute convenience dividend reflects the notion that in the long term there is an equilibrium rate of benefit obtained by storing the commodity .
we thus assume that the dividend process satisfies a stochastic equation of the form @xmath18 where @xmath19 is a @xmath4-brownian motion that is independent of @xmath15 .
we allow for the possibility that the dividend rate may occasionally be negative .
the mean reversion level @xmath20 , the reversion rate @xmath21 , and the dividend volatility @xmath22 are assumed to be constant in the present discussion , although the results can be readily generalised to the time - dependent cases .
before we proceed to calculate the expectation ( [ eq:1 ] ) and hence work out the commodity value , it will be useful to remark on various properties of the ornstein - uhlenback process .
these properties , some of which will be well known , but others perhaps less so , will help us simplify subsequent calculations .
it is an elementary exercise to check that the solution to ( [ eq:4 ] ) is @xmath23 the gaussian process @xmath16 is characterised by its mean and auto - covariance , given by @xmath24 = { \mbox{$\rm e$}}^{-\kappa t } x_0 + \theta ( 1-{\mbox{$\rm e$}}^{-\kappa t } ) , \quad { \rm cov}[x_t , x_t ] = \frac{\psi^2}{2\kappa}\ , { \mbox{$\rm e$}}^{-\kappa t } ( { \mbox{$\rm e$}}^{\kappa t } - { \mbox{$\rm e$}}^{- \kappa t})\end{aligned}\ ] ] for @xmath25 .
the ornstein - uhlenback process has the property that if we `` reinitialise '' the process at time @xmath5 then its value at some later time @xmath26 can be expressed in the form @xmath27 since @xmath16 is gaussian , by use of the variance - covariance relations one can verify that the random variables @xmath28 and @xmath29 are independent .
this property corresponds to an orthogonal decomposition of the form @xmath30 for @xmath31 .
it should be evident that if the reversion rate is set to zero , then this relation reduces to the familiar independent - increments decomposition of a brownian motion .
interestingly , there is another orthogonal decomposition of an ou process , somewhat less obvious than ( [ eq:8 ] ) .
this is given by the identity @xmath32 the process @xmath33 defined for a fixed @xmath26 by @xmath34 appearing in ( [ eq:9 ] ) is an ornstein - uhlenbeck ( ou ) bridge .
the ou bridge interpolates between the fixed values @xmath35 and @xmath36 . by use of the covariance relations
, one can check that @xmath37 and @xmath38 are independent , an important property that will be exploited in the calculation below . in figure [ fig:1a ]
we show sample paths of the ou bridge , as well as for its mean and variance , which are given , respectively , by @xmath39 = \frac{\sinh(\kappa(t - t))}{\sinh(\kappa t ) } x_0 + \left [ 1 - \frac{\sinh(\kappa t)+\sinh(\kappa(t - t))}{\sinh(\kappa t ) } \right ] \theta.\end{aligned}\ ] ] and @xmath40 = \frac{\psi^2}{2\kappa } \left [ \frac{\cosh(\kappa t ) - \cosh(\kappa(t-2t))}{\sinh(\kappa t ) } \right ] .\end{aligned}\ ] ] , @xmath41 , @xmath42 , @xmath43 , @xmath44 .
the mean ( in red ) and variance ( in purple ) of the ou bridge are also plotted for comparison .
[ fig:1a],height=264 ]
having listed some of the elementary properties of the mean - reverting process that we propose to use as a model for the dividend rate , we proceed to work out the expectation in ( [ eq:1 ] ) to determine the commodity price .
the following result will facilitate the calculation .
the information process @xmath2 and the dividend rate @xmath16 are jointly markovian , and hence @xmath45 = { \mathbb e } \left[\left .
\int_t^\infty p_u x_u { \mbox{$\rm d$}}u \right| \xi_t , x_t \right ] .
\label{eq:13 } \end{aligned}\ ] ] let us define an alternative information process @xmath46 by setting @xmath47 it should be evident that
@xmath48={\sigma}[\{x_s\}_{0\leq s\leq t } , \{\eta_s\}_{0\leq s\leq t } ] .
\end{aligned}\ ] ] we observe that @xmath46 is markovian with respect to its own filtration . to see this
, it suffices to verify that @xmath49 for any collection of times @xmath50 such that @xmath51 .
now it is an elementary property of brownian motion that for any times @xmath52 satisfying @xmath53 the random variables @xmath54 and @xmath55 are independent . more generally , for @xmath56 , we find that @xmath55 and @xmath57 are independent .
bearing in mind that @xmath58 , we have @xmath59 but since @xmath60 and @xmath61 are independent of @xmath55 , @xmath62 , the markov property ( [ eq:15 ] ) follows .
let us now define @xmath63 note that @xmath60 is independent of @xmath64 , and furthermore that @xmath65 is independent of @xmath64 .
thus , we conclude that @xmath66 \nonumber \\ & = & { \mathbb e } \left[\left
. \int_t^\infty p_u x_u { \mbox{$\rm d$}}u \right| \{\eta_s\}_{0 \leq s \leq t } , \{x_{s}\}_{0 \leq s \leq t } \right ] \nonumber \\ & = & { \mathbb e } \left[\left .
\int_t^\infty p_u x_u { \mbox{$\rm d$}}u \right| \eta_t , { \mathcal g}_t , \{x_{s}\}_{0 \leq s \leq t } \right ] \nonumber \\ & = & { \mathbb e } \left[\left . \int_t^\infty p_u x_u { \mbox{$\rm d$}}u \right| \eta_t , \{x_{s}\}_{0 \leq s \leq t } \right ] \nonumber \\ & = & { \mathbb e } \left[\left .
\int_t^\infty p_u x_u { \mbox{$\rm d$}}u \right| \xi_t , \{x_{s}\}_{0 \leq s \leq t } \right].\end{aligned}\ ] ] on the other hand we clearly have that @xmath67={\sigma } [ \xi_t , x_t , \{b_{st}\}_{0\leq s\leq t } ] , \end{aligned}\ ] ] and furthermore that the ou bridge @xmath68 is independent of @xmath69 .
therefore @xmath70 is independent of @xmath71 and @xmath72 , from which the claimed result follows .
the joint markov property ( [ eq:13 ] ) allows one to reduce the problem of working out the commodity price ( [ eq:1 ] ) to that of calculating @xmath74 .
\label{eq:19 } \end{aligned}\ ] ] one observes in particular that from the orthogonal decomposition ( [ eq:8 ] ) we can isolate the dependence of the commodity price on the current level of the convenience dividend rate @xmath28 .
remarkably , this dependence turns out to be linear in our model .
that is , we have the following relation : @xmath75 substituting this relation in equation ( [ eq:19 ] ) we thus obtain an expression of the form @xmath76 + { \mathbb e } \left[\left .
\left ( \int_t^\infty p_u { \mbox{$\rm e$}}^{-\kappa(u - t ) } { \mbox{$\rm d$}}u \right ) x_t \right| \xi_t , x_t \right ] \nonumber \\ & = & { \mathbb e } \left[\left
. a_t \right| \sigma t a_t + b_t \right ] + \left ( \int_t^\infty p_u { \mbox{$\rm e$}}^{-\kappa(u - t ) } { \mbox{$\rm d$}}u \right ) x_t . \label{eq:21 } \end{aligned}\ ] ] here @xmath77 and @xmath54 is the brownian noise at @xmath5 .
note that the conditioning with respect to @xmath78 in the first term above drops out since @xmath79 and @xmath78 are independent by ( [ eq:8 ] ) .
the problem of finding the price is thus reduced to that of calculating a conditional expectation of the form @xmath80 $ ] for @xmath81 , where @xmath82 is given in ( [ eq:21.5 ] ) and @xmath83 . we observe that @xmath82 and @xmath84 are independent and that both are gaussian random variables . in order to compute the conditional expectation above
, we recall another result concerning an orthogonal decomposition of independent gaussian random variables : if @xmath85 and @xmath86 are independent gaussian random variables , then the random variables @xmath87 and @xmath88 are orthogonal and thus independent if we set @xmath89/ ( { \rm var}[a]+{\rm var}[c])$ ] .
this follows from the fact that if @xmath85 and @xmath86 are independent gaussian , then there is a value of @xmath90 such that @xmath91 = 0 $ ] . in view of this observation ,
let us express @xmath82 in the form @xmath92 where @xmath93}{{\rm var}[a_t]+{\rm var}[c_t ] } .
\end{aligned}\ ] ] then we find that @xmath94 = z_t(a_t+c_t ) + ( 1-z_t){\mathbb e}[a_t ] - z_t { \mathbb e}[c_t ] .
\label{eq:22 } \end{aligned}\ ] ] clearly , we have @xmath95=0 $ ] . furthermore ,
if we set @xmath96 in equation ( [ eq:7 ] ) we deduce that @xmath97 & = & { \mathbb e}\left [ \theta \int_t^\infty p_{u } \left(1-{\mbox{$\rm e$}}^{-\kappa ( u - t)}\right ) { \mbox{$\rm d$}}u + \psi \int_t^\infty { \mbox{$\rm e$}}^{-\kappa u}p_{u } \int^{u}_{t } { \mbox{$\rm e$}}^{\kappa s } { \mbox{$\rm d$}}\beta_{s } { \mbox{$\rm d$}}u \right ] \nonumber \\ & = & \theta \int_t^\infty p_u { \mbox{$\rm d$}}u - \theta\int_t^\infty p_u { \mbox{$\rm e$}}^{-\kappa(u - t ) } { \mbox{$\rm d$}}u . \label{eq:23}\end{aligned}\ ] ] the final step in deriving the commodity price is to determine the variances of @xmath82 and @xmath84 . to simplify the notation
let us write @xmath98 then a short calculation shows that @xmath99 = \psi^2 \int_t^\infty q_s^2 { \mbox{$\rm d$}}s , \qquad { \rm var}[c_t ] = \frac{1}{\sigma^{2 } t } , \label{eq:24}\end{aligned}\ ] ] and hence that @xmath100 putting these results together , we deduce that the price of the commodity at @xmath5 is given by @xmath101 observe that the first term in ( [ eq:26 ] ) is essentially the annuity valuation of a constant dividend rate set at the reversion level @xmath20 , together with a correction term to adjust for the present level of the dividend rate . the second term , on the other hand , represents the contribution from the noisy observation of the future dividend flow .
several interesting observations can be made regarding the weight factor @xmath102 given by ( [ eq:25 ] ) , which clearly lies between zero and one for all @xmath5 . for large @xmath22 and/or large @xmath14 , the value of @xmath102 tends to unity ; whereas for small @xmath22 and/or small @xmath14 , the value of @xmath90 tends to 0 .
hence , if the market information has a low noise content , or if the volatility of the convenience dividend is high , then market participants also rely heavily on the information available about the future in their determination of the price , rather than assuming that the current value of the dividend is a good indicator for the future .
conversely , in the absence of a strong signal concerning the future dividend flow , an annuity valuation based on the current dividend level will dominate the price .
we see therefore that important intuitive characteristics are encoded explicitly in the pricing formula ( [ eq:26 ] ) .
indeed , ( [ eq:26 ] ) captures rather well the idea of information - based asset pricing , showing how varying amounts of information about the future can affect the development of prices , and that prices typically represent a kind of compromise between what we know for sure at some given time , and the less trustworthy but nevertheless significant intelligence that we may possess regarding events that lay ahead . in the special case for which
the interest rate is constant , the valuation formula ( [ eq:26 ] ) simplifies somewhat to give the following : @xmath103 + z_{t}\ , \frac{{\mbox{$\rm e$}}^{rt}}{\sigma t}\ , \xi_t , \label{eq:27}\end{aligned}\ ] ] where the weight factors are @xmath104 we have performed monte carlo simulation studies to gain further intuition concerning the dynamical behaviour of the commodity price .
furthermore , we have calibrated the model parameters to the prices of crude oil , and we have compared the resulting simulated sample paths to real market data . in the case of the crude oil markets we are able to estimate the expected long - term future spot price from historical average of spot prices , since there exists a supply - demand equilibrium price level to which the long - run price tend to converge .
the results indicate that even in the constant - parameter model considered above , the model is sufficiently rich to capture elements of the behaviour of real market data , as illustrated in figure [ fig:2 ] . indeed , the behaviour of the simulated sample paths is close to the actual price path in the real market , even though the model parameters are taken to be constant . ,
@xmath105 , @xmath42 , @xmath106 , and @xmath107 .
[ fig:2],height=245 ]
having obtained the price ( [ eq:27 ] ) for the commodity in the constant parameter case , we shall find it of interest to deduce the stochastic differential equation to which ( [ eq:27 ] ) is a solution .
one of the advantages of the information - based approach is that one is effectively working in the space of solutions to a dynamical equation ; hence we have circumvented the often cumbersome process of solving differential equations .
nevertheless , having obtained an explicit expression for the price it is of interest to identify the associated dynamical equation , which in turn gives us an expression for the innovations process , the underlying brownian motion that `` derives '' the price dynamics .
we begin by rewriting ( [ eq:27 ] ) in the form @xmath108 where @xmath109 this can be verified by observing that @xmath110 + b_t \nonumber \\ & = & \omega_t + \sigma t \left[\frac{rx_0+\kappa \theta}{r\left(r+\kappa\right ) } + \frac{\psi}{r+\kappa } \int_0^t { \mbox{$\rm e$}}^{-ru } { \mbox{$\rm d$}}\beta_u - \int_0^t { \mbox{$\rm e$}}^{-ru } x_u { \mbox{$\rm d$}}u\right ] .
\label{eq:31}\end{aligned}\ ] ] by using the representation ( [ eq:29 ] ) , taking the stochastic differential , and rearranging terms , we obtain @xmath111 substituting the relations @xmath112 into ( [ eq:32 ] ) , and making a further rearrangement of terms , we deduce that @xmath113\right\ } , \label{eq:34}\end{aligned}\ ] ] where @xmath114 note that @xmath115 is the conditional variance of the discounted aggregation of the future dividends , given by @xmath116 .
the apparently redundant appearance of @xmath84 in ( [ eq:34 ] ) will be clarified shortly . from ( [ eq:34 ] )
we obtain : the dynamical equation satisfied by the commodity price process is @xmath117 where @xmath118 , defined by @xmath119 and @xmath120 , \label{eq : x45}\end{aligned}\ ] ] is an @xmath10-brownian motion on @xmath121 , and @xmath122 we begin by introducing an @xmath123-martingale @xmath124 by the relation @xmath125 = p_t s_t + \int_0^t p_u x_u { \mbox{$\rm d$}}u .
\label{eq:38}\end{aligned}\ ] ] if we substitute @xmath126 and ( [ eq:30 ] ) in ( [ eq:38 ] ) , we find @xmath127 an application of ito s lemma then yields @xmath128 but this is just the term appearing in ( [ eq:32 ] ) . now from the martingale representation theorem , if we let @xmath118 be the @xmath129-brownian motion , then there exist a unique adapted process @xmath130 such that @xmath131 hence , on one hand we have @xmath132 while on the other hand we have ( [ eq:34 ] ) . squaring these relations we find , respectively , @xmath133 since
@xmath134 is unique , it must be that @xmath135 , and for @xmath118 we obtain ( [ eq : x45 ] ) , as claimed .
@xmath73 note that @xmath118 is the innovations process for the underlying filtering problem . in general
, it is difficult to obtain an explicit expression for the innovations process in such problems .
however , by an indirect approach using the martingale representation theorem we were able to obtain the underlying brownian motion in closed form .
we now return to the price process ( [ eq:27 ] ) and consider the problem of the valuation of a european - style call option with strike @xmath136 and maturity @xmath137 , for which @xmath138 \label{eq:46}.\end{aligned}\ ] ] observe from ( [ eq:27 ] ) that the random variable @xmath139 consists additively of three random components @xmath140 , @xmath141 , and @xmath142 , and that all three are gaussian .
it follows that @xmath139 is also gaussian , and thus that it holds that @xmath143 } } \int_k^\infty ( z - k ) \exp\left(-\frac{\left(z-{\mathbb e}[s_t]\right)^{2 } } { 2\,{\rm var}[s_t ] } \right){\mbox{$\rm d$}}z .
\label{eq:47}\end{aligned}\ ] ] performing this integral , we obtain @xmath144}{2\pi}}\exp\left ( -\frac{({\mathbb e}[s_t]-k)^2}{2{\rm var}[s_t]}\right ) + ( { \mathbb e}[s_t]-k ) n\left(\frac{{\mathbb e}[s_t]-k}{\sqrt{{\rm var}[s_t]}}\right)\right ] , \label{eq:48}\end{aligned}\ ] ] where @xmath145 is the normal distribution function . therefore , the problem of option valuation reduces to the determination of the mean and the variance of @xmath139 . after a lengthy but straightforward calculation
, we deduce that @xmath146 = \frac{1}{r}\left[\frac{\kappa}{r+\kappa}\theta + \frac{r}{r+\kappa } \left[{\mbox{$\rm e$}}^{-\kappa t } x_{0 } + \theta ( 1 - { \mbox{$\rm e$}}^{-\kappa t } ) \right]\right ] , \label{eq:49}\end{aligned}\ ] ] and that @xmath147 = \frac{\psi^{2}}{2\kappa \left(r+\kappa\right)^2}(1-{\mbox{$\rm e$}}^{-2\kappa t } ) + z_t^2 \left[\frac{\psi^{2}}{2r(r+\kappa)^2 } + \frac{{\mbox{$\rm e$}}^{2rt}}{\sigma^{2 } t}\right ] .
\label{eq:50}\end{aligned}\ ] ] substitution of ( [ eq:49 ] ) and ( [ eq:50 ] ) in ( [ eq:48 ] ) then gives the option price . in figure [ fig:3 ]
we have plotted the option surface as a function of the initial price @xmath148 and the option maturity @xmath137 . and the time to maturity @xmath137 in the ou model .
the parameters are set as @xmath149 , @xmath150 , @xmath151 , @xmath152 , @xmath153 , and @xmath154 .
the value of @xmath20 ranges from 0.3 to 0.8 , corresponding to the range of the initial price @xmath148 .
the maturity @xmath137 ranges from @xmath155 to @xmath156 .
[ fig:3],height=264 ]
we have derived the option price when the underlying is the spot instrument . however , many of the market traded commodity products are futures contracts , and their derivatives , and it is here where the greatest liquidity often lies .
the model under consideration will therefore be of more practical use if we extend the analysis to cover derivatives on futures contracts .
let us write @xmath157 for futures price at time @xmath5 for a futures contract that matures at time @xmath137 , where @xmath25 .
clearly , from the theory of futures prices we have @xmath158 = { \mathbb e}[s_t|{\mathcal f}_t ] , \label{eq:51}\end{aligned}\ ] ] since at maturity @xmath137 , the futures price should be equal to the spot price . writing @xmath159 for the conditional expectation , we find that @xmath160}{r(r+\kappa ) } + \frac{{\mbox{$\rm e$}}^{rt}z_t}{\sigma t}\ , { \mathbb e}_t[\omega_t ] , \end{aligned}\ ] ] where we have made use of the expression ( [ eq:29 ] ) for the spot price . to determine the futures price we are therefore required to calculate @xmath161 = { \mbox{$\rm e$}}^{-\kappa ( t- t ) } x_{t } + \theta ( 1 - { \mbox{$\rm e$}}^{-\kappa ( t - t ) } ) + \psi { \mbox{$\rm e$}}^{-\kappa t } { \mathbb e}_t \left[\int_t^t { \mbox{$\rm e$}}^{\kappa u } { \mbox{$\rm d$}}\beta_{u}\right ] \label{eq:53}\end{aligned}\ ] ] and @xmath162 = \frac{\sigma\psi t}{r+\kappa}{\mathbb e}_t \left[\int_t^\infty { \mbox{$\rm e$}}^{-ru}{\mbox{$\rm d$}}\beta_u\right ] + { \mathbb e}_t[b_t ] .
\label{eq:54}\end{aligned}\ ] ] we observe now that the filtration @xmath10 generated jointly by @xmath16 and @xmath2 is equivalent to that generated by @xmath16 and @xmath163 .
furthermore , the three processes appearing in ( [ eq:53 ] ) and ( [ eq:54 ] ) are independent of @xmath164 .
if we take note of the markov property in addition , then the conditioning reduces to the specification of the random variable @xmath165 .
we proceed to calculate these conditional expectations .
we note first that from the tower property of conditional expectation we have @xmath166 = { \mathbb e}[{\mathbb e}[b_t|\omega_t , b_t]|\omega_t]={\mathbb e}[b_t|\omega_t]$ ] .
thus , by use of ( [ eq:30 ] ) we find @xmath167 & = & { \mathbb e } \left[\left.\omega_t - \frac{\sigma\psi t}{r+\kappa}\int^{\infty}_t { \mbox{$\rm e$}}^{-ru}{\mbox{$\rm d$}}\beta_u\right|\omega_t\right ] \nonumber \\ & = & \omega_t - \frac{\sigma\psi t}{r+\kappa}\ , { \mathbb e } \left[\left.\int_t^\infty { \mbox{$\rm e$}}^{-ru}{\mbox{$\rm d$}}\beta_u\right|\omega_t\right ] \nonumber \\ & = & \omega_t - \frac{\sigma\psi t}{r+\kappa } \frac{r+\kappa}{\sigma\psi t}\ , z_t\omega_t \nonumber \\ & = & \omega_t(1-z_t).\end{aligned}\ ] ] next we consider @xmath168 & = & { \mathbb e } \left[\left.\int_t^\infty { \mbox{$\rm e$}}^{-r u } { \mbox{$\rm d$}}\beta_{u } - \int_t^t { \mbox{$\rm e$}}^{-r u } { \mbox{$\rm d$}}\beta_{u}\right|\omega_t\right ] \nonumber \\ & = & \frac{r+\kappa}{\sigma\psi t } z_t\omega_t - { \mathbb e } \left[\left.\int_t^t { \mbox{$\rm e$}}^{-r
u } { \mbox{$\rm d$}}\beta_{u}\right| \omega_t\right ] .
\end{aligned}\ ] ] finally , we observe that @xmath169 & = & { \mathbb e } \left[\left.{\mathbb e } \left[\left.\int_t^t { \mbox{$\rm e$}}^{-r u } { \mbox{$\rm d$}}\beta_{u}\right| \omega_t , \left\{\beta_u\right\}_{0\leq u\leq t}\right]\right|\omega_t\right ] \nonumber \\ & = & { \mathbb e } \left[\left.{\mathbb e } \left[\left.\int_0^t { \mbox{$\rm e$}}^{-r u } { \mbox{$\rm d$}}\beta_{u } -\int_0^t { \mbox{$\rm e$}}^{-r u } { \mbox{$\rm d$}}\beta_u\right|\omega_t , \left\{\beta_u\right\}_{0\leq u\leq t } \right]\right|\omega_t\right ] \nonumber \\ & = & { \mathbb e } \left[\left.\int_0^t { \mbox{$\rm e$}}^{-r u } { \mbox{$\rm d$}}\beta_u- \int_0^t { \mbox{$\rm e$}}^{-r u } { \mbox{$\rm d$}}\beta_u \right|\omega_t\right ] \nonumber \\ & = & 0 .\end{aligned}\ ] ] putting these results together , we deduce that @xmath170}{\sigma tt}\ , \omega_t . \end{aligned}\ ] ] to derive a price for options on futures we are required to determine the mean and the variance of the random variable @xmath157 .
one can easily verify that @xmath171=f_0^t$ ] .
for the variance , observe that the futures price is linear in @xmath78 and @xmath165 , and since these two random variables are independent , the variance takes the form @xmath172 = \frac{{\mbox{$\rm e$}}^{-2\kappa(t - t)}}{(r+\kappa)^2 } { \rm var}[x_t ] + \frac{{\mbox{$\rm e$}}^{2rt}z_t^2[z_t^t t+(1-z_t)t]^2}{\sigma^2 t^2 t^2 } { \rm var}[\omega_t ] .
\end{aligned}\ ] ] but from @xmath173 = \frac{\psi^2}{2\kappa}(1-{\mbox{$\rm e$}}^{-2\kappa t } ) , \quad { \rm var}[\omega_t ] = \frac{\sigma^2 \psi^2 t^2}{2r(r+\kappa)^2{\mbox{$\rm e$}}^{2rt } } + t = \frac{t}{1-z_t } , \end{aligned}\ ] ] we deduce that @xmath172 = \frac{\psi^2({\mbox{$\rm e$}}^{-2\kappa(t - t)}-{\mbox{$\rm e$}}^{-2\kappa t } ) } { 2\kappa(r+\kappa)^2 } + \frac{{\mbox{$\rm e$}}^{2rt}[z_t z_t t + ( 1-z_t)z_t t]^2 } { \sigma^2 t^2 ( 1-z_t)t } .
\label{eq:61}\end{aligned}\ ] ] finally , we are in the position to write down the pricing formula for options on futures .
the payout is @xmath174 , where @xmath5 is the option maturity and @xmath136 is the strike .
since @xmath157 is gaussian , direct integration gives @xmath175}{2\pi } } \exp\left ( - \frac{({\mathbb e}[f_t^t]-k)^2}{2{\rm var}[f_t^t]}\right)+ \left({\mathbb e}[f_t^t]-k\right ) n\left ( \frac{{\mathbb e}[f_t^t]-k}{\sqrt{{\rm var}[f_t^t]}}\right)\right],\end{aligned}\ ] ] where @xmath171=f_0^t$ ] and @xmath176 $ ] is as given in ( [ eq:61 ] ) .
we conclude by remarking that one advantage of the model introduced in this paper is that it can readily be extended to the time - inhomogeneous case .
for instance , writing @xmath177 , the time - inhomogeneous ou process @xmath178
\label{eq:2c}\ ] ] admits the orthogonal decomposition property such that the random variables @xmath78 and @xmath179 are independent .
we can then exploit the properties of the time - inhomogeneous ou bridge @xmath180 to work out an expression for the commodity price . following a line of argument similar to that presented here , we find that the commodity price in the time - inhomogeneous situation is given by the expression @xmath181 + z_t \frac{\xi_{t}}{\sigma t } , \label{eq:3.34}\end{aligned}\ ] ] where @xmath182 further analysis shows that closed - form expressions for option prices can be obtained in the time - inhomogeneous setup , leading to a rather more flexible class of models .
the authors are grateful to participants at the amamef ( advanced mathematical methods in finance ) conferences in alesund , norway ( 2009 ) and bled , slovenia ( 2010 ) , the workshop on derivatives pricing and risk management at the fields institute , toronto ( 2010 ) , and meetings at the university at warwick ( 2011 ) and the london school of economics ( 2012 ) , where parts of this work were presented , for helpful comments .
xy would like to acknowledge financial support from shell international limited .
brody , d. c. , l. p. hughston , and a. macrina ( 2007 ) : beyond hazard rates : a new framework for credit - risk modelling . in : _ advances in mathematical finance : festschrift volume in honour of dilip madan _ ( m. fu , r. jarrow , ju - yi yen , & r. elliott , eds . ) , 231257 .
basel : birkhuser . | this paper introduces an information - based model for the pricing of storable commodities such as crude oil and natural gas .
the model makes use of the concept of market information about future supply and demand as a basis for valuation .
physical ownership of a commodity is regarded as providing a stream of `` convenience dividends '' equivalent to a continuous cash flow .
the market filtration is assumed to be generated jointly by ( i ) current and past levels of the dividend rate , and ( ii ) partial information concerning the future of the dividend flow .
the price of a commodity is given by the expectation of the totality of the discounted risk - adjusted future convenience dividend , conditional on the information provided by the market filtration .
in the situation where the dividend rate is modelled by an ornstein - uhlenbeck process , the prices of options on commodities can be derived in closed form , both in the case when underlying is the spot price , and in the case when underlying is a futures price .
the approach presented can be applied to other assets that can yield potentially negative effective cash flows , such as real estate , factories , refineries , mines , and power generating plants .
+ * keywords * : commodity markets ; commodity derivatives ; crude oil ; natural gas ; convenience yield ; information - based asset pricing ; market microstructure . |
let @xmath3 be a partition of @xmath0 , denoted as usual by @xmath4 .
let @xmath5 denote the number of standard tableaux of shape @xmath6 .
@xmath5 is also the number of paths in the young graph @xmath7 from the root @xmath8 to @xmath6 , and it is also the degree of the irreducible character @xmath9 of the symmetric group @xmath10 .
a partition @xmath11 is said to be _ strict _ if @xmath12 for some @xmath13 . in that case
we write @xmath14 .
let @xmath15 be the subgraph of the young graph @xmath7 formed by the strict partitions .
if @xmath14 , let @xmath16 denote the number of paths in @xmath15 from the root @xmath8 to @xmath6 . according to a theorem by schur , @xmath16 is the degree of the projective representation of @xmath10 corresponding to @xmath6 .
it is also the number of standard young tableaux of _ shifted shape _ @xmath6 .
vershik and kerov @xcite have determined the asymptotic shape of the partition @xmath6 that maximizes @xmath5 as @xmath17 .
they have also shown that the same shape is also the asymptotic expected shape of a random partition with respect to the plancherel measure .
the latter result was reached independently by logan and shepp @xcite as well . through the robinson - schensted algorithm
, the expected shape relates to the expected length of the longest increasing subsequence in a random permutation .
for some recent developments related to this problem and the probability distributions involved , see @xcite more precisely , given the young diagram of a partition @xmath4 where each box is @xmath18 , shrink it along both axes by a factor of @xmath19 to obtain the re - scaled diagram @xmath20 of total area @xmath21 . for each @xmath0 , let @xmath22 be a partition @xmath4 with maximal @xmath5 , that is , @xmath23 . through slight abuse of notation , where the maximizing partition is not unique for a given @xmath0 , we shall take @xmath22 to read `` any @xmath4 of maximal degree '' .
[ th : vk ] the limit shape as @xmath24 of the re - scaled diagrams @xmath25 exists , and is given by the two axes and by the parametric curve @xmath26 figure [ fig : vk ] the vershik - kerov limit shape of theorem [ th : vk ] is shown in figure [ fig : vk ] .
the problem of determining the asymptotic shape of the partition @xmath6 which maximizes @xmath16 remains unsolved , and we are unaware of even partial characterizations of the shape
. however , recently the following was conjectured .
[ conj : bhr ] the limit shape @xmath27 of the @xmath14 maximizing @xmath28 and possibly maximizing @xmath16 is given by the two axes and by the parametric curve @xmath29 figure [ fig : bhr ] the conjectured limit shape of conjecture [ conj : bhr ] is shown in figure [ fig : bhr ] .
it was obtained from the vershik - kerov shape by bisecting it along the line @xmath30 , taking the upper half , dilating it ( so that its area become @xmath21 ) and applying the shearing transformation @xmath31 ( to bring the line @xmath30 to the @xmath32 axis ) . in the next section ,
we give the results of computing the partition maximizing @xmath16 over all @xmath14 for @xmath33 .
we observe a property of successive maximizing partitions in the range @xmath33 , and conjecture that it holds for all @xmath0 . assuming the conjecture ,
we compute the maximizing partitions for @xmath34 . our results are consistent with conjecture [ conj : bhr ] for both @xmath16 and @xmath35 .
to compute @xmath16 , we used the following formula , due to schur . let @xmath37 .
then @xmath38 all computations were done with _
mathematica_. for @xmath33 ,
@xmath16 was computed for every @xmath14 and the partitions attaining the maximal value for each @xmath0 were identified . for every @xmath39\setminus\{3,11\}$ ] , there exists a partition @xmath40 such that @xmath41 for all @xmath42 .
@xmath43 figure [ fig:100 ] figure [ fig:100 ] shows the normalized diagram @xmath44 overlaid with the conjectured limit shape .
recall that by conjecture [ conj : bhr ] , the partitions maximizing @xmath16 are asymptotically equal to the partitions maximizing @xmath35 . for all strict partitions @xmath6 of @xmath33 ,
@xmath35 was computed and the partition maximizing it was denoted @xmath45 . comparing the results with the results for @xmath16 , the following was observed . for all @xmath46 @xmath47 .
let @xmath48 and let @xmath49 .
the @xmath50-successors of @xmath51 are the elements of the set @xmath52 we have observed the following . for @xmath53 , @xmath54 . when @xmath55 is not unique , read the above to mean `` every @xmath56 of maximal projective degree is a @xmath21-successor of every @xmath14 of maximal degree ''
based on the above observation , we conjecture the following . assuming that the conjecture holds , @xmath57 was computed for @xmath58 as follows : statring with @xmath59 , for every @xmath60 , the ratio @xmath61 was computed and the @xmath51 maximizing the ratio was selected . | the projective degrees of strict partitions of @xmath0 were computed for all @xmath1 and the partitions with maximal projective degree were found for each @xmath0 . it was observed that maximizing partitions for successive values of @xmath0 `` lie close to each other '' in a certain sense . conjecturing that this holds for larger values of @xmath0 ,
the partitions of maximal degree were computed for all @xmath2 .
the results are consistent with a recent conjecture on the limiting shape of the strict partition of maximal projective degree . |
polarised deep inelastic scattering experiments at cern @xcite , desy @xcite and slac @xcite have revealed an apparent two ( or more ) standard deviations violation of ozi in the flavour - singlet axial charge @xmath1 which is extracted from the first moment of @xmath0 ( the nucleon s first spin dependent structure function ) .
this discovery has inspired much theoretical and experimental effort to understand the internal spin structure of the nucleon . in this article
we review the theory and present status of the proton spin problem in qcd .
we start with a simple sum - rule for the spin of the proton ( @xmath2 ) in terms of the angular momentum of its quark and gluonic constituents : @xmath3 here , @xmath4 and @xmath5 are the quark and gluonic intrinsic spin contributions to the nucleon s spin and @xmath6 is the orbital contribution .
one would like to understand the spin decomposition , eq.(1 ) , both in terms of the fundamental qcd quarks and gluons and also in terms of the constituent quark quasi - particles of low - energy qcd . in deep inelastic processes the internal structure of the nucleon
is described by the qcd parton model @xcite .
the deep inelastic structure functions may be written as the sum over the convolution of `` soft '' quark and gluon parton distributions with `` hard '' photon - parton scattering coefficients .
the ( target dependent ) parton distributions describe a flux of quark and gluon partons carrying some fraction @xmath7 of the proton s momentum into the hard ( target independent ) photon - parton interaction which is described by the hard scattering coefficients . in low energy processes
the nucleon behaves like a colour neutral system of three massive constituent quark quasi - particles interacting self consistently with a cloud of virtual pions which is induced by spontaneous chiral symmetry breaking @xcite .
one of the most challenging problems in particle physics is to understand the transition between the fundamental qcd `` current '' quarks and gluons and the constituent quarks of low - energy qcd .
the fundamental building blocks are the local qcd quark and gluon fields together with the non - local structure @xcite associated with gluon topology @xcite .
the large mass of the constituent quarks is usually understood in terms of dynamical chiral symmetry breaking and the interaction of the current quarks with a scalar condensate ( in the nambu - jona - lasinio model @xcite ) or in terms of scalar confinement ( in the bag model @xcite ) . through the axial anomaly in qcd @xcite ,
some fraction of the spin of the nucleon and of the constituent quark is carried by its quark and gluon partons and some fraction is carried by gluon topology @xcite .
the topological winding number is a global property of the theory ; it is independent of the local quark and gluon fields .
when we take the fourier transform to momentum space any topological contribution to the nucleon s spin has support only at @xmath8 .
this means that whereas the nucleon s momentum is given by the sum of the momenta of the partons @xmath9 one has to be careful about writing equations such as @xmath10 some fraction of the nucleon s spin may reside at bjorken @xmath8 .
this effect may be viewed as a deformation of the qcd @xmath11 vacuum @xcite inside a nucleon due to tunneling processes between vacuum states with different topological winding number . in semi - classical quark models
the quark spin content @xmath12 is equal to the nucleon s flavour singlet axial charge @xmath1 .
relativistic quark - pion coupling models predict @xmath13 with about 40% of the proton s spin being carried by orbital angular momentum @xcite . the value of @xmath1 which is extracted from polarised deep inelastic scattering experiments is @xmath14
@xmath15 [ 1 - 9 ] ; significantly less than the semi - classical prediction for @xmath1 . in qcd
the axial anomaly induces various gluonic contributions to the flavour singlet axial charge @xmath1 .
explanations of the small value of @xmath16 have been proposed based on the qcd parton model @xcite and non - perturbative chiral @xmath17 dynamics [ 25 - 31 ] . one finds @xcite @xmath18 where @xmath19 ( =3 ) is the number of light flavours liberated into the final state . here
@xmath20 and @xmath5 are the amount of spin carried by quark and gluon partons in the polarised proton and @xmath21 measures the gluon - topological contribution to @xmath1 @xcite .
the topological term @xmath21 has support only at @xmath8 ; it is missed by polarised deep inelastic scattering experiments which measure the combination @xmath22 . if some fraction of the spin of the constituent quark is carried by gluon topology in qcd , then the constituent quark model predictions for @xmath1 are not necessarily in contradiction with the small value of @xmath23 extracted from deep inelastic scattering experiments . in section 5
we explain a simple dynamical mechanism for producing such an effect .
if there is no topological @xmath8 term , then the small value of @xmath16 would be consistent with the semi - classical prediction for @xmath12 if @xmath5 is both large and positive ( @xmath24 at @xmath25gev@xmath26 ) .
if discovered in future experiments , such a large @xmath5 would pose a challenge for constituent quark models which do not naturally include such an effect .
in this article we review our present understanding of the spin structure of the nucleon , both at the constituent quark level and in qcd .
we emphasise present and future experiments which could help to unravel the nucleon s internal spin structure by separately measuring the @xmath27 , @xmath28 and @xmath21 contributions to @xmath1 .
the structure of the paper is as follows . in section 2
we give a brief introduction to deep inelastic scattering and the polarised deep inelastic measurements of the nucleon s axial charges @xmath29 . in section 3
we review the theoretical interpretation of the axial charges @xmath29 and the ozi violation observed in deep inelastic measurements of @xmath1 .
the connection between chiral symmetry and the spin structure of the nucleon is emphasised in section 3.4 .
sections 4 - 6 are more technical . in section 4
we review the axial anomaly in qcd and its role in our understanding of the physics of @xmath1 . in section 5
we discuss the relationship between gluon topology and the spin structure of the nucleon .
section 6 gives an overview of the qcd parton model and its application to polarised deep inelastic scattering . in section 7
we discuss the shape of @xmath0 and the @xmath30 dependence of the various contributions to the nucleon s axial charges and @xmath31 .
finally , in section 8 , we summarise the information about the spin structure of the nucleon which can be learnt in present and future experiments .
we refer to refs.[32 - 38 ] for further reviews on the spin structure of the nucleon in polarised deep inelastic scattering and to ref.@xcite for a review of polarised photoproduction and the transition to polarised deep inelastic scattering .
our present knowledge about the spin structure of the nucleon comes from polarised deep inelastic scattering experiments .
these experiments involve scattering a high - energy charged lepton beam from a nucleon target at large momentum transfer squared .
one measures the inclusive cross - section .
the lepton beam ( electrons at desy and slac and muons at cern ) is longitudinally polarised .
the nucleon target may be either longitudinally or transversely polarised .
consider polarised @xmath32 scattering .
we work in one photon exchange approximation .
whilst the electron photon vertex is described by perturbative qed , the internal qcd structure of the proton means that the photon proton interaction is described in terms of various structure functions ( form - factors ) .
let @xmath33 , @xmath34 and @xmath35 denote the momentum , mass and spin of the target proton and @xmath36 denote the momentum of the exchanged photon .
define @xmath37 and @xmath38 .
deep inelastic scattering involves working in the bjorken limit : @xmath39 and @xmath40 both @xmath41 with the bjorken variable @xmath42 held fixed .
we specialise to the target rest frame and let @xmath43 denote the energy of the incident electron which is scattered through an angle @xmath11 to emerge in the final state with energy @xmath44 .
let @xmath45 denote the longitudinal polarisation of the beam and @xmath46 denote a longitudinally polarised proton target .
the unpolarised and polarised differential cross - sections are : @xmath47\ ] ] and @xmath48 .\ ] ] here @xmath49 and @xmath50 denote the nucleon s first and second spin independent structure functions ; @xmath0 and @xmath51 denote the first and second spin dependent structure functions .
the structure functions contain all of the target dependent information in the deep inelastic process .
the unpolarised structure functions have been measured in experiments at cern , desy , fnal and slac .
we refer to @xcite for recent reviews of these data and the interpretation of @xmath49 and @xmath50 .
polarised deep inelastic scattering experiments with longitudinally polarised targets provide the cleanest probe of @xmath0 . in a fixed target experiment the contribution of the second spin structure function @xmath51 to the differential cross section , eq.(6 ) , is suppressed relative to the @xmath0 contribution by the kinematic factor @xmath52 . in deep inelastic experiments
@xmath52 is typically less than 0.03 and the @xmath51 contribution to eq.(6 ) becomes lost among the experimental errors .
the structure function @xmath51 can be measured using a transversely polarised target . in this case
the two spin structure functions @xmath0 and @xmath51 contribute to the spin dependent part of the total cross section with equal weight : @xmath53 .\ ] ] the experimental programme in polarised deep inelastic scattering has so far mainly focussed on measurements of @xmath0 .
the first measurements of @xmath51 have recently been reported by the smc and slac e-143 and e-154 collaborations @xcite . in this review
we focus on the physics of @xmath0 .
we refer to jaffe @xcite for a theoretical review of @xmath51 . when @xmath54 , the light - cone operator product expansion relates the first moment of the structure function @xmath0 to the scale - invariant axial charges of the target nucleon by @xcite @xmath55 here @xmath56 , @xmath57 and @xmath58 are the isotriplet , su(3 ) octet and scale - invariant flavour - singlet axial charges respectively .
the flavour non - singlet @xmath59 and singlet @xmath60 coefficients are calculable in @xmath61-loop perturbation theory and have been calculated to @xmath62 precision @xcite .
the first moment of @xmath0 is constrained by low energy weak interactions . for proton states
@xmath63 with momentum @xmath64 and spin @xmath65 latexmath:[\ ] ] where @xmath342 and @xmath343 . the splitting functions @xmath344 in eq.(63 )
have been calculated at next - to - leading order by mertig , zijlstra and van neervan @xcite and by vogelsang @xcite .
the size of @xmath28 is one of the key issues in qcd spin physics at the present time .
the polarised gluon distribution @xmath345 contributes to @xmath0 through the convolution @xmath346 .
depending on the choice of factorisation scheme , the gluonic coefficient @xmath300 has at most a @xmath308 singularity when @xmath309 .
in contrast , the leading term in the quark coefficient @xmath299 is @xmath347 .
the convolution involving @xmath300 has the practical effect that @xmath5 makes a direct contribution to @xmath0 only at @xmath348 @xcite . at the same time , the @xmath303 evolution of the flavour - singlet quark distribution involves the polarised gluon distribution @xmath349 at values of @xmath350 in the range ( @xmath351 ) @xcite see eq.(63 ) .
thus , through evolution , the polarised gluon distribution is relevant to the shape of @xmath0 over the complete @xmath30 range .
this result enables one to carry out next - to - leading order qcd fits to polarised deep inelastic data with the hope of extracting some information about @xmath28 . here
, one starts with an ansatz for the shape of @xmath352 and @xmath353 at some particular input scale @xmath354 .
the input distributions are evolved to the range of @xmath39 covered by the deep inelastic experiments .
finally , one chooses a particular factorisation scheme ( see below ) and makes a best fit to the @xmath0 data in terms of the input shape parameters and the scale @xmath354 .
several groups have followed this approach [ 4,104,105,112 - 117 ] .
different qcd motivated fits to the polarised deep inelastic data yield values of @xmath355 between zero and + 2 at @xmath356gev@xmath26 .
the value of @xmath28 which is extracted from these fits depends strongly on the functional form which is assumed for the input distributions with only small changes in the overall @xmath357 for the fits @xcite we refer to de florian et al .
@xcite for a nice overview of qcd fits to @xmath0 data .
three schemes are commonly used in the analysis of experimental data : the @xmath358 cut - off , @xmath359 and ab schemes .
these schemes correspond to different procedures for separating the phase space for photon - gluon fusion into `` hard '' and `` soft '' contributions . in the parton model that we discussed in section 6.1 using the cut - off on the transverse momentum squared , the polarised gluon contribution to the first moment of @xmath0
is associated with two - quark jet events carrying @xmath133 .
the hard coefficient is given by @xmath360 in eq.(58 ) with @xmath361 and @xmath326 .
this `` parton model scheme '' is sometimes called the `` chiral invariant '' ( ci ) @xcite or jet @xcite scheme .
different schemes can be defined relative to this `` parton model scheme '' by the transformation @xmath362 where @xmath363 equals @xmath364 times a polynomial in @xmath30 .
the parton distributions transform under ( 64 ) as @xmath365 so that @xmath0 is left invariant by the change of scheme .
the virtuality and invariant - mass cut - off versions of the parton model that we discussed in section 6.1 correspond to different choices of scheme .
the @xmath359 and ab schemes are defined as follows . in the @xmath359 scheme
the gluonic hard scattering coefficient is calculated using the operator product expansion with @xmath359 renormalisation @xcite .
one finds @xcite : @xmath366 in this scheme @xmath367 so that @xmath368 decouples from @xmath369 .
this result corresponds to the fact that there is no gauge - invariant twist - two , spin - one , gluonic operator with @xmath370 to appear in the operator product expansion for the first moment of @xmath0 . in the @xmath359 scheme the contribution of @xmath371 to the first moment of @xmath0
is included into @xmath372 .
the ab scheme @xcite is defined by the formal operation of adding the @xmath30-independent term @xmath373 to the @xmath359 gluonic coefficient , viz .
@xmath374 in both the parton model and ab schemes @xmath375 .
we refer to cheng @xcite and llewellyn smith @xcite for a critical discussion of these schemes and their application to polarised deep inelastic scattering .
the smc and slac e-154 experiments quote values of @xmath5 for their own data .
the smc values are @xcite @xmath376 and @xmath377 in the @xmath378 @xcite and ab @xcite schemes respectively each at 1gev@xmath26 : the e-154 values are @xcite @xmath379 and @xmath380 each at 5gev@xmath26 .
dedicated experiments have been proposed to measure @xmath28 more precisely .
the compass @xcite and hermes @xcite experiments will measure charm production in polarised deep inelastic scattering ; a further experiment is proposed for slac @xcite .
polarised rhic @xcite will measure prompt photon production in polarised @xmath174 collisions through the process @xmath381 , thus enabling a different measurement of @xmath28 . in the longer term
there is a proposal to polarise the proton beam at hera @xcite .
a polarised proton beam at hera would allow precision measurements of @xmath0 at small @xmath30 where it becomes increasingly more sensitive to the polarised gluon distribution and to study the two - quark - jet cross - section associated with the axial anomaly .
there have been many theoretical papers proposing possible explanations of the small value of @xmath16 extracted from polarised deep inelastic scattering .
besides offering an explanation of the size of @xmath1 it is important to understand how the different possible effects contribute to the shape of @xmath0 , which contains considerably more information than just the first moment .
early constituent quark model calculations of the shape of @xmath0 @xcite still provide a reasonable description of the @xmath0 data in the `` valence '' region ( @xmath30 greater than about 0.2 ) .
more recent quark model calculations @xcite include qcd evolution from the bag `` input scale '' @xmath123 to deep inelastic @xmath39 . in their next - to - leading order bag model
fits to deep inelastic data steffens et al .
@xcite found that the model `` input scale '' increased corresponding to a change in the coupling @xmath382 from 0.8 to 0.6 when pion cloud corrections were included into the model input .
semi - inclusive measurements of @xmath0 will enable us to disentangle the separate @xmath383 and @xmath384 contributions to @xmath0 @xcite .
the first semi - inclusive measurements have been published by the smc @xcite ; more precise data will soon be available from hermes .
if the sum of @xmath385 and @xmath386 extracted from semi - inclusive measurements of @xmath0 falls short of the constituent quark model prediction for @xmath1 , then the `` discrepancy '' could be interpreted as a first hint that some of the nucleon s spin might reside at @xmath8 .
( recall from section 5 that instanton tunneling processes have the potential to shift some fraction of @xmath1 from valence partons carrying @xmath137 to the topological term at @xmath8 . )
the first semi - inclusive data ( from smc ) yield @xmath387 and @xmath388 . combining the errors in quadrature we find @xmath389 ( cf . the constituent quark model prediction @xmath13 ) and @xmath390 ( cf .
. it will be interesting to see how these results hold up in the light of more accurate data from hermes .
wislicki @xcite has analysed the polarised semi - inclusive data from smc and hermes looking for possible evidence of the axial anomaly at large @xmath30 .
the data shows no evidence of any deviation between the charge parity @xmath391 and @xmath392 polarised quark distributions in the `` valence '' region @xmath393 .
perturbative qcd counting rules @xcite make predictions for the large @xmath30 behaviour of @xmath0 .
the small @xmath30 extrapolation of @xmath0 data is presently the largest source of experimental error on deep inelastic measurements of the nucleon s axial charges .
the small @xmath30 extrapolation is usually motivated either by regge theory or by perturbative qcd arguments .
we now outline what is known about @xmath0 at large @xmath30 ( section 7.1 ) and at small @xmath30 ( section 7.2 ) . in section 7.3
we collect this theory and describe how it explains the shape of the measured spin dependent and spin independent isotriplet structure functions as a function of @xmath30 .
perturbative qcd counting rules predict that the parton distributions should behave as a power series expansion in @xmath394 when @xmath309 @xcite .
we use @xmath395 and @xmath396 to denote the parton distributions polarised parallel and antiparallel to the polarised proton .
one finds @xcite @xmath397 here , @xmath253 is the number of spectators and @xmath398 is the difference between the polarisation of the struck quark and the polarisation of the target nucleon . when @xmath309 the qcd counting rules predict that the structure functions should be dominated by valence quarks polarised parallel to the spin of the nucleon .
the ratio of polarised to unpolarised structure functions should go to one when @xmath309 .
the small @xmath30 extrapolation of @xmath0 data is important for precise measurements of the nucleon s axial charges from deep inelastic scattering .
the slac data @xcite has the smallest experimental error in the @xmath30 range @xmath399 .
we show these data in fig.1 .
there are several important properties of the @xmath0 data at small @xmath30 . 1 .
the magnitude of @xmath400 is significantly greater than the magnitude of @xmath401 in the measured small @xmath30 region .
this is in contrast to unpolarised deep inelastic scattering where the small @xmath30 region is dominated by isoscalar pomeron exchange .
the isosinglet @xmath401 is small and consistent with zero in the measured small @xmath30 range ( @xmath402 ) .
polarised gluon models @xcite predict that @xmath401 may become strongly negative at smaller values of @xmath30 ( @xmath403 ) but this remains to be checked experimentally .
3 . polarised deep inelastic data from cern and slac consistently indicate a strong isotriplet term in @xmath0 which rises at small @xmath30 .
we consider the isotriplet part of @xmath0 in more detail .
combining the proton data from e-143 @xcite together with the neutron data from e-154 @xcite , one finds a good fit @xcite to the slac data on @xmath400 : @xmath404 in the @xmath30 range @xmath399 at @xmath405gev@xmath26 .
regge theory makes a prediction for the large @xmath406 dependence of the spin dependent and spin independent parts of the total photoproduction @xmath407 cross - section .
it is often used to describe the small @xmath30 behaviour of deep inelastic structure functions ( @xmath39 larger than about 2gev@xmath26 ) .
the regge prediction for the isotriplet part of @xmath0 is @xcite @xmath408 here @xmath409 is the intercept of the @xmath410 regge trajectory .
if one makes the usual assumption that the @xmath410 trajectory is a straight line running parallel to the @xmath411 trajectories , then one finds @xmath412 .
clearly , regge theory does not provide a good description of @xmath400 in the measured @xmath30 range @xmath399 . at first glance , this result is surprising since regge theory provides a good description @xcite of the nmc measurements @xcite of both the isotriplet and isosinglet parts of @xmath50 in the same small @xmath30 range @xmath413 at @xmath405gev@xmath26 . in practice , the shape of @xmath0 at small @xmath30 is @xmath39 dependent .
the @xmath39 dependence is driven by dglap evolution and , at very small @xmath30 ( @xmath414 ) , by the resummation of @xmath415 radiative corrections see eg .
@xcite . to understand this evolution ,
let us define an effective intercept @xmath416 to describe the small @xmath30 behaviour of @xmath0 at finite @xmath39 : @xmath417 .
the net @xmath39 dependence of @xmath418 depends strongly on the value of @xmath418 which is needed to describe the leading twist part of @xmath400 at low momentum scales for example @xmath122gev@xmath26 .
let @xmath419 denote the leading twist ( = 2 ) part of @xmath400 .
dglap evolution of @xmath419 from @xmath123 to deep inelastic @xmath39 shifts the weight of the distribution from larger to smaller values of @xmath30 whilst keeping the area under the curve , @xmath56 , constant .
qcd evolution has the practical effect of `` filling up '' the small @xmath30 region increasing the value of @xmath418 with increasing @xmath39 .
the scale independence of @xmath56 provides an important constraint on the change in @xmath418 under qcd evolution .
the closer that @xmath420 is to the regge prediction -0.4 , the more that @xmath416 will grow in order to preserve the area under @xmath419 when we increase @xmath39 to values typical of deep inelastic scattering .
badelek and kwiecinski @xcite have investigated the effect of dglap and @xmath421 resummation on the small @xmath30 behaviour of @xmath400 . they find a good fit to the data using a flat small-@xmath30 input distribution at @xmath422gev@xmath26 . in their optimal nlo qcd fit
to polarised deep inelastic data glck et al .
@xcite used a rising input at @xmath423gev@xmath26 .
the isosinglet part of @xmath0 is more complicated because of possible gluonic exchanges in the @xmath424channel .
there have been several suggestions how the isosinglet part of @xmath0 should behave at small @xmath30 based on non - perturbative @xcite and perturbative @xcite qcd arguments . to understand the shape of @xmath400 it is helpful to compare the isotriplet part of @xmath0 with the isotriplet part of @xmath50 ( the nucleon s spin independent structure function ) .
in the qcd parton model @xmath425 \otimes \delta c_{ns}\ ] ] and @xmath426 \otimes c_{ns}.\ ] ] here @xmath427 and @xmath428 denote the up and down flavoured quark distributions polarised parallel @xmath429 and antiparallel @xmath430 to the target proton and @xmath431 and @xmath432 denote the spin - dependent and spin - independent perturbative qcd coefficients @xcite and @xmath431 have the perturbative expansion @xmath433 .
they are related ( in the @xmath378 scheme ) by @xcite @xmath434 the difference between @xmath432 and @xmath431 makes a non - negligible contribution to the deep inelastic structure functions only at @xmath348 . ] .
there is no gluonic or pomeron contribution to the isotriplet structure functions @xmath400 and @xmath435 . in fig.2
we show the slac data @xcite on @xmath436 together with the nmc measurement @xcite of @xmath437 .
the nmc data are quoted at @xmath438gev@xmath26 .
our slac data set is obtained by combining the published e-143 data on @xmath439 with the e-154 data on @xmath440 both at @xmath441gev@xmath26 .
we combine the @xmath0 measurements to produce one @xmath400 data point for each @xmath30 bin listed by the nmc .
clearly , @xmath442 is greater than @xmath435 for @xmath443 in this data ( see also @xcite ) .
sum - rules for the first moments of @xmath400 and the unpolarised structure function @xmath444 provide important constraints for our understanding of the structure of the nucleon .
the gottfried integral @xcite @xmath445 measures any su(2 ) flavour asymmetry in the sea .
the integral @xmath446 has been measured by the nmc in deep inelastic scattering ( @xmath447 ) @xcite and by the fermilab e-866 nusea collaboration in drell - yan production ( @xmath448 ) @xcite .
possible explanations of this effect include the pion cloud of the nucleon @xcite and the pauli - principle @xcite at work in the nucleon s sea .
we refer to thomas and melnitchouk @xcite for a pedagogical review of the physics involved in understanding the gottfried integral .
when we convolute the polarised and unpolarised parton distributions with the perturbative coefficients @xmath432 and @xmath431 the difference between the two coefficients makes a negligible contribution to the structure functions at @xmath449 the bulk of the @xmath30 range in fig.2 . dividing @xmath450 by the central values of @xmath446 measured by the nmc and fermilab e-866 experiments we obtain 1.78 and 1.57 respectively . whilst the physical values of @xmath56 and @xmath446 differ markedly from the simple su(6 ) predictions , it is interesting to observe that the ratio @xmath451 is consistent with the su(6 ) prediction ( @xmath452 ) .
in fig.3 we plot the ratio @xmath453 .
we also plot the su(6 ) prediction for @xmath454 , together with the result of dividing twice the value of @xmath455 by the central values of the measured gottfried integral @xmath446 from nmc ( top line ) and e-866 ( bottom line ) .
there are several observations to make .
first , the deep inelastic data is consistent with @xmath456 in the @xmath30 range @xmath457 @xcite . at large @xmath30
the data is consistent with the qcd counting rules prediction @xmath458 the data exhibits no evidence of the simple regge prediction @xmath459 when @xmath460 ( say @xmath461 ) .
the smc have recently published their measurements of @xmath462 down to @xmath463 .
this new data is consistent with the slac measurements and the observation @xmath464 .
we make some phenomenological observations which may help to understand the relative shapes of the isotriplet parts of @xmath0 and @xmath50 .
first , the total area under @xmath400 is fixed by the bjorken sum - rule .
soffer and teryaev @xcite have observed that @xmath465 of the bjorken sum - rule comes from small @xmath30 ( @xmath466 ) if the shape ( 68 ) is extrapolated to @xmath8 .
suppose that we pivot @xmath400 about its measured value at @xmath467 and impose the regge behaviour @xmath468 at smaller values of @xmath30 instead of the observed small @xmath30 behaviour @xmath469 .
we shall call this the `` regge modified '' @xmath400 . for
the `` regge modified '' @xmath400 the fraction of the total bjorken sum - rule which comes from @xmath30 less than 0.12 would be reduced to @xmath470 .
that is , @xmath471 of the total bjorken sum - rule would be shifted to larger values of @xmath30 .
a priori , one would expect the simple su(6 ) prediction for @xmath462 to come closest to the ratio of the leading twist contribution to the measured structure functions at a value @xmath472 close to @xmath473 _ after _ the leading twist parton distributions have been evolved to the quark model scale @xmath123 .
qcd evolution shifts the value of @xmath472 to slightly smaller @xmath30 at deep inelastic @xmath39 .
deep inelastic structure functions fall rapidly to zero when @xmath309 .
suppose we combine the `` regge modified '' @xmath400 together with the measured @xmath435 data .
the resultant `` regge modified ''
@xmath462 would considerably exceed the su(6 ) prediction @xmath474 in the intermediate @xmath30 region because of the extra area that has been shifted under @xmath400 at @xmath475 .
however , this contains the @xmath30 range where we would most expect the su(6 ) prediction to work ( if it is to work at all ) . in summary , _ if _ @xmath462 takes the su(6 ) value @xmath103 at some intermediate value @xmath472 and decreases towards one when we increase @xmath30 greater than @xmath472 , and if @xmath400 has a power law behaviour @xmath476 at small @xmath30 , then @xmath400 must rise at small @xmath30 ( in contradiction to regge theory ) with @xmath477 so that @xmath400 saturates the bjorken sum - rule with the physical value of @xmath56 .
to summarise this section , constituent quark model calculations provide a reasonable description of @xmath0 in the intermediate @xmath30 region .
the soft regge theory predictions for the small @xmath30 behaviour of the @xmath0 spin structure function seem to fail badly at deep inelastic @xmath39 .
the shape of the measured isotriplet spin structure function @xmath400 may be understood in terms of perturbative qcd counting rules ( at @xmath478 ) and constituent quark model ideas ( in the @xmath30 range @xmath479 ) .
relativistic constituent - quark pion coupling models predict @xmath480 .
the value of @xmath1 extracted from polarised deep inelastic scattering experiments is @xmath14 0.35 .
this result has inspired many theoretical ideas about the internal spin structure of the nucleon .
central to these ideas is the role that the axial anomaly plays in the transition from parton to constituent quark degrees of freedom in low energy qcd . in qcd
some fraction of the spin of the nucleon and of the constituent quark may be carried by gluon topology .
if the topological contribution @xmath21 is indeed finite , then the constituent quark model predictions for @xmath1 are not necessarily in contradiction with the small value of @xmath22 extracted from polarised deep inelastic scattering .
* semi - inclusive measurements of @xmath0 in the current fragmentation region ( hermes ) will enable more accurate measurements of the valence and sea quark contributions to @xmath0 .
if a polarised proton beam becomes available at hera it will be possible to extend this programme into the target fragmentation region and to study the target ( in-)dependence of the small value of @xmath16 .
* measurements of open charm production in polarised deep inelastic scattering ( compass , hermes and slac ) will enable a direct measurement of the polarised gluon distribution @xmath298 in deep inelastic scattering .
a complementary measurement of @xmath298 will come from studies of prompt photon production in polarised @xmath174 collisions at rhic .
it will be interesting to compare the deep inelastic and polarised @xmath174 measurements of @xmath28 .
a polarised proton beam at hera would enable us to measure @xmath0 at small @xmath30 where it is most sensitive to @xmath298 and to study the two - quark - jet cross - section associated with the axial anomaly . *
a precision measurement of @xmath135 elastic scattering or parity violation in light atoms would enable us to make an independent determination of @xmath58 .
the value of @xmath58 which is extracted from these elastic z@xmath289 exchange processes includes any contribution from topological polarisation at @xmath8 whereas the deep inelastic measurement does not thus , enabling us to measure the topological term @xmath21 .
the physics of the flavour - singlet axial - charge @xmath1 provides a bridge between the internal spin structure of the nucleon and chiral @xmath17 dynamics .
when combined with experimental and theoretical studies of the @xmath145 system , qcd spin physics offers a new window on the role of gluons in dynamical chiral symmetry breaking .
* acknowledgements : * + i thank b. povh for the invitation to write this review and for his hospitality at the mpi , heidelberg .
many of the ideas in this review have been developed in fruitful collaboration with m.m .
brisudova , s.j .
brodsky , r.j .
crewther , a. de roeck , p.v .
landshoff , i. schmidt , f.m .
steffens and a.w .
i also thank v.n .
gribov , h. fritzsch , p. minkowski , b. povh , d. schtte , g.m .
shore and w. weise for helpful discussions about the proton spin problem in qcd and constituent quark models , and n. bianchi , a. brll , g. garvey , m. moinester , j. pochodzalla and r. windmolders for discussions about experimental data .
99 emc collaboration ( j ashman et al . )
b206 * ( 1988 ) 364 ; nucl . phys .
* b328 * ( 1989 ) 1 . the spin muon collaboration ( d. adams et al . ) , phys .
* b396 * ( 1997 ) 338 ; + ( b. adeva et al . ) , phys .
b412 * ( 1997 ) 414 . the spin muon collaboration ( b. adeva et al ) , phys
* d58 * ( 1998 ) 112001 the spin muon collaboration ( b. adeva et al . ) , phys . rev .
* d58 * ( 1998 ) 112002 .
the hermes collaboration ( k. ackerstaff et al . ) , phys .
* b404 * ( 1997 ) 383 ; + the hermes collaboration ( a. airapetian et al . ) , phys . lett .
* b442 * ( 1998 ) 484 .
the e-130 collaboration ( g. baum et al . ) , phys .
* 51 * ( 1983 ) 1135 .
the e-142 collaboration ( p.l .
anthony et al . ) , phys . rev .
* d54 * ( 1996 ) 6620 . the e-143 collaboration ( k. abe et al . ) , phys .
* 74 * ( 1995 ) 346 ; + phys .
* d58 * ( 1998 ) 112003 . the e-154 collaboration ( k. abe et al . ) , phys .
* 79 * ( 1997 ) 26 .
g. altarelli , phys .
* 81 * ( 1982 ) 1 .
thomas , adv .
* 13 * ( 1984 ) 1 .
u. vogl and w. weise , prog .
* 27 * ( 1991 ) 195 .
klevansky , rev .
* 64 * ( 1992 ) 649 .
callan , r.f .
dashen and d.j .
gross , phys .
b63 * ( 1976 ) 334 ; + r. jackiw and c. rebbi , phys .
* 37 * ( 1976 ) 172 .
crewther , effects of topological charge in gauge theories , in facts and prospects of gauge theories , schladming , austria , february 1978 , ed .
p. urban , acta physica austriaca suppl .
* 19 * ( 1978 ) 47 .
y. nambu and g. jona - lasinio , phys . rev .
* 122 * ( 1961 ) 345 .
adler , phys . rev . * 177 * ( 1969 ) 2426 .
bell and r. jackiw , nuovo cimento * 60a * ( 1969 ) 47 .
bass , mod .
a13 * ( 1998 ) 791 .
a.w . schreiber and a.w .
thomas , phys . lett .
* b215 * ( 1988 ) 141 .
k. steininger and w. weise , phys .
* d48 * ( 1993 ) 1433 ; + k. suzuki and w. weise , nucl . phys . *
a634 * ( 1998 ) 141 .
efremov and o.v .
teryaev , jinr report e288287 ( 1988 ) , and in proceedings of the international hadron symposium , bechyn 1988 , eds .
j. fischer et al .
( czechoslovakian academy of science , prague , 1989 ) p. 302 ; + a.v .
efremov , j. soffer and o.v .
teryaev , nucl . phys .
* 346 * ( 1990 ) 97 .
g. altarelli and g.g .
ross , phys .
* b212 * ( 1988 ) 391 .
carlitz , j.c .
collins , and a.h .
mueller , phys . lett .
* b214 * ( 1988 ) 229 .
jaffe and a. manohar , nucl .
b337 * ( 1990 ) 509 .
g. veneziano , mod .
lett . * a4 * ( 1989 ) 1605 ; + g.m .
shore and g. veneziano , nucl .
* b381 * ( 1992 ) 23 .
s. narison , g.m .
shore and g. veneziano , nucl . phys .
* b433 * ( 1995 ) 209 .
h. fritzsch , phys .
b229 * ( 1989 ) 122 , _ ibid _
* b256 * ( 1991 ) 75 .
s. forte , phys . lett .
* b224 * ( 1989 ) 189 ; nucl .
b331 * ( 1990 ) 1 .
s. forte and e.v .
shuryak , nucl .
b357 * ( 1991 ) 153 .
ioffe and m. karliner , phys . lett .
* b247 * ( 1990 ) 387 . g. altarelli and g. ridolfi , nucl .
b ( proc . suppl . )
* 39b , c * ( 1995 ) 106 .
cheng , int .
a11 * ( 1996 ) 5109 .
m. anselmino , a. efremov and e. leader , phys . rept . * 261 * ( 1995 ) 1 .
jaffe , hep - ph/9602236 , _ in _ proc .
erice school _ the spin structure of the nucleon _ , eds . b. frois and v. hughes ( world scientific , 1997 ) .
j. ellis and m. karliner , hep - ph/9601280 , _ in _ proc .
erice school _ the spin structure of the nucleon _ , eds .
b. frois and v. hughes ( world scientific , 1997 ) .
b. lampe and e. reya , hep - ph/9810270 .
g.m . shore , zuoz lecture hep - ph/9812354 and erice lecture hep - ph/9812355 .
s.d . bass , mod .
* a12 * ( 1997 ) 1051 .
roberts , `` the structure of the proton '' ( cambridge up , 1990 ) a.m.
cooper - sarker , r.c.e .
devenish and a. de roeck , int j. mod .
* a13 * ( 1998 ) 3385 . the e-143 collaboration ( k. abe et al . ) , phys . rev
* 76 * ( 1996 ) 587 ; + the e-154 collaboration ( k. abe et al . ) , phys .
b404 * ( 1997 ) 377 ; + the smc collaboration ( b. adeva et al . ) , phys . lett . * b336 * ( 1994 125
jaffe , comm .
nucl . part .
* 19 * ( 1990 ) 239 .
bjorken , phys . rev . * 148 * ( 1966 ) 1467 ; phys . rev . * d1 * ( 1970 ) 1376 .
j. ellis and r.l .
jaffe , phys . rev .
* d9 * ( 1974 ) 1444 ; ( e ) * d10 * ( 1974 ) 1669 . j. kodaira , nucl . phys .
* b165 * ( 1980 ) 129 .
larin , phys .
* b334 * ( 1994 ) 192 ; _ ibid _ * b404 * ( 1997 ) 153 .
the particle data group ( c. caso et al . ) , euro .
phys . j. * c3 * ( 1998 ) 1 .
close and r.g .
roberts , phys .
b316 * ( 1993 ) 165 .
p. minkowski , _ in _ proc .
workshop on _ effective field theories of the standard model _ ,
dobogk , hungary 1991 , ed .
meissner ( world scientific , singapore , 1992 ) .
r. kberle and n.k .
nielsen , phys . rev .
* d8 * ( 1973 ) 660 ; + r.j .
crewther , s .- s .
shei and t .- m .
yan , phys .
* d8 * ( 1973 ) 3396 ; + r.j .
crewther and n.k .
nielsen , nucl .
b87 * ( 1975 ) 52 .
kaplan and a.v .
manohar , nucl .
b310 * ( 1988 ) 527 .
i thank r.j .
crewther for emphasising this point .
s.j . brodsky and f. schlumpf ,
* b329 * ( 1994 ) 111 .
w. koepf , e.m .
henley and s.j .
pollock , phys . lett .
* b288 * ( 1992 ) 11 .
m. glck , e. reya and a. vogt , z physik * c67 * ( 1995 ) 433 .
jaffe and g.g .
ross , phys .
b93 * ( 1980 ) 313 ; + a.w .
schreiber , a.i .
signal and a.w .
thomas , phys .
rev * d44 * ( 1991 ) 2653 .
steffens , h. holtmann and a.w .
thomas , phys . lett.*b358 * ( 1995 ) 139 .
gribov , physica scripta * t15 * ( 1987 ) 164 ; + lund preprint lu - tp-91 - 7 ( 1991 ) unpublished .
garvey , w.c .
louis and d.h .
white , phys .
* c48 * ( 1993 ) 761 .
campbell , j. ellis and r.a .
flores , phys .
* b225 * ( 1989 ) 419 .
d. bruss , t. gasenzer and o. nachtmann , phys .
a239 * ( 1998 ) 81 ; hep - ph/9802317 . g. altarelli and b. lampe , z. phys .
* c47 * ( 1990 ) 315 ; + g. altarelli , proc .
physics at hera , vol . 1 ( 1991 ) 379 , eds
. w. buchmller and g. ingelman .
g.c . fox and d.z .
freedman , phys .
* 182 * ( 1969 ) 1628 ; + d. broadhurst , j. gunion and r.l .
jaffe , phys .
* d8 * ( 1973 ) 566 ; ann . phys .
* 81 * ( 1973 ) 88 .
adler and r.f .
dashen , _ current algebras and applications to particle physics _
( w.a . benjamin , 1968 ) .
h. fritzsch and p. minkowski , nuovo cim . * 30a * ( 1975 ) 393 .
g. christos , phys .
* 116 * ( 1984 ) 251 .
e. witten , nucl .
* b156 * ( 1979 ) 269 ; + g. veneziano , nucl .
* b159 * ( 1979 ) 213 .
g. t hooft , phys .
* 37 * ( 1976 ) 8 ; phys . rev . *
d14 * ( 1976 ) 3432 .
g. t hooft , phys . rept .
* 142 * ( 1986 ) 357 .
p. minkowski , phys .
b237 * ( 1990 ) 531 ; * b423 * ( 1998 ) 157 .
brodsky , j. ellis and m. karliner , phys .
b206 * ( 1988 ) 309 .
shore and g. veneziano , nucl .
* b516 * ( 1998 ) 333 .
d. de florian , g.m .
shore and g. veneziano , hep - ph/9711353 , _ in _ proc .
workshop on _ physics with polarized protons at hera _ , eds .
a. de roeck and t. gehrmann , desy - proceedings-1998 - 01 a. de roeck and t. gehrmann , hep - ph/9711512 , _ in _ proc .
workshop on _ physics with polarized protons at hera _ ,
a. de roeck and t. gehrmann , desy - proceedings-1998 - 01 the phoenics collaboration ( a. bock et al ) , phys .
* 81 * ( 1998 ) 534 .
the taps collaboration ( b. krusche et al . ) , phys .
* 74 * ( 1995 ) 3736 .
cebaf experiments e-89 - 039 ( s. dytman et al . ) and e-91 - 008 ( b.g .
ritchie et al . ) .
s. tapprogge , talk at the workshop _ pomeron and odderon in theory and experiment _
( heidelberg , march 1998 ) ; + http://www.tphys.uni-heidelberg.de/ws/ celsius ( h. calen et al . ) , phys . rev . lett .
* 80 * ( 1998 ) 2069 ; + the wasa collaboration ( r. bilger et al . ) , nucl . phys . * a626 * ( 1997 ) 93c . the cosy-11 collaboration ( p. moskal et al . ) , phys .
* 80 * ( 1998 ) 3202 . the wa102 collaboration ( d. barberis et al . ) , phys . lett .
* b427 * ( 1998 ) 398 . the cleo collaboration ( j. gronberg et al . ) , phys . rev .
* d57 * ( 1998 ) 33 . the cleo collaboration ( b.h .
behrens et al . ) , phys .
lett . * 80 * ( 1998 ) 3710 ; + the cleo collaboration ( t.e .
browder et al . ) , phys .
* 81 * ( 1998 ) 1786 ; + j.g .
smith , hep - ex/9803028 .
h. fritzsch , phys . lett .
* b415 * ( 1997 ) 83 .
jaffe , phys .
b365 * ( 1996 ) 359 .
manohar , phys .
* 65 * ( 1990 ) 2511 .
j. kogut and l. susskind , phys . rev .
* d11 * ( 1974 ) 3594 .
c. cronstrm and j. mickelsson , j. math .
* 24 * ( 1983 ) 2528 .
adler , in brandeis _
lectures on elementary particles and quantum field theory _ , eds . s. deser , m. grisaru and h. pendleton ( mit press , 1970 ) .
crewther , riv .
nuovo cimento * 2 * ( 1979 ) 63 , section 7 .
j. bardeen , l. cooper and j. schrieffer , phys . rev . * 108 * ( 1957 ) 1175 .
anderson and w.f .
brinkman , theory of anisotropic superfluidity in he@xmath285 , _ in _ the helium liquids : proc .
15@xmath481 scottish universities summer school in physics 1974 , eds .
j.g.m . armitage and i.e. farquhar ( academic press , new york , 1975 ) .
s. grebenev , j.p .
toennies and a.f .
vilesov , science * 279 * ( 1998 ) 2083 .
shifman , a.i .
vainshtein and v.i .
zakharov , nucl .
b147 * ( 1979 ) 385 , 448 ; + e.v .
shuryak , phys . rept .
* 115 * ( 1984 ) 151 .
bass , b.l .
ioffe , n.n .
nikolaev and a.w .
thomas , j. moscow phys . soc .
* 1 * ( 1991 ) 317 .
l. mankiewicz and a. schfer , phys .
b242 * ( 1990 ) 455 ; + l. mankiewicz , phys . rev .
* d43 * ( 1991 ) 64 .
bodwin and j. qiu , phys . rev .
* d41 * ( 1990 ) 2755 .
manohar , phys .
* 66 * ( 1991 ) 289 .
p. ratcliffe , nucl .
b223 * ( 1983 ) 45 .
bass , n.n .
nikolaev and a.w .
thomas , adelaide university preprint adp-133-t80 ( 1990 ) unpublished ; + s.d .
bass , ph.d .
thesis ( university of adelaide , 1992 ) .
s.d . bass , z physik * c55 * ( 1992 ) 653 .
cheng , phys .
b427 * ( 1998 ) 371 .
m. stratmann , hep - ph/9710379 _ in _ proc .
workshop on _ deep inelastic scattering off polarized targets : theory meets experiment _ , desy - zeuthen 1997 , eds .
j. blmlein et al .
( desy report 97 - 200 , 1997 ) . m. gluck , e. reya , m. stratmann and w. vogelsang , phys . rev .
* d53 * ( 1996 ) 4775 .
j. ellis , m. karliner and c.t .
sachrajda , phys . lett .
* b231 * ( 1989 ) 497 .
lepage and s.j .
brodsky , phys .
* d22 * ( 1980 ) 2157 .
bass , s.j .
brodsky and i. schmidt , phys .
* b437 * ( 1998 ) 417 .
g. altarelli and g. parisi , nucl . phys . *
b126 * ( 1977 ) 298 .
r. mertig and w.l .
van neervan , z phys .
* c70 * ( 1996 ) 637 ; + e.b .
zijlstra and w.l .
van neervan , nucl . phys . *
b417 * ( 1994 ) 61 ; ( e ) * b426 * ( 1994 ) 245 .
w. vogelsang , phys . rev . * d54 * ( 1996 ) 2023
. e. leader , a.v .
sidorov and d.b .
stamenov , phys .
b445 * ( 1998 ) 232 . t. gehrmann and w.j . stirling , phys . rev .
* d53 * ( 1996 ) 6100 .
g. altarelli , r.d .
ball , s. forte and g. ridolfi , nucl .
b496 * ( 1997 ) 337 .
d. de florian , o.a .
samapayo and r. sassot , phys .
rev * d57 * ( 1998 ) 5803 .
gordon , m. goshtasbpour and g.p .
ramsey , phys . rev .
* d58 * ( 1998 ) 094017 . the e-154 collaboration ( k. abe et al . ) , phys . lett .
* b405 * ( 1997 ) 180 .
g. t hooft and m. veltman , nucl . phys .
* b44 * ( 1972 ) 189 .
ball , s. forte and g. ridolfi , phys .
b378 * ( 1996 ) 255 .
llewellyn smith , hep - ph/9812301 .
the compass proposal , cern / spslc 96 - 14 .
the hermes charm upgrade program , hermes 97 - 004 .
p. bosted , private communication
. n. hayashi , y. goto and n. saito , hep - ex/9807033 .
workshop on _ physics with polarized protons at hera _ ,
a. de roeck and t. gehrmann , desy - proceedings-1998 - 01 l.l .
frankfurt et al . , phys lett . *
b230 * ( 1989 ) 141 .
f.e . close and r.g .
milner , phys .
* d44 * ( 1991 ) 3691 . the smc collaboration ( b. adeva et al ) , phys
b420 * ( 1998 ) 180 .
w. wislicki , mod .
* a13 * ( 1988 ) 405 .
j. kuti and v. weisskopf , phys . rev .
* d4 * ( 1971 ) 3418 .
f.e . close , nucl .
b80 * ( 1974 ) 269 .
r. carlitz and j. kaur , phys .
lett . * 38 * ( 1977 ) 673 .
h. weigel , l. gamberg and h. reinhardt , phys .
* d55 * ( 1997 ) 6910 ; + o. schroder , h. reinhardt and h. weigel , phys .
b439 * ( 1998 ) 398 .
brodsky , m. burkardt and i. schmidt , nucl .
* b441 * ( 1995 ) 197 . c. young _ in _ proc .
workshop on _ deep inelastic scattering off polarized targets : theory meets experiment _ , desy - zeuthen 1997 , eds . j. blmlein et al .
( desy report 97 - 200 , 1997 ) . j. soffer and o.v .
teryaev , phys.rev . *
d56 * ( 1997 ) 1549 .
s.d . bass and m.m .
brisudov , hep - ph/9711423 , euro .
j * a * ( in press ) . r.l .
heimann , nucl .
b64 * ( 1973 ) 429 .
j. ellis and m. karliner , phys .
* b213 * ( 1988 ) 73 .
landshoff , proc .
zuoz summer school , psi proceedings 94 - 01 ( 1994 ) 135 , hep - ph/9410250 .
the new muon collaboration ( m. arneodo et al . ) , nucl .
b483 * ( 1997 ) 3 .
b. badeek and j. kwieciski , phys . lett .
* b418 * ( 1998 ) 229 .
various contributions _ in _ proc .
workshop on _ deep inelastic scattering off polarized targets : theory meets experiment _ , desy - zeuthen 1997 , eds .
j. blmlein et al .
( desy report 97 - 200 , 1997 ) .
l. galfi , j. kuti and a. patkos , phys . lett .
* b31 * ( 1970 ) 465 ; + j. kuti , erice lectures ( 1995 ) , _ in _ proc .
erice school _ the spin structure of the nucleon _ , eds . b. frois and v. hughes ( world scientific , 1997 ) .
f.e . close and r.g .
roberts , phys .
lett . * 60 * ( 1988 ) 1471 ; phys . lett .
* b336 * ( 1994 ) 257 .
s.d . bass and p.v .
landshoff , phys . lett .
* b336 * ( 1994 ) 537 .
r. kirschner and l.n .
lipatov , nucl . phys .
* b213 * ( 1983 ) 122 .
j. bartels , b.i .
ermolaev and m.g .
ryskin , z phys * c70 * ( 1996 ) 273 ; * c72 * ( 1996 ) 627 . j. blmlein and a. vogt , phys . lett . *
b386 * ( 1996 ) 350 .
k. gottfried , phys .
* 18 * ( 1967 ) 1174 . the new muon collaboration ( m. arneodo et al . ) , phys . rev .
* d50 * ( 1994 ) r1 . the e866/nusea collaboration ( e.a .
hawker et al . ) , phys .
lett . * 80 * ( 1998 ) 3715 .
thomas , phys . lett .
* b126 * ( 1983 ) 97 .
r.d . field and r.p .
feynman , phys .
rev . * d15 * ( 1977 ) 2590 .
a. w. thomas and w. melnitchouk , in _ new frontiers in nuclear physics _ , eds . s. homma , y. akaishi and m. wada ( world scientific , singapore , 1993 ) , pp . | we review the theory and present status of the proton spin problem with emphasis on the transition between current quarks and constituent quarks in qcd .
150 mm 239 mm -11 mm 6 mm 0 cm [ cols= " < " , ] 12 mm * constituent quarks and @xmath0 * + * steven d. bass * + _ max planck institut fr kernphysik , postfach 103 980 , + d-69029 heidelberg , germany _ + _
institut fr theoretische physik , + physik department , technische universitt mnchen , + d-85747 garching , germany _ + 10 mm = 6truemm |
the _ consecutive 1 s problem _ ( ) @xcite is defined as follows . given a binary matrix @xmath0 the goal is to permute its rows in such a way that the resulting matrix enjoys the _ consecutive 1 s property _ : each of its columns is a vector @xmath1 where @xmath2 if and only if @xmath3 for two integers @xmath4 between @xmath5 and @xmath6 .
this problem has its roots in archeology and especially _ sequence dating _ where the goal is to recover the chronological order of sepultures based on artifacts found in these sepultures where the entry @xmath7 of matrix @xmath0 indicates the presence of artifact @xmath8 in sepulture @xmath9 . in his seminal work , egyptologist flinders petrie @xcite formulated the hypothesis that two sepultures should be close in the time domain if they present similar sets of artifacts .
already in the noiseless case , this problem presents an interesting algorithmic challenge and is reducible to the famous travelling salesman problem @xcite as observed by statistician david kendall @xcite who employed early tools from multidimensional scaling as a heuristic to solve it
. belongs to a more general class of so - called _ seriation _ problems that consist in optimizing various criteria over the discrete set of permutations .
while such problems are hard in general , it can be shown that a subset of the these problems , including , can be solve efficiently using spectral method @xcite or convex optimization @xcite .
however , little is known about the robustness to noise of such methods . in order to set the benchmark for the noisy case
, we propose a statistical _ seriation model _ and study optimal rates of estimation in this model .
assume that we observe an @xmath10 matrix @xmath11 , where @xmath12 is an unknown @xmath13 permutation matrix , @xmath14 is an @xmath10 noise matrix and @xmath15 is assumed to belong to a class of matrices that satisfy a certain shape constraint .
our goal is to give estimators @xmath16 and @xmath17 so that @xmath18 is close to @xmath19
. the shape constraint can be the consecutive 1 s property , but more generally , we consider the class of matrices that have unimodal columns , which also include monotonic columns as a special case .
these terms will be formally defined at the end of this section .
the rest of the paper is organized as follows . in section [ sec : problem ] we formulate the model and discuss related work .
section [ sec : result ] collects our main results , including uniform and adaptive upper bounds for the least squares estimator together with corresponding minimax lower bounds in the general unimodal case . in section [ sec : rankscore ] , for the special case of monotone columns , we propose a computationally efficient alternative to the least squares estimator and study its rates of convergence both theoretically and numerically .
appendix [ sec : upper ] is devoted to the proofs of the upper bounds , which use the metric entropy bounds proved in appendix [ sec : covering ] .
the proofs of the information - theoretic lower bounds are presented in appendix [ sec : lower ] . in appendix
[ sec : monotone ] ,
we study the rate of estimation of the efficient estimator for the monotonic case . appendix [ sec : trivial - upper ] contains a delayed proof of a trivial upper bound .
appendix [ sec : unimodal ] presents new bounds for unimodal regression implied by our analysis , which are minimax optimal up to logarithmic factors . notation . for a positive integer @xmath6 ,
define @xmath20=\{1,\ldots , n\}$ ] . for a matrix @xmath15 ,
let @xmath21 denote its frobenius norm , and let @xmath22 be its @xmath9-th row and @xmath23 be its @xmath8-th column .
let @xmath24 denote the euclidean ball of radius @xmath25 centered at @xmath26 in @xmath27 .
we use @xmath28 and @xmath29 to denote positive constants that may change from line to line . for any two sequences @xmath30 and @xmath31 , we write @xmath32 if there exists an absolute constant @xmath33 such that @xmath34 for all @xmath6 .
we define @xmath35 analogously . given two real numbers @xmath36 ,
define @xmath37 and @xmath38 .
denote the closed convex cone of increasing sequences in @xmath27 by @xmath39 .
we define @xmath40 to be the cartesian product of @xmath41 copies of @xmath42 and we identify @xmath40 to the set of @xmath10 matrices with increasing columns . for any @xmath43
$ ] , define the closed convex cone @xmath44 , which consists of vectors in @xmath27 that increase up to the @xmath45-th entry and then decrease .
define the set @xmath46 of unimodal sequences in @xmath27 by @xmath47 .
we define @xmath48 to be the cartesian product of @xmath41 copies of @xmath46 and we identify @xmath48 to the set of @xmath10 matrices with unimodal columns .
it is also convenient to write @xmath48 as a union of closed convex cones as follows . for @xmath49^m$ ] , let @xmath50 .
then @xmath48 is the union of the @xmath51 closed convex cones @xmath52^m$ ] .
finally , let @xmath53 be the set of @xmath13 permutation matrices and define @xmath54 where @xmath55 , so that @xmath56 is the union of the @xmath57 closed convex cones @xmath58^m$ ] .
in this section , we formally state the problem of interest and discuss several lines of related work .
suppose that we observe a matrix @xmath59 , @xmath60 such that @xmath61 where @xmath62 , @xmath63 and @xmath14 is a centered sub - gaussian noise matrix with variance proxy @xmath64 .
more specifically , @xmath14 is a matrix such that @xmath65=0 $ ] and , for any @xmath66 , @xmath67 \le
\exp\big(\frac{\sigma^2\|m\|_f^2}{2}\big)\ , , \ ] ] where @xmath68 is the trace operator . we write @xmath69 or simply @xmath70 when dimensions are clear from the context .
given the observation @xmath71 , our goal is to estimate the unknown pair @xmath72 .
the performance of an estimator @xmath73 , is measured by the quadratic loss : @xmath74 in particular , its expectation is the mean squared error . since we are interested in estimating @xmath75 , we can also view @xmath76 as the parameter space . in the general unimodal case , upper bounds on the above quadratic loss
do not imply individual upper bounds on estimation of the matrix @xmath77 or the matrix @xmath78 due to lack of identifiability . nevertheless , if we further assume that the columns of @xmath78 are monotone increasing , that is @xmath79 , then the following lemma holds .
[ lem : rearrange ] if @xmath80 , then for any @xmath81 , we have that @xmath82 and that @xmath83 let @xmath84 and @xmath85 where @xmath86\to [ n]$ ] is a permutation .
it is easy to check that @xmath87 , so @xmath88 . applying this inequality to columns of matrices
, we see that @xmath89 since @xmath80 .
moreover , @xmath90 , so @xmath91 by the triangle inequality and the previous display .
lemma [ lem : rearrange ] guarantees that @xmath92 is a pertinent measure of the performance of @xmath93 .
note further that @xmath92 is large if @xmath93 misplaces rows of @xmath78 that have large differences , and is small if @xmath93 only misplaces rows of @xmath78 that are close to each other .
we argue that , in the seriation context , this measure of distance between permutations is more natural than ad hoc choices such as the trivial 0/1 distance or popular choices such as kendall s @xmath94 or spearman s @xmath95 . apart from section [ sec : rankscore ] ( and appendix [ sec : monotone ] ) , the rest of this paper focuses on the least squares ( ls ) estimator defined by @xmath96 taking @xmath97 , we see that it is equivalent to define the ls estimator by @xmath98 note that in our case , the set of parameters @xmath76 is not convex , but is a union of @xmath57 closed convex cones and it is not clear how to compute the ls estimator efficiently .
we discuss this aspect in further details in the context of monotone columns in section [ sec : rankscore ] .
nevertheless , the main focus of this paper is the least squares estimator which , as we shall see , is near - optimal in a minimax sense and therefore serves as a benchmark for the statistical seriation model .
our work falls broadly in the scope of statistical inference under shape constraints but presents a major twist : the unknown latent permutation @xmath77 .
to set our goals , we first consider the case where the permutation is known and assume without loss of generality that @xmath99 . in this case
, we can estimate individually each column @xmath100 by an estimator @xmath101 and then get an estimator @xmath17 for the whole matrix by concatenating the columns @xmath101 .
thus the task is reduced to estimation of a vector @xmath102 which satisfies a certain shape constraint from an observation @xmath103 where @xmath104 . when @xmath102 is assumed to be increasing we speak of isotonic regression @xcite .
the ls estimator defined by @xmath105 can be computed in closed form in @xmath106 using the pool - adjacent - violators algorithm ( pava ) @xcite and its statistical performance has been studied by zhang @xcite ( see also @xcite for similar bounds using empirical process theory ) who showed in the gaussian case @xmath107 that the mean squared error behaves like @xmath108 where @xmath109 } \theta_i - \min_{i \in [ n ] } \theta_i$ ] is the variation of @xmath110 .
note that @xmath111 for @xmath112 so that this is the minimax rate of estimation of lipschitz functions ( see , e.g. , @xcite ) .
the rate in is said to be _ global _ has it holds uniformly over the set of monotone vectors with variation @xmath113 .
recently , @xcite have initiated the study of _ adaptive _ bounds that may be better if @xmath102 has a simpler structure in some sense . to define this structure ,
let @xmath114 denote the cardinality of entries of @xmath110 .
in this context , @xcite showed that the ls estimator satisfies the adaptive bound @xmath115 this result was extended in @xcite to a sharp oracle inequality where @xmath116 .
this bound was also shown to be optimal in a minimax sense @xcite .
unlike its monotone counterpart , unimodal regression where @xmath117 has received sporadic attention @xcite .
this state of affairs is all the more surprising given that unimodal density estimation has been the subject of much more research @xcite .
it was recently shown in @xcite that the ls estimator also adapts to @xmath113 and @xmath118 for unimodal regression : @xmath119 with probability at least @xmath120 for some @xmath121 .
the exponent @xmath122 in the second term was improved to @xmath5 in the new version of @xcite after the first version of our current paper was posted .
note that the exponents in are different from the isotonic case .
our results will imply that they are not optimal and in fact the ls estimator achieves the same rate as in isotonic regression .
see corollary [ cor : unimodal ] for more details .
the algorithmic aspect of unimodal regression has received more attention @xcite and @xcite showed that the ls estimator can be computed with time complexity @xmath106 using a modified version of pava . hence there is little difference between isotonic and unimodal regressions from both computational and statistical points of views
. when the permutation @xmath77 is unknown the estimation problem is more involved .
noisy permutation learning was explicitly addressed in @xcite where the problem of matching two sets of noisy vectors was studied from a statistical point of view . given @xmath123 matrices @xmath124 and @xmath125 , where @xmath126 is an unknown matrix and @xmath127 is an unknown permutation matrix , the goal is to recover @xmath77 .
it was shown in @xcite that if @xmath128 , then the ls estimator defined by @xmath129 recovers the true permutation with high probability .
however they did not directly study the behavior of @xmath130 . in his celebrated paper on matrix estimation @xcite
, sourav chatterjee describes several noisy matrix models involving unknown latent permutations .
one is the _ nonparametric bradley - terry - luce _ ( np - btl ) model where we observe a matrix @xmath131 with independent entries @xmath132 for some unknown parameters @xmath133 where @xmath134 $ ] is equal to the probability that item @xmath9 is preferred over item @xmath8 and @xmath135 .
crucially , the np - btl model assumes the so - called _ strong stochastic transitivity ( sst ) _
@xcite assumption : there exists an unknown permutation matrix @xmath136 such that the ordered matrix @xmath137 satisfies @xmath138 for all @xmath139 $ ] . note that the np - btl model is a special case of our model where @xmath140 and @xmath141 is taken to be bernoulli .
chatterjee proposed an estimator @xmath142 that leverages the fact that any matrix @xmath143 in the np - btl model can be approximated by a low rank matrix and proved ( * ? ? ?
* theorem 2.11 ) that @xmath144 , which was improved to @xmath145 by @xcite for a variation of the estimator .
this method does not yield individual estimators of @xmath12 or @xmath0 , and @xcite proposed estimators @xmath16 and @xmath17 so that @xmath146 estimates @xmath143 with the same rate @xmath145 up to a logarithmic factor .
the non - optimality of this rate has been observed in @xcite who showed that the correct rate should be of order @xmath147 up to a possible @xmath148 factor .
however , it is not known whether a computationally efficient estimator could achieve the fast rate .
a recent work @xcite explored a new notion of adaptivity for which the authors proved a computational lower bound , and also proposed an efficient estimator whose rate of estimation matches that lower bound .
also mentioned in chatterjee s paper is the so - called _ stochastic block model _ that has since received such extensive attention in various communities that it is futile to attempt to establish a comprehensive list of references .
instead , we refer the reader to @xcite and references therein .
this paper establishes the minimax rates for this problem and its continuous limit , the graphon estimation problem and , as such , constitutes the state - of - the - art in the statistical literature . in the stochastic block model with @xmath149 blocks
, we assume that we observe a matrix @xmath150 where @xmath151 is an unknown permutation matrix and @xmath0 has a block structure , namely , there exist positive integers @xmath152 , and @xmath153 real numbers @xmath154 ^ 2 $ ] such that @xmath0 has entries @xmath155 ^ 2}a_{s , t } { { \rm 1}\kern-0.24em{\rm i}}\{n_s\le i \le n_{s+1 } , n_t\le j \le n_{t+1}\}\ , , \qquad i , j \in [ n]\,.\ ] ] while traditionally , the stochastic block model is a network model and therefore pertains only to bernoulli observations , the more general case of sub - gaussian additive error is also explicitly handled in @xcite . for this problem ,
gao , liu and zhou have established that the least squares estimator @xmath142 satisfies @xmath156 together with a matching lower bound .
using piecewise constant approximation to bivariate hlder functions , they also establish that this estimator with a correct choice of @xmath157 leads to minimax optimal estimation of smooth graphons .
both results exploit extensively the fact that the matrix @xmath143 is equal to or can be well approximated by a piecewise constant matrix and our results below take a similar route by observing that monotone and unimodal vectors are also well approximated by piecewise constant ones .
moreover , we allow for rectangular matrices . in fact , our result can be also formulated as a network estimation problem but on a bipartite graph , thus falling at the intersection of the above two examples .
assume that @xmath6 left nodes represent items and that @xmath41 right nodes represent users .
assume further that we observe the @xmath10 adjacency matrix @xmath71 of a random graph where the presence of edge @xmath158 indicates that user @xmath8 has purchased or liked item @xmath9 .
define @xmath159 $ ] and assume sst across items in the sense that there exists an unknown @xmath13 permutation matrix @xmath77 such that @xmath160 and @xmath78 is such that @xmath161 for all users @xmath162 $ ] .
this model falls into the scope of the statistical seriation model .
for a matrix @xmath163 , let @xmath164 be the number of values taken by the @xmath8-th column of @xmath0 and define @xmath165 .
observe that @xmath166 .
the first theorem shows that the ls estimator adapts to the complexity @xmath167 .
[ thm : adaptive ] for @xmath168 and @xmath169 , let @xmath170 be the ls estimator defined in
. then the following oracle inequality holds @xmath171 with probability at least @xmath172 .
moreover , @xmath173 note that while we assume that @xmath62 in , the above oracle inequalities hold in fact for any @xmath168 even if its columns are _ not _ assumed to be unimodal .
the above oracle inequalities indicate that the ls estimator automatically trades off the approximation error @xmath174 for the stochastic error @xmath175 .
if @xmath78 is assumed to have unimodal columns , then we can take @xmath176 in and to get the following corollary .
[ cor : adaptive ] for @xmath62 and @xmath169 , the ls estimator @xmath170 satisfies @xmath177 with probability at least @xmath178 .
moreover , the corresponding bound with the same rate holds in expectation .
the two terms in the adaptive bound can be understood as follows .
the first term corresponds to the estimation of the matrix @xmath78 with unimodal columuns if the permutation @xmath77 is known .
it can be viewed as a matrix version of the adaptive bound in the vector case .
the ls estimator adapts to the cardinality of entries of @xmath78 as it achieves a provably better rate if @xmath179 is smaller while not requiring knowledge of @xmath179 .
the second term corresponds to the error due to the unknown permutation @xmath77 .
as @xmath41 grows to infinity this second term vanishes , because we have more samples to estimate @xmath77 better .
if @xmath180 , it is easy to check that the permutation term is dominated by the first term , so the rate of estimation is the same as if the permutation is known . the bounds in theorem [ thm : adaptive ] adapt to the cardinality of the oracle
. in this subsection , we state another type of upper bounds for the ls estimator @xmath170
they are called global bounds because they hold uniformly over the class of matrices whose columns are unimodal and that have bounded variation .
recall that we call _ variation _ of a vector @xmath181 the scalar @xmath182 defined by @xmath183 we extend this notion to a matrix @xmath184 by defining @xmath185 while this @xmath186-norm may seem odd at first sight , it turns out to be the correct extrapolation from vectors to matrices , at least in the context under consideration here .
indeed , the following upper bound , in which this quantity naturally appears , is matched by the lower bound of theorem [ thm : lower - global ] up to logarithmic terms .
[ thm : global ] for @xmath168 and @xmath169 , let @xmath170 be the ls estimator defined in .
then it holds that @xmath187 + \sigma^2 \frac{\log n}{n\wedge m } \,.\end{gathered}\ ] ] with probability at least @xmath172 .
moreover , the corresponding bound with the same rate holds in expectation . if @xmath62 , then taking @xmath176 in theorem [ thm : global ] leads to the following corollary that indicates that the ls estimator is adaptive to the quantity @xmath188 .
[ cor : global ] for @xmath62 and @xmath169 , the ls estimator @xmath170 satisfies @xmath189 with probability at least @xmath172 .
moreover , the corresponding bound with the same rate holds in expectation .
akin to the adaptive bound , the above inequality can be viewed as a sum of a matrix version of and an error due to estimation of the unknown permutation .
having stated the main upper bounds , we digress a little to remark that the proofs of theorem [ thm : adaptive ] and theorem [ thm : global ] also yield a minimax optimal rate of estimation ( up to logarithmic factors ) for unimodal regression , which improves the bound . we discuss the details in appendix [ sec : unimodal ] .
given the model @xmath169 where entries of @xmath14 are i.i.d .
@xmath190 random variables , let @xmath170 denote any estimator of @xmath72 , i.e. , any pair in @xmath191 that is measurable with respect to the observation @xmath71
. we will prove lower bounds that match the rates of estimation in corollary [ cor : adaptive ] and corollary [ cor : global ] up to logarithmic factors .
the combination of upper and lower bounds , implies simultaneous near optimality of the least squares estimator over a large scale of matrix classes . for @xmath192 and @xmath193 ,
define @xmath194 and @xmath195 we present below two lower bounds , one for the adaptive rate uniformly over @xmath196 and one for the global rate uniformly over @xmath197 .
this splitting into two cases is solely justified by better readability but it is worth noting that a stronger lower bound that holds on the intersection @xmath198 can also be proved and is presented as proposition [ prop : lower - stronger ] .
[ thm : lower - adaptive ] there exists a constant @xmath199 such that for any @xmath200 , and any estimator @xmath170 , it holds that @xmath201 \ge c , \ ] ] where @xmath202 and @xmath203 is the probability distribution of @xmath204 .
it follows that the lower bound with the same rate holds in expectation .
in fact , the lower bound holds for any estimator of the matrix @xmath205 , not only those of the form @xmath18 with @xmath206 . the above lower bound matches the upper bound in corollary [ cor : adaptive ] up to logarithmic factors .
note the presence of a @xmath207 factor in the second term . if @xmath208 then @xmath209 which means that each column of @xmath0 is simply a constant block , so @xmath210 for any @xmath63 . in this case ,
the second term vanishes because the permutation does not play a role .
more generally , the number @xmath211 can be understood as the maximal number of columns of @xmath0 on which the permutation does have an effect .
the larger @xmath45 , the harder the estimation .
it is easy to check that if @xmath212 the second term in the lower bound will be dominated by the first term in the upper bound .
a lower bound corresponding to corollary [ cor : global ] also holds : [ thm : lower - global ] there exists a constant @xmath199 such that for any @xmath213 , and any estimator @xmath170 , it holds that @xmath214 \ge c \,,\ ] ] where @xmath203 is the probability distribution of @xmath204 .
the lower bound with the same rate also holds in expectation .
there is a slight mismatch between the upper bound of corollary [ cor : global ] and the lower bound of theorem [ thm : lower - global ] above .
indeed the lower bound features a term @xmath215 instead of just @xmath216 . in the regime @xmath217 , where @xmath0 has very small variation , the ls estimator may not be optimal .
proposition [ prop : special - regime ] indicates that a matrix with constant columns obtained by averaging achieves optimality in this extreme regime .
a particularly interesting subset of unimodal matrices is @xmath40 , the set of @xmath218 matrices with monotonically increasing columns .
while it does not amount to the seriation problem in its full generality , this special case is of prime importance in the context of shape constrained estimation as illustrated by the discussion and references in section [ sec : related - work ] .
in fact , it covers the example of bipartite ranking discussed at the end of section [ sec : related - work ] . in the rest of this section , we devote further investigation to this important case . to that end , consider the model where we further assume that @xmath219 .
we refer to this model as the _ monotone seriation model_. in this context , define the ls estimator by @xmath220 since @xmath40 is a convex subset of @xmath48 , it is easily seen that the upper bounds in theorem [ thm : adaptive ] and [ thm : global ] remain valid in this case .
the lower bounds of theorem [ thm : lower - adaptive ] ( with @xmath207 replaced by @xmath5 ) and theorem [ thm : lower - global ] also extend to this case ; see appendix [ sec : lower ] .
although for unimodal matrices the established error bounds do not imply any bounds on estimation of @xmath78 or @xmath77 in general , for the monotonic case , however , lemma [ lem : rearrange ] yields that @xmath221 so that the ls estimator @xmath170 also leads to good individual estimators of @xmath77 and @xmath78 respectively . because it requires optimizing over a union of @xmath222 cones @xmath223 , no efficient way of computing the ls estimator
is known since . as an alternative
, we describe a simple and efficient algorithm to estimate @xmath72 and study its rate of estimation .
let @xmath224 and @xmath225 be defined as before .
moreover , for a matrix @xmath226 , let @xmath227 denote the set of pairs of indices @xmath228 ^ 2 $ ] such that @xmath22 and @xmath229 are not identical .
define the quantity @xmath230 by @xmath231 ^ 2\\ |\mathcal i|
= n } } \sum_{(i , j ) \in \mathcal i \cap \mathcal j } \big ( \frac{\|a_{i,\cdot } - a_{j,\cdot}\|_2 ^ 2}{\|a_{i,\cdot } - a_{j,\cdot}\|_\infty^2}\wedge \frac{m \|a_{i,\cdot } - a_{j,\cdot}\|_2 ^ 2}{\|a_{i,\cdot } - a_{j,\cdot}\|_1 ^ 2 } \big ) \,.\ ] ] it can be shown ( see appendix [ sec : monotone ] ) that @xmath232 . intuitively , the quantity @xmath230 is small if the difference @xmath233 of any two rows of @xmath0 is either very sparse ( @xmath234 is small ) or very dense ( @xmath235 is small ) .
indeed , for any nonzero vector @xmath236 , @xmath237 with equality achieved when @xmath238 , and @xmath239 with equality achieved when all entries of @xmath233 are the same . for matrices with small @xmath240 values ,
it is possible to aggregate the information across each row to learn the unknown permutation @xmath77 in a simple fashion .
recovering the permutation @xmath77 , is equivalent to ordering ( or ranking reversely ) the rows of @xmath241 from their noisy version @xmath71 .
one simple method to achieve this goal , which we call , is to permute the rows of @xmath71 so that they have increasing row sums .
however , it is easy to observe that this method fails if @xmath242 where @xmath243 and entries of @xmath14 are i.i.d .
standard gaussian variables , because the sum of noise in a row has order @xmath244 which is no less than the gaps between row sums of @xmath78 .
in fact , @xmath245 and it should be easy to distinguish the two types of rows of @xmath78 , for example , by looking at the first entry of a row .
this motivates us to consider the following method called . for @xmath246 $ ] , define @xmath247 } ( a^*_{i ' , j } - a^*_{i , j})\vee \frac 1{\sqrt m } \sum_{j=1}^m ( a^*_{i ' , j } - a^*_{i , j})\ ] ] and define @xmath248 analogously . the procedure is defined as follows : 1 . for each @xmath249 $ ] , define the score @xmath250 of the @xmath9-th row of @xmath71 by @xmath251 where @xmath252 for some tuning constant @xmath28 ( see appendix [ sec : monotone ] for more details ) .
2 . then order the rows of @xmath71 so that their scores are increasing , with ties broken arbitrarily .
the procedure recovers an order of the rows of @xmath71 , which leads to an estimator @xmath93 of the permutation
. then we define @xmath253 so that @xmath254 is the projection of @xmath71 onto the convex cone @xmath255 .
the estimator @xmath256 enjoys the following rate of estimation .
[ thm : monotone ] for @xmath219 and @xmath169 , let @xmath256 be the estimator defined above using the procedure with threshold @xmath257 , @xmath33 . then it holds that @xmath258 with probability at least @xmath259 for some constant @xmath260 .
the quantity @xmath261 only depends on the matrix @xmath78 .
if @xmath261 is bounded logarithmically , the estimator @xmath256 achieves the minimax rate up to logarithmic factors . in any case , @xmath262 , so the estimator is still consistent with the permutation error ( the last term ) decaying at a rate no slower than @xmath263 .
furthermore , it is worth noting that @xmath261 is not needed to construct @xmath256 , so the estimator adapts to @xmath261 automatically . in the same way that theorem [ thm : global ] follows from theorem [ thm : adaptive ]
, we can deduce from theorem [ thm : monotone ] a global bound for the estimator @xmath256 which has rate @xmath264 we conclude this section with a numerical comparison between the and procedures .
consider the model with @xmath265 and assume without loss of generality that @xmath99 . for various @xmath10 matrices
@xmath78 , we generate observations @xmath124 where entries of @xmath14 are i.i.d .
standard gaussian variables .
the performance of the estimators given by and defined above is compared to the performance of the oracle @xmath266 defined by the projection of @xmath71 onto the cone @xmath40 . for the estimator
we take @xmath267 .
the curves are generated based on @xmath268 equally spaced points on the base-@xmath269 logarithmic scale , and all results are averaged over @xmath269 replications .
the vertical axis represents the estimation error of an estimator @xmath270 , measured by the sample mean of @xmath271 unless otherwise specified . of size @xmath272 .
left : rows of @xmath78 are @xmath5-sparse ; right : columns of @xmath78 are identical . ] of size @xmath272 . left : rows of @xmath78 are @xmath5-sparse ; right : columns of @xmath78 are identical . ]
we begin with two simple examples for which we set @xmath273 . in the left plot of figure [ fig : plot1 ] , @xmath78 is defined as in .
as expected , fails to estimate the true permutation and performs very poorly . on the other hand ,
succeeds in recovering the correct permutation and has roughly the same performance as the oracle . because the difference of any two rows of @xmath78 is @xmath5-sparse , @xmath245 according to and the discussion thereafter .
hence , theorem [ thm : monotone ] predicts the fast rate , which is verified by the experiment .
the right plot illustrates another extreme case ; more precisely , we set @xmath78 to be the matrix with all @xmath41 columns equal to @xmath274 .
the difference of any two rows of @xmath78 is constant across all entries , so again we have @xmath245 by .
thus achieves the fast rate as expected .
note that also performs well in this case . and randomly generated @xmath78 of size @xmath10 .
left : @xmath275 ; right : @xmath276 . ] and randomly generated @xmath78 of size @xmath10 .
left : @xmath275 ; right : @xmath276 . ] in figure [ fig : plot2 ] , we compare the performance of to that of the oracle in three regimes of @xmath277 .
the matrices @xmath78 are randomly generated for different values of @xmath6 and @xmath41 as follows .
for the right plot , @xmath78 is generated so that @xmath276 , by sorting the columns of a matrix with i.i.d .
@xmath278 entries . for the left plot
, we further require that @xmath275 by uniformly partitioning each column of @xmath78 into five blocks and assigning each block the corresponding value from a sorted sample of five i.i.d . @xmath278 variables .
since the oracle knows the true permutation , its behavior is independent of @xmath41 , and its rates of estimation are bounded by @xmath279 for @xmath280 and @xmath281 for @xmath282 respectively by theorem [ thm : adaptive ] and [ thm : global ] .
( the difference is minor in the plots as @xmath6 is not sufficiently large ) . for , the permutation term dominates the estimation term when @xmath283 by theorem [ thm : monotone ] . from the plots , the rates of estimation are better than @xmath284 predicted by the worst - case analysis in both examples . for @xmath140
, we also observe rates of estimation faster than the worst - case rate @xmath285 and close to the oracle rates .
we could explain this phenomenon by @xmath286 , but such an interpretation may not be optimal since our analysis is based on worst - case deterministic @xmath78 .
potential study of random designs of @xmath78 is left open .
finally , for @xmath287 , the permutation term is of order @xmath288 theoretically , in between of the oracle rates for the two cases .
indeed has almost the same performance as the oracle experimentally .
overall figure [ fig : plot2 ] illustrates the good behavior of in this random scenario . to conclude our numerical experiments , we consider the @xmath13 lower triangular matrix @xmath78 defined by @xmath289 . for this matrix ,
it is easy to check that @xmath290 and @xmath291 .
we plot in figure [ fig : plot3 ] the estimation errors of @xmath254 , @xmath292 and @xmath293 given by , in addition to the oracle . by theorem [ thm :
monotone ] , the rate of estimation achieved by @xmath254 is of order @xmath285 , while that achieved by the oracle is of order @xmath294 since there is no permutation term .
the plot confirms this discrepancy .
moreover , @xmath295 is an appropriate measure of the performance of @xmath93 by lemma [ lem : approx - est ] and [ lem : rearrange ] , and the plot suggests that the rates of estimation achieved by @xmath292 and @xmath296 are about the same order .
finally @xmath293 seems to have a slightly faster rate of estimation than @xmath254 , so in practice @xmath293 could be used to estimate @xmath0 .
however we refrain from making an explicit conjecture about the rate .
while computational aspects of the seriation problem have received significant attention , the robustness of this problem to noise was still unknown to date . to overcome this limitation ,
we have introduced in this paper the statistical seriation model and studied optimal rates of estimation by showing , in particular , that the least squares estimator enjoys several desirable statistical properties such as adaptivity and minimax optimality ( up to logarithmic terms ) .
while this work paints a fairly complete statistical picture of the statistical seriation model , it also leaves many unanswered questions .
there are several logarithmic gaps in the bounds . in the case of adaptive bounds ,
some logarithmic terms are unavoidable as illustrated by theorem [ thm : lower - adaptive ] ( for the permutation term ) and also by statistical dimension consideration explained in @xcite ( for the estimation term ) . however , a more refined argument for the uniform bound , namely one that uses covering in @xmath297-norm rather than @xmath298-norm , would allow us to remove the @xmath148 factor from the estimation term in the upper bound of corollary [ cor : global ] .
such an argument can be found in @xcite for the larger class of vectors with bounded total variation ( see @xcite ) but we do not pursue sharp logarithmic terms in this work . for the permutation term , @xmath148 in the upper bound of corollary [ cor :
adaptive ] and @xmath207 in the lower bound of theorem [ thm : lower - adaptive ] do not match if @xmath299 .
we do not seek answers to these questions in this paper but note that their answers may be different for the unimodal and the monotone case .
perhaps the most pressing question is that of computationally efficient estimators . indeed , while statistically optimal , the least squares estimator requires searching through @xmath222 permutations , which is not realistic even for problems of moderate size , let alone genomics applications .
we gave a partial answer to this question in the specific context of monotone columns by proposing and studying the performance of a simple and efficient estimator called .
this study reveals the existence of a potentially intrinsic gap between the statistical performance achievable by efficient estimators and that achievable by estimators with access to unbounded computation .
a similar gap is also observed in the sst model for pairwise comparisons @xcite .
we conjecture that achieving optimal rates of estimation in the seriation model is computationally hard in general but argue that the planted clique assumption that has been successfully used to establish statistical vs. computational gaps in @xcite for example , is not the correct primitive .
instead , one has to seek for a primitive where hardness comes from searching through permutations rather than subsets .
p.r . is partially supported by nsf grant dms-1317308 and nsf career award dms-1053987 .
n.f . gratefully acknowledges partial support by nsf grant dms-1317308 , the institute for data systems and society and the mathematics department during his visit at mit .
c.m . acknowledges partial support by nsf career award dms-1053987 . p.r .
also thanks enno mammen for his help in retracing the literature on sharp entropy bounds for monotone classes and ramon van handel for a stimulating discussion on entropy numbers .
the authors thank alexandre tsybakov for bringing concurrent work on unimodal regression by pierre c. bellec to their attention .
finally , the authors thank pierre c. bellec for pointing out an error in an earlier version of lemma a.1 and suggesting that it holds for all closed sets .
before proving the main theorems , we discuss two methods adopted in recent works to bound the error of the ls estimator in shape constrained regression , in a general setting .
consider the least squares estimator @xmath300 of the model @xmath301 , where @xmath102 lies in a parameter space @xmath302 and @xmath303 is gaussian noise .
one way to study @xmath304 is to use the _ statistical dimension _
@xcite of a convex cone @xmath302 defined by @xmath305 \,.\ ] ] this has been successfully applied to isotonic and more general shape constrained regression @xcite .
another prominent approach is to express the error of the ls estimator via what is known as _
chatterjee s variational formula _
, proved in @xcite and given by @xmath306 note that the first term is related to the _ gaussian width _ ( see , e.g. , @xcite ) of @xmath302 defined by @xmath307,$ ] whose connection to the statistical dimension was studied in @xcite .
the variational formula was first proposed for convex regression @xcite , and later exploited in several different settings , including matrix estimation with shape constraints @xcite and unimodal regression @xcite .
similar ideas have appeared in other works , for example , analysis of empirical risk minimization @xcite , ranking from pairwise comparison @xcite and isotonic regression @xcite . in this latter work
, bellec has used the statistical dimension approach to prove spectacularly sharp oracle inequalities that seem to be currently out of reach for methods based on chatterjee s variational formula . on the other hand ,
chatterjee s variational formula seems more flexible as computations of the statistical dimension based on @xcite are currently limited to convex sets @xmath302 with a polyhedral structure . in this paper , we use exclusively chatterjee s variational formula .
we begin the proof by stating an extension of chatterjee s variational formula . while we only need this lemma to hold for a union of closed convex sets we present a version that holds for all closed sets .
the latter extension was suggested to us by pierre c. bellec in a private communication @xcite .
[ lem : variational ] let @xmath308 be a closed subset of @xmath309 .
suppose that @xmath310 where @xmath311 and @xmath312 .
let @xmath313 be a projection of @xmath314 onto @xmath315 .
define the function @xmath316 by @xmath317 then we have @xmath318 moreover , if there exists @xmath319 such that @xmath320 for all @xmath321 , then @xmath322 by definition , @xmath323 together with the definition of @xmath324 , this implies that @xmath325 therefore follows . furthermore , suppose that there is @xmath319 such that @xmath320 for all @xmath321 .
since @xmath326 , we have @xmath327 .
note that this structural result holds for any error vector @xmath328 and any closed set @xmath315 which is not necessarily convex .
in particular , this extends the results in @xcite and @xcite which hold for convex sets and finite unions of convex sets respectively . for our purpose , we need a standard chaining bound on the supremum of a sub - gaussian process that holds in high probability
. the interested readers can find the proof , for example , in ( * ? ? ?
* theorem 5.29 ) , and refer to @xcite for a more detailed account of the technique .
[ lem : chaining - tail ] let @xmath329 and @xmath330 in @xmath309 .
for any @xmath331 , it holds that @xmath332 with probability at least @xmath333 where @xmath28 and @xmath29 are positive constants .
let @xmath334 . to ligthen the notation
, we define two rates of estimation : @xmath335 and @xmath336 note that @xmath337 .
[ lem : f - bound ] suppose @xmath338 where @xmath339 and @xmath340 . for @xmath334 and all @xmath341 , define @xmath342 then for any @xmath343 , it holds simultaneously for all @xmath341 that @xmath344 with probability at least @xmath345 , where @xmath28 and @xmath29 are positive constants .
define @xmath346 ( see also definition ) .
in particular , @xmath347 and @xmath348 . since @xmath76 is a finite union of convex cones and
thus is star - shaped , by scaling invariance , @xmath349 by lemma [ lem : chaining - tail ] , with probability at least @xmath345 , @xmath350 moreover , it follows from lemma [ lem : cover - matrix - union ] that @xmath351 combining the previous three displays , we see that @xmath352 with probability at least @xmath345 .
therefore @xmath353 with probability at least @xmath345 simultaneously for all @xmath341 .
we are now in a position to prove the adaptive oracle inequalities in theorem [ thm : adaptive ] .
recall that @xmath170 denotes the ls estimator defined in . without loss of generality , assume that @xmath354 and @xmath338 .
fix @xmath334 and define @xmath355 as in lemma [ lem : f - bound ] .
we can apply lemma [ lem : variational ] with @xmath356 , @xmath357 , @xmath358 and @xmath359 to achieve an error bound on @xmath360 since @xmath361 . to be more precise , for any @xmath343 we define @xmath362 where @xmath363 is the constant in .
then it follows from lemma [ lem : f - bound ] that with probability at least @xmath364 , it holds for all @xmath365 that @xmath366 therefore by lemma [ lem : variational ] , @xmath367 and thus @xmath368 with probability at least @xmath369 in particular , if @xmath370 , then @xmath371 as @xmath372 .
we see that with probability at least @xmath373 , @xmath374 and thus @xmath375 finally , follows by taking the infimum over @xmath334 on the right - hand side and dividing both sides by @xmath376 . next , to prove the bound in expectation , observe that yields @xmath377 \le c \exp(- \frac{c s}{\sigma^2}),\ ] ] where @xmath378 is defined in . integrating the tail probability
, we get that @xmath379 and therefore @xmath380 dividing both sides by @xmath376 and minimizing over @xmath334 yields . in the setting of isotonic regression , @xcite derived global bounds from adaptive bounds by a block approximation method , which also applies to our setting . for @xmath139 $ ] ,
let @xmath381 define @xmath382 .
the lemma below is very similar to ( * ? ? ?
* lemma 2 ) and their proof also extends to the unimodal case with minor modifications .
we present the result with proof for completeness .
[ lem : vector - approximation ] for @xmath383 and @xmath139 $ ] , there exists @xmath384 such that @xmath385 in particular , there exists @xmath386 such that @xmath387 moreover , @xmath388 let @xmath389 , @xmath390 } a_i$ ] and @xmath391 } a_i$ ] . for @xmath392
$ ] , consider the intervals @xmath393 , \ ] ] and @xmath394 $ ] . also for @xmath395 $ ] , let @xmath396 : a_i\in i_j\}$ ] .
we define the vector @xmath397 by @xmath398 for @xmath249 $ ] , where @xmath8 is uniquely determined by @xmath399 . since @xmath26 is increasing on @xmath400 and decreasing @xmath401 ,
so is @xmath402 .
thus @xmath403 .
moreover , @xmath404 for @xmath405 $ ] , which implies .
next we prove the latter two assertions .
since @xmath406 , if @xmath386 and @xmath407 then @xmath408 and @xmath409 on the other hand , if @xmath410 , then @xmath411 and @xmath412 it is straightforward to generalize the lemma to matrices . for @xmath413^m$ ] , we write @xmath414 and let @xmath415 then @xmath416 for @xmath417 . define @xmath418 by @xmath419 [ lem : block - approximation ] for @xmath163 , there exists @xmath420 such that @xmath421 and @xmath422 applying lemma [ lem : vector - approximation ] to columns of @xmath0 , we see that there exists @xmath420 such that @xmath423 and @xmath424 summing over @xmath425 , we get that @xmath426 and similarly @xmath427 for @xmath163 , choose @xmath420 according to lemma [ lem : block - approximation ]
. then @xmath428 by noting that @xmath429 for @xmath60 , and similarly @xmath430 plugging and into the right - hand side of and , and then minimizing over @xmath163 , we complete the proof .
in this section , we study various _ covering numbers _ or _ metric entropy _ related to the parameter space of the model . first recall some standard definitions that date back at least to @xcite .
an @xmath431-net of a subset @xmath432 with respect to a norm @xmath433 is a set @xmath434 such that for any @xmath435 , there exists @xmath436 $ ] for which @xmath437 .
the covering number @xmath438 is the cardinality of the smallest @xmath431-net with respect to the norm @xmath433 .
metric entropy is defined as the logarithm of a covering number . in the following
, we will consider the euclidean norm unless otherwise specified .
lemma [ lem : cover - product ] below bounds covering numbers of product spaces and is useful in later proofs .
we start with a well - known result on the covering number of a euclidean ball with respect to the @xmath298-norm ( see e.g. ( * ? ? ?
* lemma 7.14 ) for an analogous result ) .
[ lem : cube - cover - ball ] for any @xmath439 $ ] , @xmath440 for some constant @xmath33 .
we aim at bounding the covering number of a euclidean ball by cubes .
let @xmath441 be a maximal @xmath442-packing of @xmath443 with respect to the @xmath298-norm , where a @xmath444-packing of a set @xmath445 with respect to a norm @xmath433 is a set @xmath446 such that @xmath447 for all distinct @xmath448 $ ] .
then this set is necessarily an @xmath442-net of @xmath443 by maximality , so @xmath449 .
consider the cubes with side length @xmath442 centered at @xmath450 for @xmath451 .
these cubes are disjoint and contained in the set @xmath452 , where @xmath453 is the cube with side length @xmath442 centered at the origin in @xmath454 . since @xmath455 , @xmath456 this proves the following bound on the covering number in terms of a volume ratio : @xmath457 now we study the metric entropy of a cartesian product of convex cones .
let @xmath458 be a partition of @xmath20 $ ] with @xmath459 and @xmath460 . for @xmath181 , the restriction of @xmath26 to the coordinates in @xmath461
is denoted by @xmath462 .
let @xmath463 be a convex cone in @xmath464 and @xmath465 .
[ lem : cover - product ] with the notation above , suppose that @xmath466 .
then for any @xmath467 and @xmath468 $ ] , @xmath469 for some constant @xmath33 . since a product of balls
@xmath470 is contained in @xmath471 , one could try to cover @xmath472 by such products of balls .
it turns out that this yields an upper bound of order @xmath473 , which is too loose for our purpose .
fortunately , the following argument corrects this dependency . without loss of generality , we assume that @xmath474 .
we construct a @xmath475-net of @xmath476 as follows .
first , let @xmath477 be an @xmath478-net of @xmath443 with respect to the @xmath298-norm .
define @xmath479}\mu_i \ge - \frac{1}{2 \sqrt m } \big\}.\ ] ] note that @xmath480 for @xmath481 , and let @xmath482 be a @xmath483-net of @xmath484 .
define @xmath485 , i.e. , @xmath486 we claim that @xmath487 is an @xmath488-net of @xmath476 .
fix @xmath489 .
let @xmath490 be the restriction of @xmath491 to the component space @xmath464 .
then @xmath492 . let @xmath493 be defined by @xmath494 , so @xmath495 .
hence we can find @xmath496 such that @xmath497 .
in particular , for all @xmath498 $ ] , @xmath499 , so @xmath481 .
moreover , @xmath500 and @xmath492 , so by definition of @xmath482 , there exists @xmath501 such that @xmath502 .
let @xmath503 .
since @xmath504 we conclude that @xmath505 therefore @xmath487 is a @xmath475-net of @xmath476 .
it remains to bound the cardinality of this net . by lemma [ lem : cube - cover - ball ] , @xmath506 moreover , recall that @xmath482 is a @xmath483-net of @xmath484 .
since @xmath466 , for any @xmath467 , @xmath507 hence we can choose the net so that @xmath508 as @xmath509 , therefore @xmath510 taking the logarithm completes the proof .
recall that @xmath42 denotes the closed convex cone of increasing vectors in @xmath27 .
first , we prove a result on the metric entropy of @xmath42 intersecting with a ball using lemma [ lem : cover - product ] .
[ lem : cover - vector - centered ] let @xmath511 be such that @xmath512 . then for any @xmath341 and @xmath513 , @xmath514 the majority of the proof is due to lemma 5.1 in an old version of @xcite , but we improve their result by a factor @xmath515 and provide the whole proof for completeness .
the bound holds trivially if @xmath516 , since the left - hand side is zero .
it also clearly holds when @xmath517 .
hence we can assume without loss of generality that @xmath518 and @xmath519 .
moreover , assume that @xmath520 for simplicity and the proof will work for any @xmath341 .
let @xmath521 and observe that @xmath522 let @xmath157 be the smallest integer for which @xmath523 .
we partition @xmath524 into @xmath157 blocks @xmath525 for @xmath526 $ ] and let @xmath527 . since @xmath528 , lemma [ lem : cover - product ] yields that @xmath529 we know from ( * ? ? ?
* lemma 4.20 ) that for any @xmath530 and @xmath531 , @xmath532^n \cap \mathcal b^{n}(b,1 ) , \|\cdot\|_2 , { \varepsilon}\big)\leq \frac{c\sqrt{n}(d - c)}{\varepsilon } \,.\ ] ] for each @xmath533 , it holds that @xmath534 for @xmath535 ( since either @xmath536 for all @xmath537 or @xmath536 for all @xmath538 ; see e.g. @xcite ) , so @xmath539 .
also @xmath540 , so we get that @xmath541 for all @xmath526 $ ] . substituting this bound into and noting that @xmath542
, we reach the conclusion @xmath543 next , we study the metric entropy of the set of matrices with unimodal columns . recall that @xmath44 for @xmath544 $ ] . for @xmath49^m$ ] , define @xmath545
. moreover , for @xmath546 , @xmath467 and @xmath547 , define @xmath548 note that in particular @xmath549 .
[ lem : cover - matrix - uni ] given @xmath184 and @xmath49^m$ ] , define @xmath550 and @xmath551 then for any @xmath341 and @xmath513 , @xmath552 assume that @xmath518 since otherwise the left - hand side is zero and the bound holds trivially
. for @xmath162 $ ] , define @xmath553 $ ] and @xmath554 \setminus [ l_j]$ ] .
define @xmath555 and @xmath556 .
let @xmath557 and observe that @xmath558 moreover , let @xmath559 be the partition of @xmath560 such that @xmath561 is a constant vector for @xmath562 $ ] .
note that elements of @xmath563 need not to be consecutive .
define the partition for @xmath564 analogously . for @xmath162 $ ] and @xmath562 $ ] ( resp .
$ ] ) , let @xmath566 ( resp .
@xmath567 ) denote the set of increasing ( resp .
decreasing ) vectors in the component space @xmath568 ( resp .
@xmath569 ) . lemma [ lem : cover - vector - centered ] implies that @xmath570 as a matrix in @xmath571 can be viewed as a concatenation of @xmath572 vectors of length @xmath573 , j \in [ m]$ ] , we define the cone @xmath574 in @xmath575 by @xmath576 , which is clearly a superset of @xmath577 .
it also follows that @xmath578 , and thus by lemma [ lem : cover - product ] and the previous display , @xmath579 where we used the concavity of the logarithm and jensen s inequality in the second step , and that @xmath580 in the last step .
since @xmath578 ( the cone @xmath581 is pointed at @xmath0 ) we have that @xmath582 for any @xmath583 . in view of definition , it holds @xmath584 in particular , taking @xmath585 , we get @xmath586 .
moreover , @xmath587 , so that @xmath588 . thus the metric entropy of @xmath589 is subject to the above bound as well . finally , we consider the metric entropy of @xmath590 for @xmath546 , @xmath467 and @xmath54 .
the above analysis culminates in the following lemma which we use to prove the main upper bounds .
[ lem : cover - matrix - union ] let @xmath546 and @xmath224 be defined as in the previous lemma .
then for any @xmath513 and @xmath467 , @xmath591 assume that @xmath518 since otherwise the left - hand side is zero and the bound holds trivially .
note that @xmath592^m } \mathcal c_\mathbf l^m$ ] , and that @xmath54 . thus @xmath76 is the union of @xmath593 cones of the form @xmath594 . by definition , @xmath595 is also the union of @xmath593 sets @xmath596 , each having metric entropy subject to the bound in lemma [ lem : cover - matrix - uni ] .
therefore , a union bound implies that @xmath597 where the last step follows from that @xmath598 for @xmath599 and that @xmath518 .
for minimax lower bounds , we consider the model @xmath169 where entries of @xmath14 are i.i.d .
@xmath190 . the varshamov - gilbert lemma ( * ? ? ?
* lemma 4.7 ) is a standard tool for proving lower bounds .
[ lem : vg ] let @xmath444 denote the hamming distance on @xmath600 where @xmath601 . then there exists a subset @xmath602 such that @xmath603 and @xmath604 for distinct @xmath605 .
we also need the following useful lemma .
[ lem : lower - general ] consider the model @xmath606 where @xmath607 and @xmath608 .
suppose that @xmath609 and for distinct @xmath610 , @xmath611 where @xmath612 .
then there exists @xmath260 such that @xmath613 \ge c.\ ] ] let @xmath614 denote the probability with respect to @xmath615 . then the kullback - leibler divergence between @xmath616 and @xmath617 satisfies that @xmath618 since @xmath609 . applying ( * ? ? ?
* theorem 2.5 ) with @xmath619 gives the conclusion .
we define @xmath620 and @xmath621 .
define the subset of @xmath622 containing permutations of monotonic matrices by @xmath623 .
since each estimator pair @xmath170 gives an estimator @xmath97 of @xmath624 , it suffices to prove a lower bound on @xmath625 . in fact , we prove a stronger lower bound than the one in theorem [ thm : lower - adaptive ] .
[ prop : lower - stronger ] suppose that @xmath626
. then @xmath627 \ge c'\end{gathered}\ ] ] for some @xmath628 , where @xmath629 is the probability with respect to @xmath630 .
this bound remains valid for the parameter subset @xmath631 if @xmath632 or @xmath633 .
note that the bound clearly holds for the larger parameter space @xmath634 . by taking @xmath635 and
@xmath636 large enough , we see that the assumption in proposition [ prop : lower - stronger ] is satisfied and the second term becomes simply @xmath637 , so theorem [ thm : lower - adaptive ] follows . in the monotonic case , by the last statement of the proposition , if @xmath638 then taking @xmath639 and @xmath636 large enough yields a lower bound of rate @xmath640 for the set of matrices @xmath0 with increasing columns and @xmath641 .
the proof of proposition [ prop : lower - stronger ] has two parts which correspond to the two terms respectively .
first , the term @xmath642 is derived from the proof of lower bounds for isotonic regression in @xcite .
then we derive the other term @xmath643 for any @xmath644 , which is due to the unknown permutation .
[ lem : lower - est ] suppose that @xmath626 .
for some @xmath628 , @xmath645 \ge c\,,\ ] ] where @xmath646 is the probability with respect to @xmath630 .
we adapt the proof of ( * ? ? ?
* theorem 4 ) to the case of matrices .
let @xmath647 for all @xmath162 $ ] .
since @xmath648,\ ] ] we can choose @xmath649 $ ] so that @xmath650 and @xmath651 . according to lemma [ lem : vg ] , there exists @xmath652 such that @xmath653 and @xmath654 for distinct @xmath605 .
consider the partition @xmath655 = \cup_{m=1}^j i_j$ ] with @xmath656 .
for each @xmath657 , let @xmath658 be the restriction of @xmath659 to coordinates in @xmath660 .
define @xmath661 by @xmath662 where @xmath663 .
it is straightforward to check that @xmath664 , @xmath665 and @xmath666 is increasing , so @xmath667 is in the parameter space .
moreover , for distinct @xmath605 , @xmath668 on the other hand , @xmath669 applying lemma [ lem : lower - general ] completes the proof . for the second term in
, we first note that the bound is trivial for @xmath670 since @xmath671 . the next lemma deals with the case @xmath639 .
[ lem : lower - monotone ] there exist constants @xmath628 such that for any @xmath638 and @xmath213 , @xmath672 \ge c ' \,,\ ] ] where @xmath646 is the probability with respect to @xmath630 . by lemma [ lem : vg ]
, there exists @xmath673 such that @xmath674 and @xmath675 for distinct @xmath605 . for each @xmath657 , define @xmath676 by setting the first column of @xmath677 to be @xmath678 and all other entries to be zero , where @xmath679 then 1 .
@xmath680 since @xmath681 , @xmath682 and we can permutate the rows of @xmath677 so that its first column is increasing ; 2 .
@xmath683 for distinct @xmath605 ; 3 .
@xmath684 for @xmath605 .
applying lemma [ lem : lower - general ] completes the proof . for the previous two lemmas , we have only used matrices with increasing columns
however , to achieve the second term in for @xmath685 , we need matrices with unimodal columns .
the following packing lemma is the key .
[ lem : zero - one - packing ] for @xmath686 $ ] , consider the set @xmath687 of @xmath10 matrices of the form @xmath688 \text { for each } i \in [ n ] , \\ 0 & \text{otherwise } . \end{cases}\ ] ] for @xmath689 , define @xmath690 .
then there exists an @xmath691-packing @xmath692 of @xmath687 such that @xmath693 if @xmath694 and @xmath695 if @xmath696 .
there are @xmath45 choices of entries to put the one in each row of @xmath667 , so @xmath697 .
fix @xmath698 . if @xmath699 where @xmath700 , then @xmath667 differs from @xmath701 in at most @xmath157 rows . if @xmath696 , taking @xmath702 gives the result .
if @xmath694 then @xmath703 moreover , let @xmath692 be a maximal @xmath704-packing of @xmath687 . then @xmath692 is also an @xmath691-net , so @xmath705 it follows that @xmath706 we conclude that @xmath693 . for notational simplicity ,
we now consider @xmath707 instead of @xmath708 .
[ lem : lower - approx ] there exist constants @xmath628 such that for any @xmath200 , @xmath213 and @xmath707 , @xmath709 \ge c ' \,,\ ] ] where @xmath646 is the probability with respect to @xmath630 .
set @xmath710 and let @xmath692 be the @xmath711-packing given by lemma [ lem : zero - one - packing ] . if @xmath712 , then @xmath713 .
now assume that @xmath694 . since @xmath714 is decreasing on @xmath715 $ ] , we have that @xmath716 .
hence for @xmath717 , @xmath718 moreover , for each @xmath719 , consider the rescaled matrix @xmath720 1 .
we can permute the rows of @xmath701 so that each column has consecutive ones ( or all zeros ) , so @xmath721 . moreover , @xmath722 and @xmath723 so @xmath724 for @xmath719 .
2 . for @xmath725 , @xmath726 , so @xmath727 3 . for @xmath725 , @xmath728 , so by , @xmath729 since @xmath730 for @xmath717 , applying lemma [ lem : lower - general ] completes the proof . combining lemma [ lem : lower - est ] , [
lem : lower - monotone ] and [ lem : lower - approx ] , and dividing the bound by @xmath376 , we get because the max of two terms is lower bounded by a half of their sum .
the last statement in proposition [ prop : lower - stronger ] holds since lemma [ lem : lower - est ] and [ lem : lower - monotone ] are proved for matrices with increasing columns .
the proof will only use lemma [ lem : lower - est ] and [ lem : lower - monotone ] , so the lower bound of rate @xmath731 holds even if the matrices are required to have increasing columns .
the last term @xmath732 is achieved by lemma [ lem : lower - monotone ] , so we focus on the trade - off between the first two terms .
suppose that @xmath733 , in which case the first term @xmath734 dominates the second term
. then @xmath735 . setting @xmath736
we see that @xmath737 .
lemma [ lem : lower - est ] can be applied with this choice of @xmath738 .
then the term @xmath739 is lower bounded by @xmath740 . on the other hand , if @xmath741 , then the second term @xmath742 dominates the first up to a constant . to deduce a lower bound of this rate
, we apply lemma [ lem : vg ] to get @xmath743 such that @xmath744 and @xmath745 for distinct @xmath605 . for each @xmath657 , define @xmath676 by setting every row of @xmath677 equal to @xmath746 . then 1
. @xmath747 since @xmath748 ; 2 . @xmath749
@xmath750 .
hence lemma [ lem : lower - general ] implies a lower bound on @xmath751 of rate @xmath752 .
for the model @xmath169 where @xmath219 and @xmath340 , a computationally efficient estimator @xmath256 has been constructed in section [ sec : rankscore ] using the procedure
. we will bound its rate of estimation in this section .
recall that the definition of @xmath256 consists of two steps .
first , we recover an order ( or a ranking ) of the rows of @xmath71 , which leads to an estimator @xmath93 of the permutation . then define @xmath253 so that @xmath254 is the projection of @xmath71 onto the convex cone @xmath255 . for the analysis of the algorithm
, we deal with the projection step first , and then turn to learning the permutation .
in fact , for _ any _ estimator @xmath93 , if @xmath293 is defined as above by the projection corresponding to @xmath93 , then the error @xmath753 can be split into two parts : the permutation error @xmath754 and the estimation error of order @xmath755 .
the proof of the following oracle inequality is very similar to that of theorem [ thm : adaptive ] , so we will sketch the proof without providing all the details .
[ lem : approx - est ] consider the model @xmath169 where @xmath756 and @xmath70 .
for any @xmath757 , define @xmath253 so that @xmath254 is the projection of @xmath71 onto @xmath255 .
then with probability at least @xmath758 , @xmath759 assume without loss of generality that @xmath354 .
let @xmath226 and define @xmath760 since @xmath761 with @xmath762 , by lemma [ lem : cover - matrix - uni ] , @xmath763 using the proof of lemma [ lem : f - bound ] , we see that @xmath764 with probability at least @xmath765
. then the proof of theorem [ thm : adaptive ] implies that with probability at least @xmath758 , @xmath766 minimizing over @xmath226 yields the desired result . the idea of splitting the error into two terms as in lemma [ lem : approx - est ] has appeared in @xcite . by virtue of lemma [ lem : approx - est ]
, it remains to control the permutation error @xmath767 where @xmath93 is given by the procedure defined in section [ sec : rankscore ] .
recall that for @xmath246 $ ] , @xmath247 } ( a^*_{i ' , j } - a^*_{i , j})\vee \frac 1{\sqrt m } \sum_{j=1}^m ( a^*_{i ' , j } - a^*_{i , j})\ ] ] and @xmath248 is defined analogously .
since columns of @xmath78 are increasing , @xmath768 recall that the procedure is defined as follows .
first , for @xmath249 $ ] , we associate with the @xmath9-th row of @xmath71 a score @xmath250 defined by @xmath769 for the threshold @xmath770 where @xmath444 is the probability of failure
. then we order the rows of @xmath71 so that the scores are increasing with ties broken arbitrarily .
this is equivalent to requiring that the corresponding permutation @xmath771 \to [ n]$ ] satisfies that if @xmath772 then @xmath773 .
define @xmath93 to be the @xmath13 permutation matrix corresponding to @xmath774 so that @xmath775 for @xmath249 $ ] and all other entries of @xmath93 are zero .
moreover , let @xmath776 \to [ n]$ ] be the permutation corresponding to @xmath77 .
to control the permutation error , we first state a lemma which asserts that if the gap between two rows of @xmath78 is sufficiently large , then the permutation defined above will recover their relative order with high probability . [
lem : row - gap ] there is an event @xmath777 of probability at least @xmath778 on which the following holds . for any @xmath246
$ ] , if @xmath779 , then @xmath780 . since @xmath340 ,
@xmath781 and @xmath782 are sub - gaussian random variables with variance proxy @xmath783 .
a standard union bound yields that @xmath784 , j \in [ m ] } |z_{i , j}| , \max_{i \in [ n ] } \frac 1{\sqrt m } \big|\sum_{j=1}^m z_{i , j}\big|\big ) \le \tau = 3 \sigma \sqrt{\log(nm \delta^{-1})}\ ] ] on an event @xmath777 of probability at least @xmath785 .
in the sequel , we make statements that are valid on the event @xmath777 .
since @xmath786 , by the triangle inequality , latexmath:[\[\label{eq : error - gap } suppose that @xmath779 .
we claim that @xmath788 .
if for @xmath544 $ ] , @xmath789 , then @xmath790 by . since @xmath78 has increasing columns , @xmath791 .
again by , @xmath792 . by definition , we see that @xmath793 . moreover , @xmath779 so @xmath794 . therefore @xmath788 . according to the construction of @xmath774 , @xmath780 . next ,
recall that for a matrix @xmath226 , @xmath227 denotes the set of pairs of indices @xmath228 ^ 2 $ ] such that @xmath22 and @xmath229 are not identical .
the quantity @xmath230 is defined by @xmath795 ^ 2\\ |\mathcal i| = n } } \sum_{(i , j ) \in \mathcal i \cap \mathcal j } \big ( \frac{\|a_{i,\cdot } - a_{j,\cdot}\|_2 ^ 2}{\|a_{i,\cdot } - a_{j,\cdot}\|_\infty^2}\wedge \frac{m \|a_{i,\cdot } - a_{j,\cdot}\|_2 ^ 2}{\|a_{i,\cdot } - a_{j,\cdot}\|_1 ^ 2 } \big ) \,.\ ] ] for any nonzero vector @xmath236 , @xmath237 with equality achieved when @xmath238 , and @xmath796 with equality achieved when all entries of @xmath233 are the same .
hence @xmath797 .
moreover , @xmath798 by hlder s inequality , so @xmath799 as the product of the two terms is no larger than @xmath41 .
the equality is achieved by @xmath800 where the first @xmath244 entries are equal to one .
therefore , @xmath801 \,.\ ] ] intuitively , the quantity @xmath230 is small if the difference of any two rows of @xmath0 is either very sparse or very dense .
[ lem : approx - error ] there is an event @xmath777 of probability at least @xmath778 on which @xmath802 throughout the proof , we restrict ourselves to the event @xmath777 defined in lemma [ lem : row - gap ] . to simplify the notation , we define @xmath803 . then @xmath804 where @xmath524 is the set of indices @xmath9 for which @xmath805 is nonzero . for each @xmath806 ,
@xmath807 by .
next , we proceed to showing that @xmath808 for any @xmath249 $ ] , where @xmath809 . to that end , note that if @xmath810 , in which case @xmath811 for all @xmath812\,:\ , i'\ge \nu(i)\}$ ] , then it follows from lemma [ lem : row - gap ] that on @xmath813 , @xmath814 .
note that @xmath815 .
hence @xmath814 implies that @xmath816 , which is a contradiction .
therefore , there does not exist such @xmath249 $ ] on @xmath813 .
the case where @xmath817 is treated in a symmetric manner .
combining this bound with and , we conclude that @xmath818 by the definitions of @xmath261 and @xmath94 .
the bound is an immediate consequence of lemma [ lem : approx - est ] and lemma [ lem : approx - error ] with @xmath819 for @xmath820 .
in theorem [ thm : lower - global ] , we have observed the term @xmath821 , whereas the ls estimator only has @xmath822 in the upper bounds .
the next proposition shows that in the case @xmath823 , we can simply use an averaging estimator that achieves the term @xmath824 .
[ prop : special - regime ] for @xmath169 where @xmath340 , let @xmath825 and @xmath17 be defined by @xmath826 for all @xmath228\times [ m]$ ] .
then , @xmath827 with probability at least @xmath828 and @xmath829 recall that @xmath830 .
since the @xmath297-norm of a vector is no larger than the @xmath831-norm , @xmath832 on the other hand , @xmath833 so we have that @xmath834 , j \in [ m ] } \big(\frac 1n \sum_{k=1}^n a^*_{k , j } + \frac 1n \sum_{k=1}^n z_{k , j } - a^*_{i , j } \big)^2 \\ & \le 2 \sum_{i \in [ n ] , j \in [ m ] } \big(\frac 1n \sum_{k=1}^n a^*_{k , j } - a^*_{i , j } \big)^2 + \frac 2{n^2 } \sum_{i \in [ n ] , j \in [ m ] } \big(\sum_{k=1}^n z_{k , j } \big)^2 \\ & \le 2 n \sum_{j \in [ m ] } v_j(a)^2 + \frac 2n \sum_{j \in [ m ] } \big(\sum_{k=1}^n z_{k , j } \big)^2 \\ & \le 2 n m^3 v(a)^2 + 2 \sum_{j \in [ m ] } g_j^2 \ , , \end{aligned}\ ] ] where @xmath835 for @xmath162 $ ] so that @xmath836 are centered sub - gaussian variables with variance proxy @xmath783 .
it is well - known that @xmath837 , so @xmath838 moreover , since @xmath839 is a sub - gaussian vector with variance proxy @xmath783 , it follows from ( * ? ? ?
* theorem 2.1 ) that @xmath840 with probability at least @xmath828 . on this event , @xmath841 dividing the previous two displays by @xmath376 completes the proof .
if the permutation in the main model is known , then the estimation problem simply becomes a concatenation of @xmath41 unimodal regressions .
in fact , our proofs imply new oracle inequalities for unimodal regression . recall that @xmath46 denotes the cone of unimodal vectors in @xmath27 .
suppose that we observe @xmath842 where @xmath843 and @xmath303 is a sub - gaussian vector with variance proxy @xmath783 .
define the ls estimator @xmath300 by @xmath844 moreover let @xmath845 and @xmath846 } \theta_i - \min_{i \in [ n ] } \theta_i$ ] .
[ cor : unimodal ] there exists a constant @xmath260 such that with probability at least @xmath847 , @xmath848 , @xmath849 and @xmath850 + \alpha \sigma^2 \frac{\log n}{n } \,.\ ] ] the corresponding bounds in expectation also hold .
first note that the term @xmath851 in the bound of lemma [ lem : cover - matrix - union ] comes from a union bound applied to the set of permutations , so it is not present if we consider only the set of unimodal matrices @xmath48 instead of @xmath76 .
hence taking @xmath852 in the lemma yields that @xmath853 for @xmath854 , define @xmath855 following the proof of lemma [ lem : f - bound ] and using the above metric entropy bound , we see that @xmath856 with probability at least @xmath765 .
then the proof of theorem [ thm : adaptive ] gives that with probability at least @xmath765 ,
@xmath857 taking @xmath858 for @xmath848 and @xmath28 sufficiently large , we get that with probability at least @xmath847 , @xmath859 minimizing over @xmath854 yields .
the corresponding bound in expectation follows from integrating the tail probability as in the proof of theorem [ thm : adaptive ] .
note that the bounds in corollary [ cor : unimodal ] match the minimax lower bounds for isotonic regression in @xcite up to logarithmic factors .
since every monotonic vector is unimodal , lower bounds for isotonic regression automatically hold for unimodal regression .
therefore , we have proved that the ls estimator is minimax optimal up to logarithmic factors for unimodal regression . a result similar
to was obtained by bellec in the revision of @xcite that was prepared independently and contemporaneously to this paper .
chatterjee and lafferty also improved their bounds to having optimal exponents @xcite after the first version of our current paper was posted .
interestingly bellec employs bounds on the statistical dimension by leveraging results from @xcite , and chatterjee and lafferty use both the variational formula and the statistical dimension .
moreover , their results are presented in the well - specified case where @xmath860 and @xmath861 .
n. n. anuchina , k. i. babenko , s. k. godunov , n. a. dmitriev , l. v. dmitrieva , v. f. dyachenko , a. v. zabrodin , o. v. lokutsievski , e. v. malinovskaya , i. f. podlivaev , g. p. prokopov , i. d. sofronov , and r. p. fedorenko . .
`` nauka '' , moscow , 1979 .
quentin berthet and philippe rigollet .
complexity theoretic lower bounds for sparse principal component detection . in shai
shalev - shwartz and ingo steinwart , editors , _ colt 2013 - the 26th conference on learning theory , princeton , nj , june 12 - 14 , 2013 _ , volume 30 of _ jmlr w&cp _ ,
pages 10461066 , 2013 .
constantinos daskalakis , ilias diakonikolas , and rocco a. servedio .
learning k - modal distributions via testing . in _ proceedings of the twenty - third annual acm - siam symposium on discrete algorithms
_ , soda 12 , pages 13711385 , philadelphia , pa , usa , 2012 .
society for industrial and applied mathematics .
constantinos daskalakis , ilias diakonikolas , rocco a. servedio , gregory valiant , and paul valiant . testing k - modal distributions : optimal algorithms via reductions . in _ proceedings of the twenty - fourth annual acm - siam symposium on discrete algorithms
_ , soda 13 , pages 18331852 , philadelphia , pa , usa , 2013 . society for industrial and applied mathematics .
fajwel fogel , rodolphe jenatton , francis bach , and alexandre daspremont .
convex relaxations for permutation problems . in c.j.c .
burges , l. bottou , m. welling , z. ghahramani , and k.q .
weinberger , editors , _ advances in neural information processing systems 26 _ , pages 10161024 .
curran associates , inc . , 2013 .
thomas l. gertzen and martin grtschel .
flinders petrie , the travelling salesman problem , and the beginning of mathematical modeling in archaeology . ,
x(extra volume : optimization stories):199210 , 2012 .
michel ledoux and michel talagrand . ,
volume 23 of _ ergebnisse der mathematik und ihrer grenzgebiete ( 3 ) [ results in mathematics and related areas ( 3)]_. springer - verlag , berlin , 1991 .
isoperimetry and processes .
cong han lim and stephen wright . beyond the birkhoff polytope : convex relaxations for vector permutation problems . in z.
ghahramani , m. welling , c. cortes , n.d .
lawrence , and k.q .
weinberger , editors , _ advances in neural information processing systems 27 _ , pages 21682176 .
curran associates , inc . , 2014 . | given a matrix , the seriation problem consists in permuting its rows in such way that all its columns have the same shape , for example , they are monotone increasing .
we propose a statistical approach to this problem where the matrix of interest is observed with noise and study the corresponding minimax rate of estimation of the matrices .
specifically , when the columns are either unimodal or monotone , we show that the least squares estimator is optimal up to logarithmic factors and adapts to matrices with a certain natural structure .
finally , we propose a computationally efficient estimator in the monotonic case and study its performance both theoretically and experimentally .
our work is at the intersection of shape constrained estimation and recent work that involves permutation learning , such as graph denoising and ranking .
, statistical seriation , permutation learning , minimax estimation , adaptation , shape constraints , matrix estimation . |
su(2 ) einstein - yang - mills - higgs ( eymh ) theory , with the higgs field in the adjoint representation , possesses globally regular gravitating magnetic monopole solutions and corresponding magnetically charged black hole solutions @xcite . for small gravitational constant ,
the gravitating fundamental monopole solution smoothly emerges from the corresponding flat space solution , the t hooft - polyakov monopole @xcite . with increasing gravitational constant ,
the mass of the gravitating fundamental monopole solution decreases , and it ceases to exist beyond a maximal value of the gravitational constant . the corresponding magnetically charged eymh black hole solutions represent counterexamples to the `` no - hair '' conjecture .
distinct from embedded reissner - nordstrm ( rn ) black holes with unit magnetic charge , they emerge from the globally regular magnetic monopole solutions when a finite regular event horizon is imposed .
consequently , they have been characterized as `` black holes within magnetic monopoles '' @xcite .
besides the fundamental monopole solution there are excited monopole solutions @xcite . the gauge field function of the @xmath0-th excited monopole solution possesses @xmath0 nodes , whereas the gauge field function of the fundamental monopole solution decreases monotonically to zero @xcite .
the excited monopole solutions are related to the globally regular einstein - yang - mills ( eym ) solutions , found by bartnik and mckinnon @xcite , and , like these solutions , have no flat space counterparts .
in flat space also dyon solutions exist , carrying both electric and magnetic charge @xcite . here
we show that , like the monopole solutions , these dyon solutions persist in the presence of gravity , up to some maximal value of the gravitational constant . beside the fundamental gravitating dyon solutions , we also construct excited gravitating dyon solutions and dyonic black hole solutions .
distinct from the embedded rn solutions with the same electric and magnetic charge , these `` black holes within dyons '' again represent counterexamples to the `` no - hair '' conjecture .
in contrast , pure su(2 ) eym theory possesses neither regular dyon solutions @xcite , nor dyonic black holes other than embedded rn solutions @xcite .
we consider the su(2 ) eymh action @xmath1 with @xmath2 and @xmath3 where @xmath4 @xmath5 @xmath6 is the gauge coupling constant , @xmath7 is the higgs coupling constant and @xmath8 is the higgs field vacuum expectation value .
variation of the action eq .
( [ action ] ) with respect to the metric @xmath9 , the gauge field @xmath10 and the higgs field @xmath11 leads to the einstein equations and the matter field equations . to construct static
spherically symmetric globally regular and black hole solutions we employ schwarzschild - like coordinates and adopt the spherically symmetric metric @xmath12 with @xmath13 for the gauge and higgs field we employ the spherically symmetric ansatz @xcite @xmath14 @xmath15 and @xmath16 with unit vectors @xmath17 , @xmath18 and @xmath19 . for @xmath20 , gravitating monopole solutions are obtained @xcite .
we now introduce the dimensionless coordinate @xmath21 and the dimensionless mass function @xmath22 , @xmath23 as well as the coupling constants @xmath24 and
@xmath25 , @xmath26 the @xmath27 and @xmath28 components of the einstein equations then yield the equations for the metric functions , @xmath29 and @xmath30 where the prime indicates the derivative with respect to @xmath21 . for the matter functions we obtain the equations @xmath31 @xmath32 and @xmath33 a special solution of these equations is the embedded rn solution with mass @xmath34 , unit magnetic charge and arbitrary electric charge @xmath35 , @xmath36 @xmath37 the corresponding extremal rn solution has horizon @xmath38 , @xmath39
let us first consider the globally regular particle - like solutions of the su(2 ) eymh system . requiring asymptotically flat solutions implies that the metric functions @xmath40 and @xmath22 both approach a constant at infinity .
we here adopt @xmath41 and @xmath42 represents the dimensionless mass of the solutions .
the matter functions also approach constants asymptotically , @xmath43 where for magnetic monopole solutions @xmath44 .
the asymptotic fall - off of the function @xmath45 determines the dimensionless electric charge @xmath35 ( see eq .
( [ q ] ) ) .
regularity of the solutions at the origin requires @xmath46 and @xcite @xmath47 the globally regular dyon solutions have many features in common with the globally regular monopole solutions . in the prasad - sommerfield limit , @xmath48 ,
the dyon solutions in flat space are known analytically @xcite , whereas for finite @xmath25 they are obtained numerically @xcite . in the presence of gravity ,
the corresponding gravitating dyon solutions extend up to a maximal value of the coupling constant @xmath24 . beyond this value
no dyon solutions exist . for small values of @xmath25
, the fundamental dyon branch does not end at the maximal value @xmath49 .
instead it exhibits a small spike there and bends backwards , up to the critical coupling constant @xmath50 . since variation of the coupling constant @xmath51
can be considered in two different ways , either as changing @xmath52 and keeping @xmath8 fixed , or vice versa , for small @xmath25 the fundamental dyon branch can be interpreted as obtained by first varying @xmath52 up to the maximal value @xmath49 , and then varying @xmath8 up to the critical value @xmath50 . at the critical value
@xmath50 the fundamental dyon branch reaches a limiting solution and bifurcates with the branch of extremal rn solutions of unit magnetic charge and electric charge @xmath35 .
the fundamental dyon branch is thus completely analogous to the fundamental monopole branch @xcite .
this is demonstrated for the normalized mass @xmath53 of the dyon solutions with electric charge @xmath54 and @xmath48 in fig . 1 , where for comparison also the monopole solutions ( @xmath55 , @xmath48 ) are shown , together with the extremal rn solutions of unit magnetic charge and electric charge @xmath54 as well as @xmath55 .
the normalization in fig . 1 is chosen to obtain the finite mass of the flat space solutions in the limit @xmath56 .
the adm mass @xmath57 can be read off the figure . for the dyon branch
we find a critical coupling of @xmath58 , while for the monopole branch it is @xmath59 @xcite . along the fundamental branch
the dyon functions approach limiting functions , when @xmath60 . as for the monopole solutions , the metric function @xmath61 of the dyon solutions develops a minimum , which decreases monotonically along the fundamental branch . in the limit @xmath60 , the minimum approaches zero at @xmath62 .
the limiting metric function then consists of an inner part , @xmath63 , and an outer part , @xmath64 . for @xmath64
, the limiting metric function corresponds to the metric function @xmath61 of the extremal rn black hole with @xmath50 , unit magnetic charge and electric charge @xmath35 .
likewise the other functions approach limiting functions , when @xmath60 , which for @xmath64 correspond to those of the extremal rn black hole with @xmath50 , unit magnetic charge and electric charge @xmath35 .
the limit @xmath60 is demonstrated in fig . 2 for the matter function @xmath45 of the dyon solution with electric charge @xmath54 and @xmath48 .
in the figure the occurrence of the spike is seen , since the function @xmath45 does not reach the limiting function at the maximal value @xmath49 , but instead at the critical value @xmath50 .
the limiting function is identically zero for @xmath63 and coincides with the rn function for @xmath64 .
the limiting behaviour of the other functions of the dyon solutions is analogous to those of the monopole solutions , shown in @xcite . beside the branch of fundamental dyon solutions
there are branches of excited dyon solutions .
the gauge field function @xmath65 of the @xmath0-th excited dyon solution has @xmath0 nodes , whereas the gauge field function of the fundamental dyon solution decreases monotonically to zero .
the excited dyon solutions also exist only below some maximal value of the coupling constant @xmath24 .
since these excited solutions have no flat space counterparts , the variation of @xmath24 along a branch of excited solutions must be interpreted as a variation of @xmath8 while @xmath52 is kept fixed . in the limit @xmath56
the higgs field vacuum expectation value therefore vanishes , while @xmath52 remains finite . because of the particular choice of dimensionless variables ( [ xm ] ) , in this limit
the solutions shrink to zero size and their mass @xmath22 diverges .
the coordinate transformation @xmath66 @xcite leads to finite limiting solutions in the limit @xmath56 .
in fact , the transformed excited monopole solutions approach the bartnik - mckinnon solutions @xcite , and so do the transformed excited dyon solutions .
this is seen in fig . 3 , where we show the appropriately normalized mass @xmath67 as a function of @xmath24 for the first excited dyon branch with electric charge @xmath54 and @xmath48 .
for comparison also the first excited monopole branch is shown .
further details will be given elsewhere @xcite .
we now turn to the dyonic black hole solutions of the su(2 ) eymh system . imposing again the condition of asymptotic flatness , the black hole solutions satisfy the same boundary conditions at infinity as the regular solutions .
the existence of a regular event horizon at @xmath38 requires @xmath68 and @xmath69 , and the matter functions must satisfy @xmath70 @xmath71 and @xmath72 again , the su(2 ) eymh `` black holes within dyons '' have many features in common with the `` black holes within monopoles '' .
in particular , for a given coupling constant @xmath24 , the black hole solutions corresponding to the fundamental dyon branch emerge from the globally regular solution in the limit @xmath73 and persist up to a critical maximal value of the horizon radius . in fig . 4 we exhibit the mass of `` black holes within dyons '' with electric charge @xmath54 and @xmath48 as a function of the horizon radius for several values of @xmath24 , together with the corresponding branches of rn solutions of unit magnetic charge and electric charge @xmath54 .
for smaller values of @xmath24 , the black hole solutions merge into non - extremal rn solutions at a critical value of the horizon radius . for larger values of @xmath24 ,
the black hole solutions show a critical behaviour analogous to the globally regular solutions , and bifurcate with an extremal rn solution . here , with increasing horizon radius the limiting solution is reached at a critical value of the horizon radius smaller than the horizon radius of the extremal rn black hole with the same @xmath24 , unit magnetic charge and electric charge @xmath35 .
this critical behaviour is demonstrated for the function @xmath45 of the dyon black hole solutions with electric charge @xmath54 and coupling constants @xmath74 and @xmath48 in fig .
further details will be given elsewhere @xcite .
the globally regular dyon solutions have many features in common with the globally regular monopole solutions . like the fundamental monopole branch ,
the fundamental dyon branch starts from the corresponding flat space solution and extends up to a critical value of the coupling constant @xmath24 . at the critical value , both the fundamental monopole branch and the fundamental dyon branch bifurcate with the corresponding branch of extremal rn solutions .
the critical coupling constant depends slightly on the electric charge @xmath35 of the dyons .
likewise , beside the fundamental dyon branch there are branches of excited dyon solutions , which extend up to a maximal value of @xmath24 . in the limit @xmath56 , the branches of excited dyon solutions tend to the corresponding bartnik - mckinnon solutions , like their monopole counterparts .
starting from the stable monopole solutions in flat space , the solutions on the fundamental monopole branch remain stable for @xmath75 @xcite .
in contrast , the classical dyon solution in flat space is unstable , since the mass can be lowered continuously , by lowering the electric charge , as long as there is no charge quantization @xcite .
this indicates , that the gravitating fundamental dyon solutions are also unstable . in analogy to `` black holes within monopoles ''
also `` black holes within dyons '' exist . for a given value of the coupling constant @xmath24 , the `` black holes within dyons '' emerge from the globally regular solutions in the limit @xmath73 and persist up to a maximal value of the horizon radius . for small @xmath24
the black hole solutions merge into the corresponding non - extremal rn solutions at a critical value of the horizon radius , whereas for larger @xmath24 ( but @xmath76 ) the black hole solutions bifurcate at a critical value of the horizon radius with the corresponding extremal rn solutions .
the static spherically symmetric `` black holes within monopoles '' and `` black holes within dyons '' provide counterexamples to the `` no - hair conjecture '' .
eymh theory also possesses counterexamples of a different type , namely static aspherical black holes , which represent `` black holes within multimonopoles '' @xcite . beside static axially symmetric black holes , whose
eym counterparts have recently been obtained non
perturbatively @xcite , there are static black holes with only discrete symmetries @xcite .
it presents a challenge to construct such black holes with only crystal symmetries non - perturbatively .
000 k. lee , v.p . nair and e.j .
weinberg , black holes in magnetic monopoles , phys .
d45 ( 1992 ) 2751 .
p. breitenlohner , p. forgacs and d. maison , gravitating monopole solutions , nucl .
b383 ( 1992 ) 357 ; + gravitating monopole solutions ii , nucl .
b442 ( 1995 ) 126 .
aichelburg and p. bizon , magnetically charged black holes and their stability , phys .
d48 ( 1993 ) 607 .
g. t hooft , nucl .
b79 ( 1974 ) 276 ; + a.m. polyakov , jetp lett . 20
( 1974 ) 194 .
r. bartnik , and j. mckinnon , particlelike solutions of the einstein - yang - mills equations , phys .
61 ( 1988 ) 141 .
poles with both magnetic and electric charges in non - abelian gauge theory , phys .
d11 ( 1975 ) 2227 ; + m.k .
prasad and c.m .
sommerfield , exact classical solution for the t hooft monopole and the julia - zee dyon , phys .
( 1975 ) 760 .
ershov and d.v .
galtsov , phys .
lett . 150a ( 1990 ) 159 .
galtsov and a.a .
ershov , non - abelian baldness of colored black holes , phys .
lett . a138 ( 1989 ) 160 .
p. bizon and o.t .
popp , no - hair theorem for spherical monopoles and dyons in su(2 ) einstein - yang - mills theory , class .
quantum grav . 9 ( 1992 ) 193 .
y. brihaye , b. kleihaus and d.h .
tchrakian , dyon - skyrmion lumps , preprint hep - th/9805059 .
b. hartmann , j. kunz and y. brihaye , in preparation .
volkov , and d.v .
galtsov , black holes in einstein - yang - mills theory , sov .
( 1990 ) 747 ; + p. bizon , colored black holes , phys .
64 ( 1990 ) 2844 ; + h. p. knzle and a. k. m. masoud - ul - alam , spherically symmetric static su(2 ) einstein - yang - mills fields , j. math .
( 1990 ) 928 .
h. hollmann , on the stability of gravitating nonabelian monopoles , phys .
b338 ( 1994 ) 181 . s. a. ridgway and e. j. weinberg , static black hole solutions without rotational symmetry , phys .
d52 ( 1995 ) 3440 .
b. kleihaus and j. kunz , static black hole solutions with axial symmetry , phys .
79 ( 1997 ) 1595 ; static axially symmetric einstein - yang - mills - dilaton solutions : ii .
black hole solutions , phys .
d57 ( 1998 ) 6138 . | we study static spherically symmetric gravitating dyon solutions and dyonic black holes in einstein - yang - mills - higgs theory .
the gravitating dyon solutions share many features with the gravitating monopole solutions . in particular ,
gravitating dyon solutions and dyonic black holes exist only up to a maximal coupling constant , and beside the fundamental dyon solutions there are excited dyon solutions .
= 6.125truein = 8.125truein preprint hep - th/9807169 |
most astrophysical systems , e.g. accretion disks , stellar winds , the interstellar medium ( ism ) and intercluster medium are turbulent with an embedded magnetic field that influences almost all of their properties .
this turbulence which spans from km to many kpc ( see discussion in armstrong , rickett , & spangler 1995 ; scalo 1987 ; lazarian , pogosyan , & esquivel 2002 ) holds the key to many astrophysical processes . for instance , propagation of cosmic rays and their acceleration is strongly affected by mhd turbulence .
recent research has shown that a substantial part of the earlier results in the field require revision .
earlier research used ad hoc models of mhd turbulence and this entailed erroneous conclusions .
before we start a discussion of mhd turbulence let us recall some basic properties of the hydrodynamic turbulence .
all turbulent systems have one thing in common : they have a large reynolds number " ( @xmath0 ; l= the characteristic scale or driving scale of the system , v = the velocity difference over this scale , and @xmath1=viscosity ) , the ratio of the viscous drag time on the largest scales ( @xmath2 ) to the eddy turnover time of a parcel of gas ( @xmath3 ) .
a similar parameter , the magnetic reynolds number " , @xmath4 ( @xmath5 ; @xmath6=magnetic diffusion ) , is the ratio of the magnetic field decay time ( @xmath7 ) to the eddy turnover time ( @xmath3 ) .
the properties of the flows on all scales depend on @xmath8 and @xmath4 .
flows with @xmath9 are laminar ; chaotic structures develop gradually as @xmath8 increases , and those with @xmath10 are appreciably less chaotic than those with @xmath11 . observed features such as star forming clouds and accretion disks are very chaotic with @xmath12 and @xmath13 .
let us start by considering incompressible hydrodynamic turbulence , which can be described by the kolmogorov theory ( kolmogorov 1941 ) .
suppose that we excite fluid motions at a scale @xmath14 .
we call this scale the _ energy injection scale _ or the _ largest energy containing eddy scale_. for instance , an obstacle in a flow excites motions on scales of the order of its size . then the energy injected at the scale @xmath14 cascades to progressively smaller and smaller scales at the eddy turnover rate , i.e. @xmath15 , with negligible energy losses along the cascade .. ] ultimately , the energy reaches the molecular dissipation scale @xmath16 , i.e. the scale where the local @xmath17 , and is dissipated there .
the scales between @xmath14 and @xmath16 are called the _ inertial range _ and it typically covers many decades .
the motions over the inertial range are _ self - similar _ and this provides tremendous advantages for theoretical description .
the beauty of the kolmogorov theory is that it does provide a simple scaling for hydrodynamic motions .
if the velocity at a scale @xmath18 from the inertial range is @xmath19 , the kolmogorov theory states that the kinetic energy ( @xmath20 as the density is constant ) is transferred to next scale within one eddy turnover time ( @xmath21 ) .
thus within the kolmogorov theory the energy transfer rate ( @xmath22 ) is scale - independent , @xmath23 and we get the famous kolmogorov scaling @xmath24 the one - dimensional , which , for isotropic turbulence , is given by @xmath25 . ] energy spectrum @xmath26 is the amount of energy between the wavenumber @xmath27 and @xmath28 divided by @xmath29 .
when @xmath26 is a power law , @xmath30 is the energy _ near _ the wavenumber @xmath31 . since @xmath32 , kolmogorov
scaling implies @xmath33 kolmogorov scalings were the first major advance in the theory of incompressible turbulence .
they have led to numerous applications in different branches of science ( see monin & yaglom 1975 ) .
however , astrophysical fluids are magnetized and the a dynamically important magnetic field should interfere with eddy motions . paradoxically , astrophysical measurements are consistent with kolmogorov spectra ( see lpe02 ) .
for instance , interstellar scintillation observations indicate an electron density spectrum is very close to @xmath34 for @xmath35 - @xmath36 ( see armstrong et al .
1995 ) . at larger scales
lpe02 summarizes the evidence of -@xmath37 velocity power spectrum over pc - scales in hi .
solar - wind observations provide _ in - situ _ measurements of the power spectrum of magnetic fluctuations and leamon et al .
( 1998 ) also obtained a slope of @xmath38 -@xmath37 .
is this a coincidence ?
what properties is the magnetized compressible ism expected to have ?
we will deal with these questions , and some related issues , below . here
we discuss a focused approach which aims at obtaining a clear understanding on the fundamental level , and considering physically relevant complications later .
the creative synthesis of both approaches is the way , we think , that studies of astrophysical turbulence should proceed . certainly an understanding of mhd turbulence in the most ideal terms is a necessary precursor to understanding the complications posed by more realistic physics and numerical effects . for review of general properties of mhd ,
see a recent book by biskamp ( 1993 ) .
in what follows , we first consider observational data that motivate our study ( 2 ) , then discuss theoretical approaches to incompressible mhd turbulence ( 3 ) .
we move to the effects of compressibility in 4 and discuss implications of our new understanding of mhd turbulence for cosmic ray dynamics in 5 .
we present the summary in 6 .
kolmogorov turbulence is the simplest possible model of turbulence .
since it is incompressible and not magnetized , it is completely specified by its velocity spectrum . if a passive scalar field , like `` dye particles '' or temperature inhomogeneities , is subjected to kolmogorov turbulence , the resulting spectrum of the passive scalar density is also kolmogorov ( see lesieur 1990
; warhaft 2000 ) . in compressible and magnetized turbulence
this is no longer true , and a complete characterization of the turbulence requires not only a study of the velocity statistics but also the statistics of density and magnetic fluctuations . direct studies of turbulence have been done mostly for interstellar medium and for the solar wind . while for the solar wind _ in - situ _ measurements are possible , studies of interstellar turbulence require inverse techniques to interpret the observational data .
attempts to study interstellar turbulence with statistical tools date as far back as the 1950s ( von horner 1951 ; kamp de friet 1955 ; munch 1958 ; wilson et al .
1959 ) and various directions of research achieved various degree of success ( see reviews by kaplan & pickelner 1970 ; dickman 1985 ; armstrong et al .
1995 ; lazarian 1999a , 1999b ; lpe02 ) .
solar wind ( see review goldstein & roberts 1995 ) studies allow pointwise statistics to be measured directly using spacecrafts .
these studies are the closest counterpart of laboratory measurements .
the solar wind flows nearly radially away from the sun , at up to about 700 km / s .
this is much faster than both spacecraft motions and the alfvn speed .
therefore , the turbulence is `` frozen '' and the fluctuations at frequency @xmath39 are directly related to fluctuations at the scale @xmath27 in the direction of the wind , as @xmath40 , where @xmath41 is the solar wind velocity ( horbury 1999 ) .
usually two types of solar wind are distinguished , one being the fast wind which originates in coronal holes , and the slower bursty wind .
both of them show , however , @xmath42 scaling on small scales .
the turbulence is strongly anisotropic ( see klein et al .
1993 ) with the ratio of power in motions perpendicular to the magnetic field to those parallel to the magnetic field being around 30 .
the intermittency of the solar wind turbulence is very similar to the intermittency observed in hydrodynamic flows ( horbury & balogh 1997 ) .
studies of turbulence statistics of ionized media ( see spangler & gwinn 1990 ) have provided information on the statistics of plasma density at scales @xmath43-@xmath44 cm .
this was based on a clear understanding of processes of scintillations and scattering achieved by theorists ( see narayan & goodman 1989 ; goodman & narayan 1985 ) . a peculiar feature of the measured spectrum ( see armstrong et al .
1995 ) is the absence of the slope change at the scale at which the viscosity by neutrals becomes important .
scintillation measurements are the most reliable data in the `` big power law '' plot in armstrong et al .
however there are intrinsic limitations to the scintillations technique due to the limited number of sampling directions , its relevance only to ionized gas at extremely small scales , and the impossibility of getting velocity ( the most important ! ) statistics directly . therefore with the data one faces the problem of distinguishing actual turbulence from static density structures .
moreover , the scintillation data does not provide the index of turbulence directly , but only shows that the data are consistent with kolmogorov turbulence . whether the ( 3d ) index can be -4 instead of -11/3 is still a subject of intense debate ( higdon 1984 ; narayan & goodman 1989 ) . in physical terms the former corresponds to the superposition of random shocks rather than eddies .
additional information on the electron density is contained in the faraday rotation measures of extragalactic radio sources ( see simonetti & cordes 1988 ; simonetti 1992 ) . however , there is so far no reliable way to disentangle contributions of the magnetic field and the density to the signal .
we feel that those measurements may give us the magnetic field statistics when we know the statistics of electron density better .
spectral line data cubes are unique sources of information on interstellar turbulence .
doppler shifts due to supersonic motions contain information on the turbulent velocity field which is otherwise difficult to obtain .
moreover , the statistical samples are extremely rich and not limited to discrete directions .
in addition , line emission allows us to study turbulence at large scales , comparable to the scales of star formation and energy injection . however , the problem of separating velocity and density fluctuations within hi data cubes is far from trivial ( lazarian 1995 , 1999b ; lazarian & pogosyan 2000 ; lpe02 ) .
the analytical description of the emissivity statistics of channel maps ( velocity slices ) in lazarian & pogosyan ( 2000 ) ( see also lazarian 1999b ; lpe02 for reviews ) shows that the relative contribution of the density and velocity fluctuations depends on the thickness of the velocity slice .
in particular , the power - law asymptote of the emissivity fluctuations changes when the dispersion of the velocity at the scale under study becomes of the order of the velocity slice thickness ( the integrated width of the channel map ) .
these results are the foundation of the velocity - channel analysis ( vca ) technique which provides velocity and density statistics using spectral line data cubes .
the vca has been successfully tested using data cubes obtained via compressible magnetohydrodynamic simulations and has been applied to galactic and small magellanic cloud atomic hydrogen ( hi ) data ( lazarian et al .
2001 ; lazarian & pogosyan 2000 ; stanimirovic & lazarian 2001 ; deshpande , dwarakanath , & goss 2000 ) .
furthermore , the inclusion of absorption effects ( lazarian & pogosyan 2002 ) has increased the power of this technique . finally , the vca can be applied to different species ( co , h@xmath45 etc . ) which should further increase its utility in the future . within the present discussion
a number of results obtained with the vca are important .
first of all , the small magellanic cloud ( smc ) hi data exhibit a kolmogorov - type spectrum for velocity and hi density from the smallest resolvable scale of 40 pc to the scale of the smc itself , i.e. 4 kpc .
similar conclusions can be inferred from the galactic data ( green 1993 ) for scales of dozens of parsecs , although the analysis has not been done systematically .
deshpande et al . ( 2000 ) studied absorption of hi on small scales toward cas a and cygnus a. within the vca their results can be interpreted as implying that on scales less than 1 pc the hi velocity is suppressed by ambipolar drag and the spectrum of density fluctuations is shallow @xmath46 . such a spectrum ( deshpande 2000 )
can account for the small scale structure of hi observed in absorption .
magnetic field statistics are the most poorly constrained aspect of ism turbulence .
the polarization of starlight and of the far - infrared radiation ( fir ) from aligned dust grains is affected by the ambient magnetic fields . assuming that dust grains are always aligned with their longer axes perpendicular to magnetic field ( see the review lazarian 2000 ) , one gets the 2d distribution of the magnetic field directions in the sky .
note that the alignment is a highly non - linear process in terms of the magnetic field and therefore the magnetic field strength is not available . the statistics of starlight polarization ( see fosalba et al .
2002 ) is rather rich for the galactic plane and it allows to establish the spectrum @xmath47 , where @xmath48 is a two dimensional wave vector describing the fluctuations over sky patch .. for sufficiently small areas of the sky analyzed the multipole analysis results coincide with the fourier analysis . ] for uniformly sampled turbulence it follows from lazarian & shutenkov ( 1990 ) that @xmath49 for @xmath50 and @xmath51 for @xmath52 , where @xmath53 is the critical angular size of fluctuations which is proportional to the ratio of the injection energy scale to the size of the turbulent system along the line of sight . for kolmogorov turbulence @xmath54 .
however , the real observations do not uniformly sample turbulence .
many more close stars are present compared to the distant ones . thus the intermediate slops are expected .
indeed , cho & lazarian ( 2002b ) showed through direct simulations that the slope obtained in fosalba et al .
( 2002 ) is compatible with the underlying kolmogorov turbulence . at the moment
fir polarimetry does not provide maps that are really suitable to study turbulence statistics .
this should change soon when polarimetry becomes possible using the airborne sofia observatory
. a better understanding of grain alignment ( see lazarian 2000 ) is required to interpret the molecular cloud magnetic data where some of the dust is known not to be aligned ( see lazarian , goodman , & myers 1997 and references therein ) .
another way to get magnetic field statistics is to use synchrotron emission .
both polarization and intensity data can be used .
the angular correlation of polarization data ( baccigalupi et al . 2001 ) shows the power - law spectrum @xmath55 and we believe that the interpretation of it is similar to that of starlight polarization .
indeed , faraday depolarization limits the depth of the sampled region .
the intensity fluctuations were studied in lazarian & shutenkov ( 1990 ) with rather poor initial data and the results were inconclusive .
cho & lazarian ( 2002b ) interpreted the fluctuations of synchrotron emissivity ( giardino et al .
2001 , 2002 ) in terms of turbulence with kolmogorov spectrum .
attempts to describe magnetic turbulence statistics were made by iroshnikov ( 1963 ) and kraichnan ( 1965 ) .
their model of turbulence ( ik theory ) is isotropic in spite of the presence of the magnetic field . for simplicity ,
let us suppose that a uniform external magnetic field ( @xmath56 ) is present
. in the incompressible limit , any magnetic perturbation propagates _ along _ the magnetic field line . since wave packets are moving along the magnetic field line , there are two possible directions for propagation .
if all the wave packets are moving in one direction , then they are stable to nonlinear order ( parker 1979 ) . therefore , in order to initiate turbulence , there must be opposite - traveling wave packets and the energy cascade occurs only when they collide .
the ik theory starts from this observation , one of the consequences of which is the modification of the energy cascade timescale : @xmath57 , where @xmath58 is alfven velocity of the mean field . here
, the ik theory assumes that opposite - traveling isotropic wave packets of similar size interact . from this and the scale - invariance of energy cascade rate
, they obtained @xmath59 however , the presence of the uniform magnetic component has non - trivial dynamical effects on the turbulence fluctuations .
one obvious effect is that it is easy to mix field lines in directions perpendicular to the local mean magnetic field and much more difficult to bend them .
the ik theory assumes isotropy of the energy cascade in fourier space , an assumption which has attracted severe criticism ( montgomery & turner 1981 ; shebalin , matthaeus , & montgomery 1983 ; montgomery & matthaeus 1995 ; sridhar & goldreich 1994 ; matthaeus et al .
mathematically , anisotropy manifests itself in the resonant conditions for 3-wave interactions : @xmath60 where @xmath61 s are wavevectors and @xmath62 s are wave frequencies .
the first condition is a statement of wave momentum conservation and the second is a statement of energy conservation .
alfvn waves satisfy the dispersion relation : @xmath63 , where @xmath64 is the component of wavevector parallel to the background magnetic field .
since only opposite - traveling wave packets interact , @xmath65 and @xmath66 must have opposite signs . then from equations ( [ k123 ] ) and ( [ w123 ] ) ,
either @xmath67 or @xmath68 must be equal to 0 and @xmath69 must be equal to the nonzero initial parallel wavenumber .
that is , zero frequency modes are essential for energy transfer ( shebalin et al .
therefore , in the wavevector space , 3-wave interactions produce an energy cascade which is strictly perpendicular to the mean magnetic field .
however , in real turbulence , equation ( [ w123 ] ) does not need to be satisfied exactly , but only to within an an error of order @xmath70 ( goldreich & sridhar 1995 ) .
this implies that the energy cascade is not strictly perpendicular to @xmath56 , although clearly very anisotropic .
we assume throughout this discussion that the rms turbulent velocity at the energy injection scale is comparable to the alfvn speed of the mean field and consider only scales below the energy injection scale .
this is called _ strong _ turbulence regime .
note that , as a consequence , the regime of @xmath71 is not considered in this review . however , the regime of @xmath72 is still relevant to the strong turbulence regime because scales below the energy equipartition scale is expected to fall in the strong turbulence regime ( cho & vishniac 2000a ) .
an ingenious model very similar in its beauty and simplicity to the kolmogorov model has been proposed by goldreich & sridhar ( 1995 ; hereinafter gs95 ) for incompressible strong mhd turbulence .
they pointed out that motions perpendicular to the magnetic field lines mix them on a hydrodynamic time scale , i.e. at a rate @xmath73 , where @xmath74 is the wavevector component perpendicular to the local mean magnetic field and @xmath75 .
these mixing motions couple to the wave - like motions parallel to magnetic field giving a _ critical balance _
condition @xmath76 where @xmath64 is the component of the wavevector parallel to the local magnetic field .
when the typical @xmath64 on a scale @xmath74 falls below this limit , the magnetic field tension is too weak to affect the dynamics and the turbulence evolves hydrodynamically , in the direction of increasing isotropy in phase space .
this quickly raises the value of @xmath64 . in the opposite limit ,
when @xmath64 is large , the magnetic field tension dominates , the error @xmath77 in the matching conditions is reduced , and the nonlinear cascade is largely in the @xmath74 direction , which restores the critical balance .
if conservation of energy in the turbulent cascade applies locally in phase space then the energy cascade rate ( @xmath78 ) is constant : @xmath79 combining this with the critical balance condition we obtain an anisotropy that increases with decreasing scale @xmath80 and a kolmogorov - like spectrum for perpendicular motions @xmath81 which is not surprising since the magnetic field does not influence motions that do not bend it . at the same time , the scale - dependent anisotropy reflects the fact that it is more difficult for the weaker , smaller eddies to bend the magnetic field .
gs95 shows the duality of motions in mhd turbulence .
those perpendicular to the mean magnetic field are essentially eddies , while those parallel to magnetic field are waves . the critical balance condition couples these two types of motions . numerical simulations ( cho & vishniac 2000b ; maron & goldreich 2001 ; cho , lazarian , & vishniac 2002 ) show reasonable agreements with the gs95 model .
.notations for compressible turbulence [ cols= " < , < " , ] [ cho_table1 ] for the rest of the review , we consider mhd turbulence of a single conducting fluid .
while the gs95 model describes incompressible mhd turbulence well , no universally accepted theory exists for compressible mhd turbulence despite various attempts ( e.g. , higdon 1984 ) . earlier numerical simulations of compressible mhd turbulence covered a broad range of astrophysical problems , such as the decay of turbulence ( e.g. mac low 1998 ; stone , ostriker , & gammie 1998 ) or turbulent modeling of the ism ( see recent review vazquez - semadeni 2002 ; see also passot , pouquet , & woodward 1988 ; vazquez - semadeni , passot , & pouquet 1995 ; passot , vazquez - semadeni , & pouquet 1995 ; vazquez - semadeni , passot , & pouquet 1996 for earlier pioneering 2d simulations and ostriker , gammie , & stone 1999 ; ostriker , stone , & gammie 2001 ; padoan et al . 2001
; klessen 2001 ; boldyrev 2002 for recent 3d simulations ) . in what follows ,
we concentrate on the fundamental properties of compressible mhd . and
@xmath82 represent the directions of displacement of fast and slow modes , respectively . in the fast basis ( @xmath83 )
is always between @xmath84 and @xmath85 . in the slow basis ( @xmath82 )
lies between @xmath86 and @xmath87 . here
, @xmath86 is perpendicular to @xmath84 and parallel to the wave front .
all vectors lie in the same plane formed by @xmath88 and @xmath89 . on the other hand ,
the displacement vector for alfvn waves ( not shown ) is perpendicular to the plane .
( b ) directions of basis vectors for a very small @xmath90 drawn in the same plane as in ( a ) .
the fast bases are almost parallel to @xmath85 .
( c ) directions of basis vectors for a very high @xmath90 .
the fast basis vectors are almost parallel to @xmath89 . the slow waves become pseudo - alfvn waves . ]
lies between @xmath86 and @xmath91 and @xmath83 between @xmath84 and @xmath85 .
again , @xmath86 is perpendicular to @xmath84 and parallel to the wave front .
note also that , for the fast wave , for example , density ( inferred by the directions of the displacement vectors ) becomes higher where field lines are closer , resulting in a strong restoring force , which is why fast waves are faster than slow waves . ]
let us start by reviewing different mhd waves .
in particular , we describe the fourier space representation of these waves . the real space representation can be found in papers on modern shock - capturing mhd codes ( e.g. brio & wu 1988 ; ryu & jones 1995 ) . for the sake of simplicity
, we consider an isothermal plasma .
figure [ fig_modes ] and figure [ fig_modes - real ] give schematics of slow and fast waves . for slow and fast waves ,
@xmath88 , @xmath92 ( @xmath93 ) , and @xmath89 are in the same plane . on the other hand , for alfvn waves , the velocity of the fluid element @xmath94 is orthogonal to the @xmath95 plane . as before , the alfvn speed is @xmath96 where @xmath97 is the average density .
fast and slow speeds are @xmath98^{1/2},\ ] ] where @xmath99 is the angle between @xmath88 and @xmath89 .
see table [ cho_table1 ] for the definition of other variables .
when @xmath90 ( @xmath100=@xmath101 ; @xmath102= gas pressure , @xmath103= magnetic pressure ; hereinafter @xmath104 average @xmath105 ) goes to zero , we have @xmath106 figure [ fig_modes ] shows directions of displacement ( or , directions of velocity ) vectors for these three modes .
we will call them the basis vectors for these modes .
the alfvn basis is perpendicular to both @xmath84 and @xmath87 , and coincides with the azimuthal vector @xmath107 in a spherical - polar coordinate system . here
hatted vectors are unit vectors .
the fast basis @xmath108 lies _ between _ @xmath84 and @xmath85 : @xmath109 ^ 2 k_{\| } \hat{\bf k}_{\| } + k_{\perp } \hat{\bf k}_{\perp},\ ] ] where @xmath110 , and @xmath111 is the averaged @xmath90 ( = @xmath112 ) .
the slow basis @xmath113 lies _ between _ @xmath114 and @xmath87 ( = @xmath115 ) : @xmath116 ^ 2 k_{\perp } \hat{\bf k}_{\perp}.\ ] ] the two vectors @xmath108 and @xmath113 are mutually orthogonal .
proper normalizations are required for both bases to make them unit - length .
when @xmath90 goes to zero ( i.e. the magnetically dominated regime ) , @xmath108 becomes parallel to @xmath85 and @xmath113 becomes parallel to @xmath87 ( fig .
[ fig_modes]b ) .
the sine of the angle between @xmath87 and @xmath113 is @xmath117 .
when @xmath90 goes to infinity ( i.e. gas pressure dominated regime ) ) but finite , so that @xmath118 means the gas pressure @xmath119 .
] , @xmath108 becomes parallel to @xmath84 and @xmath113 becomes parallel to @xmath114 ( fig .
[ fig_modes]c ) .
this is the incompressible limit . in this limit , the slow mode is sometimes called the pseudo - alfvn mode ( goldreich & sridhar 1995 ) . here we address the issue of mode coupling in a low @xmath90 plasma .
it is reasonable to suppose that in the limit where @xmath120 turbulence for mach numbers ( @xmath121 ) less than unity should be largely similar to the exactly incompressible regime .
thus , lithwick & goldreich ( 2001 ) conjectured that the gs95 relations are applicable to slow and alfvn modes with the fast modes decoupled .
they also mentioned that this relation can carry on for low @xmath90 plasmas . for @xmath122 and @xmath123
, we are in the regime of hydrodynamic compressible turbulence for which no theory exists , as far as we know . in the diffuse interstellar medium @xmath90
is typically less than unity . in addition ,
it is @xmath124 or less for molecular clouds .
there are a few simple arguments suggesting that mhd theory can be formulated in the regime where the alfvn mach number ( @xmath125 ) is less than unity , although this is not a universally accepted assumption .
alfvn modes describe incompressible motions .
arguments in gs95 are suggestive that the coupling of alfvn to fast and slow modes will be weak .
consequently , we expect that in this regime the alfvn cascade should follow the gs95 scaling .
moreover the slow waves are likely to evolve passively ( lithwick & goldreich 2001 ) . for @xmath126
their nonlinear evolution should be governed by alfvn modes so that we expect the gs95 scaling for them as well .
the phase velocity of alfvn waves and slow waves depend on a factor of @xmath127 and this enables modulation of the slow waves by the alfvn ones . however , fast waves do not have this factor and therefore can not be modulated by the changes of the magnetic field direction associated with alfvn waves .
the coupling between the modes is through the modulation of the local alfvn velocity and therefore is weak . for alfvn mach number ( @xmath128 ) larger than unity a shock - type regime
is expected . however , generation of magnetic field by turbulence ( cho & vishniac 2000a ) is expected for such a regime .
it will make the steady state turbulence approach @xmath129 .
therefore in cho & lazarian ( 2002a ) we consider turbulence in the limit @xmath123 , @xmath130 , and @xmath131 .
for these simulations , we mostly used @xmath132 , @xmath133 , and @xmath134 .
the alfvn speed of the mean external field is similar to the rms velocity ( @xmath135 ) , and we used an isothermal equation of state .
although the scaling relations presented below are applicable to sub - alfvnic turbulence , we cautiously speculate that small scales of super - alfvnic turbulence might follow similar scalings .
boldyrev , nordlund , & padoan ( 2001 ) obtained energy spectra close to @xmath136 in solenoidally driven super - alfvnic supersonic turbulence simulations .
the spectra are close to the kolmogorov spectrum ( @xmath137 ) , rather than shock - dominated spectrum ( @xmath138 ) .
this result might imply that small scales of super - alfvnic mhd turbulence can be described by our sub - alfvnic model presented below , which predicts kolmogorov - type spectra for alfvn and slow modes .
alfvn modes are not susceptible to collisionless damping ( see spangler 1991 ; minter & spangler 1997 and references therein ) that damps slow and fast modes .
therefore , we mainly consider the transfer of energy from alfvn waves to compressible mhd waves ( i.e. to the slow and fast modes ) . , @xmath139 .
( _ right _ ) the ratio of @xmath140 to @xmath141 .
the stronger the external field ( @xmath142 ) is , the more suppressed the coupling is .
the ratio is not sensitive to @xmath90 . from cho &
lazarian ( 2002a ) , title="fig : " ] , @xmath139 .
( _ right _ ) the ratio of @xmath140 to @xmath141 . the stronger the external field ( @xmath142 ) is , the more suppressed the coupling is .
the ratio is not sensitive to @xmath90 . from cho &
lazarian ( 2002a ) , title="fig : " ] in cho & lazarian ( 2002a ) , we carry out simulations to check the strength of the mode - mode coupling .
we first obtain a data cube from a driven compressible numerical simulation with @xmath143 .
then , after turning off the driving force , we let the turbulence decay . we go through the following procedures before we let the turbulence decay .
we first remove slow and fast modes in fourier space and retain only alfvn modes .
we also change the value of @xmath88 preserving its original direction .
we use the same constant initial density @xmath97 for all simulations .
we assign a new constant initial gas pressure @xmath102 . and
@xmath102 preserve the alfvn character of perturbations .
in fourier space , the mean magnetic field ( @xmath88 ) is the amplitude of @xmath144 component .
alfvn components in fourier space are for @xmath145 and their directions are parallel / anti - parallel to @xmath146 (= @xmath147 ) .
the direction of @xmath146 does not depend on the magnitude of @xmath142 or @xmath102 . ] after doing all these procedures , we let the turbulence decay . we repeat the above procedures for different values of @xmath142 and @xmath102 .
[ fig_coupling]a shows the evolution of the kinetic energy of a simulation .
the solid line represents the kinetic energy of alfvn modes .
it is clear that alfvn waves are poorly coupled to the compressible modes , and do not generate them efficiently therefore , we expect that alfvn modes will follow the same scaling relation as in the incompressible case .
[ fig_coupling]b shows that the coupling gets weaker as @xmath142 increases : @xmath148 the ratio of @xmath149 to @xmath150 is proportional to @xmath151 . ) for alfven velocity shows anisotropy similar to the gs95 .
conturs represent eddy shapes . from cho &
lazarian ( 2002a ) .
, title="fig : " ] ) for alfven velocity shows anisotropy similar to the gs95 .
conturs represent eddy shapes . from cho &
lazarian ( 2002a ) .
, title="fig : " ] this marginal coupling is in good agreement with a claim in gs95 , as well as earlier numerical studies where the velocity was decomposed into a compressible component @xmath152 and a solenoidal component @xmath153 .
the compressible component is curl - free and parallel to the wave vector @xmath89 in fourier space .
the solenoidal component is divergence - free and perpendicular to @xmath89 .
the ratio @xmath154 is an important parameter that determines the strength of any shock ( passot et al .
1988 ; pouquet 1999 ) .
porter , woodward , & pouquet ( 1998 ) performed a hydrodynamic simulation of decaying turbulence with an initial sonic mach number of unity and found that @xmath155 evolves toward @xmath156 .
matthaeus et al .
( 1996 ) carried out simulations of decaying weakly compressible mhd turbulence ( zank & matthaeus 1993 ) and found that @xmath157 , where @xmath158 is the sonic mach number . in boldyrev
( 2001 ) a weak generation of compressible components in solenoidally driven super - alfvnic supersonic turbulence simulations was obtained .
[ fig_alf ] shows that the spectrum and the anisotropy of alfvn waves in this limit are compatible with the gs95 model : @xmath159 and scale - dependent anisotropy @xmath160 that is compatible with the gs95 theory .
slow waves are somewhat similar to pseudo - alfvn waves ( in the incompressible limit ) .
first , the directions of displacement ( i.e. @xmath161 ) of both waves are similar when anisotropy is present .
the vector @xmath162 is always between @xmath114 and @xmath115 . in figure [ fig_modes ]
, we can see that the angle between @xmath114 and @xmath115 gets smaller when @xmath163 .
therefore , when there is anisotropy ( i.e. @xmath163 ) , @xmath113 of a low @xmath90 plasma becomes similar to that of a high @xmath90 plasma .
second , the angular dependence in the dispersion relation @xmath164 is identical to that of pseudo - alfvn waves ( the only difference is that , in slow waves , the sound speed @xmath165 is present instead of the alfvn speed @xmath166 ) .
goldreich & sridhar ( 1997 ) argued that the pseudo - alfvn waves are slaved to the shear - alfvn ( i.e. ordinary alfvn ) waves in the presence of a strong @xmath88 , meaning that the energy cascade of pseudo - alfvn modes is primarily mediated by the shear - alfven waves .
this is because pseudo - alfvn waves do not provide efficient shearing motions .
similar arguments are applicable to slow waves in a low @xmath90 plasma ( cho & lazarian 2002a ) ( see also lithwick & goldreich 2001 for high-@xmath90 plasmas ) . as a result
, we conjecture that slow modes follow a scaling similar to the gs95 model ( cho & lazarian 2002a ) : @xmath167 fig .
[ fig_slow]a shows the spectra of slow modes . for velocity ,
the slope is close to @xmath168 .
[ fig_slow]b shows the contours of equal second - order structure function ( @xmath169 ) of slow velocity , which are compatible with @xmath160 scaling . in low @xmath90 plasmas ,
density fluctuations are dominated by slow waves ( cho & lazarian 2002a ) . from the continuity equation
@xmath170 @xmath171 we have , for slow modes , @xmath172 hence , this simple argument implies @xmath173 where we assume that @xmath174 and @xmath158 is the mach number . on the other hand ,
only a small amount of magnetic field is produced by the slow waves .
similarly , using the induction equation ( @xmath175 ) , we have @xmath176 which means that equipartition between kinetic and magnetic energy is not guaranteed in low @xmath90 plasmas .
in fact , in fig .
[ fig_slow]a , the power spectrum for density fluctuations has a much larger amplitude than the magnetic field power spectrum .
since density fluctuations are caused mostly by the slow waves and magnetic fluctuation is caused mostly by alfvn and fast modes , we _ do not _ expect a strong correlation between density and magnetic field , which agrees with the ism simulations ( padoan & nordlund 1999 ; ostriker et al .
2001 ; vazquez - semadeni 2002 ) .
figure [ fig_fast ] shows fast modes are isotropic .
the resonance conditions for interacting fast waves are : @xmath177 since @xmath178 for the fast modes , the resonance conditions can be met only when all three @xmath89 vectors are collinear .
this means that the direction of energy cascade is _ radial _ in fourier space , and we expect an isotropic distribution of energy in fourier space . using the constancy of energy cascade and uncertainty principle
, we can derive an ik - like energy spectrum for fast waves .
the constancy of cascade rate reads @xmath179 on the other hand , @xmath180 can be estimated as @xmath181 if contributions are random , the denominator can be written by the square root of the number of interactions ( @xmath182 ) times strength of individual interactions ( @xmath183 ) . , where @xmath99 is the angle between @xmath89 and @xmath88 .
thus marginal anisotropy is expected .
it will be investigated elsewhere . ] here we assume locality of interactions : @xmath184 . due to the uncertainly principle
, the number of interactions becomes @xmath185 , where @xmath186 is the typical transversal ( i.e. not radial ) separation between two wave vectors @xmath187 and @xmath188 ( with @xmath189 ) .
therefore , the denominator of equation ( [ eq14_fast ] ) is @xmath190 .
we obtain an independent expression for @xmath180 from the uncertainty principle ( @xmath191 with @xmath192 ) . from this and equation ( [ eq14_fast ] )
, we get @xmath193 which yields @xmath194 combining equations ( [ cas_rate_fast ] ) and ( [ t_cascade ] ) , we obtain @xmath195 or @xmath196 .
this is very similar to acoustic turbulence , turbulence caused by interacting sound waves ( zakharov 1967 ; zakharov & sagdeev 1970 ; lvov , lvov , & pomyalov 2000 ) .
zakharov & sagdeev ( 1970 ) found @xmath197 . however , there is debate about the exact scaling of acoustic turbulence .
here we cautiously claim that our numerical results are compatible with the zakharov & sagdeev scaling : @xmath198 magnetic field perturbations are mostly affected by fast modes ( cho & lazarian 2002a ) when @xmath90 is small : @xmath199 if @xmath200 .
the turbulent cascade of fast modes is expected to be slow and in the absence of collisionless damping they are expected to propagate in turbulent media over distances considerably larger than alfvn or slow modes .
this effect is difficult to observe in numerical simulations with @xmath201 .
a modification of the spectrum in the presence of the collisionless damping is presented in yan & lazarian ( 2002 ) .
many astrophysical problems require some knowledge of the scaling properties of turbulence . therefore we expect a wide range of applications of the established scaling relations . here
we show how recent progress in understanding mhd turbulence affects cosmic ray propagation .
the propagation of cosmic rays is mainly determined by their interactions with electromagnetic fluctuations in interstellar medium .
the resonant interaction of cosmic ray particles with mhd turbulence has been repeatedly suggested as the main mechanism for scattering and isotropizing cosmic rays . in these analysis
it is usually assumed that the turbulence is _ isotropic _ with a kolmogorov spectrum ( see schlickeiser & miller 1998 ) .
how should these calculations be modified ?
consider resonance interaction first .
particles moving with velocity @xmath41 get into resonance with mhd perturbations propagating along the magnetic field if the resonant condition is fulfilled , namely , @xmath202 ( @xmath203 ) , where @xmath204 is the wave frequency , @xmath205 is the gyrofrequency of relativistic particle , @xmath206 , where @xmath207 is the pitch angle of particles . in other words ,
resonant interaction between a particle and the transverse electric field of a wave occurs when the doppler shifted frequency of the wave in the particle s guiding center rest frame @xmath208 is a multiple of the particle gyrofrequency . for cosmic rays , @xmath209 ,
so the slow variation of the magnetic field with time can be neglected .
thus the resonant condition is simply @xmath210 . from this resonance condition , we know that the most important interaction occurs at @xmath211 .
it is intuitively clear that resonant interaction of particles in isotropic and anisotropic turbulence should be different .
chandran ( 2001 ) obtained strong suppression of scattering by alfvenic turbulence when he treated turbulence anisotropies in the spirit of goldreich - sridhar model of incompressible turbulence .
his treatment was improved in yan & lazarian ( 2002 , henceforth yl02 ) who used a more rigorous description of magnetic field statistics .
moreover , they took into account cr scattering by compressible mhd modes and found that fast modes absolutely dominate cosmic ray scattering . in our description
we shall follow yl02 treatment of the problem .
we employ quasi - linear theory ( qlt ) to obtain our estimates .
qlt has been proved to be a useful tool in spite of its intrinsic limitations ( chandran 2000 ; schlickeiser & miller 1998 ; miller 1997 ) . for moderate energy cosmic rays ,
the corresponding resonant scales are much smaller than the injection scale .
therefore the fluctuation on the resonant scale @xmath212 even if they are comparable at the injection scale .
qlt disregards diffusion of cosmic rays that follow wandering magnetic field lines ( jokipii 1966 ) and this diffusion should be accounted separately . obtained by applying the qlt to the collisionless boltzmann - vlasov equation , the fokker - planck equation is generally used to describe the evolvement of the gyrophase - average distribution function @xmath213 , @xmath214,\ ] ] where @xmath215 is particle momentum .
the fokker - planck coefficients @xmath216 are the fundamental physical parameter for measuring the stochastic interactions , which are determined by the electromagnetic fluctuations ( schlickeiser & achatz 1993 ) . from ohm
s law @xmath217 we can get the electromagnetic fluctuations from correlation tensors of magnetic and velocity fluctuations @xmath218 @xmath219 . here ,
@xmath220 for alfven modes , cho , lazarian and vishniac ( 2002 ) obtained @xmath221 where @xmath222 is a 2d matrix in x - y plane , @xmath223 is the wave vector along the local mean magnetic field , @xmath224 is the wave vector perpendicular to the magnetic field and the normalization constant @xmath225 .
we assume that for the alfven modes @xmath226 @xmath227 where the fractional helicity @xmath228 is independent of @xmath229 ( chandran 2000 ) . according to cho & lazarian ( 2002a ) ,
fast modes are isotropic and have one dimensional spectrum @xmath197 . in low @xmath230 medium ,
the velocity fluctuations are always perpendicular to @xmath231 for all @xmath229 , while the magnetic fluctuations are perpendicular to @xmath229 .
thus @xmath232 @xmath233 of fast modes are not equal , @xmath234={\frac{l^{-1/2}}{8\pi } } j_{ij}k^{-7/2}\left[\begin{array}{c } \cos ^{2}\theta \\ \sigma \cos ^{2}\theta \\
1\end{array } \right],\label{fast_tensor_lowb}\ ] ] where @xmath235 is also a 2d tensor in @xmath236 plane .
@xmath237 are 3d matrixes . however , the third dimension is not needed for our calculations .
@xmath233 is different from that in schlickeiser & miller ( 1998 ) .
the fact that the fluctuations @xmath238 in fast modes are in the @xmath229-@xmath239 plane place another constrain on the tensor so that the term @xmath240 does nt exist . ] in high @xmath230 medium , the velocity fluctuations are radial , i.e. , along the direction of @xmath241 .
fast modes in this regime are essentially sound waves compressing magnetic field ( gs95 ; lithwick & goldreich 2001 , cho & lazarian , in preparation ) .
the compression of magnetic field depends on plasma @xmath230 .
the corresponding x - y components of the tensors are @xmath234={\frac{l^{-1/2}}{8\pi } } \sin ^{2}\theta j_{ij}k^{-7/2}\left[\begin{array}{c } \cos ^{2}\theta /\beta \\
\sigma \cos \theta /\beta ^{1/2}\\ 1\end{array } \right].\label{fast_tensor_highb}\ ] ] adopting the approach in schlickeiser & achatz ( 1993 ) , we can obtain the fokker - planck coefficients in the lowest order approximation of @xmath242 , @xmath243={\frac{\omega ^{2}(1-\mu ^{2})}{2b_{0}^{2}}}\left[\begin{array}{c } 1\\ mc\\ m^{2}c^{2}\end{array } \right]{\mathcal{r}}e\sum _ { n=-\infty } ^{n=\infty } \int _ { k_{min}}^{k_{max}}dk^{3 } & & \nonumber \\ \int _ { 0}^{\infty } dte^{-i(k_{\parallel } v_{\parallel } -\omega + n\omega ) t}\left\ { j_{n+1}^{2}({\frac{k_{\perp } v_{\perp } } { \omega } } ) \left[\begin{array}{c } p_{{\mathcal{rr}}}({\mathbf{k}})\\ t_{{\mathcal{rr}}}({\mathbf{k}})\\ r_{{\mathcal{rr}}}({\mathbf{k}})\end{array } \right]\right . &
& \nonumber \\ + j_{n-1}^{2}({\frac{k_{\perp } v_{\perp } } { \omega } } ) \left[\begin{array}{c } p_{{\mathcal{ll}}}({\mathbf{k}})\\ -t_{{\mathcal{ll}}}({\mathbf{k}})\\ r_{{\mathcal{ll}}}({\mathbf{k}})\end{array } \right]+j_{n+1}({\frac{k_{\perp } v_{\perp } } { \omega } } ) j_{n-1}({\frac{k_{\perp } v_{\perp } } { \omega } } ) & & \nonumber \\
\left.\left[e^{i2\phi } \left[\begin{array}{c } -p_{{\mathcal{rl}}}({\mathbf{k}})\\ t_{{\mathcal{rl}}}({\mathbf{k}})\\ r_{{\mathcal{rl}}}({\mathbf{k}})\end{array } \right]+e^{-i2\phi } \left[\begin{array}{c } -p_{{\mathcal{lr}}}({\mathbf{k}})\\ -t_{{\mathcal{lr}}}({\mathbf{k}})\\ r_{{\mathcal{lr}}}({\mathbf{k}})\end{array } \right]\right]\right\ } & & \label{genmu}\end{aligned}\ ] ] where @xmath244 , @xmath245 corresponds to the dissipation scale , @xmath246 is the relativistic mass of the proton , @xmath247 is the particle s velocity component perpendicular to @xmath231 , @xmath248 @xmath249 represent left and right hand polarization , the expression is only true for alfv\en modes .
there are additional compressional terms for compressable modes . ] . noticing that the integrand for small @xmath224 is substantially suppressed by the exponent in the anisotropic tensor ( see eq .
( [ anisotropic ] ) ) so that the large scale contribution is not important , we can simply use the asymptotic form of bessel function for large argument .
then if the pitch angle @xmath207 is not close to 0 , we can derive the analytical result for anisotropic turbulence ( yl02 ) , @xmath250=\frac{v^{2.5}\cos \alpha ^{5.5}}{2\omega ^{1.5}l^{2.5}\sin \alpha } \gamma [ 6.5,k_{max}^{-2/3}k_{res}l^{1/3}]\left[\begin{array}{c } 1\\ \sigma mv_{a}\\ m^{2}v_{a}^{2}\end{array } \right],\label{ana}\ ] ] where @xmath251 is the injection sale , @xmath245 corresponds to the dissipation scale , @xmath252 $ ] represents the incomplete gamma function .
the scattering frequency @xmath253 is plotted for different models in fig.([fig : incom]a ) .
it is clear that anisotropy suppresses scattering .
although our results are larger than those obtained in chandran ( 2001 ) using an _ ad hoc _ tensor with a step function , they are still much smaller than the estimates for isotropic model . unless we consider very high energy crs ( @xmath254 ) with corresponding larmor radius comparable to the injection scale , we can neglect scattering by the alfvnic turbulence .
what is the alternative way to scatter cosmic rays ? for compressible modes we discuss two types of resonant interaction : gyroresonance and transit - time damping ;
the latter requires longitudinal motions .
the contribution from slow modes is not larger than that by alfvn modes since the slow modes have the similar anisotropies and scalings .
more promising are fast modes , which are isotropic ( cho & lazarian 2002a ) . however , fast modes are subject to collisionless damping if the wavelength is smaller than the proton mean free path or by viscous damping if the wavelength is larger than the mean free path . according to cl02 , fast modes cascade over time scales @xmath255 where @xmath256 is the eddy turn - over time , @xmath257 is the turbulence velocity at the injection scale .
consider gyroresonance scattering in the presence of collisionless damping .
the cutoff of fast modes corresponds to the scale where @xmath258 and this defines the cutoff scale @xmath259 . using the tensors given in eq .
( [ fast_tensor_lowb ] ) we obtain the corresponding fokker - planck coefficients for the crs interacting with fast modes by integrating eq.([genmu ] ) from @xmath260 to @xmath261 ( see fig.([fig : incom]b ) ) . when @xmath259 is less than @xmath262 , the results of integration for damped and undamped turbulence coincides .
since the damping increases with @xmath230 , the scattering frequency decreases with @xmath230 . adopting the tensors given in eq .
( [ fast_tensor_highb ] ) , it is possible to calculate the scattering frequency of crs in high @xmath230 medium .
for instance , for density @xmath263@xmath264 temperature @xmath265k , magnetic field
@xmath266 g , the mean free path is smaller than the resonant wavelength for the particles with energy larger than @xmath267 , therefore collisional damping rather than landau damping should be taken into account .
nevertheless , our results show that the fast modes still dominate the crs scattering in spite of the viscous damping .
apart from the gyroresonance , fast modes potentially can scatter crs by transit - time damping ( ttd ) ( schlickeiser & miller 1998 ) .
ttd happens due to the resonant interaction with parallel magnetic mirror force @xmath268 . for small amplitude waves
, particles should be in phase with the wave so as to have a secular interaction with wave .
this gives the cherenkov resonant condition @xmath269 , corresponding to the @xmath270 term in eq.([genmu ] ) .
from the condition , we see that the contribution is mostly from nearly perpendicular propagating waves ( @xmath271 ) . according to eq .
( [ fast_tensor_lowb]),we see that the corresponding correlation tensor for the magnetic fluctuations @xmath233 are very small , so the contribution from ttd to scattering is not important .
self - confinement due to the streaming instability has been discussed by different authors(see cesarsky 1980 , longair 1997 ) as an effective alternative to scatter crs and essential for cr acceleration by shocks .
however , we will discuss in our next paper that in the presence of the turbulence the streaming instability will be partially suppressed owing to the nonlinear interaction with the background turbulence .
thus the gyroresonance with the fast modes is the principle mechanism for scattering cosmic rays .
this demands a substantial revision of cosmic ray acceleration / propagation theories , and many related problems may need to be revisited . for instance , our results may be relevant to the problems of the boron to carbon abundances ratio .
we shall discuss the implications of the new emerging picture elsewhere .
recently there have been significant advances in the field of compressible mhd turbulence and its implications to cosmic ray transport : 1 .
simulations of compressible mhd turbulence show that there is a weak coupling between alfvn waves and compressible mhd waves and that the alfvn modes follow the goldreich - sridhar scaling .
fast modes , however , decouple and exhibit isotropy .
2 . scattering of cosmic rays by alfvenic modes is suppressed and therefore the scattering by fast modes is the dominant process provided that turbulent energy is injected at large scales .
3 . the scattering frequency by fast modes depends on collisionless damping or viscous damping and therefore on plasma @xmath230 . *
acknowledgments : * we acknowledge the support of the nsf through grant ast-0125544 .
this work was partially supported by national computational science alliance under ast000010n and utilized the ncsa sgi / cray origin2000 .
lazarian , a. 1999b , in plasma turbulence and energetic particles in astrophysics , ed .
m. ostrowski & r. schlickeiser ( cracow , poland : obserwatorium astronomiczne , uniwersytet jagiellonski ) , 28 ( astro - ph/0001001 ) stone , j. m. , ostriker , e. c. , & gammie , c. f. 1998 , apj , 508 , l99 swordy , s. p. , 2001
, space science rev .
, 99 , 85 vazquez - semadeni , e. 2002 , in seeing through the dust , ed .
r. taylor , t. landecker , & a. willis ( san francisco : asp ) ( astro - ph/0201072 ) | turbulence is the most common state of astrophysical flows . in typical astrophysical fluids ,
turbulence is accompanied by strong magnetic fields , which has a large impact on the dynamics of the turbulent cascade .
recently , there has been a significant breakthrough on the theory of magnetohydrodynamic ( mhd ) turbulence . for the first time we have a scaling model that is supported by both observations and numerical simulations .
we review recent progress in studies of both incompressible and compressible turbulence .
we compare iroshnikov - kraichnan and goldreich - sridhar models , and discuss scalings of alfvn , slow , and fast waves .
we discuss the implications of this new insight into mhd turbulence for cosmic ray transport . |
in recent years , the standard model of electroweak interactions has been confirmed experimentally with outstanding success .
not only was almost all the particle content discovered at accelerator experiments , but their properties and interactions have been measured with high precision , in agreement with the model prediction . the only missing piece is the higgs boson , which is responsible for electroweak symmetry breaking .
however , even today we can obtain meaningful constraints on the higgs boson mass from electroweak precision measurements . due to the impressive accuracy of some of these experimental results ,
they are sensitive to electroweak radiative corrections at the next - to - leading ( nlo ) and sometimes next - to - next - to - leading ( nnlo ) level , and thus depend on the impact of the higgs boson entering in the loops .
two of the most important quantities in this respect are the mass of the @xmath0 boson , @xmath1 and the sine of the leptonic effective weak mixing angle @xmath2 .
the @xmath0-boson mass can be inferred from the muon decay constant @xmath3 , which is generated through virtual w - boson exchange , so that @xmath4 .
the effective weak mixing angle , on the other hand , reflects the ratio of the vector and axial - vector couplings , @xmath5 and @xmath6 , of the @xmath7 boson to fermions ( @xmath8 ) at the @xmath7 boson pole : @xmath9 since these couplings can be measured most precisely for leptons , the _ leptonic _ effective weak mixing angle @xmath2is usually taken as a reference .
the current experimental world average for the @xmath0-boson mass is @xmath10 gev @xcite .
recently , a lot of progress has been made towards establishing accurate theoretical prediction for @xmath1 .
the best result @xcite includes the complete two - loop corrections @xcite and some three - loop contributions @xcite .
the remaining theoretical error is estimated to be @xmath11 mev , which is well below the current experimental uncertainty .
still , the electroweak two - loop corrections total to @xmath12 mev and are thus mandatory for electroweak precision analyses . the effective weak mixing angle @xmath2 is mainly derived from various asymmetries measured around the @xmath7 boson peak at @xmath13 colliders after subtraction of qed effects .
the current experimental accuracy , @xmath14 @xcite implies strong indirect constraints on the allowed range for the higgs boson mass @xmath15 .
therefore it is important to develop precise theoretical calculations for this quantity .
usually , @xmath2 is computed as a function of the electromagnetic coupling @xmath16 , the muon constant @xmath3 and the masses of the @xmath7 boson , @xmath17 , and the top quark , @xmath18 ( other fermion masses are numerically irrelevant ) .
as explained before , @xmath1 is calculated from @xmath3 , but in addition to these corrections , the computation of @xmath2 involves the corrections to the @xmath7 vertex form factors .
the latter are expressed by the quantity @xmath19 , in such a way that the effective weak mixing angle can also be written as : @xmath20 and at tree - level , @xmath21 .
higher - order corrections to @xmath2 have been under extensive theoretical study over the last two decades . besides the one - loop result @xcite , two- and three - loop qcd corrections are available @xcite , but for the electroweak two - loop contributions only partial results were known . by means of a large mass expansion in the heavy top quark mass , the formally leading @xmath22 @xcite and next - to - leading @xmath23 @xcite terms were computed .
a part of the missing two - loop contributions was incorporated by the complete electroweak two - loop corrections to @xmath1 @xcite . while these corrections effected a shift in @xmath1 of 4 mev compared to the previously known @xmath24 contributions , the induced shift in @xmath2 was very sizable , @xmath25 , thus implying that the missing two - loop terms in the form factor @xmath26 can be of similar order .
as a first step towards completing the electroweak two - loop corrections to @xmath2 , results for the fermionic ( i.e. diagrams with closed fermion loops ) two - loop corrections were presented recently @xcite .
the genuine two - loop vertex diagrams are represented by the generic topologies in fig .
[ fig : diags ] .
higher - order corrections to the process @xmath27 near the @xmath7 pole can be consistently computed by performing an expansion of the amplitude around the complex pole @xmath28 , @xmath29 = \frac{r}{s-{\cal m}_{\rm z}^2 } + s + ( s-{\cal m}_{\rm z}^2 ) s ' + \dots \label{eq : polexp}\ ] ] here @xmath30 is the @xmath7 decay width . after subtracting contributions from s - channel photon exchange and @xmath31-@xmath7 interference , the vertex corrections form factor at nnlo
is derived to be @xmath32 where the superscripts in parentheses indicate the loop order . in this quantity ,
ir - divergencies from qed contributions drop out , which involves a delicate interplay between one- and two - loop terms in the form factors .
the uv - divergencies are cancelled by on - shell renormalization .
the relevant counterterms are derived using the methods of ref .
@xcite .
the new part of this work is the computation of the two - loop @xmath33 vertex corrections , which are treated with two independent technical methods .
the first method uses large mass expansions for the diagrams with internal top - quark lines and the differential equation method for diagrams with only light fermions @xmath34 , the masses of which are neglected .
contrary to previous work @xcite , the expansion in @xmath35 is performed to high precision , by executing the series to the order @xmath36 , reaching an overall relative precision of @xmath37 of the final result .
the coefficients of the expansion are 2-loop tadpole and 1-loop vertex diagrams , which can be evaluated efficiently with well - known analytical formulae .
the contributions from diagrams without top - quark propagators involve only two independent scales , @xmath17 and @xmath1 , allowing a fully analytical treatment . even for the limited set of diagrams with closed fermions loops ,
a large number of scalar integrals with non - trivial structures in the numerator are involved .
they can be reduced to a set of scalar master integrals by using integration - by - parts identities @xcite . owing to size of the linear equation system associated with this reduction procedure
, the algorithm has been implemented in the dedicated c++ library diagen / idsolver @xcite , which performs the necessary steps in a highly automized way .
analytical results for these master integrals are obtained by the differential equation method @xcite .
this is illustrated by the following example , @xmath38 = \frac{p^2}{p^2+m^2 } \biggl ( \frac{4-d}{2}(4 + 5 \frac{m^2}{p^2 } ) \left [ \raisebox{-5.4mm}{\psfig{figure = v2p1.ps , width=1.6cm}}\;\right ] + \frac{10 - 3d}{2 } \left [ \raisebox{-5.4mm}{\psfig{figure = v2p0.ps , width=1.6cm}}\;\right ] - \frac{2-d}{2 } \left [ \raisebox{-3mm}{\psfig{figure = t134.ps , width=0.95cm}}\right ] \biggr).\ ] ] here the thick lines represent massive propagators with mass @xmath39 , the thin lines denote massless propagators and @xmath40 is the momentum flowing into the vertex .
@xmath41 is the dimension of dimensional regularization .
the momentum derivate of the scalar integrals on the left - hand side results in the same integral and simpler integral topologies on the right - hand side .
feeding in analytical results for these simpler integrals , the differential equation can be solved in terms of generalized polylogarithms .
all integrals were also checked by using low - momentum expansions .
the second method makes use of numerical integrations based on dispersion relations .
a scalar two - loop integral with a self - energy sub - loop as in fig . [ fig : disp ] ( a ) can be expressed as @xcite ' '' '' [ cols="^,^ " , ] @xmath42 where @xmath43 is the discontinuity of the scalar one - loop self - energy function .
the second integral can be evaluated into a standard @xmath44-point one - loop function , leaving the integration over @xmath45 to be performed numerically . in general
, one can also introduce dispersion relations for triangle sub - loops @xcite , but it is often technically easier to reduce them to self - energy sub - loops by introducing feynman parameters @xcite , @xmath46^{-1 } \ ; [ ( q+p_2)^2-m_2 ^ 2]^{-1 } = \int_0 ^ 1 { \rm d}x \ ; [ ( q+\bar{p})^2 - \overline{m}^2]^{-2 } \\
\bar{p } = x\,p_1 + ( 1-x)p_2 , \qquad \overline{m}^2 = x \ , m_1 ^ 2 + ( 1-x ) m_2 ^ 2 - x(1-x)(p_1-p_2)^2 .
\end{array}\ ] ] this is indicated diagrammatically in fig .
[ fig : disp ] ( b ) .
the integration over the feynman parameters is also performed numerically . as a result
, all master integrals for the vertex topologies can be evaluated by at most 3-dim . numerical integrations . before performing the numerical integrations , possible
uv- and ir - divergencies need to be subtracted from the integrals . while this second method is applicable to two - loop vertex corrections with an arbitrary number of mass scales , it is slower and leads to much large expressions than the first method
. nevertheless it provides an important check of the result .
special caution is needed for the diagrams with a fermion triangle loop ( see the third diagram in fig .
[ fig : diags ] ) , which involve the @xmath47 matrix . in dimensional regularization , it is not possible to fulfill the two relations @xmath48 and tr@xmath49 simultaneously . as in ref .
@xcite , the contributions resulting in @xmath50-tensors were therefore evaluated in four dimensions , based on the observation that these terms are free of uv - divergencies .
potential soft and collinear divergencies of single diagrams are regulated using a photon mass , with a subsequent careful expansion for zero photon masses .
the new result for the fermionic two - loop corrections is combined with corrections of order @xmath51 , @xmath52 @xcite , @xmath53 @xcite and leading three - loop terms of @xmath54 and @xmath55 @xcite .
reducible terms of the same order are taken into account , but no resummations are preformed .
the most precise prediction of @xmath2 is obtained as a function of the muon decay constant @xmath3 , from which the @xmath0-boson mass is calculated by including the radiative corrections to @xmath1 as given in ref .
@xcite .
[ fig : res ] shows the final result for @xmath2 as a function of the higgs mass compared to the current experimental value .
included in the plot are the error bands due to the uncertainties of the experimental input parameters entering into the theoretical prediction and of the direct measurement of @xmath2 . as evident from the figure , the identification of computation and measurement for @xmath2 favor relatively small values for @xmath15 . with the new two - loop corrections , the best - fit value for @xmath15 moved from 148 gev ( using formulae of ref .
@xcite ) to 168 gev ( for @xmath56 gev ) .
the numerical result for the leptonic effective weak mixing angle has been published in ref .
@xcite as a parametric fitting formula that is accurate in the range 10 gev @xmath57 1 tev .
it has also been implemented into the newest version 6.42 of the program zfitter @xcite , with some changes recently discussed in ref .
@xcite , and was used in the latest release of electroweak precision global fits of the standard model .
the impact of the new result for @xmath2 shifts the 95% confidence level upper bound on the higgs mass upwards by 23 gev to 260 gev @xcite .
together with the inclusion of the new two - loop result in the prediction for @xmath2 , a assessment of the uncertainties from missing higher order contributions is required . since for practical purposes
@xmath2 is given as a function of the muon decay constant @xmath3 , it is useful to evaluate the theoretical error for this parametrization , i.e. combining the radiative corrections to @xmath1 and the @xmath7 vertex .
a simple method to estimate the higher order uncertainties assumes that the perturbation series follows roughly a geometric progression .
this presumption implies relations like @xmath58 with this method one obtains the following errors for @xmath15 between 10 and 1000 gev in units of @xmath59 : between 2.3 and 2.0 for the @xmath60 contributions beyond the leading @xmath61 term , between 1.8 and 2.5 for @xmath62 , between 1.1 and 1.0 for @xmath63 and between 1.7 and 2.4 for @xmath64 .
the missing bosonic @xmath65 corrections can not be appraised from geometric progression . however , considering they have a prefactor @xmath66 but no specific enhancement factor , they are estimated to be about @xmath67 . to account for possible deviations from the geometric series behavior , an overall factor
@xmath68 was included to arrive at a total error of @xmath69 .
alternatively , the error from a higher - order qcd loop can be assessed by varying the scale of the strong coupling constant @xmath70 or the top - quark mass @xmath18 in the @xmath71 scheme in the highest available perturbation order .
the scale variation leads to an error estimate of 0.1 to @xmath72 for the @xmath73 corrections and of less than @xmath74 for the @xmath75 contributions .
these numbers are of the same order as the estimated errors from the geometric progression method , so that the total error given above seems to be fairly reliable .
the new error estimate was used in the latest electroweak global fits @xcite and lead to a reduction of the width of the well - known _ blue band _ , which indicates the theoretical error in the indirect determination of the higgs mass .
in this contribution , recent progress in the calculation of higher - order corrections to the most important electroweak precision observables and their impact on the indirect determination of the higgs mass was reported .
the complete fermionic @xmath76 corrections to the leptonic effective weak mixing angle @xmath2 have been calculated and numerical results were presented . as an additional check ,
the computation of the two - loop vertex integrals was performed with two independent methods .
the new result , together with an estimate of the remaining theoretical error , was included in the latest version of the program zfitter and used for the latest global electroweak fits of the standard model .
the calculation of the remaining bosonic electroweak two - loop corrections is currently in progress and will be available soon @xcite .
furthermore , we are working on adapting the new results for quark final states , where particular attention has to be paid to the @xmath77 vertex , since it includes additional massive top - quark propagators @xcite .
the authors are grateful to g. weiglein for valuable contributions , comments and communications concerning this project .
m. c. was supported by the sofja kovalevskaja award of the alexander von humboldt foundation sponsored by the german federal ministry of education and research .
m. a. and m. c. were supported by the polish state committee for scientific research ( kbn ) for the research project in years 2004 - 2005 .
99 the lep electroweak working group and the sld electroweak and heavy flavour groups , hep - ex/0412015 .
m. awramik , m. czakon , a. freitas and g. weiglein , phys .
d * 69 * , 053006 ( 2004 ) .
a. freitas , w. hollik , w. walter and g. weiglein , phys .
b * 495 * , 338 ( 2000 ) [ erratum - ibid .
b * 570 * , 260 ( 2003 ) ] , nucl .
b * 632*,189 ( 2002 ) [ erratum - ibid .
b * 666 * , 305 ( 2003 ) ] ; m. awramik and m. czakon , phys .
b * 568 * , 48 ( 2003 ) .
m. awramik and m. czakon , phys .
lett . * 89 * , 241801 ( 2002 ) ; a. onishchenko and o. veretin , phys .
b * 551 * , 111 ( 2003 ) ; m. awramik , m. czakon , a. onishchenko and o. veretin , phys .
d * 68 * , 053004 ( 2003 ) .
k. g. chetyrkin , j. h. khn and m. steinhauser , phys .
lett . * 75 * , 3394 ( 1995 ) .
m. faisst , j. h. khn , t. seidensticker and o. veretin , nucl .
b * 665 * , 649 ( 2003 ) .
a. sirlin , phys .
d * 22 * , 971 ( 1980 ) ; + w. j. marciano and a. sirlin , phys . rev .
d * 22 * , 2695 ( 1980 ) [ erratum - ibid .
d * 31 * , 213 ( 1980 ) ] .
g. degrassi and a. sirlin , nucl .
b * 352 * , 342 ( 1991 ) ; p. gambino and a. sirlin ,
d * 49 * , 1160 ( 1994 ) .
a. djouadi and c. verzegnassi , phys .
b * 195 * , 265 ( 1987 ) ; a. djouadi , nuovo cim .
a * 100 * , 357 ( 1988 ) ; b. a. kniehl , nucl .
b * 347 * , 86 ( 1990 ) ; f. halzen and b. a. kniehl , nucl .
b * 353 * , 567 ( 1991 ) ; b. a. kniehl and a. sirlin , nucl .
b * 371 * , 141 ( 1992 ) , phys .
d * 47 * , 883 ( 1993 ) ; a. djouadi and p. gambino , phys .
d * 49 * , 3499 ( 1994 ) [ erratum - ibid .
d * 53 * , 4111 ( 1996 ) ] .
k. g. chetyrkin , j. h. khn and m. steinhauser , nucl .
b * 482 * , 213 ( 1996 ) .
j. j. van der bij and f. hoogeveen , nucl .
b * 283 * , 477 ( 1987 ) .
r. barbieri , m. beccaria , p. ciafaloni , g. curci and a. vicere , phys .
b * 288 * , 95 ( 1992 ) [ erratum - ibid .
b * 312 * , 511 ( 1993 ) ] , nucl .
b * 409 * , 105 ( 1993 ) ; j. fleischer , o. v. tarasov and f. jegerlehner , phys .
b * 319 * , 249 ( 1993 ) , phys .
d * 51 * , 3820 ( 1995 ) .
g. degrassi , p. gambino and a. sirlin , phys .
b * 394 * , 188 ( 1997 ) .
m. awramik , m. czakon , a. freitas and g. weiglein , phys .
* 93 * , 201805 ( 2004 ) .
k. g. chetyrkin and f. v. tkachov , nucl .
b * 192 * , 159 ( 1981 ) ; s. laporta and e. remiddi , phys .
b * 379 * , 283 ( 1996 ) ; s. laporta , int .
j. mod .
a * 15 * , 5087 ( 2000 ) .
m. czakon , diagen / idsolver library ( unpublished ) .
a. v. kotikov , phys .
b * 259 * , 314 ( 1991 ) ; e. remiddi , nuovo cim .
a * 110 * , 1435 ( 1997 ) .
s. bauberger , f. a. berends , m. bhm and m. buza , nucl .
b * 434 * ( 1995 ) 383 .
s. bauberger and m. bhm , nucl .
b * 445 * ( 1995 ) 25 .
a. czarnecki , u. kilian and d. kreimer , nucl .
b * 433 * , 259 ( 1995 ) ; a. frink , b. a. kniehl , d. kreimer and k. riesselmann , phys .
d * 54 * , 4548 ( 1996 ) ; a. frink , u. kilian and d. kreimer , nucl .
b * 488 * , 426 ( 1997 ) .
a. ghinculov and j. j. van der bij , nucl .
b * 436 * ( 1995 ) 30 .
a. b. arbuzov , m. awramik , m. czakon , a. freitas , m. w. grnewald , k. mnig , s. riemann , t. riemann , _ zfitter : a semi - analytical program for fermion pair production in @xmath78 annihilation , from version 6.21 to version 6.42 _ , hep - ph/0507146 , submitted to comp .
+ [ ` http://www-zeuthen.desy.de/theory/research/zfitter/ ` ] . | electroweak precision observables allow stringent tests of the standard model at the quantum level and imply interesting bounds on the mass of the higgs boson through higher - order loop effects .
very significant constraints come especially from the determination of the mass of the @xmath0 boson and from the effective leptonic weak mixing angle .
after shortly reviewing the status of theoretical computations of the @xmath0 mass , the new calculation of two - loop corrections with closed fermion loops to the effective leptonic weak mixing angle is discussed in detail .
the phenomenological implications of the new result are analyzed including an estimate of remaining uncertainties . |
for a range of scientific studies including quantum computation , quantum information and quantum simulation , an important task is to learn and engineer quantum systems [ @xcite , @xcite ( @xcite , @xcite ) , @xcite , @xcite,@xcite , @xcite , and wang ( @xcite , @xcite ) ] .
a quantum system is described by its state characterized by a density matrix , which is a positive semidefinite hermitian matrix with unit trace . determining a quantum state , often referred to as quantum state tomography , is an important but difficult task [ @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , and @xcite ] .
it is often inferred by performing measurements on a large number of identically prepared quantum systems .
more specifically , we describe a quantum spin system by the @xmath0-dimensional complex space @xmath1 and its quantum state by a complex matrix on @xmath1 .
when measuring the quantum system by performing measurements on some observables which can be represented by hermitian matrices , we obtain the measurement outcomes for each observable , where the measurements take values at random from all eigenvalues of the observable , with the probability of observing a particular eigenvalue equal to the trace of the product of the density matrix and the projection matrix onto the eigenspace corresponding to the eigenvalue . to handle the up and down states of particles in a quantum spin system , a common approach is to employ the well - known pauli matrices as observables to perform measurements and obtain the so - called pauli measurements [ @xcite , @xcite , @xcite , @xcite , @xcite , and wang ( @xcite , @xcite ) ] .
since all pauli matrices have @xmath2 eigenvalues , pauli measurements takes discrete values @xmath3 and @xmath4 , and the resulted measurement distributions can be characterized by binomial distributions . the goal is to estimate the density matrix based on the pauli measurements .
traditional quantum tomography employs classical statistical models and methods to deduce quantum states from quantum measurements .
these approaches are designed for the setting where the size of a density matrix is greatly exceeded by the number of quantum measurements , which is almost never the case even for moderate quantum systems in practice because the dimension of the density matrix grows exponentially in the size of the quantum system .
for example , the density matrix for @xmath5 spin-@xmath6 quantum systems is of size @xmath7 . in this paper
, we aim to effectively and efficiently reconstruct the density matrix for a large - scale quantum system with a relatively limited number of quantum measurements .
quantum state tomography is fundamentally connected to the problem of recovering a high - dimensional matrix based on noisy observations [ @xcite ] .
the latter problem arises naturally in many applications in statistics and machine learning and has attracted considerable recent attention . when assuming that the unknown matrix of interest is of ( approximately ) low - rank , many regularization techniques have been developed .
examples include @xcite , @xcite , @xcite ( @xcite , @xcite ) , @xcite , @xcite , bunea , she and wegkamp ( @xcite , @xcite ) , klopp ( @xcite , @xcite ) , @xcite , @xcite , @xcite , @xcite , @xcite , and @xcite , among many others .
taking advantage of the low - rank structure of the unknown matrix , these approaches can often be applied to estimate unknown matrices of high dimensions .
yet these methods do not fully account for the specific structure of quantum state tomography . as
demonstrated in a pioneering article , @xcite argued that , when considering quantum measurements characterized by the pauli matrices , the density matrix can often be characterized by the sparsity with respect to the pauli basis . built upon this connection , they suggested a compressed sensing [ @xcite ] strategy for quantum state tomography [ @xcite and @xcite ] .
although promising , their proposal assumes exact measurements , which is rarely the case in practice , and adopts the constrained nuclear norm minimization method , which may not be an appropriate matrix completion approach for estimating a density matrix with unit trace ( or unit nuclear norm ) .
we specifically address such challenges in the present paper .
in particular , we establish the minimax optimal rates of convergence for the density matrix estimation under both the spectral and frobenius norm losses when assuming that the true density matrix is approximately sparse under the pauli basis .
furthermore , we show that these rates could be achieved by carefully thresholding the coefficients with respect to the pauli basis .
because the quantum pauli measurements are characterized by the binomial distributions , the convergence rates and minimax lower bounds are derived by asymptotic analysis with manipulations of binomial distributions instead of the usual normal distribution based calculations . the rest of paper proceeds as follows .
section [ sec2 ] gives some background on quantum state tomography and introduces a thresholding based density matrix estimator .
section [ sec3 ] develops theoretical properties for the density matrix estimation problem .
in particular , the convergence rates of the proposed density matrix estimator and its minimax optimality with respect to both the spectral and frobenius norm losses are established .
section [ sec4 ] features a simulation study to illustrate finite sample performance of the proposed estimators .
all technical proofs are collected in section [ proofs ] .
in this section , we first review the quantum state and density matrix and introduce pauli matrices and pauli measurements . we also develop results to describe density matrix representations through pauli matrices and characterize the distributions of pauli measurements via binomial distribution before introducing a thresholding based density matrix estimator . for a @xmath0-dimensional quantum system
, we describe its quantum state by a density matrix @xmath8 on @xmath0 dimensional complex space @xmath1 , where density matrix @xmath8 is a @xmath0 by @xmath0 complex matrix satisfying ( 1 ) hermitian , that is , @xmath8 is equal to its conjugate transpose ; ( 2 ) positive semidefinite ; ( 3 ) unit trace , that is , @xmath9 . for a quantum system ,
it is important but difficult to know its quantum state .
experiments are conducted to perform measurements on the quantum system and obtain data for studying the quantum system and estimating its density matrix . in physics literature
, quantum state tomography refers to reconstruction of a quantum state based on measurements for the quantum systems .
statistically , it is the problem of estimating the density matrix from the measurements .
common quantum measurements are on observable @xmath10 , which is defined as a hermitian matrix on @xmath1 .
assume that the observable @xmath10 has the following spectral decomposition : @xmath11 where @xmath12 are @xmath13 different real eigenvalues of @xmath10 , and @xmath14 are projections onto the eigenspaces corresponding to @xmath12 .
for the quantum system prepared in state @xmath8 , we need a probability space @xmath15 to describe measurement outcomes when performing measurements on the observable @xmath10 .
denote by @xmath16 the measurement outcome of @xmath10 . according to the theory of quantum mechanics , @xmath16 is a random variable on @xmath15 taking values in @xmath17 , with probability distribution given by @xmath18 we may perform measurements on an observable for a quantum system that is identically prepared under the state and obtain independent and identically distributed observations .
see @xcite , @xcite , and @xcite .
the pauli matrices as observables are widely used in quantum physics and quantum information science to perform quantum measurements .
let @xmath19 where @xmath20 , @xmath21 and @xmath22 are called the two - dimensional pauli matrices .
tensor products are used to define high - dimensional pauli matrices .
let @xmath23 for some integer @xmath5 .
we form @xmath5-fold tensor products of @xmath24 , @xmath20 , @xmath21 and @xmath22 to obtain @xmath0 dimensional pauli matrices @xmath25 we identify index @xmath26 with @xmath27 .
for example , @xmath28 corresponds to @xmath29 . with the index identification
we denote by @xmath30 the pauli matrix @xmath31 , with @xmath32 .
we have the following theorem to describe pauli matrices and represent a density matrix by pauli matrices . [ thm-1 ]
pauli matrices @xmath33 are of full rank and have eigenvalues @xmath2 .
denote by @xmath34 the projections onto the eigen - spaces of @xmath30 corresponding to eigenvalues @xmath2 , respectively .
then for @xmath35 , @xmath36 denote by @xmath37 the space of all @xmath0 by @xmath0 complex matrices equipped with the frobenius norm .
all pauli matrices defined by ( [ pauli - matrix ] ) form an orthogonal basis for all complex hermitian matrices .
given a density matrix @xmath8 , we can expand it under the pauli basis as follows : @xmath38 where @xmath39 are coefficients .
for @xmath40 , @xmath41 suppose that an experiment is conducted to perform measurements on pauli observable @xmath30 independently for @xmath42 quantum systems which are identically prepared in the same quantum state @xmath8 . as @xmath30 has eigenvalues @xmath2 , the pauli measurements take values @xmath3 and @xmath4 , and thus the average of the @xmath42 measurements for each @xmath30 is a sufficient statistic .
denote by @xmath43 the average of the @xmath42 measurement outcomes obtained from measuring @xmath30 , @xmath44 .
our goal is to estimate @xmath8 based on @xmath45 .
the following proposition provides a simple binomial characterization for the distributions of @xmath43 .
[ thm-2 ] suppose that @xmath8 is given by ( [ beta - representation ] ) .
then @xmath45 are independent with @xmath46 and @xmath47 follows a binomial distribution with @xmath42 trials and cell probabilities @xmath48 , where @xmath49 denotes the projection onto the eigenspace of @xmath30 corresponding to eigenvalue @xmath3 , and @xmath39 is the coefficient of @xmath30 in the expansion of @xmath8 in ( [ beta - representation ] ) .
since the dimension of a quantum system grows exponentially with its components such as the number of particles in the system , the matrix size of @xmath8 tends to be very large even for a moderate quantum system .
we need to impose some structure such as sparsity on @xmath8 in order to make it consistently estimable .
suppose that @xmath8 has a sparse representation under the pauli basis , following wavelet shrinkage estimation we construct a density matrix estimator of @xmath8 .
assume that representation ( [ beta - representation ] ) is sparse in a sense that there is only a relatively small number of coefficients @xmath50 with large magnitudes .
formally , we specify sparsity by assuming that coefficients @xmath51 satisfy @xmath52 where @xmath53 , and @xmath54 is a deterministic function with slow growth in @xmath0 such as @xmath55 .
pauli matrices are used to describe the spins of spin-@xmath6 particles along different directions , and density matrix @xmath8 in ( [ beta - representation ] ) represents a mixture of quantum states with spins along many directions .
sparsity assumption ( [ csparsity ] ) with @xmath56 indicates the mixed state involving spins along a relatively small number of directions corresponding to those pauli matrices with nonzero @xmath50 .
the sparsity reduces the complexity of mixed states .
sparse density matrices often occur in quantum systems where particles have sparse interactions such as location interactions .
examples include many quantum systems in quantum information and quantum computation [ @xcite , @xcite , @xcite , @xcite , @xcite , and wang ( @xcite , @xcite ) ] . since @xmath57 are independent , and @xmath58 .
we naturally estimate @xmath50 by @xmath57 and threshold @xmath57 to estimate large @xmath50 , ignoring small @xmath59 , and obtain @xmath60 \label{threshold1 } \\[-8pt ] \nonumber \hat{\beta } _ k & = & \operatorname{sign}(n_k ) \bigl(|n_k|- \varpi\bigr)_+ , \qquad k = 2 , \ldots , d^2,\end{aligned}\ ] ] and then we use @xmath61 to construct the following estimator of @xmath8 , @xmath62 where the two estimation methods in ( [ threshold1 ] ) are called hard and soft thresholding rules , and @xmath63 is a threshold value which , we reason below , can be chosen to be @xmath64 for some constant @xmath65 .
the threshold value is designed such that for small @xmath50 , @xmath57 must be bounded by threshold @xmath63 with overwhelming probability , and the hard and soft thresholding rules select only those @xmath57 with large signal components @xmath50 .
as @xmath66 , an application of bernstein s inequality leads to that for any @xmath67 , @xmath68 and @xmath69^{d^2 - 1 } = \bigl [ 1 - 2 d^{-2 \hbar/(1+o(1 ) ) } \bigr]^{d^2 - 1 } \rightarrow1 , \ ] ] as @xmath70 and @xmath71 , that is , with probability tending to one , @xmath72 uniformly for @xmath73 .
thus , we can select @xmath64 to threshold @xmath57 and obtain @xmath61 in ( [ threshold1 ] ) .
we fix matrix norm notation for our asymptotic analysis .
let @xmath74 be a @xmath0-dimensional vector and @xmath75 be a @xmath0 by @xmath0 matrix , and define their @xmath76 norms @xmath77 denote by @xmath78 the frobenius norm of @xmath79 . for the case of matrix , the @xmath80 norm
is called the matrix spectral norm or operator norm .
@xmath81 is equal to the square root of the largest eigenvalue of @xmath82 , @xmath83 and @xmath84 for a real symmetric or complex hermitian matrix @xmath79 , @xmath85 is equal to the largest absolute eigenvalue of @xmath79 , @xmath86 is the square root of the sum of squared eigenvalues , @xmath87 , and ( [ ell-1infty - norm])([ell-12infty - norm ] ) imply that @xmath88 .
the following theorem gives the convergence rates for @xmath89 under the spectral and frobenius norms .
[ thm-3 ] denote by @xmath90 the class of density matrices satisfying the sparsity condition ( [ csparsity ] ) .
assume @xmath91 for some constants @xmath92 and @xmath93 . for density matrix estimator @xmath89 defined by ( [ threshold1])([threshold - estimator1 ] ) with threshold @xmath94 for some constant @xmath65
, we have @xmath95 & \leq & c_2 \pi^2_n(d ) \frac{1}{d^2 } \biggl ( \frac { \log d } { n } \biggr)^{1-q } , \\ \sup_{{\bolds{\rho}}\in\theta } e\bigl[\| \hat{{\bolds{\rho } } } - { \bolds{\rho}}\|_f^2 \bigr ] & \leq & c_3 \pi_n(d ) \frac{1}{d } \biggl ( \frac { \log d } { n } \biggr)^{1-q/2},\end{aligned}\ ] ] where @xmath96 and @xmath97 are constants free of @xmath42 and @xmath0 . [ rem1 ] theorem [ thm-3 ] shows that @xmath89 achieves the convergence rate @xmath98 under the squared frobenius norm loss and the convergence rate @xmath99 under the squared spectral norm loss .
both rates will be shown to be optimal in the next section .
similar to the optimal convergence rates for large covariance and volatility matrix estimation [ @xcite and @xcite ] , the optimal convergence rates here have factors involving @xmath54 and @xmath100 .
however , unlike the covariance and volatility matrix estimation case , the convergence rates in theorem [ thm-3 ] have factors @xmath101 and @xmath102 for the squared spectral and frobenius norms , respectively , and go to zero as @xmath0 approaches to infinity .
in particular , the result implies that mses of the proposed estimator get smaller for large @xmath0 .
this is quite contrary to large covariance and volatility matrix estimation where the traces are typically diverge , the optimal convergence rates grow with the logarithm of matrix size , and the corresponding mses increase in matrix size
. the new phenomenon may be due to the unit trace constraint on density matrix and that the density matrix representation ( [ beta - representation ] ) needs a scaling factor @xmath101 to satisfy the constraint . also for finite sample @xmath89
may not be positive semidefinite , we may project @xmath89 onto the cone formed by all density matrices under a given matrix norm @xmath103 , and obtain a positive semidefinite density matrix estimator @xmath104 .
since the underlying true density matrix @xmath8 is positive semidefinite with unit trace , and the representation ( [ threshold - estimator1 ] ) ensures that @xmath89 has unit trace , the projection implies @xmath105 .
thus , @xmath106 .
taking @xmath103 as the spectral norm or the frobenius norm and using theorem [ thm-3 ] , we conclude that @xmath104 has the same convergence rates as @xmath89 .
the following theorem establishes a minimax lower bound for estimating @xmath8 under the spectral norm .
[ thm-4 ] we assume that @xmath54 in the sparsity condition ( [ csparsity ] ) satisfies @xmath107 for some constant @xmath108 and @xmath109 . then @xmath110 \geq c_4 \pi^2_n(d ) \frac { 1}{d^2 } \biggl ( \frac { \log d } { n } \biggr)^{1-q},\ ] ] where @xmath111 denotes any estimator of @xmath8 based on measurement data @xmath45 , and @xmath112 is a constant free of @xmath42 and @xmath0
. [ rem2 ] the lower bound in theorem [ thm-4 ] matches the convergence rate of @xmath89 under the spectral norm in theorem [ thm-3 ] , so we conclude that @xmath89 achieves the optimal convergence rate under the spectral norm . to establish the minimax lower bound in theorem [ thm-4 ] , we construct a special subclass of density matrices and then apply le cam s lemma . assumption ( [ cond - pi ] )
is needed to guarantee the positive definiteness of the constructed matrices as density matrix candidates and to ensure the boundedness below from zero for the total variation of related probability distributions in le cam s lemma .
assumption ( [ cond - pi ] ) is reasonable in a sense that if the right - hand side of ( [ cond - pi ] ) is large enough , ( [ cond - pi ] ) will not impose very restrictive condition on @xmath54 .
we evaluate the dominating factor @xmath113 on the right - hand side of ( [ cond - pi ] ) for various scenarios .
first , consider @xmath56 , the assumption becomes @xmath114 , @xmath115 , and so assumption ( [ cond - pi ] ) essentially requires @xmath54 grows in @xmath0 not faster than @xmath116 , which is not restrictive at all as @xmath54 usually grows slowly in @xmath0 .
the asymptotic analysis of high - dimensional statistics usually allows both @xmath0 and @xmath42 go to infinity .
typically , we may assume @xmath0 grows polynomially or exponentially in @xmath42 .
if @xmath0 grows exponentially in @xmath42 , that is , @xmath117 for some @xmath118 , then @xmath119 is negligible in comparison with @xmath120 , and @xmath113 behavior like @xmath120 .
the assumption in this case is not very restrictive . for the case of polynomial growth , that is , @xmath121 for some @xmath122 , then @xmath123 . if @xmath124 , @xmath125 grows in @xmath0 like some positive power of @xmath0 .
since we may take @xmath126 arbitrarily close to @xmath127 , the positiveness of @xmath128 essentially requires @xmath129 , which can often be quite realistic given that @xmath130 is usually very small .
the theorem below provides a minimax lower bound for estimating @xmath8 under the frobenius norm .
[ thm-5 ] we assume that @xmath54 in the sparsity condition ( [ csparsity ] ) satisfies @xmath131 for some constants @xmath132 and @xmath133
. then @xmath134 \geq c_5 \pi_n(d ) \frac{1}{d } \biggl ( \frac { \log d } { n } \biggr)^{1-q/2},\ ] ] where @xmath111 denotes any estimator of @xmath8 based on measurement data @xmath45 , and @xmath135 is a constant free of @xmath42 and @xmath0 .
[ rem3 ] the lower bound in theorem [ thm-5 ] matches the convergence rate of @xmath89 under the frobenius norm in theorem [ thm-3 ] , so we conclude that @xmath89 achieves the optimal convergence rate under the frobenius norm .
similar to the remark [ rem2 ] after theorem [ thm-4 ] , we need to apply assouad s lemma to establish the minimax lower bound in theorem [ thm-5 ] , and assumption ( [ cond - pi-0 ] ) is used to guarantee the positive definiteness of the constructed matrices as density matrix candidates and to ensure the boundedness below from zero for the total variation of related probability distributions in assouad s lemma .
also the appropriateness of ( [ cond - pi-0 ] ) is more relaxed than ( [ cond - pi ] ) , as @xmath136 and the right - hand side of ( [ cond - pi-0 ] ) has main powers more than the square of that of ( [ cond - pi ] ) .
it is interesting to consider density matrix estimation under a schatten norm , where given a matrix @xmath79 of size @xmath0 , we define its schatten @xmath137-norm by @xmath138 are the eigenvalues of the square root of @xmath139 .
spectral norm and frobenius norm are two special cases of the schatten @xmath137-norm with @xmath140 and , respectively , and the nuclear norm corresponds to the schatten @xmath137-norm with @xmath141 .
the following result provides the convergence rate for the proposed thresholding estimator under the schatten @xmath137-norm loss for @xmath142 .
[ schatten.prop ] under the assumptions of theorem [ thm-3 ] , the density matrix estimator @xmath89 defined by ( [ threshold1])([threshold - estimator1 ] ) with threshold @xmath94 for some constant @xmath65 satisfies @xmath143 \leq c \bigl[\pi_n(d)\bigr ] ^{2 - 2/\max(s,2 ) } \frac{1}{d^{2 - 2/s } } \biggl(\frac { \log d}{n } \biggr ) ^{1-q+q/\max(s,2)}\ ] ] for @xmath142 , where @xmath144 is a constant not depending on @xmath42 and @xmath0 .
the upper bound in ( [ schatten - s1 ] ) matches the minimax convergence rates for both the spectral norm and frobenius norm .
moreover , for the case of the nuclear norm corresponding to the schatten @xmath137-norm with @xmath141 , ( [ schatten - s1 ] ) leads to an upper bound with the convergence rate @xmath145 .
we conjecture that the upper bounds in ( [ schatten - s1 ] ) are rate - optimal under the schatten @xmath137-norm loss for all @xmath146 .
however , establishing a matching lower bound for the general schatten norm loss is a difficult task , and we believe that a new approach is needed for studying minimax density matrix estimation under the schatten @xmath137-norm , particularly the nuclear norm .
[ rem4 ] the pauli basis expansion ( [ beta - representation ] ) is orthogonal with respect to the usual euclidean inner product , and as in the proof of lemma [ lem - norm ] we have @xmath147 where @xmath148 and @xmath89 are threshold estimators of @xmath149 and @xmath8 , respectively .
the sparse vector estimation problem is well studied under the gaussian or sub - gaussian noise case [ @xcite and @xcite ] and can be used to recover the minimax result for density matrix estimation under the frobenius norm loss , because of orthogonality .
in fact , our relatively simple proof of the minimax results for the frobenius norm loss is essentially the same as the sparse vector estimation approach .
however , such an equivalence between sparse density matrix estimation and sparse vector estimation breaks down for the general schatten norm loss such as the commonly used spectral norm and nuclear norm losses .
for the spectral norm loss , lemma [ lem - norm ] in section [ proofs ] provides a sharp upper bound for @xmath150 $ ] through the @xmath151-norm of @xmath152 , and the proof of the minimax lower bound in theorem [ thm-4 ] relies on the property that the spectral norm is determined by the largest eigenvalue only .
such a special property allows us to reduce the problem to a simple subproblem and establish the lower bound under the spectral norm loss .
the arguments can not be applied to the case of the general schatten norm loss in particular the nuclear norm loss .
moreover , instead of directly applying lemma [ lem - norm ] and remark [ rem5 ] in section [ proofs ] to derive upper bounds for the general schatten norm loss , we use the obtained sharp upper bounds for the spectral norm and frobenius norm losses together with moment inequalities to derive sharper upper bounds in proposition [ schatten.prop ] . however , similar lower bounds are not available .
our analysis leads us to believe that it is not possible to use sparse vector estimation to recover minimax lower bound results for the general schatten norm loss in particular for the spectral norm loss .
a simulation study was conducted to investigate the performance of the proposed density matrix estimator for the finite sample .
we took @xmath153 and generated a true density matrix @xmath8 for each case as follows .
@xmath8 has an expansion over the pauli basis @xmath154 where @xmath155 , @xmath40 . from @xmath51
, we randomly selected @xmath156 $ ] coefficients @xmath39 and set the rest of @xmath39 to be zero .
we simulated @xmath157 $ ] values independently from a uniform distribution on @xmath158 $ ] , assigned the simulated values at random to the selected @xmath39 , and then constructed @xmath8 from ( [ rho - simulation ] ) .
the constructed @xmath8 always has unit trace but may not be positive semi - definite .
the procedure was repeated until we generated a positive semi - definite @xmath8 .
we took it as the true density matrix .
the simulation procedure guarantees the obtained @xmath8 is a density matrix and has a sparse representation under the pauli basis . for each true density matrix @xmath8 ,
as described in section [ sec-2 - 2 ] we simulated data @xmath43 from a binomial distribution with cell probability @xmath39 and the number of cells @xmath159 .
we constructed coefficient estimators @xmath160 by ( [ threshold1 ] ) and obtained density matrix estimator @xmath89 using ( [ threshold - estimator1 ] ) .
the whole estimation procedure is repeated @xmath161 times .
the density matrix estimator is measured by the mean squared errors ( mse ) , @xmath162 and @xmath163 , that are evaluated by the average of @xmath164 and @xmath165 over @xmath161 repetitions , respectively .
three thresholds were used in the simulation study : the universal threshold @xmath166 for all @xmath39 , the individual threshold @xmath167 for each @xmath39 , and the optimal threshold for all @xmath39 , which minimizes the computed mse for each corresponding hard or soft threshold method .
the individual threshold takes into account the fact in theorem [ thm-2 ] that the mean and variance of @xmath43 are @xmath39 and @xmath168 , respectively , and the variance of @xmath43 is estimated by @xmath169 . .
are plots of mses based on the spectral norm for @xmath153 , respectively , and are plots of mses based on the frobenius norm for @xmath170 , respectively . ]
[ figure1 ] . are plots of mses based on the spectral norm for @xmath153 , respectively , and are plots of mses based on the frobenius norm for @xmath171 , respectively . ] . are plots of mses based on the spectral norm for @xmath172 , respectively , and are plots of mses based on the frobenius norm for @xmath173 , respectively . ] or @xmath174 against matrix size @xmath0 for the proposed density estimator with hard and soft threshold rules for @xmath175 . are plots of @xmath174 times of mses based on the spectral norm for @xmath175 , respectively , and are plots of @xmath0 times of mses based on the frobenius norm for @xmath175 , respectively . ] figures [ figure1 ] and [ figure2 ] plot the mses of the density matrix estimators with hard and soft threshold rules and its corresponding density matrix estimator without thresholding [ i.e. , @xmath39 are estimated by @xmath43 in ( [ threshold - estimator1 ] ) ] against the sample size @xmath42 for different matrix size @xmath0 , and figures [ figure3 ] and [ figure4 ] plot their mses against matrix size @xmath0 for different sample size .
the numerical values of the mses are reported in table [ table1 ] .
figures [ figure1 ] and [ figure2 ] show that the mses usually decrease in sample size @xmath42 , and the thresholding density matrix estimators enjoy superior performances than that the density matrix estimator without thresholding even for @xmath176 ; while all threshold rules and threshold values yield thresholding density matrix estimators with very close mses , the soft threshold rule with individual and universal threshold values produce larger mses than others for larger sample size such as @xmath177 and the soft threshold rule tends to give somewhat better performance than the hard threshold rule for smaller sample size like @xmath178 .
figures [ figure3 ] and [ figure4 ] demonstrate that while the mses of all thresholding density matrix estimators decrease in the matrix size @xmath0 , but if we rescale the mses by multiplying it with @xmath174 for the spectral norm case and @xmath0 for the frobenius norm case , the rescaled mses slowly increase in matrix size @xmath0 .
the simulation results largely confirm the theoretical findings discussed in remark [ rem1 ] .
let @xmath179 . denote by @xmath180 s generic constants whose values are free of @xmath42 and @xmath181 and may change from appearance to appearance .
let @xmath182 and @xmath183 be the maximum and minimum of @xmath184 and @xmath126 , respectively . for two sequences @xmath185 and @xmath186 , we write @xmath187 if @xmath188 as @xmath189 , and write @xmath190 if there exist positive constants @xmath191 and @xmath192 free of @xmath42 and @xmath181 such that @xmath193 . without confusion
we may write @xmath54 as @xmath194 .
proof of proposition [ thm-1 ] in two dimensions , pauli matrices satisfy @xmath195 , and @xmath196 , @xmath197 have eigenvalues @xmath2 , the square of a pauli matrix is equal to the identity matrix , and the multiplications of any two pauli matrices are equal to the third pauli matrix multiplying by @xmath198 , for example , @xmath199 , @xmath200 , and @xmath201 . for @xmath206 , @xmath207 , @xmath208 and @xmath209 , @xmath210\otimes[{\bolds{\sigma}}_{\ell_2}{\bolds{\sigma}}_{\ell^\prime_2 } ] \otimes \cdots\otimes[{\bolds{\sigma}}_{\ell_b } { \bolds{\sigma}}_{\ell^\prime_b}],\ ] ] is equal to a @xmath0 dimensional pauli matrix multiplying by @xmath211 , which has zero trace .
thus , @xmath212 , that is , @xmath30 and @xmath213 are orthogonal , and @xmath214 form an orthogonal basis .
@xmath215 . in particular @xmath32 , and @xmath216 .
@lcd4.3d1.3d1.3d1.3d1.3d1.3d1.3d2.3d2.3d2.3@ & & & + & & & + & & & & & + & & & & & & & & & + & & & & & & & & & + & & & & & & & + & & & & & & & + & & & & & & & & & & & + 32 & 100 & 348.544 & 4.816 & 4.648 & 5.468 & 4.790 & 6.104 & 4.762 & 24.782 & 15.180 & 0.619 + & 200 & 175.034 & 4.449 & 4.257 & 5.043 & 4.708 & 5.293 & 4.667 & 17.524 & 7.739 & 0.562 + & 500 & 70.069 & 2.831 & 3.054 & 3.344 & 4.130 & 3.260 & 4.071 & 11.083 & 2.397 & 0.373 + & 1000 & 35.028 & 1.537 & 1.974 & 1.875 & 3.201 & 1.875 & 3.155 & 7.837 & 1.099 & 0.212 + & 2000 & 17.307 & 0.785 & 1.195 & 1.001 & 2.230 & 0.989 & 2.200 & 5.541 & 0.551 & 0.116 + 64 & 100 & 368.842 & 1.583 & 1.572 & 1.744 & 1.583 & 1.954 & 1.586 & 27.148 & 16.660 & 0.395 + & 200 & 183.050 & 1.565 & 1.534 & 1.669 & 1.575 & 1.833 & 1.571 & 19.196 & 9.252 & 0.376 + & 500 & 73.399 & 1.175 & 1.228 & 1.367 & 1.490 & 1.347 & 1.476 & 12.141 & 2.900 & 0.307 + & 1000 & 36.692 & 0.566 & 0.807 & 0.747 & 1.249 & 0.722 & 1.233 & 8.585 & 1.308 & 0.177 + & 2000 & 18.402 & 0.186 & 0.443 & 0.255 & 0.832 & 0.251 & 0.820 & 6.070 & 0.657 & 0.061 + 128 & 100 & 381.032 & 0.543 & 0.542 & 0.574 & 0.543 & 0.705 & 0.545 & 29.323 & 17.500 & 0.237 + & 200 & 190.113 & 0.541 & 0.539 & 0.570 & 0.542 & 0.594 & 0.542 & 20.734 & 10.246 & 0.235 + & 500 & 75.824 & 0.471 & 0.480 & 0.514 & 0.525 & 0.509 & 0.522 & 13.114 & 3.547 &
0.213 + & 1000 & 38.010 & 0.309 & 0.350 & 0.355 & 0.470 & 0.354 & 0.466 & 9.273 & 1.613 & 0.146 + & 2000 & 18.907 & 0.142 & 0.216 & 0.194 & 0.359 & 0.194 & 0.356 & 6.557 & 0.725 & 0.080 + @lcd4.3d1.3d1.3d1.3d1.3d1.3d1.3d2.3d2.3d2.3@ & & & + & & & + & & & & & + & & & & & & & & & + & & & & & & & & & + & & & & & & & + & & & & & & & + & & & & & & & & & & & + 32 & 100 & 317.873 & 6.052 & 5.119 & 6.195 & 5.274 & 7.050 & 5.246 & 24.782 & 11.004 & 9.936 + & 200 & 159.679 & 5.217 & 4.629 & 5.616 & 5.187 & 5.874 & 5.143 & 17.524 & 5.681 & 3.771 + & 500 & 63.823 & 3.165 & 3.229 & 3.732 & 4.575 & 3.642 & 4.512 & 11.083 & 2.286 & 0.954 + & 1000 & 31.856 & 1.722 & 2.053 & 2.119 & 3.540 & 2.119 & 3.492 & 7.837 & 1.100 & 0.401 + & 2000 & 15.967 & 0.894 & 1.219 & 1.155 & 2.424 & 1.141 & 2.394 & 5.541 & 0.546 & 0.182 + 64 & 100 & 641.437 & 3.909 & 3.528 & 3.951 & 3.563 & 4.463 & 3.562 & 27.148 & 13.719 & 13.234 + & 200 & 319.720 & 3.706 & 3.401 & 3.755 & 3.548 & 4.082 & 3.536 & 19.196 & 7.042 & 5.515 + & 500 & 127.958 & 2.691 & 2.551 & 3.069 & 3.342 & 3.023 & 3.309 & 12.141 & 2.800 & 1.275 + & 1000 & 63.845 & 1.335 & 1.628 & 1.765 & 2.791 & 1.717 & 2.756 & 8.585 & 1.277 & 0.548 + & 2000 & 31.952 & 0.433 & 0.882 & 0.610 & 1.842 & 0.596 & 1.817 & 6.070 & 0.647 & 0.258 + 128 & 100 & 1283.182 & 2.370 & 2.240 & 2.370 & 2.242 & 2.924 & 2.245 & 29.323 & 15.989 & 16.128 + & 200 & 639.556 & 2.349 & 2.219 & 2.354 & 2.238 & 2.444 & 2.238 & 20.734 & 8.218 & 7.799 + & 500 & 255.954 & 1.990 & 1.906 & 2.125 & 2.172 & 2.102 & 2.160 & 13.114 & 3.355 & 1.773 + & 1000 & 127.714 & 1.221 & 1.341 & 1.463 & 1.943 & 1.448 & 1.924 & 9.273 & 1.546 & 0.729 + & 2000 & 63.921 & 0.581 & 0.815 & 0.798 & 1.471 & 0.798 & 1.456 & 6.557 & 0.719 & 0.327 + denote by @xmath34 the projections onto the eigenspaces corresponding to eigenvalues @xmath2 , respectively .
then for @xmath217 , @xmath218 and solving the equations we get @xmath219 for @xmath220 , @xmath221 , @xmath222 and @xmath213 are orthogonal , @xmath223 and @xmath224 which imply @xmath225 for a density matrix @xmath8 with representation ( [ beta - representation ] ) under the pauli basis ( [ pauli - matrix ] ) , from ( [ eq1 ] ) we have @xmath226 and @xmath227 , and thus @xmath228 \label{tr - qrho2 } \\[-8pt ] \nonumber & = & \frac{1}{2 } + \frac{\beta_k}{d } \operatorname{tr}({\mathbf{b}}_{k } { \mathbf{q}}_{k \pm } ) = \frac{1 \pm\beta_k}{2}.\end{aligned}\ ] ] proof of proposition [ thm-2 ] we perform measurements on each pauli observable @xmath229 independently for @xmath42 quantum systems that are identically prepared under state @xmath8 .
denote by @xmath230 the @xmath42 measurement outcomes for measuring @xmath229 , @xmath231 .
@xmath232 @xmath233 , @xmath234 , @xmath235 , are independent , and take values @xmath2 , with distributions given by @xmath236 as random variables @xmath230 are i.i.d . and take eigenvalues @xmath2
, @xmath237 is equal to the total number of random variables @xmath230 taking eigenvalue @xmath3 , and thus @xmath238 follows a binomial distribution with @xmath42 trials and cell probability @xmath239 . from ( [ tomography3])([tomography4 ] ) and proposition [ thm-1 ] ,
we have for @xmath234 , @xmath240 from proposition [ thm-2 ] and ( [ tomography3])([tomography4 ] ) , we have that @xmath43 is the average of @xmath249 , which are i.i.d .
random variables taking values @xmath2 , @xmath250 , @xmath251 and @xmath252 . applying bernstein s inequality ,
we obtain for any @xmath67 , @xmath253 both @xmath254 and @xmath255 are less than @xmath256 , which is bounded by @xmath257 [ lem - norm ] @xmath258 + \biggl\ { \sum _ { j=2}^p e\bigl[|\hat{\beta}_j - \beta_j|\bigr ] \biggr\}^2 \nonumber \\[-8pt ] \label{e - spectral - norm - square } \\[-8pt ] \nonumber & & { } - \sum_{j=2}^p \bigl\{e\bigl(|\hat { \beta}_j - \beta _ j|\bigr)\bigr\}^2.\end{aligned}\ ] ] since pauli matrices @xmath30 are orthogonal with respect to the usual euclidean inner product , with @xmath259 , and @xmath260 , we have @xmath261 \label{trace - norm } \\[-8pt ] \nonumber & = & \sum_{j=2}^p |\hat { \beta}_j - \beta_j|^2/d , \\ p^{1/2 } \| \hat{{\bolds{\rho } } } - { \bolds{\rho}}\|_2 & = & \biggl\| \sum _ { j=2}^p ( \hat{\beta } _ j - \beta_j ) { \mathbf{b}}_j \biggr\|_2 \leq\sum _ { j=2}^p |\hat{\beta}_j - \beta _
j| \| { \mathbf{b}}_j \|_2 \nonumber \\[-8pt ] \label{spectral - norm } \\[-8pt ] \nonumber & = & \sum_{j=2}^p |\hat { \beta}_j - \beta_j| , \\
p \| \hat{{\bolds{\rho } } } - { \bolds{\rho}}\|_2 ^ 2 & = & \biggl\| \sum _ { j=2}^p ( \hat{\beta}_j - \beta_j ) { \mathbf{b}}_j \biggr\|_2 ^ 2 \nonumber\\ & \leq & \sum_{j=2}^p |\hat { \beta}_j - \beta_j|^2 \| { \mathbf{b}}_j \|_2 ^ 2 + 2 \sum_{i < j } ^p \bigl|(\hat{\beta}_i - \beta_i ) ( \hat{\beta}_j - \beta_j)\bigr| \| { \mathbf{b}}_i { \mathbf{b}}_j \|_2 \nonumber \\[-8pt ] \label{spectral - norm - square } \\[-8pt ] \nonumber & \leq & \sum_{j=2}^p |\hat { \beta}_j - \beta_j|^2 \| { \mathbf{b}}_j \|_2 ^ 2 + 2 \sum_{i < j } ^p \bigl|(\hat{\beta}_i - \beta_i ) ( \hat{\beta}_j - \beta_j)\bigr| \| { \mathbf{b}}_i\|_2 \| { \mathbf{b}}_j \|_2 \nonumber \\ & = & \sum_{j=2}^p |\hat { \beta}_j - \beta_j|^2 + 2 \sum _
{ i < j } ^p\bigl|(\hat{\beta}_i - \beta_i ) ( \hat{\beta}_j - \beta_j)\bigr| . \nonumber\end{aligned}\ ] ] as @xmath262 are independent , @xmath263 are independent .
thus , from ( [ trace - norm])([spectral - norm - square ] ) we obtain ( [ e - trace - norm])([e - spectral - norm ] ) , and @xmath264 + \biggl\ { \sum _ { j=2}^p e\bigl[|\hat{\beta}_j - \beta_j|\bigr ] \biggr\}^2 - \sum_{j=2}^p \bigl\{e\bigl(|\hat{\beta}_j - \beta_j|\bigr)\bigr \}^2.\end{aligned}\ ] ] [ rem5 ] since pauli matrices @xmath30 have eigenvalues @xmath2 , the schatten @xmath137-norm @xmath265 .
similar to ( [ spectral - norm])([spectral - norm - square ] ) , we obtain that @xmath266 \label{s - norm - square } \\[-8pt ] \nonumber & = & d^{2/s } \biggl [ \sum_{j=2}^p | \hat{\beta}_j - \beta_j|^2 + 2 \sum _
{ i < j } ^p \bigl|(\hat{\beta}_i - \beta_i ) ( \hat{\beta}_j - \beta_j)\bigr| \biggr].\hspace*{-12pt } \nonumber\end{aligned}\ ] ] using ( [ threshold1 ] ) , we have @xmath268 + |\beta_j| p\bigl(|n_j| \leq\varpi \bigr ) \nonumber \\ & & \qquad \leq \bigl[e| n_j - \beta_j|^2 p\bigl(|n_j| \geq\varpi\bigr ) \bigr]^{1/2 } + \varpi p\bigl(|n_j| \geq\varpi\bigr ) + |\beta_j| p\bigl(|n_j| \leq\varpi\bigr ) \nonumber \\ & & \qquad \leq \bigl [ n^{-1 } \bigl(1-\beta_j^2\bigr ) p\bigl(|n_j| \geq\varpi\bigr ) \bigr]^{1/2 } + \varpi p\bigl(|n_j| \geq\varpi\bigr ) + |\beta_j| p\bigl(|n_j| \leq\varpi \bigr ) \nonumber \\ & & \qquad\leq 2 \varpi \bigl[p\bigl(|n_j| \geq\varpi\bigr ) \bigr]^{1/2 } + | \beta _
j| p\bigl(|n_j| \leq\varpi\bigr ) \nonumber \\ & & \qquad= 2 \varpi \bigl[p\bigl(|n_j| \geq\varpi\bigr ) \bigr]^{1/2 } \bigl\{1\bigl(|\beta _ j| > a_1 \varpi\bigr)+ 1\bigl(|\beta_j| \leq a_1 \varpi\bigr)\bigr\ } \nonumber \\ & & \qquad\quad{}+ |\beta_j| p\bigl(|n_j| \leq\varpi\bigr ) \bigl\{1\bigl(| \beta_j| >
a_2 \varpi\bigr)+1\bigl(|\beta_j| \leq a_2 \varpi\bigr)\bigr\ } \nonumber \\ & & \qquad\leq 2\varpi1\bigl(|\beta_j| > a_1 \varpi\bigr ) + 2 \varpi \bigl[p\bigl(|n_j| \geq\varpi\bigr ) \bigr]^{1/2 } 1\bigl(| \beta_j| \leq a_1 \varpi\bigr ) \nonumber \\ & & \qquad\quad{}+ p\bigl(|n_j|\leq\varpi\bigr ) 1\bigl(|\beta_j| > a_2
\varpi\bigr ) + |\beta _
j| 1\bigl(|\beta_j| \leq a_2 \varpi\bigr),\end{aligned}\ ] ] where @xmath269 and @xmath270 are two constants satisfying @xmath271 whose values will be chosen later , and @xmath272^{1/2 } 1\bigl(| \beta_j| \leq a_1 \varpi\bigr)\label{beta-1 } \\ & & { } + \sum_{j=2}^p p\bigl(|n_j| \leq\varpi\bigr ) 1\bigl(|\beta _
\varpi\bigr ) + \sum _ { j=2}^p |\beta_j| 1\bigl(| \beta_j| \leq a_2 \varpi\bigr ) . \nonumber\end{aligned}\ ] ] similarly , @xmath273
^ 2\\ & & \qquad \leq e\bigl[|\hat{\beta}_j - \beta _ j|^2\bigr ] \nonumber \\ & & \qquad\leq e\bigl
[ 2\bigl(| n_j - \beta_j|^2 + \varpi^2 \bigr ) 1\bigl(|n_j| \geq\varpi\bigr)\bigr ] + | \beta_j|^2 p\bigl(|n_j| \leq\varpi\bigr ) \nonumber \\ & & \qquad \leq 2 \bigl[e| n_j - \beta_j|^4 p\bigl(|n_j| \geq\varpi\bigr ) \bigr]^{1/2}\\ & & \qquad\quad { } + 2 \varpi ^2 p\bigl(|n_j| \geq\varpi\bigr ) + |\beta_j|^2 p\bigl(|n_j| \leq\varpi\bigr ) \nonumber \\ & & \qquad\leq c \varpi^2
\bigl[p\bigl(|n_j| \geq\varpi\bigr ) \bigr]^{1/2 } + |\beta_j|^2 p\bigl(|n_j| \leq\varpi\bigr ) \nonumber \\ & & \qquad = c \varpi^2 \bigl[p\bigl(|n_j| \geq\varpi\bigr ) \bigr]^{1/2 } \bigl\{1\bigl(|\beta_j| > a_1 \varpi\bigr)+ 1\bigl(|\beta_j| \leq a_1 \varpi\bigr)\bigr\ } \nonumber \\ & & \qquad\quad{}+ |\beta_j|^2 p\bigl(|n_j| \leq\varpi\bigr ) \bigl[1\bigl(|\beta_j| > a_2 \varpi\bigr ) + 1\bigl(| \beta_j| \leq a_2 \varpi\bigr)\bigr ]
\nonumber \\ & & \qquad \leq c \varpi^2 1\bigl(|\beta_j| > a_1 \varpi\bigr ) + c \varpi^2 \bigl[p\bigl(|n_j| \geq\varpi\bigr ) \bigr]^{1/2 } 1\bigl(|\beta_j| \leq a_1
\varpi\bigr ) \nonumber \\ & & \qquad\quad{}+ p\bigl(|n_j|\leq\varpi\bigr ) 1\bigl(|\beta_j| > a_2
\varpi\bigr ) + |\beta_j|^2 1\bigl(|\beta_j| \leq a_2 \varpi\bigr),\end{aligned}\ ] ] and
@xmath274\nonumber\\ & & \qquad \leq c \varpi^2 \sum_{j=2}^p 1\bigl(| \beta_j| >
a_1 \varpi\bigr)\nonumber \\[-8pt ] \label{beta-2 } \\[-8pt ] \nonumber & & \qquad\quad { } + c \varpi^2 \sum_{j=2}^p \bigl[p\bigl(|n_j| \geq\varpi \bigr ) \bigr]^{1/2 } 1\bigl(| \beta_j| \leq a_1 \varpi\bigr ) \\ & & \qquad\quad{}+ \sum_{j=2}^p p\bigl(|n_j|
\leq\varpi\bigr ) 1\bigl(|\beta_j| > a_2 \varpi\bigr ) + \sum _ { j=2}^p |\beta_j|^2 1\bigl(| \beta_j| \leq a_2 \varpi \bigr ) .
\nonumber\end{aligned}\ ] ] by lemma [ lem - sparse ] , we have @xmath275 \label{beta - sparse-11 } \\[-8pt ] \nonumber & & \qquad \leq ( a _ 2 \varpi)^{2-q } \sum _ { j=2}^{p } |\beta_j|^q 1\bigl(| \beta_j| \leq a_2 \varpi\bigr ) \leq a_2^{2-q } \pi_n(d ) \varpi^{2-q } , \\
\label{beta - sparse-2 } & & \varpi \sum_{j=2}^p 1\bigl(| \beta_j| \geq a_1 \varpi\bigr ) \leq a_1^{-q}\pi_n(d ) \varpi^{1-q}.\end{aligned}\ ] ] on the other hand , @xmath276 \\ & & \qquad\leq4 p^{1- \vafrac{\hbar^2 |a_2 - 1|^2}{1+o(1 ) } } = 4 p^{-1 - ( 2-q)/(2 c_0 ) } \leq4 p^{-1 } n^{-(q-2)/2}\nonumber\\ & & \qquad = o\bigl(\pi_n(d ) \varpi^{2-q}\bigr ) , \nonumber\end{aligned}\ ] ] where the third inequality is from lemma [ lem - tail ] , the first equality is due the fact that we take @xmath277 so that @xmath278 , and @xmath279 is the constant in assumption @xmath280 . finally , we can show @xmath281^{1/2 } 1\bigl(| \beta_j| \leq a_1 \varpi\bigr ) \\ & & \qquad\leq\varpi \sum_{j=2}^p \bigl[p ( n_j - \beta_j \leq- \varpi- \beta _ j ) \nonumber \\ \label{beta - sparse-4 } & & \quad\qquad{}+ p(n_j - \beta_j \geq\varpi- \beta_j ) \bigr]^{1/2 } 1\bigl(|\beta_j| \leq a_1 \varpi\bigr ) \\ & & \qquad\leq\varpi \sum_{j=2}^p \bigl[p\bigl ( n_j - \beta_j \leq- |1-a_1| \varpi \bigr ) + p\bigl(n_j - \beta_j \geq|1-a_1| \varpi\bigr ) \bigr]^{1/2 } \nonumber \\ & & \qquad \leq2\varpi p^{1- \hbar^2 ( 1-a_1)^2/(2(1+o(1 ) ) ) } = 2\varpi p^{-1 } = o\bigl ( \pi_n(d )
\varpi^{1-q}\bigr ) , \nonumber\end{aligned}\ ] ] where the third inequality is from lemma [ lem - tail ] , and the first equality is due to the fact that we take @xmath282 so that @xmath283 . plugging ( [ beta - sparse-1])([beta - sparse-4 ] ) into ( [ beta-1 ] ) and ( [ beta-2 ] ) , we prove the lemma .
proof of theorem [ thm-3 ] combining lemma [ lem - thm3 ] and ( [ e - trace - norm])([e - spectral - norm ] ) in lemma [ lem - norm ] , we easily obtain @xmath284 & \leq & c_1 \frac{\pi_n(d)}{p^{1/2 } } \biggl ( \frac { \log p } { n } \biggr)^{\vfrac{1-q}{2 } } , \\ e\bigl[\| \hat{{\bolds{\rho } } } - { \bolds{\rho}}\|_f^2\bigr ] & \leq & c_0 \pi_n(d ) \frac{1}{d } \biggl ( \frac { \log p } { n } \biggr)^{1-q/2}.\end{aligned}\ ] ] using lemma [ lem - thm3 ] and ( [ e - spectral - norm - square ] ) in lemma [ lem - norm ] , we conclude @xmath285 & \leq & c_2 \biggl [ \pi^2_n(d ) \frac{1}{p } \biggl ( \frac { \log p } { n } \biggr)^{1-q } + \pi_n(d ) \frac{1}{p } \biggl ( \frac { \log p } { n } \biggr)^{1-q/2 } \biggr ] \nonumber \\[-8pt ] \label{equation - spectral - norm - square } \\[-8pt ] \nonumber & \leq & c \frac{\pi^2_n(d)}{d^2 } \biggl ( \frac { \log p } { n } \biggr)^{1-q},\end{aligned}\ ] ] where the last inequality is due to the fact that the first term on the right - hand side of ( [ equation - spectral - norm - square ] ) dominates its second term . proof of proposition [ schatten.prop ] applying the lyapunov s moment inequality to the schatten @xmath137-norm , we have for @xmath286 $ ] and @xmath287 , @xmath288 & \leq & d^{-1 + 2/s } e \bigl [ \| \hat{{\bolds{\rho } } } -{\bolds{\rho}}\|_{*2 } ^2 \bigr ] \\ & = & d^{-1 + 2/s } e \bigl [ \| \hat{{\bolds{\rho } } } -{\bolds{\rho}}\|_{f } ^2 \bigr ] \\ & \leq & c_1 \pi_n ( d ) d^{-2 + 2/s } \biggl(\frac{\log d}{n } \biggr ) ^{1-q/2},\end{aligned}\ ] ] where the last inequality is due to theorem [ thm-3 ] . on the other hand , applying hlder s inequality by interpolating between schatten @xmath137-norms with @xmath289 and @xmath290
, we obtain for @xmath291 $ ] and @xmath287 , @xmath292 & \leq & e \bigl [ \| \hat{{\bolds{\rho}}}- { \bolds{\rho}}\|_{*2}^{4/s } \| \hat{{\bolds{\rho}}}- { \bolds{\rho}}\| _ { * \infty } ^{2 - 4/s } \bigr ] \\ & \leq & \bigl [ e \| \hat { { \bolds{\rho}}}- { \bolds{\rho}}\|_{*2}^{2 } \bigr ] ^{2/s } \bigl[e \| \hat{{\bolds{\rho}}}- { \bolds{\rho}}\|_{*\infty } ^{2 } \bigr ] ^{1 - 2/s } \\ & \leq & c_7 \pi_n ^{2 - 2/s}(d ) d^{-2 + 2/s } \biggl(\frac{\log d}{n } \biggr ) ^{1-q+q / s},\end{aligned}\ ] ] where the last inequality is due to theorem [ thm-3 ] , and @xmath293 .
the result follows by combining the above two inequalities together .
proof of theorem [ thm-4 ] we first define a subset of the parameter space @xmath90 .
it will be shown later that the risk upper bound under the spectral norm is sharp up to a constant factor , when the parameter space is sufficiently sparse .
consider a subset of the pauli basis , @xmath294 , where @xmath295 or @xmath296 .
its cardinality is @xmath297 .
denote each element of the subset by @xmath298 , @xmath299 , and let @xmath300 .
we will define each element of @xmath90 as a linear combination of @xmath298 .
let @xmath301 , @xmath302 , and denote @xmath303 .
the value of @xmath304 is either @xmath305 or @xmath306 , where @xmath306 is the largest integer less than or equal to @xmath307 . by assumption ( [ cond - pi ] )
, we have @xmath308 let @xmath309 and set @xmath310 .
now we are ready to define @xmath90 , @xmath311 note that @xmath90 is a subset of the parameter space , since @xmath312 and its cardinality is @xmath313 .
we need to show that @xmath314 note that for each element in @xmath90 , its first entry @xmath315 may take the form @xmath316 .
it can be shown that @xmath317 it is then enough to show that @xmath318 which immediately implies @xmath319 we prove equation ( [ klowerbd ] ) by applying le cam s lemma . from observations
@xmath320 , @xmath321 , we define @xmath322 , which is @xmath323 .
let @xmath324 be the joint distribution of independent random variables @xmath325 .
the cardinality of @xmath326 is @xmath313 . for two probability measures
@xmath327 and @xmath328 with density @xmath329 and @xmath330 with respect to any common dominating measure @xmath331 , write the total variation affinity @xmath332 , and the chi - square distance @xmath333 . define @xmath334 the following lemma is a direct consequence of le cam s lemma [ cf .
@xcite and @xcite ] .
it is enough to show that @xmath342 which implies @xmath343 , then we have @xmath344 .
let @xmath345 denote the number of overlapping nonzero coordinates between @xmath346 and @xmath347 .
note that @xmath348 when @xmath349 , we have @xmath350 \biggr ) ^{j } \\
& = & \biggl ( \sum_{l=0}^{n } \biggl [ \pmatrix{n \cr l } \biggl ( \frac { ( 1+a ) ^{2}}{2 } \biggr ) ^{l } \biggl ( \frac { ( 1-a ) ^{2}}{2 } \biggr )
^{n - l } \biggr ] \biggr ) ^{j } \\ & = & \biggl ( \frac { ( 1+a ) ^{2}}{2}+\frac { ( 1-a ) ^{2}}{2 } \biggr ) ^{nj } \\ & = & \bigl ( 1+a^{2 } \bigr ) ^{nj}\leq\exp \bigl ( na^{2}j \bigr),\end{aligned}\ ] ] which implies @xmath351 since @xmath352 ^{2}\cdot ( d-1-k ) \cdot \ldots\cdot ( d-2k+j ) } { j!\cdot ( d-1 ) \cdot \ldots \cdot ( d - k ) } \\ & \leq&\frac{k^{2j } ( d-1-k ) ^{k - j } } { ( d - k ) ^{k}}\leq \biggl ( \frac{k^{2}}{d - k } \biggr ) ^{j},\end{aligned}\ ] ] and @xmath309 , we then have @xmath353 ^{j } \leq\sum_{1\leq j\leq k } \biggl [ \frac{d^{2v+ ( 1 - 2v ) /2}}{d - k } \biggr ] ^{j}\rightarrow0.\end{aligned}\ ] ] apply assouad s lemma , and we show below that @xmath134 \geq c \pi_n(p ) \frac{1}{d } \biggl ( \frac { \log p } { n } \biggr)^{1-q/2},\ ] ] where @xmath111 denotes any estimator of @xmath8 based on measurement data @xmath262 , and @xmath180 is a constant free of @xmath42 and @xmath181 . to this end
, it suffices to construct a collection of @xmath356 density matrices @xmath357 such that ( i ) for any distinct @xmath358 and @xmath359 , @xmath360 where @xmath191 is a constant ; ( ii ) there exists a constant @xmath361 such that @xmath362 where @xmath363 denotes the kullback
leibler divergence . by the gilbert
varshamov bound [ cf .
@xcite ] , we have that for any @xmath364 , there exist @xmath365 binary vectors @xmath366 , @xmath367 , such that ( i ) @xmath368 , ( ii ) @xmath369 , and ( iii ) @xmath370 .
let @xmath371 where @xmath372 since @xmath373 , @xmath374 whenever @xmath375 .
moreover , @xmath376 on the other hand , @xmath377 now the lower bound can be established by taking @xmath378 and then @xmath379 which are allowed by the assumption @xmath380 for @xmath381 . | quantum state tomography aims to determine the state of a quantum system as represented by a density matrix .
it is a fundamental task in modern scientific studies involving quantum systems . in this paper , we study estimation of high - dimensional density matrices based on pauli measurements . in particular , under appropriate notion of sparsity , we establish the minimax optimal rates of convergence for estimation of the density matrix under both the spectral and frobenius norm losses ; and show how these rates can be achieved by a common thresholding approach .
numerical performance of the proposed estimator is also investigated .
./style / arxiv - general.cfg , , , |
let @xmath0 be an @xmath1-dimensional manifold , and let @xmath2 be a hyperplane distribution .
we say the pair @xmath3 is a * contact manifold * if @xmath4 is maximally non - integrable .
the primary example of a contact manifold we will use is the * one - jet manifold of @xmath5 * , @xmath6 , where @xmath7 , @xmath8 is the coordinate of @xmath9 and @xmath10 is the pullback to @xmath11 of the tautological one - form of @xmath12 .
let @xmath13 ; we say an @xmath14-dimensional submanifold @xmath15 is * legendrian * if @xmath16 .
this is equivalent to : @xmath17 we define the * lagrangian projection * @xmath18 , * front projection * @xmath19 , and * base projection * @xmath20 to be the projections : @xmath21@xmath22@xmath23 let @xmath24 be a @xmath25-dimensional manifold , and let @xmath26 be a two - form on @xmath24 .
we say @xmath26 is a * symplectic form * and @xmath24 is a * symplectic manifold * if @xmath26 is closed and non - degenerate .
we say @xmath27 is * exact symplectic * if @xmath26 is exact .
the * symplecticization * of @xmath11 is given by @xmath28 , where @xmath29 is the coordinate of @xmath9 ; this is then a symplectic manifold .
we say an @xmath30-dimensional submanifold @xmath31 is * lagrangian * if : @xmath32 we say @xmath33 is * exact lagrangian * if : @xmath34 for some function @xmath35 .
we say two legendrian submanifolds are * legendrian isotopic * if they are isotopic through legendrians , and similarly we say two exact lagrangian submanifolds are * exact lagrangian isotopic * if they are isotopic through exact lagrangians .
we are interested in calculating invariants of compact legendrians , one of which is the * legendrian contact homology * , which is the homology of a unital graded algebra @xmath36 over a field @xmath37 generated by the reeb chords of @xmath38 and with a differential given by counting rigid pseudoholomorphic disks .
( we will ordinarily assume @xmath39 , in which case we will suppress the subscript , but our primary results hold for arbitrary field . )
we first prove a seifert - van kampen theorem for legendrian contact homology . if @xmath40 , we define @xmath41 to be @xmath42 restricted to reeb chords lying over @xmath43 .
this may or may not be a well - defined differential graded algebra .
however : [ theorem1.1 ] let @xmath15 be a compact legendrian submanifold , and let @xmath44 be a hypersurface that divides @xmath5 into two components , @xmath45 and @xmath46 .
let @xmath47 be an arbitrarily small neighborhood of @xmath48 , and let @xmath49 . then , after a legendrian isotopy that does not change @xmath38 outside of @xmath50 , @xmath51 are well - defined differential graded algebras , and the following diagram is a push - out square : @xmath52^{i_1}\\ \mathcal{a}_k(\lambda)|_{q_2 } \ar[u]^{i_2 } & \mathcal{a}_k(\lambda)|_n \ar[u]^{j_1}\ar[l]^{j_2}}\ ] ] where @xmath53 are the inclusion maps .
the legendrian isotopy in theorem [ theorem1.1 ] is called the * pinching isotopy * ( also referred to as * dipping * in @xcite , and similar to the * splashing * of @xcite ) .
this result has been known for @xmath54 since @xcite , and for @xmath55 and for higher dimensions where the front projection has no codimension-2 singularities since @xcite .
we prove it for all cases .
note that , since pinching does not change @xmath38 outside of @xmath47 , we can apply multiple pinching to separate @xmath38 into more then two components .
in addition , we can pinch inside the pinch zone , to separate @xmath47 itself into multiple components .
this separation also descends to the linearized contact homologies if they exist , providing a mayer - veitoris theorem : [ theorem1.3 ] let @xmath15 be a compact legendrian submanifold that has been pinched along a neighborhood @xmath47 of a hyper surface @xmath48 that divides @xmath5 into @xmath56 .
if @xmath38 has an augmentation @xmath57 , then @xmath57 induces augmentations @xmath58 of @xmath59 , and there is a long exact sequence : @xmath60@xmath61 where @xmath62 denotes the linearized contact homology . next ,
let @xmath63 be legendrian submanifolds .
we define an * exact lagrangian cobordism * from @xmath64 to @xmath65 to be an exact lagrangian submanifold @xmath31 such that : * there exists @xmath66 such that : @xmath67@xmath68 * if @xmath69 , then @xmath35 is constant on @xmath70 and on @xmath71 .
* @xmath72 is compact with boundary @xmath73 .
we refer to @xmath74 as the * positive and negative cones of @xmath33*. an exact lagrangian cobordism @xmath33 defines a homomorphism @xmath75 , as discussed in section [ subsec : linearized ] .
we have an equivalent of our sievert - van kampen theorem for these cobordisms as well : [ theorem1.4 ] let @xmath31 be an exact lagrangian cobordism from @xmath64 to @xmath65 , and let @xmath76 be a hypersurface that divides @xmath77 into two components @xmath78 and @xmath79 , such that @xmath80 is also a hypersurface for any choice of @xmath81 , and such that @xmath82 does not cross any cusp edges of the legendrian lift of @xmath33 .
let @xmath83 , @xmath84 , let @xmath85 be an arbitrarily small neighborhood of @xmath86 , and let @xmath87 .
then @xmath33 is exact lagrangian isotopic to an exact lagrangian cobordism @xmath88 from @xmath89 to @xmath90 , where @xmath91 are legendrian isotopic to @xmath92 , and such that the image of the restriction of @xmath93 to @xmath94 lies in @xmath95 . as applications of theorems [ theorem1.1 ] and [ theorem1.4 ]
, we then prove that : [ theorem1.5 ] let @xmath15 be a legendrian submanifold , and let @xmath96 be the product of a critical legendrian ambient surgery on @xmath38 .
then the differential graded algebras @xmath42 and @xmath97 are stable tame isomorphic to a pair of differential graded algebras @xmath98 and @xmath99 , such that : * let @xmath100 be the generators of @xmath99 .
then the generators of @xmath98 are @xmath101 . * for @xmath102 ,
@xmath103 * @xmath104 . in section [ sec : overview ]
, we provide an overview of the concepts behind legendrian contact homology and exact lagrangian cobordisms . in section [
sec : surgery ] , we will use theorem [ theorem1.1 ] to prove theorem [ theorem1.5 ] . in section [ sec :
proof ] , we will prove theorems [ theorem1.1 ] , [ theorem1.3 ] , and [ theorem1.4 ] .
a * contact manifold * is a @xmath1-dimensional manifold @xmath0 equipped with a hyperplane distribution @xmath4 such that there exists no embedded @xmath25-dimensional submanifold that is everywhere tangent to @xmath4 . a * contact form * @xmath105 is a one - form in @xmath106 such that @xmath107 .
although a contact form can always be found locally , a global contact form does not always exist .
however , in our case we are limiting our interest to one - jet manifolds @xmath108 , where @xmath4 is the kernel of the global contact form @xmath7 , where @xmath10 is the tautological one - form of @xmath12 and @xmath8 is the @xmath9 coordinate . in addition , we are particularly interested in the case @xmath109 , @xmath110 . in this subcase , we use @xmath111 for coordinates on @xmath5 and @xmath112 to be the corresponding cotangent coordinates . then : @xmath113 an @xmath14-dimensional submanifold @xmath114 is called * legendrian * if @xmath115 .
this is equivalent to @xmath116 .
a * symplectic manifold * is a @xmath25-dimensional manifold @xmath24 equipped with a closed 2-form @xmath26 such that @xmath117 is a volume form .
we are interested in symplectic manifolds in part because we will frequently be working in @xmath12 , which is equipped with the canonical symplectic form @xmath118 .
recall that an @xmath14-dimensional submanifold @xmath33 of a symplectic manifold is called * lagrangian * if @xmath119 .
if @xmath15 is an embedded legendrian submanifold , then @xmath120 is an immersed lagrangian submanifold , because we can define a function @xmath121 that is simply the @xmath8 coordinate of that point in @xmath38 , and : @xmath122 an * almost complex structure * on a @xmath25-dimensional manifold @xmath24 is a smooth linear map @xmath123 such that @xmath124 .
the standard example is the complex plane @xmath125 with real coordinates @xmath126 and canonical contact form @xmath127 , where : @xmath128@xmath129 given two manifolds with almost complex structures @xmath130 and @xmath131 , we say that a function @xmath132 is * pseudoholomorphic * if @xmath133 .
let @xmath134 be a @xmath25-dimensional manifold equipped with an almost complex structure @xmath135 , a riemannian metric @xmath136 , and a symplectic form @xmath26 .
we say that the triple @xmath137 is * tame * if @xmath136 is complete and there exists constants @xmath138 such that : * every loop @xmath139 contained in a ball @xmath140 with @xmath141 bounds a disc in @xmath142 of area less then @xmath143 . * @xmath144 for all @xmath145 * for every vector @xmath146 , @xmath147
. we can always equip @xmath12 with a tame triple @xmath137 ( @xcite , ch . 5 , sec .
4.1 ) , and from now on will assume we have done so .
furthermore , the space of almost complex structures @xmath135 on @xmath12 which is part of a tame triple is contractible ( @xcite , proposition 4.1 . )
we say that a legendrian submanifold is * front generic * if it has the following properties : * the base space projection @xmath148 is an immersion outside of a codimension-1 submanifold @xmath149 . *
we define @xmath150 inductively to be a codimension-1 subset of @xmath151 , such that the map @xmath152 is an immersion .
* at points @xmath153 , @xmath154 has a standard * cusp edge singularity*. that is , there exist coordinates @xmath155 of @xmath38 around @xmath156 and coordinates @xmath111 of @xmath5 around @xmath157 such that , if @xmath8 is the fiber coordinate in @xmath158 , then @xmath159 , where : @xmath160@xmath161@xmath162 where @xmath163 are constants .
as discussed in @xcite , section 2.2.1 , any legendrian submanifold @xmath38 can be made front generic after an arbitrarily small legendrian isotopy .
we will therefore assume from here on that our legendrian submanifolds are all front generic .
a * reeb vector field * @xmath164 is the vector field of @xmath11 such that : @xmath165@xmath166 for @xmath11 , @xmath167 .
a * reeb chord * of a legendrian submanifold @xmath38 is a trajectory of @xmath164 that begins and ends on @xmath38 . for @xmath167 ,
reeb chords correspond precisely to self - intersections of @xmath168 .
these intersections will generically be transverse double points .
we label the quadrants of each reeb chord in @xmath168 with the signs shown in figure [ fig : signsofquadrants ] .
[ fig : signsofquadrants ] ] we define @xmath36 to be the formal unital algebra over @xmath37 , where @xmath39 or @xmath169 , generated by the reeb chords of @xmath38 - that is , elements of @xmath36 are sentences made of words @xmath170 , where @xmath171 and @xmath172 are reeb chords .
we provide a differential to @xmath36 by counting rigid pseudoholomorphic disks in @xmath12 . first
, for every reeb chord @xmath173 of @xmath38 , let @xmath174 be the upper and lower end points of the reeb chord .
we choose a * capping path * for each chord @xmath173 , which is a path @xmath175 \to \lambda$ ] such that @xmath176 .
we define @xmath177 to be the unit disk in @xmath125 with the canonical complex structure , with @xmath178 punctures on its boundary labeled @xmath179 .
then , if @xmath180 are reeb chords and @xmath98 is an element in @xmath181 , we define the * moduli space * @xmath182 to be the space of maps @xmath183 such that : * @xmath184 is pseudoholomorphic . * as @xmath8 limits to @xmath185 , @xmath186 limits to a positive quadrant of @xmath187 . * as @xmath8 limits to @xmath188 , @xmath189 , @xmath186 limits to a negative quadrant of @xmath190 . *
the restriction of @xmath184 to @xmath191 has a continuous lift @xmath192 . *
the homology class of @xmath193 equals @xmath194 .
we define @xmath195 in the same fashion , except that we drop the requirement on the homology class .
these moduli spaces are open manifolds with boundary .
we define the differential @xmath196 of @xmath36 by counting 0-dimensional moduli spaces .
if @xmath39 , then : @xmath197 if @xmath198 , then : @xmath199 we extend these differentials to @xmath200 using the linearity and the product rules : @xmath201@xmath202 we say that a pseudoholomorphic disk in a 0-dimensional moduli space is * rigid*. note that we could have zero negative punctures of a rigid pseudoholomorphic disk , in which case the disk contributes 1 to the differential . and : [ theorem2.1 ] @xmath203 , so the homology of @xmath36 is well - defined .
* proof : * see @xcite , proposition 2.6 .
the differential graded algebra @xmath36 is _ not _ an invariant of the legendrian isotopy class of @xmath38 .
however , its stable tame isomorphism class _ is_. a * stabilization * of @xmath36 , which we denote @xmath204 , is a differential graded algebra whose generators consist of the generators of @xmath36 , plus two generators @xmath205 , and whose differential @xmath206 is defined to be : @xmath207@xmath208@xmath209 an * elementary automorphism * of a differential graded algebra generated by @xmath210 is an automorphism such that there exists some @xmath211 so that @xmath212 for @xmath213 and @xmath214 , where @xmath184 is a word in @xmath215 .
a * tame automorphism * is an automorphism that is the composition of a series of elementary automorphisms
. a * tame isomorphism * between a differential graded algebra generated by @xmath210 and a differential graded algebra generated by @xmath216 is the composition of a tame automorphism with the isomorphism sending @xmath217 .
we say that two differential graded algebras @xmath218 are * stable tame isomorphic * if they are tame isomorphic after some series of stabilizations . and
: [ theorem2.2 ] the stable tame isomorphism class of @xmath36 is invariant under legendrian isotopy . *
proof : * see @xcite , proposition 2.7 . the homology of @xmath36 is invariant under stable tame isomorphism , and so is in turn an invariant of the legendrian isotopy class .
this homology is called the legendrian contact homology of @xmath38 . when @xmath219 , it is easy to find these pseudoholomorphic disks , because , thanks to the riemann mapping theorem , they correspond to polygons in @xmath12 whose vertices are reeb chords and whose edges lie on @xmath168 .
it is significantly harder to calculate the differential when @xmath220 .
fortunately , as we discuss in the next subsection , @xcite provides a technique for calculating the differential in a very broad special case .
@xcite provides a technique for calculating @xmath196 provided that either @xmath54 or 2 or the front projection @xmath19 has no codimension-2 or higher singularities that are not cusp edge intersections .
this is a large and very interesting class of legendrian submanifolds .
he does this using an object called a gradient flow tree .
consider the front projection of a legendrian submanifold .
we can regard this as the graph of some collection of * height functions * @xmath221 .
a path @xmath222 \to m$ ] is a * flow line * of the height functions @xmath223 if @xmath224 .
if @xmath222\to m$ ] is a flow line , the * 1-jet lift * of @xmath225 is an unordered pair @xmath226 \to \lambda \subset j^1(m)$ ] of continuous paths such that @xmath227 .
we define the * flow orientation * of @xmath228 at a point @xmath229 to be the unique lift of the vector @xmath230 to @xmath231 .
let @xmath232 be a tree with finitely many edges . for @xmath233 , at each @xmath234-valent vertex , place a cyclic ordering on the edges .
a map @xmath235 is a * gradient flow tree * if it satisfies the following conditions : * if @xmath236 is an edge of @xmath232 then the restriction of @xmath237 to @xmath236 is a flow line .
* if @xmath238 is a @xmath234-valent vertex with ordered edges @xmath239 , and @xmath240 are the 1-jet lifts of @xmath237 restricted to @xmath241 , then we require that @xmath242 , with the convention @xmath243 .
in addition , we require that the flow orientation of @xmath244 at @xmath238 is directed towards @xmath238 if and only if the flow orientation of @xmath245 is directed away . *
the lagrangian projection of the 1-jet lifts of the edges of @xmath232 fit together to form a closed oriented curve @xmath246 in @xmath168 .
let @xmath247 be the space of gradient flow trees with a positive puncture at @xmath187 and negative punctures at @xmath172 .
let @xmath248 be the lift of @xmath246 to @xmath38 , and let @xmath249 be the restriction of @xmath247 to trees such that the homology class of @xmath250 equals @xmath98 in @xmath181 . [ theorem2.3 ] if @xmath54 or 2 or if the only singularities of @xmath19 are cusp edges , then there exists an almost complex structure such that there is a bijection between @xmath247 and @xmath195 , and between @xmath249 and @xmath182 , if the disks are rigid .
* proof : * see @xcite , theorem 1.1 . for a reeb chord @xmath251 , we will use @xmath252 to denote the morse index of the height difference function at @xmath251 . if @xmath253 we say @xmath251 is a * minimum * , and if @xmath254 we say @xmath251 is a * maximum*. the vertices of a gradient flow tree may correspond to reeb chords or they may correspond to other joinings of flow lines .
we label the following kinds of non - chord vertices : * * end : * one - valent vertex where a flow line between two sheets meets a cusp edge of that sheet .
* * switches : * two - valent vertex where a flow line between sheets @xmath127 and @xmath211 , where @xmath127 lies above @xmath211 , meets a cusp edge between sheets @xmath127 and @xmath234 ( or @xmath234 and @xmath211 ) , and leaves as a flow line between sheets @xmath234 and @xmath211 ( or @xmath127 and @xmath234 ) . * * @xmath255 vertices : * three - valent vertex where a flow line between sheets @xmath127 and @xmath211 splits into two flow lines , one between sheets @xmath127 and @xmath234 and one between sheets @xmath234 and @xmath211 , where the @xmath8 coordinate of sheet @xmath234 lies between the @xmath8 coordinates of sheets @xmath127 and @xmath211 . * * @xmath256 vertices : * three - valent vertex where a flow line between sheets @xmath127 and @xmath211 meets a cusp edge between sheets @xmath234 and @xmath257 , where @xmath127 lies above @xmath234 and @xmath257 lies above @xmath211 , and leaves as two flow lines , one between sheets @xmath127 and @xmath257 and one between sheets @xmath234 and @xmath211 . ]
( _ solid lines are gradient flow lines , dotted lines are cusp edges _ ) [ theorem2.4 ] let @xmath232 be a gradient flow tree with positive reeb chords @xmath258 , negative reeb chords @xmath259 , and non - chord vertices @xmath260 .
the dimension of the moduli space containing @xmath232 is then : @xmath261 where @xmath262 is the morse index of @xmath263 if @xmath263 is a reeb chord , @xmath264 if @xmath263 is a positive special puncture , @xmath265 if @xmath263 is a negative special puncture , @xmath266 if @xmath267 is an end , 0 if @xmath267 is a @xmath255 vertex , and -1 if @xmath267 is a switch or @xmath256 vertex .
* proof : * see @xcite , prop .
[ theorem2.5 ] any vertex of a rigid gradient flow tree with one positive puncture must be a one - valent vertex at a positive or negative reeb chord ; a two - valent vertex at a positive minimum or a negative maximum ; a cusp edge ; a switch ; a @xmath255 vertex ; or a @xmath256 vertex .
* proof : * see @xcite , lemma 3.7 .
we will also make use of objects called * partial flow trees * , which are gradient flow trees that have one or more * special punctures * , which are 1-valent vertices that are neither cusp edges nor reeb chords . a special puncture may be considered to be either positive or negative , depending on the orientation of the 1-jet lift . if @xmath238 is a special puncture , then we treat @xmath268 if @xmath238 is positive , @xmath269 if @xmath238 is negative .
critically , given any gradient flow tree @xmath232 , and any point @xmath251 in a flow line ( not a vertex ) of that tree , we can separate @xmath232 into a pair of partial flow trees , @xmath270 and @xmath271 , by breaking @xmath232 at @xmath251 . then : @xmath272 the advantage of gradient flow trees is that they reduce finding the rigid pseudoholomorphic disks from an infinite - dimensional problem to a finite - dimensional problem .
although we can not always use them , given the requirements on @xmath19 , they are a very valuable tool when they are useable . the essence of the proof of theorems 1.1 through 1.5 will be to show that , although we can not prove that pseudoholomorphic disks are globally equivalent to gradient flow trees if @xmath19 has higher - codimension singularities , we can show they are _ locally _ equivalent .
the legendrian contact homology of a legendrian submanifold @xmath15 is , generally , infinite - dimensional , and often quite difficult to work with .
we therefore define algebraic invariants of the legendrian contact homology which will be finite - dimensional .
first , we split the differential @xmath196 into components @xmath273 , etc . , where @xmath274 denotes the restriction of @xmath196 to words of length @xmath234 .
since @xmath275 , we obtain : @xmath276@xmath277@xmath278 etc .
this means that , if @xmath279 , then @xmath280 , which means that we can define a * linearized chain complex * given by the restriction of @xmath42 to words of unit length , and equipped with the differential @xmath281 .
the homology of this algebra will then be finite - dimensional .
unfortunately , @xmath282 in general .
however , @xmath42 may be stably tame isomorphic to a dga where @xmath283 _ does _ equal zero .
finding such an isomorphism is equivalent to finding a graded algebra map @xmath284 such that @xmath285 and @xmath286 .
we call such a map an * augmentation*. given an augmentation , we define a dga map by @xmath287 if @xmath263 is a reeb chord , and extend to products multiplicatively - that is : @xmath288 @xmath289 is a stable tame isomorphism to the differential graded algebra with the same generators as @xmath42 , but with differential @xmath290 .
@xmath291 will always equal zero , so we can define a linearized chain complex for this augmented algebra .
the homology of this complex is called the * linearized contact homology * , or @xmath292 .
since @xmath57 is not necessarily unique , a given legendrian may have multiple linearized contact homologies , and two different maps @xmath57 may produce the same linearized contact homology .
however : [ theorem2.6 ] the set of linearized contact homologies of @xmath38 is an invariant of the stable tame isomorphism class of @xmath42 , and therefore is an invariant of the legendrian isotopy class of @xmath38 .
* proof : * see @xcite , theorem 5.2 .
let @xmath293 .
the symplectization @xmath294 is symplectomorphic to the cotangent bundle with canonical symplectic form @xmath295 by the map : @xmath296 although this is not an _ exact _ symplectomorphism , the difference between the primitive of the symplectic form of @xmath297 and the tautological one - form of @xmath298 _ is _ exact : @xmath299 therefore , if @xmath300 , then @xmath301 , where @xmath10 is the tautological one - form of @xmath302 .
so @xmath303 is still an exact lagrangian cobordism .
we will use @xmath33 to refer to @xmath303 from here on , and work exclusively in @xmath302 .
we prefer this environment because it is the native environment for our pseudoholomorphic disks .
an exact lagrangian cobordism @xmath33 from @xmath64 to @xmath65 induces a dga map @xmath304 , given by counting rigid pseudoholomorphic disks whose punctures limit to the reeb chords at infinity .
that is , if @xmath187 is a reeb chord in @xmath64 and @xmath172 are reeb chords in @xmath65 , we define the moduli space @xmath305 to be the space of maps @xmath306 such that : * @xmath184 is pseudoholomorphic , that is , @xmath307 , where @xmath308 is the almost complex structure on the domain of @xmath184 . * @xmath309 limits to @xmath310 as @xmath311 , where @xmath187 is a reeb chord of @xmath64 . * @xmath309 limits to @xmath312 as @xmath313 , where @xmath190 is a reeb chord of @xmath65 .
we then define @xmath314 by : @xmath315 let @xmath316 denote the differential of @xmath92 . then : [ theorem2.7 ] @xmath317 * proof : * see @xcite , theorem 1.2 .
[ theorem2.8 ] if @xmath318 and @xmath319 are exact lagrangian isotopic , then @xmath320 is chain homotopic to @xmath321 .
* proof : * see @xcite , theorem 1.3 .
note that , given an exact lagrangian cobordism @xmath322 from @xmath323 to @xmath324 , and a second exact lagrangian cobordism @xmath325 from @xmath324 to @xmath326 , we can form a new exact lagrangian cobordism @xmath327 from @xmath323 to @xmath326 by concatenating @xmath322 and @xmath325 : we delete the negative cone of @xmath322 and positive cone of @xmath325 and glue them together .
then : [ theorem2.9 ] @xmath328 * proof : * see @xcite , theorem 1.2 .
this section is derived from @xcite , section 2 , with minor modifications to adapt the concept to higher dimensions - principally adding extra coordinates as appropriate .
recall that , according to the definition of an exact lagrangian cobordism @xmath329 , there exists a function @xmath330 such that @xmath331 .
we can lift @xmath332 to @xmath333 , with @xmath334 unique up to @xmath9-translation , by mapping @xmath335 to @xmath336 .
let @xmath337 be the canonical contact form of @xmath338 ; then : @xmath339 so @xmath334 is legendrian , though it is definitely _ not _ compact .
however , as we will see in section [ sec : proof ] , we do not _ need _ compactness .
we will prove convergence of the boundary of pseudoholomorphic curves to gradient flow lines in @xmath38 , where @xmath38 may denote a compact legendrian submanifold of @xmath11 or it may denote the legendrian lift @xmath340 of an exact lagrangian cobordism , and where @xmath5 may mean @xmath5 or may mean @xmath341 . this does create some confusion between the @xmath8 coordinate of @xmath92 and the new @xmath8 coordinate of @xmath334 given by integrating @xmath342 . in section [ sec : proof ] , when we refer to the @xmath8 coordinate of @xmath38 , we will always mean the @xmath9 coordinate of the one - jet space , whether that one - jet space happens to be @xmath11 or @xmath338 .
legendrian ambient surgery is defined in @xcite , sec . 3 , as an analogue of morse surgery for legendrians .
we begin with the following data : let @xmath15 be an @xmath14-dimensional legendrian submanifold , and let @xmath343 .
let @xmath344 be an embedded @xmath234-sphere with a choice of framing @xmath237 of its normal bundle @xmath345 .
we define the * symplectic normal bundle of @xmath346 * over @xmath48 to be : @xmath347 let @xmath348 be an isotropic embedded @xmath349-disk that satisfies the following properties : * @xmath350 * @xmath351 * any outward normal vector field to @xmath352 is nowhere tangent to @xmath38 * for any vector field @xmath353 in @xmath346 such that @xmath354 for any vector @xmath355 that is outward normal to @xmath352 , we require that the frame given by adjoining @xmath353 to the lagrangian frame of @xmath356 is homotopic to @xmath237 .
@xcite uses the above data to construct a standard model of a neighborhood of @xmath344 , then performs the surgery on this standard model , obtaining a surgered legendrian submanifold @xmath357 .
he calculates the change in @xmath42 for @xmath358 , but is unable to calculate the change for @xmath359 .
using pinching techniques , we can calculate the change in the algebra for @xmath359 , using a modified version of the standard model , which is legendrian isotopic to [ ri ] s version .
we begin with some preliminary morse lemmas , which will be used in constructing our local neighborhood model .
first , we define : @xmath360 we then have the following morse lemmas : [ lemma3.1 ] let @xmath361 be a morse function , where @xmath5 is a compact manifold . for any @xmath362 , there exists @xmath363 such that , for any @xmath364 , there is a bijection between the critical points of @xmath365 and @xmath366 , the ascending manifold of every critical point of @xmath366 lies within @xmath57 of the ascending manifold for the corresponding critical point of @xmath365 , and the descending manifold of every critical point of @xmath366 lies within @xmath57 of the descending manifold of the corresponding critical point of @xmath365 .
* proof : * see appendix [ sec : morselemmas ] .
[ lemma3.2 ] let @xmath367 be a morse function , where @xmath5 is a closed manifold .
let @xmath368 be a compact codimension-0 subset of @xmath5 that includes no critical points of @xmath35 .
then for any @xmath362 there exists @xmath363 such that for any critical point @xmath369 and any points @xmath370 , if : @xmath371@xmath372 then : @xmath373 and , for any @xmath362 , there exists @xmath363 such that , for any points @xmath370 , if @xmath371@xmath374 then : @xmath375 * proof : * see appendix [ sec : morselemmas ] .
[ lemma3.3 ] let @xmath376 be a morse function , where @xmath5 is a compact manifold .
for any choice of @xmath362 and any compact codimension-0 submanifold @xmath377 that lies in the descending manifold of maxima of @xmath365 and contains no critical points of @xmath365 , there exists @xmath378 such that , for any generic choice of @xmath364 and any @xmath379 , the ascending manifold of @xmath251 for @xmath380 will lie within an @xmath57-neighborhood of the ascending manifold of @xmath251 for @xmath381 .
* proof : * see appendix [ sec : morselemmas ] .
[ lemma3.4 ] let @xmath15 be a front - generic legendrian submanifold whose front projection is defined by sheet functions @xmath382 , where @xmath383 .
let @xmath96 be a second front - generic legendrian submanifold whose front projection is defined by sheet functions @xmath384 .
then , for any choice of @xmath362 , there exists @xmath363 such that , if : @xmath385 then , there exists a bijection between the rigid gradient flow trees @xmath38 with one positive reeb chord and the rigid gradient flow trees of @xmath357 with one positive reeb chord , such that a tree @xmath232 of @xmath38 shares the same reeb chords with its corresponding tree @xmath386 of @xmath357 , and such that the projection of @xmath386 to @xmath5 lies within an @xmath57-neighborhood of the projection of @xmath232 to @xmath5 . * proof : * see appendix [ sec : morselemmas ] .
we will use ordinary letters @xmath263 to denote reeb chords prior to the surgery , and we will use a superscript @xmath387 to denote reeb chords after the surgery .
we will use a superscript @xmath388 to denote reeb chords prior to the perturbation ( see below for an explanation of the perturbation ) . since
our surgery sphere @xmath48 lies on @xmath149 , we can assume without loss of generality that , after a legendrian isotopy , @xmath352 is arbitrarily small , because we can contract it in the front projection through any intervening sheets .
we may therefore further assume that , after a legendrian isotopy , all sheets of the front projection @xmath389 that lie over a neighborhood @xmath40 of @xmath390 are flat hypersurfaces that do not cross @xmath391 or each other .
label the sheets in @xmath391 with ( possibly negative ) integers by @xmath8 order , so that sheet 0 and 1 are the sheets containing @xmath48 .
pick a coordinate patch of @xmath5 so that @xmath43 corresponds to the @xmath14-disk centered at 0 of radius 4 .
this gives us a natural projection @xmath392 .
we will use @xmath393 to denote @xmath394 .
we perform a legendrian isotopy so that , over @xmath395 , the front projection of sheet @xmath127 is the graph of a sheet function @xmath396 which will be specified below .
we will use @xmath397 to denote the sheet functions after the surgery .
note that , for @xmath398 , @xmath399 .
we begin by defining a preliminary function @xmath400 : * @xmath400 has a maximum at @xmath401 , with the value @xmath402 .
* let @xmath37 be a 0.01-neighborhood of @xmath251 .
@xmath403 has no other critical points in @xmath37 . * over @xmath404 , @xmath403 has the value : @xmath405 where @xmath406 is the projection to @xmath407 and @xmath408 $ ] is a function that has a maximum at the north pole , a minimum at the south pole , and no other critical points . * over @xmath409 , @xmath403 has the value : @xmath410 where @xmath411 denotes the length of the straight line from @xmath251 to @xmath412 through @xmath263 . *
@xmath403 has no other critical points in @xmath395 .
define : @xmath413 then define further preliminary functions @xmath414 .
( it is intentional that @xmath415 at this stage . )
we will use * formally rigid * to denote the gradient flow trees on these height difference functions , excluding @xmath416 , that have formal dimension zero ( they are not actually rigid because @xmath417 are not generic ) . pick an arbitrarily small @xmath362 , and let @xmath418 be an @xmath57-neighborhood of the image in the base space @xmath5 of the formally rigid gradient flow trees of all of the height difference functions @xmath419 , excluding @xmath420 .
note that @xmath418 includes , in particular , the gradient flow line from @xmath421 to @xmath422 , which coincide for all @xmath423 ; this fact will be important later .
find a disk @xmath424 of radius @xmath425 , choosing @xmath425 to be arbitrarily small .
in particular , choose @xmath425 to be small enough that we can approximate the height functions @xmath426 linearly over @xmath427 .
let @xmath428 be the infinum of @xmath429 over @xmath427 .
let @xmath430 be a quantity small enough that : @xmath431 where @xmath432 is the supremum of @xmath433 .
by lemma [ lemma3.4 ] , we can find @xmath434 such that , if we perturb our height functions by an amount less then @xmath435 , the image of the rigid gradient flow trees of the perturbed height functions will be within @xmath57 of the images of the rigid gradient flow trees of the height functions @xmath417 , excepting the undefined flow trees on @xmath420 .
let @xmath378 be the minimum of @xmath436 . for @xmath437 ,
define @xmath438 , and define @xmath439 .
observe that @xmath440 are @xmath441-small perturbations in the @xmath442 metric .
define @xmath397 to be generic perturbations of @xmath443 that are @xmath441-small in the @xmath442-metric .
the height difference functions @xmath444 have three clusters of critical points : maxima around @xmath251 , which we denote by @xmath445 ; index-@xmath446 critical points around the north pole of @xmath447 , which we denote by @xmath448 ; and minima around the south pole of @xmath447 , which we denote by @xmath449 .
we will use @xmath450 to denote a critical point of type @xmath263 for the height difference function between sheets @xmath127 and @xmath211 .
a cross - section of @xmath395 is shown in figure [ fig : crosssectionofd3 ] .
[ fig : crosssectionofd3 ] ] let @xmath451 denote the corresponding critical points of the height difference functions @xmath452 ( we will use these in the following subsection ) .
observe that , by construction , the rigid gradient flow trees of @xmath453 coincide with the rigid gradient flow trees of @xmath454 .
therefore , the rigid gradient flow trees of @xmath397 lie within @xmath418 for all @xmath127 , and thus avoid @xmath427 .
let @xmath369 be the center of @xmath427 , and let @xmath455 be the natural projection .
define @xmath456 .
define the set : @xmath457 we define @xmath458 as follows : * outside of @xmath459 , @xmath460 . * within @xmath461
, @xmath462 has the form : @xmath463@xmath464 * @xmath465 are not defined inside @xmath466 . * within @xmath467 : @xmath468 where @xmath428 is the infinum of @xmath469 , over @xmath427 .
we define @xmath470 to be generic @xmath441-small perturbations of @xmath440 , and for @xmath398 we define @xmath471 .
then , prior to the legendrian ambient surgery , outside of @xmath427 we will have reeb chords @xmath472 corresponding to the chords @xmath473 . within @xmath427 , we will have an index-1 reeb chord @xmath474 between sheets 0 and 1 .
however , observe that , for @xmath475 , over @xmath427 : @xmath476@xmath477 since @xmath478 is equal to @xmath397 on the boundary , we can bound @xmath479 by @xmath480 : @xmath481@xmath482 where @xmath428 is strictly less then the supremum of @xmath469 over @xmath427 .
therefore , no other reeb chords are generated besides @xmath474 .
an `` overhead '' view of @xmath389 will have the form shown in figure [ fig : overheadofd3 ] .
[ fig : overheadofd3 ] ] we will do this in several steps .
first of all : [ lemma3.5 ] if @xmath483 are small enough , then there exists no partial flow tree @xmath232 in @xmath38 or @xmath357 with a positive special puncture over @xmath427 and a negative puncture at an @xmath187 or @xmath445 reeb chord .
* proof : * let @xmath484 be the union of the ascending manifolds of every reeb chord @xmath421 for every pair of height difference functions @xmath485 ; let @xmath486 be the union of the ascending manifolds of @xmath484 for every pair of height difference functions @xmath485 ; let @xmath487 be the union of the ascending manifolds of @xmath486 ; etc .
let @xmath432 be the total number of sheets of @xmath38 .
we claim that @xmath488 are disjoint from @xmath427 if @xmath489 are small enough .
recall that @xmath443 denotes the sheet height functions before they are genericized .
let @xmath490 be a morse neighborhood of @xmath491 for @xmath492 . recall that @xmath491 denotes the critical point of the height difference functions @xmath452 corresponding to the cluster of critical points @xmath493 , where @xmath494 are the height functions before the perturbation .
( since these functions are all equal to each other multiplied by a constant , they will have a common morse neighborhood , except that the functions will take the form @xmath495 instead of @xmath496 . )
if @xmath378 is small enough , the critical points @xmath497 will lie in @xmath490 .
define an outward - pointing radial vector field of @xmath490 : @xmath498 and observe @xmath499 on @xmath500 for all @xmath423 , so all gradient flows of the height difference functions @xmath501 are leaving @xmath490 .
therefore , if @xmath378 is small enough , then @xmath502 on @xmath500 for all @xmath423 , so all gradient flows of @xmath503 on @xmath500 are also leaving .
what this means is that , for every @xmath504 and every height difference function @xmath505 , the ascending manifold of @xmath251 for that height difference function will lie inside @xmath490 . therefore , if @xmath378 is small enough , the ascending manifold of every point @xmath506 for every choice of height difference function @xmath505 will lie in @xmath490 .
therefore , @xmath507 . since @xmath490 is disjoint from @xmath427
if @xmath483 are small enough , we conclude that @xmath488 is disjoint from @xmath427 .
why does this matter ?
well , consider a partial flow tree @xmath232 in @xmath38 or @xmath357 with a positive special puncture over @xmath427 and a negative puncture at an @xmath187 or @xmath445 chord .
to simplify notation , assume without loss of generality this tree is in @xmath38 .
let @xmath508 be an arbitrarily small open neighborhood of @xmath490 that is also disjoint from @xmath427 .
let @xmath509 be the subtree of @xmath232 that is obtained by restricting @xmath232 to @xmath508 and deleting the components of the restricted tree that do not contain our reeb chord at @xmath421 .
@xmath509 will then have a positive special puncture @xmath369 on @xmath510 , a negative puncture @xmath421 , and possibly some @xmath255 vertices and/or negative special punctures on @xmath511 .
consider the sequence of edges @xmath239 in @xmath509 between @xmath421 and @xmath369 : @xmath421 is a boundary point of @xmath512 , define @xmath513 to be the joint boundary point of @xmath512 and @xmath514 , @xmath515 to be the joint boundary point of @xmath514 and @xmath516 , and so on , up to @xmath517 is a boundary point of @xmath518 , as shown in figure [ fig : edgesinua ] .
note that , since each edge @xmath519 must lie between fewer sheets then the edge @xmath520 , we can conclude that @xmath521 .
[ fig : edgesinua ] ] then @xmath522 , so @xmath369 lies in @xmath523 .
but @xmath524 , and @xmath525 , which is a contradiction .
therefore @xmath232 can not have a negative puncture at an @xmath187 chord .
[ lemma3.6 ] if @xmath483 are small enough , then there exists no partial flow tree @xmath232 in @xmath38 or @xmath357 with a positive special puncture over @xmath427 and a negative puncture at a @xmath173 or @xmath448 reeb chord .
* proof : * let @xmath526 denote the union of the ascending manifolds of @xmath422 for @xmath527 for all @xmath423 .
let @xmath528 denote the union of the ascending manifolds of @xmath526 for @xmath529 for all @xmath530 ; let @xmath531 denote the union of the ascending manifolds of @xmath528 ; and so on , up to @xmath532 , where @xmath432 is the number of sheets of @xmath389 .
let @xmath533 be a morse neighborhood of @xmath534 , and assume @xmath378 is small enough that @xmath535 for all @xmath423 .
as in equation [ eqn : defines ] in the proof of lemma [ lemma3.1 ] , define @xmath48 to be a @xmath536-neighborhood of the ascending manifolds of @xmath534 for @xmath537 : @xmath538 consider @xmath539 as the union of subsets @xmath540 , as in equation [ eqn : definevw ] @xmath541@xmath542 and let @xmath543 be vector fields on @xmath540 : @xmath544@xmath545 then , if @xmath378 is small enough , for every @xmath546 : @xmath547@xmath548 therefore , for any point @xmath549 , and any choice @xmath423 , we know that @xmath550 , and in particular the ascending manifold must leave @xmath533 through @xmath551 .
therefore , @xmath552 all lie in @xmath48 , and @xmath553 all lie in @xmath551 .
we can separate @xmath551 into two components : @xmath554 , whose ascending manifolds lie in @xmath555 , and @xmath556 , whose ascending manifolds are disjoint from @xmath555 .
now consider the ascending manifold of @xmath554 for any choice of @xmath546 .
define @xmath557 to be a @xmath425-neighborhood of the ascending manifold of @xmath558 for @xmath537 , as shown in figure [ fig : wandq ] , and define @xmath559 to be the closure of @xmath560 .
@xmath559 then contains no critical points of @xmath561 for any @xmath530 , and lies in the descending manifold of @xmath491 for @xmath537 .
in addition , @xmath562 if @xmath563 . and @xmath554[fig : wandq ] ] by lemma [ lemma3.3 ] , if @xmath378 is small enough , the ascending manifold of every point in @xmath554 for @xmath546 will lie within @xmath564 of the ascending manifold of that point for @xmath537 . by lemma [ lemma3.2 ] ,
if @xmath536 is small enough , then the ascending manifold of every point of @xmath551 for @xmath537 will lie within @xmath564 of the ascending manifold of @xmath534 for @xmath537 .
therefore , if @xmath565 are small enough , the ascending manifold of @xmath554 for every @xmath566 will lie within @xmath567 . therefore ,
if @xmath565 are small enough , then @xmath568 .
now consider @xmath528 .
since @xmath526 consists of the union of the ascending manifolds of @xmath554 for every choice of height difference functions @xmath569 , then for any choice @xmath570 , there exists some ascending manifold in @xmath526 of that function , and every other point in @xmath526 lies within @xmath571 of that ascending manifold .
therefore , by lemma [ lemma3.2 ] , we can choose @xmath572 to be small enough that @xmath528 lies within @xmath573 of @xmath526 for any choice of @xmath574 , and therefore @xmath575
. we can then repeat this process for @xmath576 , until we obtain that @xmath577 .
what we obtain from this is that , if @xmath578 and @xmath378 are all small enough , then @xmath579 will be disjoint from @xmath427 .
therefore , for the same reasons as for the @xmath187 chords in lemma [ lemma3.5 ] , there can exist no partial flow tree with a positive special puncture over @xmath427 and a negative puncture at a @xmath173 chord .
this concludes the proof .
[ lemma3.7 ] if @xmath489 are small enough , then there exist no rigid gradient flow trees of @xmath38 whose image passes through @xmath427 and which does not have either a switch or a puncture at @xmath474 .
* proof : * let @xmath232 be a rigid flow tree in @xmath38 which passes through @xmath427 and which has no switch or puncture at @xmath474 .
let @xmath580 denote the image of @xmath232 in the base space @xmath5 .
we can break @xmath232 at some points @xmath581 into a connected partial flow tree @xmath509 and a collection of partial flow trees @xmath582 , so that @xmath509 is disjoint from @xmath427 , @xmath509 has negative special punctures at @xmath583 , and @xmath584 has a positive special puncture at @xmath585 .
let @xmath586 be the sheets of @xmath232 at the point @xmath585 .
generically , every point @xmath585 , if translated to @xmath357 , will be in the ascending manifold of the minimum @xmath449 of @xmath586 .
therefore , we can find a partial flow tree @xmath587 in @xmath357 - _ not _ in @xmath38 - consisting of a flow line on @xmath588 from @xmath585 to the minimum @xmath449 of sheets @xmath586 . since , by construction , the gradient flow @xmath588 is entering @xmath427 at @xmath585
, @xmath587 must cross @xmath427 .
observe that @xmath509 is also a valid partial flow tree in @xmath357 , because it is disjoint from @xmath427 , and @xmath589 agree except over @xmath427 .
define @xmath386 to be the flow tree in @xmath357 obtained by joining @xmath590 . by construction
, the rigid flow trees of @xmath357 do not cross @xmath427 , so @xmath386 can not be rigid .
recall from section [ subsec : flowtrees ] that the formal dimension of the moduli space containing a gradient flow tree @xmath591 is given by : @xmath592 where @xmath13 , @xmath593 are the positive punctures , @xmath594 are the negative punctures , @xmath267 are the other vertices , @xmath262 is the morse index of the reeb chord @xmath263 , and @xmath595 is the maslov content of a vertex @xmath8 .
recall further that , if @xmath263 is a positive special puncture , then @xmath264 , and if @xmath263 is a negative special puncture , then @xmath265 . finally , recall that , if we divide a flow tree @xmath591 into two partial flow trees @xmath596 , then : @xmath597 therefore , if we divide @xmath591 into @xmath598 partial flow trees @xmath599 , then : @xmath600 we now apply these formulas to our current case .
each @xmath587 consists of a positive special puncture , an edge , and a negative puncture at a minimum ; therefore : @xmath601 and by equation [ eqn : sumdimti ] : @xmath602 therefore : @xmath603 now consider @xmath604 .
since @xmath584 does not have a switch or a negative puncture at @xmath474 , and since by lemma [ lemma3.4 ] @xmath584 can not have negative punctures at our @xmath187 and @xmath173 reeb chords , @xmath584 can have no reeb chords other than minima .
besides minima , it can have @xmath255 and @xmath256 vertices .
let @xmath234 be the number of minima in @xmath584 , and let @xmath257 be the number of @xmath256 vertices . observe that @xmath605 . therefore : @xmath606@xmath607 therefore , combining equations [ eqn : sumdimti ] , [ eqn : dimgammaprime ] , and [ eqn : dimgammadoubleprime ] , we obtain : @xmath608@xmath609 therefore , @xmath232 is not rigid .
[ lemma3.8 ] if @xmath489 are small enough , then if @xmath232 is a partial flow tree in @xmath38 which has a positive special puncture followed by a switch inside @xmath427 and does not have a negative puncture at @xmath474 , then @xmath232 does not have a @xmath256 vertex . *
proof : * the edge emerging from the switch must be on the height difference function @xmath610 or @xmath611 .
suppose without loss of generality that it is @xmath610 .
since there are no cusp edges within @xmath612 except inside @xmath427 , if the partial flow tree has a @xmath613 vertex , the height difference function of the edge entering the @xmath256 vertex must be on @xmath614 , where @xmath615 .
but , for a @xmath256 vertex , we must have @xmath616 .
this is a contradiction .
[ lemma3.9 ] if @xmath489 are small enough , than for every rigid gradient flow tree @xmath232 in @xmath38 which has a switch inside @xmath427 and does not have a negative puncture at @xmath474 , there is a unique , second rigid gradient flow tree with the same positive and negative punctures .
* proof : * we can find a point @xmath617 in @xmath232 immediately above a switch , so that we can break @xmath232 at @xmath618 into partial flow trees @xmath619 , where @xmath509 has a negative special puncture at @xmath617 , @xmath620 has a positive special puncture at @xmath617 , and @xmath620 has only a single switch .
suppose the switch is from the sheets whose height difference function is @xmath621 to the sheets whose height difference function is @xmath622 ( the following argument works equivalently if it switches from or to other sheets ) .
let @xmath623 be the tangency locus of the cusp edge inside @xmath427 for the relevant sheets , and let @xmath624 be the component of the moduli space containing @xmath620 . since @xmath620 does not have a negative puncture at @xmath474 , by lemmas [ lemma3.4 ] and [ lemma3.6 ] it must consist of a single positive puncture at @xmath617 , the switch on @xmath625 , one or more minima at @xmath626 , and possibly @xmath255 vertices .
therefore , for any point @xmath627 , we can find a partial flow tree @xmath628 that has a switch at @xmath156 .
let @xmath629 denote the set of points @xmath229 such that there is a partial flow tree in @xmath624 with a special positive puncture at @xmath251 . since we can find a partial flow tree @xmath628 for any @xmath627 , the ascending manifold of @xmath625 for the vector field @xmath630 is contained in @xmath631 . because we have chosen @xmath425 small enough that we can treat @xmath632 as linear over @xmath427 , and because @xmath633 on the cusp edge , @xmath625 will consist of those points under the cusp edge in @xmath427 where @xmath634 , which will generically be diffeomorphic to @xmath635 . therefore the ascending manifold is diffeomorphic to the cylinder @xmath636 ; let @xmath637 denote the restriction of @xmath631 to this ascending manifold .
furthermore , since any two points on @xmath625 are at most @xmath638 apart , and @xmath425 is arbitrarily small , by lemma [ lemma3.2 ] @xmath637 must lie within an arbitrarily small neighborhood of the ascending manifold of @xmath639 , where @xmath640 is the switch in @xmath620 .
this ascending manifold is a 1-dimensional curve between @xmath640 and a maximum chord .
the upshot is that , since @xmath425 is arbitrarily small , @xmath637 is arbitrarily narrow " .
let @xmath641 denote the component of the moduli space containing @xmath509 , and let @xmath642 denote the set of points @xmath229 such that there is a partial flow tree in @xmath641 with a special negative puncture at @xmath251 . if @xmath232 is rigid , then the intersection @xmath643 must be zero - dimensional .
since @xmath637 is codimension-1 , @xmath644 must therefore be codimension-@xmath446 , that is , 1-dimensional . and
, if @xmath425 is small enough , @xmath637 will be narrow enough that the boundary points of @xmath645 will not lie in the solid cone whose boundary is @xmath646 , where @xmath647 is the disk whose boundary is the cusp edge inside @xmath427 .
then , @xmath648 must be even . and ,
for any point @xmath649 , we can find partial flow trees @xmath650 which have special punctures at @xmath651 , and connect them together to obtain a rigid gradient flow tree @xmath652 which has the same negative and positive punctures as @xmath232 .
this concludes the proof .
[ lemma3.10 ] if @xmath489 are small enough , than for every rigid gradient flow tree @xmath232 in @xmath38 which has a negative puncture at @xmath474 , there is a unique , second rigid gradient flow tree with the same positive and negative punctures .
* proof : * the proof works broadly similarly to the proof of lemma [ lemma3.9 ] . because @xmath474 is of morse index 1 , @xmath232 can not have a 2-valent vertex there .
since the ascending manifolds of @xmath474 of @xmath653 do not intersect any cusp edges , that means that @xmath232 must either be the flow line from @xmath654 to @xmath474 with no other vertices , or the edge of @xmath232 leading to the vertex at @xmath474 must begin with a @xmath255 vertex , @xmath617 .
if @xmath232 is the flow line , there is a second flow line approaching the other side of @xmath474 , so assume it is not .
break @xmath232 at @xmath617 into @xmath655 , where @xmath270 has a negative special puncture at @xmath617 , @xmath271 and @xmath656 have positive special punctures at @xmath617 , and @xmath656 is the flow line from @xmath617 to @xmath474 .
recall that @xmath657 denotes the component of the moduli space containing the partial flow tree @xmath658 , and that @xmath659 denotes the set of points where a partial flow tree in @xmath657 has a special puncture . since @xmath232 is rigid , we know that : @xmath660 further , since @xmath474 is index-1 , its ascending manifold on @xmath653 is a pair of 1-dimensional curves , @xmath661 , and these are submanifolds of @xmath662 ; call them @xmath663 .
these are codimension-@xmath446 , so , generically , @xmath664 must be codimension-1 .
consider @xmath665 .
this consists of a pair of curves , each of which has one boundary point at @xmath654 , and the other boundary point on @xmath666 ; label the boundary points on @xmath666 by @xmath667 .
@xmath667 must be within @xmath638 of each other . since @xmath471 outside of @xmath427 , the ascending manifolds of @xmath667 by @xmath653 equal the ascending manifolds of @xmath667 by @xmath668 , and for the function @xmath669 , @xmath667 lie in the same component of the descending manifold of @xmath670 and ascending manifold of @xmath671 .
therefore , by lemma [ lemma3.2 ] , for any @xmath672 there exists @xmath673 such that if @xmath674 , then the two components of @xmath665 lie within @xmath675 of each other .
therefore , since we can make @xmath425 arbitrarily small , we can ensure that the two curves in @xmath663 are arbitrarily close together outside of @xmath427 - and since they must necessarily be within @xmath638 of each other inside @xmath427 , we can bound the distance between the two curves everywhere . therefore , the points in @xmath676 will appear in pairs .
for every such point @xmath651 , we can find partial flow trees @xmath677 that have special punctures at @xmath651 , and then connect them together with a @xmath255 vertex to form a new rigid gradient flow tree @xmath509 with the same positive and negative punctures as @xmath232 .
[ lemma3.11 ] if @xmath489 are small enough , then for @xmath102 , the differential after the legendrian ambient surgery is given by : @xmath678 * proof : * because the only difference between @xmath38 and @xmath357 is over @xmath427 , and by construction all of the rigid flow trees of @xmath357 avoid @xmath427 , all of the rigid flow trees of @xmath357 are also rigid flow trees of @xmath38 . however , it is possible that there are now new rigid flow trees for @xmath38 that pass through @xmath427 . by lemmas [ lemma3.7 ] , [ lemma3.9 ] , and [ lemma3.10 ]
, any such new rigid flow trees appear in pairs , and so their contribution to the differential is canceled out .
this concludes the proof .
* proof of theorem [ theorem1.4 ] : * this is immediate from lemma [ lemma3.11 ] .
recall that , in theorem 1.1 , we defined @xmath48 to be a hypersurface in @xmath5 that divides @xmath5 into @xmath45 and @xmath46 , @xmath47 to be an arbitrarily small neighborhood of @xmath48 , and @xmath49 .
we can generically assume that @xmath48 intersects cusp edges transversely , and that @xmath47 does not contain any codimension-2 or higher singularities of @xmath19 .
pick a quantity @xmath363 that is less then the action of any existing reeb chord ( new reeb chords will be created by the pinching isotopy ) , and choose a tubular neighborhood @xmath679 $ ] of @xmath48 such that @xmath680 $ ] , and let @xmath681 denote the tubular coordinate . we define @xmath682 \to \mathbb{r}$ ] to be a smooth inverted bump function such that : @xmath683@xmath684@xmath685@xmath686 we will ordinarily suppress the subscripts .
a graph of this function has the form shown in figure [ fig : graphoflambda ] .
[ fig : graphoflambda ] ] then define @xmath687 be a smooth function such that : @xmath688\\ \lambda(\mu)&\mbox{for } x\in s \times [ -1 , 1 ] \end{array}\right\}\ ] ] we then define a legendrian isotopy @xmath689 in terms of the front projection @xmath389 .
let @xmath690 be the sheet functions @xmath691 of the front projection , and define : @xmath692 this defines a legendrian isotopy @xmath693 , where @xmath694 .
although this isotopy captures the primary features we need , we need to make some further small modifications in order to preserve front genericity . choose a quantity @xmath695 and perform a @xmath696-small perturbation of @xmath693 of order @xmath675 in an @xmath675-neighborhood of any cusp edges that intersect @xmath47 , to make it front generic , as defined in section [ subsec : dga ] .
this gives us a new legendrian isotopy @xmath689 .
we call this isotopy a * pinching isotopy along s*. we need to modify our approach to cope with exact lagrangian cobordisms , since these are not compact and do not have reeb chords . recall that we define @xmath334 to be the legendrian lift of an exact lagrangian cobordism @xmath33 ; that @xmath697 is a hypersurface such that each time - slice is a hypersurface in @xmath5 and @xmath698 is disjoint from @xmath149 ; and that @xmath699 is an arbitrarily small neighborhood of @xmath82 .
a legendrian isotopy of @xmath334 descends to an exact lagrangian _ homotopy _ of @xmath33 .
this is an exact lagrangian isotopy if and only if there are no reeb chords at any time in the legendrian isotopy .
this is the reason for the restriction that @xmath82 can not cross any cusp edges or self - intersections of @xmath700 in the statement of theorem [ theorem1.4 ] ; we claim that this is a sufficient ( though not necessary ) condition to allow us to find a pinching isotopy of @xmath334 that descends to an exact lagrangian isotopy . instead of @xmath701 ,
we use @xmath702 , where @xmath703 is the cylindrical coordinate .
let @xmath704 be the cylindrical coordinates such that @xmath705 are the cones over @xmath706 .
let @xmath363 be some quantity smaller then the action of the smallest reeb chord of @xmath707 .
we require that @xmath702 obey the same requirements as @xmath701 , and , in addition : @xmath708@xmath709@xmath710 we then define an isotopy on @xmath334 in a precisely analogous fashion as we did for @xmath38 . [ lemma5.20 ] for any exact lagrangian cobordism @xmath33 , given a hypersurface @xmath697 such that each time - slice is a hypersurface in @xmath5 , and there is a neighborhood @xmath699 of @xmath82 such that no cusp edges lie above @xmath699 , the isotopy induced by @xmath334 descends to an exact lagrangian isotopy .
* proof : * let @xmath711 be sheet functions of @xmath700 prior to the pinching isotopy , @xmath712 . since , by definition , @xmath699 may not cross a cusp edge , we further know that : @xmath713 during the pinching , within @xmath714 $ ] @xmath711 are replaced by @xmath715
. then , within @xmath699 , observe that : @xmath716 by construction , all of the terms of on the right hand side of the equation are positive , so there are no reeb chords within @xmath699 at any value of @xmath29 .
once we have performed the pinching isotopy , we can then prove the main theorem , outsourcing the detailed analysis to appendix [ sec : maintechnical ] : * proof of theorem [ theorem1.1 ] and [ theorem1.4 ] : * define @xmath717 , where @xmath718 is the base space coordinate and @xmath719 is the cofiber coordinate .
as discussed in appendix [ sec : maintechnical ] , after adding some additional punctures to the domain @xmath720 of @xmath721 to obtain @xmath722 , we can divide @xmath722 of @xmath721 into overlapping sets @xmath723 , where @xmath724 maps to a neighborhood away from the singularities of @xmath19 , @xmath725 maps to a neighborhood of the cusp edges , and @xmath726 maps to a neighborhood of the higher - codimension singularities of @xmath19 .
we can further find @xmath727 . then according to lemma [ lemma5.16 ]
, @xmath728 converges to a collection of discrete points as @xmath729 , and according to lemma [ lemma5.17 ] , over @xmath730 , as @xmath729 : @xmath731@xmath732 since @xmath721 is pseudoholomorphic , this in turn gives us : @xmath733@xmath734 where @xmath190 are the gradients of sheet height functions .
this tells us that @xmath735 converges to a gradient flow over @xmath730 .
similarly , we define @xmath736 to be the projection of @xmath721 restricted to @xmath725 to @xmath737 , where @xmath738 is the set of cusp edges . according to lemmas
[ lemma5.18 ] and [ lemma5.19 ] this converges to a gradient flow in the projection of @xmath739 to @xmath737 .
a secondary consequence of this is also that @xmath740 .
now consider @xmath718 as @xmath729 .
outside of an arbitrarily small neighborhood of the codimension-2 subset @xmath741 , @xmath742 converges to a collection of gradient flow lines .
since the neighborhood of @xmath743 is arbitrarily small , we can assume that it is disjoint from any points in @xmath744 which correspond to a crossing of sheets in the front projection @xmath745 .
recall from section [ subsec : dga ] that @xmath746 will have a continuous lift to @xmath15 . as a consequence ,
if @xmath722 has only a single positive puncture , @xmath742 may not converge to a flow line along a pair of sheets whose front projections cross each other , since this would imply the existence of a second positive puncture .
now suppose we have pinched @xmath38 along @xmath48 to form @xmath747 using arbitrarily small @xmath748 , and suppose we have a pseudoholomorphic disk @xmath721 such that @xmath721 has punctures over both @xmath749 and @xmath750 . by construction , @xmath751 will be disconnected ; let @xmath752 be two components of @xmath753 .
this implies that , as @xmath729 , @xmath754 converges to the embedding of a tree whose edges follow the gradient flow of height difference functions and which intersects both @xmath755 and @xmath756 .
( note that , although this is a tree that follows the gradient flows , it is technically not necessarily a _ gradient flow tree _
, in the sense of @xcite . ) we can therefore truncate the tree to find a piecewise smooth map @xmath757 \to n$ ] such that @xmath758 , and except at finitely many points @xmath759 , @xmath760 for some height functions @xmath711 , and @xmath712 .
let @xmath761 - \{v_1 , ... , v_m\ } ) \to \mathbb{r}^+$ ] be the function that maps @xmath762 $ ] to @xmath763 , where @xmath711 are the corresponding height functions at @xmath764 .
because @xmath765 , @xmath766 .
furthermore , by construction , @xmath767 for @xmath362 . therefore , @xmath768 is monotone decreasing . the argument in this section
is taken essentially from @xcite , sec .
proof of theorem [ theorem1.3 ] : * @xmath38 is a legendrian submanifold pinched as in theorem 1.1 along a neighborhood @xmath47 of a hypersurface @xmath48 dividing @xmath5 into @xmath56 , and that @xmath57 is an augmentation of @xmath42 . define @xmath53 to be the inclusion maps : @xmath769^{i_1}\\ \mathcal{a}(\lambda)|_{q_2 } \ar[u]^{i_2 } & \mathcal{a}(\lambda)|_n \ar[u]^{j_1}\ar[l]^{j_2}}\ ] ] observe that @xmath770 .
define @xmath771 .
these are obviously augmentations , since @xmath772 .
now , consider the maps @xmath773 and @xmath774 .
observe that : @xmath775 since we are working over @xmath776 .
therefore , the image of @xmath773 lies in the kernel of @xmath774 .
now observe that , if @xmath777 is in the kernel of @xmath774 , that implies that @xmath778 , meaning that @xmath263 is in the intersection of @xmath779 and @xmath780 , which is @xmath781 .
therefore the kernel of @xmath774 equals the image of @xmath773 and we have a long exact sequence . just as we have a seifert - van kampen theorem for both legendrian submanifolds and exact lagrangian cobordisms , there is an extension of the mayer - veitoris theorem to exact lagrangian cobordisms .
let @xmath33 be an exact lagrangian cobordism from @xmath64 to @xmath65 , and let @xmath314 be the cobordism map .
if @xmath65 has an augmentation @xmath57 , this induces an augmentation @xmath782 on @xmath64 .
then , after splashing , theorem 1.4 holds for the linearized chain complexes of @xmath783 as well as the differential graded algebra .
this section is derived , with some modifications and a great many omissions , from @xcite .
sections 3.1 and 3.2 are derived entirely from [ ek , sec .
section 3.3 is derived from @xcite , sec .
5.1 , but we make the reasoning more explicit . the remaining sections are derived from @xcite , sec .
5.2 and 5.3 , but with modifications to allow higher - dimensional singularities in the front projection . for simplicity s sake
, we will provide the original source for each lemma in parentheses next to the lemma .
let @xmath784 be the scaling of the cofiber and @xmath8 components by @xmath785 .
we define : @xmath786 we begin by defining some arbitrarily small modifications to @xmath787 in sections [ subsec : deformation_metric ] and [ subsec : deformation_cusprounding ] .
these modifications will produce @xmath739 , which is what we will work with after those sections .
@xmath739 allows us to obtain a modified version of the monotonicity lemma in section [ subsec : monotonicitylemma ] , which will still hold even as @xmath788 varies . in section [ subsec :
slitmodel ] , we define the slit model of the disk . in section [ subsec :
boundingderivative ] , we use the slit model and the monotonicity lemma to bound the derivatives of the map of the pseudoholomorphic disk in terms of @xmath788 . in section [ subsec : domainsubdivision ] , we add punctures to the boundary of the disk , which allows us to restrict our attention to the disk away from an arbitrarily small neighborhood of the codimension-2 singularities of @xmath19 . in section [ subsec : convergenceofdiskboundaries ] , we show that , away from this neighborhood , the boundary of the disk converges to a flow line of a height difference function as @xmath729 . in section [ subsec :
proofofmaintheorem ] , we use this result to prove theorems [ theorem1.1 ] and [ theorem1.4 ] . in section [ subsec : proofoflinearizedtheorem ] , we extend this result to linearized contact homology , proving theorem [ theorem1.3 ]
recall that section [ subsec : dga ] defines @xmath789 as the codimension-@xmath234 components of the singularities of the front projection @xmath19 . define @xmath790 to be a product neighborhood of @xmath150 of radius @xmath626 inside @xmath38 , and define @xmath791 to be a product neighborhood of @xmath150 of radius @xmath626 in @xmath12 .
let @xmath792 be a metric on @xmath5 such that the self - intersections of @xmath793 are orthogonal .
let @xmath794 be a deformation parameter .
define @xmath173 to be the restriction of @xmath792 to @xmath795 . using @xmath173 and @xmath792 we can produce a metric @xmath136 on @xmath796 which can be made arbitrarily close to @xmath792 by choosing @xmath572 small enough , and such that if @xmath797 then there exists a coordinate patch around @xmath432 disjoint from @xmath798 with coordinates @xmath799 , such that : @xmath800@xmath801 now that we have established this product metric and coordinates , we move on to cusp rounding inside @xmath796 .
let @xmath802 , and pick a coordinate patch @xmath803 around @xmath432 .
we have two kinds of sheets over @xmath432 : those on which @xmath20 is an immersion , and those with a cusp edge singularity .
we call the former an * unfolded sheet * , and the latter a * folded sheet*. from section [ subsec : deformation_metric ] , we have local coordinates @xmath804 of @xmath11 , where @xmath805 , @xmath806 is the cofiber coordinate of @xmath188 , @xmath807 is the cofiber coordinate of @xmath156 , and @xmath8 is the @xmath9 coordinate .
then an unfolded sheet can be parameterized locally as the graph of a function @xmath35 : @xmath808 we perform a small legendrian isotopy for @xmath809 , replacing @xmath35 with its taylor polynomial of degree 1 in @xmath156 : @xmath810@xmath811 this is a legendrian isotopy that introduces no new reeb chords .
next we perform a similar isotopy for folded sheets .
we assume without loss of generality that the projection of the folded sheet lies in @xmath812 .
then the sheet can be parameterized locally as the graph of a function @xmath35 : @xmath813 where : @xmath814@xmath815 we perform an isotopy over @xmath816 to replace @xmath35 with its taylor polynomial of degree 3 in @xmath156 , obtaining : @xmath817@xmath818 where @xmath819 depend only on @xmath369 , and @xmath820 .
we follow this with an additional isotopy inside an even smaller neighborhood of the cusp edge . for notational simplicity , we begin with the case @xmath54 , where the cusp edge has the form : @xmath821 where @xmath822 .
fix @xmath823 .
we define @xmath824 to be the projection of the above cusp edge to @xmath825 for @xmath826 , where @xmath263 is the base space coordinate and @xmath827 is the cofiber coordinate , and where @xmath828 .
we define @xmath829 to be the curve in @xmath825 given by a half - circle of radius @xmath830 centered at @xmath831 , connected to a pair of horizontal line segments @xmath832 from @xmath833 to @xmath834 .
explicitly , this has the formula : @xmath835 restricted to @xmath826 . in @xmath825 ]
the area bounded by the curve @xmath829 is @xmath837 .
the area bounded by the curve @xmath824 is @xmath838 .
the two curves intersect each other at @xmath839 .
therefore , we can find a hamiltonian isotopy supported in @xmath840 which deforms @xmath824 into a smoothed version of @xmath829 for @xmath841 and so that the first and second derivatives of the new curve are bounded by @xmath842 for some @xmath843 .
therefore , we can lift this isotopy to a legendrian isotopy , giving us a new curve @xmath844 parameterized by : @xmath845 where @xmath846 have the properties : @xmath847@xmath848@xmath849 we can extend this legendrian isotopy naturally to @xmath220 , giving us a legendrian isotopy restricted to a very small neighborhood of our cusp edge which changes our folded sheets from equation [ eqn : foldedsheetform ] to the form : @xmath850@xmath851 formally , we denote this deformed legendrian by @xmath852 .
we will ordinarily omit the deformation parameters , and write it simply as @xmath739 . to sum up
: our deformed legendrian submanifold @xmath739 will be equal to the cofiber - scaling of @xmath38 away from the cusp edges .
near the cusp edges , sheets that are not folded in the cusp edge will be approximated linearly , while sheets that are folded will be approximated by their degree-3 taylor polynomial .
finally , in an even smaller neighborhood inside of that neighborhood , the cusp edges will be replaced by semicircles of radius @xmath853 in the lagrangian projection .
the purpose of these deformations is to ensure that we have a useable version of the monotonicity lemma .
this section is based on the treatment in @xcite , sec .
5.1 , though made more explicit , and in @xcite , ch .
5 , sec . 4
, though @xcite considers only fixed lagrangians .
let @xmath854 , and let @xmath855 be an @xmath178-ball around @xmath251 .
the standard monotonicity lemma says that if we have a @xmath135-holomorphic map @xmath856 , then there exists a constant @xmath857 such that : @xmath858 the problem is that @xmath859 depends on @xmath744 , which in turn depends on @xmath788 .
even with our deformations , we can not retrieve the full version of the monotonicity lemma .
what we _ can _ do , however , is get a version that is good enough . recall that by the definition of a tame almost complex symplectic manifold , there exists constants @xmath860 and @xmath861 such that if @xmath225 is a loop in @xmath12 contained in a ball @xmath862 , then @xmath225 bounds a disc in @xmath863 of area less then @xmath864 .
then : [ lemma5.2 ] there exists a constant @xmath865 such that , for any riemannian surface with boundary @xmath612 and any pseudoholomorphic map @xmath866 , if @xmath867 then : @xmath868 in addition , there exists a constant @xmath869 such that , if @xmath870 , and @xmath871 , then : @xmath872 * proof : * see @xcite , ch .
5 , prop . 4.3.1.i and ii . in particular , the second statement is the general form of the monotonicity lemma for symplectic manifolds .
note that @xmath873 do not depend on @xmath788 , or even on @xmath38 .
they are properties of the symplectic manifold @xmath12 and its associated symplectic form , almost complex structure , and riemannian metric .
[ lemma5.4 ] if @xmath574 is small enough , then there exists a constant @xmath874 independent of @xmath788 such that , if @xmath612 is a disk in @xmath125 and @xmath875 is a pseudoholomorphic map , @xmath876 , where @xmath860 is the @xmath860 defined in remark 5.1 , then : @xmath877 * proof : * in what follows , we will refer to the cusp edges " and sheets " of the legendrian , even though we are working with the lagrangian projection @xmath168 .
these both refer to the projection of the corresponding objects in the front projection to the lagrangian projection
. choose @xmath878 small enough that no solid ball in @xmath12 of radius @xmath378 intersects more then one sheet of @xmath879 unless it contains a cusp edge or a double point , in which case it intersects at most two sheets .
in addition , choose @xmath378 small enough that we can always find a coordinate chart of @xmath12 containing any solid ball of radius @xmath378 .
the subscript @xmath323 is used to indicate @xmath880 .
if @xmath574 is small enough , and @xmath881 , then any ball @xmath882 , will intersect @xmath744 in at most one sheet unless it contains a cusp edge or a double point , in which case it will intersect at most two sheets , thanks to the cusp rounding conducted in section [ subsec : deformation_cusprounding ] .
we can regard @xmath744 as the union of the graphs of a collection of sheet functions : @xmath883 where @xmath884 . because @xmath739 is compact , outside of the @xmath830-neighborhood of the cusp edges
, we can bound @xmath885 for some constant @xmath859 .
in addition , inside the @xmath830-neighborhood of @xmath886 , but outside of a @xmath830-neighborhood of @xmath743 , we can instead describe @xmath744 as the graph of a function @xmath887 , and bound @xmath888 .
let @xmath889 .
we begin with the case where @xmath855 does not intersect a double point or the @xmath830-neighborhood of a cusp edge .
let @xmath890 , where @xmath891 , be the sheet function . pick a coordinate chart in @xmath5 around @xmath892 , with @xmath893 .
then we can define a path @xmath222 \to \pi_\mathbb{c}(\lambda_\sigma)$ ] that connects @xmath894 to @xmath895 by : @xmath896@xmath897@xmath898 therefore , @xmath899 .
a similar argument allows us to bound the length of a path by @xmath900 if the ball intersects a double point .
if the ball intersects the @xmath830-neighborhood of the cusp edge , then pick a coordinate chart in @xmath12 corresponding to the product neighborhood of the cusp edge , with @xmath901 , and let @xmath136 be the cusp edge function .
we define @xmath902 to be the projection from @xmath903 .
we can define a path @xmath222 \to \pi_\mathbb{c}(\lambda_\sigma)$ ] that connects @xmath895 to @xmath894 by : @xmath904@xmath905@xmath906@xmath907 therefore , @xmath908 .
@xmath909 is independent of @xmath788 because , as a result of the cusp rounding and the fact we are working in an arbitrarily small neighborhood of the cusp edge , @xmath136 is independent of @xmath788
. therefore , we can find @xmath910 such that , for any ball of radius @xmath911 , any any two points @xmath912 , there exists a path in @xmath744 linking @xmath894 to @xmath895 of length less then or equal to @xmath913 .
now let @xmath914 be an arbitrary ball intersecting @xmath739 , and let @xmath184 be a pseudoholomorphic map of an open disk @xmath915 .
then @xmath916 is a union of arcs @xmath917 in @xmath918 from @xmath744 to @xmath744 , which have length less then or equal to @xmath919 .
let @xmath920 be arcs in @xmath921 closing @xmath917 .
then they have length less then or equal to @xmath913 .
therefore , by lemma [ lemma5.2 ] , @xmath922 bound a union of disks @xmath551 of area less then or equal to @xmath923 , where @xmath865 is a constant that depends only on @xmath12 .
therefore , @xmath924 is a surface with boundary on @xmath925 .
@xmath926 must be a disk or pair of disks by our restrictions on @xmath178 , so it is contractible .
by the fact that @xmath927 is contractible and the fact that @xmath928 , we obtain that the area of @xmath929 is less then or equal to @xmath930 , where @xmath874 is another constant that depends only on @xmath12 .
[ lemma5.5]if @xmath574 is small enough , then there exists constants @xmath931 which are independent of @xmath788 such that , if @xmath932 , then for any non - constant @xmath135-holomorphic map @xmath856 , such that @xmath251 is in the image of @xmath184 , and where @xmath612 is an open surface , then : @xmath933 * proof : * define @xmath934 for @xmath935 , @xmath936 . since @xmath184 is smooth , sard s theorem ( @xcite , theorem ii.3.1 ) implies that @xmath937 is a subsurface with @xmath442 boundary @xmath938 for almost all @xmath29
. define @xmath939 .
@xmath940 is an absolutely continuous function , and @xmath941 for almost all @xmath29 . then by lemma [ lemma5.4 ]
, we have @xmath942 for almost all @xmath935 . since @xmath184 is not constant and the image of @xmath184 contains @xmath251 , @xmath943 for @xmath944 .
thus , for @xmath944 we know : @xmath945 therefore , integrating the above from @xmath946 to @xmath178 , we obtain : @xmath947 which implies equation [ eqn : scaledmonotonicitylemma ] .
recall that @xmath948 is the unit disk in @xmath125 with @xmath432-punctured boundary .
we define @xmath720 to be the subset of @xmath949 $ ] given by deleting @xmath950 horizontal slits of width @xmath951 , where each slit ends in a half - circle . @xmath720
has a canonical complex and symplectic structure inherited from @xmath952 .
( @xcite , lemma 2.2)[lemma5.6 ] @xmath720 is biholomorphic to @xmath948 .
* proof : * see @xcite , lemma 2.2 .
we will work in the slit model of the disk for most of what follows .
recall from the introduction to this section that @xmath953 is the scaling of the cofiber and @xmath8 components by @xmath785 .
let @xmath954 .
then , there is a bijection between @xmath135-holomorphic disks on @xmath168 and @xmath955-holomorphic disks on @xmath744 .
let @xmath956 be a pseudoholomorphic map .
let @xmath721 be the corresponding map @xmath957 , which can be parameterized by @xmath958 , where @xmath719 is the map to the base space and @xmath718 is the map to the cofiber .
recall from section [ subsec : dga ] that @xmath789 denotes the codimension-@xmath234 component of the singularities of the front projection @xmath19 , that @xmath959 is the product neighborhood of radius @xmath626 around @xmath150 in @xmath38 , and that @xmath960 is the product neighborhood of radius @xmath626 around @xmath150 in @xmath12 .
we analogously have @xmath961 ; however , in what follows we will omit the subscript .
there is a projection map @xmath962 .
define @xmath963 , with @xmath964 restricted to the pre - images of @xmath965 .
define @xmath966 , where @xmath967 is an extremely small deformation parameter . throughout all that follows
, we will restrict @xmath721 to vertical lines contained in the pre - image of @xmath968 , where @xmath969 is arbitrarily small .
we denote the restricted domain of vertical lines by @xmath970 .
next , we put bounds on @xmath971 . to do this , we will use @xmath972-notation , where @xmath973 means that : @xmath974 ( @xcite , lemma 5.2 and 5.4)[lemma5.7 ] for any @xmath135-holomorphic map @xmath975 , there exist constants @xmath976 independent of @xmath788 such that the symplectic area of @xmath977 is less then @xmath978 and the length of @xmath979 is less then @xmath909 for any @xmath788 .
* proof : * by stokes theorem : @xmath980 since @xmath981 is the tautological 1-form , and @xmath982 is scaled by @xmath788 , this integral equals @xmath978 , giving the first result .
we obtain the second result as follows : let @xmath983 denote the length of @xmath984 .
for @xmath985 small enough we can find @xmath432 disjoint solid balls of radius @xmath986 whose centers lie on @xmath987 , where @xmath988 rounded down .
lemma [ lemma5.5 ] then shows that : @xmath989 since @xmath990 , this shows that @xmath983 must have some maximum .
( @xcite , theorem 9.4)[lemma5.8 ] fix @xmath991 and @xmath992 .
let @xmath98 be the open unit disk in @xmath125 or the half disk with boundary on @xmath9 .
for any compact @xmath993 , there exsts a constant @xmath910 such that for all holomorphic maps @xmath994 , we have : @xmath995 to be the open @xmath178-disk centered at the origin in @xmath125 , and @xmath996 .
there exists @xmath997 such that , if @xmath998 , and @xmath999 , then the distance between @xmath1000 is greater then @xmath626 when @xmath880 .
let @xmath1001 be an arbitrary open set with diameter less then @xmath626 .
choose @xmath363 small enough that @xmath1002 has a tubular neighborhood of radius @xmath378 .
then let @xmath1003 denote the restriction to @xmath1002 of a @xmath378-tubular neighborhood in the product neighborhood .
define @xmath1004 to be the connected segment of @xmath1005 containing @xmath48 .
recall from section [ subsec : boundingderivative ] that @xmath1006 is the product neighborhood of radius @xmath378 around @xmath150 , and that @xmath1007 .
( @xcite , lemma 5.6)[lemma5.9 ] there exists @xmath1008 such that if @xmath1009 is pseudo - holomorphic then : @xmath1010 in addition , for all @xmath843 large enough there exists @xmath1008 such that if @xmath1011 is pseudoholomorphic , and if @xmath1012 lies outside a @xmath842-neighborhood of @xmath1013 , then : @xmath1014 finally , for all @xmath997 small enough , if @xmath1015 is a pseudoholomorphic map such that @xmath1016 then : @xmath1017 * proof : * we begin with the first case . if @xmath1018 the inequality is trivially true , so we assume it does not . recall from @xcite , theorem 27.18 , that the sobolev inequality for @xmath1019 is : @xmath1020 where @xmath14 is the dimension of the space , @xmath43 is a bounded open subset of @xmath1021 with @xmath442 boundary , @xmath1022 is a holder space , and @xmath225 is defined by : @xmath1023 where @xmath1024 denotes the floor of @xmath263 . in our case @xmath1025
, @xmath1026 is a subset of @xmath1027 , so @xmath1028 , and we choose @xmath1029 .
therefore , for @xmath1030 : @xmath1031 since @xmath1032 for any @xmath1033 , this gives us : @xmath1034 then , we apply lemma [ lemma5.8 ] to @xmath1035 using @xmath1036 . from this
we obtain : @xmath1037@xmath1038 , [ eqn : wofg1 ] , and [ eqn : supofgr ] we have : @xmath1039 where @xmath1040 is defined by : @xmath1041 the limit of @xmath1042 as @xmath1043 is given by : @xmath1044 combining equation [ eqn : c3(r ) ] , equation [ eqn : limc2(r ) ] , and the fact that the supremum of @xmath1045 will be strictly declining as @xmath178 shrinks , tells us that the limit of @xmath1040 as @xmath1043 must be defined and finite . therefore the same inequality holds with @xmath1046 , giving us : @xmath1047 the process is precisely analogous for the other cases .
recall from section [ subsec : boundingderivative ] that @xmath970 is the closed set in @xmath720 made of vertical lines @xmath257 such that @xmath1048 lies outside an arbitrarily small neighborhood of @xmath798 .
let @xmath1049 denote the points in @xmath970 which are connected to @xmath251 by a path in @xmath970 of length at most @xmath178 .
( @xcite , lemma 5.7)[lemma5.10 ] if @xmath574 and @xmath788 are small enough , if @xmath1050 are small enough , and if @xmath1051 is big enough , then there exists @xmath1052 such that : * let @xmath1053 , and at least @xmath1054 distance from @xmath1055 .
then : @xmath1056 * if @xmath1057 and @xmath1058 is outside of a @xmath1059-neighborhood of @xmath1013 , then : @xmath1060 * if @xmath1057 and @xmath1061 and @xmath1062 then : @xmath1063 * proof : * we begin with the second case . there is a biholomorphic map with uniformly bounded derivatives from @xmath1064 to @xmath1065
. then , by lemma [ lemma5.9 ] , there is a constant @xmath859 such that for all @xmath1066 , and : @xmath1067 therefore @xmath1068 is the square root of the symplectic area of @xmath1069 . since @xmath1070 , and the symplectic area of @xmath977 is less than the action of the positive reeb chords , and the action of the positive reeb chords is scaled by @xmath788 , we obtain @xmath1071 for all @xmath1072 .
we define a norm on the space of linear operators . for a given point @xmath251 and a linear operator @xmath1073 ,
we define : @xmath1074 and for an operator @xmath1075 , we define : @xmath1076 pick a complex coordinate chart around @xmath1077 that agrees with the ambient almost complex structure at @xmath1077 .
we denote the complex structure from the coordinate chart by @xmath1078 .
then , for any @xmath1079 , @xmath1080 . let the sheet of @xmath744 in @xmath12 containing @xmath1077 be the graph @xmath1081 of a 1-form @xmath1082 .
we claim that for all @xmath788 small enough , we can find a diffeomorphism @xmath1083 such that : * @xmath1084 equals the 0-section , and is therefore real analytic * @xmath1085 along @xmath1081 * let @xmath1086 be the sup norm metric on @xmath442 .
@xmath1087 for arbitrarily small @xmath428 .
we prove the existence of @xmath1083 by calculating its inverse .
let @xmath1088 be the coordinates of the local complex chart such that the 0-section corresponds to @xmath1089 .
define @xmath1090 .
then , for very small @xmath827 , define : @xmath1091 and extend the map arbitrarily to the rest of the chart .
then @xmath1092 , so it sends the zero section to the graph @xmath1081 , and it is @xmath442-close to the identity . and , along the zero - section : @xmath1093@xmath1094 so @xmath1095 .
this gives us @xmath1083 .
define @xmath1096 .
then @xmath1097 , and @xmath1098 is @xmath1099-holomorphic .
define @xmath1100 and @xmath1101 .
then @xmath1102 and @xmath1103 is @xmath1104-holomorphic .
furthermore , @xmath1103 is @xmath1104-holomorphic on @xmath1105 .
we extend @xmath1103 by copying it over the 0-section and the boundary , obtaining a map @xmath1106 , where @xmath612 is a disc , @xmath1107 is @xmath442 , and @xmath1108 , where @xmath1104 is extended over the new area by @xmath1109 , where the bars denote the complex conjugate .
now define a map @xmath1110 by : @xmath1111 and define @xmath1112 .
let @xmath1113 be the imaginary components of @xmath1114 .
since @xmath1103 is pseudoholomorphic , the function @xmath1115 is subharmonic , and is therefore bounded .
since @xmath1107 is the doubling of @xmath1103 , the same is true of @xmath1116 .
therefore , @xmath1117 is uniformly bounded , and the derivatives of @xmath237 are uniformly bounded in a neighborhood of the image of @xmath1107 . furthermore : @xmath1118 where @xmath1119 . since @xmath1120 , this implies : @xmath1121@xmath1122 define @xmath1123 .
this then becomes : @xmath1124 since @xmath237 is @xmath1078-holomorphic : @xmath1125 we obtain : @xmath1126 now let @xmath1127 be a cutoff function which equals 1 on @xmath1128 and equals 0 outside @xmath1129 , and define @xmath1130 .
then , define : @xmath1131 since @xmath981 is compactly supported , and therefore its derivatives are bounded , and since @xmath1132 , we conclude @xmath1133 . therefore : latexmath:[\[\label{eqn : barpartialf1 } provided @xmath428 is small enough , we can rearrange equation [ eqn : barpartialf1 ] to obtain : @xmath1135 then add @xmath1136 to both sides to obtain : @xmath1137 since @xmath26 is exact , by stokes theorem : @xmath1138 therefore , if we integrate both sides of equation [ eqn : finequality ] , we obtain : @xmath1139 which is less than or equal to some constant @xmath1140 since @xmath1141 is bounded . note that
, since the derivatives of @xmath237 are uniformly bounded on a neighborhood of the image of @xmath1107 , we have : @xmath1142 therefore @xmath1143 is bounded by some constant , and therefore , by scaling , @xmath1144 .
combining this with lemma [ lemma5.10 ] , we obtain the result .
the first and third part of the lemma follow analogously .
let @xmath1145 .
pick @xmath1146 .
let @xmath1147 denote the points in @xmath1148 such that @xmath1149 and @xmath1150 crosses @xmath1151 at @xmath1147 , ordered by position on the boundary .
puncture the disk @xmath720 at every @xmath1152 such that there is @xmath1153 so that @xmath1154 for @xmath1155 or @xmath1156 .
( this is a correction of the version of boundary puncturing that appears in @xcite ) .
this gives us a new disk with more punctures but the same map that we will denote @xmath722 .
we define the * boundary minimum * of a component @xmath1157 of @xmath1158 to be the point @xmath1159 with minimum @xmath263 value . in addition , recall from section [ subsec : boundingderivative ] that @xmath970 denotes the restriction of @xmath722 to vertical lines contained in the pre - image of @xmath1160 , where @xmath969 is arbitrarily small .
( @xcite , lemma 5.9)[lemma5.11 ] there exists a constant @xmath1161 that does not depend on @xmath788 such that the number of added punctures is less than or equal to @xmath910 for fixed @xmath378 and any @xmath788 .
* proof : * each added puncture of @xmath722 corresponds to @xmath1162 such that there exists @xmath1153 so that @xmath1154 on @xmath1148 for @xmath1163 .
the length of the segment between @xmath1164 is bounded from below by @xmath378 , and by lemma [ lemma5.7 ] the length of @xmath1165 is bounded above for any @xmath788 , so the result follows .
we now label the boundary components of @xmath722 by the following types : * out : * @xmath1166 * 0 : * @xmath1167 * in : * @xmath1168 then , for @xmath1157 a boundary component and @xmath1169 , define @xmath1170 to be a @xmath425-neighborhood of @xmath1157 in @xmath720 , and : @xmath1171 fix a small @xmath362 .
( @xcite , lemma 5.10)[lemma5.12 ] there exists @xmath910 such that if @xmath785 is small enough , then : @xmath1172@xmath1173 * proof : * this is a direct consequence of lemma [ lemma5.10 ] .
let @xmath1174 be the subset containing all vertical line segments in @xmath970 which connect a point in a boundary component of type * in * to some other boundary point .
note @xmath1175 is a collection of vertical line segments .
( @xcite , lemma 5.11)[lemma5.13 ] for any @xmath1176 and sufficiently small @xmath785 , the distance from any @xmath1177 , where @xmath1157 is of type * out * , to @xmath1178 is larger than @xmath1179 . in particular ,
if @xmath257 is a vertical line segment in @xmath1175 and @xmath1180 then @xmath369 is either a boundary minimum on a segment of type * in * or it lies on a boundary segment of type * 0*. * proof : * suppose @xmath1177 , @xmath1157 of type * out * , and the distance from @xmath251 to @xmath1178 is less than @xmath1179 .
then there exists a path in @xmath1181 of length less than @xmath1182 , where @xmath178 is the number of punctures of @xmath722 , from @xmath251 to a point @xmath369 midway between two horizontal boundary segments of length 1 , at least one of which is of type * in*. call the segment of type * in * @xmath1183 and the other @xmath1184 . since @xmath1185 along this path and @xmath1186 , we know that : @xmath1187 therefore , for @xmath788 small enough , @xmath1188 . let @xmath1189 be a horizontal segment of length 1 that intersects @xmath369 .
for @xmath788 small enough , @xmath1190 . by definition ,
call the region between @xmath1183 and @xmath1192 @xmath1193 , and observe that @xmath1194 \times [ 0 , \frac{1}{2}]$ ] .
observe further that the image under @xmath721 of every vertical path in @xmath1193 has length at least @xmath1195 .
therefore , if @xmath29 is the vertical coordinate of @xmath722 : @xmath1196 from this , we conclude that the @xmath1197-norm of @xmath1045 is bounded below .
this contradicts lemma [ lemma5.7 ] , which states that the symplectic area of @xmath1198 is bounded above by @xmath1199 for some constant @xmath910 .
let @xmath1200 be the region of points of distance @xmath257 or less from @xmath1178 .
choose @xmath1201 so that @xmath1202 and @xmath1203 are vertical line segments disjoint from boundary minima .
define @xmath1204 .
note that if @xmath1205 then @xmath251 is in a boundary component of type * 0 * or * out * , and if @xmath1206 then @xmath251 is in a boundary component of type * 0 * or *
in*. lemma [ lemma5.12 ] implies that : @xmath1207 ( @xcite , lemma 5.12)[lemma5.14 ] for small enough @xmath788 , @xmath1208 and @xmath1209 . *
proof : * let @xmath1210 . then @xmath369 is linked by a path @xmath225 of length less than 5@xmath178 , where @xmath178 is the number of punctures of @xmath722 , to a point @xmath1211 .
@xmath251 must lie in a component of type * 0 * or * out * , so @xmath1212 .
since @xmath1213 along the path , for @xmath788 small enough the second statement follows .
@xmath1214 consists of boundary segments of types * 0 * and * in * and vertical lines ending on boundary components of type * 0*. we know from the definition that the image of the boundary segments will lie in @xmath1215 , while the bound on @xmath971 allows us to ensure that @xmath1216 for @xmath788 small enough .
then , if @xmath1217 does not lie inside @xmath1218 , we may bound its area from below for any @xmath788 by lemma [ lemma5.5 ] , our modified monotonicity lemma
. we can then repeat this process : we add additional punctures at the intersections of @xmath1219 with @xmath1220 .
we label the boundary components of @xmath1221 with * out , 0 , in*. we bound @xmath736 in a neighborhorhood @xmath1222 , and use equivalents of lemmas 5.13 and 5.14 for @xmath736 to split @xmath1223 into @xmath725 , which stays away from @xmath798 , and @xmath1224 , which does not .
we only do this once , unlike in @xcite .
we have thus divided @xmath1225 , so that @xmath724 maps to a neighborhood away from the singularities , @xmath725 maps to a neighborhood of the cusp edges , and @xmath726 maps to a neighborhood of @xmath798 .
let @xmath1226 be a neighborhood of the boundary minima of @xmath1227 such that : * @xmath1228 is a union of arcs in @xmath1229 and of vertical line segments ; * each component of @xmath1226 contains at least one boundary minimum ; * and the width of each component of @xmath1226 is at most @xmath1230 .
now consider a sequence : @xmath1231 [ lemma5.16]@xmath728 converges to a point as @xmath729 .
* proof : * by lemma [ lemma5.12 ] , @xmath1185 in @xmath1232 , so the area of @xmath728 is @xmath1233 .
let @xmath1234 be the value of @xmath788 for which the number of components of @xmath730 reaches its maximum .
( since the total number of added punctures is bounded by lemma [ lemma5.11 ] , we must be able to select @xmath1232 in such a way that a maximum is eventually achieved . )
restrict @xmath788 to @xmath1235 , and consider a vertical line segment @xmath1236 .
let @xmath24 be the component of @xmath730 containing @xmath257 .
if @xmath24 is an infinite or half - infinite rectangle , it can be parameterized by @xmath1237 \times [ 0,1]$ ] or @xmath1238 $ ] or @xmath1239 $ ] . in each case , @xmath257 is specified by a choice of @xmath263 , so our choice of @xmath257 is well - defined for any @xmath1235 .
if @xmath24 is a finite rectangle , it can be parameterized by @xmath1240 \times [ 0 , 1]$ ] , where @xmath1241 depends on @xmath788 .
pick some value of @xmath788 , which we call @xmath1242 . then , for @xmath1243 , @xmath257 is specified by a choice of @xmath1244 $ ] .
then we can define a function @xmath1245 .
this lets us consider our choice of @xmath257 to be well - defined for any value of @xmath1235 .
we can similarly show that a choice of vertical line segment @xmath1246 is well - defined for any @xmath1235 .
recall from section [ subsec : boundingderivative ] that we can parameterize @xmath721 by @xmath1247 , where @xmath718 is the point in the base space and @xmath719 is the cofiber coordinate .
. by the definition of @xmath724 , the image of @xmath257 is outside of a neighborhood of @xmath149 .
we can therefore find a neighborhood of the image of @xmath257 in which @xmath120 can be parameterized as the graph of some collection of functions @xmath1248 .
we refer to these graphs as sheets , by analogy to the sheets of @xmath389 .
let @xmath1249 be functions of the sheets containing the image of @xmath1250 .
( @xcite , lemma 5.13)[lemma5.17 ] for all sufficiently small @xmath785 , along any vertical line segment @xmath1236 : @xmath1251@xmath1252 where @xmath1253 denotes the connection , the subscripts @xmath1254 are used to indicate @xmath1255 , and @xmath1254 are the vertical , horizontal coordinates respectively of @xmath730 .
* proof : * let @xmath1256 \times [ 0 , 1 ] \subset d_0(\sigma ) - w_0(\sigma)$ ] be an arbitrarily small neighborhood around @xmath257 , with @xmath1257 . by lemma [ lemma5.10 ] , @xmath1213 on @xmath1258 .
therefore , we can pick a radius @xmath1193 such that @xmath1259 , where @xmath1260 is the geodesic disk in @xmath5 of radius @xmath1261 centered around @xmath1262 .
we think of @xmath1263 as a subset of @xmath1264 , @xmath1265 .
let @xmath1266 be the flat metric on @xmath1267 , and let @xmath1268 be the corresponding standard complex structure on @xmath1263 , with the coordinates chosen such that @xmath1269 and @xmath1270 . recall that @xmath718 is the base space component and @xmath719 is the cofiber component of @xmath721 ; there exists some @xmath37 such that @xmath1271 for all @xmath788 .
define @xmath1272 to be @xmath1273 .
define an almost complex structure @xmath1274 , and define @xmath1275 .
@xmath1103 is @xmath1276-holomorphic .
we claim that @xmath1277 .
we will prove the statement in dimension 2 ; the proof can be extended to higher dimensions by adding appropriate summation signs .
define @xmath223 by : @xmath1278 we know that @xmath1279 because @xmath1280 .
we will ordinarily omit the @xmath788 coordinate of @xmath223 . since @xmath1272 is compact , @xmath1281 all obtain some maximum ; let @xmath1282 be some number greater then all of them . using the taylor expansion ,
we obtain : @xmath1283 therefore @xmath1284 . from this , we obtain : @xmath1285@xmath1286@xmath1287 therefore @xmath1288 . from this
we further obtain : @xmath1289@xmath1290@xmath1291@xmath1292@xmath1293 therefore @xmath1294 . combining these ,
we obtain : @xmath1295 we can then repeat this calculation for @xmath1296 . combining the two ,
we obtain @xmath1297 .
this calculation can be extended to higher dimension in a straight - forward manner by adding appropriate summation signs .
recall that we defined the functions @xmath1298 to be the two sheets of @xmath744 over @xmath1267 corresponding to the restriction of @xmath1299 to @xmath1300 . after scaling @xmath1264 by @xmath1301
, we replace @xmath1302 with @xmath1303 .
define @xmath1304 ; note that these are lagrangian subspaces .
observe that : @xmath1305 in the next step , we will assume again that we are working in two dimensions without loss of generality .
we can rewrite the righthand side of equation [ eqn : pminusb1 ] using its taylor expansion around @xmath1306 to obtain : @xmath1307 therefore , for any @xmath703 , @xmath1308 , as does its derivative and double derivative by @xmath703 .
we can repeat this calculation for @xmath1309 .
therefore there exists a function @xmath1310 such that : @xmath1311@xmath1312@xmath1313@xmath1314 and @xmath1315 is @xmath1268-holomorphic to first order on @xmath1316 , that is , @xmath1317 .
define @xmath1318 , and define @xmath1319 to be the linear solution to : @xmath1320@xmath1321@xmath1322@xmath1323 ( there is a typo in the definition of @xmath1324 in @xcite , and we believe this is what he means . ) define @xmath1325 .
then @xmath1326 , @xmath1327 is @xmath1268-holomorphic to first order on @xmath1316 , @xmath1328 , and : @xmath1329 let @xmath1330 , \mathbb{c}^n)$ ] denote the hilbert space with the weight function @xmath1331 for @xmath1332 and @xmath1333 for @xmath1334 .
let @xmath1335 .
define @xmath1336 , \mathbb{c}^n ; \mathbb{r}^n , 0_2)$ ] to be the space of functions @xmath237 with boundary on @xmath1021 and with three derivatives in the weighted sobolev space @xmath1197 , and such that the restriction to the boundary of @xmath1337 and its first derivatives vanishes .
we similarly define @xmath1338 to be the space of functions with 2 derivatives in the weighted sobolev space @xmath1197 and which vanish to first order along the boundary . as shown in @xcite , prop .
6.3 , @xmath1339 , \mathbb{c}^n ; \mathbb{r}^n , 0_2 ) \to \mathcal{h}_{2,-\gamma}(\mathbb{c}^n;0_1)$ ] is a fredholm operator of index @xmath14 with kernel spanned by the constant functions
. we will generally write @xmath1340 .
let @xmath1341,\mathbb{c}^n;\mathbb{r}^n,0_2)$ ] be the subspace of non - constant functions @xmath237 with @xmath1342 . then there exists a constant @xmath910 such that : @xmath1343 now let @xmath1344 be a cutoff function such that : * @xmath142 is real - valued and holomorphic to first order on @xmath1345 * @xmath1346 on @xmath1347\times[0,1]$ ] * @xmath1348 outside @xmath1349\times[0,1]$ ] then : @xmath1350 } \left(|\bar{\partial}v_\sigma|^2 + |d\bar{\partial}v_\sigma|^2 + |d^2\bar{\partial}v_\sigma|^2\right)\ ] ] we split this into cases . first , observe that : @xmath1351\times[-\frac{1}{4}c,\frac{1}{4}c ] } \right|\right|^2_{h_{2,2,-\gamma}}\]]@xmath1352 given that @xmath1353 for @xmath1354 , the first term is equal to : @xmath1355\times[-\frac{1}{4}c,\frac{1}{4}c ] } \right|\right|^2_{h_{2,2,-\gamma } } = \left|\left|\left .
\bar{\partial}v_\sigma\right|_{[0,1]\times[-\frac{1}{4}c,\frac{1}{4}c ] } \right|\right|^2_{h_{2,2,-\gamma}}\ ] ] since @xmath1356 , this tells us that the first term of @xmath1357 equals @xmath1358 .
the second term , we calculate similarly ; the bound on @xmath1359 tells us that it also equals @xmath1358 , and then no other term depends on @xmath29 , giving us : @xmath1360 finally , the third term we calculate as follows : since @xmath1361 are bounded , @xmath1362 are bounded , and thus @xmath1363 . therefore : @xmath1364 this equals @xmath1365 . since @xmath1366 , this equals @xmath1367 .
we can then repeat this process for @xmath1368 to show that : @xmath1369 since this controls the supremum norm over @xmath1347 \times [ 0,1]$ ] , we obtain : @xmath1370\times[-\frac{1}{4}c_\sigma,\frac{1}{4}c_\sigma ] } |d^kv_\sigma| = \mathcal{o}(\sigma ) , k = 0,1\ ] ] if we write @xmath1371 , we obtain : @xmath1372@xmath1373 where @xmath232 denotes the linear operator in @xmath1374 . recall that @xmath1375 .
further , @xmath1376 , where @xmath1377 for @xmath1378 , and @xmath1379 .
therefore : @xmath1380@xmath1381 [ lemma5.18]@xmath1382 converges to a point as @xmath729 . [ lemma5.19]for all sufficiently small @xmath785 , then along any vertical line segment @xmath1246 :
@xmath1383@xmath1384 * proof of lemmas [ lemma5.18 ] and [ lemma5.19 ] : * the proofs are precisely analogous to the proofs of lemmas [ lemma5.16 ] and [ lemma5.17 ] , but with @xmath721 replaced with @xmath736 ; the base space replaced with @xmath1385 ; @xmath190 replaced by their restriction to @xmath1386 ; @xmath739 replaced with its projection to @xmath1387 ; and @xmath955 replaced by its restriction to @xmath1387 .
in this section we prove [ lemma3.1 ] , [ lemma3.2 ] , [ lemma3.3 ] , and [ lemma3.4 ] . first , recall that we define : @xmath1388 [ tubularflowtheorem ] let @xmath418 be a vector field on an @xmath14-dimensional manifold @xmath5 . then , at any point @xmath229 such that @xmath1389 , there exists a neighborhood @xmath43 of @xmath251 and a diffeomorphism @xmath1390 such that the trajectories of @xmath418 in @xmath43 are mapped to the trajectories of @xmath1391 in @xmath1392 .
* lemma [ lemma3.1 ] : * _ let @xmath361 be a morse function , where @xmath5 is a compact manifold . for any @xmath362 , there exists @xmath363 such that , for any @xmath364 , there is a bijection between the critical points of @xmath365 and @xmath366 , the ascending manifold of every critical point of @xmath366 lies within @xmath57 of the ascending manifold for the corresponding critical point of @xmath365 , and the descending manifold of every critical point of @xmath366 lies within @xmath57 of the descending manifold of the corresponding critical point of @xmath365 . _ * proof : * it is well known that , if @xmath365 is morse , then there exists @xmath1393 such that all @xmath1394 have the same number of critical points as @xmath365 , and each critical point of @xmath366 has the same morse index as the corresponding critical point of @xmath365 .
restrict @xmath378 to @xmath1395 .
now , since @xmath365 is smooth , @xmath1396 is continuous , and @xmath1397 is equal to the set of critical points of @xmath365 .
further , we know that : @xmath1398 therefore , if @xmath364 , then @xmath1399 . therefore , if @xmath363 is small enough , the critical points of @xmath366 will lie arbitrarily close to the corresponding critical points of @xmath365 .
let @xmath258 denote the critical points of @xmath365 , and let @xmath1400 denote the critical point of @xmath366 corresponding to @xmath593 .
pick open ball - shaped morse neighborhoods @xmath1401 for every critical point @xmath593 , and assume that @xmath378 is small enough that @xmath1402 for every @xmath364 and every critical point @xmath1400 .
define : @xmath1403 observe that @xmath368 is compact . by the tubular flow theorem , for every point @xmath251 that is not a critical point , we can find a radius @xmath1404 such that there is an open neighborhood of @xmath251 of radius @xmath1405 that is diffeomorphic to @xmath1406 , and such that the diffeomorphism carries flow lines of @xmath381 to flow lines of @xmath1391 .
let @xmath1407 denote the tubular flow neighborhood of @xmath251 , and observe that @xmath1407 is an infinite open cover of the compact submanifold @xmath368 .
we can therefore find a finite subcover @xmath1408 .
let @xmath1409^n \to u_1 , ... , \phi_m:[-1,1]^n\to u_m$ ] denote the diffeomorphisms .
equip each copy of @xmath1410^n$ ] with the pullback of the metric of @xmath5 , rather then the standard euclidean metric .
pick an arbitrary flow line @xmath1411\to m$ ] parameterized by arc length , so that @xmath1412 . assume without loss of generality that @xmath1413 form an open cover of the image of @xmath225 ( see figure [ fig : coverofflowline ] ) .
let @xmath185 be the point where the pullback of @xmath225 to @xmath1414 intersects @xmath1415 .
@xmath1416 will be open in @xmath1414 , so there is some disk of radius @xmath536 around @xmath185 in @xmath1415 which is contained within @xmath1417 .
define the set : @xmath1418 where @xmath1419 is the morse index of @xmath593 .
the boundary of @xmath48 can be considered as the union of two ( possibly disconnected , possibly empty ) overlapping parts : @xmath1420@xmath1421 schematically , this takes the form shown in figure [ fig : schematicofsvw ] .
now observe that , if we pull @xmath366 back to @xmath1414 with @xmath1431 : @xmath1432@xmath1433 and we know that : @xmath1434 therefore , if @xmath378 is small enough : @xmath1435 for all @xmath364 . and , for @xmath378 small enough , @xmath1436 .
what this means is that the flow lines of @xmath1437 for any @xmath364 will enter @xmath48 through @xmath418 and leave through @xmath551 , but not vice - versa , if @xmath378 is small enough .
this means , first , that @xmath1438 .
in addition , it means that @xmath1439 .
define @xmath1440 \in h_{k_i-1}(w)$ ] , where @xmath1441 is the singular homology of @xmath551 with @xmath1442 coefficients . since @xmath551 is diffeomorphic to @xmath1443^{n - k_i+1}$ ] , we know that @xmath1444
. if @xmath1445 is trivial as an element of the homology of @xmath1446 , then it bounds some disk @xmath1447 , so that @xmath1448 . since the descending manifold of @xmath1400 is transverse to the ascending manifold at @xmath1400 , and intersects it at no other point , that implies that @xmath1449 , which is a contradiction .
therefore , @xmath1445 can not be trivial .
this , in turn , implies that for any choice of @xmath1450 , there exists some point @xmath1451 .
therefore , if @xmath378 is small enough to ensure equation [ eqn : nablaf2rvandrw ] holds , then the intersection @xmath1452 will be non - empty , and contain at least some open @xmath1419-disk .
call the image of this disk in @xmath5 @xmath1453 .
now consider @xmath1454 , as shown in figure [ fig : u1 ] . from the tubular flwo theorem ,
we know that , in @xmath1410^n$ ] , the pullback of @xmath381 has the form : @xmath1455 where @xmath1456 .
since @xmath1410^n$ ] is compact , we can find @xmath1457 such that @xmath1458 for all @xmath263 .
let @xmath1459 be the point where @xmath1460 leaves @xmath1410^n$ ] .
we also know that : @xmath1461@xmath1462 now , pick an arbitrary point in @xmath1454 , and let @xmath1463 \to [ -1,1]^n$ ] be the flow line beginning at that point for @xmath1464 , parameterized by @xmath1465 , such that @xmath1466^n$ ] .
then : @xmath1467 let @xmath1468 denote the @xmath1469 coordinate of @xmath1470 .
then equations [ eqn : nablaph1f2 ] , [ eqn : boundnablaphi2f2minusf1 ] , and [ eqn : dgamma1dt ] together imply that : @xmath1471@xmath1472 therefore , since the width of @xmath1410^n$ ] is 2 , we can bound @xmath1473 by : @xmath1474 and , for @xmath1475 : @xmath1476 therefore , for any choice of @xmath794 , if @xmath536 and @xmath378 are small enough , then @xmath1477 will be within @xmath572 of @xmath1459 . since this is an open condition , if @xmath536 and @xmath378 are small enough , there will be some @xmath1419-disk in @xmath1478 within @xmath572 of @xmath1459 . since @xmath1479
, this means that , if @xmath572 is small enough , this @xmath1419-disk will lie within @xmath1480 .
call the image of this disk in @xmath5 @xmath1481 .
now repeat this process for @xmath1481 in @xmath1482 .
if @xmath572 and @xmath378 are small enough , there will be some @xmath395 within @xmath574 of @xmath1483 , which is the point where @xmath225 leaves @xmath1484 . then repeat the process for @xmath395 in @xmath1485 , and so on , until we reach an open @xmath1419-disk @xmath1486 . then @xmath1487 will be a @xmath1419-disk in @xmath1488 .
analogous to @xmath1414 , we can define : @xmath1489@xmath1490@xmath1491 and , if @xmath378 is small enough , then @xmath1492 , the ascending manifold of @xmath1493 intersects @xmath1494 in @xmath1495 , and the descending manifold of @xmath1493 intersects @xmath1496 in @xmath1497 . since the pullback of @xmath225 lies within @xmath1498 , which is within @xmath1499 , then if @xmath1500 is small enough , @xmath1487 will lie within @xmath1499 .
generically , the intersection of @xmath1487 with @xmath1501 will be @xmath1502-dimensional .
therefore , for any choice of flow line @xmath225 from @xmath593 to @xmath1503 for @xmath365 , we can find some @xmath1504 such that if @xmath1505 , then for any @xmath364 there exists some flow line from @xmath1400 to @xmath1493 for @xmath366 that lies within @xmath57 of @xmath225 .
since the set of flow lines emerging from @xmath593 is diffeomorphic to @xmath1506 , we can find some @xmath1507 such that if @xmath1508 , then for any flow line @xmath225 emerging from @xmath593 and for any @xmath364 there exists some flow line from @xmath1400 to @xmath1493 for @xmath366 that lies within @xmath57 of @xmath225 . and since there are only finitely many critical points of @xmath365 , we can find @xmath378 such that , if @xmath364 , this holds for any critical point .
this concludes the proof . * lemma [ lemma3.2 ] : * _ let @xmath367 be a morse function , where @xmath5 is a closed manifold .
let @xmath368 be a compact codimension-0 subset of @xmath5 that includes no critical points of @xmath35 .
then for any @xmath362 there exists @xmath363 such that for any critical point @xmath369 and any points @xmath370 , if : @xmath1509@xmath1510 then : @xmath373 and , for any @xmath362 , there exists @xmath363 such that , for any points @xmath370 , if @xmath1509@xmath1511 then : @xmath375 _ * proof : * we begin by proving that , given a _ specific _ point @xmath1512 , we can find @xmath378 depending on @xmath513 such that if @xmath1513 , then @xmath1514 .. we will then use compactness to extend the result to all of @xmath368 .
we prove the statement for the descending manifold , since @xmath1515 .
assume @xmath513 is in the ascending manifold of @xmath369 .
let @xmath1516 be a morse neighborhood of @xmath369 , so that @xmath1517 is an open @xmath14-ball and the image @xmath1518 is contained in an open ball of radius @xmath1519 centered at @xmath369 .
let @xmath111 be the coordinates of @xmath1520 . by definition , @xmath1521 .
therefore : @xmath1522 this means that if @xmath1523 is in the ascending manifold of @xmath1524 , its descending manifold in @xmath1517 is simply a straight line starting at @xmath251 and ending at 0 .
now , let @xmath1411 \to m$ ] be a parameterization of the descending manifold of @xmath513 by arc length , @xmath1525 . by the tubular flow theorem , for every point @xmath1526 , we can find a local neighborhood @xmath1527 of @xmath764 , where @xmath1528 is diffeomorphic to @xmath1406 , where the flow lines of @xmath1529 are mapped to lines with constant @xmath1530 coordinates , and where the image @xmath1531 is contained in a @xmath1532-neighborhood of the image of @xmath225 .
then , with the addition of @xmath1518 , @xmath1531 form an infinite cover of the image of @xmath225 , as shown in figure [ fig : gammaanditscovering ] .
since the image of @xmath225 is compact , we can find a finite subcover @xmath1533 - that is , @xmath1534 is close to the critical point @xmath1459 , while @xmath1535 contains our starting point @xmath513 .
we write @xmath1536 and let @xmath1431 denote the embedding @xmath1537 for succinctness .
now , consider the preimage @xmath1538 .
as shown in figure [ fig : tubularneighborhood ] , let @xmath1539 denote the projection of @xmath1538 to @xmath1540 , dropping the @xmath1541 coordinate , and let @xmath1542 denote the ascending manifold of @xmath1538 in @xmath1543 . since the pullback of the vector field @xmath1529 to @xmath1543 is @xmath1544 , @xmath1542 will be diffeomorphic to @xmath1545 : @xmath1546 note that @xmath1542 will contain @xmath1547 . note
also that , since @xmath1548 is a diffeomorphism onto its image and its domain is open , @xmath1518 is open .
therefore @xmath1549 is open , so @xmath1550 is open , so @xmath1542 is open . now consider @xmath1551 .
let @xmath1552 be the projection of @xmath1551 that drops the @xmath1541 coordinate , let @xmath1553 be the ascending manifold of @xmath1551 in @xmath1554 , and observe that @xmath1555 , that @xmath1553 contains @xmath1556 , and that @xmath1553 is open . continue repeating this process until you reach @xmath1557 .
now consider @xmath1558 .
since @xmath1559 is a diffeomorphism onto its image , @xmath1558 will be open . since @xmath1557 contains the @xmath1560
, @xmath1558 contains @xmath513 .
therefore , @xmath1558 contains an open ball @xmath1561 of radius @xmath378 around @xmath513 .
suppose @xmath1562 , and @xmath515 lies in the same component of @xmath1563 as @xmath513 .
then @xmath1564 , so its descending manifold in @xmath1565 lies in @xmath1557 .
we can continue forward from here , showing that the preimage in @xmath1414 of the descending manifold @xmath1566 lies in @xmath1567 for all @xmath127 .
then , once we reach @xmath1517 , since @xmath1568 , the descending manifold terminates in a curve to @xmath369 .
since @xmath1414 lie in an @xmath57-neighborhood of @xmath1569 , we may conclude that , for any @xmath1570 , we can find @xmath1571 such that , if : @xmath1572@xmath1573 then : @xmath1574 let @xmath1575 denote an open ball of radius @xmath1576 around @xmath251 , where @xmath379 .
these balls form an infinite open cover of @xmath368 .
since @xmath368 is compact , we can find a finite subcover @xmath1577 .
define : @xmath1578
suppose @xmath370 and @xmath1513 .
@xmath513 must be contained in some ball @xmath1579 .
then , the distance from @xmath515 to @xmath1580 is : @xmath1581 so @xmath1582 .
therefore : @xmath1583@xmath1584 which in turn implies that : @xmath373 this concludes the proof . * lemma [ lemma3.3 ] : * _ let @xmath376 be a morse function , where @xmath5 is a compact manifold . for any choice of @xmath362 and any compact codimension-0 submanifold @xmath377 that lies in the descending manifold of maxima of @xmath365 and contains no critical points of @xmath365 , there exists @xmath378 such that , for any generic choice of @xmath364 and any @xmath379 , the ascending manifold of @xmath251 for @xmath380 will lie within an @xmath57-neighborhood of the ascending manifold of @xmath251 for @xmath381 .
_ * proof : * the proof of this lemma is essentially analogous to the proof of lemma [ lemma3.1 ] . pick any point @xmath379 , and let @xmath1585 \to m$ ] be the flow line from @xmath251 to its maximum @xmath369 for @xmath1586 parameterized by arc length , so that @xmath1587 .
let @xmath1588 be a morse neighborhood of @xmath369 of radius less then @xmath57 .
find a cover of @xmath1589 by tubular flow neighborhoods , @xmath1409^n \to u_1 , ... ,
\phi_m:[-1,1]^n \to u_m$ ] .
restrict @xmath378 to be small enough that the critical points of any @xmath364 will be in one - to - one correspondence with the critical points of @xmath365 , and the critical point of @xmath366 corresponding to @xmath369 will lie inside @xmath1590 . note that we require @xmath366 to always be generic , to ensure that @xmath251 will still lie in the descending manifold of the maximum of @xmath366 corresponding to @xmath369 .
consider @xmath1591 . by the tubular flow theorem
, this will equal : @xmath1592 for some function @xmath1593^n \to \mathbb{r } , \lambda_1(x ) > 0 $ ] .
since @xmath1410^n$ ] is compact , we can find @xmath1457 such that @xmath1458 for all @xmath263 .
let @xmath1459 be the point where @xmath1460 leaves @xmath1410^n$ ] .
we know that , for @xmath364 : @xmath1594@xmath1595 define @xmath1596\to[-1,1]^n$ ] to be the flow line beginning at @xmath1597 and ending on @xmath1598^n$ ] for @xmath1599 , parameterized by @xmath1600 .
then : @xmath1601 let @xmath1602 denote the @xmath1469 coordinate of @xmath1603 .
then equations [ eqn : lemma3 - 3eq1 ] , [ eqn : lemma3 - 3eq2 ] , and [ eqn : lemma3 - 3eq3 ] together imply that : @xmath1604@xmath1605 therefore , since the width of @xmath1410^n$ ] is 2 , for any @xmath364 we can bound @xmath1473 by : @xmath1474 and , for @xmath1606 : @xmath1607 since the pullback of @xmath764 to @xmath1410^n$ ] has constant @xmath1530 coordinates for all @xmath29 , we may conclude that , for any choice of @xmath794 , if @xmath378 is small enough then @xmath1608 will lie within @xmath572 of @xmath1459 for any @xmath364 - and , indeed , all points of @xmath1603 will lie within @xmath572 of @xmath225 within @xmath1609 . and , if @xmath572 is small enough , @xmath1610 will lie inside @xmath1611^n)$ ] for any choice of @xmath364 .
now define @xmath1453 to be a closed @xmath572-ball around @xmath1612 in @xmath5 , and consider @xmath1613^n$ ] .
define @xmath1483 to be the point in @xmath1410^n$ ] where @xmath1614 leaves the cube . for any point @xmath1615 ,
define @xmath1616 \to [ -1,1]^n$ ] to be the flow line of @xmath1617 starting at @xmath515 and ending on @xmath1598^n$ ] . by an analogous argument we can again bound : @xmath1618 where : @xmath1619^n\]]@xmath1620 since @xmath1453 is compact , for any choice of @xmath1621 , for any point @xmath1615 , and for any @xmath364 , if @xmath378 is small enough then the flow line of @xmath1622 starting at @xmath515 and ending on @xmath1598^n$ ] will be within @xmath1623 of @xmath369 .
therefore , for any @xmath1624 and any @xmath364 , if @xmath378 is small enough then the flow line of @xmath1625 beginning at @xmath251 and ending on @xmath1626^n)$ ] will be within @xmath1623 of @xmath225 .
proceeding in this way , we can ultimately show that , if @xmath378 is small enough , then for any @xmath364 , the flow line of @xmath1625 beginning at @xmath251 and ending on @xmath1627^n)$ ] will lie within @xmath1628 of @xmath225 .
if we choose @xmath1629 to be small enough , then the flow line will end on @xmath1630 , where @xmath1588 is the morse neighborhood of the maximum @xmath369 .
now , consider @xmath1631 .
the pullback of @xmath365 to @xmath1631 is : @xmath1632 therefore , if we define @xmath1633 , then in the ambient metric of @xmath1631 : @xmath1634 and , if we choose @xmath378 small enough , we can therefore ensure that : @xmath1635 therefore , if @xmath378 is small enough , the ascending manifold of any point on @xmath1636 for any @xmath1637 , will lie inside @xmath1631 . and
since @xmath1631 has radius less then @xmath57 , the image of this ascending manifold in @xmath5 must lie within @xmath57 of @xmath225 .
therefore , for any @xmath379 we can find @xmath1571 such that , for any generic @xmath1638 , the ascending manifold of @xmath251 for @xmath366 lies within @xmath57 of the ascending manifold of @xmath251 for @xmath365 . and ,
since @xmath368 is compact , we can therefore find @xmath378 such that @xmath1639 for all @xmath251 .
this concludes the proof . * lemma [ lemma3.4 ] : * _ let @xmath15 be a front - generic legendrian submanifold whose front projection is defined by sheet functions @xmath382 , where @xmath383 .
let @xmath96 be a second front - generic legendrian submanifold whose front projection is defined by sheet functions @xmath384 .
then , for any choice of @xmath362 , there exists @xmath363 such that , if : @xmath1640 then , there exists a bijection between the rigid gradient flow trees @xmath38 with one positive reeb chord and the rigid gradient flow trees of @xmath357 with one positive reeb chord , such that a tree @xmath232 of @xmath38 shares the same reeb chords with its corresponding tree @xmath386 of @xmath357 , and such that the projection of @xmath386 to @xmath5 lies within an @xmath57-neighborhood of the projection of @xmath232 to @xmath5 . _ * proof : * restrict @xmath378 to be small enough that the reeb chords of @xmath357 are in one - to - one correspondence with the reeb chords @xmath38 .
consider some rigid gradient flow tree @xmath232 of @xmath38 , which has only one positive reeb chord and whose vertices are only reeb chords and @xmath255 vertices .
begin by considering the case where @xmath232 has no 2-valent vertices at reeb chords .
@xmath232 is equipped with a natural direction , given by the direction of the gradient flow functions .
label each vertex of @xmath232 with the number of edges of the longest directed path from that vertex to a negative reeb chord - so , e.g. , reeb chords would be labeled with 0 .
let @xmath1567 denote the set of vertices of @xmath232 which are labeled with @xmath127 . for a @xmath255 vertex @xmath1641 ,
define @xmath1642 to be the pair of vertices in @xmath232 such that there is an edge in @xmath232 connecting @xmath238 to them , directed from @xmath238 to @xmath1642 .
define @xmath1643 to be the sheet height functions such that the edge from @xmath238 to @xmath1644 flows on the gradient flow @xmath1645 .
let @xmath1646 denote the corresponding sheet height functions of @xmath357 .
we will define sets @xmath1647 iteratively .
we begin by defining , for any @xmath1648 , @xmath1649 .
then , for any @xmath238 - not just in @xmath1650 - we define : @xmath1651 and for @xmath1652 , we define : @xmath1653 now , for any reeb chord of @xmath38 designated by @xmath1648 , there is a corresponding reeb chord of @xmath357 .
we denote this reeb chord by @xmath1654 , and we define : @xmath1655 then , for @xmath1652 , we define : @xmath1656@xmath1657 now , consider any @xmath1658 . by lemma [ lemma3.1 ] ,
if @xmath363 is small enough , then @xmath1659 lies within an @xmath536-neighborhood of @xmath1660 for all @xmath1648 , for any choice of @xmath1661 .
then , let @xmath1662 denote an @xmath536-neighborhood of @xmath1660 .
for @xmath1663 , consider @xmath1664 .
if @xmath536 is small enough , then for any choice of @xmath430 , every point in @xmath1665 will lie within a @xmath1666-neighborhood of @xmath1667 .
since @xmath1668 , we know @xmath1669 is non - empty , and @xmath1670 . from this
we conclude that if @xmath536 is small enough , then @xmath1669 lies within a @xmath1666-neighborhood of @xmath1671 . now consider @xmath1659 for @xmath1663 . we know that @xmath1671 must be at least codimension-1 , because if they are codimension-0 , then @xmath232 is not rigid ; and since @xmath378 is small enough that our reeb chords correspond , this means @xmath1669 must also be at least codimension-1 . therefore , we can pick @xmath377 containing @xmath1671 but not containing any reeb chords of @xmath1672 . therefore , by lemma [ lemma3.3 ] , if @xmath1666 is small enough , then ascending manifold of @xmath1669 for @xmath1673 will lie within @xmath564 of @xmath1659 . and by lemma [ lemma3.2 ] , if @xmath1666 is small enough , the ascending manifold of @xmath1669 for @xmath1673 will lie within @xmath564 of @xmath1660 .
therefore , if @xmath1666 is small enough , then @xmath1659 will lie within @xmath572 of @xmath1660 .
now , let @xmath1674 , and consider @xmath1675 .
note that @xmath1676 . if @xmath572 is small enough ,
then for any choice of @xmath434 , every point in @xmath1665 will lie within @xmath435 of @xmath1667 . since @xmath1677 , we conclude that if @xmath572 is small enough , then @xmath1669 is non - empty and lies within a @xmath435-neighborhood of @xmath1678 .
we keep repeating this process until we reach @xmath1679 , where @xmath187 is the positive reeb chord of @xmath232 , and @xmath1680 is the @xmath255 vertex connected by an edge to @xmath187 . by this process
, we show that @xmath1681 lies within an @xmath1500-neighborhood of @xmath1682 .
let @xmath445 denote the reeb chord of @xmath357 corresponding to @xmath187
. now define @xmath1683 to be : @xmath1684@xmath1685 since @xmath232 is rigid , @xmath1683 must be one - dimensional , and @xmath1686 must be equal to the image in the base space of the edge of @xmath232 emerging from @xmath187 . by lemma [ lemma3.1 ] ,
if @xmath378 is small enough , then @xmath1687 will lie within @xmath1688 of @xmath1689 .
and we already know that if @xmath1690 are small enough , then @xmath1682 will lie within @xmath1688 of @xmath1681 .
therefore , if @xmath1688 is small enough , then @xmath1686 will lie within @xmath1691 of @xmath1692 . and ,
since @xmath1683 are one - dimensional , we know that @xmath1693 and @xmath1694 . once again , since @xmath232 is rigid , we can conclude that @xmath1695 are one - dimensional .
we know that @xmath1698 lies within @xmath1699 of @xmath1700 . and
, by lemma [ lemma3.2 ] , if @xmath1701 are small enough , @xmath1702 will lie within @xmath1703 of @xmath1704 .
therefore , if @xmath1705 are small enough , @xmath1706 will lie within @xmath1707 of @xmath1708 .
repeat this process until we reach the reeb chords .
the trace of the @xmath1709 form a unique rigid gradient flow tree @xmath386 of @xmath357 with reeb chords corresponding to the reeb chords of @xmath232 , and lying within an @xmath57-neighborhood of @xmath232 .
therefore , for every rigid @xmath232 there exists some @xmath1711 such that if @xmath1712 for all @xmath127 , then there exists a unique rigid gradient flow tree @xmath386 of @xmath357 that lies within an @xmath57-neighborhood of @xmath232 .
since there are only finitely - many rigid flow trees , we can therefore find @xmath378 such that , if @xmath1712 for all @xmath127 , then @xmath386 lies within an @xmath57-neighborhood of @xmath232 for all @xmath232 .
this completes the proof . | we prove a seifert - van kampen theorem for legendrian submanifolds and exact lagrangian cobordisms , and use it to calculate the change in the dga caused by critical legendrian ambient surgery .
* keywords : * symplectic , legendrian , reeb chords , exact lagrangian cobordism , gradient flow tree , contact homology * ams subject code : * 53d12 , 53d42 |
lattice qcd is the ideal approach not only for computing @xmath3 meson spectrum , but also for hybrids and glueballs .
however , the lattice technique is not free of systematic errors .
the wilson gauge and quark actions suffer from significant lattice spacing errors , which are smaller only at very large @xmath4 , and very large lattice volume is required to get rid of finite size effects .
there have been several quenched lattice calculations of hybrid meson masses , part of them are listed in tab .
[ tab6 ] . in ref .
@xcite , the wilson gluon action and quark action were used . in refs .
@xcite , the authors used wilson gauge action and sw improved quark action . for the hybrid mesons containing heavy quarks @xmath5 , the nrqcd action@xcite and
the lbo action@xcite have also been applied .
there is also a recent work using the improved ks quark action@xcite . in this work ,
we employ _ both improved gluon and quark actions on the anisotropic lattice _ , which should have smaller systematic errors , and should be more efficient in reducing the lattice spacing and finite volume effects .
we will present data for the @xmath2 hybrid mass and the splitting between the @xmath2 hybrid mass and the spin averaged s - wave mass for charmonium .
details can be found in ref .
the total lattice action is @xmath6 .
the improved gluonic action @xmath7 is @xcite : @xmath8 where @xmath9 stands for a @xmath10 plaquette and @xmath11 for a @xmath12 rectangle .
the sw improved action for quarks@xcite is @xmath13 \nonumber \\ & - & \kappa_s \sum_{x , j } [ { \bar \psi}(x ) ( 1-\gamma_j)u_{j}(x ) \psi ( x+\hat{j } ) \nonumber \\ & + & { \bar \psi } ( x ) ( 1+\gamma_j)u^{\dag}_{j}(x-\hat{j})\psi(x-\hat{j } ) ] \nonumber \\ & + & i { \kappa}_s { c}_s^{ti } \sum_{x , j < k } { \bar \psi } ( x ) \sigma_{jk}{\hat f_{jk}}(x)\psi ( x ) \nonumber \\ & + & i \kappa_s c_t^{ti } \sum_{x , j}{\bar \psi } ( x ) \sigma_{j4 } { \hat f_{j4 } } ( x ) \psi ( x ) , \label{s_q}\end{aligned}\ ] ] where @xmath14 stands for the clover - leaf construction@xcite for the gauge field tensor .
tadpole improvement is carried out so that the actions are more continuum - like .
on our pc cluster@xcite , the su(3 ) pure gauge configurations were generated with the gluon action in eq .
( [ s_g ] ) using cabibbo - marinari pseudo - heatbath algorithm .
the configurations are decorrelated by su(2 ) sub - group over - relaxations .
we calculated the tadpole parameter @xmath15 self - consistently .
90 independent gauge configurations at @xmath0 and @xmath16 on the @xmath17 lattice were stored .
although such an ensemble is not very big , it is bigger than earlier simulations by ukqcd and milc collaborations@xcite on isotropic lattices .
the quark propagator was obtained by inverting the matrix @xmath18 in @xmath19 in eq .
( [ s_q ] ) by means of bicgstab algorithm .
the residue is of @xmath20 .
we computed the correlation functions with various sources and sinks@xcite , at four values of the wilson hopping parameter ( @xmath21 , 0.4199 , 0.4279 , 0.4359 ) . in fig .
[ mprf1t.eps ] , we plot the effective masses for the @xmath22 , @xmath23 , @xmath24 ordinary mesons and @xmath2 exotic meson at @xmath25 . for the ordinary mesons , we used the their corresponding operator as both source and sink .
for the exotic meson , we tried two different cases : ( 1 ) the @xmath2 operator as both source and sink ; ( 2 ) the @xmath26 source and @xmath2 sink , which give consistent results within error bars .
the cp - pacs , milc and ukqcd collaborations used the unimproved wilson gauge action to generate configurations .
they had to work on very large @xmath27 ) , corresponding to very small @xmath28 0.1 fm ) , to get rid of the finite spacing errors .
they had also to use very large lattices @xmath29 , to avoid strong lattice size effects at such small @xmath30 . in comparison ,
our lattices are much coarser ( @xmath31 fm ) , and the number of lattice sites is much smaller .
our finite size effects could be ignored , for the physical size of the spatial lattice is @xmath32 and should be big enough .
our results for the effective mass indicate the existence of a much wider plateau than in the previous work on isotropic lattices .
by extrapolating the effective mass of the @xmath2 hybrid meson to the chiral limit , and using @xmath33 determined from the @xmath23 mass , we get @xmath34 mev . in tab .
[ tab6 ] , we compare the results from various lattice methods . our result is consistent with the milc data@xcite , obtained using the wilson gluon action and clover quark action on much larger isotropic lattices and much smaller @xmath30 .
we also show our results in tab .
[ tab6 ] for the @xmath2 hybrid meson mass in the charm quark sector , using the method discussed in refs .
our corresponding @xmath35 is obtained by tuning @xmath36=3067.6 mev , where on the right hand side , the experimental inputs @xmath37 mev and @xmath38 mev are used .
the @xmath2 hybrid meson mass at our @xmath39 is @xmath40 mev , is consistent with the milc data@xcite .
the splitting between the hybrid meson mass and the spin averaged s - wave mass [ @xmath41 , at our @xmath42 is @xmath43 mev , consistent with the cp - pacs data , obtained using the wilson gluon action and nrqcd quark action on much larger anisotropic lattices and much smaller @xmath30 .
as a byproduct , we give the @xmath24 p - wave @xmath44 meson in the chiral limit , as well as their experimental values@xcite .
if we assume that the pion is massive and @xmath45 is made of @xmath46 , the @xmath24 p - wave @xmath44 meson mass would be @xmath47 mev .
to summarize , we have used the tadpole - improved gluon action and clover action to compute the hybrid meson masses on much coarser anisotropic lattices .
the main results are given in tab .
[ tab6 ] and compared with other lattice approaches . in our opinion ,
our approach is more efficient in reducing systematic errors due to finite lattice spacing . | we study hybrid mesons from the clover and improved gauge actions at @xmath0 on the anisotropic @xmath1 lattice using our pc cluster .
we estimate the mass of @xmath2 light quark hybrid as well as the mass of the charmonium hybrid .
the improvement of both quark and gluonic actions , first applied to the hybrid mesons , is shown to be more efficient in reducing the lattice spacing and finite volume errors . |
the identification of the higgs boson(s ) is one of the main goals of the large hadron collider ( lhc ) being built at cern .
there are expectations that there exists a ` light ' higgs boson with mass @xmath6 gev . in this mass range
, its detection at the lhc will be challenging .
there is no obvious perfect detection process , but rather a range of possibilities , none of which is compelling on its own . _
either _ large signals are accompanied by a huge background , _ or _ the processes have comparable signal and background rates for which the number of higgs events is rather small . here
we wish to draw attention to the exclusive signal @xmath7 , where the + sign indicates the presence of a rapidity gap .
it may be possible to install proton taggers so that the ` missing mass ' can be measured very accurately .
the experimental challenge is to provide a set - up in which the bulk of the proton - tagged signal is deposited in a small missing mass window @xmath8 @xcite .
the exclusive process allows the mass of the higgs to be measured in two independent ways .
first the tagged protons give @xmath9 and second , via the @xmath10 decay , we have @xmath11 , although now the resolution is much poorer with @xmath12 gev or more .
the existence of matching peaks , centered about @xmath13 , is a unique feature of the exclusive diffractive higgs signal .
besides its obvious value in identifying the higgs , the mass equality also plays a key role in reducing background contributions .
another crucial advantage of the exclusive process @xmath14 , with @xmath10 , is that the leading order @xmath15 background subprocess is suppressed by a @xmath16 , p - even selection rule @xcite .
the basic mechanism for the exclusive process , @xmath14 , is shown in fig . @xmath17 . since the dominant contribution comes from the region @xmath18
the amplitude may be calculated using perturbative qcd techniques@xcite @xmath19 where the overall normalization constant @xmath20 can be written in terms of the @xmath21 decay width@xcite , and where the @xmath22 vertex factors for cp @xmath23 higgs production are , after azimuthal - averaging , v_h(0^+ ) q_t^2 , and v_a(0 ^ - ) ( p_1
t p_2t)n_0 , [ eq : rat4 ] expressions ( [ eq : m],[eq : rat4 ] ) hold for small @xmath24 , where the @xmath25 are the transverse momenta of the outgoing protons , and @xmath26 is a unit vector in the beam direction . the @xmath27 s are the skewed unintegrated gluon densities at the hard scale @xmath28 , taken to be @xmath29 . since @xmath30 , it is possible to express @xmath31 , to single log accuracy , in terms of the conventional integrated density @xmath32 .
the @xmath27 s embody a sudakov suppression factor @xmath33 , which ensures that the gluon does not radiate in the evolution from @xmath34 up to the hard scale @xmath29 , and so preserves the rapidity gaps .
the apparent infrared divergence of ( [ eq : m ] ) is nullified for @xmath35 production by these sudakov factors .
however the amplitude for @xmath36 production is much more sensitive to the infrared contribution .
indeed let us consider the case of small @xmath24 of the outgoing protons .
then we see , from ( [ eq : rat4 ] ) , that the @xmath37 integration for @xmath35 is replaced by @xmath38 for @xmath36 , and now the sudakov suppression is not enough to prevent a significant contribution from the @xmath39 domain .
= 0.4 the radiation associated with the @xmath40 hard subprocess is not the only way to populate and to destroy the rapidity gaps .
there is also the possibility of soft rescattering in which particles from the underlying event populate the gaps .
the probability , @xmath41 at the lhc , that the gaps survive the soft rescattering was calculated using a two - channel eikonal model , which incorporates high mass diffraction@xcite . including this factor , and the nlo @xmath42 factor ,
the cross section is predicted to be @xcite @xmath43 for the production of a standard model higgs boson of mass 120 gev at the lhc .
it is estimated that there may be a factor of 2.5 uncertainty ( up or down ) in this prediction@xcite .
if we include a factor 0.6 for the efficiency associated with proton tagging , 0.67 for the @xmath10 branching fraction , 0.6 for @xmath44 and @xmath45 tagging , 0.5 for the @xmath46 jet polar angle cut , @xmath47 , ( necessary to reduce the @xmath0 qcd background)@xcite , then , for a luminosity of @xmath48 , the original @xmath49 events are reduced to an observable signal of 11 events .
the advantage of the @xmath50 signal is that there exists a @xmath16 selection rule , which requires the leading order @xmath51 background subprocess to vanish in the limit of massless quarks and forward outgoing protons .
( the @xmath52 superscript is to note that each gluon comes from colour - singlet @xmath53-channel exchange . ) however , in practice , lo background contributions remain .
the prolific @xmath54 subprocess may mimic @xmath0 production since we may misidentify the outgoing gluons as @xmath44 and @xmath45 jets . assuming the expected 1% probability of misidentification , and applying @xmath47 jet cut , gives a background - to - signal ratio @xmath55 .
( here , for reference , we assume that the bulk of the higgs signal can be collected within an interval @xmath56 gev . ) secondly , there is an admixture of @xmath57 production , arising from non - forward going protons which gives @xmath58 .
thirdly , for a massive quark there is a contribution to the @xmath16 cross section of order @xmath59 , leading to @xmath55 , where @xmath60 is the transverse energy of the @xmath44 and @xmath45 jets .
next , we have the possibility of nlo @xmath61 background contributions .
of course , the extra gluon may be observed experimentally and these background events eliminated . however , there are exceptions .
the extra gluon may go unobserved in the direction of a forward proton .
this background may be effectively eliminated by requiring the equality @xmath62 . moreover ,
soft gluon emissions from the initial @xmath63 state factorize and , due to the overriding @xmath16 selection rule , these contributions to the qcd @xmath0 production are also suppressed .
the remaining danger is large angle hard gluon emission which is collinear with either the @xmath44 or @xmath45 jet , and therefore unobservable .
if the cone angle needed to separate the @xmath64 jet from the @xmath44 ( or @xmath45 ) jet is @xmath65 then the expected background from unresolved three jet events leads to @xmath66 .
the nnlo @xmath67 background contributions are found to be negligible ( after requiring @xmath68 ) , as are soft pomeron - pomeron fusion contributions to the background ( and to the signal ) @xcite .
also note that radiation off the screening gluon , in fig .
@xmath17 , is numerically small@xcite .
so , in total , for the exclusive production of a 120 gev ( sm ) higgs boson at the lhc with the integrated luminosity @xmath69 fb@xmath70 , the signal - to - background ratio is s / b ( 1 /m_missing ) 11/4 events , after cuts and acceptance .
this corresponds to a statistical significance of roughly @xmath71 .
that is , if almost the whole higgs signal can be collected within the interval @xmath56 gev , then @xmath72 , corresponding to a @xmath73 signal . in the case of a gaussian missing mass distribution of width @xmath74 ,
about 87% of the signal is contained in a bin @xmath75 , that is @xmath76 .
we could consider higgs production in other diffractive channels , such as diffractive production accompanied by proton dissociation ( @xmath77 ) , or central inelastic production ( @xmath78 ) .
however they are worse than the usual totally inclusive production
there is no precise missing mass measurement , no selection rule to suppress the background and more serious pile - up problems
. the somewhat smaller density of soft secondary hadrons in the higgs rapidity region does not compensate for the much smaller statistics ( cross sections ) in diffractive processes .
to be specific , we discuss the three neutral higgs bosons of the mssm model : @xmath79 with cp=1 and @xmath80 with cp=1 .
there are regions of mssm parameter space where the conventional signals ( @xmath81 decays ) are suppressed , but where the exclusive subprocess @xmath82 is strongly enhanced@xcite .
for example , for @xmath83 = 130 gev and tan@xmath84 = 50 , we have @xmath85 = 124.4 gev with @xmath86 events , @xmath87 = 135.5 gev with @xmath88 events and @xmath83 = 130 gev with @xmath89 events , so both @xmath90 and @xmath91 should be clearly visible .
( again , for reference , we assume that @xmath56 gev can be achieved . )
the decoupling regime ( @xmath92 and tan@xmath93 ) is another example where the exclusive signal is of great value . in this case
@xmath90 is indistinguishable from a sm higgs , and so the discovery of @xmath91 is crucial to establish the underlying dynamics .
the plot of fig .
@xmath94 , with tan@xmath95 , shows that a @xmath96 signal is possible up to quite large values of @xmath87 . if the exclusive cross sections for scalar and pseudoscalar higgs production were comparable , it would be possible to separate them readily by a missing mass scan , and by the study of azimuthal correlations between the outgoing protons .
unfortunately pseudoscalar exclusive production is strongly suppressed by the p - even selection .
maybe the best chance to identify the @xmath36 boson is through the double - diffractive process , @xmath97 , where both protons dissociate@xcite .
the exclusive higgs signal is particularly clean , and the signal - to - background ratio is especially favourable , in comparison with the other proposed detection modes . however
the expected number of events is low .
therefore it is important to check the predictions for exclusive higgs production by studying processes mediated by the same mechanism , but with rates which are sufficiently high that they may be observed at the tevatron ( as well as at the lhc ) .
the most obvious examples are those in which the higgs of fig . @xmath17 is replaced by either a dijet system , a @xmath3 or @xmath4 meson , or by a @xmath5 pair .
first , we discuss the exclusive production of a pair of high @xmath60 jets , @xmath98 @xcite .
this would provide an effective @xmath63 ` luminosity monitor ' just in the kinematical region of the higgs production .
the corresponding cross section was evaluated to be about 10@xmath99 times larger than that for the sm higgs boson .
thus , in principle , this process appears to be an ideal ` standard candle ' .
the expected cross section is rather large , and we can study its behaviour as a function of the mass of the dijet system . unfortunately , in the present cdf environment , the background from ` inelastic pomeron - pomeron collisions ' is large as well . theoretically the exclusive dijets should be observed as a narrow peak , sitting well above the background , in the distribution of the ratio r_jj = e_dijet / e_pp at @xmath100 , where @xmath101 is the energy of the incoming pomeron - pomeron system . in practice
the peak is smeared out due to hadronization and the jet - searching algorithm . for jets with @xmath102 gev and a jet cone @xmath103 ,
more than 1 gev will be lost outside the cone , leading to ( i ) a decrease of the measured jet energy of about 1 - 2 gev , and , ( ii ) a rather wide peak ( @xmath104 ) in the @xmath105 distribution .
the estimates based on ref .
@xcite give an exclusive cross section for dijet production with @xmath106 gev ( and cdf cuts ) of about 40 pb , which is very close to the recent cdf measurement@xcite , ( r_jj>0.8 , e_t>25 gev ) = 34 5(stat ) 10(syst ) pb .
however there is no ` visible ' peak in the cdf data for @xmath105 close to 1 .
the contribution from other channels ( called central inelastic in ref .
@xcite ) is too large , and matches with the expected peak smoothly jet searching algorithm , rather than the jet cone algorithm , would improve the selection of the exclusive events .
this is in accord with the studies in ref .
@xcite . ] .
an alternative possibility is to measure exclusive double - diffractive @xmath107 production with high @xmath60 photons , that is @xmath108 @xcite .
here there are no problems with hadronization or with the identification of the jets . on the other hand
the exclusive cross section is rather small . as usual ,
the perturbative qcd pomeron is described by two ( reggeized ) gluon exchange .
however the photons can not be emitted from the gluon lines directly .
we need first to create quarks .
thus a quark loop is required , which causes an extra coupling @xmath109 in the amplitude .
the predictions of the cross section for exclusive @xmath5 production are shown in fig .
@xmath110 .
= 0.8 recently the first ` preliminary ' result on exclusive @xmath3 production has been reported@xcite .
although it is consistent with perturbative qcd expectations@xcite , the mass of the @xmath3-boson , which drives the scale of the process , is too low to justify just the use of perturbative qcd and azimuthal angular distributions appear to be in good agreement with the perturbatively based expectations@xcite . ] .
however , in ref .
@xcite , it was found that both a regge formalism and perturbative qcd predict essentially the same qualitative behaviour for the central double - diffractive production of ` heavy ' @xmath111 and @xmath112 mesons . due to the low scale , @xmath113 ,
there is a relatively small contribution coming from the process in which the incoming protons dissociate . therefore simply selecting events with a rapidity gap on either side of the @xmath114 ,
almost ensures that they will come from the exclusive reaction , @xmath115 .
although exclusive @xmath114 production is expected to dominate , the predicted@xcite event rates are large enough to select double - diffractive dissociative events with large transverse energy flows in the proton fragmentation regions .
such events are particularly interesting .
first , in this case , the large value of @xmath60 provides the scale to justify the validity , and the reasonable accuracy , of the perturbative qcd calculation of the cross section .
next , by measuring the azimuthal distribution between the two @xmath60 flows , the parity of the centrally produced system can be determined .
another possible probe of the exclusive double - diffractive formalism would be to observe central open @xmath0 production ; namely @xmath116 jets with @xmath117 .
again , this would put the application of perturbative qcd on a sounder footing .
it would allow a check of the perturbative formalism , as well as a study of the dynamics of @xmath0 production .
if the higgs is light , @xmath118 gev , it will be experimentally challenging to study it in detail at the lhc .
all possible processes should be considered .
here we have emphasised the unique advantages of exclusive double - diffractive higgs production , _ provided _ the forward outgoing protons can be precisely tagged .
the missing mass , @xmath119 , measured by the forward proton detectors can then be matched with the mass @xmath120 from the main decay mode , @xmath2 .
moreover the qcd @xmath0 background is suppressed by a @xmath16 selection rule .
the events are clean , but the predicted yield is low : about 10 events , after cuts and acceptance , for an integrated luminosity of @xmath121 .
the signal - to - background ratio is about 1 or better , depending crucially on the accuracy with which @xmath119 can be measured .
we have emphasized the importance of checking these perturbative qcd predictions by observing analogous double - diffractive processes , with larger cross sections , at the tevatron .
adm thanks the leverhulme trust for an emeritus fellowship .
this work was supported by the uk pparc , by a royal society fsu grant , by grant rfbr 04 - 02 - 16073 .
xx a. de roeck , v.a .
khoze , a.d .
martin , r. orava and m.g .
ryskin , eur .
j. * c25 * ( 2002 ) 391 .
khoze , a.d .
martin and m.g .
ryskin , eur .
j. * c19 * ( 2001 ) 477 .
khoze , a.d .
martin and m.g .
ryskin , eur .
j. * c14 * ( 2000 ) 525 .
khoze , a.d .
martin and m.g .
ryskin , eur .
j. * c18 * ( 2000 ) 167 .
khoze , a.d .
martin and m.g .
ryskin , eur .
j. * c23 * ( 2002 ) 311 .
kaidalov , v.a .
khoze , a.d .
martin and m.g .
ryskin , eur .
j. * c33 * ( 2004 ) 261 .
khoze , a.d .
martin and m.g .
ryskin , eur .
j. * c26 * ( 2002 ) 229 .
m. gallinaro ( representing the cdf collaboration ) , fermilab - conf-03 - 403-e , november 2003 , arxiv : hep - ph/0311192 . b. cox , j.r . forshaw and a. pilkington , in preparation .
khoze , a.d .
martin , m.g .
ryskin and w.j .
stirling , arxiv : hep - ph/0409037 .
khoze , a.d .
martin , m.g .
ryskin and w.j .
stirling , eur .
j. * c35 * ( 2004 ) 211 .
wa102 collaboration : d. barberis et al .
b467 * ( 1999 ) 165 ; ibid . *
b474 * ( 2000 ) 423 ; ibid . *
b484 * ( 2000 ) 198 ; ibid . *
b488 * ( 2000 ) 225 ; ibid . *
b453 * ( 1999 ) 305,316 .
khoze , a.d .
martin and m.g .
ryskin , eur .
j. * c24 * ( 2002 ) 581 . | we review the signal , and the @xmath0 background , for higgs production by the exclusive double - diffractive process , @xmath1 , and its subsequent @xmath2 decay , at the lhc .
we discuss the production of higgs bosons in both the sm and mssm .
we show how the predicted rates may be checked at the tevatron by observing the exclusive double - diffractive production of dijets , or @xmath3 or @xmath4 mesons , or @xmath5 pairs . |
broad absorption lines ( bal ) quasars constitute about @xmath8 10 % of the overall quasar population .
observationally , they are characterized by very broad ( @xmath9 at zero intensity ) and blue shifted absorption lines mostly from highly ionized species such as nv@xmath101240 , civ@xmath101550 , .
approximately 10 % of the balqso display in addition absorption lines from low ionisation species such as siii , mgii and aliii ( weymann et al .
these are the so - called low ionisation balqso , or lo - bal .
balqso in general are also heavily absorbed in the x - rays ( green & mathur 1996 ) though it is not clear that the hot and highly ionized gas needed to produce the ionization edges seen in the kev range is the same as that which produces the uv and optical absorption troughs .
the balqso phenomenon is poorly understood theoretically .
in particular , its relation to the general qso phenomenon is unclear .
there are basically two classes of models : in the first class , balqso represent a transitory `` cocoon '' phase in an evolutionary process whereby a temporary excess of accretion fuel is expelled by radiation pressure from the central engine through a wind ( voit , weymann & korista 1993 ) .
this model predicts that eventually balqsos will evolve into `` normal '' quasars once the excess gas has been removed . in the second class of models ,
balqso are normal quasars but viewed at a particular inclination such that our line - of - sight intercept a wind of material blown - off the surface of the accretion disk or the top of the molecular torus ( e.g. murray et al .
spectropolarimetric observations tend to support this second hypothesis in that they often reveal significantly polarized emission which is best explained by the scattering of a partially obscured but otherwise normal quasar spectrum ( e.g. wills et al .
1992 , cohen et al .
1995 , hines & wills 1995 , goodrich & miller 1995 ) .
the fact that the emission lines and continuum properties of balqsos are indistinguishable from that of non bal qsos ( weymann et al .
1991 ) also lend credit to the viewing direction hypothesis . an intriguing and important property of balqsos
is that , up to now , there were no known radio - loud broad absorption line quasars ( stocke et al .
the criterion commonly used for radio - loudness is @xmath11 , where r is the k corrected ratio of 5 ghz to 4400 ( b band ) flux ( stocke et al . 1992 ) .
there is some intriguing suggestion that balqsos show an excess of radio emission over non - balqsos such that there is an apparent accumulation of balqsos close - to but slightly below the @xmath12 dividing line ( francis , hooper and impey 1993 ) .
the quasar 1556 + 3517 ( z = 1.48 ) was recently discovered in the course of the first vla radio survey ( becker white and helfand 1995 ) .
1556 + 3517 belongs to the lo - bal class , and , with @xmath13 , unambiguously qualifies as the first and so far only radio - loud balqso ( becker et al .
because it holds the promise of shedding light on the radio - loudness / balqso dichotomy and , by extension , on the difference between radio loud and radio quiet quasars , i initiated a program to observe 1556 + 3517 in the mid - ir with the iso satellite ( kessler et al .
one particular aim was to obtain a better estimate of the amount of dust on the line of sight to the qso and check that 1556 + 3517 remains radio - loud even after reddening correction .
another was to verify the prediction that lo - bal qso in general are heavily obscured .
more generally , it is important to measure the spectral energy distribution ( sed ) of balqso over as wide a frequency range as possible in order to compare it to that of non - bal qso in the hope of finding differences that hold clues to the origin of the bal phenomenon .
the quasar 1556 + 3517 was observed with the @xmath14 px long wavelength ( lw ) array of the isocam instrument ( cesarsky et al .
1996 ) on board the iso satellite ( kessler et al . 1996 ) .
the observation was executed nominally with the c01 observing mode ( aot ) on 14 march 1997 ( revolution 484 ) , starting at 06:17:10 ( ut ) and lasting 46 minutes and 26 second .
it consisted of a series of @xmath15 raster maps centered on the optical position of the qso ( r.a .
= 15h 56 m 33.8s , dec = 35@xmath16 17@xmath17 58@xmath18 , epoch = j2000 ) , with one map per filter .
the magnification selected was 3@xmath18 per pixel and the distance between two raster points was 6@xmath18 .
the axis of the rasters were oriented parallel to the celestial north and east .
the number of exposures per raster dwell point was 19 , with a unit integration time of 2.1 s per exposure and an amplifier gain of 2 .
the 6 filters used were , in sequential order : lw3 [ 1218 @xmath0 ] , lw2 [ 58.5 @xmath0 ] , lw9 [ 1416 @xmath0 ] , lw8 [ 10.712 @xmath0 ] , lw7 [ 8.510.7 @xmath0 ] and lw1 [ 45 @xmath0 ] , where the numbers in brackets refer to the wavelength range covered by each of the filter .
this sequence corresponds to decreasing pixel illuminations such that it minimizes the detector stabilization time .
the six monochromatic maps were processed with the cam interactive analysis ( cia ) software starting from the edited raw data ( erd ) .
the first few ( @xmath8 4 ) exposures of each frame were discarded so as to retain only those data for which the detector throughput had reached 90 % of its stabilisation level .
the remainder of the processing was standard and involved the usual steps , i.e. dark current subtraction , flat - fielding , removal of particle hits ( `` de - glitching '' ) and flux calibration . the source is clearly detected in all six maps at a very high significance level ( e.g. 63 sigma for the lw3 map ) .
the 1218 @xmath0 map is shown in fig.[map ] as an example .
the monochromatic intensities were computed by integration of the flux in a 6 px radius ( 18@xmath18 ) circle centered on the source and subtraction of the normalized background integrated in an annulus of 8 and 13 px internal and external external radii , respectively .
the results are displayed in table [ results ] .
the errors quoted derive mainly from the uncertainty in the calibration and background , the latter being induced by residual flat - fielding errors .
the full widths at half maximum ( fwhm ) of the source images were computed by fitting a two dimensional gaussian to the flux spatial distribution .
the results are listed in column 5 of table [ results ] . for comparison ,
the theoretical value of the full width at 41 % of the maximum intensity of a diffraction limited telescope of diameter d is @xmath20 , i.e. 0.343@xmath21 in the case of iso .
this yields e.g. 5.0 @xmath18 for the lw3 image , consistent with the value in table [ results ] . as a sanity check
, the lw3 images of point sources obtained with the same pixel magnification were retrieved from the set of iso calibration files and analysed in the same fashion , yielding fhwm=4.4 @xmath18 , in good agreement with the fwhm of 1556 + 3517 . for all six filters , the source width is consistent with that of a point source .
neither of the six maps show extended structures nor close - by objects at least down to the detection limits quoted in column 6 of table [ results ] .
these limits were obtained by computing the standard deviation of the background in areas of the map well outside the image of the source .
cccccc lw & @xmath22 & @xmath23 & flux & fwhm & limit + & ( @xmath0 ) & ( @xmath0 ) & ( mjy ) & ( @xmath18 ) & ( @xmath24 ) + 1 & 4.005.00 & 1.81 & @xmath25 & @xmath26 & 14 + 2 & 5.008.50 & 2.75 & @xmath27 & @xmath28 & 8 + 7 & 8.5010.7 & 3.90 & @xmath29 & @xmath30 & 11 + 8 & 10.712.0 & 4.56 & @xmath31 & @xmath32 & 20 + 3 & 12.018.0 & 5.85 & @xmath33 & @xmath34 & 8 + 9 & 14.016.0 & 6.01 & @xmath35 & @xmath36 & 18 +
the overall rest frame spectral energy distribution ( sed ) of 1556 + 3517 , from 1.5 ghz to the b band , is plotted in fig.[sed ] , where the optical and radio data are taken from becker et al .
all observed fluxes have been converted to rest frame luminosities @xmath37 ( in ergs@xmath38 ) assuming @xmath39 , @xmath40 and using the standard equations ( e.g. weedman 1986 ) : where l@xmath43 is the monochromatic luminosity per unit frequency in the rest frame of the qso , f@xmath44 is the monochromatic flux per unit frequency in the observer s frame , d is the distance to the qso and the other symbols have their usual meanings .
the sed of the quasar iras 14026 + 4341 is also shown in fig.[sed ] for comparison , the data being taken from low et al .
the comparison is illustrative since iras 14026 + 4341 is an iras discovered balqso with a large ir - to - optical luminosity ratio ( low et al .
it is immediately apparent that , despite similar ultraviolet luminosities , 1556 + 3517 is almost an order of magnitude more luminous in the near to mid - ir than 14026 + 4341 .
this presumably indicates the presence of a large amount of warm dust in 1556 + 3517 which causes a depression of the optical - uv continuum by reddening and an enhancement of the near and mid - ir flux because of grain thermal emission .
the 1556 + 3517 spectral index ( @xmath45 ) from the b band ( @xmath46 ) to the lw1 filter ( @xmath47 ) is @xmath48 . using the r band ( @xmath49 ) flux instead yields @xmath50
. both indices are significantly steeper than those of `` normal '' non - bal quasars .
note that only the hottest dust grains at @xmath51 k yield a significant contribution to the 1.81 @xmath0 flux .
since such a temperature is higher or equal to the grain sublimation temperature , relatively little dust thermal emission is expected at this wavelength .
hence , to a first approximation , the steep uv - to - near - ir spectral index of 1556 + 3517 is mostly a consequence of dust extinction . 1 .
the 1774.2 band lies in a region relatively free of any strong emission or absorption lines .
this is not the case of the 2823 band which is contaminated by the @xmath53 emission and absorption .
2 . more importantly , the 2823 band is definitely enhanced by balmer continuum ( bac ) emission and blended low - contrast feii emission lines which create a pseudo - continuum in quasar spectra often referred to as the `` small bump '' . 1 .
the `` intrinsic '' , i.e. un - reddened , optical - uv spectral index of 1556 + 3517 is -0.5 . within the uncertainty ,
this is equal to the mean uv spectral index @xmath54 of the sample of 33 high redshift ( z@xmath551 ) qsos of obrien , gondhalekar and wilson ( 1988 ) as well as to the mean optical spectral index @xmath56 of the large 227 qso sample of cheney and rowan - robinson ( 1981 ) or that obtained by neugebauer et al .
( 1987 ) from their sample of 104 quasars ( @xmath57 ) .
the reddening law is as parametrized by seaton ( 1979 ) in the optical - ultraviolet range and as tabulated by rieke and lebofsky ( 1985 ) from 1 to 13@xmath0 . where @xmath60 $ ] is the ( rest ) wavelength range over which the spectral index is measured ( 0.1771.810 @xmath0 in the present case ) and the function @xmath61 is the reddening law at wavelength @xmath10 .
this yields a visual extinction @xmath62 magnitudes which implies that the continuum is reddened by 4.01 magnitudes at a rest wavelength of 1774 but only 0.256 magnitudes at 1.81 @xmath0 .
note that the foreground galactic extinction in the direction of 1556 + 3517 ( @xmath63 ) is only 0.05 v magnitudes and can be neglected .
after correction for extinction , the radio to optical flux ratio becomes @xmath64 if the b band data are used to derive the flux at @xmath65 ( k correction ) and @xmath66 if one uses the r flux instead .
the two estimates differ at the 1.1 sigma level only .
i therefore adopt the mean of the two @xmath67 .
such a ratio is less than 1.3 sigma away from the @xmath3 dividing line between radio - loud and radio - quiet quasars .
one therefore concludes that the case for 1556 + 3517 being radio - loud is marginal , at best .
it is also worth pointing - out that de - reddening brings 1556 + 3517 close to the @xmath3 line where data suggests that there is an apparent excess of balqsos over non - balqsos ( francis , hooper and impey 1993 ) .
the near to mid - ir luminosity @xmath68 of 1556 + 3517 was computed by integrating all the de - reddened rest - frame emission over the range @xmath69 to @xmath70 after subtraction of an -0.5 index power - law matching the de - reddened b band and lw1 band luminosities : @xmath71
. one can use @xmath68 to estimate the mass of dust responsible for near to mid - ir emission in 1556 + 3517 .
the mean rest wavelength of the mir emission corresponds to a black - body temperature of @xmath8740 k. under the assumptions that the dust grains have a mean size of 0.1 @xmath0 and radiate like black - bodies and following rudy and puetter ( 1982 ) , i obtain a dust mass @xmath72 .
this should in fact be seen as a lower limit since actual grains are likely to emit less efficiently than black - body thereby requiring an even larger mass to produce the observed mir luminosity . for a normal galactic dust to gas mass ratio of 200 , this implies a total mass of gas of about @xmath7 .
the color temperature of the dust is higher than that of typical hii regions which suggests that it is closely associated to the active nucleus .
it is therefore legitimate to ask whether the dusty gas responsible for the mir emission could be the bal wind itself . in the disk wind model of murray
et al . ( 1995 ) , the bal region covering factor is about 10 % , its column density @xmath73 and its inner radius is of the order of @xmath74 cm .
this yields a bal mass @xmath75 , more than 4 orders of magnitude smaller than the above estimate .
of course , there is more mass available as the wind will accumulate material over the years . for an axially symmetrical wind
, the mass flow rate is given by : @xmath76 where , @xmath77 is the solid angle sustained by the wind as seen from the central source , @xmath78 its column density in units of @xmath79 , @xmath80 its radius in units of @xmath81 , and @xmath82 its velocity in units of @xmath83 . assuming parameters value as in murray et al s model yields @xmath84 .
this implies that the bal phenomenon should have lasted at least @xmath8 250,000 years to accumulate the mir emitting gas .
while such a value in itself is not unreasonable , there are problems with this scenario : * on the one hand , the dust must somehow `` see '' the central uv source in order to retain a temperature of @xmath8 700 k for such a long time . however , given the mass involved and the geometry of the wind , the opacity will be extremely high and the bulk of the dust and gas will be ( self-)shielded from the central source . * this gas would presumably be at least partially ionized .
since the total column would far exceed @xmath85 , the gas would be opaque to free - free absorption and the continuum totally suppressed , contrary to what is observed .
it therefore seems unlikely that the observed mir emission originates from the wind itself .
a more plausible origin would be e.g. , the molecular torus , the narrow line region , or a recent burst of star formation .
it is a pleasure to thank the time allocation committee for awarding discretionary iso observing time to this project .
i am also grateful to the referee , dr .
ray weymann , for valuable comments which improved the manuscript . | to date , the quasar 1556 + 3517 is the only radio loud broad absorption line qso ( balqso ; becker et al . 1997 ) .
this prompted narrow band filter imaging observations in the range 415 @xmath0 with the iso satellite .
the source is clearly detected in all filters and appears point - like and isolated at the resolution of iso .
the overall spectral energy distribution ( sed ) in @xmath1 peaks at a rest wavelength @xmath2 and is reddened by 1.6 visual magnitudes . correction for reddening brings 1556 + 3517 within 1.3 sigma from the @xmath3 dividing line between radio - loud and radio - quiet sources .
the mid - ir luminosity integrated over the 1.86 @xmath0 rest wavelength range is @xmath4 ( @xmath5 ) , which requires at least @xmath6 of dust and @xmath7 of associated gas .
it is unlikely that such a large mass stems from the bal wind itself . |
the generally accepted distance modulus to the large magellanic cloud ( lmc ) is @xmath4mag ( for recent discussion see westerlund 1997 , madore & freedman 1998 ) . however , there is a long standing @xmath5mag discrepancy between the `` long '' distance determined using cepheids ( e.g. laney & stobie 1994 ) and the `` short '' distance determined using rr lyr stars ( e.g. walker 1992 , layden et al .
a similar discrepancy is present in the distance to the lmc derived with the supernova sn1987a ( @xmath6mag , gould & uza 1998 ; @xmath7mag , panagia et al . 1997 ) .
recently udalski et al . (
1998 ) used red clump stars observed in the lmc by the ogle 2 project ( udalski et al .
1997 ) and obtained a value of @xmath8mag ( statistical plus systematic error ) .
this distance modulus is @xmath2mag smaller than the `` long '' distance modulus used , for example , by the _ hst _ extragalactic distance scale key project team ( e.g. rawson et al .
1997 and references therein ) . because errors in the distance to the lmc can propagate into errors in such key quantities as distances , luminosities , masses , and sizes of extragalactic objects , it is important to check the result of udalski et al .
( 1998 ) using independent data , in order to investigate possible systematic errors .
red clump stars are the metal rich equivalent of the better known horizontal branch stars , and theoretical models predict that their absolute luminosity only weakly depends on their age and chemical composition ( seidel , demarque , & weinberg 1987 ; castellani , chieffi , & straniero 1992 ; jimenez , flynn , & kotoneva 1998 ) . indeed , the absolute magnitude - color diagram from _ hipparcos _ data ( perryman et al .
1997 , their figure 3 ) clearly shows a compact red clump the variance in the @xmath9-band magnitude is only @xmath10mag ( stanek & garnavich 1998 ; udalski et al .
1998 ) . despite their large number and the theoretical understanding of their evolution ,
red clump stars have seldom been used as distance indicators . however , stanek ( 1995 ) and stanek et al .
( 1994 , 1997 ) used these stars to map the galactic bar .
paczyski & stanek ( 1998 ) used the red clump stars observed by the ogle project ( udalski et al .
1993 ) to obtain the distance to the galactic center .
stanek & garnavich ( 1998 ) used red clump stars observed by the _ hst _ in m31 to obtain a one - step distance to this galaxy . in this paper
we follow the approach of paczyski & stanek ( 1998 ) and present an estimate of the distance to the lmc based on the comparison between the red clump giants observed locally by the _ hipparcos _
( perryman et al . 1997 ) satellite and those observed in the lmc by the @xmath11 digital photometric survey of the magellanic clouds ( zaritsky et al .
1997 ) . in section 2
we describe the data used in this paper and select low extinction regions for further analysis . in section 3
we analyze the red clump distribution in the lmc and derive the distance to this galaxy . in section 4 we discuss the possible reasons for the discrepancy with the cepheid distance to the lmc and how it can be resolved .
zaritsky , harris & thompson ( 1997 ) have undertaken a large scale @xmath11 photometric ccd survey of the magellanic clouds , with the ultimate goal of imaging the central @xmath12 of the lmc and @xmath13 of the smc .
the initial results for a @xmath14 region were presented by zaritsky et al .
( 1997 ) , and the extinction map for the same region was constructed by harris , zaritsky & thompson ( 1997 ) .
the area of the lmc observed by zaritsky et al .
( 1997 ) is shown in figure [ fig : lmc ] , along with the four fields used by udalski et al .
( 1998 ) to determine the red clump distance to the lmc . as noted by udalski et al .
( 1998 ) , uncertainties in the extinction estimates are the largest contributor to their systematic error .
we therefore select low reddening regions from the area observed by zaritsky et al .
harris et al .
( 1997 ) used @xmath15 ob main sequence stars to construct a map of the reddening in the region observed by zaritsky et al .
they find a mean reddening of @xmath16mag , with a non - gaussian tail to high values . as discussed by harris et al .
( 1997 ) , it is possible that the reddening map is biased to high reddening values , since it is based on the reddening toward ob stars , a population that may reside in dustier - than - average regions of the ism . on the other hand , the nature of their interpolation may partially counter this bias by smoothing over reddening spikes .
recently schlegel , finkbeiner & davis ( 1998 ; hereafter : sfd ) published a new all - sky reddening map , based on the cobe / dirbe and iras / issa maps . with its high spatial accuracy , the sfd map might be potentially useful for selecting low - reddening regions in the magellanic clouds . however , as discussed by sfd , the lmc , smc and m31 are not removed from their map , nor are sources within their holmberg radius .
the reddening estimates from the sfd map are unreliable within these objects because of the lack of spatial temperature resolution from dirbe and confusion with internal ir sources . however , because neither of these two effects artificially creates large areas of low redenning in the map , the map can be used as a guide to select regions of low total extinction .
if the total column density of dust is low , then the dispersion of reddening values of a population distributed along the line - of - sight will also be low .
we therefore decided to select regions of low dirbe / iras reddening from the area observed by zaritsky et al .
( 1997 ) and then determine the reddening using the map of harris et al .
( 1997 ) . examining the sfd map in the region observed by zaritsky et al .
( 1997 ) we selected two circular regions with @xmath17 radii ( defined by the low extinction region ) , which we hereafter call a1 and a2 .
these regions are centered on ( ra , dec)@xmath18 and ( ra , dec)@xmath19 , respectively , and are shown as small circles in figure [ fig : lmc ] .
the sfd map gives an average reddening of @xmath20mag for each of these regions .
the harris et al .
( 1997 ) map gives @xmath21 for the region a1 and @xmath22 for the region a2 .
we find the good agreement between the maps reassuring .
however , we assume a conservative error of the @xmath23 reddening to be @xmath24mag , following the discussion by harris et al .
this leads to the values of the extinction @xmath25mag and reddening @xmath26mag for the region a1 and @xmath27mag and @xmath28mag for the region a2 . in figure
[ fig : cmd ] we show the red clump dominated parts of the @xmath29 color - magnitude diagrams ( cmds ) for both regions a1 and a2 , corrected for the extinction and the reddening using the above values of @xmath30 and @xmath31 .
the dashed rectangle corresponds to the region of the cmd selected for comparison with the local red clump stars observed by _
( see the next section ) .
the lmc red clump is clearly bluer than the local one ( paczyski & stanek 1998 , their figure 2 ) , indicating a lower average metallicity of the lmc ( see jimenez et al .
1998 , their figure 5 ) . however , there is a sufficient overlap with the _ hipparcos _ color range to allow for meaningful comparison , which we perform in the next section .
following paczyski & stanek ( 1998 ) , we selected the red clump stars in the color range @xmath32 and the magnitude range @xmath33 in the a1 region ( 1725 stars ) and in the a2 region ( 1273 stars ) .
the color range was selected to correspond to the color range of the red clump stars observed locally by the _ hipparcos _ ( paczyski & stanek 1998 , their figure 2 ) .
following garnavich & stanek ( 1998 ) , we fitted both distributions with a function @xmath34.\ ] ] the first three terms describe a fit to the `` background '' distribution of the red giant stars , and the gaussian term represents a fit to the red clump itself .
@xmath35 corresponds to the peak magnitude of the red clump population .
we obtained the values of @xmath36 for the a1 region and @xmath37 for the a2 region .
the distribution of the lmc red clump stars as a function of their @xmath38 magnitude is shown in figure [ fig : dist ] along with the fitting function described by the eq.1 .
the gaussian fitted to the a1 field red clump distribution has a smaller dispersion , @xmath39mag , than the gaussian fitted to the a2 red clump , @xmath40mag .
the red clump is less pronounced , and has a correspondingly larger dispersion , in the a2 field because the red clump is bluer and mostly falls outside the color cut .
indeed , when we select the red clump stars within the color range @xmath41 , both fitted distributions have the same @xmath42mag .
this different color selection also moves the red clump peaks @xmath35 to slightly fainter magnitudes , @xmath43 for the a1 region ( 1907 stars ) and @xmath44 ( 2497 stars ) for the a2 region .
this slight color dependence in the @xmath9-band magnitudes of the red clump stars is consistent with the results of paczyski & stanek ( 1998 ) , stanek & garnavich ( 1998 ) & udalski et al .
( 1998 ) .
we now proceed to obtain the lmc distance modulus using the red clump , by assuming that the absolute @xmath9-band brightness of the red clump stars is the same for the local stars observed by _
hipparcos _ and those in the lmc .
the straight average of the red clump peak apparent magnitudes @xmath35 for the two regions is @xmath45 and the weighted mean is @xmath46 . combining @xmath47 with the distribution of local red clump stars , which have @xmath48 ( stanek & garnavich 1998 )
, we obtain the distance modulus for the lmc , @xmath49mag , or @xmath50 ( statistical error only ) . after adding the systematic error of @xmath51mag due to the uncertainty in the @xmath30 determination , and @xmath24mag due to the zero - point uncertainty in the @xmath9-band photometry ( zaritsky et al .
1997 ) , we arrive at the final value of @xmath52mag ( statistical plus systematic error ) .
this is indistinguishable from the value of @xmath53mag determined by udalski et al .
( 1998 ) , but @xmath2mag below generally accepted value of @xmath54mag ( madore & freedman 1998 ) .
we discuss possible reasons for this discrepancy and how it can be resolved in the next section .
as with all distance - ladder techniques , our analysis includes the assumption that the calibrating and target objects being compared are intrinsically similar . in our red clump analysis , this assumption is manifested by the assertion that the @xmath9-band brightness of red clump stars is independent of the age , chemical composition , and mass differences that may exist between the red clump stars near the sun and those in the lmc .
indeed , the lmc red clump is systematically bluer than the local one , indicating the somewhat different properties of these stars .
however , paczyski & stanek ( 1998 ) , stanek & garnavich ( 1998 ) and udalski et al .
( 1998 ) found that the @xmath9-band peak magnitude of the red clump depends very weakly on their @xmath55 color in the range @xmath56 , and therefore is independent of the metallicity ( jimenez et al .
this is confirmed in this paper as well , by comparing the peak brightness of the red clump for two color ranges @xmath57 and @xmath58 ( see the previous section ) .
the fact that the observed red clump distributions are so narrow ( @xmath59mag ) indicates that the age dependence of the red clump @xmath9-band peak luminosity is also small ( @xmath60mag ) . otherwise , in a system with a complex star formation history , such as the lmc ( holtzman et al .
1997 ; geha et al . 1998 ) , the resulting red clump should have considerable width .
stanek & garnavich ( 1998 ) compared three different lines - of - sight that probe a large range of m31 galactocentric distances and locations , and hence a range of metallicities and possibly ages and star formation histories .
the fact that the derived distance moduli for their three fields varied by only @xmath61mag indicates that the red clump is a potentially stable standard candle .
the mostly empirical support for using the red clump stars as a distance indicator should also be verified using modern theory of the stellar structure and evolution . in particular ,
@xmath9-band predictions are seldom given by such theoretical calculations .
so why does the red clump distance to the lmc disagree with the cepheid distance ( madore & freedman 1998 ) ? as usual , there are several possible answers .
contrary to our arguments given above , there might still be something `` unusual '' about the red clump population in the lmc .
although the red clump and cepheid distances to the lmc are discrepant , the distances to m31 derived from the two methods are in excellent agreement ( @xmath62 from stanek & garnavich 1998 and 24.44@xmath63 from freedman & madore 1990 ) .
another possibility is that the cepheid distance to the lmc is simply poorly determined , as there are few cepheids with well determined parallaxes in the _ hipparcos _ catalog . in their recent study , madore & freedman ( 1998 )
find @xmath64mag , from a sample of 19 cepheids observed by _ hipparcos _ with good @xmath65 data , and @xmath66mag , from a sample of only 7 cepheids with good @xmath67 data .
yet a third possibility , as discussed by madore & freedman ( 1998 ) , is that there are other effects on the cepheid pl relation ( e.g. extinction , metallicity , and statistical errors ) , which are as significant as any reassessment of the zero point based on _
hipparcos_. the metallicity effect on the cepheid pl relation , determined by kennicutt et al .
( 1998 ) ( @xmath68 = -0.24 \pm 0.16 $ ] ) , reduces the discrepancy between the red clump and cepheid distances to the lmc by @xmath69mag , while the somewhat larger metallicity dependence found by sasselov et al .
( 1997 ) and kochanek ( 1997 ) reduces it by @xmath70mag . to illustrate the effect of the assumed reddening on the derived distance modulus
, we note that the value of the lmc distance modulus , @xmath71mag , derived recently by salaris & cassisi ( 1998 ) and based on the @xmath72-band brightness of the rr lyr stars , becomes @xmath73mag if their assumed reddening of @xmath74mag is increased to @xmath75mag , corresponding to the mean reddening found by harris et al .
it is disturbing that the distance to a key calibrator of the entire distance scale is uncertain by up to 20% . as described by udalski et al .
( 1998 ) , the @xmath2mag discrepancy between their ( and now our as well ) `` short '' distance to the lmc and the `` long '' distance to the lmc from the cepheids can be resolved by using detached eclipsing binaries as a direct distance indicator ( paczyski 1997 ) .
the cepheids in the lmc can also be used to get a direct distance estimate through a modified baade - wesselink method ( e.g. krockenberger 1996 ; krockenberger , sasselov , & noyes 1997 ) .
both these methods require no intermediate steps in the distance ladder , therefore avoiding the propagation of errors usually crippling the distance scale . with the 6.58 meter telescopes now being built in the southern hemisphere
the necessary spectroscopy of the detached eclipsing binaries and cepheids can be quite easily obtained for these 1418 magnitude stars .
it is worth mentioning here that the effort to obtain direct distances with the detached eclipsing binaries and cepheids to the m31 and m33 galaxies is already under way and the first results look promising ( project direct : kaluzny et al .
1998 , stanek et al .
1998 , krockenberger et al .
1998 , sasselov et al . 1998 ) . to summarize , among the various stellar distance indicators the red clump giants might be the best for determining the distance to the lmc and other nearby galaxies because there are so many red clump stars . in particular , _
hipparcos _ provided accurate distance determinations for almost 2,000 such stars , but unfortunately @xmath9-band photometry is available for only @xmath76 of them , so it is important to obtain @xmath9-band photometry for all _ hipparcos _ red clump giants .
we also need to test the metallicity dependence of _ hipparcos _ red clump giant absolute luminosities and colors .
there are many stars within @xmath77 of the sun for which very high - resolution spectroscopy is possible . | following an approach developed by paczyski & stanek , we derive a distance to the large magellanic cloud ( lmc ) by comparing red clump stars from the _ hipparcos _ catalog with the red clump stars observed in two fields in the lmc that were selected from the ongoing photometric survey of the magellanic clouds to lie in low extinction regions .
the use of red clump stars allows a single step determination of the distance modulus to the lmc , @xmath0mag ( statistical plus systematic error ) , and the corresponding distance , @xmath1 .
this measurement is in excellent agreement with the recent determination by udalski et al .
, also based on the red clump stars , but is @xmath2mag smaller than the generally accepted value of @xmath3mag .
we discuss possible reasons for this discrepancy and how it can be resolved . |
it is widely accepted that mechanical feedback from active galactic nuclei ( agns ) plays a key role in the evolution of galaxy clusters , suppressing cooling flows and the associated fast growth of central galaxies ( e.g. , @xcite ; @xcite ) .
one of the most compelling evidences comes from mounting detections of surface brightness depressions in x - ray images of galaxy clusters , often referred as x - ray cavities " or agn bubbles " .
many cavities are associated with radio jets and spatially coincident with radio lobes ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ) , confirming the common note that they are evolved from the interaction of agn jets with the intracluster medium ( icm ) .
taking the agn jet origin for x - ray cavities , they are widely used to estimate the energetics of mechanical agn feedback . while travelling through the icm , the radio jets displace and
do @xmath5 work on the surrounding hot gas , producing x - ray cavities .
the total energy required to create a cavity is around @xmath6 where @xmath7 is the pressure of the gas surrounding the cavity , @xmath8 is the cavity s volume , and @xmath9 is the ratio of specific heats of the gas inside the cavity ( @xmath10 for relativistic gas ) .
as these quantities are relatively well understood , this method is currently the most reliable to measure the non - radiative energy output from agn jets . in this paper , we argue for the first time that in addition to the cavity s volume ( @xmath8 in equation 1 ) , its shape is also important .
x - ray cavities are produced by the interaction of agn jets with the surrounding gas , and subsequently rise in the icm .
thus the cavity shape potentially encodes important information about the properties of agn jets and the icm .
the icm viscosity significantly affects the long - term evolution of x - ray cavities , but as we will demonstrate in this paper , it does not appreciably affect the shape of young cavities , which may be used to probe the properties of agn jets .
agn feedback events may be triggered as supermassive black holes ( smbhs ) accrete hot or cold gas from various origins @xcite .
the hot icm is the dominant baryon component in galaxy clusters , and cold gas has also been observed in central regions of many cool core clusters ( e.g. , @xcite ) .
furthermore , the accretion of hot gas is directly affected by its heating and cooling balance , and if cooling dominates over heating intermittently ( see @xcite for more discussions ) , the accretion of cooling hot gas alone could be in either hot or cold mode , as seen in both steady state models ( @xcite ; @xcite ) and hydrodynamic simulations @xcite . thus due to potential variations in accretion modes ,
the properties of agn jets may vary substantially in different systems , leaving signatures in the shapes of x - ray cavities .
the observed cavities are the line - of - sight projected x - ray surface brightness depressions on the sky . during the last decade ,
a large number of x - ray cavities have been detected by _
chandra _ and _ xmm - newton _ x - ray space telescopes ( see @xcite for a review ) . while some cavities are nearly circular , many cavities are not , as seen in the cavity samples of @xcite and @xcite . to evaluate the cavity volume used in equation ( [ eq1 ] ) , observers often approximate x - ray cavities as ellipses ( ellipsoids in 3-dimensional space ) .
almost all non - circular cavities in @xcite and @xcite are elongated along either the jet direction ( defined as the radial direction from the cluster center to the cavity center ) or the perpendicular direction ( defined as the direction perpendicular to the jet axis ) , suggesting that x - ray cavities are not subject to significant rotation during their evolution in the icm .
otherwise , a large number of the cavities would be elongated along random directions with respect to the jet direction . in this paper
, we study the shape of x - ray cavities in hydrodynamic simulations , and particularly investigate the effects of jet properties and viscosity on the cavity shape .
our primary aim is to find some general trends connecting the cavity shape with some fundamental jet properties and the icm viscosity , motivating future observational studies on this important topic .
we describe our numerical setup in section 2 and present the numerical results in section 3 .
we conclude and discuss our results in section [ section : discussion ] .
we characterize the cavity shape for the first time with two geometrical parameters @xmath0 and @xmath1 .
radial elongation @xmath11 , defined as the ratio of the cavity axis along the jet direction ( @xmath12 ) to that along the perpendicular direction ( @xmath13 ) , is a measure of the cavity elongation ( the direction of the major axis ) with respect to the jet axis .
the value of @xmath0 can be used to classify cavities into two types : type i elongated along the perpendicular direction ( @xmath14 ) and type ii elongated along the jet direction ( @xmath15 ) .
the second parameter , top wideness @xmath1 , is a parameter to describe the deviation of a cavity s shape from a perfect ellipse by looking at the variation of its perpendicular size along the jet axis .
the value of @xmath1 is defined as the relative location of the cavity s widest perpendicular size in its axis along the jet direction from the cavity bottom , @xmath16 , where @xmath17 is the distance from the cavity bottom to the location of its widest perpendicular size in the jet axis .
the value of @xmath1 further separates each type of cavities into three subgroups : top - wide ( @xmath18 ) , center - wide ( @xmath19 ) , and bottom - wide ( @xmath20 ) . a simple sketch of two representative types of cavities is shown in figure [ plot1 ] .
we consider the formation and evolution of x - ray cavities in the hot intracluster medium by a pair of bipolar jets released from the cluster center . including viscosity , the basic hydrodynamic equations governing the gas evolution in the three - dimensional eulerian space may be written as @xmath21 @xmath22 @xmath23 + where @xmath24 is the lagrangian time derivative , * @xmath25 * is the viscous stress tensor @xmath26 @xmath27 is the dynamic viscosity coefficient , @xmath28 is the gas density , @xmath29 is the gas pressure , @xmath30 is the gas energy density , @xmath31 is the gas velocity , and @xmath32 is the gravitational potential .
we adopt the ideal gas law for the hot gas and the gas pressure and energy density are related via @xmath33 , where @xmath34 .
the molecular weight is assumed to be @xmath35 .
we ignore radiative cooling , which is unimportant during the short jet duration ( @xmath36 myr ) in our simulations and may be offset by heating sources ( e.g. , agn feedback or thermal conduction ) on longer timescales .
the simplest picture of the cavity formation contains a rotational symmetry with respect to the jet axis .
we thus solve the above equations in @xmath37 cylindrical coordinates using a two - dimensional eulerian code similar to zeus 2d @xcite .
the code has also been successfully used in many previous studies , e.g. , @xcite , @xcite , @xcite .
our numerical implementation of shear viscosity has been described in detail in the appendix of @xcite .
the computational grid consists of @xmath38 equally spaced zones in both coordinates out to @xmath39 kpc plus additional @xmath39 logarithmically - spaced zones out to @xmath40 mpc .
we have repeated several simulations with double resolution , and find that the shapes of young cavities converge quite well .
the long - term evolution of cavities converges relatively well in viscous runs , but in non - viscous runs , old cavities are more disrupted in higher - resolution simulations , as discussed in more detail in @xcite .
we adopt zero - gradient " boundary conditions at the outer boundary and reflective boundary conditions at the inner boundary .
our method is generally applicable to all reasonably relaxed clusters , but for concreteness , we adopt simulation parameters appropriate for the well - observed nearby system m87 and its halo within the virgo cluster .
the initial radial profiles of gas density and temperature are taken from the analytic fits to the observations by @xcite .
the static gravitational field is determined to establish exact hydrostatic equilibrium for the initial gas pressure profile .
the same initial cluster setup has been used in @xcite , where the readers are referred to for more details . in section [ section :
perseus ] , we applied our method to a more massive cluster perseus , and tested the robustness of our results .
the jet is introduced along the @xmath41-axis with the radius @xmath42 and an initial opening angle of @xmath43 degree .
we initialize the jet at @xmath44 by adding gas fluxes of mass , energy , and momentum corresponding to a uniform jet with density @xmath45 , energy density @xmath46 and velocity @xmath47 .
real agn jets on sub - kiloparsec scales ( @xmath48 kpc ) often move at relativistic velocities and are too narrow to be resolved by our numerical simulations .
thus we initialize jets in our simulations at @xmath49 kpc , and explore the dependence of the resulting cavity shape on @xmath42 , @xmath47 , and other jet parameters .
similar off - center jet injection methods have also been adopted in previous studies ( e.g. , @xcite ; @xcite ) .
although real agn jets may contain internal structures , our uniform jet setup allows us to study the connection of the cavity shape with simple jet parameters in great details .
the jet density and energy density can be normalized to the ambient icm gas density @xmath50 and energy density @xmath51 at the jet base ( here @xmath52 g/@xmath53 and @xmath54 erg/@xmath53 at @xmath44 along the jet axis ) by a density contrast @xmath55 and energy density contrast @xmath56 , respectively .
an important jet parameter affecting its evolution is the internal mach number at the jet base , which is defined as @xmath57 .
the jet is active for a duration of @xmath58 and is turned off afterwards . to explore the parameter space , we presented in this paper a series of fourteen representative simulations , four of which include nonzero shear viscosity ( @xmath27 ; see section 3.2 ) .
the jet in each simulation injects both thermal and kinetic energies into the icm .
the kinetic power can be written as @xmath59 , and the thermal power is @xmath60 .
the jet parameters , energetics , and the shapes of resulting young cavities ( values of @xmath0 and @xmath1 ) in all our simulations are listed in table 1 . [ cols="<,^,^,^,^,^,^,^,^,^,^,^ " , ]
the shape of x - ray cavities is determined by the properties of the jet and the icm . in this paper
we focus on the dependence of the cavity shape on the jet properties ( jet parameters listed in table 1 ) and the icm viscosity .
after the jet was turned off , the cavity rises away from the cluster center due to buoyancy and possibly the momentum injected by the original jet . at this late stage ,
the cavity shape still changes with time . in section 3.1 ,
we focus on the shape of young cavities in our non - viscous simulations , which is directly related to the jet parameters .
we then study the shape evolution of x - ray cavities in the icm in section 3.2 , where the role of shear viscosity is also explored .
we note that the cavity shape discussed in our hydrodynamic simulations is intrinsic " . for
real observed cavities , the projection effect may make them appear more circular ( i.e. , with the value of @xmath0 closer to 1 ) , as also discussed in @xcite .
x - ray cavities are produced by agn jets .
we consider young cavities to be those that have not significantly risen away from their initial locations where they are created .
in particular , here we study the shape of young x - ray cavities in our simulations at @xmath61 myr , shortly after the jets are turned off at @xmath62 .
we do not study the early evolution of agn jets and radio lobes , which have been explored in some previous studies ( e.g. , @xcite ; @xcite ) .
we first investigate how the cavity shape depends on the jet density and energy density .
figure [ plot2 ] shows central slices of gas density in a series of four simulations at @xmath61 myr with varying values of the initial jet density @xmath45 or energy density @xmath46 . in each panel , clearly seen are a cavity ( a 10@xmath6320-kpc - size region with relatively lower gas densities compared to the ambient value ) and a surrounding bow shock ( the outer discontinuity annotated in each panel ) produced by the agn jet event .
due to the high sound speed in the hot icm , the bow shock is usually weak , with the mach number in our fiducial run ( run r1 ) dropping from @xmath64 at @xmath65 myr to @xmath66 at @xmath67 myr .
the jet in run r1 is very light with density contrast @xmath68 , and is over - pressured with respect to the ambient gas by a factor of @xmath69 .
the r1 jet is internally subsonic with @xmath70 .
the resulting cavity ( the top left panel ) is elongated more significantly along the perpendicular direction than the jet direction with the value of @xmath71 .
however , the cavity is not exactly an ellipse .
if one looks at its perpendicular size @xmath72 measured along the direction perpendicular to the jet axis , it is clear that @xmath73 drops with the distance @xmath74 from the cluster center along the jet axis .
the value of top wideness is @xmath75 , corresponding to a bottom - wide " type - i cavity in our classification as shown in figure 1 .
this is an important feature , which could be used to disentangle different jet parameters for cavities with similar values of @xmath0 as further discussed below .
the very light internally - subsonic jet in run r2 also produces a similar bottom - wide type - i cavity ( the bottom left panel ) , which is much smaller than the r1 cavity due to the much less powerful jet in run r2 . in right panels ,
the jets in runs r3 and r4 are internally supersonic with @xmath76 and @xmath77 respectively , and are more massive ( @xmath78 ) , resulting in cavities with higher values of radial elongation ( @xmath79 and @xmath80 respectively ) than in runs r1 and r2 .
the r3 cavity is type - i and center - wide ( @xmath81 ) .
notably , the r4 cavity is elongated along the jet direction ( a type - ii cavity ) and its perpendicular size roughly increases with the distance from the cluster center ( @xmath82 ; top - wide " ) .
thus figure [ plot2 ] indicates that bottom - wide " type - i cavities can be produced by very light internally - subsonic jets , while top - wide " type - ii cavities can be produced by heavier jets with internally supersonic velocities . in figure [ plot3 ]
, we show the synthetic x - ray surface brightness maps of two typical jet simulations ( runs r1 & r4 ) at @xmath61 myr .
the x - ray surface brightness is computed as the line of sight projection of the cooling rate @xmath83 along a direction perpendicular to the jet axis . here
@xmath84 is the electron number density , @xmath85 is the ion number density , and @xmath86 is the @xcite cooling function at the gas metallicity @xmath87 .
type i and ii cavities are clearly seen in the left ( run r1 ) and right ( run r4 ) panels , respectively . the internally - subsonic jet in run r1 is energetically dominated by the thermal power , and the internally - supersonic jet in run r4 is energetically dominated by the kinetic power ( see table 1 ) .
type - i cavities can also be produced by very light jets energetically dominated by cosmic rays , as seen in @xcite .
we then study how the cavity shape depends on the jet velocity in runs rv1 and rv2 , as shown in figure [ plot4 ] .
the jet parameters in these two runs are identical to those in run r2 except for higher values of the jet velocity . comparing the bottom left panel of figure [ plot2 ] and figure [ plot4 ] ,
it is clear that the cavity becomes more radially elongated ( i.e. , the value of @xmath0 increases ) as the jet velocity increases .
although type ii cavities can be created by very light jets with high jet velocities ( e.g. , run rv2 with @xmath68 , @xmath88 cm / s , and @xmath89 ) , these cavities are not top wide " as those created by heavier jets ( e.g. , in run r4 with @xmath78 ) . in fact , the cavities produced by very light internally - supersonic jets in runs rv1 and rv2 are close to cylindrical " , as also found in @xcite and a relativistic simulation in @xcite .
in homogenous environments , very light internally - supersonic jets even produce bottom - wide cavities , as shown in @xcite .
the cavity shape also depends on the jet radius @xmath42 at the jet base , which is explored in runs rr1 and rr2 shown in figure [ plot5 ] .
the jet parameters in these two runs are identical to those in our fiducial run r1 except that @xmath42 is changed to @xmath90 and @xmath91 kpc in runs rr1 and rr2 respectively .
the very light jets ( @xmath68 ) in these two runs also produce bottom - wide " cavities as in run r1 , and the cavity becomes more elongated along the perpendicular direction ( i.e. , the value of @xmath0 decreases ) as the jet radius increases .
the dependence of the cavity shape on the jet duration @xmath58 is investigated in runs rt1 and rt2 , which have the same jet parameters as in run r1 except that @xmath58 is increased to @xmath92 and @xmath93 myr in runs rt1 and rt2 respectively .
as shown in figure [ plot6 ] , the cavity shape in run rt1 is similar to that in run r1 , but the cavity in run rt2 is elongated along the jet direction with a much larger radial elongation @xmath0 than in run r1 .
thus the very light jets in these two runs also produce bottom - wide " cavities as in run r1 , but the cavity becomes more radially elongated as the jet duration increases . in this section
, we presented a suite of 10 simulations , showing that the shape of young x - ray cavities is significantly affected by various jet parameters , including @xmath45 , @xmath46 , @xmath47 , @xmath42 and @xmath58 .
our simulations explicitly demonstrated that it is feasible to produce both type - i and -ii cavities by agn jets and found some general trends .
very light jets ( e.g. , @xmath68 in run r1 ) produce very strong backflows , which transport jet materials away from jet heads .
the back - flowed jet materials tend to accumulate near the bottom of cavities and expand laterally
. if jets are very - light and internally subsonic , the forward motions of jet heads are relatively slow and the cavities lateral expansion is often significant enough to produce bottom - wide cavities elongated along the perpendicular direction ( type i ) .
if these jets are active for a very long duration , they tend to create bottom - wide cavities elongated along the jet direction ( type ii ) . on the other hand ,
if the velocity of a very light jet is very large ( internally supersonic ) , the jet head s motion tends to be comparable to or even more important than the cavity s lateral expansion .
therefore , the resulting cavity may also be elongated along the jet direction , but it appears more cylindrical than bottom - wide . for heavier jets ( e.g. , the light jet with @xmath78 in run r4 ) , the backflows are less strong , and the jet heads moves relatively faster , resulting in cavities with larger values of @xmath0 and @xmath1 ( e.g. , center - wide or top - wide ) . in summary
, bottom - wide type - i cavities are produced by very light internally - subsonic jets , while top - wide type - ii cavities are produced by heavier jets with internally supersonic velocities , which may also produce center - wide cavities with @xmath4 if the jets are slightly supersonic ( @xmath94-@xmath91 ) . in our simulations
, we adopt very light jets with @xmath68 and heavier jets with @xmath78 .
a general threshold @xmath95 may be used to distinguish between very light ( @xmath96 ) and heavier ( @xmath97 ) jets .
bottom - wide type - ii cavities can be produced by very light jets with very long durations ( @xmath58 ) .
cylindrical cavities ( both type i and ii ) are produced by very light internally - supersonic jets , and their radial elongations increase ( from type - i to type - ii cavities ) with the jet s internal mach number @xmath98 .
we study the late evolution of x - ray cavities and the role of shear viscosity in this subsection .
after the jet is turned off , the cavity moves away from the cluster center ( i.e. , rises in the icm " ) due to the combined effects of buoyancy and the original momentum injected by the jet . during this process , the cavity shape , and in particular its radial elongation @xmath0 and top wideness @xmath1 , may evolve with time .
however , it is not easy to accurately predict the shape evolution of cavities in this process , as they are subject to the disruption by rayleigh - taylor ( rt ) and kelvin - helmholtz ( kh ) instabilities developed at the cavity surface .
this can be seen in the left panels of figures [ plot7 ] and [ plot8 ] for runs r1 and r4 respectively , as also studied in @xcite for initially static cavities . to keep the cavity intact ( such as the outer cavities in the perseus cluster observed by @xcite ;
also see @xcite for another potential evidence in cygnus a ) , additional physics , e.g. , shear viscosity ( @xcite ; @xcite ; @xcite ) or magnetic tension ( @xcite ; @xcite ; @xcite ) , was proposed to suppress the instabilities . here
we briefly investigate the role of shear viscosity on the cavity evolution .
our simulations are the first set of simulations that study the evolution in the viscous icm of x - ray cavities _ directly created by agn jets _ , while previous studies ( e.g. , @xcite ) follow the evolution of an initially - static cavity which neglects the backflows and the internal circulating motions produced during the interaction of agn jets with the icm .
following @xcite , we introduce a constant dynamic viscosity coefficient @xmath27 in our four viscous runs : r1-visc1 , r1-visc2 , r4-visc1 , r4-visc2 ( see table 1 for model parameters ) .
the jet parameters in the first two runs are the same as those in our fiducial run r1 except that @xmath27 is increased from zero in run r1 to @xmath92 g @xmath99 s@xmath100 in run r1-visc1 and @xmath39 g @xmath99 s@xmath100 in r1-visc2 . similarly , runs r4-visc1 and r4-visc2 are very similar to run r4 except for the different values of @xmath27 . as a comparison , the theoretical braginskii viscosity in a fully ionized , unmagnetized , thermal plasma is @xmath101 g @xmath99 s@xmath100 @xcite .
figure [ plot7 ] shows the time evolution of type - i cavities in runs r1 , r1-visc1 , and r1-visc2 .
it is clear that viscosity does not appreciably affect the shape of young cavities ( e.g. , top panels at @xmath65 myr ) .
however , viscosity does affect the late evolution of x - ray cavities and suppresses the development of kh and rt instabilities , consistent with previous studies ( e.g. @xcite ) . the cavity evolution in run r1-visc2 ( the right panels )
suggests that type - i cavities becomes more elongated along the perpendicular direction ( i.e. , @xmath0 decreases ) as they rise in the icm .
figure [ plot8 ] shows the time evolution of type - ii cavities in runs r4 , r4-visc1 , and r4-visc2 .
similar to type - i cavities , viscosity does not appreciably affect the shape of young type - ii cavities , but does affect their long - term evolution by suppressing interface instabilities .
another important feature is that the cavities in these runs ( figure [ plot8 ] ) rise faster than those in runs r1 , r1-visc1 , and r1-visc2 ( figure [ plot7 ] ) , suggesting that higher momentum fluxes injected by more massive jets in the r4-series runs contribute significantly to the late cavity rise in the icm . in other words ,
bottom - wide type - i cavities rise in the icm mainly due to buoyancy , while top - wide type - ii cavities may rise in the icm due to the combined effects of buoyancy and the original momenta injected by agn jets .
figures [ plot7 ] and [ plot8 ] show that shear viscosity significantly changes the late evolution of x - ray cavities in runs r1-visc2 and r4-visc2 , effectively suppressing rt and kh instabilities .
cavities in our non - viscous runs evolve into torus - like structures ( as also seen in @xcite ) , while viscosity tends to suppress the formation of torus - like morphology , making cavities more coherent .
the viscosity coefficients in these two runs are @xmath102 g @xmath99 s@xmath100 , a significant fraction of the braginskii viscosity ( @xmath103 g @xmath99 s@xmath100 at @xmath104 kev ) . a level of @xmath105 g @xmath99 s@xmath100 in runs r1-visc1 and r4-visc1 produces a much smaller effect .
our 2d simulations assumed axisymmetry , and thus could not explore the roles of non - axisymmetric features potentially present in both agn jets and the ambient icm .
as studied in @xcite in detail , non - axisymmetric motions may affect turbulent mixing within cavities and the development of interface instabilities , but we do not expect that they significantly affect the overall shape of young x - ray cavities and their long - term evolution in a viscous icm .
future 3d viscous simulations including a small level of non - axisymmetry in the initial icm setup ( e.g. , as adopted in @xcite ) are required to explore this interesting topic . in this subsection
, we explore the robustness of our results by studying the formation of type i and ii x - ray cavities in another system the perseus cluster .
perseus is much more massive than our fiducial system virgo , and the icm temperature is about 2 3 times higher .
the gas pressure in perseus is also higher than that in virgo ( by a factor of @xmath106 at @xmath107 kpc and @xmath108 at @xmath109 kpc according to observational profiles in @xcite and @xcite ) . here
we present two representative perseus simulations , runs p1 and p4 , producing type i and ii cavities respectively .
we use the same numerical method as described in section [ section2 ] .
the initial gas temperature and density profiles in perseus are adopted from observational profiles given in @xcite .
the static gravitational potential well is also chosen to establish hydrostatic equilibrium at @xmath110 .
the jet parameters ( @xmath111 , @xmath112 , @xmath47 , @xmath58 , @xmath42 ) in runs p1 and p4 are the same as those in runs r1 and r4 ( see table 1 ) , respectively .
figure [ plot9 ] shows the cavities produced in runs p1 and p4 at @xmath61 myr . the very light internally - subsonic jet in run p1 ( @xmath113 , @xmath114 ) produces a bottom - wide type - i cavity in perseus ( similar to run r1 for virgo ) , while the heavier , internally - supersonic jet in run p4
( @xmath78 , @xmath115 ) produces a top - wide type - ii cavity ( similar to run r4 ) .
this confirms our main results presented in section [ section : young ] , suggesting that the connection between the shapes of young cavities and jet properties may be quite general .
it should also be noted that the sizes of cavities in perseus are slightly smaller than the sizes of corresponding virgo cavities produced by the same jet properties , due to higher icm gas pressures in perseus .
x - ray observations indicate that the scaled angular - averaged pressure profiles in galaxy clusters are quite universal @xcite .
we thus expect that the main results in the paper should hold quite generally .
the shapes of young cavities , particularly radial elongation and top wideness , may be useful probes of the properties of agn jets producing those cavities .
however , we also caution that local variations in the surrounding gas pressure may lead to asymmetries and irregularities in the bubble shape .
angular variations in the icm distribution and the differences in the jet entrainment history may cause that a pair of opposite jets produce two cavities with different shapes .
these effects should be considered while inferring jet properties from the shapes of observed young x - ray cavities .
the propagation of agn jets and the formation of radio lobes and x - ray cavities in the icm have been previously studied with hydrodynamic simulations by many authors in the literature ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) , sometimes including additional physics , e.g. , magnetic fields ( @xcite ; @xcite ; @xcite ) , cosmic rays @xcite , viscosity ( @xcite ; @xcite ) , etc .
many previous studies focus on the thermodynamic impact of mechanical agn feedback on the icm ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) and how to suppress the cavity disruption by interface instabilities ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ) . in the current paper
, we present a very detailed and extensive study on the cavity shape produced by agn jets with many different jet parameters , including very light internally - subsonic jets largely ignored in previous studies .
in particular , we propose two very important parameters to characterize the shape of x - ray cavities : radial elongation @xmath0 and top wideness @xmath1 .
our results are consistent with @xcite , both demonstrating that massive ( high - density ) jets tend to form cavities elongated along the jet direction and fat " type - i cavities are formed by very light jets .
the major difference is that our jets producing type - i cavities are energetically dominated by thermal energy density , while the jets in @xcite are dominated by the cosmic ray energy density .
@xcite showed that fat cavities can also be produced by massive jets with very large half - opening angles ( @xmath116 ) , although radio jets are usually observed to be well collimated ( e.g. , @xcite ) .
@xcite studied the long - term evolution of initially - static cavities in the icm and the role of shear viscosity .
our results in section 3.2 are generally consistent with @xcite on the role of viscosity in suppressing kh and rt instabilities . however , starting with initially - static cavities , @xcite could not probe the role of the jet momentum , which likely plays an important role in the rise of type ii cavities in the icm as shown in figure [ plot8 ] .
our simulations directly probe the formation process of x - ray cavities by agn jets , which produce jet backflows and circulating motions in cavities ( e.g. , @xcite ; @xcite ) . for type - i cavities shown in figure [ plot7 ] ,
some backflows reaching the cluster central regions detach from the cavity at late times ( e.g. , at @xmath117 myr ) , even in run r1-visc2 with quite strong viscosity @xmath102 g @xmath99 s@xmath100 .
this feature is not seen in simulations of initially - static cavities @xcite , and it is unclear if it represents reality .
the detached cavity material leads to a small x - ray - deficient area in x - ray surface brightness maps ( bottom panels ) , and may not be easily seen in real x - ray images .
if the non - existence of this feature is observationally confirmed , further investigations are necessary to explain it .
_ chandra _ and _ xmm - newton _ observations of galaxy clusters have detected a large number of x - ray cavities in galaxy groups and clusters , providing one of the most compelling evidences for the importance of jet - mode agn feedback ( e.g. , @xcite ) . through the cavity s
volume and the surrounding gas pressure , x - ray cavities have been successfully used to estimate the energetics of mechanical agn feedback ( e.g. , @xcite ; @xcite ; @xcite ) .
here in this paper we argue for the physical importance of the shapes of x - ray cavities , whose importance has been largely ignored in previous studies .
x - ray cavities are created by the interaction of agn jets with the surrounding icm gas , and thus the cavity shape potentially contains important information about the properties of agn jets and the icm .
in particular , the cavity elongation with respect to the jet direction may be of great interest .
x - ray observations indicate that most cavities are elongated along either the perpendicular direction ( including nearly circular cavities ) or the jet direction ( e.g. , @xcite ; @xcite ) , suggesting that x - ray cavities are usually not subject to substantial rotation during their evolution in the icm .
this further implies that typical turbulent motions in galaxy clusters do not effectively rotate the cavities .
however , it should be noted that coherent gas motions ( e.g. , gas sloshing ) may bend ( or rotate ) radio lobes and x - ray cavities ( e.g. , @xcite ; @xcite ) , as numerically simulated in @xcite . to quantitatively characterize the cavity shape , we define two important geometrical parameters : radial elongation @xmath0 and top wideness @xmath1 .
we refer cavities with @xmath14 ( i.e. , nearly circular or elongated along the perpendicular direction ) as type i cavities , and those with @xmath15 ( i.e. , elongated along the jet direction ) as type ii cavities .
both types of x - ray cavities have been observed in the cavity samples of @xcite and @xcite .
the value of @xmath1 further separates each type of cavities into three subgroups : top - wide ( @xmath18 ) , center - wide ( @xmath19 ) , and bottom - wide ( @xmath20 ) . using a suite of 14 axisymmetric hydrodynamic simulations , we study how the cavity shape is affected by various jet properties and viscosity in the icm .
the x - ray cavities in our simulations are directly created by agn jets , and we also follow their evolution in the icm . after the jet is turned off , the produced cavity rises in the icm and its shape continues to evolve .
we thus study the shapes of young cavities and old cavities separately to better distinguish the roles of the different physics parameters .
our simulations indicate that viscosity does not significantly affect the shapes of young cavities , which can be directly used to probe jet properties .
we show that it is feasible to produce both type i and ii cavities in hydrodynamic simulations .
the value of radial elongation @xmath0 increases with the jet density , velocity , and duration , but decreases with the jet energy density and radius . to disentangle the effects of these jet parameters , we suggest to study the second geometrical parameter top wideness .
bottom - wide type - i cavities ( e.g. , the southern inner cavity in perseus ) are produced by very light internally - subsonic jets , while top - wide type - ii cavities are produced by heavier jets with internally supersonic velocities , which may also produce center - wide cavities with @xmath4 if the jets are slightly supersonic ( @xmath94-@xmath91 ) .
bottom - wide type - ii cavities can be produced by very light jets with very long durations , and cylindrical cavities ( both type i and ii ) are produced by very light internally - supersonic jets .
we confirm the above connection between the shapes of young cavities and jet properties with two additional simulations of the perseus cluster , which is much more massive than our fiducial virgo cluster .
viscosity plays a significant role in the long - term evolution of x - ray cavities in the icm . in non
- viscous runs , both type i and ii cavities are subject to rt and kh instabilities and the major part of a cavity tends to evolve into a torus - like structure .
the role of viscosity becomes more important as the dynamic viscosity coefficient increases . with @xmath102 g @xmath99 s@xmath100 ( a significant fraction of the braginskii viscosity @xmath118 g @xmath99 s@xmath100 ) , viscosity
significantly affects the long - term cavity evolution , effectively suppressing both interface instabilities and the formation of torus - like morphology .
we also find that the top - wide type - ii cavity in run r4 rises much faster than the bottom - wide type - i cavity in run r1 , suggesting that the momenta injected by agn jets , in addition to buoyancy , also contribute significantly to the late rise of top - wide type - ii cavities in the icm . for bottom - wide type - i cavities , buoyancy may play the dominant role , similar to initially static cavities in @xcite .
our numerical study has important observational implications and we encourage x - ray observers to study the shapes of x - ray cavities
. the shapes of young cavities can be used to probe the properties of agn jets , while the shapes of old cavities may be a useful probe of the icm viscosity level .
one can look at individual cavities .
for example , the outer northwestern x - ray cavity in perseus ( see figure 3 in @xcite ) is significantly flattened ( cap - shaped ) and elongated along the perpendicular direction . in our simulations ,
the most similar cavity is the one in the viscous run r1-visc2 at @xmath119 myr ( see figure [ plot7 ] ) , suggesting that the cavity was created by a very light internally - subsonic jet , and viscosity in the icm of perseus is significant . while further studies are necessary to make conclusive statements , the suggested significant viscosity level is consistent with the constraint derived from coherent cold filaments in the wake of this cavity ( @xmath120 @xmath121 s@xmath100 or @xmath122 g @xmath99 s@xmath100 taking @xmath123 @xmath124 for the electron number density ; @xcite ) .
a substantial viscosity level in the icm is also consistent with the stability of cold fronts observed in some galaxy clusters ( e.g. , @xcite ) .
furthermore , it may be even more interesting to statistically study the cavity shape in a large sample of x - ray cavities , investigating if there is a bimodality in the distributions of @xmath0 and @xmath1 , and how the values of @xmath0 and @xmath1 vary with other cavity properties ( e.g. , the cavity distance from the cluster center , the mechanical agn power , etc ) .
f.g . acknowledges generous support by the zwicky prize fellowship and the observational cosmology group ( pi : simon lilly ) at eth zrich .
f.g . thanks william mathews for helpful discussions and two anonymous referees for very insightful comments , which significantly improved the paper . | x - ray observations of galaxy clusters have detected numerous x - ray cavities , evolved from the interaction of agn jets with the intracluster medium ( icm ) and providing compelling evidence for the importance of jet - mode agn feedback . here
we argue for the physical importance of the cavity shape , which we characterize with two geometric parameters : radial elongation @xmath0 and top wideness @xmath1 .
we study the cavity shape with 16 hydrodynamic jet simulations in two representative clusters , and find that the shapes of young cavities are mainly determined by various jet properties .
our simulations successfully reproduce two observed types of young cavities elongated along either the jet ( @xmath2 ; type - ii ) or perpendicular ( @xmath3 ; type - i ) direction .
bottom - wide type - i cavities are produced by very light internally - subsonic jets , while top - wide type - ii cavities are produced by heavier , internally - supersonic jets , which may also produce center - wide cavities with @xmath4 if the jets are only slightly supersonic .
bottom - wide type - ii cavities can be produced by very light jets with very long durations and cylindrical cavities are produced by very light internally - supersonic jets .
while not appreciably affecting the shapes of young cavities , viscosity significantly affects the long - term cavity evolution , suppressing both interface instabilities and the formation of torus - like morphology .
we encourage observers to study the shapes of young and old x - ray cavities separately , the former probing the properties of agn jets and the latter potentially probing the icm viscosity level . |
in the collision of two heavy ions one can generate for a very short time an atomic system with an effective charge which is the sum of the charges of the colliding nuclei .
this is possible because electrons with a wavelength of 1/@xmath0 fm can not resolve structures far below this scale . for the short time
such a collision takes an electron becomes very strongly bound .
its binding energy can actually exceed twice its rest mass .
this happens if the combined atomic charge is larger than roughly 173 , depending on the amount of shielding by the other electrons .
fig.1 gives an impression of the relevant atomic processes in such an ` overcritical ' heavy ion collisons .
these processes have been extensively studied , both theoretically and experimentally . with the exception of @xmath1-electron production in u+pd collisions at 6.1 mev / u the theoretical and experimental results
have been found to agree perfectly within their uncertainties , which are typically 10 - 20 percent . + the technical treatment of strong field problems is basically always the same .
because the coulomb - potential is so strong it can not be treated as perturbation but has to be included into the fundamental hamiltonian .
for the special case of two colliding heavy ions this reads @xmath2 with @xmath3 here @xmath4 is the nuclear charge density and the indices @xmath5 and @xmath6 stand for projectile and target .
it is also practical to include the dominant part of the electron - eclectron interaction , which describes shielding of the ion charges , in the potential .
thus @xmath7 can , e.g. , be chosen as @xmath8 implying a slater approximation for the exchange term . in writing down this hamiltonian
we have made the monopole approximation , i.e. higher multipoles of the coulomb - field are neglected .
one can go beyond this approximation , but it was found that the contributions from higher multipoles are negligable .
@xmath9 is used to calculate the time - independent eigenstates of the two - ion system .
time dependence enters through the varying distance vector between the two ions @xmath10 .
the time - dependent dirac - equation can be converted to a system of coupled differential equations by expanding the wave functions according to @xmath11 @xmath12 @xmath13 these equations have to be solved for a given trajectory @xmath10 .
the square of the expansion amplitudes @xmath14 taken at large @xmath6 leads to the occupation probability for the state @xmath15 . in this way many experiments were described with high accuracy . figure 2 and 3 are typical examples and one could show dozens of similar plots .
however , the most interesting process in figure 1 , namely the process c which is called ` spontaneous pair production ' could not be observed so far and theoretical calculations show that to observe it would require the two ions to stick together much longer than usual .
it is generally agreed that for the rather broad class of nuclear reactions summed over in the experiments so far , such long sticking times can not be expected . + in spite of this fact positron experiments at gsi found pronounced positron lines @xcite and even extremely sharp electron - positron coincidence lines @xcite . at the time of their discovery it was tempting to interpret these lines as signal for the decay of a new particle , possibly the axion .
new experiments and further theoretical investigations of many low - energy phenomena showed , however , conclusively that this is no valid possibility ( for a review see @xcite ) .
we summarized these results already early on as folllows : ` ... we have to conclude that we did not find any scenario that would allow to describe the production of an x particle with the required properties without very unnatural assumptions . '
( 1986 , @xcite ) and ` ... therefore , it is tempting to conjecture that some of the experiments are unreliable . '
( 1988 , @xcite ) .
+ the present situation of these lines is the following : recent epos @xcite and orange @xcite experiments at gsi could not reproduce the line observed early although the detectors were improved to give much better statistics ( especially for epos ) . a completely independent experiment at argonne national laboratory called apex
did not find any indication for lines @xcite . at the same time it became clear that nuclear pair - conversion was less well understood than assumed , such that at least some of the data from recent years might be due to conversion , a possibility which is still excluded for the early much larger line structure .
it is unclear , how this has to be interpreted , but definitely there is presently no compelling evidence for any process more exotic than nuclear pair conversion .
a very interesting possibility opened by high - z atoms is to test qed in a non - perturbative regime .
qed is a divergent field theory , and the three basic divergences : vacuum polarization , self energy , and vertex correction have to be renormalized .
this renormalization procedure in a coulomb - potential differs from that in free space , which results in finite corrections to basically all physical quantities .
the complete understanding of these corrections is indispensible for a reliable description of processes taking place in external fields . in frankfurt
we have a long tradition in calculating these @xcite and presently a number of projects are under work .
let me just make a few comments on one of them , which is related to our work on channeling .
+ _ delbrck scattering _ ,
i.e. the scattering of a photon off the coulomb - field of an atom ( see figure 6 ) was carefully measured already some time ago @xcite .
systematic deviations from the lowest - order theoretical prediction were found for high - z atoms , indicating that for these the higher order terms , proportional to @xmath16 , are important .
a direct perturbative calculation of the next order graph is extremely involved and anyway one would like to calculate this graph exactly in all orders of @xmath17 .
this is in fact possible if the perturbative propagator is substituted by the exact propagator for the dirac - operator including the coulomb - field @xmath18 we decomposed the propagator again into contributions of definite angular momentum and inserted these into the calculation of the vacuum polarization graph .
the resulting calculation is very demanding , involving thousends of angular momentum states , and requiring many numerical and analytic tricks to ensure convergence .
details can be found in ref .
some results are given in the table for photons with the energy @xmath19 for the scattering angle @xmath20 obviously one finds corrections with respect to the lowest - order result for sufficently large @xmath21 .
these investigations are still continued to get complete systematic results for all photon energies and scattering angles . + in another investigation we treated _ channeling _ by solving the dirac equation , starting from first principles .
again this calculation was extremely involved .
however , it constitutes the first completely relativistic calculation for this problem .
if the physics of channeling were qualitatively similar to that of strong fields this should show up in specific relativistic effects .
in fact we got rather satisfactory results ( see figure 7 and 8 , and @xcite ) , but we found no specifically relativistic effects .
it is usually argued that the relevant mass is very large @xmath22 , such that one can perform quasi - classical calculations .
our results confirm this and show at the same time that for such effectively very heavy particles no phenomena occur which would be similar to overcritical binding in heavy atoms . as a by - product of this calculation we obtained the distribution of the radial distance from the channeling axes at which photons are emitted ( see figure 9 ) .
again details can be found in our publication @xcite .
+ a similar conclusion was also reached by the following investigation .
we studied again the delbrck graph from figure 6 , but this time for the axially symmetric field of a channeling axis .
the calculation was much simplified by the fact that this problem is two - dimensional rather than three - dimensional .
the physical motivation was that a photon couples also to the channeling fields due to virtual production of electron - positron pairs , and this interaction could lead to a channeling of photons with a very small angle relative to the channeling axis , see @xcite .
one result is shown in figure 10 , where the average rotation angle around the channeling axis during the passage of a photon through the central part of the channeling potential is shown as a function of the angle of incidence @xmath23 .
@xmath24 would imply complete channeling , i.e. the photon would be bound to the channel .
obviously we are very far from this point , even for very small angles .
for details see @xcite .
again , no substantial increase of the effect with increasing photon energy was observed .
such an increase would have indicated a prolonged existence of the virtual electron - positron pair and thus a stronger interaction with the channeling potential ) was observed .
+ cutting a long discussion short , we would like to argue that the channeling effect is basically different from the effects in over - critical coulomb - fields for the following reason : the fundamental process underlying all effects is the separation of an electron - positron pair in a strongly space - dependent potential . the energy gained by this charge separation is a measure for the size of the novel effets up to the point where it allows the electron and positron to become real , which would be spontaneous pair production . in the channeling situation
the electric fields becomes , however , only strong due to a lorentz boost ( into the rest system of , e.g. , the electron ) .
the same boost introduces also an magnetic field which confines the electron and positron to a radial distance of the size of a landau orbit @xmath25 .
thus the crucial quantity ( electric field strength ) @xmath26 ( radial separation of electron and positron ) is proportional to @xmath27 , which is basically independent of the electron velocity , i.e. the lorentz factor @xmath28 .
thus no typical overcritical field effects can be expected .
+ * acknowledgement : * i have reported results of the whole frankfurt atomic physics group , most notably j. augustin , j. klenner , u. mller - nehler , j. reinhardt , a. scherdin , g. soff . this group is headed by w. greiner .
i also thank the experimentalists of the apex , epos , and orange collaboration for communicating their results prior to publication .
w. koenig et al . , phys . lett . * b 218 * ( 1989 ) 12 a. schfer , j. phys . *
g 15 * ( 1989 ) 373 a. schfer , j. reinhardt , b. mller , and w. greiner , z. phys . * a 324 * ( 1986)243 a. schfer , j. reinhardt , b. mller , and w. greiner , int .
phys . * a 3 * ( 1988 ) 1751 h. bokemeyer private communication and r. ganz et al .
, to be published in phys .
lett * b * w. knig private communication apex collaboration , i. ahmed et al.,argonne preprint phy-8030-hi-95 , to be published in phys .
p. rullhusen et al . , phys . rev .
* c 27 * ( 1983 ) 559 a. scherdin , a. schfer , w. greiner , and g. soff , phys . rev .
* d45 * ( 1992 ) 2982 yu.v .
kononets , v.a .
ryabov , nucl .
* b 48 * ( 1990 ) 269 and 274 ; yu.v .
kononets , j. moscow phys .
* 2 * ( 1992 ) 71 r. medenvaldt et al . , phys .
* b242 * ( 1990 ) 517 j. augustin , a. schfer , and w. greiner , phys . rev . * a 51 * ( 1995 ) 1367 j. klenner , j. augustin , a. schfer , and w. greiner , phys .
rev . * a 50 * ( 1994 ) 1019 | we discuss the status of atomic physics in strong field .
we focus on the problem of electron - positron lines observed in heavy - ion collisions and on qed effects , calculated in strong coulomb fields , especially delbrck - scattering .
we discuss the similarities and differences between these effects and channeling respectively beamstrahlung .
we investigated the prospects for photon - channeling , calculated channeling from first principles on the basis of the dirac equation , and determined the rate for electron - positron pair production in the collision of two high - energy particle pulses . in this contribution
i would like to put the channeling phenomenon in a larger context , namely the physics of strong electromagnetic fields , a subject extensively investigated in frankfurt since many years .
the common technical feature of the many different effects investigated under this label is that the interaction of ( usually ) electrons with the electromagnetic fields of the problem can not be treated in perturbation theory , but has to be taken into account exactly @xcite .
+ to illustrate the diversity of this field i would like to address very shortly three topic , namely the status of the electron - positron coincidences observed at gsi for quite a number of years , some problems of qed in heavy atoms , and our contribution to the understanding of channeling and beamstrahlung . |
ring lasers ( rl ) are top sensitivity devices able to measure absolute rotations .
the principle of operation of a ring laser is based on the sagnac effect @xcite .
rls are very reliable instruments , with large bandwidth and very high duty cycle .
the most advanced rls are indeed used for accurate metrology in geophysics ( rotational seismology ) , and in geodesy for monitoring the fast variations of the earth rotation rate . for a laboratory on earth ,
the signal of a rl is proportional to the instantaneous s , much smaller than the typical times of other time depending phenomena in the lab and on the planet altogether . ]
norm of the vector sum of the diurnal rotation rate of the planet , @xmath0 , and the local rotation rate of the device , @xmath1 ; the two components together may be called _ kinematic rotations_. in principle
the superposition of local and global kinematic rotations produces a time varying signal , since local rotations are referred to a non - inertial reference frame , so that the rotation rate with respect to distant inertial observers turns out to cyclically depend on time . in many practical applications
the latter time dependence is _ de facto _ negligible .
when the effects of non - newtonian gravity are included , an additional contribution may appear ; let us call it @xmath2 .
if general relativity ( gr ) is used , @xmath2 is in turn the sum of two contributions : the lense - thirring drag term @xmath3 and the de sitter geodetic precession @xmath4 .
if the rl is carried on a vehicle the dominant term is @xmath1 ; in a laboratory fixed to the ground @xmath0 prevails ; the gr terms are @xmath5 rad / s , nine orders of magnitude below the earth rotation rate . as for @xmath1 , in an earth based laboratory it is either negligible or known and modelled so that it can be accounted for and subtracted .
the present best sensitivity of a rl is @xmath6 rad / s in one day of integration time @xcite , not far from the threshold to be crossed in order to detect the gr terms . in short ,
the response of the rl is a beat frequency @xmath7 proportional to the scalar product between the total angular rotation vector and the area vector , @xmath8 , of the ring : @xmath9 .
the proportionality factor @xmath10 is called scale factor and depends on the geometry of the ring .
it is @xmath11 , where @xmath12 is the area and @xmath13 the perimeter of the ring , @xmath14 is the wavelength of the light of the laser . with an appropriate construction and location of the apparatus and for long enough integration time we may assume @xmath15 to be negligible , even with respect to the gr terms , and other effects to be modelled and subtracted accurately , so
that , in the framework of general relativity , we write @xmath16.the purpose of the ginger experiment ( gyroscopes in general relativity ) is to measure the gr components of the gravitational field of the earth at @xmath17 or better accuracy level , by means of an array of ring - lasers . in 2011 a first proposal was presented based on an octahedral configuration @xcite .
the three - dimensional array would permit to reconstruct the modulus of the total angular rotation vector in the laboratory .
the gr terms in this scheme would be evaluated by subtracting the earth rotation rate measured independently by the international earth rotation and reference systems service ( iers ) , @xmath18 .
the proposed approach would require long term stability and very high accuracy , since it would be necessary to subtract the contribution of @xmath0 , which , as said , is about nine orders of magnitude bigger than the gr terms .
so far the gravitomagnetic field of the earth has been measured by spaceborne experiments , being the present accuracy limit @xmath19 lares .
the experimental goal to measure @xmath20 down to @xmath17 , remains an important challenge .
ginger would provide the first measurement of the general relativistic features of the gravitational field , on the surface of the earth ( not considering the gravitational redshift ) .
though not in free fall condition , it would be a direct local measurement independent from the global distribution of the gravitational field , which is the principle difference with the space experiments where the result is the consequence of an averaging of the effects along whole orbits . in the following we shall discuss the ways an actual measurement based on ring lasers can be done , evidencing criticalities and the role of physical and geometrical parameters and the related uncertainties .
as we have already written in the introduction , the beat frequency @xmath7 of a rl is proportional to the flux of a total rotation vector @xmath21 across the area of the ring .
in general we may write : f = s_t [ flusso ] where @xmath22 is the unit vector perpendicular to the plane of the ring ( provided , of course , that it is contained in a plane ) .
if we wish to fully recover @xmath21 from the measurement of frequencies we need in principle three independent rings , which form a local three - dimensional reference frame , as it was proposed in @xcite .
if we have reasons to think that a couple of rings may be oriented so that the plane of their @xmath22 s contains @xmath21 , the problem becomes bi - dimensional and two rings are enough ; we shall comment on this later , but let us assume for the moment that this is the case . making the scalar products explicit and calling @xmath23 the angle between @xmath24 and @xmath25 ( see fig .
[ twodt ] ) we may write : f_1&=&s_1 _ t + [ system0 ] + f_2&=&s_2 _ t here @xmath26 is the angle between @xmath21 and @xmath24 . system ( [ system0 ] ) may be transformed into : & = & + _
t&=&= provided @xmath23 is known ( directly measured ) , we may solve and obtain : & = & + [ generale ] + _ t&= & of course everything simplifies if we may assume @xmath27 and @xmath28 . it
would then be : & = & [ tanz ] + s_t&= & [ omega2 ] in order not to get in trouble with signs , we should specify a couple of assumptions : 1 ) @xmath21 is in between @xmath24 and @xmath25 ; 2 ) it is @xmath29 .
either in the form ( [ generale ] ) or ( [ tanz ] ) ( [ omega2 ] ) the two rings give @xmath21 without any reference to the composition of the vector and specifying the orientation in the meridian plane with respect to themselves .
the metric of the external space - time of a spherical rotating mass is written in the simplest form in the reference frame of an inertial observer located at infinity and at rest with respect to the center of the source of gravity .
it is also convenient to work in weak field approximation , where terms down to the smallest interesting contribution are kept .
the smallest term we keep is linear in the angular momentum of the central mass @xmath30 .
the line element , using space `` polar '' coordinates , is : @xmath31 variables marked by pedix @xmath32 will change when passing to the final frame .
it has been assumed that @xmath33 and negligible .
it is @xmath34 the last assumption includes the hypothesis that the earth is a rigid body whose relevant moment of inertia is @xmath35 .
the numerical values , when considering the surface of the earth , i.e. @xmath36 m , confirm that the approximation adopted is correct .
the measurement is intended to be performed in a terrestrial laboratory , so it is appropriate to rewrite the line element in its reference frame .
this is made through two steps @xcite : * rotation of the axes at the angular velocity of the earth @xmath37 ; * boost at the peripheral speed of the earth whose absolute value is @xmath38 where @xmath39 is the colatitude of the laboratory and @xmath40 is the ( average ) radius of the earth . in the process
, we keep an approximation level consistent with the weak field hypothesis , extended to kinematical rotation terms .
the size of the latter is expressed by the ratio @xmath41 .
the result is : @xmath42 for short we write @xmath43 the frame is non - inertial and comoving with the laboratory ; the origin remains in the center of the earth .
considering a null line - element ( i.e. a light ray : @xmath44 ) from ( [ ellinea ] ) we deduce the coordinated travel time of flight element @xmath45 . assuming a path closed in the laboratory ( it is not closed for an inertial external observer ) and integrating along the path once to the right ( @xmath46 ) , once to the left ( @xmath47 ) , then subtracting the two results , we arrive to the difference in the coordinated times of flight ( expressed in arbitrary coordinates ) : @xmath48 it is possible to convert the result to the proper time of the observer at rest in the lab , @xmath49 , just multiplying by @xmath50 at his / her position @xmath51 considering the symmetry of the problem we may interpret @xmath52 as the only non - zero component of a three - vector @xmath53 aligned with the axis of rotation of the earth so that ( introducing the unit vector @xmath54 aligned with the trajectory of light ) the formula becomes : @xmath55 the quantity is an observable , i.e. a true scalar : the same for any observer . in a ring laser
the time of flight asymmetry is converted into a difference in the frequency of stationary light beams and one obtains a beat frequency : @xmath56 the line integral may be transformed into a flux using gauss s theorem ( classically it would be called stoke s theorem ) .
if the change of the values of the curl of @xmath57 across the area of the closed integration path is negligible , the result becomes simply @xmath58where @xmath12 is the area contoured by the beams , @xmath13 is the length of the path , @xmath59 is the wavelength in the active cavity and the curl is evaluated in any point within the ring . besides the physical and geometrical parameters , to be controlled experimentally , the signal depends on three quantities : @xmath60 ( proportional to the mass of the source @xmath61 ) , @xmath35 ( contained in @xmath62 ) , and @xmath63 .
after a few manipulations , the expected signal becomes @xmath64 \cdot \hat{u}_{n } \label{base } \\ & = & s\left ( \vec{\omega}_{\oplus } + \vec{\omega}_{ds}+\vec{\omega}_{lt}\right ) \cdot \hat{u}_{n } \label{baseb}\end{aligned}\ ] ] as we see , according to gr , @xmath21 is the sum of three vectors , all contained in one plane ( the meridian plane ) , thus explaining the special attention paid to the two rings system in the previous section . let us introduce the angle @xmath65 between the direction of the axis of the earth and the axis of the ring ( see fig .
[ orientation ] ) . then the relevant dot products become : @xmath66 and of the two gr effective rotations @xmath3 and @xmath4 is represented .
the sum @xmath21 of all rotation vectors is also shown .
the amplitude of @xmath0 is down - scaled by about 10 orders of magnitude .
the graph is not on scale . ]
introducing the shorthand notations @xmath67 eq .
( [ base ] ) is converted into : @xmath68 the absolute value bars @xmath69 have been introduced just to remember that the frequency is of course always a positive quantity .
the factor in front of the bar is the scale factor @xmath10 of the rl .
it is also important to remark that @xmath70 acts as a global multiplication factor .
eq.s ( [ base ] ) and ( [ frequenza ] ) , unlike eq.s ( [ system0 ] ) , present the expected orientation of the rl referring to the external frame formed by @xmath71 and @xmath72 ; this is manifested by the presence of the colatitude @xmath73 and the angle @xmath65 ( in the configuration represented in fig .
[ orientation ] the two angles have opposite signs ) .
as stated at the beginning , we have treated throughout the earth as a sphere , but we know that our planet is not a sphere . without entering into the details of the geoid , we should better treat the earth as an ellipsoid : how would this better approximation affect formulae like ( [ base ] ) and ( [ frequenza ] ) ?
we can recall that the shape parameter of the terrestrial reference ellipsoid is @xmath74 .
it should produce locally a deviation of the vertical direction ( i.e. the local gravitoelectric field ) from the radial direction ( which appears in the formulae ) in the order of @xmath75 rad at most .
such a deviation affects the gr terms in eq .
( [ base ] ) on the corresponding amount of @xmath76 part in @xmath77 at most , so being below the target accuracy for the experiment .
going back to eq . ( [ base ] )
we may express @xmath37 in terms of @xmath71 and @xmath78 .
we then obtain an explicit formula for the modulus of @xmath21 : _ t = _ [ omt ] equating ( [ omt ] ) to the second equation in ( [ generale ] ) we get a relation between the experimental quantities and a combination of the parameters of the theory : _ = remember that @xmath79 and @xmath80 are expected to be of the order of @xmath81 , so that we may keep the only first order corresponding terms : 1 + 2b^2+(a - b)^2 [ solparz ] if , for simplicity , we assume that the earth is spherical with an internal uniform mass distribution , the relationship between @xmath35 and @xmath82 is : @xmath83 ( @xmath84 ) . for the real earth the numerical factor is closer to @xmath85 rather than @xmath86 ( @xmath87)(see the appendix ) . apart from the modulus of @xmath21 the theory tells us also the orientation of the vector in the meridian plane : it will be at an angle @xmath88 with respect to the axis of the earth .
when the ring has its axis oriented as @xmath21 , the signal reaches its maximum value : f_max = s_t angle @xmath88 can be obtained applying the maximum condition to eq .
( [ frequenza ] ) .
in fact , starting from eq .
( [ frequenza ] ) , using @xmath65 as the independent variable , calling @xmath88 the value of the angle at the maximum , we find : @xmath89 considering the orders of magnitude , the first order approximation is : @xmath90 posing @xmath91 in ( [ frequenza ] ) , then using ( [ alpha ] ) or ( [ sima ] ) , we may obtain from ( [ frequenza ] ) the same @xmath92 as the one written in ( [ omt ] ) . in order to establish a correspondence between the predictions of the theory and the results obtained from a couple of rings ,
let us remark that eq .
( [ frequenza ] ) can also be written putting @xmath93 .
looking at ( [ generale ] ) , let us consider a configuration where ring 1 is aligned at the maximum ; this corresponds to @xmath94 .
if the second ring is horizontal ( i.e. @xmath95 ) , it is @xmath96 ( see fig . [ orientation ] ) .
we may write : = finally , calling in ( [ alpha ] ) or ( [ sima ] ) , we arrive at : = [ secex ] or a-3b [ secap ] the end point of this process is reached putting ( [ secap ] ) and ( [ solparz ] ) in a system : a-3b & & + [ sist ] + 1 + 2b^2+(a - b)^2 & & choosing @xmath79 and @xmath80 as unknowns , all other parameters must be measured . in principle system ( [ sist ] ) can be solved .
the formal result is : a&=&-++ [ aaa ] + b&=&--+ [ bbb ] the practical difficulty with this solution is that it is composed of strongly differing parts , scaling over at least eleven orders of magnitude ( if the aim is a @xmath17 accuracy in @xmath79 and @xmath80 ) , so requiring a corresponding accuracy in all parameters , including @xmath37 .
going back to system ( [ sist ] ) , we see that the first equation is apparently purely geometrical , without calling in @xmath37 .
now the basis is the measurement of @xmath23 , which in fact is equivalent to the direct measurement of @xmath88 ; remember that @xmath23 is the angle between @xmath21 ( i.e. the direction of the maximum along which the first ring is oriented ) and @xmath71 ( which is perpendicular to the plane of the second ring ) . using the first equation
only , it would even be unnecessary to have the second ring , but of course the angles must be measured with an accuracy better than @xmath76 nrad and the result would be the combination @xmath97 . to have @xmath97 , instead of the two parameters separately , would not be a problem since we know the relation between the two .
of course the same results may be obtained from ( [ aaa ] ) and ( [ bbb ] ) .
the configuration first considered in 2011 , as recalled in the introduction , has been the octahedron @xcite .
this configuration has been extensively discussed in previous papers ; it measures the three components of @xmath21 in all three spatial directions and reconstructs the norm of the vector combining together different measurements .
this approach allows the comparison of different co - located rings , giving the possibility of precisely measuring the systematics of the laser .
if the orientation of the octahedron with respect to the rotation axis of the earth and the meridian plane are not given , all information that can be retrieved by the experiment is contained in the norm of @xmath21 confronted with the theory .
if also the external orientation is given , the full @xmath21 vector , including the angle with respect to the axis of the earth , is obtained . in general
, using multiple independent rings ( two , three or more ) has several advantages : the statistics would be improved since the shot noise of each ring , in the set of many , is independent from the others .
co - locating more than three rings would be a powerful tool to keep the systematics of an experimental apparatus under control .
an array of at least four co - located rings would have the very interesting feature that the angular rotation vector could be reconstructed with different combinations @xmath98 by @xmath98 .
the comparison of different results would give information on the systematics of the lasers .
redundancy would be allowed , which is always welcome in this kind of experiment .
the weak point is that the detection of the gr terms requires the knowledge of @xmath0 which in practice is provided by iers and @xmath99 is given with an uncertainty too high , as for now , to allow to reconstruct the lense - thirring effect at the @xmath17 accuracy level .
in fact , the length of day ( lod ) is measured with different methods by the iers , but , in the best case , with a @xmath100 @xmath101 error .
this is compatible with a @xmath102 test , more or less ; improvements are not foreseen in the next five years plan ( iers annual report 2014 @xcite ) .
it is however true that prolonging the measurement time would reduce the uncertainty ( 1 order of magnitude in 10 days ) , but of course one must insure the stability of the apparatus over the whole extension of the run . as we have already seen , a single ring measures the projection of the total @xmath21 on a direction perpendicular to the plane containing the ring .
in principle a single ring could give all the information ( first equation in ( [ system0 ] ) ) besides the orientation of the total vector with respect to the axis of the earth , but the knowledge of the absolute angle @xmath26 between the normal to the ring and the direction of the maximum signal is required .
the normal to the ring is assumed to lie in the meridian plane . using the first approximation of eq .
( [ omt ] ) it is : 2b^2+(a - b)^2 - 1 [ uno ] a special case is obtained when the ring is oriented to the maximum signal .
it is then @xmath94 and @xmath103 .
an advantage of this configuration is that , being in a maximum condition , the sensibility to orientation inaccuracy is second order : an uncertainty of the order of a @xmath104 affects the frequency at the @xmath105 level .
furthermore the orientation with respect to the axis of the earth is directly given by the theory , eq .
( [ alpha ] ) or ( [ sima ] ) . in any case using the simple proportionality relation mentioned at the end of sect .
[ exptheor ] , the unknown is reduced to one ( for instance @xmath79 ) and we may solve for it .
another special case is a horizontal ring ( normal in the meridian plane and aligned with the local newtonian field ) .
the output is obtained from eq .
( [ frequenza ] ) putting @xmath106 : f_h = s_(1 + 2b)cos[oriz ] whence the gravito - magnetic parameter @xmath80 immediately stems .
the advantage of this configuration is that , unlike other orientations , it corresponds to an angle materially defined in the laboratory .
unfortunately this simplification is not so strong as it looks : horizontality is not an extremal condition , so it has to be reached with an accuracy of the same order of magnitude as the one required for the gr term . in practice
the maximum tolerable deviation from the horizontal plane is of the order of a @xmath107 .
we must mention that the horizontality condition is affected by the shape of the geoid , which has not regular surface , according to the comments we have put at the end of section [ signal ] . here too , however , the effect on the gr terms as such is negligible .
unfortunately the isolation of the lense - thirring term requires the subtraction of a contribution of the order of @xmath108 which must be known in 1 part in @xmath109 and this is the real reason for the @xmath107 requirement mentioned above .
it is worth remarking that eq .
( [ base ] ) tells also that gr terms could , in principle , be obtained keeping the normal in the meridian plane , but orienting the ring so that its plane contains the direction of the axis of the earth ( orthogonality condition between the axis and the normal to the plane of the ring ) . in that configuration the kinematic frequency would be zero .
unfortunately , this method is not viable for two reasons : 1 ) ring - lasers must be operated with a bias in order to avoid the locking of the two counter - propagating modes ( and the gr terms alone would probably be too small to give the necessary bias ) ; 2 ) the accuracy required in the alignment in order to insure that the ring keeps its normal in the meridian plane becomes extremely severe .
the latter statement may be verified by an example : considering a square ring - laser @xmath110 m in side , and in the same time with its normal perpendicular to the axis of the earth , a tilt of @xmath111 @xmath107 ( from the zero kinematical contribution orientation ) would mimic the expected gr signal .
in general a single ring requires a control of its absolute orientation and , based on the above arguments , the practically viable solution is the ring oriented at the maximum signal .
a single ring parallel to the axis of the earth has no practical application : the rl does not work properly , and the requirements on the accuracy of @xmath65 become exceedingly severe .
these considerations hold also in the case of a pair of nested rings perpendicular to each other ( @xmath28 ) , where one is oriented to the maximum : the second ring would not work . keeping @xmath79 and
@xmath80 as separate unknowns , the dependence on the co - latitude can be exploited .
a couple of measurements performed by rings in laboratories located at different latitudes could give both unknowns .
the general equations are : @xmath112 if both rings are oriented to the maximum , it is : @xmath113 the formal solution of the system ( [ latit ] ) is : @xmath114 here too , a useful combination obtainable from ( [ solu ] ) is a-3b2 [ a-3b ]
the advantage of this equation is that @xmath70 ( actually its inverse ) appears only as a global multiplier , which means that the requirement on its absolute accuracy is much less severe than in the cases where it has to be subtracted from something else .
another possibility worth mentioning is with a ring to the maximum at latitude @xmath115 and the other horizontal at latitude @xmath116 .
the equations are : @xmath117 \nonumber \\ f_{h2}&\simeq & s_2\omega_\oplus ( 1 + 2b)\cos\theta_2 \nonumber \label{mh}\end{aligned}\ ] ] the formal solution is now : @xmath118 finally the convenient combination : a-3b [ solhor ] again the impact of the accuracy on @xmath108 is reduced , but the second ring needs to be laid in the horizontal plane within a @xmath107 or so .
this case has already been treated in sect.s [ ii ] and [ exptheor ] .
a constraint for this configuration is to insure that the normals to both rings lay in the meridian plane .
it is a manageable condition , since the meridian is a symmetry plane ; the additional contributions coming for an out - of - the plane component of @xmath21 would be proportional to the cosine of an angle @xmath119 whose value on the plane would be @xmath120 . in practice , as for the orientation to the maximum , an uncertainty of @xmath121 rad on @xmath119 would affect the measured frequencies at the @xmath122 rad / s level only . the retrievable information is contained in system ( [ sist ] ) and its solutions .
recalling the discussion made in sect .
[ exptheor ] we remark again that in principle the only ring at the maximum could be enough ( eq . ( [ secap ] ) ) , provided one is able to measure the angle to the radial direction @xmath23 , then exploiting the relation between @xmath79 and @xmath80 . in the discussions presented so far ,
always appear frequencies @xmath7 , either as expected values obtained from the knowledge of the other physical parameters and the angles , or as input data to find the gr quantities @xmath79 and @xmath80 .
it must be recalled that the @xmath7 s are not the measured frequencies given by the rls system , since in that experimental output other effects are contained too
. the measured frequencies also account for the movements of the axis of the earth , originated from external perturbations ; the rotational motions of the crust of the planet ; the local movements of the ground and of the laboratory originating from various possible causes , etc .
last but not least there are the instabilities of the laser , down to the shot noise .
the random components tend to average to zero , prolonging the duration of each experimental run ; the other contributions need to be identified and then can be described as best as possible by appropriate modeling . in any case whatever is not earth rotation and gr must carefully be subtracted from the raw data in order to obtain the frequencies to be used in our formulae .
the cleaning process must of course be as accurate as the required final accuracy in the @xmath7 s .
we have analyzed and discussed various possible configurations and orientations of up to three rls , located either in one place or at different latitudes .
we have laid down the basic equations to be used in the various cases and shown the level of accuracy required , if the gr effects are aimed at .
a convenient choice is to have one ring oriented to the maximum signal so that the orientation accuracy can be relaxed to approximately @xmath76 @xmath123rad .
apart from the above , the critical parameters are the angles , which , with the exception of the orientation at the maximum , should be known within the prad .
so far , the physical principles and constraints have been treated , laying down the fabric of possible experiments and the constraints to be abided by in order to give relevant results .
next come the measurement strategies and all the features of a real experiment .
the behaviour of the laser and of the resonant loop need to be discussed , then the behaviour of the mirrors used to obtain the closed path for light , back scattering effects , etc .. the mechanical and thermal stability of the whole setup have to be taken into account , and so on .
it is clear that , for practical reasons , it will be convenient to work with a redundancy of rings , allowing to mutually control and quantize the intrinsic uncertainties .
all these aspects are under consideration and will be treated in a technical paper , now in preparation . for sure , the experiment is not an easy one , but both the general considerations we have presented , and the ongoing technological trends tell us that the objective of using ring laser arrays for fundamental physics is a viable one and the ginger project and collaboration moves on the right path .
an additional bonus for this type of measurements is that they also provide a good amount of important information on the behaviour of the geophysics and geodesy of our planet .
from the general definitions we may express @xmath80 as a function of @xmath79 : @xmath124 . the present knowledge about the size and shape of the earth is thoroughly exposed in @xcite and the most recent data are in @xcite .
it is : iers annual report 2014 . edited by wolfgang r. dick and daniela thaller .
international earth rotation and reference systems service , central bureau .
frankfurt am main : verlag des bundesamts fr kartographie und geodsie , 2015 . 165 pp . ,
isbn 978 - 3 - 86482 - 084 - 7 ( print version ) , in press .
online at https://www.iers.org/iers/en/publications/annualreports/annualreport2014.html?nn=94830 | the paper discusses the optimal configuration of one or more ring lasers to be used for measuring the general relativistic effects of the rotation of the earth , as manifested on the surface of the planet .
the analysis is focused on devices having their normal vector lying in the meridian plane .
the crucial role of the evaluation of the angles is evidenced .
special attention is paid to the orientation at the maximum signal , minimizing the sensitivity to the orientation uncertainty .
the use of rings at different latitudes is mentioned and the problem of the non - sfericity of the earth is commented . |
position and momentum of a _ classical _ non - relativistic particle are intimately linked since the momentum variable generates spatial translations .
mathematically , this important relation is embodied in the structure of the galilei group .
it turns out to be even more fundamental for a _ quantum mechanical _ particle , where it takes the form of the commutation relation of its position and momentum operators .
the associated heisenberg - weyl group of phase - space translations continues to be relevant for quantum systems with only a _
finite _ number of orthogonal states , providing a basis of the space @xmath2 . for each dimension
@xmath3 , there is a set of unitary operators which give rise to a discrete equivalent of the heisenberg - weyl group schwinger60 .
physicists would expect the state spaces @xmath2 to be structurally identical , at least with respect to properties closely related to the heisenberg - weyl group .
it thus comes as a surprise that the heisenberg - weyl group allows one to construct @xmath0 so - called mutually unbiased ( mu ) bases of the space @xmath2 if @xmath1 is the power of a prime number @xcite while the construction fails in all other composite dimensions .
no other successful method to construct @xmath4 mu bases in all dimensions is known @xcite . given @xmath0 orthonormal bases in the space @xmath5 , they are _ mutually unbiased _ if the moduli of the scalar products among the @xmath6 basis vectors take these values : @xmath7 where @xmath8 .
such _ complete _ sets of mu bases are ideally suited to reconstruct quantum states @xcite while sets of up to @xmath0 mu bases have applications in quantum cryptography @xcite and in the solution of the mean king s problem aharonov+01 , for example .
the methods to construct complete sets of mu bases typically deal with all prime or prime - power dimensions simultaneously .
they either make use of the heisenberg - weyl group @xcite , exploit identities from number theory @xcite , or they are couched in the language of finite fields @xcite .
all these methods are _ constructive _ and effectively lead to the same bases .
the existence of other , inequivalent sets of complete mu bases remains unclear . in this paper , we choose a different method to study mu bases in dimensions two to five .
it allows us to directly conclude that the corresponding complete sets of @xmath0 mu bases are unique ( up to some irrelevant equivalence , cf . below ) .
what is more , we are also able to exhaustively list all inequivalent classes of @xmath1 or less mu bases .
this approach is attractive because it uses elementary methods only , except in dimension five . in order to present the details of this method we need to briefly discuss the relation between mu bases and complex hadamard matrices .
each mu basis in the space @xmath2 consists of @xmath1 orthogonal unit vectors which , collectively , will be thought of as a unitary @xmath9 matrix .
two ( or more ) mu bases thus correspond to two ( or more ) unitary matrices , one of which can always be mapped to the identity @xmath10 of the space @xmath5 , using an overall unitary transformation .
it then follows from the conditions ( [ mub conditions ] ) that the remaining unitary matrices must be complex hadamard matrices : the moduli of all their matrix elements equal @xmath11 .
this representation of mu bases links their classification to the classification of complex hadamard matrices @xcite .
it is , for example , possible to list all pairs of mu bases @xmath12 in @xmath2 once all complex hadamard matrices are known . using this observation as a starting point
, we will extend the classification from pairs to sets of @xmath13 mu bases .
as complex hadamard matrices have been classified for @xmath14 , we expect to obtain an exhaustive list of sets of @xmath15 mu bases in these low dimensions .
the approach we take is inspired by recent works aimed at dimension six grassl04,brierley+09,jaming+09 .
starting with a pair of mu bases @xmath12 we will find _ all _ vectors @xmath16 which are mu to @xmath10 and @xmath17 .
only these vectors represent candidates to form additional bases and by analysing their inner products we will be able to obtain _ all _ mu bases in low dimensions . in dimension six
it is not trivial to determine all candidate vectors .
it becomes necessary to use grbner bases @xcite or to discretise the underlying space @xcite . for @xmath18 , however , we find that elementary properties of the complex plane are sufficient to solve ( most of ) the relevant equations in closed form . the task to find all mu bases
is complicated by the fact that , actually , many sets of apparently different mu bases are identical to each other . for the desired classification ,
it is sufficient to enumerate all _ dephased _ sets of @xmath19 mu bases .
this _ standard form _
tadej+06 is given by @xmath20 where @xmath10 is the identity in @xmath2 and the other matrices are complex hadamard matrices of a particular form : the components of the first column of the matrix @xmath21 are given by @xmath22 , and the first row of each hadamard matrix @xmath23 has entries @xmath22 only ( see eqs .
( [ f5 ] ) and ( [ d=5 complete set ] ) for an explicit example with @xmath24 ) .
the possibility of dephasing is based on the notion of _ equivalence classes _ for mu bases , explained in more detail in appendix a. roughly speaking , two mu bases are equivalent to each other if one can be obtained from the other by changing the overall phases of individual vectors and by permuting them .
the results of this paper have been arranged as follows . in sec .
[ dim 23 ] we deal with dimensions two and three .
the complete list of sets of mu bases in dimension four is derived in sec .
[ dim 4 ] .
then , all sets of mu bases of @xmath25 are constructed , and in sec .
[ concl ] we summarize and discuss our results .
in this section , we construct all sets of mu bases in dimensions two and three using only simple properties of the complex plane . the direct approach to construct all mu bases for @xmath28 in sec .
[ dim 4 ] will be based on similar arguments .
the matrices consisting of the eigenvectors of the heisenberg - weyl operators form a set of three mu bases in dimension two which are unique up to the equivalences specified in appendix a. we present a simple proof of this well - known fact .
let us begin by noting that there is only one dephased complex hadamard matrix in @xmath26 ( up to equivalences ) , the discrete @xmath29 fourier matrix @xmath30 a vector @xmath31 is mu to the standard basis @xmath10 ( constructed from the eigenstates of the @xmath32-component of a spin @xmath33 ) if its components have modulus @xmath22 . applying the transformation given in eq .
( diagunitaries ) , the dephased form of such a a vector reads @xmath34 , with a real parameter @xmath35 $ ] .
the vector @xmath16 is mu to the columns of @xmath36 if the phase @xmath37 satisfies two conditions , @xmath38 these equations hold simultaneously only if @xmath39 .
thus , there are only two vectors which are mu to both @xmath10 and @xmath36 , given by @xmath40 .
since this is a pair of orthogonal vectors , they form a hadamard matrix @xmath41 and , therefore , the three sets @xmath42 represent all ( equivalence classes of ) one , two , or three mu bases in dimension two . in dimension three
there is also only one dephased complex hadamard matrix up to equivalence .
it is given by the @xmath43 discrete fourier matrix@xmath44 defining @xmath45 again , we search for dephased vectors @xmath46 , @xmath47 , which are mu with respect to the matrix @xmath48 .
this leads to the following three conditions @xmath49 removing an overall factor of @xmath50 they can be rewritten @xmath51 where @xmath52 . by considering a plot in the complex plane , fig .
[ threeplane ] , we see that these three equations hold simultaneously only if two of the cosine terms are equal .
this implies that the only possible values of the parameter @xmath53 are @xmath54 or @xmath55 , leading to the requirement @xmath56 .
consequently , the eqs .
( d=3 constraints ) have exactly six solutions which give rise to vectors @xmath57 examining their inner products shows that there is only one way to arrange them ( after normalization ) into two orthonormal bases , namely @xmath58 and @xmath59 .
it is useful to note that one can write @xmath60 where @xmath61 is a diagonal unitary matrix with entries identical to the components of the vector @xmath62 , i.e. the first column of @xmath63 .
the triples obtained from adding either @xmath64 or @xmath65 to the pair @xmath66 are equivalent , @xmath67 as follows from first applying the unitary @xmath68 globally from the left , rephasing the first basis with @xmath69 , and finally rearranging the last two bases .
we therefore conclude that the sets constitute a complete classification of all sets of mu bases in dimension @xmath27 . ) in the complex @xmath32-plane .
the real numbers @xmath70 and @xmath71 each represent one of the three numbers @xmath72 and @xmath73 ; it follows that at least two of these three expressions must be equal.,scaledwidth=50.0% ]
in dimension @xmath28 , a one - parameter family of complex hadamard matrices exists , @xmath74\ , . \label{d=4 fourier hadamard}\ ] ] it is elementary to show that all @xmath75 complex hadamard matrices are equivalent to a member of the family @xmath76 . when @xmath77 , the resulting matrix is the discrete fourier transform @xmath78 on the space @xmath79 , with matrix elements given by @xmath80 .
the matrix @xmath81 is equivalent to a direct product of the matrix @xmath36 with itself while for other values of @xmath82 it can be written as a hadamard product of @xmath78 with an @xmath82-dependent matrix .
after dephasing , any vector mu to the standard basis takes the form @xmath83 @xmath84 where @xmath85 . for convenience
, we will use an _ enphased _ variant of @xmath16 . multiplying through by the phase factor @xmath86 and defining
@xmath87 $ ] , @xmath88 $ ] , and similarly for @xmath89 , we consider the parametrization @xmath90 @xmath91 instead . the conditions for @xmath92 to be mu to the columns of @xmath76 lead to four equations , @xmath93 where complex numbers @xmath94 have been introduced .
we will now construct all solutions of these equations as a function of the value of @xmath82 .
we treat the cases ( i ) @xmath95 , ( ii ) @xmath96 , and ( iii ) @xmath97 separately since the eqs .
( [ zetaplus ] ) and ( [ zetaminus ] ) take different forms for these values .
( [ zetaminus ] ) simplify to the pair @xmath99 which only hold simultaneously if @xmath100 so that eqs .
( [ zetaplus ] ) are satisfied automatically .
thus , solutions exist for any value of @xmath82 whenever @xmath101 , and the resulting vectors can be written as @xmath102 , with @xmath103 $ ] .
it will be convenient to divide this family of states into two sets,@xmath104 introducing @xmath105 and @xmath106 .
( [ zetaplus ] ) and ( [ zetaminus ] ) now reverse their roles : the conditions @xmath108 require @xmath109 , with ( zetaplus ) being satisfied since @xmath110 follows immediately .
hence , there is another one - parameter family of mutually unbiased vectors for all values of @xmath82 if @xmath111 .
this family can be written as @xmath112 , @xmath113 $ ] , after dephasing and absorbing a factor of @xmath114 in the definition of the phase , @xmath115 .
again , we express these solutions as a set of pairs , @xmath116 where @xmath117 and @xmath118 . a plot in the complex plane ( see fig .
[ fourplane ] ) reveals that one must have @xmath120 if eqs .
( [ zetaplus ] ) are to hold with @xmath121 .
thus , the real part of @xmath122 vanishes , @xmath123 with @xmath124 being the only acceptable solution : the other solution , @xmath125 leads to @xmath126 , producing a contradiction since @xmath127 .
thus , using @xmath124 , we obtain @xmath128 find the following relation between @xmath53 and @xmath129 : @xmath130 similarly , eqs .
( [ zetaminus ] ) for @xmath131 imply that @xmath132 . using @xmath124 in the definition of @xmath133 , we find @xmath134 , so that @xmath135 the right - hand - side of this equation only vanishes if @xmath136 : both @xmath137 and @xmath138 would , according to ( [ final cond 1 ] ) , require @xmath139 which we currently exclude .
therefore , solutions to eqs .
( [ zetaplus],[zetaminus ] ) with @xmath127 or @xmath140 only exist for @xmath136 if a second relation between @xmath53 and @xmath129 holds , @xmath141 the form of the additional mu vectors is determined by eqs .
( [ final cond 1 ] ) and ( [ final cond 2 ] ) which have four solutions .
first , for @xmath142 we obtain mu vectors of the form @xmath143 @xmath144 or @xmath145 after dephasing . splitting this family into two subsets as before ,
we find @xmath146 similarly , the choice @xmath147 leads to two sets of dephased mu vectors , @xmath148 next , when proceeding in an entirely analogous manner for the remaining two choices @xmath149 and @xmath150 , we obtain the following four families of dephased vectors mu to @xmath151 , @xmath152 with @xmath153 . ) in the @xmath154-plane implying that , for @xmath155 , their solutions are given by @xmath156.,scaledwidth=40.0% ] knowing all vectors that are mu to both the identity and @xmath78 , we now determine those combinations which form other bases . to begin , consider the mu vectors @xmath158 in eqs .
( [ h1 and h2],[h3 and h4 ] ) which exist for all values of @xmath159 $ ] . calculating their inner products
, one finds that they only form an _ orthonormal _ basis of @xmath160 if @xmath161 and @xmath162 .
thus , for each value of @xmath82 , the pair @xmath163 may be complemented by a third mu basis taken from the two - parameter family @xmath164 in other words , there is a _ three parameter - family of triplets _ of mu bases @xmath165 in dimension @xmath28 .
this family was not known before .
if @xmath136 , additional mu vectors @xmath166 , and @xmath167 , have been identified , cf .
( [ k1 k2]-[j family ] ) . calculating the scalar products within each group
, one sees that two further orthonormal two - parameter bases emerge , @xmath168 if the conditions @xmath169 , @xmath170 , and @xmath171 , @xmath172 , respectively , are satisfied . no other combinations of the mu vectors can form inequivalent bases so that the matrices in eqs .
( h(y , z)-[k4 ] ) represent all possible choices of a mu basis . permuting appropriate rows and columns of the matrices @xmath173 and @xmath174 transforms them into @xmath175
; thus , the triples @xmath176 and @xmath177 are equivalent to @xmath178 .
let us begin by noting that sets of four mu bases can not exist away from @xmath179 .
no two matrices @xmath180 and @xmath181 are mu since @xmath182 only hold if @xmath183 however , these equations have no solution for any values of @xmath184 and @xmath185 .
a similar argument shows that there are no values of @xmath32 and @xmath186 such that the matrices @xmath180 and @xmath187 are mu .
we now show that for @xmath136 the bases @xmath188 and @xmath189 give rise to four and five mu bases if the free parameters are chosen appropriately .
an argument similar to the one just presented shows that no two bases within either the family @xmath190 or @xmath191 are mu .
thus , any quadruple of mu bases must contain bases from different families . the inner products @xmath192 and
@xmath193 have modulus @xmath33 if there are values for @xmath184 and @xmath194 such that the equations @xmath195 hold simultaneously . upon introducing a factor of @xmath196 , eqs .
( d=4 constraints ) are equivalent to the constraints @xmath197 consequently , one must have @xmath198 , and thus @xmath199 or @xmath200 since @xmath201 .
an entirely analogous argument restricts the values of @xmath202 and @xmath32 : checking the inner products @xmath203 etc .
tells us that the matrices @xmath204 and @xmath190 are mutually unbiased only if @xmath205 .
we also find that the pairs @xmath206 @xmath207 and @xmath208 are mu only when all six parameters take the value @xmath209 we are now in the position to list all possible sets of mu bases in @xmath210 beyond @xmath211 , @xmath212 there is one three - parameter family of _ triples _ consisting of the one - parameter fourier family @xmath76 combined with two - parameter set @xmath204 ; neither @xmath190 nor @xmath191 give rise to other triples since each of these sets of hadamard matrices is equivalent to @xmath213 . the three - dimensional set ( [ d=4 list ] ) of mu bases in dimension @xmath28 may be visualized as a cuboid defined by @xmath214 and @xmath215 .
the reduction in the parameter range of @xmath32 is due to the equivalence @xmath216 which follows from an overall complex conjugation .
each of the points in the cuboid corresponds to one triple while both the quadruple and the quintuple are located at the point , @xmath217 .
only one set of _ four _ mu bases exists , @xmath218 , since the other two candidates obtained by combining @xmath219 with either @xmath220 or @xmath221 are permutations of this quadruple .
finally , there is a unique way to a construct _ five _ mu bases which is easily seen to be equivalent to the standard construction of a complete set of mu bases in dimension four .
as in dimensions two and three , there is a unique choice of a @xmath223 dephased complex hadamard matrix @xcite , @xmath224 equal to the discrete @xmath223 fourier matrix , with @xmath225 denoting a fifth root of unity .
the uniqueness of @xmath226 is obtained by an analytical method that is far from elementary @xcite .
we have not found an elementary method to obtain a list of all vectors which are mu to the fourier matrix @xmath227 .
instead , we will rely on earlier work @xcite where those vectors have been constructed _ analytically _ by means of a computer program .
the vector @xmath228 is mu to @xmath226 if it satisfies the conditions @xmath229defining @xmath230 . according to @xcite ,
the solutions of these equations give rise to 20 vectors which can be arranged in four mu bases , @xmath231to obtain this result , eqs .
( [ d=5 constraints ] ) have been expressed as a set of coupled quadratic polynomials in eight real variables . using an implementation @xcite of buchberger s
algorithm @xcite on the computer program maple @xcite , a grbner basis of these equations has been constructed which leads to the 20 vectors given by the columns of the four hadamard matrices above .
it is important to note that no other solutions of eqs .
( [ d=5 constraints ] ) exist , a result which does _ not _ follow from the known methods to construct a complete set of six mu bases in @xmath232 .
each of the four matrices in ( [ d=5 complete set ] ) is related to the fourier matrix in a remarkably simple manner . in analogy to the unitary diagonal matrix used in eq .
( [ powers of d ] ) , define a diagonal unitary matrix @xmath233with entries given by the first column of @xmath234 and you find that @xmath235using this observation , we can express the unique complete set of six mu bases for dimension @xmath222 as follows @xmath236which will be useful later on .
next , we proceed to classify all smaller sets of mu bases of @xmath237 by combining subsets of the four hadamard matrices @xmath238 in ( [ d=5 complete set ] ) with the pair @xmath239 . for clarity ,
we now list the set of inequivalent classes which we will obtain .
in addition to the pair @xmath240 and the complete set given in ( [ complete d=5 ] ) there are _ two _ inequivalent triples as well as _ one _ quadruple and _ one _ quintuple : @xmath241 select one of the four matrices given in ( [ d=5 complete set ] ) and adjoin it to the pair @xmath239 .
you obtain four triples of mu bases with two immediate equivalences , namely , @xmath242 on the one hand , and @xmath243 on the other .
the equivalence ( [ triples in c5 a ] ) follows from multiplying the set @xmath244 with @xmath245 from the left , rephasing the first basis with @xmath246 from the right , using @xmath247 and swapping the last two matrices .
a similar argument establishes the equivalence ( [ triples in c5 b ] ) , using @xmath248 instead of @xmath245 .
thus , it remains to check whether the triples @xmath249 and @xmath250 @xmath251 are equivalent to each other .
it turns out that these two triples are , in fact , _
inequivalent_. more explicitly , this means that no unitary matrix @xmath252 and no monomial matrices @xmath253 and @xmath254 can be found which would map @xmath255 into @xmath256 according to @xmath257 a proof of this statement is given in appendix b. there are six possibilities to form quadruples by selecting two of the four matrices in eq .
( [ d=5 complete set ] ) and adding them to the pair @xmath240 . recalling that @xmath258 , we identify the following equivalences which relate three quadruples each , @xmath259 and @xmath260 to show the first equivalence in eq .
( [ d=5equivquintquad ] ) , for example , multiply its left - hand - side with @xmath248 from the left , use the identity @xmath261 and rearrange the bases appropriately .
the other equivalences follow from analogous arguments .
thus , there are at most two inequivalent sets of four mu bases in @xmath237 , with representatives @xmath262 and @xmath263 , say .
interestingly , these two classes of mu bases are _ equivalent _ to each other leaving us with a single equivalence class of quadruples in dimension five , with representative @xmath264 , say . to show this equivalence
, we multiply the first quadruple with the adjoint of @xmath265 from the left @xmath266using the identity @xmath267 , with some permutation matrix @xmath268 , and swapping the first two bases . the action of @xmath269 on the other two elements is surprisingly simple : the hadamard matrix @xmath234 is mapped to itself , @xmath270up to a monomial matrix @xmath271 , while @xmath272 is sent to @xmath273 , @xmath274again up to some monomial matrix @xmath275 .
both relations simply follow from working out the product on the left and factoring the result , e.g. @xmath276where the @xmath277 entry of the diagonal matrix @xmath278 is given by the sum of the @xmath277 column of @xmath272 , denoted by @xmath279 , and @xmath268 permutes the columns . using these identities in eq .
( [ d=5last equivalence ] ) we find that @xmath280the two quadruples are equivalent .
four sets of mu bases can be obtained by adding any three of the four matrices in eq .
( [ d=5 complete set ] ) to the pair @xmath239 .
it is not difficult to show that the four resulting sets of quintuples are equivalent to each other .
thus , there is effectively only one possibility to choose five mu bases in @xmath237 , with representative @xmath281 .
let us show now that this representative , which has been obtained by leaving out @xmath282 , is equivalent to the set @xmath283 , for example .
indeed , the equivalence @xmath284 follows immediately from multiplying the second set by @xmath246 from the left and using @xmath285 , @xmath286 reordering the set of five matrices on the right reveals the desired equivalence with the quintuple @xmath287 .
effectively , the four matrices different from @xmath10 undergo a cyclic shift under multiplication with @xmath246 , and the remaining equivalences follow from shifts induced by @xmath68 and @xmath248 , respectively .
we have constructed all inequivalent sets of mutually unbiased bases in dimension two to five .
our approach is based on the fact that all complex hadamard matrices are known in these dimensions . for dimensions up to @xmath28 ,
elementary arguments suffice to classify the existing sets of mu bases while dimension five requires some analytic results which have been found earlier using algebraic computer software .
.the number of inequivalent mu bases for dimensions two to six where @xmath288 denotes a @xmath289-parameter set ; see text for details . [ cols=">,^,^,^,^,^,^,^",options="header " , ] the first four columns of table [ mu classes ] summarize the results obtained in this paper .
all _ pairs _ of mu bases in dimensions two to five are listed in the first row , effectively reflecting the known classification of inequivalent hadamard matrices ; a continuous ( one - parameter ) set of inequivalent mu pairs only exists in dimension four . the main results concern _
triples _ of mu bases in dimension four where we find a _ three - parameter family _ and in dimension five where we obtain _ two _ inequivalent triples . finally , we have shown that there is only one class of both mu _ quadruples _ and mu _ quintuples _ in dimensions four and five . in all dimensions considered ,
there is a unique @xmath1-tuple which can be extended to a complete set of @xmath0 mu bases using a construction presented in @xcite .
the last column of table [ mu classes ] contrasts these results with dimension six where the classification of all complex hadamard matrices is not known to be complete .
the first entry shows that there is a three - parameter family of pairs of mu bases @xcite ( it has been conjectured that the parameter space has , in fact , four dimensions @xcite ) .
furthermore , it is possible to construct a two - parameter family of triples @xcite using an idea taken from @xcite .
there is strong numerical @xcite and analytical @xcite evidence to suggest that triples are the largest sets of mu bases in dimension six but this problem remains open .
the notion of equivalence used in this paper ( see appendix [ app a ] ) is mathematical in nature ; it captures all possible operations that leave invariant the conditions ( [ mub conditions ] ) for two bases to be mutually unbiased .
motivated by experiments , there is a finer equivalence of complete sets of mu bases based on the entanglement structure of the states contained in each basis @xcite . for dimensions that are a power of two , a complete set of mu bases can be realized using pauli operators acting on each two - dimensional subsystem .
two sets of mu bases are then called equivalent when they can be factored into the same number of subsystems . for @xmath290
this notion of equivalence also leads to a unique set of @xmath0 mu bases .
however , for @xmath291 complete sets of mu bases can have different entanglement structures even though they are equivalent up to an overall unitary transformation @xcite .
the traditional approach to find _ complete _ sets of mu bases in prime - power dimensions via the heisenberg - weyl group or by using finite fields is constructive and , therefore , does not exclude the existence of other inequivalent complete sets .
the present approach is , in contrast , _ exhaustive _ : we are able to affirm that the known complete sets for @xmath292 are unique ( up to equivalence ) . their uniqueness has been shown earlier for @xmath293 @xcite while @xcite contains a proof for @xmath294 in a lie algebraic setting .
we find it appealing that it is possible to prove the uniqueness of complete sets of mu bases in low dimensions by _ elementary _ methods .
the authors would like to thank m. kibler for pointing out the complex conjuation equivalence relation given in eq .
( [ conjugationequiv ] ) .
sb and sw gratefully acknowledge financial support through qnet , a network funded by the epsrc , and ib has been supported by the swedish research council .
99 j. schwinger , proc .
, * 46 * , 560 , ( 1960 ) i. d. ivanovi , j. phys .
a * 14 * , 3241 ( 1981 ) w. k. wootters and b. d. fields , ann .
( n.y . ) * 191 * , 363 ( 1989 ) c. archer , j. math .
phys . * 46 * , 022106 ( 2005 ) m. planat , h. rosu , and s. perrine , found .
phys . * 36 * , 1662 ( 2006 ) a. roy , a. scott , math . phys . * 48 * , 072110 ( 2007 ) n. cerf , m. bourennane , a. karlsson , and n. gisin : phys . rev
. lett . * 88 * , 127902 ( 2002 ) s. brierley : _ quantum key distribution highly sensitive to eavesdropping _ , arxiv:0910.2578 y. aharonov and b. g. englert , z. naturforsch . a : phys .
* 56a * , 16 ( 2001 ) s. bandyopadhyay , p. o. boykin , v. roychowdhury and f. vatan , algorithmica , * 34 * , 512 , ( 2002 ) a. klappenecker and m. rtteler , lect .
notes comput . sc .
* 2948 * , 137 ( 2004 ) t. durt : _ a new expression for mutually unbiased bases in prime power dimensions _
, arxiv : quant - ph/0409090 w. tadej and k. yczkowski , open systems and infor .
* 13 * 133 - 177 ( 2006 ) _ on - line catalogue of known hadamard matrices _ maintained by w. tadej and k. yczkowski at ` http://chaos.if.uj.edu.pl/karol/hadamard/ ` m. grassl , _ on sic - povms and mubs in dimension 6 _ , in : _ proc .
erato conference on quantum information science ( equis 2004 ) _ , j. gruska ( ed . ) s. brierley and s. weigert , phys .
a * 79 * , 052316 ( 2009 ) p. jaming , m. matolcsi , p. mra , f. szllsi and m. weiner , j. phys .
theor . * 42 * 245305 ( 2009 ) u. haagerup , _ orthogonal abelian * -subalgebras of the @xmath295 matrices and cyclic @xmath296-roots _ , in : _ proc .
operator algebras and quantum field theory ( rome 1996 ) _ , s. dplicher ( ed . ) salsa : solvers for algebraic systems and applications ; software available from http://fgbrs.lip6.fr/salsa/ b. buchberger , _ an algorithm for finding the basis elements of the residue class ring of a zero dimensional polynomial ideal .
dissertation , university of innsbruck ( 1965 ) ( english translation by m. abramson in j. symb . comp . *
41 * , 471 ( 2006 ) ) b. buchberger , _ grbner bases and applications _ in _ lms lecture notes series , 251 _ ( 1998 ) , b. buchberger and f. winkler ( eds . ) maple 11 , waterloo maple inc .
waterloo , ontario , canada m. weiner : _ a gap for the maximum number of mutually unbiased bases _ , arxiv:0902.0635 b. karlsson : _ three - parameter complex hadamard matrices of order 6 _ , arxiv:1003.4177 f. szllsi , _ a two - parameter family of hadamard matrices of order 6 induced by hypocycloids _
, arxiv:0811.3930v2 a. skinner , v. newell , and r. sanchez , j. math .
* 50 * , 012107 ( 2009 ) g. zauner , _ quantendesigns .
grundzuge einer nichtkommutativen designtheorie .
_ phd thesis , university of wien , 1999 .
p. butterley and w. hall , phys .
a * 369 * , 5 ( 2007 ) s. brierley and s. weigert , phys .
a * 78 * , 042312 ( 2008 ) j. lawrence , c. brukner , and a. zeilinger , phys .
a * 65 * , 032320 ( 2002 ) j. romero , g. bjrk , a. klimov , and l. snchez - soto , phys .
a * 72 * , 062310 ( 2005 ) .
ericsson ( unpublished , 2004 ) p.o .
boykin , m. sitharam , p.h tiep and p. wocjan , quantum inf . comp . * 7 * , 371 ( 2007 )
many sets of mu bases are identical to each other . to simplify the enumeration of all sets of mu bases we introduce _ equivalence classes _ and a _ standard form _ of sets of mu bases .
each set of @xmath19 mu bases in @xmath5 corresponds to a list of @xmath297 ( with @xmath298 ) complex matrices @xmath299 , @xmath300 of size @xmath301 .
two such lists @xmath302 and @xmath303 are _ equivalent _ to each other , @xmath304 if they can be transformed into each other by a succession of the following four transformations : 1 .
an _ overall unitary _
transformation @xmath252 applied from the left , @xmath305 which leaves invariant the value of all scalar products ; 2 .
@xmath19 _ diagonal unitary _
transformations @xmath306 from the right which attach phase factors to each column of the @xmath19 matrices , @xmath307 these transformations exploit the fact that the overall phase of a quantum state drops out of from the conditions of mu bases ; 3 . @xmath19 _ permutations _ of the elements within each basis , @xmath308 which amount to relabeling the elements within each basis by means of unitary permutation matrices @xmath309 satisfying @xmath310 ; 4 . _ pairwise exchanges _ of two bases , @xmath311 which amounts to relabeling the bases . 5 .
an _ overall complex conjugation _
@xmath312which leaves the values of all scalar products invariant .
these equivalence relations allow us to _ dephase _ a given set of mu bases . the resulting _ standard form _
@xmath313 @xmath314 is characterized by four properties : ( i ) the first basis is chosen to be the standard basis of @xmath5 described by @xmath315 , where @xmath10 is the @xmath301 identity matrix ; ( ii ) the remaining bases are described by ( complex ) _ hadamard _ matrices : each of their matrix elements has modulus @xmath22 ; ( iii ) the components of the first column of the matrix @xmath21 are given by @xmath22 ; ( iv ) the first row of each of the hadamard matrices @xmath21 to @xmath316 has entries @xmath22 only .
we show that the two classes of triples of mu bases given by @xmath317 and @xmath318 are _ inequivalent_. in a first step , we explain that it is sufficient to search for equivalence transformations generated by matrices of a special form . in a second step
we show that a contradiction arises if one assumes that the triples @xmath319 and @xmath256 are equivalent .
let us begin with a general remark about the structure of equivalence classes of sets of mu bases @xmath320 of @xmath321 for all @xmath322 . for convenience ,
we assume that the first basis equals the identity , i.e. the set is given in standard form .
as explained in appendix [ app a ] all sets of mu bases equivalent to @xmath323 are obtained as follows , @xmath324 with a unitary @xmath252 and @xmath19 monomial matrices @xmath325 being a product of diagonal unitaries with permutation matrices ; to keep the notation simple we do not reorder the ( @xmath326 ) bases within @xmath327 .
for the set @xmath328 to be in standard form , one of the bases in @xmath323 , say @xmath329 , must be mapped to the identity . as a consequence , the overall unitary transformation @xmath252 must have a particular form , namely @xmath330 where @xmath331 is some monomial matrix and @xmath329 is one of the matrices contained in the set @xmath323 . in view of eq .
( [ u is special ] ) we are lead to determine the action of @xmath332 and @xmath333 on the triple @xmath334 as well as the action of @xmath269 and @xmath335 on the triple @xmath256 .
it turns out that both triples are _ invariant _ under these global transformations as we have the equivalences @xmath336and @xmath337the first equivalence in ( [ invariant triple 1 ] ) follows from using @xmath338 and eq .
( [ fdaggeraction2 ] ) while the second one also requires the identity @xmath339with some monomial matrix @xmath271 .
the equivalences ( [ invariant triple 2 ] ) are derived in a similar way .
consequently , we can always remove the effect of the matrices @xmath340 in the global transformations ( [ u is special ] ) which leaves us with @xmath341 where @xmath342 and @xmath254 are monomial matrices , and up to rearranging terms .
the non - zero entries of the monomial matrix @xmath331 must , in fact , be fifth roots of unity but we will not need this fact has a nonzero element different from a fifth root , say @xmath343 .
this makes it impossible to transform @xmath319 into standard form using right multiplication by monomial matrices unless the other nonzero elements of @xmath331 also equal @xmath344 .
it follows that @xmath331 must be a permutation matrix @xmath268 apart from a phase factor , @xmath345 .
thus the matrices @xmath346 must have a common factor of @xmath347 which , however , is irrelevant for the definition of mu bases ] . using the restricted transformations shown in eqs .
( [ limited trfs ] ) , the triples @xmath348 and @xmath349 are equivalent to each other only if either @xmath350 or @xmath351 hold for some monomial matrices @xmath352 and @xmath254 . the choice @xmath353 in eqs .
( [ limited trfs ] ) ensures that the identity will be mapped to the identity .
( [ poss 1 ] ) will now be shown to imply the identity @xmath354 for some monomial matrix @xmath271 while @xmath355 is a diagonal matrix with fifth roots of unity as nonzero entries , _ not _ proportional to the identity , @xmath356 .
however , eq . ( [ n removed ] ) only holds if @xmath355 _ is _ a multiple of the identity .
this contradiction implies that there are no matrices @xmath357 such that eqs .
( [ poss 1 ] ) hold . since eqs .
( [ poss 2 ] ) also imply eq .
( [ n removed ] ) with a ( possibly different ) diagonal matrix @xmath358 , the triples @xmath319 and @xmath256 can not be equivalent .
use @xmath359 , to express the second equation in ( poss 1 ) as @xmath360 introducing @xmath361 .
thus , the matrix @xmath362 is obtained from @xmath246 by reordering its diagonal elements according to the permutation @xmath268 defined via @xmath363 , with some unitary diagonal matrix @xmath364 . combining this equation with the first one in ( poss 1 ) leads to @xmath365 , or @xmath366 which is identical to ( [ n removed ] ) upon defining @xmath367 and @xmath368 which , as a product of two monomial matrices , is another monomial matrix .
since no permutation of the elements on the diagonal of @xmath369 produces the inverse of @xmath68 or a multiple thereof , we have @xmath370 . using the pair ( [ poss 2 ] ) instead of ( [ poss 1 ] ) also leads to an equation of the form ( [ n removed ] ) with @xmath371 replaced by @xmath372 which
, however , can not be a multiple of the inverse of @xmath68 , leading again to @xmath370 .
we now show that eq .
( [ n removed ] ) only holds if the matrix @xmath355 is proportional to the identity .
write the monomial matrix @xmath271 in ( [ n removed ] ) in the form @xmath373 where @xmath268 is a permutation matrix and @xmath374 is a _
diagonal _ matrix with entries having modulus one only .
denoting the inverse of @xmath374 by @xmath375 , eq .
( [ n removed ] ) takes the form @xmath376 let us write @xmath377 with phase factors @xmath378 , etc , and similarly for @xmath379 , and consider the simplest case @xmath380 .
then the matrix relation ( [ n removed new ] ) reads explicitly @xmath381 the conditions resulting from the first row immediately imply that the elements on the diagonal of @xmath379 are all equal to @xmath382 , or @xmath383 .
the conditions of the first column imply that the matrix @xmath355 is also a multiple of the identity , namely @xmath384 .
this contradicts the fact that the matrix @xmath355 is different from a multiple of the identity .
let us now drop the restriction the @xmath385 .
the effect of @xmath268 acting on @xmath227 from the right is to permute its columns .
the first row of @xmath386 will not change under this operation . under the action of @xmath268
, the first column will either stay where is is or it will be mapped to one of the four others . in the first case , we can immediately apply the argument given above to derive a contradiction . in the second case , it it straightforward to see that a similar argument still applies involving the first row of the matrices and that column which is the image of the first column .
thus , all possible choices of the monomial matrix @xmath271 in ( [ n removed ] ) require @xmath355 to be a multiple of the identity which it is not .
finally , we consider the action of an overall complex conjugation ( conjugation ) on either of the triples .
we find that the set of three mu bases , @xmath387 remains invariant under complex conjugation @xmath388similarly , complex conjugation maps @xmath256 to itself , @xmath389 . in summary
, we have shown that the equivalence relations ( [ overallunitary ] ) to ( conjugation ) can not transform the triple @xmath319 into @xmath390 or vice versa , i.e. these triples are inequivalent . | all complex hadamard matrices in dimensions two to five are known .
we use this fact to derive all inequivalent sets of mutually unbiased ( mu ) bases in low dimensions .
we find a three - parameter family of triples of mu bases in dimension four and two inequivalent classes of mu triples in dimension five .
we confirm that the complete sets of @xmath0 mu bases are unique ( up to equivalence ) in dimensions below six , using only elementary arguments for @xmath1 less than five . |
coin flipping by telephone was first introduced with these words by manuel blum in 1981 : `` alice and bob [ ] have just divorced , live in different cities , want to decide who gets the car '' @xcite .
they agree that the best thing to do is to flip a coin , but neither of them trusts the other and they are unable to agree on a mutually trusted third party to do the flip for them .
more generally , coin - flipping protocols ( also known as `` coin - tossing '' ) can be used whenever two players need to pick a random bit even though it could be to the advantage of one of them ( or perhaps both ) to choose , or at least bias , the outcome of the protocol .
the original coin - flipping protocol introduced by blum is _ asynchronous _ in the sense that it consists of a sequence of rounds in which the two players alternate in sending messages to each other .
the security of blum s protocol depends on the assumed difficulty of factoring large numbers .
such an assumption is of course of little value in our quantum world , owing to peter shor s algorithm @xcite , but classical coin flipping can be based on more general one - way functions , which could potentially be immune to quantum attacks . nevertheless , any coin - flipping protocol that takes place by the asynchronous transmission of classical messages has the property that one of the has complete control over the outcome , given sufficient computing power . in the best case
, such protocols can be computationally secure , and even _ that _ depends on unproven computational complexity assumptions .
unconditionally secure classical coin - flipping protocols are possible in the _ synchronous _ model , in which the players are requested to send messages to each other with sufficient simultaneity to force their complete independence .
such protocols are called _ relativistic _ because special relativity must be invoked to prevent alice from to receive bob s message before choosing her own ( and vice versa ) .
relativistic protocols must be implemented carefully their security depends crucially on the physical distance between the players , and either of them could try to fool the other by pretending to be away than they really are .
such cheating can be thwarted if each player has a trusted agent near the other player @xcite . for the rest of this paper , we only consider asynchronous protocols and `` coin - flipping '' will systematically mean `` asynchronous coin - flipping '' . in _ quantum _ coin - flipping protocols , alice and bob
are allowed to exchange quantum states .
such protocols were first investigated in 1984 by charles h. bennett and gilles brassard @xcite . in that paper ,
a protocol was presented and it was shown that `` ironically [ it ] can be subverted by use of a still subtler quantum phenomenon , the einstein - podolsky - rosen paradox '' , making it the first use of entanglement @xcite in quantum cryptography .
we shall refer to it henceforth as the `` bb84 protocol '' ( not to be confused with the better - known quantum key distribution protocol introduced in the same paper ) .
the question was left open : can there be a perfect quantum coin - flipping protocol ?
the proof that this is impossible was given more than a decade later by hoi - kwong lo and hoi fung chau @xcite , whose result was further clarified by mayers , louis salvail and yoshie chiba - kohno @xcite .
nevertheless , if quantum coin - flipping protocols can not be perfect , can they at least be better than anything classically possible ? to make this question more precise
, we say that one player enjoys an if a cheating strategy exists by which that player could choose either bit and influence the outcome of the protocol to be that bit with probability at least @xmath0 , assuming that the other player follows the protocol honestly .
this definition is _ unconditional _ in the sense that we allow the would - be cheater to enjoy unlimited computational power and a technology limited only by the laws of physics .
the _ bias _ of a protocol is the largest value of @xmath1 so that at least one player an .
a _ perfect _ protocol would be one whose bias is @xmath2 , but they can not exist , classical or quantum . at the other end of the spectrum , a protocol
whose bias is @xmath3 is considered to be _ completely broken_. all classical protocols are completely broken by this definition , and so is the bb84 quantum protocol .
the question at the end of the previous paragraph was therefore : is there a quantum coin - flipping protocol whose bias is strictly less than @xmath3 ? = + 1 the first such protocol was discovered in 2000 by dorit aharonov , amnon ta - shma , umesh vazirani and andrew c .- c .
yao @xcite , who proved that the bias of their protocol ( atvy ) is at most ( without any claim concerning the tightness of their bound ) .
it was subsequently proven by robert w. spekkens and terry rudolph @xcite that the atvy protocol is even better than its inventors had thought : its bias is in fact exactly . in the same paper , spekkens and
rudolph gave an amazingly simple coin - flipping protocol that achieves the same bias as well as another one whose bias is merely . according to an earlier paper of theirs @xcite , that s the smallest bias possible for a coin - flipping protocol in which the quantum communication is limited to a single qubit .
meanwhile , andris ambainis @xcite and , independently , spekkens and rudolph @xcite discovered quantum coin - flipping protocols whose bias @xmath4 is even smaller , but they require the transmission of a qu__trit _ _ ( or one qubit and two qutrits in the case of spekkens and rudolph ) .
on the other hand , alexei kitaev @xcite proved that no quantum coin - flipping protocol can have a bias below . very recently , andr chailloux and iordanis kerenidis @xcite have announced a quantum coin - flipping protocol whose bias is arbitrarily close to kitaev s bound , but it requires an unlimited number of rounds of interaction as it this bound . despite the theoretical success of quantum coin - flipping protocols , compared to classical protocols , severe practical problems inherent to their implementation have been discussed by jonathan barrett and serge massar @xcite , who argued that quantum coin flipping is problematic in any realistic scenario in which noise and loss can occur in the processing ( preparation , transmission and measurement ) of quantum information .
for this reason , they proposed random bit-_string _ generation of _ single - shot _ coin flipping .
however , this is not interesting from a quantum cryptographic perspective the same goal can be achieved with purely classical means @xcite . in a subsequent paper ( nfpm ) written in collaboration with anh tuan nguyen , julien frison and kien phan huy
, massar has defined a _ figure of merit _ on which quantum coin - flipping protocols can outperform any possible classical protocol even in a realistic setting and they have verified their concept experimentally @xcite . even though their protocol is not broken in the presence of loss ,
however , alice can choose the outcome with near certainty in a realistic setting .
we claim that , in order to be of practical use , a protocol should be _ loss tolerant _ , which we define as being completely impervious to loss of quantum information . in this sense , the nfpm protocol is not loss tolerant because its bias increases asymptotically towards 0.5 as losses more and more severe , which is unavoidable in practice ( with current technology ) over increasing distance between alice and bob . in this paper
, we concentrate on this most likely source of imperfection in actual implementations , namely _
losses_. with the exception of the nfpm protocol mentioned above ( which is not loss tolerant ) , all previously proposed quantum coin - flipping protocols _
completely _ insecure even in the absence of noise as soon as the quantum channel between alice and bob is lossy .
we introduce the first loss - tolerant quantum coin - flipping protocol .
we prove that our protocol is _ fair _ in the sense that either alice or bob can enjoy a bias of exactly @xmath5 with an optimal cheating strategy , independently of the channel s transmission and other sources of losses , provided quantum information that is not lost is prepared , transmitted and measured faithfully .
after this introduction , the structure of the paper is as follows .
we begin in section [ bb84:prot ] with a review of the original 1984 quantum coin - flipping protocol of bennett and brassard @xcite and we explain why it is completely vulnerable to a so - called epr - attack . this is interesting not only for historical reasons , but also because our novel loss - tolerant protocol follows the same template .
section [ ambprot ] reviews perhaps the most famous of all quantum coin - flipping protocols , due to ambainis @xcite , whose theoretical bias is @xmath4 .
however , we demonstrate in section [ ambweak ] that the security of that protocol is completely compromised in the presence of arbitrarily small channel loss . moreover , we argue that this problem is to the protocol in the sense that it can not be with small corrections .
( the same would be true of the qutrit - based coin - flipping protocol due to spekkens and rudolph @xcite as well as of their optimal single - qubit protocol @xcite . )
this is due to the notion of _ conclusive measurements _ , which we review in section [ mmt ] .
we show in section [ ourprot ] how to combine the strengths of the original bb84 protocol with those of the atvy protocol ( which is not loss tolerant either in its published form ) to finally achieve loss tolerance in quantum coin flipping and we analyse the security of our protocol .
conclusions and open problems are presented in section [ concl ] .
we review the original bb84 quantum coin - flipping protocol as well as the way it can be broken @xcite . here
are the so - called bb84 states : @xmath6{\large x}}{| \psi_{0,1}\rangle}={| 1\rangle}\phantom{{| + \rangle}}\!\!\!\!\!\!\cr \end{array}\right\ } & & a=0\nonumber\\[1ex ] \left .
\begin{array}{l } { | \psi_{1,0}\rangle}={| + \rangle}\phantom{{| 0\rangle}}\!\!\!\!\!\!\cr \phantom{\framebox[0mm]{\large x}}{| \psi_{1,1}\rangle}={| -\rangle}\phantom{{| 0\rangle}}\!\!\!\!\!\!\cr \end{array}\right\ } & & a=1\;,\end{aligned}\ ] ] where .
we say of @xmath7 that @xmath8 is the _ basis _ and @xmath9 is the _
bit_. we define measurement bases @xmath10 for . in the full bb84 quantum coin - flipping protocol @xcite , alice would prepare and send bob a large number of qubits , all in the same randomly chosen basis @xmath8 . to emphasize the essential features of the protocol , however , we outline below a simplified version in which a single qubit is used . 1 . [ stepone ] alice prepares one of the four bb84 states @xmath7 with basis @xmath8 and bit @xmath9 chosen independently at random
; she transmits that qubit to bob .
[ stepthree ] bob chooses a random and measures the received qubit in basis @xmath11 ; let bob s result be @xmath12 .
[ steptwo ] bob sends a randomly chosen bit @xmath13 to alice .
[ stepfour ] alice reveals her original @xmath8 and @xmath9 to bob .
[ stepfive ] if and , bob aborts the protocol , _ calling alice a cheater _ ; if , bob has no way to verify alice s honesty . 6 .
[ stepsix ] if bob did not abort the protocol , the outcome of the coin flip is , where `` @xmath14 '' denotes the sum modulo @xmath15 ( also known as the `` exclusive or '' ) . in this protocol , bob can not cheat at all .
the only strategy for a cheating bob would be to make an educated guess on alice s choice of @xmath8 before deciding on the @xmath13 to send her at step [ steptwo ] , so as to bias the coin - flip outcome .
however , @xmath8 corresponds to alice s random choice of basis .
hence , the state @xmath16 received by bob at step [ stepone ] is either or .
it follows from the fact that that it is impossible for bob to guess the value of @xmath8 better than at random . on the other hand , it is obvious that alice can bias the protocol if she does not mind the risk of being called a cheater .
the simplest approach is to be honest in the first step . when she receives @xmath13 from bob , the probability is @xmath17 that she is happy with the outcome , in which case she proceeds honestly with the protocol .
on the other hand , if she is unhappy with , she can lie on @xmath8 at step [ stepfour ] and send a random @xmath9 . in that case ,
her probability of being caught is @xmath18 since bob chose with probability @xmath17 , in which case he obtained also with probability @xmath17 .
all counted , this allows her to a .
a slightly more interesting cheat is for her to send state for a random at step [ stepone ] and declare the @xmath8 that suits her wish ( with the appropriately chosen @xmath9 ) at step [ stepfour ] .
this her to enjoy a bias of , with a probability of being called a cheater .
it was to make the probability of undetected cheating exponentially small that the full bb84 protocol required the transmission and measurement of a large number of qubits @xcite .
a much more remarkable kind of cheating is possible for alice , as explained in the same paper that introduced the bb84 protocol itself @xcite , which allows her to break the protocol completely ( i.e. enjoy a ) with no fear of ever being caught .
let us say she wishes the outcome of the coin flip to be bit @xmath19 . instead of sending a legitimate bb84 state or any other pure state at step [ stepone ]
, alice sends half an epr pair to bob and keeps the other half for herself .
she waits until step [ steptwo ] , when she learns bob s choice of @xmath13 , to measure in basis the half she had kept ; let @xmath9 be her measurement outcome .
this tells her that bob has obtained ( or will obtain , if he has not yet measured ) @xmath20 in case he has measured ( or will measure ) in basis @xmath8 .
( alice does not care about the value of @xmath12 if bob chooses to measure in the other basis . ) hence , she can always obtain her desired outcome by sending those @xmath8 and @xmath12 to bob in step [ stepfour ] .
= + 1 as subsequently discovered independently by mayers @xcite and by lo and chau @xcite in the context of quantum bit commitment , this kind of cheating is _ always _ possible for alice in any quantum coin - flipping protocol that has the structure of the bb84 protocol , regardless of the actual set of quantum states , whenever the density matrices used to signal or at step [ stepone ] are identical ( ) .
this is due to the striking quantum process known as `` '' , discovered by erwin schrdinger @xcite as early as 1936 and better known as the hjw @xcite .
( see ref .
@xcite for an entertaining history of this . )
note that the remote steering attack works just as well if bob postpones his measurement until after alice reveals @xmath8 and @xmath9 , and _ almost _ just as well if @xmath21 and @xmath22 , although different , are exponentially indistinguishable @xcite .
this last remark caused the demise of the bit commitment scheme proposed in ref .
@xcite and we shall henceforth not differentiate between density matrices that are equal and those that are merely exponentially indistinguishable .
in order to escape the remote steering attack , aharonov , ta - shma , vazirani and yao @xcite introduced a coin - flipping protocol in which . to reduce even further alice s possible bias , they shuffled the order of steps [ stepthree ] , [ steptwo ] and [ stepfour ] so that bob delays his measurement of alice s supplied state until after she tells him what she claims to have sent . in this way , he can measure systematically in the basis rather than having to measure in a randomly chosen basis @xmath23 whose outcome had probability @xmath17 of being useless .
this allowed atvy to design a quantum coin - flipping protocol with bias , as subsequently proven by spekkens and rudolph @xcite . to achieve his smaller bias of @xmath4 , ambainis used the following states on qutrits ( rather than on qubits ) : @xmath24{\large x}}{| \phi_{0,1}\rangle}=\frac{1}{\sqrt{2}}{| 0\rangle}-\frac{1}{\sqrt{2}}{| 1\rangle}\cr \end{array}\right\ } & & a=0\nonumber\\[1ex ] \left .
\begin{array}{l } { | \phi_{1,0}\rangle}=\frac{1}{\sqrt{2}}{| 0\rangle}+\frac{1}{\sqrt{2}}{| 2\rangle}\cr \phantom{\framebox[0mm]{\large x}}{| \phi_{1,1}\rangle}=\frac{1}{\sqrt{2}}{| 0\rangle}-\frac{1}{\sqrt{2}}{| 2\rangle}\cr \end{array}\right\ } & & a=1\,.\end{aligned}\ ] ] again we say of @xmath25 that @xmath8 is the basis and @xmath9 is the bit . this time , we define measurement bases @xmath26 for .
here is ambainis protocol . 1 . [ astepone ] alice prepares one of the four ambainis states @xmath25 with basis @xmath8 and bit @xmath9 chosen independently at random ; she transmits that qutrit to bob , who stores it in his quantum memory .
[ asteptwo ] bob sends a randomly chosen bit @xmath13 to alice .
[ astepthree ] alice reveals her original @xmath8 and @xmath9 to bob .
[ astepfour ] bob takes alice s qutrit out of quantum memory and measures it in basis @xmath27 ; let bob s result be @xmath12 .
if , bob aborts the protocol , _ calling alice a cheater_. 6 .
[ astepsix ] if bob did not abort the protocol , the outcome of the coin flip is .
alice can not use remote steering to gain complete control over the outcome of the coin flip because the density matrices that could be by bob at step [ astepone ] , corresponding to her choice @xmath8 of basis , @xmath28 are distinct .
it is easy to see that alice can enjoy a if she sends state at step [ astepone ] and declares @xmath8 and @xmath9 appropriately at step [ astepthree ] .
the proof that this is alice s optimal cheating strategy is nontrivial but has been worked out in detail by ambainis @xcite . on the other hand , the fact that makes it possible for bob to cheat by measuring alice s qutrit before step [ asteptwo ] in order to learn information about @xmath8 and bias his choice of @xmath13 accordingly .
the most obvious strategy for bob is to measure alice s qutrit in the computational basis , which allows him to enjoy a as well .
the fact that this is bob s optimal cheating strategy follows directly from carl w. helstrom s optimal measurement theory @xcite , which we review in section [ mmt ] . for completeness
, we mention that the independently - discovered quantum coin - flipping protocol of spekkens and rudolph @xcite requires alice to choose a random bit @xmath8 , prepare a one - qubit - and - two - qutrit entangled state @xmath29 and send bob one of the two qutrits in step [ astepone ] .
step [ asteptwo ] is the same as in ambainis protocol . in step [ astepthree ] , alice sends the qubit and the other qutrit to bob , which allows him to verify her honesty in step [ astepfour ] by performing a povm on the combined alice - provided state whose outcome is either @xmath30 , @xmath31 or `` alice cheated '' .
again , the result of the coin flip is provided bob did not catch alice cheating . the essential ( but not sufficient ) reason why this protocol gives the same bias as ambainis
is that the density matrix @xmath16 received by bob at step [ astepone ] is exactly the same in both protocols , as given by eq .
( [ rhoamb ] ) .
even though ambainis analysis of his protocol is mathematically impeccable , there is a practical problem that can not be neglected if one is ever to implement such protocols in real life : there will be unavoidable losses in the quantum channel between alice and bob .
this is true in particular if photons are used to carry quantum information and if the quantum channel is an optical .
further losses are to be expected in bob s quantum memory and detection apparatus .
the situation is even worse for the protocol of spekkens and rudolph because it requires both alice and bob to have quantum memory since they must store one qutrit each between steps [ astepone ] and [ astepthree ] ( there is no real need for alice to store also the qubit since it contains only classical information )
. please remember that one of the main appeals of quantum cryptography , from its very beginnings @xcite , has been to offer protocols _ that can be implemented with current technology _ , yet remain secure against any potentially future attack so long as quantum mechanics is not violated .
it follows that there is a possibility , when bob tries to measure alice s qutrit at step [ astepfour ] of ambainis protocol ( or the one - qubit - two - qutrit system in the protocol of spekkens and rudolph ) , that he does not register anything _ even though both alice and bob have been entirely honest_. how should bob react in this case ? he could hardly call alice a cheater if the most likely cause for loss is in his own detectors ! there seem to be only two reasonable responses from bob , as pointed out already by barrett and massar @xcite concerning a quantum coin - flipping protocol inspired by the quantum gambling protocol of won young hwang , doyeol ahn and sung woo hwang @xcite : bob can ( 1 ) accept alice s declared @xmath8 and @xmath9 on faith or ( 2 ) request alice to restart the coin - flipping protocol from scratch . but both these `` solutions '' are unacceptable . if alice knows that bob will believe her on faith in case he gets no detection signal , she can totally bias the coin flip with the most maddeningly simple cheating strategy : she does _ nothing _ at all during step [ astepone ] and lets bob `` store '' the empty signal .
after having received bob s choice of @xmath13 , she is then free to `` reveal '' whichever @xmath8 would produce her desired outcome . since bob s measurement of the empty signal will yield nothing at step [ astepfour ]
, he has no other choice but to believe alice on faith . on the other hand ,
if alice and bob have agreed to restart the coin - flipping protocol in case bob does not detect anything at step [ astepfour ] , it is bob who can totally bias the coin flip without any need to manipulate quantum states . whenever he receives alice s qutrit , he does nothing at all with it and sends his random choice of @xmath13 . if alice reveals @xmath8 at step [ astepthree ] so that he is happy with outcome , with probability @xmath17 , bob pretends to measure alice s long - lost qutrit and claims to be satisfied with her honesty . but
if he is not happy with the outcome , bob simply tells alice that he has not registered anything and requests a new instance of the protocol .
this continues until the outcome is to bob s liking .
this cheating strategy can not be detected by alice whenever the expected probability @xmath32 of registering an outcome when both alice and bob are honest is at most @xmath17 , provided bob asks to restart the protocol with probability even when progressing to step [ astepsix ] would have produced his desired outcome .
unless bob has the technological ability to make sure he received a quantum state from alice at step [ astepone ] without disturbing it , and because alice will not accept to restart the protocol after she has revealed @xmath8 and @xmath9 , he has only one line of defence against her `` send nothing '' cheating strategy : he must measure alice s signal immediately upon reception and be allowed to ask her to restart the protocol from scratch ( with an independent random choice of state ) until he actually registers a measurement outcome .
this means that we must revert to the original bb84 template in which bob measures before alice reveals @xmath8 and @xmath9 . as explained at the beginning of section [ ambprot ] , this will make it easier for a cheating alice to escape detection since bob s measurement can no longer depend on her claimed state .
however , provided bob chooses his measurement basis at random , he will carry out with probability @xmath17 the same measurement he would have performed in ambainis original protocol . as we prove in subsection [ bobhonest ] , it follows that bob s probability of catching alice s eventual cheating is reduced , but not by more than a factor of 2 , compared to the original protocol .
hence , alice s bias is at most @xmath33 rather than @xmath4 .
the important point is that this bias remains below @xmath3 .
nevertheless , this modification in ambainis protocol , which is made necessary by practical considerations ( with current technology ) , reopens the door for bob to completely break the revised protocol ! when alice tells him she transmitted a qutrit , bob measures it immediately in the computational basis .
if he obtains @xmath34 or @xmath35 , bob _ knows _ alice s choice of @xmath8 .
this allows him to choose @xmath13 so that suits his desired outcome . on the other hand
, if bob either obtains @xmath36 or if he does not register anything , then he tells alice that the transmission has been unsuccessful and he qutrit from her . in effect
, the protocol will proceed to the step in which bob sends his choice of @xmath13 to alice _ only _ when he already knows alice s choice of @xmath8 .
this means that bob enjoys a bias of @xmath3 and the protocol is completely broken .
to camouflage his ( in case alice might wonder why the measurement does not succeed more often ) , bob can pretend at the outset that his detectors are half as efficient as they really are .
alternatively , he could surreptitiously replace the quantum channel that links him to alice with a sufficiently better one .
this fatal flaw in any practical implementation of ambainis protocol comes from one simple consideration . even though the mixed states @xmath21 and @xmath22 ( see eq .
[ rhoamb ] ) used at step [ astepone ] by alice to partially commit to either or , respectively , are non - orthogonal , hence they can not be distinguished with certainty by bob all the time , _ they can be distinguished conclusively with positive probability_. below the notion of conclusive measurements , we introduce our new protocol in section [ ourprot ] and prove that its bias is exactly @xmath5 even when arbitrarily severe losses are taken into account .
consider two non - orthogonal density matrices @xmath21 and @xmath22 ( they could be pure states ) .
there are several figures of merit in measurements that attempt to distinguish them @xcite .
helstrom has studied the optimal measurement to output a guess that minimizes the error probability @xcite .
assuming both @xmath21 and @xmath22 were equally likely _ a priori _ , helstrom s measurement outputs the correct guess with probability @xmath37 where @xmath38 is the _ trace distance _ between @xmath21 and @xmath22 , `` @xmath39 '' denotes the trace and . in particular , if @xmath40 is a diagonal real matrix , then . when the spans of @xmath21 and @xmath22 are distinct , there another type of measurement , known as _ conclusive _ measurement or
_ unambiguous state discrimination _ ( usd ) @xcite .
these measurements have three possible outcomes , `` @xmath2 '' , `` @xmath41 '' and `` ? '' , the latter of which being called the _
inconclusive outcome_. whenever outcome is obtained , it is guaranteed that the measured state was indeed @xmath16 ( assuming of course that is was either @xmath21 or @xmath22 and that there were no experimental errors ) . furthermore , the probability of a _ conclusive _ outcome ( not `` ? '' ) must be strictly positive for either input state
. a well - known example of conclusive measurement can distinguish between @xmath36 and @xmath42 with conclusive outcome probability .
more to the point of our paper , the obvious measurement in computational basis distinguishes between ambainis @xmath21 and @xmath22 ( eq .
[ rhoamb ] ) with conclusive probability @xmath17 . in general ,
any coin - flipping protocol that follows the template of bb84 is vulnerable to the attack described in section [ ambweak ] when a conclusive measurement exists between the corresponding @xmath21 and @xmath22 .
it is tempting to think that this line of attack will not apply when conclusive measurements do not .
however , this is not necessarily the case .
erika , stephen m. barnett , chefles , sarah croke , claire r. gilson and john jeffers have studied `` maximum confidence quantum measurements '' ( mcqm ) , which somehow interpolate between helstrom and conclusive measurements @xcite . like conclusive measurements ,
mcqms have some probability of yielding the `` ? '' inconclusive outcome .
however , when the outcome is either `` @xmath2 '' or `` @xmath41 '' , it is correct with probability at least .
helstrom s measurement maximizes @xmath43 subject to whereas conclusive measurements ( when they ) minimize @xmath32 subject to .
there can be a trade - off between those two probabilities in the sense that it is sometimes possible to achieve an increase in @xmath43 ( compared to the accuracy of helstrom s measurement ) by tolerating a larger @xmath32 .
the relevance of mcqms is that bob could use them to increase his coin - flip bias at the cost of increasing his probability of asking alice to restart the protocol ( pretending he has not registered an outcome ) .
if mcqms exist for @xmath43 arbitrarily close to @xmath41 , bob can come as near as he wants to choosing the coin - flip outcome , provided alice has enough patience ( navety ? ) to accept restarting the protocol indefinitely until bob tells her that he is ready to continue . to demonstrate that this is a legitimate worry ,
consider an ( admittedly contrived ) quantum coin - flipping protocol based on ambainis states ( see the beginning of section [ ambprot ] ) , except that we add and . in step [ astepone ] ,
alice chooses basis at random with uniform probability , but is chosen so that whereas . the density matrix received by bob would be either @xmath44 these two mixed states can not be distinguished conclusively .
nevertheless , a measurement in the computational basis yields either @xmath36 , which is interpreted as the inconclusive outcome `` ? '' , or it yields either @xmath34 or @xmath35 , which are interpreted as either @xmath45 or @xmath46 , respectively .
this mcqc is inconclusive with probability .
when it is not inconclusive , however , the verdict is correct with probability .
this is much better than helstrom s measurement , which would give an answer but be correct only with probability @xmath47 since .
therefore , a quantum coin - flipping protocol that uses these states would allow bob to enjoy a maximal theoretical bias of @xmath48 if we did not take losses into account .
however , bob s bias becomes larger than @xmath49 if alice agrees to restart the protocol whenever he claims to have not registered an outcome , which he would do whenever his measurement outcome is the inconclusive @xmath36 .
now , we introduce our novel quantum coin - flipping protocol and we prove that its bias is exactly , regardless of the extent of losses that are unavoidable with current technology . for this , we use the states introduced by atvy in their protocol @xcite but revert to the original bb84 template @xcite . consider the states @xmath50{\large x}}{| \varphi_{1,0}\rangle}=\alpha { | 0\rangle}-\beta { | 1\rangle}\cr \end{array}\right\ } & & x=0\nonumber\\[1ex ] \left . \begin{array}{l } { | \varphi_{0,1}\rangle}=\beta { | 0\rangle}-\alpha { | 1\rangle}\cr \phantom{\framebox[0mm]{\large x}}{| \varphi_{1,1}\rangle}=\beta { | 0\rangle}+\alpha { | 1\rangle}\cr \end{array}\right\ } & & x=1\end{aligned}\ ] ] where @xmath51 and @xmath52 are real numbers such that and .
see fig .
[ fig1 ] . as
always , we say of @xmath53 that @xmath8 is the basis and @xmath9 is the bit .
we define measurement bases _ mutatis mutandis _ as we had done in eq .
( [ eq : basis ] ) for the bb84 states : @xmath54 for .
we shall soon see why we have regrouped the states according to the value of @xmath9 rather than that of @xmath8 as we had done previously .
here is our loss - tolerant quantum coin - flipping protocol . 1 . [ nstepone ] alice prepares one of the four states @xmath53 with basis @xmath8 and bit @xmath9 chosen independently at random ; she transmits that qubit to bob .
[ nsteptwo ] bob chooses a random and measures the received qubit in basis @xmath55 . if his apparatus does not register an outcome , he requests alice to start over at step [ nstepone ] ; otherwise , let the measurement result be @xmath12 .
[ nstepthree ] bob sends a randomly chosen bit @xmath13 to alice .
[ nstepfour ] alice reveals her original @xmath8 and @xmath9 to bob .
[ nstepfive ] if and , bob aborts the protocol , _ calling alice a cheater _ ; if , bob has no way to verify alice s honesty . 6 .
[ nstepsix ] if bob did not abort the protocol , the outcome of the coin flip is .
there are three differences , compared to the original bb84 protocol described in section [ bb84:prot ] : ( 1 ) in addition to having been globally rotated to facilitate subsequent analysis , the states used are more general in the sense that bases @xmath56 and @xmath57 need not be mutually unbiased ; ( 2 ) step [ nsteptwo ] allows bob to ask alice to restart the protocol in case his measurement apparatus fails to register an outcome ; and ( 3 ) the final coin - flip result is rather than .
the first modification will allow us to fine - tune the protocol in subsection [ fair ] .
the second modification makes it possible for the protocol to be loss tolerant as we have explained in section [ ambweak ] .
the effect of the third modification , which corresponds to the main original contribution of the atvy protocol , is that two _ distinct _ density matrices @xmath58 and @xmath59 ( see below ) are now used by alice to partially commit to either or by the transmission of @xmath53 at step [ nstepone ] with a randomly @xmath8 .
= + 1 @xmath60 in sharp contrast , the intuition behind the original bb84 protocol was for alice to `` commit '' to either or with a randomly chosen @xmath9 , but that was doomed by schrdinger s remote steering process because the corresponding density matrices were equal .
furthermore , we shall see in subsection [ alicehonest ] that , contrary to the mixed states @xmath21 and @xmath22 used in ambainis protocol ( see eq .
[ rhoamb ] ) , this time @xmath58 and @xmath59 can not be distinguished conclusively , nor even by a maximum confidence quantum measurement better than helstrom s measurement , which is the key to loss tolerance .
although we have presented our protocol as a modification of the original bb84 protocol , it is useful for analysis purposes to contrast it also with the atvy protocol .
the key difference between the atvy protocol and ours is that they chose to introduce a new template ( used in most subsequent protocols such as ambainis ) in which bob stores the quantum state sent by alice at step [ nstepone ] in order to delay measurement until alice has given him its classical description .
the intention was to make it harder for alice to cheat because bob would know to measure the state in basis @xmath61 rather than to choose a random basis @xmath55 .
unfortunately , as we have explained , it was this feature that made their protocol incapable of tolerating channel loss unless bob has the technological ability to detect if a signal has been received by alice without otherwise disturbing its state .
= + 1 next , we prove that the maximum biases that alice and bob can enjoy are @xmath62 respectively , and we give explicit cheating strategies to achieve those .
we conclude that the choice of @xmath51 and @xmath52 that makes those two biases equal corresponds to a fair protocol whose bias is 0.4 .
it would be relatively easy to determine alice s optimal cheating strategy were she restricted to sending some pure state to bob in the first step of the protocol .
however , we have learned from the demise of the original bb84 protocol @xcite that it might be to her dishonest advantage to prepare an entangled state , send one qubit from it to bob at step [ nstepone ] , and wait until bob s announcement of @xmath13 at step [ nstepthree ] to measure in the most informative way what she had kept .
this could increase her chances of deciding on her best choice of @xmath8 and @xmath9 to `` reveal '' at step [ nstepfour ] in order to maximize her probability of successfully biasing the coin flip .
it is significantly more complicated to take all possible such strategies into account .
fortunately , most of the work has already been done by spekkens and rudolph in their thorough analysis of the atvy protocol @xcite .
let us begin our analysis by briefly pretending that bob does not measure the state alice has sent him after she reveals her choice of basis and bit , and that he measures it in that basis . as we have already pointed out , this becomes the atvy protocol .
it follows directly from the analysis of spekkens and rudolph [ choosing what they call @xmath63 in their eq .
( 20 ) so that and ] that any cheating strategy alice may deploy gives her bias @xmath64 now , to analyse our protocol , we must take into the fact that we require bob to measure alice s state _ before _
she tells him what she claims to have sent .
( this is how we make our protocol loss tolerant . )
this difference makes it easier for alice to cheat because bob s measurement can not be chosen to maximize his probability of discriminating between her claimed state and any other state that she might have sent .
however , as we show below , this modification does not reduce bob s probability of catching alice cheating by more than a factor of @xmath15 .
recall that bob is required by step [ nstepthree ] of the protocol to send a _
random _ choice of @xmath13 , which must be uncorrelated with his former choice @xmath23 of measurement and its outcome @xmath12 .
the crucial observation is that the randomness of @xmath13 deprives alice from any information she might otherwise have obtained from bob concerning @xmath23 and @xmath12 .
( the importance of this observation is illustrated in section [ cunning ] . )
it follows that her choice of which @xmath8 and @xmath9 to declare at step [ nstepfour ] can not depend on what has happened at bob s after she transmitted her quantum state at step [ nstepone ] . in particular , with probability @xmath17 since @xmath23 is chosen at random by an honest bob , in which case he has chosen by chance at step [ nsteptwo ] precisely the measurement he would have performed in the atvy protocol , had he been allowed to wait until alice s announcement of @xmath8 and @xmath9 before deciding on his measurement .
the above implies that any cheating strategy that alice might deploy against our protocol translates into an _
identical _ cheating strategy against the atvy protocol , possibly with a different success probability , since it makes no measurable difference to alice whether bob measures before ( as in our protocol ) or after ( as in the atvy protocol ) she has to declare @xmath8 and @xmath9 .
now , consider an arbitrary cheating strategy on the part of alice and let @xmath65 ( resp .
@xmath66 ) be the bias that she enjoys with this strategy against our protocol ( resp .
the atvy protocol ) .
consider an arbitrary run of our protocol , when alice uses this cheating strategy .
with probability @xmath17 , independently from anything else , bob randomly chooses the same measurement he would have performed in the atvy protocol , in which case alice succeeds with probability . in the other @xmath17 of the cases
, bob can not verify the state alice claims to have sent and the probability that alice succeeds is at most @xmath41 .
all counted , the success probability of alice in our protocol is @xmath67 it follows that @xmath68 where the last inequality follows from eq .
( [ sralice ] ) .
we now proceed to show that this bound can be saturated . surprisingly
( when we recall the demise of the bb84 quantum coin - flipping protocol ) , the optimal cheating strategy for alice does _ not _ require her to prepare an entangled state from which she would send one qubit at step [ nstepone ] .
instead , it suffices for her to send either or to bob .
let us say she sent @xmath42 ( the other case is similar ) and bit @xmath13 from bob at step [ nstepthree ] .
if her desired outcome is @xmath19 , she sets and claims at step [ nstepfour ] to have sent @xmath69 at step [ nstepone ] .
with probability @xmath17 , bob had already chosen , in which case he can not catch alice cheating . with complementary probability @xmath17 , he had chosen , in which case alice escapes detection with probability @xmath70 ( the last equality is because ) . putting it all together , alice obtains her desired outcome with probability @xmath71 and her bias is therefore @xmath72 which saturates the bound given in eq .
( [ alicebias ] ) . in the analysis of bob s bias in quantum coin - flipping protocols ,
it is usual to apply eqs .
( [ eq : helstrom ] ) and ( [ eq : tracedist ] ) to compute the trace distance @xmath73 in to determine that the optimal helstrom measurement that bob could perform to best guess alice s committed bit ( here , @xmath9 ) gives him the correct answer with probability @xmath74 . from this
, we would `` normally '' conclude that bob s maximal bias is . as we have seen , however , this approach is not appropriate in our context because it does not take into the eventual possibility for bob to increase his bias by making conclusive or maximum confidence quantum measurements on the state sent by alice at step [ nstepone ] , so that he could ask her to restart the protocol whenever he is not sufficiently satisfied with the probability that his guess of @xmath9 be correct .
fortunately , the analysis of bob s optimal cheating strategy is straightforward in this case . for either value of that alice
may have chosen in step [ nstepone ] , eq .
( [ rhonous ] ) shows the mixed state @xmath75 that bob would receive from her .
the key observation is that , algebraically speaking , and .
it follows that , from a cheating bob s perspective , whose only concern is in guessing @xmath9 as best as possible ( including the possibility of asking alice to restart the protocol if he is not satisfied with his own confidence in his guess ) , this is _ strictly identical _ to an alternative scenario in which alice would have sent @xmath36 with probability @xmath74 or @xmath34 with probability @xmath76 in case , and vice versa in case . in other words , this is exactly as if alice had communicated her choice of @xmath9 to bob as the _ purely classical _ signal @xmath77 through a binary symmetric channel with error probability @xmath76 .
seen this way , there is _ nothing quantum _ about the situation .
hence , a complete measurement in the computational basis of the state received by alice provides bob with an optimal cheating strategy since it yields _ all _ the information classically available in the `` quantum '' signal !
not surprisingly , this is indeed precisely helstrom s measurement . in particular
, the outcome of this measurement does not give bob any indication on whether or not it would be to his advantage to ask alice to restart the protocol .
( note that it is not simply because alice s signal can be thought of as classical that there can not be a conclusive or a maximum confidence measurement ; in particular , _ erasure channels _ provide the classical equivalent to quantum conclusive measurements but binary symmetric channels do nt . ) = + 1 to summarize , bob s optimal cheating strategy is to measure alice s qubit in the computational basis in to learn the value of @xmath9 with error probability @xmath76 .
this him to choose @xmath13 appropriately and obtain his outcome with complementary probability @xmath74 , thus enjoying his optimal bias . in this subsection
, we find the value of @xmath51 so that the bias either player can achieve by cheating is the same .
we simply need to fulfil the condition @xmath78 which for our protocol amounts to @xmath79 subject to . solving this system yields @xmath80 as illustrated in fig .
these values correspond to @xmath81 which defines a fair loss - tolerant quantum coin - flipping protocol whose bias is @xmath5 . in other words ,
either player can obtain a desired outcome with probability @xmath82 by using an optimal cheating strategy , provided of course the other player is honest .
but what if they both try to cheat ?
an interesting phenomenon is illustrated if we take the unusual step of considering the case when both alice and bob cheat .
the most example of double - cheating occurs when alice is convinced that bob is so greedy that he will try his best to control the coin flip . in this case
, alice can send an honest quantum state in step [ nstepone ] . to maximize his chances of guessing alice s choice of @xmath9 , bob measures the state in the computational basis , as we have seen .
this allows him to send alice a value of @xmath13 that produces his desired outcome with probability @xmath82 ( assuming they use the fair version of the protocol ) .
but at the same time , he has lost his ability to verify the honesty of alice since he did not measure her qubit in a legitimate basis . suspecting this , alice is free to claim whatever suits her best at step [ nstepfour ] , probably lying about @xmath9 , and bob
can not look her in the face and call her a cheater ! for an amusing variation on this theme , consider what happens if both parties attempt to perform their optimal cheating strategies simultaneously .
in this case , alice sends either @xmath42 or @xmath83 , which bob measures in basis .
when bob attempts influencing the coin flip by choosing @xmath84 as a function of his measurement outcome , little does he suspect that his choice is in fact totally random : ironically , bob follows the honest protocol at step [ nstepthree ] !
= -1 for a more subtle example , consider a scenario according to which bob is alice s young son .
they agree to flip a quantum coin to determine who will decide on the film to be seen tonight : alice will have the choice if the outcome is @xmath41 and bob if it is @xmath2 .
alice knows that her little rascal will do his best to cheat and win , but also that he will follow step [ nsteptwo ] properly ( rather than measuring in the computational basis ) because he will not want to relinquish his possibility to catch his mother cheating at verification step [ nstepfive]or so he thinks .
hence , bob s cheating will consist in sending at step [ nstepthree ] instead of choosing @xmath13 at random .
after a calculation , we find that this gives him probability @xmath85 that , thus producing his winning outcome .
unbeknownst to bob , however , his mother wants him to win , but she does not wish him to know this for fear of undermining her authority ! assuming she knows her son as well as she thinks , she proceeds as follows . at step [ nstepone ]
, she honestly sends some random state @xmath53 .
when she receives bob s choice of @xmath13 at step [ nstepthree ] , there are two possibilities . if , she can relax and continue honestly since bob wins , according to her wish . on the other hand ,
if , then alice can deduce that bob used the wrong basis in his measurement : .
this allows her to `` reveal '' @xmath8 together with any @xmath86 of her choice at step [ nstepfour ] since she knows ( or suspects ) that bob will be unable to call her a cheater .
according to our story , she chooses to make bob happy .
admittedly , the above tale is unlikely at best .
nevertheless , it serves the purpose of demonstrating the importance in our proof of security against alice ( subsection [ bobhonest ] ) that bob s bit @xmath13 be chosen randomly despite the fact that he has significant information about alice s @xmath9 after an honest measurement at step [ nsteptwo ] . indeed , our proof relied in a crucial way on the fact that alice would have no information on the measurement basis used by bob .
our tale shows how wrong and damaging it would have been had the protocol asked an _
honest _ bob to send a @xmath13 correlated to his measurement outcome .
we have seen that bob could exploit the loss of quantum information to cheat in previous protocols , whereas our protocol is loss tolerant .
nevertheless , our protocol could become susceptible to loss if it were implemented with insufficient care . a _ side channel _
is any source of information that bob could exploit about the quantum state sent by alice above and beyond its theoretical definition as a pure qubit .
for example , the protocol would be obviously insecure if the apparatus used by alice to generate her quantum states produced each of the four legitimate states as photons of significantly different wavelengths or spatial position .
these issues have been extensively in the context of quantum key distribution .
see ref .
@xcite for a compelling example of successful hacking of a commercially available apparatus .
= -1 an interesting example of side channel would in a careless implementation of our quantum coin - flipping protocol if alice used an attenuated laser pulse to generate her states , as is done in most current implementations of quantum key distribution .
the problem stems from the fact that one can distinguish conclusively between @xmath58 and @xmath59 ( eq . [ rhonous ] ) when a pulse consists of two ( or more ) identical atvy states .
for this , it suffices for bob to measure one photon in basis @xmath56 and the other in basis @xmath57 ( eq . [ double - prime ] ) .
if the two measurements produce the same outcome , this is necessarily the correct bit @xmath9 encoded by alice since one of the measurement has been performed in the correct basis .
this occurs with probability . with the fair states ,
bob s probability of conclusive outcome is therefore a substantial @xmath87 each time a pulse contains two photons , which means that this _ implementation _ of the protocol is completely insecure because bob can another state from alice until this event . to make the case of attenuated laser pulses even worse , a perfectly natural measurement apparatus that an bob could use in the implementation of our protocol @xcite would produce such conclusive outcome with half the probability derived above , namely @xmath88 of the double - photon pulses .
it follows that bob , who started the protocol with pure intentions , might find it difficult to resist sliding to the dark side when his ( ) measurement apparatus reveals with certainty the value of @xmath9 chosen by alice . a practical solution to this problem is for alice to use a source of entangled photons , as discussed in ref .
quantum coin flipping is a cryptographic primitive that has been studied extensively .
several approaches have been considered since the very beginnings of quantum cryptography . however , previous protocols are either totally insecure @xcite , or they are highly sensitive to ( if not completely broken by ) technologically unavoidably losses of quantum information on the channel between the players or in their storage and detection apparatus @xcite . in this paper
, we introduced the first loss - tolerant quantum coin - flipping protocol , which means that it is completely impervious to such losses .
we proved that our protocol is _ fair _ in the sense that both alice and bob have an optimal cheating strategy of producing their desired outcome with @xmath82 probability of success ( assuming the other player is honest ) and we provided those strategies explicitly . even though our protocol can tolerate arbitrary loss of quantum information
, it would fail in case of _ noise _ it would be impossible for bob to know , in case of a mismatch between his measurement outcome and alice s claimed state , if that is due to a genuine error ( on his part or alice s ) or to alice s dishonesty . that this may be unavoidable ;
indeed and massar have argued that single - shot quantum coin - flipping protocols are problematic when both noise and loss can @xcite .
if secure single - shot quantum coin - flipping protocols are indeed impossible in the presence of noise , is there something useful that quantum coin flipping can do above and beyond anything classically possible ?
we already know that random bit - string generation @xcite is not the answer given that it can be done classically @xcite .
we are currently investigating this issue @xcite along a different line from that proposed in ref .
@xcite .
another open question concerns the optimality of our protocol .
could there be a loss - tolerant quantum coin - flipping protocol whose bias is smaller than @xmath5 ?
alternatively , can we prove that @xmath5 is the smallest bias possible among all loss - tolerant quantum coin - flipping protocols ?
finally , we mention that this paper was entirely concerned with the task known as _ strong _ coin flipping
. there exists a similar task , _ weak _ coin flipping , in which only one outcome is favourable for alice , the other outcome being the only one favourable for bob
. protocols with arbitrarily small bias are known to exist for weak coin flipping @xcite , in stark contrast with kitaev s lower bound for the strong case @xcite .
can _ loss - tolerant _ quantum weak coin - flipping protocols exist with arbitrarily small bias ?
this question will be addressed in a subsequent paper .
we have recently implemented our own loss - tolerant protocol and we have successfully tested it against alice s and bob s optimal cheating strategies .
we report on those experiments elsewhere @xcite .
this was in a sense going full circle because it was our wish to implement ambainis quantum coin - flipping protocol that lead to the realization that we could never succeed and thus to the development of our new protocol .
the authors are grateful to hassne bada for pointing out to us the fact that spekkens and rudolph had worked out a tight analysis of the atvy protocol ( which was lacking in the atvy paper ) and to serge massar , carlos mochon and robert spekkens for discussing their papers very patiently with us .
somshubhro bandyopadhyay , sbastien gambs and tal mor have helped the authors concerning the issue of maximum confidence quantum measurements .
is supported in part by canada s natural sciences and engineering research council ( nserc ) , the canada research chair program , the canadian institute for advanced research ( cifar ) , quantum__works _ _ and the institut transdisciplinaire dinformatique quantique ( intriq ) .
is supported in part by the fonds qubcois de la recherche sur la nature et les technologies ( fqrnt ) , the canadian institute for photonics innovations ( cipi ) and an nserc canada . n.g .
is supported in part by the centre doptique , photonique et lasers ( copl ) , quantum__works _ _ , nserc , cipi and intriq .
bennett and g. brassard , `` quantum cryptography : public key distribution and coin tossing '' , _ proceedings of international conference on computers , systems and signal processing _ ,
bangalore , india , pp . 175179 , 1984 .
g. brassard , c. crpeau , r. jozsa and d. langlois , `` a quantum bit commitment scheme provably unbreakable by both parties '' , _ proceedings of 34@xmath90 annual ieee symposium on the foundations of computer science _ , palo alto , ca , pp . 362371 , 1993 .
h. buhrman , m. christandl , m. kouck , z. lotker , b. patt - shamir and n.k .
vereshchagin , `` high entropy random selection protocols '' , _ proceedings of 11th international workshop on randomization and computation ( random 2007 ) _ , princeton , nj , usa , pp . 366379 , 2007 .
y. zhao , c .- h.f .
fung , b. qi , c. chen and h .- k .
lo , `` quantum hacking : experimental demonstration of time - shift attack against practical quantum key distribution systems '' , _ physical review a _ * 78*:042333 , 2008 . | coin flipping is a cryptographic primitive in which two spatially separated players , who in principle do not trust each other , wish to establish a common random bit .
if we limit ourselves to classical communication , this task requires either assumptions on the computational power of the players or it requires them to send messages to each other with sufficient simultaneity to force their complete independence . without such assumptions , all classical protocols are so that one dishonest player has complete control over the outcome .
if we use quantum communication , on the other hand , protocols have been introduced that limit the maximal bias that dishonest players can produce .
however , those protocols would be very difficult to implement in practice because they are susceptible to realistic losses on the quantum channel between the players or in their quantum memory and measurement apparatus . in this paper
, we introduce a novel quantum protocol and we prove that it is completely impervious to loss .
the protocol is _ fair _ in the sense that either player has the same probability of success in cheating attempts at biasing the outcome of the coin flip
. we also give explicit and optimal cheating strategies for both players . |
this paper is the third of a series presenting results from studies of the qsos discovered in the apm survey for z 4 quasars .
a study of the evolution of lyman limit absorption systems over the redshift range 0.04 @xmath28 z @xmath28 4.7 was presented in storrie - lombardi ( 1994 ) [ paper i ] . the intermediate resolution ( 5 )
qso spectra and the survey for high redshift damped absorbers are presented in storrie - lombardi ( 1996 ) [ paper ii ] .
the evolution of the cosmological mass density of neutral gas at high redshift and the implications for galaxy formation theories are discussed in storrie - lombardi , mcmahon & irwin ( 1996 ) [ paper iv ] . in separate papers we will describe the intrinsic properties of the qsos and studies of the forest clouds at high redshift . a high resolution study of the forest region in a redshift @xmath29 qso has been completed by williger ( 1994 )
. how and when galaxies formed are questions at the forefront of work in observational cosmology .
absorption systems detected in quasar spectra provide the means to study galaxy formation and evolution up to redshifts of approximately five , back to when the universe was less than 10 percent of its present age .
surveys for absorption features have several advantages over trying to directly detect galaxies at high redshift .
much shorter exposure times are required because the qsos are relatively bright ( r@xmath30 18 - 19.5 ) and the large equivalent width systems are easily detected in the spectra .
this provides good absorption candidates to follow up with higher resolution spectra .
the redshift and column density can be accurately determined from the wavelength of the absorption system and the line profile .
this is far easier and more reliable than trying to directly get a spectrum of a very faint high redshift galaxy .
while the baryonic content of spiral galaxies that are observed in the present epoch is concentrated in stars , in the past this must have been in the form of gas .
the principal gaseous component in spirals is neutral hydrogen which has led to surveys for absorbers detected by the damped lines they produce ( wolfe 1986 , hereafter wtsc ; lanzetta 1991 , hereafter lwtlmh ; lanzetta , wolfe & turnshek 1995 , hereafter lwt ; wolfe 1995 ; paper ii ) .
though damped systems are observationally very rare objects with @xmath3140 confirmed examples known , the mass per unit comoving volume they contain is roughly comparable to the mass density of baryonic matter in present - day spirals , ie . a major constituent of the universe ( wolfe 1987 , lwt ) .
their metal abundances are much lower than galactic values ( pettini , boksenberg & hunstead 1990 ; rauch 1990 ; pettini 1994 ) and they are characterised by low molecular content and low , but not negligible , dust content ( fall , pei & mcmahon 1989 ; pei , fall & bechtold 1991 ; pettini 1994 ) , features consistent with an early phase of galactic evolution
. they may be the progenitors of spiral galaxies like our own and are clearly important for the study of the formation and evolution of galaxies .
they have been detected across a very large redshift range z@xmath30[0.5,4.5 ] providing the means to pinpoint the epoch of formation of disk galaxies and study their evolution .
eleven candidate damped absorption systems out of 32 measured features were identified in 27 spectra of the mainly non - bal quasars from the apm z 4 survey ( paper ii ) .
the eleven candidates cover the redshift range @xmath0 ( 8 with @xmath1 ) and have estimated column densities @xmath2 10@xmath3 atoms @xmath4 . in this paper the qso br1144@xmath60723 with a candidate absorber at [email protected] is removed from further consideration in the sample .
it has been observed with the anglo - australian telescope at high resolution and the damped candidate has been found to be all absorption at [email protected] ( r. hunstead , private communication ) .
high resolution echelle spectra ( 0.8 fwhm ) were obtained by s. dodorico as part of the eso key programme studying high redshift quasars for four of the qsos in the apm sample ( bri0952@xmath60115 , br1033@xmath60327 , bri1108@xmath60747 , br1202@xmath60725 ) .
the signal - to - noise ratio in bri0952@xmath60115 was very poor but the other spectra have been used to confirm two features as damped with another falling just below the log @xmath2 20.3 threshold .
we have discovered the highest redshift damped absorber known at [email protected] in qso br1202@xmath60725 .
the confirmation of the absorption systems is discussed in section 2 .
these data have been combined with data from previous surveys ( wtsc , lwtlmh , and lwt ) and the results for the lyman limit systems obtained in paper i to study the column density distribution for @xmath12@xmath7 17.2 and redshift evolution of these systems for 0.008 @xmath8 z @xmath8 4.7 .
numerous authors have studied the distribution of column densities , @xmath32 , for absorption lines .
the first determination was by carswell ( 1984 ) for lines with @xmath33 @xmath34 atoms @xmath4 .
they found @xmath35 ( @xmath36 ) .
damped absorption ( dla ) systems comprise the high column density tail of neutral hydrogen absorbers with column densities of @xmath37 atoms @xmath4 .
they dominate the baryonic mass contributed by .
when damped systems are included in the column density distribution function for a single power law fit the exponent is @xmath38 = 1.41.7 ( tytler 1987 ; petitjean 1993 and references therein ) . assuming the baryonic mass is proportional to the column density and takes the form @xmath39 for the column density distribution function , the mass contribution from the damped systems can be estimated as @xmath40.\end{aligned}\ ] ] @xmath41 one problem with the power law representation is that if @xmath42 , as all current estimates indicate , then the total mass in damped systems diverges unless an upper bound to the hi column density is assumed .
for example , if we take 20.3 log @xmath43 22 , the fractional contribution to the total hi mass for damped systems , @xmath44 , is then @xmath45 for @xmath46 and @xmath47 for @xmath48 .
however , there is no a priori reason for assuming this upper limit and hence there is no strict upper bound to any estimate of the total hi mass in damped systems .
an alternative parameterisation using a gamma function to describe the hi column density distribution was adopted by pei & fall and provides an elegant solution to the diverging mass problem .
we discuss these points in more detail in section 3 and the redshift evolution of the absorbers in section 4 .
echelle spectra of four qsos were obtained in march , 1993 by s. dodorico as part of an eso key programme studying high redshift quasars .
they were taken at la silla with the 3.5 m ntt telescope using the emmi instrument in echelle mode using a 2048@xmath492048 pixel loral ccd as the detector .
a slit of 15 `` in length was used and generally the slit width was 1.2 '' .
two grating setups were used , one covering 4700 - 8300 and the other covering 5800 - 9500 with a resolution of @xmath3140 km s@xmath50 ( 1 ) .
[ see giallongo 1994 for more details . ]
the observations are summarised in table [ t_obs_eso ] .
.eso observations march , 1993 [ cols= " < , < , > , < , < " , ] a single power law fit to the combined data set described above of the form in equation [ fnzeqn ] with @xmath51 results in @xmath52 and @xmath53 with similar values for @xmath54 ( @xmath55 ) .
the quoted error for the normalisation constant @xmath56 is large because for a small change in the value of @xmath38 , @xmath56 can change by 2 orders of magnitude .
these results are in good agreement with the results found by lwtlmh ( @xmath57 , log k@xmath58 ) and are plotted for the entire data set in figure [ f_dlafn_all ] .
the data is binned for display purposes only with the vertical error bars plotted at the mean column density for each bin .
we will see in the next section that a single power law is not a good fit to the data .
as shown in paper i for the lyman - limit system evolution , the arbitrary binning of the data for presentation in differential plots includes a subjective component that can mask exactly what is happening in the underlying data .
a cumulative distribution plot is far better at revealing the true nature of the distribution and this approach is examined now .
the @xmath12 of the cumulative number of damped systems detected versus @xmath12 is plotted in figure [ f_dlacum](a ) . a point for the expected number of lyman - limit systems that would be detected down to @xmath12 @xmath59
is shown with a circled star .
this is calculated by integrating the number density per unit redshift ( @xmath60 ) over the redshift path covered by the @xmath61 qsos in the dla sample , @xmath62 @xmath63 it is obvious from figure 3(a ) that a power law will not fit the entire column density range 17.2 @xmath28 log @xmath28 22 .
a kolmogorov - smirnov ( k - s ) yields a probability of less than @xmath64 that the fit represents the underlying data set . in figures [ f_dlacum](b - d )
the same distribution is overplotted with single power law fits for different values of @xmath38 that were fit to the graph by eye .
( b ) shows that @xmath65 will fit from the lyman limit column density through the damped systems with @xmath12 @xmath66 , a flatter slope than the canonical 1.5 - 1.7 range .
( c ) shows that @xmath67 fits the damped distribution with @xmath68 @xmath69 well but does not describe the high or low column density tails of the distribution .
( d ) shows a fit to the sharp drop off in numbers for damped systems with @xmath12 21.3 .
this can be expected from looking at the estimated column densities for the damped systems in table [ t_dla ] or by looking at the spectra .
there are not a lot of heavily damped systems .
clearly the results for a single power law fit depend critically on the range of column densities included .
this characteristic can explain much of the variation in the results previously seen by various authors that were summarised in section 3.1 . to qualitatively study the redshift evolution of the column density distribution of the damped systems the cumulative distribution shown in figure [ f_dlacum ] has been split in half at redshift 2.5 and each set plotted individually ( figure [ f_zsplit ] ) .
the damped systems with @xmath70 are shown by the solid line and the absorbers with @xmath71 are shown by the dashed line .
the higher redshift absorbers appear to have a slightly flatter slope up to @xmath12 @xmath521 and then a sharper drop in the number of very high column density systems , though a k - s test shows that this is not a statistically significant difference .
the evolution with redshift in the slope of the column density distribution is also apparent when looking at the differential @xmath72 .
lwt plotted this in 3 redshift bins z@xmath5[0.008,1.5],[1.5,2.5 ] , and [ 2.5,3.5 ] . in the highest redshift bin
there was a flattening of the column density distribution slope towards higher column densities . in figure [ f_dlafn ]
we have plotted our combined data set with this same binning with the addition of one higher redshift bin z@xmath5[3.5,4.7 ] .
the flattening of the distribution function towards higher column density systems in the z@xmath5[2.5,3.5 ] bin in the lwt data is no longer pronounced .
the most striking feature is the steepness of the distribution in the highest redshift bin .
it is not just steeper due to a decrease in the highest column density systems ( log @xmath721 ) , but there is also an increase in the number of lower column density systems relative to the other bins . even if 15 - 20% of the candidate systems with log @xmath3020.3 turn out not to be damped when observed at higher resolution , as we expect , this result still holds
there are two strong motivating factors to find an alternative model for describing the hi column density distribution .
first , as shown in section 3.4 , there is direct evidence for an apparent variation in the power law slope as a function of @xmath43 .
this implies that a higher order functional form other than a power law is needed to describe the column density distribution .
second , as noted in the introduction , with a power law model the integral mass contained within damped ly@xmath73 systems is divergent for realistic values of @xmath38 .
this in turn means that it is impossible to assign a formal upper limit to any estimate of the neutral gas content of the early universe .
consequently , following pei and fall ( 1995 ) , we have chosen to model the data with a gamma distribution of the form @xmath74 where @xmath75 is the characteristic number of absorbing systems at the column density @xmath76 , and @xmath76 is a parameter defining the turnover , or ` knee ' , in the number distribution .
both @xmath75 and @xmath76 may in general vary with redshift but for the moment we treat them as constants .
this functional form is similar to the schechter luminosity function ( schechter 1976 ) .
for @xmath77 the gamma function tends to the same form as the single power law , @xmath35 ; whilst for @xmath78 @xmath76 , the exponential term begins to dominate .
we can understand how a gamma function might provide a better description of the damped ly@xmath73 data by considering the differential logarithmic slope which is given by @xmath79 as the column density approaches @xmath76 the slope begins to steepen and rapidly turns over at higher column densities , qualitatively similar to what we observe in figures [ f_dlacum ] and [ f_zsplit ] .
furthermore , the integral hi over the column density distribution ( equation [ himass1 ] ) for @xmath80 is now given by @xmath81 where @xmath82 denotes the standard gamma function .
this function is bounded if @xmath83 . the maximum likelihood technique outlined in appendix a
can readily be modified to incorporate this form .
we note that the likelihood solution can be found over a two - dimensional grid of pairs of values of @xmath76 and @xmath38 , since the constant @xmath75 can be directly computed using the constraint @xmath84 where m is the total number of observed systems .
this is computationally much less intensive than doing a 3-d grid search .
the results of a single functional fit to the entire dataset are log @xmath15 , @xmath16 , and @xmath17 .
the log - likelihood function results with confidence contours are shown in figure [ f_gamconf](a ) .
the best fit is overplotted on the differential form of @xmath72 in figure [ f_gamconf](b ) and on the cumulative distribution in figure [ f_gamconf](c ) .
( the single power law form of @xmath72 was shown fitted to the same data in figures [ f_dlafn_all ] and [ f_dlacum ] ) .
the differential form of the plots ( figures [ f_dlafn_all ] and [ f_gamconf](b ) ) show little difference between the single power law and @xmath85-distribution fits .
when displayed with the cumulative number of absorbers in figure [ f_gamconf](c ) , the @xmath85-distribution now clearly fits the entire data set with column densities log @xmath2 20.3 .
if the expected number of lyman - limit systems are included in the fit , the results are log @xmath86 , @xmath87 , and @xmath88 .
this also provides a reasonable fit to the data as shown in figure [ f_gamconf2](a - c ) .
differential evolution in the number density of damped absorbers has been described by lwt and wolfe ( 1995 ) .
while the change in number density per unit redshift is consistent with no intrinsic evolution of the absorbers over the range 0 @xmath8 z @xmath8 3.5 , they find that the systems with log @xmath89 disappear at a much faster rate from [email protected] to z@xmath50 than does the population of damped absorbers as a whole .
we now examine the redshift evolution of the damped absorbers in our combined data set by determining the number density of absorbers per unit redshift , @xmath90 . in a standard friedmann universe for absorbers with cross section @xmath91 and number density @xmath92 per unit comoving volume @xmath93
it is customary to represent the number density as a power law of the form @xmath94 where @xmath95 .
this yields @xmath25 = 1 for @xmath96 = 0 and @xmath25 = 1/2 for @xmath96 = 1/2 for the case of no evolution with redshift in the product of the number density and cross section of the absorbers @xcite
. a maximum likelihood fit to the data yields @xmath97 which is consistent with no intrinsic evolution even though the value of @xmath25 is similar to that found for the lyman limit systems where evolution is detected at a significant level ( paper i ; stengler - larrea 1995 ) .
the log - likelihood function for @xmath25 and @xmath98 with @xmath768.3% and @xmath795.5% confidence contours is plotted in figure [ f_maxz ] .
we also find redshift evolution in the higher column density systems but with a decline in @xmath26 for [email protected] .
these results are displayed in figure [ f_dndz ] .
the entire data set is plotted as dashed lines with the above fit .
the results for only the absorbers with log n(hi)@xmath99 are shown as solid lines .
figure [ f_zhi ] shows hi column density versus redshift , and the paucity of absorbers with log @xmath7 21 at z @xmath7 4 is apparent .
three qsos from apm survey have been observed at 0.8 resolution .
two have damped systems with confirmed hi column densities of @xmath2 10@xmath3 atoms @xmath4 , with a third absorber falling just below this threshold .
we have discovered the highest redshift damped absorber known at [email protected] in qso br1202@xmath60725 . the two systems with @xmath2 10@xmath3 atoms @xmath4 , and remaining nine candidate damped absorbers from the apm survey
have been combined with data from previous surveys to study the column density distribution and number density evolution for absorbers with @xmath2 17.2 .
if the hi column density distribution function is fit with a power law , @xmath10 , we find evidence for breaks in the power law , flattening for 17.2 @xmath28 log 21 and steepening for log 21.2 .
the column density distribution function for the data with [email protected] is better fit with the @xmath85-distribution form @xmath14 with log @xmath15 , @xmath16 , and @xmath17 . for the number density evolution of the damped absorbers
( log @xmath2 20.3 ) over the redshift range 0.008 @xmath8 z @xmath8 4.7 we find the best fit of a single power law form for @xmath18 yields @xmath23 and @xmath24 .
this is consistent with no intrinsic evolution in the absorbers even though the value of @xmath25 is similar to that found for the lyman limit systems where evolution is detected at a significant level .
evolution is evident in the highest column density absorbers with the incidence of systems with log n(hi)@xmath221 decreasing for z 3.5 .
* acknowledgments * we would like to thank bob carswell for providing software for and assistance with the data reduction and profile fitting of the spectra .
lsl acknowledges support from an isaac newton studentship , the cambridge overseas trust , and a university of california president s postdoctoral fellowship .
rgm acknowledges the support of the royal society .
, peebles p.j.e . , 1969 ,
apj , 156 , l7 r.f . , morton d.c . , smith m.g . ,
stockton a.n . ,
turnshek d.a .
, weymann r.j . , 1984 ,
apj , 278 , 486 a.j . , 1994 , ph.d . thesis .
cambridge university s.m .
, pei y.c . , mcmahon r.g . , 1989 ,
apj , 341 , l5
e. , dodorico s. , fontana a. , mcmahon r.g .
, savaglio s. , cristiani s. , molaro p. , trevese d. , 1994 , apj , 425 , l1 k.m . , wolfe a.m. , turnshek d.a . , 1995 , apj , 440 , 435 k.m . ,
wolfe a.m. , turnshek d.a .
, lu l. , mcmahon r.g . ,
hazard c. , 1991 , apjs , 77 , 1 lu l , sargent w.l.w .
, womble d.s . ,
barlow t.a . , 1996 ,
apj , 457 , l1 y.c .
, fall s.m . ,
bechtold j. , 1991 , apj,378 , 6 y.c .
, fall s.m . , 1995 , apj , 454 , 69 p. , webb j.k . , rauch m. , carswell r.f . , lanzetta k.m . , 1993 , mnras , 262 , 499 m. , boksenberg a. , hunstead r.w . , 1990 , apj , 348 , 48 m. , smith l.j . , hunstead r.w . , king d.l . , 1994 , apj , 426 , 79 m. , carswell r.f .
, robertson j.g . , shaver p.a .
, webb j.k . , 1990 , mnras , 242 , 698 w.l.w . , steidel c.c . ,
boksenberg a. , 1989 , apjs , 79 , 703 w.l.w .
, young p.t .
, boksenberg a. , tytler d. , 1980 , apjs , 42 , 41 p. , 1976 ,
apj , 203 , 297 p. , press w.h . , 1976 , apj , 203 , 557 stengler - larrea e.a . , , 1995 , apj , 444 , 64 l.j . ,
mcmahon r.g .
, irwin m.j . ,
hazard c. , 1994 , apj , 427 , l13 ( paper i ) l.j .
, mcmahon r.g .
, irwin m.j . ,
hazard c. , 1996 , apj , 468 , 121 ( paper ii ) l.j . ,
mcmahon r.g . , irwin m.j . , 1996 , mnras , in press ( paper iv
) d. , 1987 , apj , 321 , 49 e.j .
, williger g.m .
, baldwin j.a . , carswell r.f . ,
hazard c. , mcmahon r.g .
, 1996 , a&a , in press g.m . , baldwin j.a .
, carswell r.f . , cooke a.j . ,
hazard c. , irwin m.j . ,
mcmahon r.g .
, storrie - lombardi l.j . , 1994 , apj , 428 , 574 a.m. , 1987 , proc.phil.trans.roy.soc .
, 320 , 503 a.m. , turnshek d.a .
, smith h.e . ,
cohen r.d . , 1986 ,
apjs , 61 , 249 a.m. , lanzetta k.m . ,
foltz c.b . ,
chaffee f.h . , 1995 ,
apj , 454 , 698
using equation [ fnzeqn ] for the column density distribution function , the damped absorbers will be found randomly distributed according to this function along the qso line - of - sight in @xmath100 space .
if the space is divided into @xmath101 cells each of volume @xmath102 , the expected number of points in cell @xmath103 is given by @xmath104 the probability of observing @xmath105 points in cell @xmath103 is @xmath106 the likelihood function for qso@xmath107 taking the product over all the cells is then @xmath108 if the volume of each cell @xmath102 becomes very small such that there is either 1 or 0 points in each cell , @xmath109 then the likelihood can be rewritten separating out the terms for full and empty cells . for @xmath110 empty cells @xmath111 @xmath112 full cells @xmath113 taking the @xmath12 of the likelihood function gives @xmath114 @xmath115 ( schechter & press 1976 ) . ignoring the constant terms , in the limit where @xmath116 this becomes @xmath117 \end{aligned}\ ] ]
@xmath118 to get the overall log likelihood for @xmath61 qsos we evaluate the integrals in equation [ dlal4eqn ] and additively combine the @xmath12 l s resulting in @xmath119 \end{aligned}\ ] ] @xmath120 where @xmath121 is the number of detected dlas in qso@xmath122 and @xmath123 is the minimum column density . | eleven candidate damped absorption systems were identified in twenty - seven spectra of the quasars from the apm z 4 survey covering the redshift range @xmath0 ( 8 with @xmath1 ) .
high resolution echelle spectra ( 0.8 fwhm ) have been obtained for three quasars , including two of the highest redshift objects in the survey .
two damped systems have confirmed hi column densities of @xmath2 10@xmath3 atoms @xmath4 , with a third falling just below this threshold .
we have discovered the highest redshift damped absorber known at [email protected] in qso br1202@xmath60725 .
the apm qsos provide a substantial increase in the redshift path available for damped surveys for z@xmath73 .
we combine this high redshift sample with other quasar samples covering the redshift range 0.008 @xmath8 z @xmath8 4.7 to study the redshift evolution and the column density distribution function for absorbers with log @xmath9 . in the column density distribution @xmath10
we find evidence for breaks in the power law , flattening for @xmath1121 and steepening for @xmath12@xmath13 .
the breaks are more pronounced at higher redshift . the column density distribution function for the data with log
@xmath220.3 is better fit with the form @xmath14 with log @xmath15 , @xmath16 , and @xmath17 .
we have studied the evolution of the number density per unit redshift of the damped systems by fitting the sample with the customary power law @xmath18 . for a population with no intrinsic evolution in the product of the absorption cross - section and comoving spatial number density
this will give @xmath19 ( @xmath20 ) or @xmath21 ( @xmath22 ) .
the best maximum likelihood fit for a single power law is @xmath23 and @xmath24 , consistent with no intrinsic evolution even though the value of @xmath25 is also consistent with that found for the lyman limit systems where evolution is detected at a significant level .
however , redshift evolution is evident in the higher column density systems with an apparent decline in @xmath26 for [email protected] . 0q@xmath27 1 2 2oi+siii .
^m .
^s .
. cosmology galaxies : evolution galaxies : formation quasars : absorption lines
quasars : individual ( br1033@xmath60327 , bri1108@xmath60747 , br1202@xmath60725 ) |
the 4@xmath5-metals , and their alloys and intermetallic compounds exhibit rich pressure - temperature - composition phase diagrams . often , transitions between different phases are accompanied by a dramatic change in the material properties , including abrupt volume variation , and in particular the extent to which the @xmath5 electron localization changes under such phase transitions is of great scientific interest .
the transition between localized and itinerant behaviour is at the forefront of physics of strongly correlated systems .
such a transition occurs in the 5@xmath5 electron shell of the pure actinide metals - the 5@xmath5 electrons behave in a localized fashion in the heavy actinides but in a more delocalized manner in the light actinides - centering near pu and am .
a quantitative and conclusive understanding of this phenomenon is still missing.@xcite in the lanthanide series , where the 4@xmath5 orbitals are spatially less extended than the 5@xmath5 orbitals in the actinides , localized - itinerant transitions occur at the beginning ( around ce ) , in the middle ( sm , eu ) and at the end ( yb ) . application of high pressure leads to a 4@xmath5 delocalization and , in some cases , to structural phase transformations accompanied by a volume discontinuity .
the most famous example is cerium , which exhibits the isostructural @xmath6 volume - collapse phase transition upon either cooling or application of external pressure .
several interpretations have been suggested to explain the origin of this phenomenon however its nature is still widely debated.@xcite the intermediate - valence compound ceni represents another interesting example of a system experiencing a pressure - induced structural instability .
ceni has the crb - type orthorhombic crystal structure ( space group @xmath7 ) with the room - temperature crystal lattice parameters @xmath8 , @xmath9 , @xmath10 .@xcite the orthorhombic @xmath7 crystal structure repeatedly appears in rare - earth and actinide metals under pressure such as @xmath11-ce , pa , nd , and pr .
the @xmath12-u metal also has this type of structure which is usually stable up to very high pressures . of course , at hundreds of gpa of pressures , closed - packed metal structures , such as fcc , hcp , and bcc ,
are once again favored due to large electrostatic repulsions ruling out the more open and lower symmetry structures .
ceni shows a different behavior . already at ambient pressure ceni displays clear signatures of lattice instability upon cooling,@xcite but a structural transformation
does not occur down to the lowest temperature . in 1985
gignoux and voiron found a pressure - induced first - order structural phase transition in ceni and determined its phase @xmath1 diagram up to @xmath130.5 gpa and @xmath13150 k.@xcite later the ceni phase diagram was extended up to @xmath132 gpa and 300 k using neutron diffraction and magnetic measurements.@xcite however , neither the space group nor atomic positions of the pressure - induced ceni phase were established at the time .
here we report the results of two diffraction studies performed to investigate the pressure - induced structural phase transition in ceni .
the first study was performed at room temperature using x - ray diffraction , and the second study employed neutron powder diffraction at 100 k. in both the cases we observed a pressure - induced structural phase transition in ceni accompanied by a volume jump .
the results obtained show that the structure of pressure - induced ceni phase ( phases ) can be described with the @xmath0 space group .
we show that density functional theory ( dft ) calculations successfully reproduce the ground state volume , the phase transition pressure , and the volume collapse associated with the phase transition .
stoichiometric amounts of elemental reactants , ce ( chunks , ames ) and ni ( alfa ) were loaded into alumina crucibles , which were baked at 800@xmath14 c prior to use , and then placed inside quartz tubes .
additional alumina crucibles containing zr pieces were placed on top as potential oxygen sponges during sample synthesis process .
the quartz tubes were sealed under vacuum , and subsequently heated in box furnaces at 800@xmath14 c for 6 hours , followed by 1@xmath14 c / h cooling to 600@xmath14 c. the phase purity of the synthesized samples was verified using x - ray powder diffraction and magnetization measurements which showed that the desired ceni phase ( @xmath7 space group ) had formed with a small amount of an impurity phase ( @xmath15% of ceni@xmath16 ) .
magnetization measurements performed with a commercial physical properties measurement system ( quantum design ) showed good agreement with the published data.@xcite room temperature high - pressure synchrotron powder diffraction measurements were carried out at beamline 17-bm at the advanced photon source using a 100 micron monochromated x - ray beam at a wavelength of @xmath17 .
variable pressure diffraction data were collected _ in situ _ using a perkin elmer amorphous - si flat panel detector centered on the x - ray beam .
the sample - to - detector distance was nominally set at 300 mm , yielding an available 2@xmath18 scattering angle of 27.5 deg .
, corresponding to access of bragg reflections with @xmath19-spacing as low as 1.52 .
the diffractometer geometrical parameters ( such as precise sample - to - detector distance and tilt of the detector ) were optimized with respect to a nist a lab@xmath20 ( 660a ) standard .
the ceni powdered sample was loaded in a membrane - driven diamond anvil cell fitted with a pair of 800 micron culet diamonds .
the gasket was made from 250 @xmath21 m thick full - hard 301 stainless steel , pre - indented to about 120 @xmath21 m and drilled with a 400 @xmath21 m hole .
gold was added to the sample as a pressure manometer and silicone oil was used as pressure transmitting medium .
the pressure on the dac membrane was increased in 25 steps up to a maximum sample pressure of @xmath137.8 gpa .
the pressure values were determined by fitting the measured gold unit cell volume to a third - order birch - murnaghan equation of state ( eos)@xcite using the parameters v@xmath22
@xmath23 , b@xmath24 gpa , and b@xmath25 , where v@xmath26 is the au primitive cell volume at ambient conditions , b@xmath27 is the bulk modulus , and b@xmath28 is its first pressure derivative.@xcite time - of - flight neutron diffraction measurements at 100 k were performed at the spallation neutrons and pressure ( snap ) beamline of the spallation neutron source at oak ridge national laboratory .
neutron detectors were located at 50 deg . and 90 deg .
with respect to the incident beam .
the sample - to - detector distance was 50 cm .
the accessible @xmath29 range in this configuration extended from 30 deg . to 115 deg .
, up to 7.5 deg . out of the scattering plane .
the wavelength range of the neutrons was between 0.3 and 3.7 , which allows a sampling of bragg reflections between @xmath30 .
the incident beam final collimation is a rectangular aperture measuring 1@xmath315 mm , designed to match the sample shape and the beam divergence . for sample loading a pellet
was made from powdered ceni mixed with pb , which serves a dual role of pressure calibration standard and pressure transmitting medium .
pressure was generated in a paris - edinburgh high - pressure press fitted with boron nitride ( bn ) toroidal anvils@xcite and using a null scattering ( tizr ) alloy that adds no coherent bragg scattering to the data .
the equatorial configuration of the cell was chosen allowing a wide angular aperture for the scattered beam . for reference
, we also measured the room temperature neutron diffraction pattern of the pressure cell filled with ceni and pb at snap and used the standard value of the pb crystal lattice constant @xmath32 @xcite for fitting . to cool the pressure cell down to 100
k we used liquid nitrogen flow .
after cooling , the pb crystal lattice parameter was refined for every measurement to determine the actual pressure values using a third - order birch - murnaghan eos@xcite with measured value v@xmath26(@xmath33=100 k)@xmath34 @xmath23 and the parameters b@xmath35 gpa and b@xmath36 from work of schulte and holzapfel.@xcite to index the high - pressure ( hp ) phases we used the powdercell@xcite and dicvol06@xcite program packages .
neutron diffraction data reduction was done using the mantid software suite.@xcite structural parameters were refined with the fullprof software.@xcite
figure [ fig1 ] shows the integrated x - ray diffraction ( xrd ) intensities for the ceni sample along with the gold pressure standard as a function of increasing pressure at room temperature .
the peaks observed around 18 and 21 degrees are ( 111 ) and ( 200 ) bragg reflections of au used for pressure determination .
the low pressure ( lp ) ceni structure remains stable from ambient pressure up to @xmath37 gpa . between 0.7 and 2.1 gpa the xrd patterns of ceni exhibit remarkable changes .
in particular , the appearance of new peaks around 15 , 17.5 deg . and the suppression of diffraction peaks at 8 , 12 , 15.5 , 16 deg .
clearly indicate a pressure - induced phase transition . above 2.1 gpa only subtle variations of peak positions and intensities are observed , however upon further increase in pressure a new diffraction pattern develops suggesting one more structural transformation starting at @xmath38 gpa .
phase.[fig1 ] ] figure [ fig2 ] shows typical xrd patterns obtained for the lp and hp phases of ceni .
the ceni lp phase ( below @xmath39 gpa ) adopts the crb - type orthorhombic structure described by the @xmath7 space group .
the ceni primitive cell contains two ce and two ni ions , both located at ( 4c ) @xmath40 wyckoff positions . a rietveld refinement at ambient pressure resulted in the following structural parameters : @xmath41 , @xmath42 , @xmath43 , @xmath44(ce)@xmath45 , @xmath44(ni)@xmath46 , which is in good agreement with values previously reported in the literature.@xcite the measured xrd pattern of the hp phase ( pressure range between 2.1 and 4.9 gpa ) was indexed to belong to the feb - type orthorhombic crystal structure ( @xmath0 space group ) . for example , the rietveld refinement resulted in the following structural parameters at @xmath47 gpa : @xmath48 , @xmath49 , @xmath50 , @xmath51(ce)@xmath52 , @xmath53(ce)@xmath54 , @xmath51(ni)@xmath55 , @xmath53(ni)@xmath56 ( fig . [ fig2 ] ) .
within the structural transition region at @xmath57 gpa the x - ray diffraction pattern is described as a superposition of diffraction patterns resulting from 21@xmath582 volume percent of crb - phase and 79@xmath587 volume percent of feb - phase .
attempts to refine the ceni crystal structure from the powder diffraction patterns at pressures above 4.9 gpa ( hp@xmath59 phase ) did not enable us to extract reliable structural parameters or atomic positions .
thus we can not quantitatively describe the structure evolution in this pressure range .
it seems that above 4.9 gpa the diffraction patterns can no longer be ascribed to a single phase .
this issue will be discussed in section [ disc ] .
gpa ( b ) .
the red circles are the measured scattering intensity , and the black solid line represents the rietveld refinement fit to the data .
the vertical bars indicate bragg reflection positions of the main phase ( top ) and au ( bottom ) used for pressure determination .
the inserts show schematic view of the low pressure crb - type ( @xmath7 ) ( a ) and high pressure feb - type ( @xmath0 ) ( b ) of the ceni crystal structure.[fig2 ] ] in order to extend the phase diagram to lower temperatures we performed neutron powder diffraction study of the ceni structure under pressures up to @xmath60 gpa at @xmath61 k. our experimental data are summarized in fig .
it is seen that the initial crb - type structure is conserved at low pressures .
for example , a fullprof refinement gives the following structural parameters at @xmath62 gpa : @xmath63 , @xmath64 , @xmath65 , @xmath44(ce)@xmath66 , @xmath44(ni)@xmath67 ( fig.[fig4 ] , bottom ) . at @xmath68
gpa a modification in the ceni diffraction pattern is seen that is compatible with the expected @xmath4 symmetry change .
in fact , the @xmath7 ( 111 ) , ( 110 ) , and ( 020 ) reflections fully disappear between 0.96 and 2.94 gpa . on the other hand , the representative @xmath0 reflections , for instance , ( 211 ) and ( 011 )
appear in the same pressure region .
this means that ceni transforms into a mixed low - pressure - high - pressure state . at @xmath69 and 5.05 gpa the diffraction pattern varies again as compared to that observed at the lower pressures .
in addition to the reflections of @xmath0 feb phase , new bragg peaks appear , the most noticeable at @xmath70 .
indexing the ceni diffraction pattern at @xmath60 gpa with dicvol06@xcite leads to a new orthorhombic cell with crystal lattice parameters close to those found for the quenched modification of the tbni compound , @xmath71 , @xmath72 , @xmath73 ( @xmath0 space group).@xcite k.[fig3 ] ] this structure is closely related to the feb type structure with transformation equations @xmath74 ; @xmath75 ; @xmath76 . assuming the ceni high - pressure phase at 100 k to have this type of structure ( we will call it 3@xmath77 for a convenience ) , we performed rietveld refinement of the experimental data at 5.05 gpa and obtained a rather good agreement between the experimental and calculated diffraction patterns ( fig .
[ fig4 ] , top ) .
the resulting crystal lattice parameters are @xmath78 , @xmath79 , @xmath80 , close to that for the quenched tbni composition.@xcite moreover , we were able to describe the experimental data at @xmath68 gpa , assuming this state to be a mixture of low - pressure @xmath7 and high - pressure 3@xmath77 structures ( fig .
[ fig4 ] , middle ) . according to the fullprof analysis ,
the volume fraction of low - pressure phase is 37@xmath581% , and the high - pressure phase occupies 63@xmath581% of the sample volume .
however , we note that a complete fullprof refinement of the high - pressure diffraction patterns is hardly possible due to large number of fitting parameters ( for 3@xmath77 structure there are three nonequivalent ce ions and three nonequivalent ni ions located at ( 4c ) @xmath81 positions ) .
moreover , the peak broadening at high pressure leads to significant overlap of reflections making the fit of atomic positions ambiguous .
finally , ceni ingots always possess strong texture ( or preferred orientations ) which can not be averaged out completely by milling the ingots to powder . as a result
, a texture effect is seen clearly even in the x - ray diffraction pattern measured at ambient pressure ( fig .
[ fig2 ] ) .
because of these circumstances , the calculated diffraction patterns for @xmath68 and 5.05 gpa in fig . [ fig4 ] were obtained with the atomic coordinates equal to those reported for tbni in ref . and fixed .
a similar procedure of fixing atomic positions was also applied for the fullprof analysis of the mixed - phase room - temperature x - ray diffraction pattern at @xmath57 gpa .
structure , top panel ) , and 0.96 gpa ( mixed lp - hp state , middle panel ) .
vertical bars indicate bragg positions for lp phase ( the upper rows at 0.15 and 0.96 gpa ) , for pb ( the lower row ) , and for hp phase ( the middle row at 0.96 and the upper row at 5.05 gpa ) . [ fig4 ] ]
figure [ fig5 ] shows the pressure dependences of the crystal lattice parameters for the lp crb and hp feb phases of ceni as obtained from experimental xrd measurements at room temperature . in the pressure range below 1.2 gpa ( lp phase including phase coexistence region )
the parameters @xmath82 , @xmath83 , and @xmath84 shrink by @xmath132.8% , @xmath130.8% , and @xmath130.8% , respectively , so that the change in volume is about 4.4% .
the parameters @xmath82 , @xmath83 , and @xmath84 of the hp feb lattice are reduced by @xmath131.1% , @xmath131.1% , and @xmath130.6% , respectively , when going from @xmath57 gpa to 4.86 gpa .
the corresponding change in volume of the hp phase is about 2.8% . for a comparison , as seen from fig .
[ fig6 ] , the volume jump at the transition ( @xmath57 gpa ) is @xmath131.3% only at room temperature ( @xmath85 k ) .
however at lower temperature ( @xmath61 k , @xmath68 gpa ) the jump increases to 7.9% .
volume vs. pressure dependences shown in fig .
[ fig6 ] were analyzed using a third order birch - murnaghan eos.@xcite the obtained eos parameters ( see table [ table1 ] ) indicate that the ceni hp @xmath0 structure is less compressible than the lp @xmath7 phase . [
cols="<,^,^,^,^,^,^ " , ] the experimental results described above allow us to draw some conclusions about the @xmath1 phase diagram of ceni ( fig . [ fig7 ] ) .
first , we found that at room temperature the lp @xmath7 structure of ceni transforms into the feb - type of structure belonging to the @xmath0 space group .
the latter structure is typical for the @xmath86ni compounds where @xmath86 is the rare - earth metal from the second half of the lanthanide series , while light lanthanides , including cerium , form the crystal lattice of the crb type.@xcite therefore , the @xmath0 symmetry is more favorable for the small rare earth ionic volume than @xmath7 .
both the feb and crb type structures contain a common structural unit , the trigonal prism , which is stacked differently to form either structure.@xcite if the @xmath53 value of the @xmath87 sites in the @xmath0 structure goes to zero , one obtains the higher - symmetry @xmath7 structure.@xcite thus , the pressure - induced conversion from a crb to a feb type structure is not surprising . furthermore , hohnke and parth@xcite established the set of transformation equations between the crystal lattice parameters of these two structures : @xmath88 for this `` ideal '' case the volumes of feb and crb structures are equal .
the room temperature lattice parameters of the crb structure at @xmath57 gpa ( two phases coexist at this pressure ) are @xmath89 , @xmath90 , @xmath91 (fig .
[ fig5 ] ) . for these parameters ,
( [ eq1 ] ) predict the feb structure to have @xmath92 , @xmath93 , @xmath94 , so that deviations from the experimental values at @xmath95 k and @xmath57 gpa are -4% , -2% , and + 7% , respectively . the fact that the experimental feb crystal lattice parameters are found to be close to the expected `` ideal '' values is an additional argument in favor of @xmath0 symmetry of room temperature ceni phase with pressure range between 1 and 5 gpa .
the deviation of the experimental unit cell parameters from the `` ideal '' values is the result of a @xmath131.3% volume jump due to the transition .
similar behavior of the unit cell parameters is found also at @xmath61 k. at @xmath68 gpa we have @xmath96 , @xmath97 , @xmath98 for lp phase . equations ( [ eq1 ] ) give @xmath99 , @xmath100 , @xmath101 , while experimental values are @xmath102 , @xmath103 , @xmath104 .
the volume jump at the transition ( @xmath61 k , @xmath68 gpa ) is @xmath137.9% , that is , as expected , much bigger than at room temperature . ] now we can establish an approximate @xmath1 phase diagram of ceni ( fig .
[ fig7 ] ) . at @xmath85
k , the @xmath4 ( crb @xmath105 feb ) transformation starts at @xmath130.7 gpa and ends at @xmath132.1 gpa ( see fig .
[ fig1 ] ) .
the middle point @xmath106 gpa almost coincides with the empirical @xmath107 transition line reported in ref . .
at @xmath61 k and @xmath68 gpa the ceni sample is found to be inside the transition region ( fig . [ fig6 ] ) , however not with the pure feb @xmath0 structure but its 3@xmath77 modification .
we did not observe the pure feb structure at 100 k. thus , we can not say whether or not there is a pressure window for the feb structure between the transition line established in ref . and confirmed in ref . by magnetic measurements , and the transition to the 3@xmath77 at around 1 gpa ( see fig .
[ fig7 ] ) .
we believe that the feb structure _ may _ appear here and then transforms , almost immediately , under increase in pressure into the observed 3@xmath77 structure .
such a behavior can not be excluded since one of these structures ( crb and feb ) can transform to the other directly or via a sequence of polymorphous transformations.@xcite in fact , a very similar picture is observed at 300 k. the crb @xmath105 feb conversion does occur at 300 k , and then a second structure modification takes place at pressure above 4.9 gpa .
k ( red triangles ) and from the npd measurements at @xmath61 k ( blue circles ) .
solid lines are the results of fitting the measured unit cell volume to a third - order birch - murnaghan eos@xcite with the parameters given in table [ table1 ] .
note , that for the hp phase at 100 k the reduced to two formula units primitive cell @xmath108 is plotted .
[ fig6 ] ] as mentioned in section [ experiment ] , the x - ray diffraction patterns of ceni above this pressure can no longer be described by a single phase , neither pure feb nor its polymorphous modifications of the @xmath109 type ( @xmath110 etc . ) .
therefore , it seems quite reasonable to assume that at @xmath111 gpa and at room temperature the ceni sample consists of a mixture of hp feb phase and @xmath109 structures , which can be called `` @xmath0-based mixed phase structure '' .
if this is true , then the approximate ceni @xmath1 phase diagram takes the form shown in fig .
[ fig7 ] with the second transition line separating the feb and @xmath0-based mixed phase structural states .
note , that an extrapolation of this transition line to @xmath112 k goes to a critical pressure value @xmath113 gpa , close to that obtained by specific heat measurements under pressure in ref . .
we can not exclude , however , a more complicated ceni structure at pressures above , say , 5 gpa .
the appearance of low scattering angle bragg reflections in the ceni xrd pattern at room temperature can also be considered to be the result of a coexistence of mixed feb - crb stacking variants similar to that found in @xmath114ni systems ( @xmath115 gd and @xmath116 tb).@xcite however , the presence of the feb - crb mixture at high pressures does not look plausible since it requires reappearance of crb phase . the question remains open whether such a coexistence of phases is of intrinsic origin or is a consequence of non - hydrostatic compression conditions and/or stress between grains of the powder samples .
nevertheless , based on our experimental results we can conclude that the structure of pressure - induced ceni phase has @xmath0 symmetry at least at pressures below @xmath135 gpa .
phase diagram of ceni as follows from the x - ray and neutron powder diffraction measurements .
green squares correspond to the room - temperature transitions from crb ( @xmath7 ) low - pressure to feb ( @xmath0 ) high - pressure phase and the second transition to the @xmath0-based mixed phase region ( hp@xmath117 phase in fig .
[ fig1 ] ) .
yellow triangles and blue solid line represent the experimental data ( magnetic measurements ) and their approximation , respectively , from ref . .
the results of magnetic measurements from ref . and
specific heat data from ref . are shown by red diamonds and blue circle , respectively .
dashed - dotted line indicates the conventional transition line from the feb - type of structure to the @xmath0-based mixed phase state .
[ fig7 ] ] the relative stability of the @xmath7 and @xmath0 structures of ceni under pressure can be understood by using density functional theory ( dft ) .
note , that we restrict our theoretical analysis to the crb ( @xmath7 ) @xmath105 feb ( @xmath0 ) transition .
first - principles calculations were performed using the vienna _ ab initio _ simulation package ( vasp)@xcite based on projector augmented wave ( paw ) pseudopotentials@xcite within the generalized gradient approximation ( gga ) as parameterized by perdew , burke , and ernzerhof ( pbe).@xcite after careful convergence tests , a plane wave cutoff energy of 350 ev and 15@xmath315@xmath3111 @xmath118-point monkhorst - pack mesh@xcite for the 8-atomic unit cell of ceni @xmath7 lp phase and 9@xmath3113@xmath3111 mesh for the @xmath0 hp phase were found to be sufficient to converge the total energy within 2 mev / cell .
bulk moduli have been computed by fitting to a birch - murnaghan equation of state.@xcite ( circles ) and hp @xmath0 ( diamonds ) structure as obtained from dft calculations .
arrow indicates the structural transition at pressure @xmath119 gpa .
[ fig8 ] ] first , we started from the ground states of the @xmath7 and @xmath0 structures at ambient pressure .
we determine an equilibrium state by starting from experimentally determined parameters and letting the lattice constants and the atomic positions relax .
after the systems have relaxed the total energy difference between them turns out to be only 0.049(2 ) ev / cell , with the @xmath7 structure being the lower one .
next , we carried out calculations of the total energy of ceni in both phases for different lattice parameters , which can imitate the influence of pressure .
for every pressure the atomic positions were allowed to relax keeping the lattice parameters fixed .
one sees in fig .
[ fig8 ] that with increasing pressure ( or with decrease in the lattice parameters in accordance to that found in our xrd measurements ) the energies of the lp and hp phases cross at about 0.94 gpa .
this confirms the stability of the @xmath7 ceni crystal structure at ambient pressure down to the lowest temperature and , at the same time , provides an explanation for the @xmath4 structural phase transition at @xmath131 gpa .
the computed unit cell volumes and bulk moduli are found to be in a reasonable agreement with the experimental values ( table [ table1 ] ) .
note , that the calculated bulk modulus of the @xmath7 structure turns out to be lower than that of the @xmath0 structure , consistent with the experiment .
phase at ambient pressure , ( b ) @xmath7 phase at @xmath120 gpa and ( c ) @xmath0 phase at @xmath120 gpa ( c ) .
zero in the horizontal axis represents the fermi energy ( @xmath121 ) .
the insets show the same dos in the vicinity of @xmath121 on the expanded scale .
[ fig9 ] ] the electronic densities of states ( dos ) for ceni on both sides of the structural phase transition resulting from our dft - gga calculations are shown in fig .
[ fig9](a , c ) . for comparison , the anticipated dos for a @xmath7 phase at @xmath122 gpa
is also shown ( fig .
[ fig9](b ) ) .
as expected , dft calculations give metallic behavior of ceni for both lp and hp phases , as indicated by a non - zero dos at the fermi level ( @xmath121 ) .
the site projected dos of ni 3@xmath19 and ce 4@xmath5 orbitals overlap in the vicinity of @xmath121 , suggesting their hybridization .
[ fig9 ] demonstrates a rather weak but distinct impact of a moderate pressure , @xmath123 gpa , on the electronic states of ceni .
it can be seen that the structural transformation affects the dos more than the pressure in the persistent structure , the ni 3@xmath19 states near @xmath121 being the most affected . as mentioned in introduction , ceni is a well known intermediate - valence system , characterized by the magnetic susceptibility of enhanced pauli type at low temperature and passing through a broad maximum at around 140 k , a moderately high value of the electronic specific heat coefficient @xmath124 mj mol@xmath125 k@xmath126,@xcite and the characteristic spin fluctuation energy ( kondo scale ) @xmath127 mev.@xcite the intermediate - valence behavior of ceni implies a rather strong hybridization of the 4@xmath5 electrons with ni 3@xmath19 states .
the pressure - induced volume - collapse structural transition indicates essential changes of the ceni electronic states . what can be deduced about these changes from the structural data obtained in the present study ?
let us consider the variations of interatomic distances in ceni due to the structural transition from the ambient pressure crb - type of structure to feb - type of structure ( @xmath47 gpa ) at room temperature .
ce ions in the @xmath7 structure have seven ni ions in the first coordination sphere at nearly equal distances r@xmath128(lp ) = 2.938 ( 2 ions ) , r@xmath16(lp ) = 2.948 ( 4 ions ) , and r@xmath129(lp ) = 3.011 ( 1 ion ) .
the next coordination sphere consists of eight ce ions at r@xmath130(lp ) = 3.601 ( 2 ions ) , r@xmath131(lp ) = 3.752 ( 4 ions ) , and r@xmath20(lp ) = 3.771 ( 2 ions ) . in the @xmath0 structure ,
ce ions are surrounded by six nearest neighboring ni ions at r@xmath128(hp ) = 2.357 ( 1 ion ) , r@xmath16(hp ) = 2.761 ( 1 ion ) , r@xmath129(hp ) = 2.794 ( 2 ions ) , and r@xmath130(hp ) = 2.980 ( 2 ions ) .
the second coordination sphere of the @xmath0 structure , as well as @xmath7 structure , contains eight ce ions at r@xmath131(hp ) = 3.525 (2 ions ) , r@xmath20(hp ) = 3.611 ( 2 ions ) , r@xmath132(hp ) = 3.773 ( 4 ions ) .
thus , ce - ni interatomic distances shrink essentially due to @xmath4 transition . ce - ce interatomic distances decrease also , but their variation is less pronounced .
the decrease of the ce - ni distances suggests an enhanced ce 4@xmath133 ni 3@xmath19 hybridization in the high pressure ceni phase as compared to the ambient pressure @xmath7 structure . as a result
, one can expect an increase of the characteristic magnetic fluctuation energy @xmath134 , and a correlated decrease of the electronic specific heat coefficient @xmath135 and the low temperature value of the pauli - like magnetic susceptibility ( both these values are generally inversely proportional to @xmath134 , which plays a role of an effective band width ) . obviously , these expectations agree with the experimental results.@xcite in fact , at an applied pressure of about 0.5 gpa both the electronic specific heat coefficient @xmath135 ( see fig . 3 in ref . [ ] ) and the magnetic susceptibility @xmath136(@xmath137 k ) ( red curve in fig . 1 of ref .
[ ] ) decrease by the same factor of about 1.4 .
note , that the smoothing of the peak structure of ni 3@xmath19 dos in the vicinity of @xmath121 ( fig .
[ fig9](c ) ) may also be interpreted as a result of an enhanced 4@xmath133 3@xmath19 hybridization due to the structural transition .
however , to achieve more direct and complete understanding of the evolution of the 4@xmath5 electronic states across the volume - collapse transition ( including the variation of @xmath134 and the 4@xmath5 occupation number as a function of pressure ) inelastic neutron scattering experiments are required .
by means of x - ray and neutron powder diffraction measurements we have shown that pressure induces in ceni structural phase transitions from a low - pressure crb - type of structure ( @xmath7 space group ) to high - pressure phases belonging to the @xmath0 space group .
the experimental results allow us to draw an approximate @xmath1 phase diagram of ceni according to which ceni undergoes two successive phase transitions within the @xmath1 domain @xmath2 gpa , @xmath138 k. the first transition converts the lp structure into the feb ( @xmath0 ) structure ( which was not , however , observed experimentally at @xmath61 k ) , while the second transition separates the feb structure ( hp phase ) and @xmath0-based mixed phase hp@xmath117 state .
dft calculations predict that at ambient pressure the total energies of the @xmath7 and @xmath0 crystal structures are finely balanced , the latter energy being by 0.05 ev per unit cell higher than the former one . in terms of a dft approach the pressure - induced structural phase transition in ceni
can be understood as a result of the higher compressibility of lp phase as compared to the hp phase : the lp total energy rises rapidly with pressure leading to a phase transition at the estimated pressure of 0.94 gpa .
we are grateful to m.v .
ryzhkov for enlightening discussions on the dft calculations .
research at oak ridge national laboratory s spallation neutron source was supported by the scientific user facilities division , office of basic energy sciences , a u.s .
department of energy .
part of this work was supported by the materials sciences and engineering division , a u.s .
department of energy , basic energy sciences .
this research used resources of the national energy research scientific computing center , which is supported by the office of science of the u.s .
department of energy under contract no .
de - ac02 - 05ch11231 .
lawrence livermore national laboratory is operated by lawrence livermore national security , llc , for the u.s .
department of energy , national nuclear security administration under contract de - ac52 - 07na27344 .
jgt wishes to thank the office of advanced simulation and computing for support of his travel to oak ridge national laboratory for experiments at the spallation neutron source .
work at rfnc - vniitf was supported in part by contract b601122 between llnl and rfnc - vniitf .
this research used resources of the advanced photon source , a u.s .
department of energy ( doe ) office of science user facility operated for the doe office of science by argonne national laboratory under contract no .
de - ac02 - 06ch11357 . | the pressure - induced structural phase transition in the intermediate - valence compound ceni has been investigated by x - ray and neutron powder diffraction techniques . for the first time
it is shown that the structure of the pressure - induced ceni phase ( phases ) can be described in terms of the @xmath0 space group .
equations of state for ceni on both sides of the phase transition are derived and an approximate @xmath1 phase diagram is suggested for @xmath2 gpa and @xmath3 k. the observed @xmath4 structural transition is analyzed using density functional theory ( dft ) calculations , which successfully reproduce the ground state volume , the phase transition pressure , and the volume collapse associated with the phase transition . |
the casimir effect @xcite is presently well known due to the many potential applications in both fundamental physics and nanotechnology .
the casimir force is an extension of the van der waals interaction to larger separations between macroscopic bodies where the relativistic retardation of the electromagnetic interaction becomes important due to the finite speed of light .
the casimir effect originates from the existence of zero - point and thermal oscillations in restricted quantization volumes .
the results of extensive experimental and theoretical studies on the role of the casimir force in configurations with idealized boundary surfaces ( infinitely thin , with ideal - metal boundary conditions etc . ) and also between surfaces of real material bodies with account of roughness , nonzero skin depth and nonzero temperature are presented in ref .
@xcite ( see also earlier books and reviews @xcite ) .
interrelationship between experiment and theory in the case of real materials was specially reviewed in ref .
@xcite . in condensed matter physics
the casimir force is taken into account in the investigation of various properties of thin films , surface tension , in atomic force microscopy and critical phenomena @xcite .
when the characteristic sizes of microdevices shrink below a micrometer , the role of the casimir and electric forces become comparable .
this opens new opportunities for the creation of nanoscale devices actuated by the casimir force @xcite .
investigation of the combined action of the electric and casimir forces in microdevices is vital for the resolution of issues associated with their stability @xcite .
the most universally known _
normal _ casimir force acts in the direction perpendicular to the interacting surfaces .
however , when the material properties of the interacting bodies are anisotropic or they are asymmetrically positioned , a _
lateral _ casimir force may exist which acts tangential to the surface .
similar to the normal casimir force , the lateral force originates from the modification of electromagnetic zero - point and thermal oscillations by material boundaries . for plates made of anisotropic materials , the lateral casimir force and related torques
were theoretically predicted in refs .
@xcite ( see also refs .
@xcite ) . for two parallel ideal metal plates covered with uniaxial sinusoidal corrugations of equal periods ,
the lateral casimir force was predicted@xcite and calculated@xcite in the second perturbation order with respect to the amplitudes of corrugations . in both cases
a sinusoidal dependence of the result on the respective characteristic angle was obtained . for sinusoidally corrugated plates made of real metal described by the plasma model ,
the lateral casimir force was considered in refs .
@xcite in second order perturbation theory . in the case of two ideal metal parallel plates covered with periodic uniaxial corrugations of rectangular shape
, some results for the normal casimir force were found in refs .
it is pertinent to note that the use of normal and lateral casimir forces gives the possibility to actuate both normal and lateral translations of corrugated surfaces in micromachines by means of the electromagnetic zero - point oscillations .
the lateral casimir force might be used to solve the tribological problems plaguing the microdevice industry . specifically , it was suggested @xcite to use this effect for frictionless transmission of lateral motion .
there are also proposals @xcite to measure the casimir torque arising between the anisotropic bodies due to the lateral casimir force .
both the normal and lateral casimir forces are in general nonadditive and possess a complicated dependence on the geometrical shape of the boundary surfaces connected with diffraction effects . in the configuration of a smooth al coated sphere above a sinusoidally corrugated al coated plate the nontrivial behavior of the normal casimir force
was experimentally demonstrated @xcite .
this was done in an additive regime when the sphere - plate separation is much smaller than the period of corrugations @xcite .
recently @xcite the nontrivial behavior of the normal casimir force caused by the complicated geometry of a boundary surface was observed in the configuration of a smooth au coated sphere above a si plate covered with rectangular corrugations ( trenches ) . in this case
the nonadditive regime was used where separation distances are of the order of corrugation period .
the deviations of the measurement data from the additive theory were reported @xcite .
these deviations are , however , about 50% less than those expected for ideal metal boundaries in the theory taking exact account of the geometrical shape @xcite .
the remaining discrepancy might be explained by the role of a nonzero skin depth which was not accounted for in ref .
thus , the scattering approach for grating geometries taking the exact account of surface geometry at zero temperature and describing metallic properties by means of the simple plasma model @xcite was found to be in better agreement with the data of ref .
@xcite .
the first measurement of the lateral casimir force was performed between the surfaces of an au coated sphere and an au coated plate both covered with aligned uniaxial corrugations of sinusoidal shape @xcite .
the corrugation amplitudes were @xmath0 nm on the plate and @xmath1 nm on the sphere .
the period of corrugations was equal to 1200 nm , i.e. , much larger than separation distances between the sphere and plate .
this means that the experiments of refs .
@xcite was done in the application region of the proximity force approximation ( pfa ) @xcite .
the lateral casimir force with an amplitude of @xmath2n at the shortest separation of 221 nm was found to sinusoidally oscillate as a function of the phase shift between the corrugations .
the total experimental error of the force amplitude measurement at the closest point was determined to be around 19% at the 95% confidence level @xcite .
the experimental results were compared with theory using the pfa which should supposedly be applicable at separations much smaller than the corrugation period . in so doing the dielectric properties of au
were described using the simple plasma model in fourth order perturbation theory with respect to a small parameter related to the ratio of the skin depth to the separation distance .
the experimental results were found to be in good agreement with the theory in the limits of the experimental error .
it was also predicted theoretically ( but not observed experimentally due to the use of small corrugation amplitudes and insufficient precision ) that the lateral casimir force is asymmetric , i.e. , deviates from purely sinusoidal harmonic dependence on the phase shift , through the contribution of higher order terms in the perturbation expansion with respect to the corrugation amplitudes . in this paper we present complete measurement data of the recently performed experiment on the lateral casimir force between a sinusoidally corrugated sphere and plate covered with au layers
. the much smaller period of corrugations used in this experiment provides opportunities to measure deviations of the lateral casimir force from the theoretical predictions obtained on the basis of the pfa .
much deeper corrugations on both the sphere and the plate and shorter separations where the measurements were performed make it possible to demonstrate the asymmetry of the lateral casimir force , as predicted in ref .
the measurement data are compared with the exact theory describing the rayleigh scattering of the electromagnetic oscillations on the sinusoidally corrugated boundary surfaces of a sphere and a plate with no fitting parameters .
only the configuration of a perfect sphere near a plane plate is treated using the pfa . at the separations
considered , this leads to only a negligibly small error of order 0.1% @xcite .
the dielectric properties of au are described by means of the generalized plasma - like model @xcite which takes into account the interband transitions of core electrons .
the computational results for the amplitude of the lateral casimir force as a function of separation and for the lateral casimir force as a function of the phase shift at different separations obtained using the exact theory at the laboratory temperature ( @xmath3k ) are found to be in a very good agreement with the measurement data .
the same data deviate markedly from the theoretical results computed using the pfa .
this clearly demonstrates the influence of diffraction effects on the lateral casimir force .
some of the results of this paper related to the first set of our measurements were briefly published in ref .
@xcite . here ,
we present additional phase dependences of the lateral casimir force not included in ref .
@xcite for the same corrugation amplitudes , as were used in @xcite , and also the results of further experiments performed with a much larger corrugation amplitude on the sphere .
we demonstrate that the experimental data of various measurements for different samples are in mutual agreement .
we also present additional important evidence characterizing the quality of corrugated surfaces , procedures of electrostatic calibrations , and details of error analysis .
calculations using the pfa are done by means of the exact lifshitz formula rather than with the help of perturbation theory in corrugation amplitudes , as in ref .
the role of surface roughness is estimated .
the details of exact computations of the lateral casimir force using the rayleigh scattering theory at nonzero temperature given in this paper have never been presented previously .
the paper is organized as follows . in sec .
ii we describe the experimental setup and sample preparation .
section iii contains the description of the procedures of electrostatic calibration using both the normal and lateral electric forces . in sec .
iv the measurement data and the error analysis are presented .
section v is devoted to the computation of the lateral casimir force between corrugated surfaces using the pfa . in sec .
vi the lateral casimir force is computed in a more fundamental way using the rayleigh theory . in both secs .
v and vi the theoretical results are compared with the experimental data .
section vii contains our conclusions and discussion .
a schematic diagram of the experimental setup is shown in fig . [ setup ] .
the experiments under discussion are performed using a standard atomic force microscope ( afm ) in oil - free vacuum chamber at a pressure below 10mtorr and at room temperature .
the observation of the lateral casimir force is done using two test bodies whose surfaces are covered with longitudinal sinusoidal corrugations of the same period .
the axes of the corrugations should be perfectly parallel . following @xcite
, one can see that misalignment by @xmath4 of the axes of corrugations can lead to the loss of any lateral force . as the first test body , we have used a sinusoidally corrugated grating of size @xmath5 vertically mounted on the piezotube of an afm ( see fig .
[ setup ] ) . here ,
the corrugations have an average period @xmath6 nm ( i.e. , more than two times smaller than in refs .
@xcite in order to achieve the regime where the pfa becomes inappropriate for the calculation of the lateral casimir force between the corrugated surfaces ) .
the amplitude of corrugations on the grating @xmath7 nm was a factor of 1.45 larger than in refs .
a @xmath8 m long v - shaped silicon nitride cantilever of the afm was first uniformly coated with 40 nm of al .
this modification of cantilever was performed to improve its thermal and electric conductivity and prevent deformation due to differential thermal expansion in vacuum .
a polystyrene sphere of a @xmath9 m diameter was placed at the end of the cantilever with conductive ag epoxy .
next , a freshly cleaved mica sheet of @xmath10 m length , @xmath11 m width , and a few micrometer thickness was attached to the bottom of the sphere also with ag epoxy .
then a second polystyrene sphere with a nominal diameter @xmath12 m was attached to the bottom free end of the mica sheet ( see fig . [ setup ] ) .
after imprint of corrugations on its surface ( see below ) , this last sphere was used as the second test body in the measurement of the lateral casimir force .
we emphasize that the silver epoxy is rigid at all attachments . the first sphere and mica sheet are needed to isolate the laser reflection spot on the cantilever tip from the interaction region between the two corrugated surfaces . as a result , the effect of scattered light from the top and sides of the corrugated plate is substantially reduced .
the resulting system ( cantilever , mica sheet and the attached two spheres ) was uniformly coated with a 10 nm layer of cr and then with a more than 50 nm layer of au in a thermal evaporator .
uniformity of coating was ensured through rotation of the cantilever @xcite .
the cr layer was used to improve the adhesion of the au coating and prevent its peeling during the imprint of the corrugations on the sphere .
the lateral casimir force arises when the uniaxial corrugations on the two test bodies are perfectly aligned and have the same period @xmath13 @xcite . to ensure that these conditions are satisfied ,
we have imprinted corrugations on the second sphere using the grating as a template . in order to obtain deeper corrugations than in refs .
@xcite , a more sophisticated procedure was implemented . here , the template grating had a 300 nm au coating applied by the manufacturer on top of sinusoidal corrugations made of hard epoxy on a 3-mm - thick pyrex substrate .
note that hard epoxy does not require the use of the oxide of al layer which was coated on the soft plastic grating in refs .
. the smaller period of corrugations used here required the use of a precise stepper motor and piezocontrolled imprinting technique .
the imprinting procedure was done as follows . at first
, the second sphere was brought in contact with the grating using a micromanipulator .
then a hard flat stylus ( see fig . [ setup ] ) was moved in @xmath14 m steps using a stepper motor until it comes into contact with the other side of the sphere which is now sandwiched against the grating .
next the imprinting was done by applying a voltage to the @xmath15 piezo to gently press the sphere between the grating and the hard flat stylus . to obtain deeper imprints while preserving sphericity some additional pressure was put on the sphere using the stepper motor .
special care was needed at this step .
if the applied pressure from the hard flat stylus was too much , the resulting rotation of the second sphere will cause the cantilever to break .
there is less concern on the preservation of sphericity as polystyrene spheres are elastic .
then the voltage on the @xmath15-piezo is reversed to remove the pressure on the sphere . with the help of the same stepper motor the hard flat stylus
was gently removed and the sphere was then translated horizontally by around @xmath11 m to a different position on the grating .
the last operation is necessary because the part of the grating used to imprint corrugations on the sphere might have changed its local amplitude during the imprinting process .
this translation was performed preserving the orientation of the cantilever . the same imprinting procedure , but with application of larger pressure by means of the stepper motor , was done with a similar sphere in the second set of our measurements .
after the imprint process was done , the system was left in vacuum for more than 2 hours to reach a stable equilibrium .
the corrugations on the grating and on both spheres were examined using an afm and found to be rather homogeneous .
this was done after the completion of the experiment .
an afm scan of the au - coated corrugations of the vertical template grating is shown in fig .
[ rghpl](a ) .
the value of the amplitude of these corrugations @xmath16 , also measured using the afm , is indicated above .
the variance of the stochastic roughness on the grating was estimated in the following way .
the topography of the groove from an afm scan [ see fig .
[ rghpl](a ) ] was fitted to a sine function .
note that the top and bottom of corrugations are symmetric as expected for a sine .
this is illustrated in fig .
[ rghpl](b ) where a typical section of fig .
[ rghpl](a ) along a @xmath17 plane is shown .
then the difference of heights between the groove topography and the sine function was calculated at a sufficient number of intermediate points .
the variance was found using these differences as the root - mean - square deviation between the two curves .
a total 30 grooves were used to evaluate the average value of the variance leading to @xmath18 nm . in fig .
[ rghsph](a ) we show an afm scan of the corrugations imprinted on the sphere used in the first set of our measurements .
the amplitude of these corrugations over a @xmath19 area on the sphere was measured with an afm to be @xmath20 nm ( i.e. , an increase by a factor of 1.7 as compared with refs .
figure [ rghsph](b ) is a typical section of fig .
[ rghsph](a ) along a @xmath17 plane .
it illustrates a symmetry between the top and bottom of the imprinted corrugations .
the variance of the stochastic roughness of the imprinted grooves @xmath21 was found using the same procedure as for a grating with the result @xmath22 nm . in fig .
[ sph1 ] the afm scan of a @xmath23 area of imprinted corrugations on the sphere used in the first set of measurements is demonstrated . here
, the lighter tone shows higher points , as indicated in the additional scale . as is seen in fig .
[ sph1 ] , the imprinting procedure did not destroy the sphericity of the imprinted surface .
an afm scan of the sphere surface with imprinted corrugations used in the second set of our measurements was also performed .
the amplitude of these corrugations was measured to be @xmath24 nm ( increase by a factor of 3.2 as compared with refs .
the respective variance of the stochastic roughness in this case was @xmath25 nm . note that the relatively large values of the variance of the stochastic roughness presented above ( in comparison with the case of flat surfaces @xcite ) are explained by the fact that the r.m.s .
deviation is calculated along the @xmath15-axis , i.e. , not perpendicular to the corrugated surface at most of the intermediate points . in fig .
[ sph2 ] we show a @xmath23 area of the corrugations imprinted on the sphere used in the second set of our measurements . as can be seen in this figure ,
the grooves are deeper than for the sphere used in the first set of measurements , and the sphericity of the surface is preserved .
the diameters of the spheres with the imprinted corrugations used in the first and second set of measurements were measured to be @xmath26 m and @xmath27 m , respectively , using a scanning electron microscope ( sem ) .
the sem was calibrated with a nist - traceable calibration grating which was independently cross calibrated with our interferometrically calibrated afm .
the distinctive feature of our setup is that the lateral casimir force acting along the @xmath28-axis tangential to the corrugated sphere and a grating leads to the vertical bending of the cantilever .
this bending is measured using bicell photodiodes a and b in fig .
[ setup ] . whereas a force acting normal to the test bodies ( the normal casimir force ) leads to a torsional deflection of the cantilever .
the torsional spring constant of the used cantilever @xmath29 was found to be 46 times larger than the bending spring constant @xmath30 ( see the next section ) . because of this the normal casimir force could lead to only a negligible change in the position of the second sphere and in the phase of the corrugations .
thus the setup used is much more sensitive to detecting the lateral casimir force , while simultaneously suppressing the effect of the normal casimir force .
the calibration of the cantilever ( i.e. , the determination of its spring constants ) , the measurement of the residual potential difference @xmath31 between the sphere and the plate , and the determination of the separation on contact @xmath32 were done by using both the normal and lateral electric forces .
we begin with the determination of @xmath29 and @xmath31 by means of the normal electric force . for this purpose
we have measured the torsional cantilever deflection due to the voltages @xmath33 applied to the grating while the sphere remained grounded .
this was performed through the measurement of the difference signal between the bicell photodiodes shown in fig .
[ setup ] . to determine the parameters of our interest
, the measured deflection signal @xmath34 at some separation @xmath35 between the zero levels of corrugations on the grating and on the sphere , where the casimir force is negligible , should be fitted to the theoretical expression for the normal electrostatic force between the corrugated surfaces . to attain these ends we begin with a brief derivation of such an expression in the sphere - plate configuration which is not available in the literature . in the configuration of an ideal metal sphere above an ideal metal plate with smooth boundary surfaces the exact expression for the electric force is well known @xcite . in the range of separations from @xmath36 nm to @xmath37 m it can be represented @xcite by the following polynomial with a relative error less than @xmath38% @xmath39 where @xmath40 is the permittivity of free space and the numerical coefficients @xmath41 are given by @xmath42 the approximate expression for the electric force in the case when the surfaces of a sphere and a plate are covered with sinusoidal corrugations can be obtained using the pfa , as applied to electric forces . for this purpose
we consider the separation along the @xmath15-axis between any two points @xmath43 on the corrugated sphere and @xmath44 on the corrugated plate @xmath45 here @xmath35 , as defined above , is the closest separation between a perfectly shaped sphere and a plate , and @xmath46 is the phase shift between the corrugations of both bodies .
then , according to the pfa , the normal electric force between corrugated a sphere and a plate is obtained from eq .
( [ eq1 ] ) by averaging over a period as @xmath47 the integral in ( [ eq4 ] ) can be calculated by using the following substitution of ( [ eq3 ] ) : @xmath48 , \label{eq5}\ ] ] where @xmath49 substituting eq .
( [ eq5 ] ) into eq .
( [ eq4 ] ) and performing integration , one obtains @xmath50 . \label{eq7}\end{aligned}\ ] ] it is well known that for corrugated surfaces the use of the pfa may not lead to precise expressions for the casimir force @xcite . for a static electric force , however , the pfa should work much better because in this case the diffraction effects are absent . to verify the validity of eq .
( [ eq7 ] ) , we applied a 3-dimensional finite element analysis ( fea ) using comsol multiphysics @xcite to numerically solve the poisson equation with appropriate boundary conditions in the configuration of corrugated sphere and plate . in the fea computation , it is difficult to use the whole corrugated sphere and plate due to computer limitations .
this necessitates the use of truncated spheres and parts of the plates .
such a replacement is justified as the primary contribution to the force comes from sphere - plate regions which are in close proximity .
however , the error introduced by the truncation has to be independently confirmed to be negligible by varying the size of the regions of the sphere and plate used in the computations .
computations were performed as a function of the common size @xmath51 of a truncated section of the corrugated sphere and a square corrugated plate . the separation distance and phase difference between the corrugations on both surfaces
were set to @xmath52 nm and @xmath53 , respectively .
the system was enclosed in a grounded rectangular box which represents the boundary conditions at infinity .
next , boundary conditions ( applied voltages ) identical to the experimental parameters were assigned to all objects .
the rectangular enclosure and the sphere were both set to be grounded .
the rectangular box was automatically set to be at infinity by comsol multiphysics .
next the poisson equation was solved for the given boundary conditions . then the normal electrostatic force was calculated by integrating the @xmath54-component of the maxwell stress tensor over the surface of corrugated sphere and compared with eq .
( [ eq7 ] ) . to make sure that the solution converged , the size of the corrugated objects and the number of the surface mesh elements were both varied and the force recalculated . for each @xmath51 ,
the number of the mesh elements was increased from one million to 22 millions until the calculated force converged .
the size of the truncated section of the corrugated plate and sphere was increased from @xmath55 m to @xmath56 m .
the convergence of the force was observed for @xmath57 m . the corresponding difference of eq .
( [ eq7 ] ) from the fea results varied from 36% to 2.8% when @xmath51 increased from @xmath58 m to @xmath59 m , respectively , with convergence at the largest values of @xmath51 .
thus , we confirmed from simulations that eq .
( [ eq7 ] ) is valid with an error less than 2.8% .
it turns out that this error is almost phase independent and , thus , does not propagate to the lateral electric force .
now we return to the experimental electrostatic calibrations using the normal electric force .
the deflection signal @xmath34 was measured for eight different voltages between 0.52v and 0.47v applied to the grating at the constant grating - sphere separation @xmath60 m where the casimir force is negligibly small ( see below for the determination of absolute separations using the lateral electrostatic force ) .
the respective experimental normal electrostatic force is given by @xmath61 where @xmath62 is the normal force calibration constant measured in the units of force per unit deflection signal ( note that for the attractive force the deflection signal is negative ) .
this constant is connected with the torsional spring constant discussed above as @xmath63 where the deflection coefficient @xmath64 is measured in the units of length per unit deflection signal .
the obtained experimental data for @xmath34 and for the respective force @xmath65 from eq .
( [ eq8 ] ) was fitted to eq .
( [ eq7 ] ) .
the resulting mean values of @xmath31 and @xmath62 found from the fit are @xmath66 measurement of the lateral casimir force as a function of absolute separation requires knowledge of the lateral force calibration constant @xmath67 where @xmath68 is the bending deflection coefficient .
if the separation on contact @xmath32 is determined , absolute separation between the mean values of the corrugations on both surfaces is given by @xcite @xmath69 where @xmath70 is the distance moved by the plate owing to the voltage applied to the piezoelectric actuator , @xmath71 is the photodiode difference signal due to the force ( either electric or casimir ) .
the determination of @xmath72 and @xmath32 is achieved by measuring the cantilever deflection signal due to the lateral electrostatic force which arises when a voltage is applied to the grating .
the measurements of this signal were performed at small separations from close to @xmath32 to @xmath73 nm .
note that calibrations using the lateral electrostatic force were done after the measurements of the deflection signal due to the lateral casimir force are performed , but are reported in this section for the benefit of the reader .
first , a voltage of 141.456mv was applied to the grating .
the sphere was kept at a distance 3.96 nm from @xmath32 . the phase shift between corrugations
was changed continuously at a frequency 0.103hz with the @xmath28-piezo to a maximum translation of @xmath74 m .
it is important that corrugated surfaces move parallel to each other as the phase is changed with the @xmath28-piezo .
the grating template was mounted as parallel as possible to the @xmath28-axis on the afm piezo .
however , experimentally there is always a finite nonzero angle between the grating template and the @xmath28-axis .
when the phase change is introduced by moving the @xmath28-piezo , this tilt will lead to changes in the separation distance between the corrugated surfaces , which in turn would result in a systematic error of decreasing or increasing force with phase .
this nonparallelity should be corrected before start of the experiments using the lateral electrostatic force .
a voltage is applied to the plate and the lateral electrostatic force is monitored as a function of the phase over many periods of the grating .
a correcting voltage adapted from that applied to the @xmath28-piezo using an adjustable voltage divider was synchronously applied to the @xmath15-piezo .
the applied voltage to the @xmath15-piezo was changed till the amplitude of the lateral electrostatic force become independent of the phase . as an example , in fig .
[ elfc ] we show the deflection signal due to the lateral electrostatic force versus the phase difference @xmath75 between the corrugated surfaces ( a ) before and ( b ) after the tilt correction was introduced .
this was done approximately 25 nm away from contact with a 0.2v applied voltage .
it can be seen that the amplitudes of each peak are more identical in fig .
[ elfc](b ) .
this procedure experimentally verifies that the separation distance between the corrugations remains fixed as the phase was changed with the @xmath28-piezo .
the cantilever deflection signal @xmath76 corresponding to the total force acting at the separation @xmath77 nm ( sum of the lateral electrostatic and lateral casimir forces ) was recorded at 8192 evenly spaced data points .
the sphere was moved further away from the grating by 5.40 nm to a separation 9.36 nm from @xmath32 , and the measurement was repeated .
the cantilever deflection signal @xmath76 was measured at many other @xmath35 at the same voltage and also for a second voltage of 101.202mv applied to the grating . to obtain the deflection signal due to the lateral electrostatic force alone ,
it is necessary to subtract from the obtained signal @xmath76 the deflection signal @xmath78 due to the lateral casimir force ( the signal @xmath78 was measured before @xmath76 and the respective measurement data are reported in sec .
iv ) . to perform this subtraction , @xmath78 was fitted to the following sum of harmonics : @xmath79 which takes proper account of the fact that the lateral casimir force is asymmetric ( see sec .
iv ) . as the sphere - grating separations in the measurements of the lateral casimir force and of the total force
are not identical , interpolation was used to determine the values of @xmath80 at the separations corresponding to the total lateral force . after determination of the coefficients @xmath80 from the fit , the signal @xmath78 was subtracted from the data for the total deflection signal @xmath76 .
the obtained deflection signal @xmath81 now corresponds to the lateral electrostatic force @xmath82 the experimental force data in this equation were used to determine the calibration parameter @xmath72 and @xmath32 from the fit to the theoretical expression for the lateral electric force . to derive the analytic expression for the lateral electric force acting between the sinusoidally corrugated surfaces of a sphere and a plate , we start with the electrostatic energy in sphere - plane configuration with smooth surfaces @xcite @xmath83 . \label{eq15}\ ] ] here
, the integration constant @xmath84 is equal to @xcite @xmath85 where @xmath86 , and @xmath41 are defined in eq .
( [ eq2 ] ) . using the pfa
, we obtain the electrostatic energy between sinusoidally corrugated surfaces of a sphere and a plate in the form @xmath87 where @xmath88 is given by eqs .
( [ eq3 ] ) , ( [ eq5 ] ) and ( [ eq6 ] ) . using the definition of the lateral electric force [ see the formula for the phase shift below eq .
( [ eq3 ] ) ] @xmath89 and calculating all integrals in ( [ eq17 ] ) , we arrive at the result @xmath90 . \label{eq19}\end{aligned}\ ] ] here , the parameter @xmath91 is defined in eq .
( [ eq6 ] ) and the coefficients @xmath92 are given by @xmath93 by fitting the force data in eq .
( [ eq14 ] ) to eq .
( [ eq19 ] ) , the quantities @xmath72 and @xmath32 were found .
this was repeated for four different electrostatic force measurements and the average values obtained are @xmath94 using an independent quad - cell afm measurement of the torsional movement of such a cantilever , the torsional spring constant was found to be @xmath95n / m .
the value of the bending deflection coefficient @xmath96nm / unit @xmath97 was obtained in the same way as in normal casimir force measurements @xcite from the change in position of the contact point for application of different voltages to the plate during the normal electrostatic force measurement .
then the value of the bending spring constant @xmath98n / m was found from eqs .
( [ eq11 ] ) and ( [ eq21 ] ) . by comparing the two spring constants , we find that @xmath99 which means that our system is really much more sensitive to the lateral rather than to the normal casimir force . in recent literature @xcite
there is a discussion that in several measurements of the casimir force the residual potential difference @xmath31 depends on separation , whereas in some other it is constant . because of this , we have performed an additional control measurement of @xmath31 at short separation @xmath100 nm using the parabolic dependence of @xmath101 in eq .
( [ eq19 ] ) on @xmath33 . for this purpose
five different voltages close to the residual potential difference were applied to the grating and the cantilever deflection was measured as a function of v leading to @xmath102mv .
this is consistent with the result ( [ eq10 ] ) obtained at large separation @xmath60 m and confirms that in our experiment the residual potential difference is separation independent .
the same calibration procedures , as described above for the first set of measurements , were repeated for a second set with a larger corrugation amplitude of 25.5 nm on the sphere .
here we obtained a residual potential difference @xmath103mv and the following values of the separation on contact and of the lateral force calibration constant @xmath104 typically in our second set of measurements with the larger amplitude corrugations the calibration errors are a bit larger than in the first , where smaller amplitude corrugations were used .
now we describe how the measurements of the lateral casimir force were performed . in this case the residual voltage @xmath31 , as determined in the electrostatic calibration using the normal electric force , was applied to the grating in order to make the electric force equal to zero .
the @xmath28-piezo was used to move the grating along the @xmath28-axis and thus change @xmath105 .
the @xmath15-piezo , which was independently controlled by an external voltage source , was used to change @xmath35 .
the piezo extensions with applied voltage in both directions were calibrated using optical interferometry @xcite . initially the corrugated sphere was positioned 3.79 nm from the separation on contact between the two surfaces @xmath32 determined by the corrugations and the highest roughness peaks .
the thermomechanical drift of the separation distance was measured to be 0.14nm / min from the difference in the @xmath15-piezo voltage to bring about contact of the two corrugated surfaces after a time interval around 30 minutes .
a phase shift was introduced by moving the @xmath28-piezo continuously for a total distance of @xmath74 m at 0.103hz .
the photodiode signal corresponding to the cantilever deflection was filtered with a low - pass filter with a 30ms time constant and recorded at each of the 8192 points corresponding to @xmath28-changes of 0.4 nm .
the effect of the scattered laser light which would lead to a linear modification of the signal with phase was found to be negligible in this experiment .
then the separation from @xmath32 was increased by 3.6 nm from 3.79 nm to 7.39 nm and the deflection signal @xmath78 was similarly measured as a function of @xmath105 and recorded . after this , the separation from @xmath32 was increased by 3.96 nm from 7.39 nm to 11.35 nm and the measurements repeated .
next @xmath78 due to the lateral casimir force as a function of @xmath105 was measured at separations of 20.05 , 32.48 , 45.30 , 58.01 and 70.86 nm from @xmath32 . using the values of @xmath72 and @xmath32 obtained through the electrostatic calibration ( see sec .
iii ) the measurement data for the deflection signal @xmath78 were converted into values of the lateral casimir force at every separation @xmath35 as : @xmath106 the resulting lateral casimir force for a sphere with the corrugation amplitude @xmath107 nm over four corrugation periods is shown in fig .
[ phasen](a - c ) as dots versus the normalized lateral displacement @xmath108 between the corrugated surfaces at separations @xmath109 , 128.6 , and 149.8 nm , respectively .
similar results were obtained at all other @xmath35 listed above . as can be seen in fig .
[ phasen ] , the lateral casimir force is a periodic function of the phase shift .
the maximum values of the lateral force , @xmath110 , decrease with the increase of @xmath35 .
the important characteristic feature of the periodic curves in fig .
[ phasen](a - c ) is that they are _ asymmetric _ , i.e. , the dependence of @xmath111 on @xmath108 is not strictly sinusoidal ( note that the sinusoidal dependence of the lateral force on the phase shift holds only if the calculation is restricted to the lowest order in corrugation amplitudes @xmath112 ; see sec .
v ) , the true asymmetry of the measured lateral force is obvious even without the theory curve . for example , in fig .
[ phasen](a ) the average shift of each maximum point from the midpoint of two adjacent minima is @xmath113 . with the increase of separation ,
the relative contribution of higher perturbation orders in a small parameter @xmath114 decreases and the dependence of the lateral force on the phase shift becomes more close to sinusoidal .
the solid lines in fig .
[ phasen ] are related to the exact theoretical computations .
they are discussed in sec .
the experimental values of the @xmath115 versus separation @xmath35 are shown in fig .
[ ampn ] as crosses .
the arms of the crosses indicate the total experimental errors determined at a 95% confidence level using the following procedure . from the data of the phase curves in fig .
[ phasen ] ( and related curves at all the other separations measured ) we have calculated the mean values @xmath116 and the variances @xmath117 of the quantity @xmath118 at each @xmath35 ( see columns 2 and 3 in table 1 ) .
for the first set of measurements under consideration , the averaging was performed over the four periods .
the respective student coefficient is @xmath119 where @xmath120 .
thus , the experimental random error is obtained from the data in column 3 of table 1 as @xmath121 the main sources of systematic errors in the quantity @xmath122 are the errors due to the uncertainty in @xmath72 and due to the resolution of data . according to eq .
( [ eq21 ] ) , @xmath123 . from eq .
( [ eq22 ] ) this leads to the systematic error @xmath124 .
the systematic error resulting from the resolution of data does not depend on separation @xmath125pn . combining these two errors at a 95% confidence level as random quantities characterized by a uniform distribution @xcite
, we obtain the values of the total systematic error given in column 4 of table 1 . using the rule for the combination of random and systematic errors at the same confidence level @xcite
, we obtain the total experimental error @xmath126 as a function of separation presented in the fifth column of table 1 . as can be seen in table 1 ,
all absolute errors are larger at shorter separations and decrease with increasing separation .
this might be explained by local deviations of the shape of corrugations from perfect sinusoidal form which are more influential at shorter separations .
the total relative error of @xmath115 at @xmath127 nm is equal to 22.5% ( at a 95% confidence level ) .
the relative errors at separations @xmath109 , 128.6 , and 137.3 nm are equal to 13% , 25.5% , and 14% , respectively .
the error in the measurement of the absolute separations , @xmath128 , is a combination of the error in the determination of @xmath32 , @xmath129 nm , indicated in eq .
( [ eq21 ] ) and the error in the quantity @xmath130 in eq .
( [ eq12 ] ) .
the latter is equal to the half of the first step of the @xmath15-piezo , @xmath131 nm . combining these two errors at a 95% confidence level with the help of the same rule , as applied above for the systematic errors @xcite , one obtains @xmath132 nm .
this is a factor of eight improvement as compared with ref .
note that the bands between the dashed and the solid lines in fig .
[ ampn ] are related to theoretical predictions using the pfa and the exact theory , respectively .
they are discussed in secs .
v and vi .
the same measurement procedure , as described above , was applied in the second set of measurements for a sphere covered with deeper corrugations of an amplitude @xmath133 .
the resulting lateral casimir force over five or six corrugation periods is shown in fig .
[ phasem](a - c ) as dots versus the lateral displacement at separations @xmath134 , 156.5 and 179 nm , respectively .
similar results were obtained also at separations @xmath135 , 201.6 , 224 , and 246.6 nm . in the same way , as in fig .
[ phasen ] , the lateral casimir force at each @xmath35 is a periodic function which is asymmetric ( nonsinusoidal ) . the asymmetry is more pronounced at shorter separations .
the experimental values of the @xmath115 versus separation are plotted in fig .
[ ampm ] as crosses .
similar to fig .
[ ampn ] the arms of the crosses show the total experimental errors of force and separation measurements determined at a 95% confidence level .
equation ( [ eq21a ] ) and the same statistical procedure , as in the first set of measurements , were used to determine these errors with the only difference that for the second set the averaging was performed at all separations over five periods [ the respective student coefficient is @xmath136 with @xmath120 ] .
the total error in the measurement of separation , @xmath138 nm , turned out to be a bit larger than in the first set of measurements .
the total error of the lateral force measurements demonstrates similar irregular behavior on separation distance .
this can also be explained by the influence of local deviations of groove shape from sinusoidal form keeping in mind that with larger corrugation amplitude the role of such deviations should be larger . in table 2
we present the mean values , the variances of the mean , the systematic errors , and the total experimental errors ( at a 95% confidence level ) of the maximum magnitudes of the lateral casimir force at different separations for the second set of measurements .
the total relative error of the lateral force measurements at separations @xmath134 , 145.2 , 156.5 , and 179 nm varies as 23% , 16% , 14% , and 13% , respectively , all calculated at a 95% confidence level .
in this section and in sec . vi we compare the obtained experimental results for the lateral casimir force with two theoretical approaches applicable in the case of corrugated surfaces : the pfa and the exact scattering approach , respectively , with no fitting parameters . as was mentioned in the introduction , the pfa approach was also used to compare with theory the measurement data of the first observation of the lateral casimir force in refs .
however , in those papers the real properties of au were described using the simple plasma model and fourth - order perturbation theory with respect to the relative skin depth at zero temperature . here
, we develop a more complete description of the experimental data in the framework of the pfa approach based on the lifshitz formula at the laboratory temperature .
this will help us to separate diffraction type contributions to the lateral casimir force which are beyond the pfa .
we start with the lifshitz formula for the normal casimir force acting between a sphere and a plate made of a real material ( au ) but bounded with perfectly smooth surfaces @xcite @xmath139\right .
\nonumber \\ & + & \left .
\ln\left[1-r_{\rm te}^2(i\xi_l , k_{\bot})\,e^{-2q_la}\right]\right\}. \label{eq24}\end{aligned}\ ] ] here , @xmath140 is the boltzmann constant , @xmath141 with @xmath142 are the matsubara frequencies , prime near the summation sign adds a multiple 1/2 to the term with @xmath143 .
the reflection coefficients for the transverse magnetic and transverse electric polarizations of the electromagnetic field are defined as @xmath144 where @xmath145 and @xmath146 is the dielectric permittivity for the material of the sphere and plate calculated along the imaginary matsubara frequencies .
equation ( [ eq24 ] ) is obtained from the standard lifshitz expression for the free energy in the configuration of two parallel plates by the multiplication by @xmath147 in accordance with the pfa . keeping in mind the values of the experimental parameters ( @xmath148 m , @xmath149 nm )
, one can conclude that the error introduced by the use of the pfa in this case ( of about @xmath150% @xcite ) is negligibly small .
the general expression for the lateral casimir force can be obtained from eq .
( [ eq31 ] ) in the same way as the lateral electric force in eq .
( [ eq18 ] ) . after performing the differentiation ,
we rearrange the result to the form @xmath157\,e^{-ny}\,{i}_1(n\beta y ) .
\label{eq32}\end{aligned}\ ] ] this expression is convenient for numerical computations . equations ( [ eq30 ] ) and ( [ eq32 ] ) are generalizations for the case of real materials of the previously known expressions obtained at @xmath158 for ideal metal corrugated bodies . in this case @xmath159 and one obtains @xmath160 introducing the new variable @xmath161 and performing the summation , we arrive at @xmath162 after performing the integration in these equations , the final result is @xmath163 the last expression for the lateral force was obtained in refs .
@xcite with corrections due to the skin depth .
in the first perturbation order in @xmath164 ( i.e. , for @xmath165 ) this expression was also used in refs . @xcite . equations ( [ eq30 ] ) and ( [ eq32 ] ) take into account the sphericity of the second test body and sinusoidal corrugations on both bodies in the framework of the pfa and the nonzero skin depth in the framework of the lifshitz theory . to perform numerical computations
, one should use a model to represent the dielectric permittivity . at the separations considered ( from 100 to 250 nm )
the thermal effects do not play any role in any of the theoretical approaches and reliable results can be obtained using the generalized plasma - like model @xcite @xmath166 where @xmath167ev for au is the plasma frequency .
the values of the oscillator frequencies @xmath168 , relaxation parameters @xmath169 and the oscillator strengths @xmath170 leading to approximately the same results for @xmath171 as those obtained from the optical tabulated data @xcite over the frequency range from 2 to 25ev are listed in refs .
@xcite .
an important factor that is not taken into account in eq .
( [ eq32 ] ) is the surface roughness .
stochastic roughness covering the corrugated surfaces in both the sphere and the plate is clearly seen in figs .
[ rghpl],[rghsph ] and its measured variances are presented in sec .
ii . in view of its stochastic ( nonperiodic ) character ,
surface roughness can not influence the phase dependence of the casimir energy in the configuration of corrugated surfaces .
however , it contributes to the casimir energy , and , thus , to the lateral casimir force , through a phase independent correction factor which depends only on separation .
it has been found @xcite that the correction factor to the casimir force due to stochastic roughness in the configuration of a sphere above a plate with no corrugations is given by @xmath172 now we take into account that in the framework of the multiplicative approach the correction factor due to surface roughness is the same for real and ideal metals .
keeping in mind that the casimir force between a perfectly shaped ideal metal sphere and a plate is @xmath173 , we obtain the correction to the casimir energy by integrating the casimir force with the inclusion of roughness , @xmath174 , with respect to @xmath35 @xmath175 note that for the experimental variances ( see sec .
ii ) the fourth order term in eq . ( [ eq38 ] ) is negligibly small .
thus , for the first set of measurements at the shortest separation @xmath176 nm the second order term contributes 3% of the force , but the fourth order term only 0.2% of the force . for the second set of measurement with deeper corrugations on the sphere the contributions of the second and fourth order terms at separation @xmath177
nm are 4% and 0.3% , respectively . to obtain the correction factor to the lateral casimir force due to the surface roughness with account of the sinusoidal corrugations on the sphere and on the plate ,
we should replace @xmath35 with separation @xmath88 between the corrugated surfaces in eqs .
( [ eq3 ] ) and ( [ eq5 ] ) and perform the averaging with respect to the phase shift and the period @xmath178 ^ 2}. \nonumber\end{aligned}\ ] ] calculating both integrals in the second equality in eq .
( [ eq39 ] ) and substituting the obtained result into the first equality , we arrive at @xmath179 where @xmath180 is the hypergeometric function . substituting the values of @xmath181 and @xmath182 from the first set of measurements
( see sec .
ii ) in eq . ( [ eq40 ] ) , we find that at separations @xmath183 , 150 , and 200 nm the correction factor due to surface roughness takes the values @xmath184 , 1.029 , and 1.015 , respectively . for the second set of measurements ( i.e. , using @xmath133 and @xmath185 in sec .
ii ) the correction factor in eq . ( [ eq40 ] ) takes the values @xmath186 , 1.041 , and 1.021 at separations @xmath187 , 150 , and 200 nm , respectively .
it is seen that even for the second set of our measurements , where the surface roughness on the sphere is larger than in the first one , the roughness correction is rather moderate and does not exceed 6% of the lateral casimir force . in fig .
[ ampn ] ( the lower dashed line ) we present the computational results for the maximum values of the lateral casimir force as a function of separation in the framework of the pfa [ i.e. , using eqs .
( [ eq32 ] ) and ( [ eq36 ] ) ] for the first set of our measurements .
this line does not take surface roughness into account .
the upper dashed line in fig .
[ ampn ] presents the results for @xmath188 as a function of separation , i.e. , taking into account the correction due to surface roughness . keeping in mind that the multiplicative approach provides the means to estimate the effect of roughness rather than to calculate it precisely on the basis of the fundamental theory
, we consider the band between the dashed lines in fig .
[ ampn ] as a theoretical prediction given by the pfa approach . as can be seen in fig .
[ ampn ] , the experimental data indicated as crosses are inconsistent with the prediction of the pfa approach , thus revealing the role of diffraction - type effects on the corrugations of relatively small period used in this experiment . in fig .
[ ampm ] a similar comparison between the experimental data and theory using the pfa approach is done for the second set of our measurements with corrugations of larger amplitude .
the lower dashed line is computed using eqs .
( [ eq32 ] ) and ( [ eq36 ] ) with the surface roughness disregarded .
the upper dashed line is obtained by the multiplication of the computation results by an additional factor @xmath189 computed using eq .
( [ eq40 ] ) .
again the theoretical band between the dashed lines is inconsistent with the experimental data shown as crosses due to the neglect of diffraction - like effects . in the next section we present a more fundamental theoretical approach taking the diffraction effects into account which brings theory into agreement with the experimental data
the pfa approach used to calculate the lateral casimir force in the configuration of a sphere above a plate with corrugated boundaries has an uncontrolled error which increases with decreasing period of corrugations . as was mentioned in the introduction , in such cases
a more fundamental theory is desirable which takes exact account of diffraction effects . here
, we present such a theory , based on the scattering approach , for the configuration of two plates covered with periodical corrugations of arbitrary shape but common period . this theory can be applied to our experimental configuration by using the pfa for the transition from the configuration of two plane plates to a smooth sphere above a plate . according to our discussion in sec .
v , the error in computations introduced by the use of the pfa for this restricted purpose alone , is extremely small and quite satisfactory for the needs of our experiment .
the scattering approach started with the work @xcite where it was applied to obtain the lifshitz formula and refs .
@xcite on the multi - scattering expansion . on this basis
the multipole scattering technique was developed in refs .
@xcite and was applied to several configurations with curved boundaries for scalar and electromagnetic cases .
alternative techniques within the scattering approach to the casimir effect were presented in refs .
@xcite . for objects of
spherical and cylindrical shapes the multipole scattering technique was used in refs .
this technique works well for large separations between the objects and shows poor convergence when the separations between the objects are small . at small separations the leading asymptotic expansion of the casimir energy at zero temperature within the scattering approach
was developed in ref .
the technique developed in refs .
@xcite uses another basis , the rayleigh basis , which is a generalization of the plane wave basis .
the rayleigh basis is a natural choice for grating geometries .
we consider two parallel @xmath190-dimensional longitudinal ( along the @xmath191-axis ) periodic ( along the @xmath28-axis ) dielectric ( metallic ) gratings of arbitrary form separated by a vacuum gap so that they form a waveguide and one grating is located above the other ( we assume that the edge of the corrugation region of each grating is perpendicular to the @xmath15 axis ) . the periods of both gratings @xmath13 are equal . in general , the top and bottom gratings can have different shapes and different dielectric ( metallic ) properties . in fig .
[ grat1 ] the height of the corrugation region is equal to @xmath192 for the bottom grating .
the meaning of the lateral displacement @xmath75 becomes obvious from the comparison of fig .
[ grat1 ] with fig .
[ grat2 ] and fig .
[ grat3 ] .
we suppose that the space between the two gratings is vacuum with @xmath193 and we assume @xmath194 inside the medium .
the physical problem is time and @xmath191 invariant , so that the particular solutions for the electric and magnetic fields can be written in the form : @xmath195 the solutions of maxwell equations should satisfy the quasi - periodicity conditions : @xmath196 in every casimir problem one needs to determine the complete basis of solutions .
let us suppose that the top grating is absent .
we consider a generalized conical diffraction problem @xmath197 on the lower grating for the incident wave with the @xmath28-component of the wave vector equal to some fixed value which we denote @xmath198 .
the longitudinal components of the electromagnetic field outside the corrugated region ( @xmath199 ) may be written by making use of the rayleigh expansion @xcite for an incident monochromatic wave : @xmath200 here @xmath201 with @xmath202 \equiv\tilde{k}_x-\frac{2\pi}{\lambda}p , \label{eq44b}\ ] ] where @xmath203 $ ] is the integer part of the number @xmath204 . from this
it follows that @xmath205 .
the other notation in eq .
( [ eq43 ] ) is as follows : @xmath206 the quantities @xmath207 and @xmath208 in eq .
( [ eq43 ] ) are the rayleigh coefficients .
[ note that with notations ( [ eq44a])([eq44c ] ) it holds @xmath209 , @xmath210 . ]
the solution ( [ eq43 ] ) is also valid outside any periodic structure in @xmath28 direction .
all other field components can be expressed in terms of the longitudinal components @xmath211 by using standard formulas as done in waveguide theory .
this can be done since the factor @xmath212 is conserved after the reflection of the electromagnetic wave from the grating . at @xmath213
the solution has to satisfy the expansions @xmath214 where @xmath215 which are valid for @xmath216 .
the coefficients @xmath217 are called the transmission matrix coefficients .
the @xmath218 component of the electromagnetic field in the region @xmath219 is defined as follows : @xmath220 the other components of the electromagnetic field in the region @xmath221 are defined in analogy . inside the corrugation region @xmath222 it is convenient to rewrite maxwell equations in the form of the first order differential equations , @xmath223 , where @xmath224 is a square matrix of dimension @xmath225 , @xmath226 and @xmath227 is the number of coefficients considered in the expansion ( [ eq49 ] ) for @xmath218 and similar expansions for @xmath228 and @xmath229 . for a rectangular grating
the matrix @xmath224 is constant ( independent of @xmath15 ) . from maxwell equations @xmath230
we get @xmath231 substituting into eq .
( [ eq48 ] ) the expansion ( [ eq49 ] ) and analogous expansions for all other components of @xmath232 and @xmath233 , we arrive at @xmath234 where @xmath235 is the toeplitz matrix .
the following equations follow from the remaining four maxwell equations : @xmath236 one can substitute eq .
( [ eq50 ] ) into eq .
( [ eq51 ] ) and obtain a system of the first order differential equations for the fourier components of the electromagnetic field @xmath237 in the region @xmath238 .
now we have to determine the rayleigh coefficients @xmath239 for the specific periodic geometry profile .
one can determine these coefficients by matching the solution of equations inside the corrugation region @xmath238 with rayleigh expansions ( [ eq43 ] ) at @xmath240 and expansions ( [ eq45 ] ) at @xmath213 .
this can be done by imposing the continuity conditions on each fourier component of the fields @xmath241 at @xmath213 and @xmath240 .
there is no separation of the te and tm modes for the general case .
that is why the reflection matrix @xmath242 for the reflection from the bottom grating can be defined as follows : @xmath243 to obtain the casimir energy we need to determine the eigenfrequencies of all the normal modes of the electromagnetic field between the two periodic gratings .
these eigenfrequencies can be summed up by making use of an argument principle , which states : @xmath244 here , @xmath245 are zeroes , @xmath246 are poles of the function @xmath247 inside the contour of integration , and the degenerate eigenvalues are summed over according to their multiplicities . for the casimir energy we have @xmath248 .
the equation for eigenfrequencies of the corresponding problem of classical electrodynamics is @xmath249 .
consider first the plate - plate geometry when the two dielectric parallel slabs ( slab @xmath250 : @xmath251 , slab @xmath252 : @xmath253 ) are separated by a vacuum gap ( @xmath254 ) . in this case
the te and tm modes are not coupled .
the equation for the te eigenfrequencies is : @xmath255 here @xmath256 is the reflection coefficient of a plane wave moving down which reflects on the dielectric surface of the slab @xmath250 at @xmath213 , while @xmath257 is the reflection coefficient of an upward moving plane wave which reflects on the dielectric surface of the slab @xmath252 at @xmath258 , and @xmath259 .
one can deduce from the maxwell equations that @xmath260 [ here , @xmath261 is the reflection coefficient of a downward moving te plane wave which reflects from the dielectric slab @xmath252 temporarily located at the position of the slab @xmath250 , i.e. at @xmath251 ] . from ( [ eq54 ] ) and the analogous equation for the tm modes one immediately obtains the lifshitz formula by making use of the argument principle ( [ eq53 ] ) . for two periodic dielectrics or gratings separated by a vacuum gap
one has to consider the reflection of downward and upward moving waves from the unit cell @xmath263 .
imagine that we removed the top grating in the system .
then the reflection matrix of the downward moving wave is defined as @xmath242 .
imagine now that we remove the bottom grating in the system .
then we denote the reflection matrix of the upward moving wave as @xmath264 .
the reflection matrices @xmath265 depend on the wave vectors of the incident waves , parameters of the gratings and the mutual location of the gratings .
the equation for normal modes states : @xmath266 where @xmath267 is an eigenvector describing the normal mode with the frequency @xmath268 . instead of
( [ eq54 ] ) one obtains : @xmath269 = 0 .\label{eq56}\ ] ] for every @xmath270 the solution of ( [ eq56 ] ) yields possible eigenfrequencies @xmath268 for the solutions of maxwell equations that should be substituted into the definition of the casimir energy @xmath271 .
suppose that the reflection matrix @xmath272 for the reflection from the fictitious ( imaginary ) grating located as in fig .
[ grat2 ] is known in the coordinates @xmath273 . performing a change of coordinates @xmath274 , @xmath275 in ( [ eq43 ] ) , it is possible to obtain the matrix @xmath264 for the reflection of upward waves from a grating with the same profile turned upside - down and displaced from the lower grating by @xmath276 ( see fig .
[ grat3 ] ) .
it follows that @xmath277 where @xmath278 is the reflection matrix of the downward moving waves from the grating in the system of coordinates @xmath273 depicted in fig .
[ grat2 ] . here
, @xmath279 is the diagonal @xmath280 matrix of the form : @xmath281 with the matrix elements @xmath282 $ ] @xmath283 on the main diagonal of the matrix @xmath284 .
the lateral translation @xmath280 diagonal matrix @xmath285 is defined as follows : @xmath286 with matrix elements @xmath287 on the main diagonal of the matrix @xmath288 , where @xmath289 .
the summation over the eigenfrequencies is performed by making use of the argument principle ( [ eq53 ] ) , which yields the casimir energy of two parallel gratings on a unit cell of a period @xmath13 and unit length in the @xmath191 direction : @xmath290 . \label{eq60}\ ] ] here , as was defined in sec . ii , @xmath291 and @xmath75 is a lateral displacement of the two gratings .
this is an exact expression valid at zero temperature for two arbitrary parallel gratings with equal periods @xmath13 separated by a vacuum gap .
the casimir free energy on a unit surface @xmath292 for the system of two gratings can be written as follows : @xmath293 .
\label{eq61}\ ] ] this formula is valid for an arbitrary profile and an arbitrary dielectric permittivity of each grating .
when one grating has a curvature of the sphere of the radius @xmath294 , the lateral casimir force in this system @xmath111 can be obtained by combining the use of the pfa for a sphere - plate configuration and the exact formula for the free energy of the two gratings ( [ eq61 ] ) : @xmath295 for the sinusoidally corrugated sphere and plate used in our experiment @xmath296 and the lateral casimir force is given by @xmath297 .
\label{eq63}\end{aligned}\ ] ] different methods were developed in grating theory starting with the pioneering work of rayleigh @xcite . for our case of sinusoidal gratings
we used the so - called differential method @xcite .
the integration of the first - order ordinary differential equations ( [ eq51 ] ) with the dielectric permittivity of the generalized plasma - like model ( [ eq36 ] ) in the corrugation region of each grating was based on an explicit runge - kutta @xmath298 formula , the dormand - prince pair , and was performed using the matlab package .
then the reflection matrices @xmath242 and @xmath264 were computed .
the lateral casimir force @xmath111 was computed using eq .
( [ eq63 ] ) for the experimental parameters of the first and second set of measurement .
every matsubara term in eq .
( [ eq63 ] ) was evaluated with a precision of 1.5% which is also the computational precision of the obtained @xmath111 . in fig .
[ phasen](a - c ) the exact computational results for the lateral casimir force as a function of the phase shift between the sinusoidally corrugations on both surfaces are shown by the solid lines at separations @xmath109 , 128.6 , and 149.8 nm , respectively ( the the first set of measurements with corrugation amplitude on the sphere of 13.7 nm ) .
no fitting parameters are used in the comparison as can be seen in fig .
[ phasen](a - c ) , the solid lines are in a very good agreement with the experimental data shown as dots .
these lines clearly demonstrate deviations from the sinusoidal behavior which decreases with the increase of separation .
thus , both the experimental data and the exact theory confirm the prediction of ref .
@xcite made using the pfa approach that the lateral casimir force is asymmetric . in fig .
[ ampn ] the exact computational results for the @xmath115 are plotted as a band between the solid lines versus the separation @xmath35 between the mean levels of corrugations .
the width of the band takes into account the computational errors equal to 1.5% and the correction to eq .
( [ eq63 ] ) due to surface roughness .
the lower solid line is obtained as the computed @xmath115 minus @xmath299 .
the upper solid line represents @xmath300 , where the correction factor @xmath301 for the first set of measurement was found in sec .
v. as can be seen in fig .
[ ampn ] , the exact theory is in a very good agreement with the measurement data although it deviates significantly from the theoretical band between the two dashed lines computed using the pfa approach .
again no fitting parameters are used in the theory .
this allows one to conclude that with the grating period of @xmath6 nm the diffraction - type effects on the lateral casimir force have been reliably demonstrated both experimentally and theoretically .
the results of similar computations using eq .
( [ eq63 ] ) are presented by the solid lines in fig .
[ phasem](a - c ) where the lateral casimir force is plotted as a function of the phase shift for the second set of our mesurements at @xmath134 , 156.5 , and 179 nm , respectively . as before no fitting parameters are used . in spite of the fact that the experimental data of this set of measurements with deeper corrugations on the sphere are more noisy , the exact theory is clearly consistent with data and confirms the asymmetric ( nonsinusoidal ) character of the lateral casimir force . in fig .
[ ampm ] the exact computational results for the @xmath115 in the second set of our measurements versus separation are presented as the band between the two solid lines .
the width of this band takes into account both the computational errors and the surface roughness .
it was found using the same procedure , as for the first set of measurements , with the replacement of @xmath301 by @xmath189 ( see sec .
v ) . as in previous comparisons ,
no fitting parameters are used .
it can be seen in fig .
[ ampm ] that in spite of the larger noise in the second set of measurements , the data are consistent with the exact theory .
the deviations of both the exact theory and the measurement data from the theoretical prediction of the pfa approach are even more pronounced than in the first set of measurements . thus , the second set of our measurements confirms the observation of the diffraction effects in the lateral casimir force .
in the foregoing we have presented the results of an experimental and theoretical investigation of the lateral casimir force which arises between au - coated aligned sinusoidally corrugated surfaces of a sphere and a plate spaced in close proximity to each other .
the most distinctive feature of this experiment in comparison with the one performed previously @xcite is the use of the grating with much smaller corrugation period which permitted to enter the region of parameters outside the applicability of the pfa and observe the nontrivial diffraction - type effects .
another distinctive feature of this experiment is the use of much deeper corrugations on both the plate and the sphere .
this allowed us to observe an asymmetry of the lateral casimir force ( i.e. , the deviation of its profile from perfect sine ) which was predicted @xcite in the framework of the pfa approach , but not observed up to date due to the use of small corrugation amplitudes . in this paper
we have presented details of the experimental setup , the original procedure allowing to imprint relatively deep corrugations using a metallized template grating on a sphere of about @xmath302 m radius , and the process of electrostatic calibrations giving the possibility to measure small forces and at short absolute separations with high precision .
the dependence of the lateral casimir force on the phase shift between the corrugations on the sphere and the plate was measured at many different separation distances in the two sets of measurements with different amplitudes of corrugations on the sphere .
the maximum value of the lateral force as a function of separation was also investigated in the two sets of measurements .
both the random and systematic errors of the measured forces and separations were found and the total experimental errors were calculated at a 95% confidence level . the experimental data were compared with an approximate theory using the pfa and with the exact theory based on the scattering approach .
note that exact calculation methods allowing the evaluation of the casimir force between bodies of arbitrary shape and made of any real material have been actively investigated since 2006 @xcite .
however , until the present paper and the rapid communication @xcite there was no detailed comparison of exact theoretical results beyond the lifshitz theory taking into account such experimental conditions as real material properties of the test bodies , surface roughness and nonzero temperature with measurement data of any specific experiment . using the exact theory , the computed dependences of the lateral casimir force on the phase shift and of the maximum magnitude of the lateral force as a function of separation were found to be in agreement with the experimental data . regarding the pfa approach ,
the computational results were found to be excluded by the data .
this provides the quantitative confirmation for the fact that this experiment was performed outside the applicability region of the pfa and marks the beginning of investigations of diffraction effects in the phenomenon of the lateral casimir force .
as was already noted in the introduction , the lateral casimir force might find applications in nanotechnology .
specifically , it was proposed to use this effect for the frictionless transmission of lateral motion by means of a nanoscale rack and pinion without intermeshing cogs @xcite or in a ratchet with asymmetric corrugations driven by the casimir force @xcite .
the above experimental and theoretical results bring us closer to the realization of such kinds of micromachines . in future
it is planned to perform similar experiments with corrugations of more complicated shape which provides an opportunity to modulate the respective lateral casimir force .
this work was supported by the nsf grant no .
phy0653657 ( measurement of the lateral casimir force ) and doe grant no .
de - fg02 - 04er46131 ( calculation of the electric and casimir forces ) .
v.n.m . and u.m .
were also supported by the darpa grant n66001 - 09 - 1 - 2069 .
g.l.k . and v.m.m .
were partially supported by the dfg grant ge 696/10 - 1 .
was partially supported by the grants rnp 2.1.1/1575 and rfbr 07 - 01 - 00692-a .
the authors are grateful to a. v. boukhanovsky , i. s. zudina and to the institute of computer technologies of st.petersburg state university of information technologies , mechanics and optics where the numerical computations were performed .
99 h. b. g. casimir , proc .
akad . wet . *
51 * , 793 ( 1948 ) .
m. bordag , g. l. klimchitskaya , u. mohideen , and v. m. mostepanenko , _ advances in the casimir effect _ ( oxford university press , oxford , 2009 ) .
w. milonni , _ the quantum vacuum _ ( academic press , san diego , 1994 ) .
v. m. mostepanenko and n. n. trunov , _ the casimir effect and its applications _ ( clarendon , oxford , 1997 ) .
k. a. milton , _ the casimir effect _ ( world scientific , singapore , 2001 )
. g. plunien , b. mller , and w. greiner , phys . rep . * 134 * , 87 ( 1986 ) .
m. kardar and r. golestanian , rev .
phys . * 71 * , 1233 ( 1999 ) .
m. bordag , u. mohideen , and v. m. mostepanenko , phys . rep . * 353 * , 1 ( 2001 ) .
k. a. milton , j. phys .
* 37 * , r209 ( 2004 ) . s. k. lamoreaux , rep .
68 * , 201 ( 2005 ) .
g. l. klimchitskaya , u. mohideen , and v. m. mostepanenko , rev .
* 81 * , 1827 ( 2009 ) .
m. krech , _ the casimir effect in critical systems _ ( world scientific , singapore , 1994 )
. h. b. chan , v. a. aksyuk , r. n. kleiman , d. j. bishop , and f. capasso , science , * 291 * , 1941 ( 2001 ) .
h. b. chan , v. a. aksyuk , r. n. kleiman , d. j. bishop , and f. capasso , phys .
* 87 * , 211801 ( 2001 ) .
e. buks and m. l. roukes , phys .
b * 63 * , 033402 ( 2001 ) .
f. m. serry , d. walliser , and g. j. maclay , j. appl . phys .
* 84 * , 2501 ( 1998 ) .
v. a. parsegian and g. h. weeiss , j. colloid .
interface sci . * 40 * , 35 ( 1972 ) .
s. barash , izv .
, radiofiz .
* 16 * , 1227 ( 1973 ) [ radiophys .
quantum electron . * 16 * , 945 ( 1973 ) ] .
s. j. van enk , phys .
a * 52 * , 2569 ( 1995 ) .
r. golestanian and m. kardar , phys .
. lett . * 78 * , 3421 ( 1997 ) .
r. golestanian and m. kardar , phys .
a * 58 * , 1713 ( 1998 ) .
t. emig , a. hanke , r. golestanian , and m. kardar , phys .
* 87 * , 260402 ( 2001 ) .
t. emig , a. hanke , r. golestanian , and m. kardar , phys .
a * 67 * , 022114 ( 2003 ) .
r. b. rodrigues , p. a. maia neto , a. lambrecht , and s. reynaud , phys .
* 96 * , 100402 ( 2006 ) .
r. b. rodrigues , p. a. maia neto , a. lambrecht , and s. reynaud , phys .
a * 75 * , 062108 ( 2007 ) .
t. emig , europhys .
lett . * 62 * , 466 ( 2003 ) .
r. bscher and t. emig , phys .
a * 69 * , 062101 ( 2004 ) .
a. ashourvan , m. miri , and r. golestanian , phys .
lett . * 98 * , 140801 ( 2007 ) .
a. ashourvan , m. miri , and r. golestanian , phys .
e * 75 * , 040103(r ) ( 2007 ) .
t. emig , phys .
* 98 * , 160801 ( 2007 ) .
j. n. munday , d. iannuzzi , y. barash , and f. capasso , phys .
a * 71 * , 042102 ( 2005 ) .
r. b. rodrigues , p. a. maia neto , a. lambrecht , and s. reynaud , europhys . lett . * 76 * , 0822 ( 2006 ) .
a. roy and u. mohideen , phys .
. lett . * 82 * , 4380 ( 1999 ) .
h. b. chan , y. bao , j. zou , r. a. cirelli , f. klemens , w. m. mansfield , and c. s. pai , phys .
* 101 * , 030401 ( 2008 ) .
a. lambrecht and v. n. marachevsky , phys .
* 101 * , 160403 ( 2008 ) .
a. lambrecht and v. n. marachevsky , int
a * 24 * , 1789 ( 2009 ) .
f. chen , g. l. klimchitskaya , u. mohideen , and v. m. mostepanenko , phys .
lett . * 88 * , 101801 ( 2002 ) .
f. chen , g. l. klimchitskaya , u. mohideen , and v. m. mostepanenko , phys .
a * 66 * , 032113 ( 2002 ) .
g. l. klimchitskaya , u. mohideen , and v. m. mostepanenko , j. phys . a : math
* 40 * , 13485 ( 2007 ) .
r. s. decca , d. lpez , e. fischbach , g. l. klimchitskaya , d. e. krause , and v. m. mostepanenko , eur .
j. c * 51 * , 963 ( 2007 ) .
chiu , g. l. klimchitskaya , v. n. marachevsky , v. m. mostepanenko , and u. mohideen , phys .
b * 80 * , 121402(r ) ( 2009 ) .
chiu , c .- c .
chang , r. castillo - garza , f. chen , and u. mohideen , j. phys .
theor . * 41 * , 164022 ( 2008 ) .
f. chen , g. l. klimchitskaya , u. mohideen , and v. m. mostepanenko , phys .
a * 69 * , 022117 ( 2004 ) .
w. r. smythe , _ electrostatics and electrodynamics _
( mcgraw - hill , new york , 1950 ) .
f. chen , u. mohideen , g. l. klimchitskaya , and v. m. mostepanenko , phys .
a * 74 * , 022103 ( 2006 ) .
l. boyer , f. houz , a. tonck , j .- l
. loubet , and j
.- m . georges , j. phys .
d : appl . phys .
* 27 * , 1504 ( 1994 ) .
r. s. decca , e. fischbach , g. l. klimchitskaya , d. e. krause , d. lpez , u. mohideen , and v. m. mostepanenko , arxiv:0904.4720 .
f. chen , g. l. klimchitskaya , v. m. mostepanenko , and u. mohideen , phys .
lett . * 97 * , 170402 ( 2006 ) .
w. j. kim , m. brown - hayes , d. a. r. dalvit , j. h. brownell , and r. onofrio , phys .
a * 79 * , 026102 ( 2009 ) .
f. chen and u. mohideen , rev .
instrum . * 72 * , 3100 ( 2001 ) .
s. g. rabinovich , _ measurement errors and uncertainties : theory and practice _ ( springer , new york , 2000 )
. g. l. klimchitskaya , f. chen , r. s. decca , e. fischbach , , d. e. krause , d. lpez , u. mohideen , and v. m. mostepanenko , j. phys . a : math
. gen . * 39 * , 6485 ( 2006 ) .
a. canaguier - durand , p. a. maia neto , i. cavero - pelaez , a. lambrecht , and s. reynaud , phys .
* 102 * , 230404 ( 2009 ) .
t. emig , j. stat .
p04007 ( 2008 ) .
d. e. krause , r. s. decca , d. lpez , and e. fischbach , phys .
* 98 * , 050403 ( 2007 ) . ,
e. d. palik ( academic , new york , 1985 )
. m. j. renne , physica * 56 * , 125 ( 1971 ) .
r. balian and b. duplantier , ann .
( n.y . ) * 104 * , 300 ( 1977 ) .
r. balian and b. duplantier , ann .
( n.y . ) * 112 * , 165 ( 1978 ) .
t. emig , r. l. jaffe , m. kardar , and a. scardicchio , phys .
lett . * 96 * , 080403 ( 2006 ) .
t. emig , n. graham , r. l. jaffe , and m. kardar , phys .
lett . * 99 * , 170403 ( 2007 ) .
t. emig , n. graham , r. l. jaffe , and m. kardar , phys .
d * 77 * , 025005 ( 2008 ) .
o. kenneth and i. klich , phys .
lett . * 97 * , 160401 ( 2006 ) .
o. kenneth and i. klich , phys .
b * 78 * , 014103 ( 2008 ) .
a. bulgac , p. magierski , and a. wirzba , phys .
d * 73 * , 025007 ( 2006 ) .
k. a. milton and j. a. wagner , j. phys . a : math .
theor . * 41 * , 155402 ( 2008 ) . s. j. rahi , t. emig , r. l. jaffe , and m. kardar , phys . rev .
a * 78 * , 012104 ( 2008 ) .
s. j. rahi , t. emig , n. graham , r. l. jaffe , and m. kardar , phys .
d * 80 * , 085021 ( 2009 ) .
a. maia neto , a. lambrecht , and s. reynaud , phys.rev .
a * 78 * , 012115 ( 2008 ) .
m. bordag , phys .
d * 73 * , 125018 ( 2006 ) .
o. m. rayleigh , proc .
a * 79 * , 399 ( 1907 ) . ,
d. maystre ( spie press , bellingham , 1993 ) .
.[tab1 ] the mean values ( column 2 ) , the variances of the mean ( column 3 ) , the systematic errors ( column 4 ) , and the total experimental errors at a 95% confidence level ( column 5 ) of the measured maximum magnitudes of the lateral casimir force @xmath110 at different separations ( column 1 ) for the first set of measurements . [ cols="^,^,^,^,^ " , ] | the lateral casimir force , which arises between aligned sinusoidally corrugated surfaces of a sphere and a plate , was measured for the case of a small corrugation period beyond the applicability region of the proximity force approximation .
the increased amplitudes of the corrugations on both the sphere and the plate allowed observation of an asymmetry of the lateral casimir force , i.e. , deviation of its profile from a perfect sine function .
the dependences of the lateral force on the phase shift between the corrugations on both test bodies were measured at different separations in two sets of measurements with different amplitudes of corrugations on the sphere .
the maximum magnitude of the lateral force as a function of separation was also measured in two successive experiments .
all measurement data were compared with the theoretical approach using the proximity force approximation and with the exact theory based on rayleigh expansions with no fitting parameters . in both cases real material properties of the test bodies and nonzero temperature
were taken into account .
the data were found to be in a good agreement with the exact theory but deviate significantly from the predictions of the proximity force approximation approach .
this provides the quantitative confirmation for the observation of diffraction - type effects that are disregarded within the pfa approach .
possible applications of the phenomenon of the lateral casimir force in nanotechnology for the operation of micromachines are discussed . |
paraxial propagation of linearly polarized laser beams through transparent optical media with intensity dependent refractive index is mathematically equivalent to the free evolution of the wavefunction order parameter used in the mean - field description of a two - dimensional gas of @xmath0 interacting atoms in a bose - einstein condensate ( bec ) at temperature @xmath1 @xcite .
both systems can be modeled by identical nonlinear schrdinger equations@xcite . for photons in the laser beam
, the @xmath2 component of the nonlinear optical susceptibility plays the same role as n - body interactions between atoms in the cloud and the propagation constant can be identified with a chemical potential for the light distribution .
as all the photons in a coherent wave are equal , the laser beam can be treated on equal foot as any system of @xmath0 identical interacting bosons at zero temperature@xcite .
the previous point of view , which takes into account the equivalence between laser beams and becs of ultracold atoms , has led to an interesting suggestion made by chiao@xcite , who recently proposed to verify the superfluidity of coherent light , in analogy with degenerate quantum atomic gases .
more recently , similar concepts have been successfully used to analyze condensation phenomena of nonlinear waves@xcite and quantum phase transitions of photons in periodic lattices@xcite . in chiao s model the key point of the analysis
is _ `` same equations , same predictions '' _ and therefore photons from a monochromatic laser source are considered as an ideal bosonic gas at zero temperature in which continuous phase transitions ( cpt ) can take place due to long range quantum fluctuations around the ground state@xcite .
these critical phenomena are thus called _ quantum _ phase transitions ( qpt)@xcite to distinguish them from the standard phase changes which are well - known in classical thermodynamics . as we will show in this work
, cpt may occur in any _ classical _ system at zero temperature without long range quantum correlations involved , if opposite nonlinear interactions are present . in the case of chiao s `` superfluid light ''
the phase transition is produced by effect of a defocusing intensity - dependent refractive index@xcite and thus a waveguide is used to avoid spreading of the beam ( in the same way as magneto - optical traps are employed to hold atomic becs ) . however , as superfluidity may occur both in gases and liquids , it can not be considered as a trace of the presence of a liquid state@xcite ; it is also required the appearance of surface tension effects@xcite . moreover , in chiao s model , the interactions between particles are repulsive and they can not drive a gas - liquid transition . thus , in this work we will follow the same lines of thought to suggest the possibility of obtaining a gas - to - liquid phase transition in a classical gas at @xmath1 described in a mean field theory by the so - called _ cubic - quintic _ ( cq ) model with competing nonlinearities .
several pioneering works have highlighted the interesting properties of this cq - model@xcite .
cavitation , superfluidity and coalescence have been investigated@xcite in the context of liquid he , where the model is a simple approach if nonlocal interactions are not taken into account .
stable optical vortex solitons and the existence of top - flat states have been also reported in optical materials with cq optical susceptibility@xcite .
the surface tension properties that appear in this system@xcite have been considered as a trace of a `` liquid state of light''@xcite . on the other hand , recent experiments about filamentation of high - power laser pulses in @xmath3
have shown that the cq nonlinearity is achievable in this material@xcite .
it has been also suggested that atomic coherence may be used to induce a giant cq - like refractive index of a rb gas@xcite .
thus , the practical realization of the first `` liquid of light '' state as an example of non - quantum liquefaction at zero temperature is close . in bec systems
, a cq - model can be used in the mean field description of an ultracold gas at zero temperature in the presence of efimov states with tunable two- and three - body interactions , which have been recently proposed@xcite .
the cubic - quintic model describes a coherent bosonic system of @xmath0 particles with two- and three - body interactions .
the mathematical formulation of the mean field theory yields a generalised non - linear schrdinger ( nls , also called gross - pitaevskii ) equation of the form : @xmath4 if the system modeled by the previous equation is a photon gas , the above nls describes paraxial propagation of a continuous linearly - polarized laser beam of wavelength @xmath5 in a nonlinear medium with a refractive index depending on the intensity @xmath6 in the form @xmath7 and the adimensional variables are : @xmath8 the propagation distance multiplied by @xmath9 , @xmath10 the beam irradiance multiplied by the kerr coefficient @xmath11 , @xmath12 an adimensional parameter indicating the strength of the quintic nonlinear optical susceptibility , and @xmath13 the transverse laplacian operator , where @xmath14 and @xmath15 are the transverse spatial dimensions multiplied by @xmath16 . in the case of a two - dimensional atomic bec
tightly trapped along one axis by a parabolic potential of frequency @xmath17 and thickness @xmath18 , the adimensional variables correspond to : @xmath8 the time in units of @xmath19 , @xmath10 the atomic density multiplied by two - body coefficient @xmath20 , @xmath21 an adimensional parameter indicating the strength of the three - body interactions , and @xmath13 the transverse laplacian operator , where @xmath14 and @xmath15 are the transverse spatial dimensions divided by @xmath22 .
recent experiments for beam propagation in @xmath3@xcite at @xmath23 yield the following values for the above parameters in the case of laser beams : @xmath24 , @xmath25 and @xmath26 . other materials like
air@xcite or chalcogenide glasses@xcite seem to display the c - q behavior usually accompanied by ionization and non - linear losses .
it has been also pointed the possibility of engineering this type of optical response by quantum techniques which allow to access this nonlinear regime with miliwatt ultrastabilized lasers@xcite . for atomic bec systems
it has been proposed that a combination of two - body ( attractive ) and three - body ( repulsive ) elastic interactions can yield liquid _
this behaviour can be explained in terms of the efimov states@xcite .
however , three - body scattering in becs has inelastic contributions and yields highly nonlinear losses .
this means that in the most general case the coefficient @xmath21 in equation may be complex both for laser systems as well as for atomic gases .
the above cq - nlse admits soliton - like solutions of finite size@xcite of the form @xmath27 , being @xmath28 the propagation constant in the case of light and the chemical potential for atomic becs .
these solitons can only be calculated numerically and coexist with plane waves solutions of constant amplitude @xmath29 , which lead by substitution in eq . to @xmath30 . in fig.[fig1 ] we have plotted the maximum value of @xmath31 vs. @xmath28 in units of @xmath32 , for different kinds of stationary solutions of eq . .
the continuous and dotted lines correspond respectively to stable and modulationally unstable plane waves@xcite , whereas the dashed line stands for numerically calculated localized eigenstates@xcite .
it is known@xcite that the shape of the solitons ( see dashed profiles in fig.[fig2 ] ) vary from quasi - gaussian shapes for low powers to almost square profiles for beam amplitudes close to a certain critical value of @xmath33@xcite . at this point
the size of the solutions tends to infinity whereas the amplitude of the beam stabilizes at @xmath34 .
we will now clarify the reasons of this behavior by using a thermodynamic model .
the appropriate tool to study the equilibrium condition as a function of the number of particles @xmath35 is landau s grand potential @xmath36,\end{aligned}\ ] ] where the lagrange multiplier @xmath28 is the chemical potential for a bec or the propagation constant in the case of an optical system . for our two - dimensional model
the area is @xmath37 and we have that @xmath38 is the pressure .
the equilibrium configurations are obtained when @xmath39 is minimized . in particular
@xmath40 implies that the pressure has to be zero , which for the plane waves of fig.[fig1 ] yields @xmath41 .
there are two possibilities : the trivial case @xmath42 and a uniform phase , with @xmath43 and @xmath44 . in the language of field theory , these solutions are the two possible vacuum states of the system . as it can be seen in fig .
[ fig1 ] the two vacua can be `` connected '' by the soliton solutions of the dashed line in fig .
[ fig1 ] , which constitute the _ instantons _ of the theory@xcite .
it is interesting to notice that the non - zero vacuum solution implies a spontaneously symmetry breaking of the global phase symmetry , which is preserved in the case with @xmath45 .
this result about the pressure has already been discussed by authors in ref @xcite within the framework of the madelung transformation ( mt ) , which allows one to establish a formal analogy between nonlinear optics and classical fluid dynamics@xcite . with the aid of the mt ,
an analytical expression for the effective pressure has been derived@xcite .
very remarkably , this effective pressure , which is an approximation since it does not take into account the _ quantum - mechanical pressure term_@xcite , calculated for the plane - wave solutions of the cq - nlse has the same mathematical expression as the one shown before . unlike the previous works related in the references , we have calculated the effective pressure by means of the potential given by eq . which contains all the relevant information about the nonlinear system which is currently being studied .
while the analysis of the infinite plane - wave solutions has already been considered in the literature @xcite , our discussion of the pressure also applies to the localised soliton solutions , being a central contribution of the present work .
we also find a significant difference between the low - power quasi - gaussian solutions and the high - power top - flat solutions of eq .. the expression for plane - waves and top - flat eigenstates can be approximated by the same `` reduced '' expression for @xmath46 given by both the mt and the omega density calculation .
in fact , we have shown in fig.[fig1 ] that the two branches of solutions , i.e. the localised solitons and the stable plane waves , merge as @xmath28 approaches @xmath47 , having a radius increasing to infinity , thus in this case it is justified to neglect @xmath48 in eq .. on the other hand , for the quasi - gaussian eigenstates the previous gradient term is not negligible . in this case
, we should calculate @xmath46 numerically in the lack of analytical solutions .
the analysis of the pressure can be refined calculating numerically its distribution for different eigenstates as it is plotted in fig.[fig2 ] .
the pressure of the plane wave branches in fig.[fig1 ] is plotted as a function of @xmath49 .
very remarkable is the fact that the pressure of the stable branch is higher than its counterpart of the unstable branch .
this implies that the stable plane waves free energy density is smaller than the free energy of the modulationally unstable solutions branch . in fig.[fig2 ] it is also plotted the curve of the eigenstates central pressure .
as it can be seen in the graph , the curve corresponding to the eigenstates is bounded by the two curves of the plane - waves branches , so that the existence domain of the filaments phase is limited by them . in the insets of fig.[fig2 ]
, we show the shape profiles of the eigenstates ( dashed line ) superposed with their effective pressure profiles ( solid lines ) which have been conveniently rescaled to fit in the graph .
as it can be appreciated in inset a ) , solitons with @xmath50 have smooth pressure distributions with a central maximum located at the centroid of the soliton and two negative - valued minima . as the value of @xmath51 increases , the soliton profiles and their corresponding shapes of @xmath46 narrow , reaching a minimum width at @xmath52 , which corresponds to a filament soliton solution with the same peak amplitude as the `` critical '' plane wave with @xmath53 .
this plane - wave marks the border between stable and modulationally unstable plane waves@xcite . when the chemical potential reaches the value @xmath54 ( see inset b ) , the absolute maximum of @xmath46 for eigenstates is obtained , i.e , this filament has the minimum of @xmath55 at its centroid .
we also consider that close to @xmath56 , the liquid top - flat eigenstates begin to exist@xcite .
in fact , the curve of stable localised solutions in fig.[fig1 ] has a maximum in that region and both localised solution and stable plane - waves branches seem to merge there . for @xmath57 ,
the pressure maximum is located in a flat region ( see inset c ) , tending to zero as @xmath51 approaches the critical value @xmath58 , point d ) in fig.[fig1 ] , over which no localized solutions exist@xcite .
in this section , we will provide a set of numerical simulations showing the condensation process , i.e. , the phase transition from the gaseous phase to an homogeneus coherent `` liquid '' plane wave solution corresponding to the upper branch in fig.[fig1 ] . in order to achieve this result
, we will perturb an unstable plane wave , corresponding to the lower branch in fig [ fig1 ] , with a randomly varying noise .
this will produce a filamented phase , made of coherent structures whose shape remains qualitatively unchanged , up to smaller scale fluctuations@xcite . in order to achieve the non - quantum liquefaction we have included both a linear incoherent pumping mechanism and nonlinear three - body losses@xcite .
in other words , we have considered @xmath59 and we have introduced a linear gain term @xmath60 in eq .. this corresponds to a continuous load of particles in the system .
note that in this way , we are describing a more realistic non - conservative version of eq . , which models an experimentally achievable scenario in the framework of current bec experiments .
in fact , in condensed matter systems it is possible to control the load of particles and two and three - body recombination within the coherent atomic cloud@xcite . on the other hand , in nonlinear optics , this kind of nonlinear models are well - known in the frame of complex ginzburg - landau ( g - l ) equations used to describe wide - aperture laser cavities@xcite .
although it is always possible to control the linear gain introduced in the system , three - body losses are often imposed by the nonlinear response of the material , so it is not possible to manage the dissipation terms of the system . however , by means of electromagnetic induced transparency techniques , it is possible to customize the nonlinear optical response of cold atomic ensembles like rb@xcite so that the nonlinear refractive index corresponds to the one given by the modified eq . analyzed in the current section of the paper . therefore our model , and its predicted phenomenon of liquefaction that we will demostrate below , can correspond both to realistic bec and nonlinear optical systems . in fig.[fig3 ] , the initial state consists of an incoherent set of filaments with a randomly varying phase distribution . within this apparent disorder ,
some coherent uncorrelated structures ( filaments ) exist and can be observed in fig.[fig3 ] a ) . in this simulation
, we have considered a nonlinear parameter regime where the linear gain was not enough to compensate the three - body dissipative term . as a consequence , certain degree of coherence
is lost since the filaments disappear and only the noisy background is observed , as shown by snapshot c ) in fig.[fig3 ] ) . starting from the same initial condition but increasing the linear gain term over a certain threshold
, we have performed the simulation shown in fig.[fig4 ] . in this case , we see that the system evolves towards an homogeneus plane wave by the combined effect of adding particles to the initial random state and dissipation due to many - body inelastic processes .
this relaxation process is well - known in the context of the complex g - l equations and is attributted to the non - conservative nature of the model@xcite .
nevertheless , as our theory predicts , the system will tend to form a particular plane wave of the stable `` liquid '' upper branch of fig.[fig1 ] , the one with zero pressure .
this illustrates the tendency of the system to reach the non - zero vacuum state with @xmath61 .
very remarkably , during the process pairs of vortex - antivortex with topological charges @xmath62 are formed ( see bottom phase maps in fig.[fig4 ] ) , so that the constant phase of the emerging coherent wave remains hidden by the overlapping of the different vortex rotating phase - distributions .
vortices are very robust topological structures@xcite and in our simulations they remain stable as far as we could follow the numerical simulations . finally , we have considered the effect of replacing the linear gain term by a nonlinear gain term of the form @xmath63 . in this situation ,
the numerical results are qualitatively the same as in fig.[fig4 ] , although it can be observed that in the last stage of the evolution the vortices are annihilated . in snapshot
f ) of fig.[fig4 ] , it can be seen how the underlying plane wave , which emerges after the dynamical process as described above , has a homogeneus phase distribution .
we think that this result is a very interesting example of a self - organization process , where coherence is produced from disorder , with evident practical applications .
we have shown in the present work that a system of @xmath0 equal bosons in a system with competing nonlinearities can undergo a phase transition from a gas to a liquid state .
the process takes place at zero temperature without any quantum effect and it is only ruled by nonlinear interactions .
finally , we have shown that a cubic - quintic medium with complex susceptibilities exhibiting linear gain and nonlinear losses , will tend to produce a homogeneous phase liquid distribution starting from a collection of non - correlated filaments .
this opens the door for experiments in the field of bec systems in ultracold gases .
h. m. thanks a. j. legget and a. ferrando for useful discussions and y. castin , j. ho and g. v. shlyapnikov for warm hospitality at inst . henri poincar .
this work was supported by mec , spain ( projects fis2006 - 04190 and fis2007 - 62560 ) xunta de galicia ( project pgidit04tic383001pr and d.n .
grant from consellera de educacin , xunta de galicia ) .
e. l. bolda , r. y. chiao and w. h. zurek , , 416419 ( 2001 ) ; r. y. chiao _
et al . _ , j. phys .
b : at . mol .
phys * 37 * , s81s89 ( 2004 ) ; r. y. chiao _ et al .
_ , , 063816 ( 2004 ) ; a. tanzini and s.p .
sorella a * 263 * , 4347 ( 1999 ) .
h. michinel , j. campo - tboas , m. l. quiroga - teixeiro , j. r. salgueiro , and r. garca fernndez , j. opt .
b : quantum semiclass . opt . * 3 * , 314 ( 2001 ) ; m. j. paz - alonso and h. michinel , phys .
lett . * 94 * , 093901 ( 2005 ) .
crasovan , ba . malomed and d. mihalache , , 016605 ( 2001 ) ; p. grelu , j. soto - crespo and n. akhmediev , opt .
express * 13 * , 23 , pp : 9352 - 9360 ( 2005 ) s. mancas , sr .
choudhury , chaos , solitons and fractals , * 27 * , 5 , pp : 1256 - 1271 ( 2006 ) | we show that a gas - to - liquid phase transition at zero temperature may occur in a coherent gas of bosons in the presence of competing nonlinear effects .
this situation can take place both in atomic systems like bose - einstein condensates in alkalii gases with two and three - body interactions of opposite signs , as well as in laser beams which propagate through optical media with kerr ( focusing ) and higher order ( defocusing ) nonlinear responses .
the liquefaction process takes place in absence of any quantum effect and can be formulated in the frame of a mean field theory , in terms of the minimization of a thermodynamic potential .
we also show numerically that the effect of linear gain and three - body recombination also provides a rich dynamics with the emergence of self - organization behaviour . |
in this work the 4@xmath4 pure compact u(1 ) gauge theory is revisited . this model , being the prototype of gauge abelian interaction , has received a lot of attention through the years , since it constitutes a preliminary step in approaching qed . however , the questions related with the mechanisms driving the transition and with the order of the phase transition itself , have turned out to be a more controversial issue than one might expect from its apparent simplicity . through this section
we shall summarize the main results obtained up to date .
the 4@xmath4 pure compact u(1 ) theory is known to undergo a phase transition between a region in which the photons are confined ( strong coupling ) and another region in which they carry a coulomb potential ( weak coupling ) .
the theory can be discretized in the lattice using the wilson action @xmath5 \ . \label{sw}\ ] ] most of the work has been done on the hypercubic lattice with periodic boundary conditions , namely on the hypertorus . on this topology
the pioneer studies ( see eg .
@xcite ) pointed to a continuous transition since numerical simulations did not evidence signals of metastability .
the confinement - deconfinement transition sets in at a coupling @xmath6 .
the largest lattice size simulated in those works was @xmath7 .
however , as simulations in larger lattices became available to computer resources , metastabilities were evidenced , and two - peaked energy histograms could be observed from @xmath8 on ( see eg .
@xcite ) pointing to a first order character for the phase transition .
this idea was also supported by several works that approached the problem from the point of view of the renormalization group @xcite . nowadays , though the general belief is that the transition is first order , there are still some puzzling questions .
namely , the numerical simulations up to @xmath9 do not reveal a stable latent heat , it is rather decreasing with the lattice size , and also , the associated critical exponent @xmath10 is always in the interval @xmath11 , which is different from the first order value ( @xmath12 ) and from the trivial second order one , namely @xmath13 .
these facts led some authors to wonder about the possibility of a slowly vanishing latent heat , which could eventually go to zero in the thermodynamic limit . in short ,
the 4@xmath4 pure compact u(1 ) theory could be critical provided the two peaks appearing in the simulations are a finite size effect . due to the physical implications of a critical theory with non trivial critical exponents this possibility was not completely ruled out .
now , we go back in time to introduce the extended wilson action .
besides the wilson action , an action including a term proportional to the square of the plaquettes was proposed by g. bhanot at the beginning of the eighties @xcite @xmath14 + \gamma \sum_{\rm p } [ 1 - { \mathop{\rm re}\nolimits}{\mathop{\mathrm{tr}}}u_{\rm p}^2 ] \ . \label{sew}\ ] ] in that work the extended ( @xmath15,@xmath3 ) parameter space was explored in @xmath16 , and it was suggested that the phase transition line @xmath17 became first order for @xmath18 , while for @xmath19 the transition appeared to be second order on those small lattices .
the phase diagram is represented in figure [ phd ] .
simulations in larger hypercubic lattices have shown that two metastable states are present for all , up to date , positive and negative simulated @xmath3 values .
what holds true is the observation that a positive @xmath3 value reinforces the transition towards the first order , while a negative @xmath3 value weakens the transition .
in fact , the latent heat , @xmath20 , decreases with @xmath3 , and then , a coupling where @xmath20 goes to zero was supossed to exist at some finite ( negative ) value of @xmath3 .
this behavior induced some authors to conjecture about the existence of a tricritical point ( tcp ) at some negative coupling @xmath21 where the order of the phase transition could change from being first order to be continuous @xcite .
however , every simulation performed at negative @xmath3 has exhibited double peak structures for large enough @xmath22 .
as for the wilson action ( @xmath23 ) , for negative @xmath3 a non stable latent heat is revealed by simulating in lattice sizes in the range @xmath24 . from the point of view of the renormalization group this problem was studied in @xcite and the results pointed to a phase transition line @xmath17 of first order for all @xmath3 except at the end , @xmath25 , where the transition is second order with trivial exponents .
nowadays , it is generally accepted that the transition is first order for positive @xmath3 values .
however for @xmath26 two scenarios arise : 1 ) the phase transition line @xmath17 is first order everywhere except at @xmath27 ; 2 ) there is a finite @xmath28 from which on , the phase transition is second order .
as we have already remarked , all the previous discussions refer to simulations performed on the hypertorus .
recently some authors have cast a new hue on this problem .
it was suggested some years ago that the jump in the energy observed when the confining - deconfining transition takes place , could be due to the existence of closed monopole loops wrapping around the toroidal lattice , which due to the particular topology can not be contracted to a point @xcite .
this hypothesis led authors to work on lattices homotopic to the sphere , since on this topology all monopole loops can be contracted to a single point @xcite .
the relationship between the closed monopole loops and the discontinuity in the energy is far from being proved .
the influence of those non trivial monopole loops on the phase transition has been studied by rebbi and coworkers @xcite .
their results suggest that the properties of the phase transition seem not to be affected by the existence of such closed monopole loops .
however it is claimed the two state signal does dissapear when working on a lattice constructed considering the surface of a 5@xmath4 cube , which has the topology of a sphere @xcite . this single peak structure together with the estimation of the critical exponent @xmath29 measured at @xmath30 , has induced authors in @xcite to claim that the double peak signal on the torus is a finite size effect , and when working on the sphere one observes that indeed the line @xmath17 for @xmath26 is a line of second order phase transitions , all of them in the same universality class characterized by a @xmath10 exponent around @xmath31 . assuming this second order scenario for the pure gauge theory ,
fermions have been included ( on the toroidal topology ) in order to study how fermions fit in the u(1 ) theory .
also in this case a second order phase transition is conjectured to exist @xcite .
this situation is somehow disturbing , because one would expect to get the asymptotic behavior of a phase transition earlier on a lattice which is homogeneous and translationally invariant , namely the hypertorus , than on a lattice which is manifestly non homogeneous , as is the surface of a 5d cube , and on which the translational invariance is only recovered in the limit of infinite volume .
first order phase transitions exhibit a wide variety of behaviors .
some of them do not show any pretransitional effect , the correlation length at the critical point , @xmath32 , is small , and the thermodynamic quantities diverge as the volume for reasonable lattice sizes .
however when the correlation length is very large , pretransitional effects take place , and they can also be very significant . in literature these transitions are referred to as weak first order ( wfo ) phase transitions and are characterized by small discontinuities , such as a small latent heat .
depending on whether or not @xmath22 is larger than @xmath32 , a wfo phase transition presents two differentiate regions : * _ transient region _ ( @xmath33 ) .
the transition behaves like a second order one , since @xmath32 is effectively infinite .
thermodynamic quantities diverge less than @xmath34 and hence the effective critical exponent @xmath10 is larger than @xmath35 .
* _ asymptotic region _
( @xmath36 ) .
the first order character of the transition is revealed by the onset of a stable two - peak structure and an effective @xmath10 compatible with @xmath35 .
how large should @xmath32 be for the transition to be considered wfo seems a subjective matter . since @xmath37
always , all first order transitions would be weak in some extent . the key point for a transition to be considered
wfo is the existence of pretransitional effects .
the pretransitional behavior makes the transition along the transient region to look like a second order one with a definite critical coupling @xmath38 .
this second order phase transition is preempted by the true first order one at the coupling @xmath39 .
the distance between the couplings @xmath38 and @xmath39 can be taken as a measure of the weakness of the transition @xcite .
typical examples of such wfo transitions are 2@xmath4 potts models @xcite and the deconfining transition in finite t qcd @xcite . from this description
it is clear that the observation of no double - peak structures does not prove by itself the second order character of a phase transition . in view of this
, we decided to make a systematic study of the phase transition line @xmath17 on the hyper - torus ( @xmath40 topology hereafter ) and on the lattice with spherical topology following @xcite ( which will be referred to as @xmath41 topology ) . on the hypertorus we improve on the statistics running up to @xmath42 at some negative @xmath3 values . on the sphere we have run lattices substantially larger than those employed by authors in @xcite to check if the two - state signal reappears when
larger spheres are considered .
our preliminary results , suggesting that the double peak structure indeed reappears , are published somewhere else @xcite . here , the statistics on those lattice sizes have been increased , and simulations in larger lattices are included .
new observables are used to extract our conclusions .
we shall consider the parameter space described by the extended wilson action , which can be expressed in terms of the plaquette angle in the following way : @xmath43 we use for the simulations the conventional @xmath44 hypercube with toroidal boundary conditions ( hypertorus ) , and , for comparison we also consider the surface of a @xmath45 cube which is topologically equivalent to a @xmath44 sphere .
contrarily to the hypertorus , the surface of a @xmath45 cube is not homogeneous .
there are a number of sites which do not have the maximum connectivity , namely 8 neighbors . due to this fact
, uncontrolled finite size effects are expected to turn up .
their influence can be somehow alleviated by the introduction of appropriate weight factors in those inhomogeneous sites @xcite . since the topology remains unchanged and we do not expect the order of the phase transition to be affected by the rounding , we do not use weight factors . as a notational remark ,
we label with @xmath22 and @xmath46 the side length for the @xmath40 topology and for the @xmath41 topology respectively .
next we define the energies associated to each term in the action @xmath47 where @xmath48 stands for the number of plaquettes . on the hyper - torus
this number is simply proportional to the volume , namely the forward plaquettes are @xmath49 . on the sphere
the number of plaquettes has a less simple expression , and can be computed as a function of @xmath46 as @xmath50 . in this case , the system is not homogeneous and @xmath48 is not proportional to the number of points on the four dimensional surface , which is @xmath51 , some points having a number of surrounding plaquettes less than the possible maximum @xmath52 , as opposed to what happens on the torus . in order to allow comparison , we define @xmath53 , in such a way that a hypertorus of @xmath54 has the same number of plaquettes as the corresponding hypersphere
. the specific heat and the binder cumulant are useful quantities to monitorize the properties of a phase transition . at the critical point , they are known to posses different thermodynamical limits depending on whether the transition is first order or higher order , and hence , their behavior with increasing lattice size can give some clues as to the order of the phase transition .
we have studied the finite size scaling ( fss ) of these two energy cumulants .
the specific heat is defined for both energies as : @xmath55 @xmath56 as we have observed a high correlation between both energies , the results being qualitatively the same , we shall only report on the observable defined for the plaquette energy @xmath57 . in a second order phase transition , scaling theory predicts the specific heat maximum to diverge as @xmath58 .
if the transition is first order it is expected to diverge like the volume @xmath34 ( or strictly like the number of plaquettes @xmath48 ) reflecting that the maximum of the energy fluctuation has the size of the volume .
this is expected to hold in the asymptotic region of the transition @xmath36 . in the transient region , namely @xmath59
, the specific heat is expected to grow more slowly than the volume .
we shall use the number of plaquettes @xmath48 to study the specific heat behavior in order to have a single parameter for both the @xmath40 and the @xmath41 topologies .
this is of course equivalent to doing the discussion as a function of @xmath22 on the @xmath40 and @xmath60 on the @xmath41 .
we have in this case @xmath61 .
we have also studied the behavior of the fourth cumulant of the energy @xmath62 when the energy distribution describing the system is gaussian @xmath63 in the thermodynamic limit .
this is the case for a second order phase transition .
if the transition is first order , far from the critical coupling @xmath39 , @xmath64 also tends to 2/3 , reflecting the gaussianity of the energy distribution .
however , at @xmath39 the distribution can be described by two gaussians centered about the energy of each metastable state @xmath65 and @xmath66 , and hence this quantity has the non - trivial thermodynamic limit @xmath67 another interesting quantity measurable from the density of states is the distribution of the partition function zeroes , or lee - yang zeroes @xcite . to clarify the role of the partition function zeroes let us first stress a well known fact : there are no phase transitions on a finite volume , phase transitions arise in the thermodynamical limit and are signaled by non - analyticity in the free energy : @xmath68 @xmath69 is a linear combination of exponentials , and hence an analytical function .
this implies that the free energy can be singular only where @xmath70 .
however if the coupling @xmath15 is real , and the volume is finite , that linear combination is a sum of positive terms and hence , the zeroes of @xmath69 are located in the complex plane of the coupling @xmath15 .
the onset of the phase transition in the limit of infinite volume is signaled by a clustering of the zeroes on the real axis at @xmath39 .
here follows the description of our procedure , for a general description of the method see @xcite .
as mentioned previously we work only with @xmath57 , and then we consider only the spectral density method @xcite for this variable .
since we work at fixed @xmath3 we consider only the @xmath15 coupling derivatives . from the mc simulation at @xmath15
we obtain an approximation to the density of states which allows us to compute the normalized energy distribution @xmath71 .
the energy distribution can be expressed as @xmath72 we use the standard reweighting technique to obtain from the distribution measured at @xmath15 the distribution at another coupling @xmath73 , which is complex in the more general case @xmath74 .
the standard reweighting formula is : @xmath75 using ( [ prob ] ) and the normalization condition @xmath76 one obtains @xmath77 we can factorize the contributions from the real and imaginary part : @xmath78 this is the standard reweighting formula extended to the complex parameter space of couplings . as a first observation we have a pure oscillating factor due to @xmath79 .
since @xmath80 is @xmath81 this is a rapidly oscillating function which makes it impossible to locate zeroes with large imaginary part .
the real part of the coupling contributes to the well known exponential damping , @xmath82 , which is telling us that we can trust the extrapolation only in a small neighborhood @xmath83 .
since @xmath84 has no zeroes in a finite volume , an easy way to locate the zeroes numerically is looking at the minima of the function @xmath85 , where : @xmath86 the function to minimize is : @xmath87 one interesting property of the partition function zeroes concerns the estimation of critical exponents . denoting by @xmath88 the coupling where the first zero is located , the distance to the real axis scales with the @xmath10 exponent : @xmath89 we shall use this property to compute the effective @xmath10 exponent . from @xmath90 , we can measure the free energy gap , @xmath91 , which is the difference between the minima and the local maximum of the free energy @xcite .
we use the spectral density method to get , from the measured histograms , a new histogram where both peaks have equal height .
we take the logarithm of those histograms and measure the energy gap .
a growing @xmath91 in the asymptotic region of the transition implies a first order phase transition .
an increase proportional to @xmath92 is expected @xcite . for the transition to be second order
, @xmath91 must stay constant with increasing lattice sizes .
[ tabla_esfera ] [ tabla_toro ] most of the work has been done by simulating the subgroup @xmath93 , since the phase transition associated to the discrete group lies safely far away .
the overrelax algorithm can be applied only in the simulations with @xmath23 so that the gain in statistical quality due to the overrelax effect is limited to the wilson action , where we have simulated both , the full group @xmath94 and the discrete one for the sake of comparison . for @xmath95
we have simulated the discrete group since so the simulation is considerably speeded up . for every lattice size we consider , we perform trial runs to locate the coupling @xmath96 where the specific heat shows a peak .
we use the standard reweighting techniques to extrapolate in a neighborhood of the simulated coupling .
once the peak is located within an error in the fourth digit of the coupling , we perform an intensive simulation there to get @xmath97 .
the statistics performed at these pseudo - critical couplings are reported in table [ tabla_esfera ] and table [ tabla_toro ] for the @xmath41 and the @xmath40 topology respectively .
typically we measure the energies every 10 @xmath98 sweeps in order to construct the energy histogram ( [ prob ] ) from this energy distribution we obtain the cumulants of the energy we are interested in , and the critical exponents using finite size scaling techniques .
we remark that one of the main sources of systematic error when measuring critical exponents , is the indetermination in the coupling @xmath96 where to measure .
the simulations have been done in the _ rtnn _ machine consisting of 32 pentium pro 200mhz processors .
the total cpu time used is the equivalent of 6 pentium pro years .
we have updated using a standard metropolis algorithm with 2 hits .
the acceptance has always been between 65% and 75% . in order to consider the statistical quality of the simulation , following
@xcite we define the unnormalized autocorrelation function for the energy c(t ) = _ i=1^n - t e_i e_i+t - e ^2 , as well as the normalized one ( t ) = .
the integrated autocorrelation time for the energy , @xmath99 , can be measured using the window method ^int(t ) = + _ t^ = 1^t ( t^ ) , for large enough @xmath100 , which is in practice selected self - consistently .
we use @xmath100 in the range 5@xmath99 , 10@xmath99 , and we check that the obtained @xmath99 remains stable as the window in @xmath100 is increased . we have always started from hot and cold configurations in order to make sure that the system does not remain in a long living metastable state , which could be interpreted as a dirac sheet .
the results coming from both types of starts have always been indistinguishable .
we have studied the deconfinement - confinement phase transition line at several values of @xmath3 ( figure [ phd ] ) . in the region of negative @xmath3 ,
apart from the deconfinement transition line , there is another transition line provoked by the competing interaction between the couplings ( @xmath101 ) . in the limit @xmath27
the model is not dynamical since for all finite @xmath15 , @xmath102 .
so we expect this line to end at the corner @xmath103 .
we have not performed a deep study of this transition line , however the simulations at @xmath104 revealed double - peak structures pointing to a first order character .
the limit @xmath105 is equivalent to a @xmath106 theory and the critical point can be calculated exactly by self - duality .
we have studied carefully the region between lines @xmath107 and @xmath108 finding no signatures of the existence of additional lines .
we have focused on the transition line @xmath107 , at several values of @xmath3 , on the hyper - torus and on the hyper - sphere .
the structure of this section is the following : first , we describe the results at @xmath109 on the spherical topology .
our purpose is to observe how does the @xmath41 topology behave in a non controversial region , and comparing with the known results on the hypertorus .
second , we study the point @xmath23 ( wilson action ) on the spherical lattice .
this coupling has been recently claimed to be the starting point of a critical line ( infinite correlation length at the critical point ) which extends in the range @xmath26 , with an associated critical exponent @xmath29 @xcite .
since this assertion relies on the absence of two - state signals on spherical lattices up to @xmath110 , we shall check whether or not double peak structures set in when larger spheres are considered .
third , we go to the region of negative @xmath3 .
we are aware of no previous systematic study on the hypertorus in this region , so we run simulations up to @xmath42 at @xmath111 . motivated by the results on the hyper - torus we have just run a single @xmath3 negative value on the hyper - sphere . for the sake of comparison with @xcite we choose this value to be @xmath112 .
finally , we discuss the finite size scaling behavior exhibited by both topologies . as we have pointed out ,
there is a general agreement on considering the phase transition first order in the region of positive @xmath3 .
the lattice sizes used range from @xmath114 to @xmath115 , which correspond to @xmath116 and @xmath117 respectively see ( table [ tabla_esfera ] ) . in figure
[ evol02 ] , the mc evolution for @xmath118 is shown .
we remark that no multicanonical update is needed to obtain a very high rate of flip - flops up to @xmath115 .
however , on the torus , for @xmath119 the probability of tunneling between both metastable states is so tiny that a reasonable rate of flip - flops is not accessible to ordinary algorithms .
probably the inhomogeneity of the sphere is increasing the number of configurations with energies which correspond neither to the confined nor to the deconfined phase , but in between .
these configurations make the free energy gap to decrease and hence the tunneling is easier on the @xmath41 topology . in short , the sites without maximum connectivity act as catalysts of the tunneling . in figure [ histo02 ]
the energy distributions are plotted .
the distribution at @xmath120 is distinctly non - gaussian , but a blatant two - peak structure is observed only from @xmath121 ( @xmath122 ) on .
so , when comparing with the result at @xmath123 on the torus , ( see small window in figure [ histo02 ] ) a first observation is that the onset of a two - state signal is particularly spoiled by the @xmath41 topology , at least in 4@xmath4 pure compact u(1 ) gauge theory . concerning the latent heat , we remark its stability already at @xmath121 , or , if anything , its increase from @xmath121 to @xmath115 .
one would even say that the same happens between the positions of the would - be two states in @xmath120 and the position observed in @xmath121 .
the behavior of the binder cumulant reflects this fact ( see figure [ binder02 ] ) .
we observe a rapid growth of @xmath124 for small lattices sizes , apparently towards @xmath125 ( second order value ) .
this growth stops when the splitting of the two peaks is observed .
the splitting of the two peaks is reflected by smaller values in @xmath124 .
we remark that the errors quoted for @xmath124 are calculated taking into account the indetermination in the value of the minimum , but not the displacement in the position of the coupling where the minimum appears . in our opinion , mainly two reasons can give account of this behavior .
the first one comes from general grounds : at the very asymptotic region of a first order phase transition , the energy jump gets sharper and sharper , and a slightly increase of the latent heat could be expected
. the second one would be the increasing restoration of homogeneity in the hypersphere when increasing the lattice size .
the latent heat observed in small @xmath46 might be affected by the inhomogeneity of the hyper - sphere . at this point
we can not give a single reason for this to happen .
we postpone a stronger conclusion to the section devoted to the finite size scaling discussion .
the value of the latent heat when obtained with a cubic spline fit to the peaks in @xmath115 is @xmath126 .
results obtained with mixed hot - cold starts in @xmath127 seems to give an energy jump around 0.067 .
however the reliability of this method is very limited and we do not dare to extract strong conclusions from it . taking into account
that he cubic spline at @xmath121 gives 0.053(2 ) a possible scenario would be a slowly increasing latent heat towards its asymptotic value @xmath128 .
the peak of the specific heat for the different lattice sizes is displayed in figure [ peaks02 ] .
the continuous line represents the @xmath129 extrapolation .
the plot of the peak value , @xmath130 as a function of the plaquette number reveals a linear relationship , and hence a first order character .
let us first situate the status of the studies using the toroidal topology with the wilson action .
as we have pointed out in the introduction , the transition was believed to be continuous till the simulation of a @xmath8 lattice revealed the first signs of the existence of two metastable states @xcite .
numerical simulations up to @xmath132 showed a rapid decrease of the latent heat when increasing the lattice size , suggesting that the two metastable states might superimpose in the limit @xmath133 @xcite .
however , in @xcite , azcoiti et al . suggest that the latent heat starts to stabilize at @xmath134 .
when one tries to simulate large lattices the tunneling becomes scarce and the use of multicanonical simulations for those lattice sizes seems to be in order .
however , in spite of its usefulness in spin models , we are aware of no multicanonical simulation showing a substancial flip - flop rate improvement for this model . an alternative procedure to enhance the tunneling probability has been proposed in @xcite . within this approach the use of a monopole term in the action as a dynamical variable , has been revealed as a useful method to increase the flip - flop rate .
lattice sizes up to @xmath135 have also been studied using multihistogramming techniques @xcite and rg approaches @xcite .
the results support the idea of a quasi stable latent heat for @xmath136 .
topological considerations stressed already in the introduction led some authors to use lattices homotopic to the sphere . in the case of the wilson action ( @xmath23 ) the two - state signal
is absent up to @xmath110 @xcite . following this , we have simulated on the spherical topology in lattices ranging from @xmath120 to @xmath137 , finding that the two - state signal sets in from @xmath138 on .
the preliminary results up to @xmath138 are already published @xcite .
we include here the results with higher statistics in @xmath138 and the results in @xmath137 . the mc evolution for @xmath139 is plotted in figure [ evol0 ] . for the @xmath137 lattice we have four independent runs , signaled by the dashed lines in the figure , all of them giving the same predictions . in figure [ histo0 ] the distribution of @xmath57
is plotted in lattices ranging from @xmath120 to @xmath137 .
we observe that a two - peak structure is revealed first time by the histogram in @xmath138 , which has a @xmath140 . on the toroidal topology
the equivalent signal is observed already at @xmath8 ( see small window in figure [ histo0 ] ) .
having in mind the results of the previous section at @xmath109 , it seems that the minimum @xmath60 required to observe a two - peak structure in the spherical topology is around three times the minimum lattice size @xmath22 needed in the torus to observe two peaks .
the lattice sizes used in @xcite at @xmath141 ranged from @xmath114 to @xmath110 , so it is not surprising that they did not see any two - peak structure .
our results are in that sense compatible with theirs , though ours show a faster divergence for @xmath142 with increasing @xmath46 than the one observed in @xcite in the lattice sizes we share ( n=6,8,10 ) .
however we are particularly confident on this point because our simulation has been performed closer to the peak of the specific heat @xmath96 , and so we expect the ferrenberg - swendsen extrapolation to be more precise . for the sake of comparison with the full group u(1 ) , simulations with the wilson action
have been performed at @xmath143 . at this coupling
an implementation of overrelax is possible , and the global decorrelating effect should manifest itself in a better statistical quality . in figure [ histo0o ] we show the histograms of @xmath57 from @xmath120 to @xmath138 simulating the full group .
the results are fully compatible with those obtained from simulations with @xmath144 .
this test being performed , we go back to the description of the results obtained for the discrete group .
in figure [ peaks0 ] we plot @xmath145 for different @xmath146s , together with the ferrenberg - swendsen extrapolation .
the small window is a @xmath147 plot of @xmath148 as a function of @xmath48 .
a linear behavior is observed from @xmath110 on , which is even faster than linear when @xmath137 is taken into account .
this fact by itself implies the first order character of the transition since it means that the maximum in the energy fluctuation has the size of the volume . as a further check we have measured the @xmath10 exponent from the scaling of the fisher zeroes ( see table [ tabla_esfera ] ) .
in figure [ fisher0 ] we plot @xmath85 for several lattice sizes on the spherical topology at @xmath143 .
the different curves stand for the different @xmath73 we extrapolate . for small lattice sizes @xmath149
is larger , and the damping is more severe than for the larger lattices since in the later the imaginary part contributes with a faster oscillating function .
actually , we observe that for @xmath137 even a second minimum could be measured before the signal is damped , while for @xmath110 one can measure accurately only the first one . the results for the effective @xmath10 are plotted in figure [ nuesfera ] lower window . as could be expected from the behavior of the specific heat , for small lattices
the effective exponent is somewhat larger than @xmath150 .
it gets compatible with the first order value from @xmath110 on . in the largest lattices the distance between the two peaks slightly increases , and we measure a @xmath151 slightly smaller than the first order value . as
a @xmath152 is strictly impossible we expect this to be a transient effect due to finite size effects associated to the observed splitting up of the two peaks .
concerning the latent heat , we observe a behavior completely analogous to the one observed at @xmath109 .
we do observe a two peak structure quite stable when comparing @xmath138 and @xmath137 , but it slightly increases in @xmath137 .
the plot of the binder cumulant reflects again this fact ( see figure [ binder0 ] ) .
the fast growth of @xmath124 towards @xmath125 is preempted by the onset of double peaked histograms from @xmath138 on , and it even decreases in @xmath137 .
the cubic spline fit in @xmath137 at the peaks gives for the latent heat @xmath153 .
which is compatible with the results suggested by extrapolating to infinite volume the values obtained on the torus up to @xmath154 .
our results are hence supporting the conjecture stressed in @xcite about a quasi stabilization of the latent heat on the torus from @xmath9 on .
we have performed a systematic study of the transition at several negative values of @xmath3 on the toroidal topology .
we find that the two - state signal persists for all @xmath3 values we consider .
the energy histograms reveal an increasing weakness of the transition when going to more negative @xmath3 values ( see figure [ histo_toro ] ) .
a double peak structure is clearly visible at @xmath156 in @xmath104 , while at @xmath157 one has to go to @xmath9 to observe an equivalent signal . in what concerns the behavior of the free energy gap @xmath91
, it grows for all investigated lattice sizes at all negative @xmath3 values ( see figure [ gap_lat ] lower window ) .
the value of @xmath22 at which @xmath91 starts growing is certainly larger as the value of @xmath3 is more negative .
this is another test of the increasing weakness of the transition as @xmath3 gets more negative .
the statistics performed on the torus are reported in table [ tabla_toro ] .
we also quote for the different negative @xmath3 , the value of @xmath96 and the position of partition function zero closest to the real axis .
we have computed from the imaginary part of the zeroes the effective @xmath10 exponent between consecutive lattice sizes following ( [ nu ] ) . in figure [ nu_toro ]
we plot for the different @xmath3 values the @xmath151 we measure . in all cases a @xmath158 is observed for small lattice sizes , which gets closer to 0.25 when the lattice is large enough . from this figure
the trend of @xmath151 seems rather clear towards the first order value . from the energy distributions we measure the latent heat through a cubic spline at the peaks .
taking into account the value @xmath159 we measure , we plotted it as a function of the inverse of the volume @xmath160 , which is also the expected behavior of the latent heat when the transition is first order .
the latent heat can be extrapolated to a value which is safely far from zero ( see figure [ gap_lat ] upper window ) . in figure [ peaks_toro ]
we plot @xmath161 as a function of @xmath48 .
as could be expected from the behavior of the effective exponent @xmath10 , the maximum of the specific heat for small lattices diverges slower than the volume . for the smaller lattices the effective exponent is @xmath162
this value increases monotonically with the lattice size . in the largest ones we observe @xmath163 . on the toroidal topology
we have found that the minimum lattice size required to observe a two state signal is obtained through an apropriate combination @xmath164 , with increasing @xmath165 for increasingly negative @xmath3 , the behavior being qualitatively similar for all the @xmath3 values we have investigated . in view of this
, we have studied a single @xmath155 value on the spherical lattice to check if the two state signal is absent on this topology . for the sake of comparison with @xcite
we choose this value to be @xmath112 .
we have run simulations on spheres ranging from @xmath120 ( @xmath166 ) to @xmath167 ( @xmath168 ) .
the mc evolution is plotted in figure [ evol_02 ] .
the distribution in @xmath137 is distinctly non gaussian and the splitting of the peaks occurs in @xmath167 ( see figure [ histo_02 ] ) .
we remark that the simulations in @xmath167 are extremely expensive in cpu . in order to alleviate thermalization
we have parallelized the code using shared memory in two ppro processors . in this lattice
we have run two independent simulations starting from different configurations the results being fully compatible .
the behavior of the maximum of the specific heat is shown in figure [ peaks_02 ] .
we observe the same trend than in the previous values of @xmath3 in the sphere , i.e. , an increasingly fast divergence of the specific heat with the lattice size , showing an effective @xmath169 already when the histogram width becomes constant with increasing lattice size . at this value of @xmath3
we have a worse estimation for the latent heat to be , as the two peaks have not split enough to allow an accurate measurement .
again the behavior of the binder cumulant is very significant ( see figure [ binder_02 ] .
the value @xmath124 shows a very fast trend towards 2/3 in the small lattices . when increasing the lattice size the rate gets slower , and finally the value in @xmath167 is compatible with the value in @xmath137 preempting the extrapolation to @xmath125 .
unfortunately lattices larger than @xmath167 are unaccessible to our computers nowadays , and we can not observe a decreasing @xmath124 as we did for the other @xmath3 values . however from the behavior between @xmath137 and @xmath167 an increasing @xmath124 for larger spheres seems to be very unlikely . concerning the effective critical exponent @xmath10 we have measured the position of the first fisher zero ( see table [ tabla_esfera ] ) and compute @xmath151 ( see figure [ nuesfera ] ) .
again the first order value @xmath35 is reached in the largest lattices . due to the fact that on the sphere there are a number of points with less than the maximum connectivity
, violations to standard fss in the form of uncontrolled finite size effects are expected to appear on this topology .
we have observed all along this work that working on the spherical topology retards the onset of two - state signals . in terms of @xmath60 , some qualitative prediction arising from our results would be that at fixed @xmath3 , the minimum lattice size required to observe a double peak structure , @xmath165 , is around three times larger on the sphere than on the torus .
this one being the showier difference , is not certainly the only one .
when comparing @xmath170 on both topologies one finds always smaller values on the spherical topology ( see small window in figure [ salto ] ) .
we did run on @xmath9 on the torus and @xmath115 on the sphere , which has an @xmath117 , sweeping an interval of @xmath15 values including the phase transition . in figure [ salto ]
we have plotted @xmath57 in both topologies . in the region of low
@xmath15 the system is disordered , the entropy is higher and the system is not so sensitive to inhomogeneities of the lattice .
however , in the region of high @xmath15 the energy is in general smaller than for the homogeneous lattice since the system tends to be ordered , and the influence of the sites with less than maximum connectivity is more evident .
this difference gets smaller when the lattice size is increased on the sphere , and hence the energy jump is larger .
this feature explains the observed splitting of the two states on the spherical lattices when increasing lattice size .
altogether , it is not recommended to work on spherical lattices to check the existence or not of two - state signals . however , there are a number of facts that make this topology not so dissapointing in spite of those uncontrolled finite size effects .
the first one concerns the behavior of @xmath96 and the second one the measure of the latent heat . in figure [ betac ]
we compare @xmath96 on the torus and on the sphere .
a first observation is that despite they not having the same values in finite lattices , both curves get closer when increasing the lattice size .
this supports the idea of a common thermodynamic limit for both topologies . on the other hand ,
the shift in the apparent critical coupling with @xmath22 is much less dramatic on the sphere than on the torus .
it seems that finite size corrections to @xmath39 are smaller in the case of the spherical topology .
this behavior has also been observed in the ising model on spherical lattices @xcite .
the behavior of the latent heat is very significant . in figure [ latcompa ]
we compare the latent heat in finite volumes for both topologies . on the wilson line , at @xmath23 the asymptotic value of the latent heat
is obtained in the spherical topology .
the inhomogeneity of the sphere has , paradoxically , helped us to run lattices with @xmath171 without having to worry about the tunneling rate . from this graph
we quote @xmath172 . in the upper window the same is plotted for @xmath112 .
unfortunately we have just a single lattice size on the sphere to measure the latent heat , however , from the behavior exhibited by this topology , an increase of the latent heat on spheres larger than @xmath167 seems rather likely . altogether , at @xmath112 a @xmath173 is plausible .
in what concerns the latent heat , the spherical topology seems to afford an useful complement to the results obtained for the the toroidal lattice .
in fact , on the wilson line lattices larger than @xmath135 can not be studied due to the technical problem associated to the scarce tunneling .
the spherical lattice alleviates this technical problem , and the results are supporting the value of the extrapolated latent heat as a function of the inverse of the volume , from the data on toroidal lattices up to @xmath135 .
one could consider the possibility of simulating on spherical lattices at negative @xmath3 to solve the tunneling problem , but in our opinion , the price to pay is too high .
the first order character of the deconfinement transition in pure u(1 ) has been proved , up to the limits of a rather reasonable numerical evidence , in the interval @xmath1 $ ] .
in @xmath109 we have been able to stabilize the latent heat .
we are aware of no simulation on the torus showing a stable latent heat due to the scarce tunneling in lattices larger than @xmath8 .
the spherical topology has helped to solve this problem .
however , probably any lattice with inhomogeneities would produce the same catalysing effects .
in @xmath143 we have also been able to measure the asymptotic value of the latent heat .
we have proved the suggestions of several authors about a quasi stabilization on lattices larger than @xmath174 .
the data on the spherical topology have been crucial to discard the possibility of a slowly vanishing latent heat .
it follows that the discretization of pure compact u(1 ) lgt on the lattice using the wilson action exhibits a first order phase transition with a latent heat in the thermodynamical limit @xmath175 . on the toroidal topology things happen qualitatively in the same way than in @xmath143 up to @xmath157 .
we have run on spherical lattices in @xmath112 looking for an argument similar to the one found in the wilson case concerning the stabilization of the latent heat .
the simulation in @xmath167 gives an estimation for the latent heat though rather imprecise because the splitting of the peaks is not good enough for the measurement to be accurate .
however , in view of the behavior exhibited by the spherical topology concerning the trend of the two peaks on spherical lattices to separate , we are prone to consider the value measured in @xmath167 as a lower bound for the latent heat in @xmath176 . on the other hand , we would at present
say that the latent heat extrapolated from the data on the torus up to @xmath177 is rather accurate , in view of the behavior for the wilson case .
the possibility of running larger spheres surpasses our computer resources .
a highly parallelized version of the code should be used to alleviate thermalization , which is a possibility we do not discard completely at medium term .
either proving or discarding the possibility of a tcp at some finite negative @xmath178 will be a very difficult task from the numerical point of view .
an analytical argument would be welcome .
a very tiny two - state signal , comparable with the one observed in @xmath104 at @xmath157 , is observed on the torus in @xmath135 at @xmath179 , which is the most negative value we have run on the torus .
taking into account the factor 3 in @xmath60 one would need a @xmath180 sphere to observe a tiny double peak structure .
simulation with spherical lattices in this range of @xmath3 have no sense , and nothing can be concluded from the absence of two state signals from such negative @xmath3 values . in view of these difficulties to stabilize the latent heat for very negative @xmath3 values , the only chance to discern
the order of the phase transition is the study of effective critical exponents .
cumulants of the energy , such as the binder cumulant , have been shown to behave like expected in the first order case only when the stable two - peak structure is almost setting in , and there , we do not need further evidences any longer .
we could not expect additional information since by definition the binder cumulant relies on the existence of stable latent heat in order to extrapolate to @xmath181 .
the advantage of studying the effective exponents ( which is nothing but studying the evolution of the histograms width ) is that we do not need a direct observation of latent heat to conclude that a transition is first order @xcite .
we have observed a @xmath151 which evolves monotonically until it reaches the first order value , namely 0.25 , in all cases .
the statistics needed to observe a monotonous behavior , are order @xmath182 at the coupling @xmath96 , which has to be located with high precision ( four digits in our experience ) in order to accurately measure effective exponents .
however , one has to make every effort to observe the trend of the effective exponents , since it is the only chance to discern the order of such tricky transitions . in pure u(1 ) , for the wilson case , an exponent @xmath183 was widely predicted @xcite at the beginning of the eighties when the lattice sizes where too small to reveal two peaks . that @xmath10 was shown to become @xmath184 when simulating @xmath154 @xcite
we have measured @xmath185 on spherical lattices and stated its first order character . within the approximation of effective potentials
it can be shown that along the transient region of a weak first order phase transition , everything goes like in a second order one with a thermal exponent @xmath186 @xcite .
together with josephson law ( @xmath187 ) it implies @xmath29 .
this statement has been checked in 2d potts @xcite , 3d and 4d @xmath188 models @xcite and in the 4d su(2)-higgs at t=0 with fixed higgs modulus @xcite .
our results prove that pure compact u(1 ) theory behaves in the same way .
a general remark to draw from this paper is though there is no theorem constraining transitions presenting @xmath158 to be of wfo , yet the occurrence of such exponent is a warning to consider the phase transition as suspicious of being first order .
we have benefited from comments and discussions with l.a .
fernndez and j. salas . the discussions with c.b .
lang and j. jersk at the final stage of the work have been very appreciated by the authors .
i.c . feels indebted to j. jersk for his many stimulating comments as well as to c.b .
lang for his warm hospitality .
the simulations have been done using the _ rtnn _ machine at theoretical physics department , university of zaragoza ( spain ) .
i.c . is a mec fellow
. 99 m. creutz , l. jacobs and c. rebbi .
rev . _ * d20 * ( 1979 ) 1915 b. lautrup and m. nauenberg _ phys
* b95 * ( 1980 ) 63 g. bhanot _ phys .
_ * d24 * ( 1981 ) 461 j. jersk , t. neuhaus and p.m. zerwas _ phys .
* b133 * ( 1983 ) 103 v. azcoiti , g. di carlo and a. grillo _ phys . lett . _
* b238 * ( 1990 ) 355 g. bhanot , t. lippert , k. schilling and p. ueberholz _ nuc . phys . _ * b378 * ( 1992 ) 633 c.b .
lang and c. rebbi _ phys .
rev . _ * d35 * ( 1987 ) 2510 d.j .
callaway and r. petronzio _ nuc .
* b280 * ( 1987 ) 481 p. hasenfratz and a. hasenfratz _ nuc .
phys . _ * b295 * ( 1988 ) 1 l.a .
fernndez , a. muoz - sudupe , r. petronzio and a. tarancn .
lett . _ * b267 * ( 1991 ) 100 g. bhanot _ nuc .
phys . _ * b205 * ( 1982 ) 168 h.g .
evertz , j. jersk , t. neuhaus and p.m. zerwas _ nuc .
* b251 * ( 1985 ) 279 a. hasenfratz _ phys .
lett . _ * b201 * ( 1988 ) 492 r. gupta , m.a .
novotny and r. cordery _ phys .
lett . _ * b172 * ( 1986 ) 86 c.b .
lang and t. neuhaus _ nuc .
_ * b431 * ( 1994 ) 119 m. baig and h. fort _ phys .
lett . _ * b332 * ( 1994 ) 428 j. jersk , c.b .
lang and t. neuhaus _ phys .
* 77 * ( 1996 ) 1933 w. kerler , c. rebbi and a. weber _ phys .
lett . _ * b348 * ( 1995 ) 565 j. cox , w. franzki , j. jersk , c.b .
lang and t. neuhaus _
( preprint hep - lat://9705043 ) _ l.a .
fernndez , m.p .
lombardo , j.j .
ruiz - lorenzo and a. tarancn _ phys .
_ * b277 * ( 1992 ) 485 the ape collaboration _ nuc .
* b318 * ( 1989 ) 553 w. kerler , c. rebbi and a. weber _ nuc .
phys . _ * b452 * ( 1995 ) 452 i. campos , a. cruz and a. tarancn _ to appear in phys .
b _ j. lee and j.m .
* 65 * ( 1990 ) 137 m.s .
challa , d.p .
landau , k. binder _ phys .
* d34 * ( 1986 ) 1841 c. n. yang and t. d. lee .
* 87 * ( 1952 ) 404 e. marinari in _ 1996 budapest summer school on monte carlo methods _ _
e - print : cond - mat//9612010 _ m. falcioni , g. martinelli , m.l .
paciello , g. parisi and b. taglienti _ nuc .
_ * b265 * ( 1986 ) 187 .
bosonic algorithms _ in _ quantum fields on the computer_. advanced series on direction in high energy physics vol .
11 , m. creutz editor .
world scientific , singapore 1992 .
v. azcoiti , g. di carlo and a. grillo _ phys .
_ * b268 * ( 1991 ) 101 o. diego , j. gonzalez , j. salas _ j. phys . _
* a27 * ( 1994 ) 2965 .
alonso , j.m .
carmona , j. clemente , l.a .
fernndez , d. iiguez , a. tarancn and c.l .
lett . _ * b376 * ( 1996 ) 148 h.g .
ballesteros , j.m .
carmona , l.a .
fernndez and a. tarancn _ phys .
* b419 * ( 1998 ) p 303 i. campos , l.a .
fernndez , a. tarancn _ phys .
rev . _ * d55 * ( 1997 ) 2965 j.l .
alonso , h.g .
ballesteros , l.a .
fernndez , v. martn - mayor , a. muoz sudupe and a. tarancn _ phys .
rev . _ * b53 * ( 1996 ) 2537 i. campos , _ nuc .
phys . _ * b514 * ( 1998 ) 336
as a further check we have implemented the schwinger - dyson equations ( sde ) @xcite which allow one to recover the simulated couplings from the montecarlo data .
this equation relates the values of the couplings with the expectation values we measure from the mc simulation .
we need as many independent equations as couplings we have to determine in the action . in our case , in order to measure both @xmath15 and @xmath3 , we need two operators with null expectation value in order to have two independent tests . at each lattice site @xmath193 and for every direction @xmath194
we consider the operators : @xmath195 on the hypertorus @xmath197 means the sum over the plaquettes in positive and negative directions ( @xmath198 ) bordering the link ( 12 plaquettes ) . on the sphere one has to be careful , since not all the links have 12 surrounding plaquettes , and the sum has to be understood as extended to the existing plaquettes .
these equations hold for all @xmath193 in such a way that we can sum up the equations for every single link and average over the number of links .
we quote in table [ tabla_sd ] for the tests done at @xmath112 using both topologies .
they show a perfect agreement between the simulated couplings and the ones recovered from the mc expectation values . | we have performed a systematic study of the phase transition in the pure compact u(1 ) lattice gauge theory in the extended coupling parameter space ( @xmath0 ) on toroidal and spherical lattices .
the observation of a non - zero latent heat in both topologies for all investigated @xmath1 $ ] , together with an exponent @xmath2 when large enough lattices are considered , lead us to conclude that the phase transition is first order . for negative @xmath3 ,
our results point to an increasingly weak first order transition as @xmath3 is made more negative .
_ departamento de fsica terica , facultad de ciencias , + universidad de zaragoza , 50009 zaragoza , spain + e - mail : isabel , tarancon , [email protected]_ + |
Subsets and Splits